text
stringlengths 254
1.16M
|
|---|
---
title: Pip5k1c Loss in Chondrocytes Causes Spontaneous Osteoarthritic Lesions in Aged
Mice
authors:
- Minghao Qu
- Mingjue Chen
- Weiyuan Gong
- Shaochuan Huo
- Qinnan Yan
- Qing Yao
- Yumei Lai
- Di Chen
- Xiaohao Wu
- Guozhi Xiao
journal: Aging and Disease
year: 2023
pmcid: PMC10017150
doi: 10.14336/AD.2022.0828
license: CC BY 2.0
---
# Pip5k1c Loss in Chondrocytes Causes Spontaneous Osteoarthritic Lesions in Aged Mice
## Body
Osteoarthritis (OA) is a common degenerative joint disease characterized by progressive degeneration of articular cartilage, subchondral sclerosis, synovial inflammation, and osteophyte formation [1]. The clinical symptoms of OA include chronic pain, joint swelling and stiffness, and limited range of motion, leading to disability, psychological distress, and reduced quality of life [2]. The major risk factors for developing OA include aging, joint trauma, obesity, and genetic susceptibility [3]. During the last decade, the global prevalence of OA has rapidly increased, especially in aged populations [4, 5]. For instance, in China, the number of patients with symptomatic OA has increased from 26.1 million to 61.2 million, from 1990 to 2017 [6]. To date, there are no FDA-approved medications that can effectively prevent or delay OA progression due to a limited understanding of OA pathogenesis. Thus, it is highly desirable to investigate the pathological mechanisms underlying OA initiation, development, and progression.
Type 1 phosphatidylinositol 4-phosphate 5-kinases (Pip5k1s) are a group of lipid kinases that phosphorylate the fifth hydroxyl of phosphatidylinositol 4 phosphate (Pi4p) to synthesize phospholipid phosphatidylinositol 4,5-bisphosphate (Pip2) [7]. The latter can serve as a second messenger directly or as a precursor to form other second messengers, such as inositol 1,4,5-triphosphate and diacylglycerol [8]. In mammals, there are three isoforms of Pip5k1 protein, termed Pip5k1a, Pip5k1b, and Pip5k1c [9]. Cumulative evidence has highlighted the pivotal functions of Pip5k1c in a series of physiological processes, such as focal adhesion (FA) formation and dynamics, cell migration, vesicle trafficking, intracellular calcium release, energy metabolism, and cellular signal transduction [10-17]. Moreover, alterations in Pip5k1c expression and/or activation have been linked to several disease conditions, such as osteoporosis, neural dysfunction, obesity, pain hypersensitivity, inflammation, and tumor metastasis [13, 18-21]. In humans, homozygous mutations in the PIP5K1C gene cause a rare condition termed lethal congenital contracture syndrome 3 (LCCS3), which is characterized by severe joint contracture, reduced or absent limb movement, and lethality at or soon after birth [22]. In mice, homozygous deficiency of the Pip5k1c gene causes early lethality, with a $50\%$ reduction of Pip2 in the brain and impaired synaptic transmission in cortical neurons [23]. Xu and colleagues have reported that the polarization of Pip5k1c induced by integrins is required for the recruitment of neutrophils during inflammatory responses [20]. Zhu et al. have reported that excessive Pip5k1c expression impairs osteoclast formation and bone resorption via enhancing the *Pip2* generation [24]. Our previous study has demonstrated an essential role of Pip5k1c, through its expression in mesenchymal stem cells (MSCs), in the control of bone remodeling [25]. Loss of Pip5k1c in MSCs leads to a low turn-over osteopenia-like phenotype in adult mice, by impairing Runx2-mediated osteoblast differentiation and subsequent bone formation [25]. While these studies have clearly indicated the involvement of Pip5k1c in physio-pathological conditions of the musculoskeletal system, whether Pip5k1c plays a role in the pathogenesis of OA remains unknown.
In this study, we demonstrate that inducible deletion of Pip5k1c expression in aggrecan-expressing chondrocytes causes spontaneous OA-like phenotypes in aged (15-month-old) mice, but not in adult (7-month-old) mice. Pip5k1c loss decreases chondrocyte proliferation and increases cell apoptosis in the knee joint articular cartilage of aged mice. Pip5k1c loss inhibits the expression of anabolic extracellular matrix (ECM) proteins and promotes chondrocyte hypertrophic differentiation in aged articular cartilages. Pip5k1c deletion impairs chondrocyte-ECM adhesion partially through downregulation of the expression of several FA proteins.
## Abstract
Osteoarthritis (OA) is the most common degenerative joint disease affecting the older populations globally. Phosphatidylinositol-4-phosphate 5-kinase type-1 gamma (Pip5k1c), a lipid kinase catalyzing the synthesis of phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2), is involved in various cellular processes, such as focal adhesion (FA) formation, cell migration, and cellular signal transduction. However, whether Pip5k1c plays a role in the pathogenesis of OA remains unclear. Here we show that inducible deletion of Pip5k1c in aggrecan-expressing chondrocytes (cKO) causes multiple spontaneous OA-like lesions, including cartilage degradation, surface fissures, subchondral sclerosis, meniscus deformation, synovial hyperplasia, and osteophyte formation in aged (15-month-old) mice, but not in adult (7-month-old) mice. Pip5k1c loss promotes extracellular matrix (ECM) degradation, chondrocyte hypertrophy and apoptosis, and inhibits chondrocyte proliferation in the articular cartilage of aged mice. Pip5k1c loss dramatically downregulates the expressions of several key FA proteins, including activated integrin β1, talin, and vinculin, and thus impairs the chondrocyte adhesion and spreading on ECM. Collectively, these findings suggest that Pip5k1c expression in chondrocytes plays a critical role in maintaining articular cartilage homeostasis and protecting against age-related OA.
## Animal studies
*The* generation of Pip5k1cfl/fl mice was previously described [25]. The Pip5k1cfl/fl mice were bred with the AggrecanCreERT2 knock-in transgenic mice to obtain the Pip5k1cfl/fl; AggrecanCreERT2 mice. For inducible deletion of Pip5k1c gene in aggrecan-expressing chondrocytes, 2-month-old male Pip5k1cfl/fl; AggrecanCreERT2 mice were intraperitoneally injected with tamoxifen (Sigma T5648, 100 mg/kg per body weight/day, 5 continuous injections). Age-matched male Pip5k1cfl/fl; AggrecanCreERT2 mice were treated with corn oil and served as the control group. All research protocols in this study were approved by the Institutional Animal Care and Use Committees (IACUC) of the Southern University of Science and Technology.
## Micro-computerized tomography
In vivo micro-computerized tomography (µCT) analyses of the knee joint were performed according to our previously established protocol [26, 27]. After sacrifice, the knee joints were collected, fixed in $4\%$ paraformal-dehyde for 24 hours, and scanned using a Skyscan scanner 1276 high-resolution µCT scanner (Bruker, Aartselaar, Belgium) with 60 kVp source and 100 µAmp current with a resolution of 10 µm. Three-dimensional structural reconstructions were performed using the scanned µCT images from each group at the same thresholds. Quantitative µCT parameters, including bone mineral density (BMD) and the volume of calcified meniscus and synovial tissue, were analyzed as previously described [26, 28].
## Histology
The decalcification, dehydration, and paraffin embedding of knee joint samples were performed as previously described [26, 29]. The paraffin-embedded knee joint samples were cut into 5-µm thick sections and stained with Safranin O & Fast Green (SO&FG) (Solarbio, Cat#G1371) as previously described [26, 30]. The severity of OA-like lesions was evaluated using the Osteoarthritis Research Society International (OARSI) scoring system in a double-blinded manner. The Safranin O-positive areas of articular cartilage and growth plate were analyzed by Image J (version 1.53k) as previously described [26]. Representative images were selected based on the mean values of histological scores.
## Quantitative immunofluorescent analyses
For immunofluorescent (IF) staining, 5-µm knee joint sections were hydrated and permeabilized with Immunostaining Permeabilization Solution with Saponin (Beyotime, Cat# P0095) for 5 mins at room temperature (RT), blocked with Immunol Staining Blocking Buffer (Beyotime, Cat# P0102) for 1h at RT, and then incubated with primary antibodies overnight at 4°C. Antibodies used for IF staining in this study were Pip5k1c (Santa Cruz, sc-377061, 1:50), Aggrecan (Abcam, ab36861, 1:200), Col2a1 (Abcam, ab34712, 1:200), Mmp13 (Abcam, ab39012, 1:200), Adamts5 (Abcam, ab41037, 1:200), Col10a1 (Abcam, ab58632, 1:200), Runx2 (Abcam, ab23981, 1:200), Ki67 (CST, 12202S, 1:200), 9EG7 (BD Pharmingen, 553715, 1:200), talin (Abcam, ab110080, 1:200), and vinculin (Santa Cruz, sc-73614, 1:200). After washing in PBS with $0.1\%$ Tween 20, the sections were incubated with Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 (Invitrogen, Cat# A-11008) (1:400) for 1h at RT. Isotype antibody (Normal Rabbit IgG, Sigma, NI01) controls and secondary antibody-only controls were employed to validate antibody specificity and distinguish genuine target staining from the background. The fluorescent signals in articular cartilages were determined using a Leica SP8 Confocal Microsystems. Representative images were selected based on the mean values of fluorescent signals.
## TUNEL staining
Cell apoptosis was evaluated using the One Step TUNEL Apoptosis Assay Kit (Red Fluorescence) (Beyotime, C1090) as previously described [31, 32].
## In vitro siRNA knockdown experiments
Mouse ATDC5 cells were cultured in DMEM/F12 supplemented with $5\%$ FBS, $1\%$ penicillin and streptomycin, and $1\%$ insulin-transferrin-selenium (Gibco™, Cat# 51500056) to induce chondrogenic differentiation. For in vitro knockdown of Pip5k1c expression, ATDC5 cells were transfected with Pip5k1c siRNA using a Lipofectamine RNAiMAX transfection reagent (Invitrogen, Cat# 13778075) as previously described [25, 33, 34]. ATDC5 cells transfected with negative control siRNA were used as the control group. Cell-ECM adhesion assay was performed according to our previously established protocol [26]. At 48h after siRNA transfection, protein extracts were collected and analyzed by western blotting. Antibodies used for western blotting in this study were Pip5k1c (Santa Cruz, sc-377061, 1:1000), Aggrecan (Abcam, ab36861, 1:1000), Col2a1 (Abcam, ab34712, 1:1000), Tubulin (CWBIO, CW0098, 1:1000), PI3K (CST, 4292S, 1:1000), phosphorylated-PI3K (CST, 4228S, 1:1000), Akt (CST, 4691S, 1:1000), phosphorylated-Akt (CST, 4056S, 1:1000), Erk (CST, 9102S, 1:1000), phosphorylated-Erk (CST, 9101S, 1:1000), integrin β1 (CST, 34971, 1:1000), integrin β3 (CST, 13166S, 1:1000), talin (Abcam, ab110080, 1:1000), and vinculin (Santa Cruz, sc-73614, 1:1000). The Pip5k1c siRNA sequence: 5’ primer-GCGU GCAGUCUGGUGGCAATT, 3’ primer-UUGCCACCA GACUGCACGCTT.
## Statistical Analysis
All mice used in this study were randomly assigned to each group. Statistical analyses were completed using the Prism GraphPad. Results were expressed as mean ± standard deviation (s.d.). Normality of data was tested for all variables using the Kolmogorov-Smirnov (K-S) test. For normally distributed data, a two-tailed unpaired Student’s t test was used to assess the differences between the two groups. For non-normally distributed data, an unpaired nonparametric Mann-Whitney test was used to determine the statistical difference between the two groups. A two-way ANOVA test was used for the repeated measurement data from longitudinal in vivo µCT analyses and in vitro chondrocyte-ECM adhesion assay. Differences with $P \leq 0.05$ were considered statistically significant.
## Data availability
All data generated for this study are available from the corresponding authors upon reasonable request.
## Generation of inducible chondrocyte-specific Pip5k1c knockout mice
To investigate the role of Pip5k1c in chondrocytes, the floxed Pip5k1c (Pip5k1cfl/fl) mice were bred with AggrecanCreERT2 transgenic mice to generate Pip5k1cfl/fl; AggrecanCreERT2 mice (Fig. 1A, B). For inducible deletion of Pip5k1c in aggrecan-expressing chondrocytes, 2-month-old male Pip5k1cfl/fl; AggrecanCreERT2 mice were intraperitoneally injected with tamoxifen (TAM, 100 mg kg-1 body weight) (hereafter referred to as cKO). Note: Age- and sex-matched Pip5k1cfl/fl; AggrecanCreERT2 mice treated with corn oil were used as a control group. At 5 and 13 months after TAM injections, in vivo µCT scans were performed to detect structural changes in the knee joints. At 15 months of age, all mice were sacrificed, and the knee joints were collected for further analyses. The deletion of Pip5k1c in chondrocytes was confirmed by immunofluorescent (IF) staining (Fig. 1C). Quantitative analyses showed that the percentages of Pip5k1c-positive cells were decreased by $26.7\%$ and $27.7\%$ in articular cartilage and growth plate, respectively, in cKO mice as compared to those in control mice ($P \leq 0.0001$, two-tailed unpaired Student’ t test) (Fig. 1D).
Figure 1.Genetic deletion of Pip5k1c in aggrecan-expressing chondrocytes in adult mice. ( A) PCR genotyping using tail DNA. Pip5k1c flox KO band, ~380bp; Pip5k1c flox wildtype (WT) band, ~250bp; AggrecanCreERT2, ~650bp. ( B) A schematic diagram illustrating the breeding strategy and experimental design. ( C) Immunofluorescent (IF) staining of knee joint sections showing the reduced expression of Pip5k1c in articular cartilage (AC) and growth plate (GP) after TAM injections. White dashed boxes indicate the higher magnification images in the right panels. Scale Bar: 50 mm. ( D) Percentages of Pip5k1c-positive cells in AC and GP. $$n = 6$$ mice per group. Results are expressed as mean ± standard deviation (s.d.). The exact P values are shown in the figures.
## Pip5k1c loss causes subchondral bone sclerosis and osteophyte formation in aged mice
Results from in vivo µCT analyses showed no marked difference in knee joint structure between control and cKO groups at 5 months after TAM injections (Fig. 2A, left panels). Interestingly, we observed a significant increase in subchondral bone sclerosis and osteophyte formation in cKO mice at 13 months after TAM injections ($$P \leq 0.0011$$, two-way ANOVA test) (Fig. 2A, right panels). Moreover, quantitative µCT parameters, including bone mineral density (BMD) and bone volume of calcified meniscus and synovium, were comparable between the two groups at 5 months after TAM injections, but were markedly increased in cKO mice at 13 months after TAM injections relative to control group ($$P \leq 0.0003$$, two-way ANOVA test) (Fig. 2B, C).
Figure 2.Pip5k1c loss induces subchondral bone sclerosis and osteophyte formation in aged mice. ( A) In vivo μCT scans of knee joints from control and cKO mice at 5- and 13-months post-TAM injections. Scale bar, 1 mm. Red arrowheads indicate the formation of osteophytes. Yellow arrowheads indicate subchondral bone sclerosis. ( B, C) Quantitative analyses of bone mineral density (BMD) (B) and the volume of calcified meniscus and synovial tissue (C). $$n = 6$$ mice per group. Results are expressed as mean ± standard deviation (s.d.). The exact P values are shown in the figures.
## Pip5k1c loss causes spontaneous OA-like lesions in aged mice
Next, we performed safranin O & fast green (SO&FG) staining on knee joint sections from control and cKO mice at 13 months after TAM injections. Strikingly, the cKO mice displayed a series of severe OA-like lesions, including spontaneous surface fissures of articular cartilage (Fig. 3A, black arrowheads), loss of integrity in growth plate (Fig. 3A, blue arrowheads), loss of safranin O staining in articular cartilage and growth plate (Fig. 3A, green arrowheads), and excessive osteophyte formation (Fig. 3A, red arrowheads). Moreover, meniscus deformation and synovial hyperplasia were observed in cKO mice (Fig. 3A). Quantitative histological analyses revealed significantly higher Osteoarthritis Research Society International (OARSI) scores, osteophyte scores, and synovitis scores in cKO mice as compared with those in control mice (Fig. 3B-D) ($P \leq 0.05$ for all indicated parameters, two-tailed unpaired Student’s t test). In addition, the safranin O-stained cartilage areas were decreased by $22.72\%$ and $20.41\%$ in articular cartilage and growth plate, respectively, in cKO versus control mice (Fig. 3E, F).
Figure 3.Loss of Pip5k1c in chondrocytes promotes OA-like lesions in aged mice. ( A) Representative images of safranin O & fast green (SO&FG)-stained knee joint sections from control and cKO mice at 13 months after TAM injections. Black dashed boxes indicate the higher magnification images of AC, GP, meniscus, and synovium in lower panels. Black arrowheads indicate the degradation of AC. Blue arrowheads indicate the loss of integrity of GP. Red arrowheads indicate the formation of osteophytes. Scale bar, 50 μm. ( B) The severity of OA-like lesions was analyzed using the Osteoarthritis Research Society International (OARSI) scoring system. ( C, D) Quantitative analyses of safranin O-positive areas in the AC (c) and GP (d). ( E, F) Osteophyte score (E) and synovitis score (F) were performed using histological sections. $$n = 6$$ mice per group. Results are expressed as mean ± standard deviation (s.d.). The exact P values are shown in the figures.
Figure 4.Pip5k1c loss causes ECM degradation and chondrocyte hypertrophic differentiation in aged mice. ( A) IF staining for expressions of aggrecan, Col2a1, Mmp13, Adamts5, Col10a1, and Runx2 using knee joint sections from control or cKO mice at 13 months after TAM injections. White dashed boxes indicate the higher magnification images in the right panels. White dashed lines indicate the cartilage surfaces. Scale bar: 50 µm. ( B-G) Quantitative analyses of the percentages of aggrecan-, Col2a1-, Mmp13-, Adamts5-, Col10a1-, and Runx2-positive cells in AC. $$n = 6$$ mice per group. Results are expressed as mean ± standard deviation (s.d.). The exact P values are shown in the figures.
## Pip5k1c loss promotes ECM degradation and chondrocyte hypertrophic differentiation in aged mice
IF staining analyses revealed that Pip5k1c loss significantly decreased the expression levels of anabolic ECM proteins, including aggrecan and Col2a1, in the articular cartilage of cKO mice (Fig. 4A-C). Notably, the percentages of aggrecan- and Col2a1- positive cells were decreased by $30.3\%$and $37.7\%$, respectively, in cKO versus control cartilages ($P \leq 0.001$, two-tailed unpaired Student’ t test) (Fig. 4A-C). Interestingly, we found that the expression levels of catabolic ECM enzymes, including Mmp13 and Adamts5, were comparable between the two groups (Fig. 4A, E, F). We next determined the expression levels of chondrocyte hypertrophic markers, including Col10a1 and Runx2, by IF staining analyses. Results showed that, while Col10a1 was barely detectable in the superficial and middle layers of the articular cartilages in control mice, its expression was dramatically upregulated in these areas in cKO mice ($P \leq 0.0001$, two-tailed unpaired Student’ t test) (Fig. 4A, F). Interestingly, Runx2 expression was not markedly increased in the articular cartilages of cKO mice compared with that in control mice (Fig. 4A, G).
## Pip5k1c loss decreases chondrocyte proliferation and increases chondrocyte apoptosis in aged mice
We further performed IF staining of cell proliferation marker Ki67 to assess whether the proliferative activity of articular chondrocytes could be affected by Pip5k1c deficiency. In control mice, Ki67 was strongly detected in cells of the superficial and middle layers of articular cartilage (Fig. 5A). However, the percentage of Ki67-positive chondrocytes was decreased by $22.17\%$ in these areas of cKO mice compared to that in control mice (Fig. 5B) (56.17±$10.72\%$ in control group vs 34±$9.08\%$ in Pip5k1c-cKO group, $$P \leq 0.0031$$, two-tailed unpaired Student’ t test). Moreover, results from the terminal deoxynucleotidyl transferase-mediated nick-end labelling (TUNEL) staining revealed that Pip5k1c loss markedly increased the number of apoptotic chondrocytes in the superficial and middle layers of articular cartilage in cKO mice ($$P \leq 0.0132$$, two-tailed unpaired Student’ t test) (Fig. 5C, D). In vitro studies from cultured ATDC5 cells showed that siRNA knockdown of Pip5k1c dramatically reduced the protein level of aggrecan and Col2a1 in these cells ($P \leq 0.05$, unpaired nonparametric Mann-Whitney test) (Fig. 5E, F). Moreover, Pip5k1c siRNA treatment significantly downregulated the total and phosphorylated protein levels of PI3K, Akt, and Erk ($P \leq 0.05$ for all indicated parameters, unpaired nonparametric Mann-Whitney test), without affecting the phosphorylated/total ratios of these proteins (Fig. 5E, F).
Figure 5.Pip5k1c loss inhibits chondrocyte proliferation and induces chondrocyte apoptosis in aged mice. ( A) IF staining for expression of Ki67 using knee joint sections from control or cKO mice at 13 months after TAM injections. White dashed boxes indicate the higher magnification images in the right panels. White dashed lines indicate the cartilage surfaces. Scale bar: 50 µm. ( B) *Quantitative data* of (A). ( C) Fluorescent TUNEL staining. Scale bar: 50 µm. ( D) *Quantitative data* of (d). ( E) Western blotting. Protein extracts were isolated from cultured ATDC5 cells transfected with negative control (NC) siRNA or Pip5k1c siRNA and subjected to western blot analyses with indicated antibodies. t: total; p: phosphorylated. ( F) Relative protein levels normalized to the NC siRNA group. In vitro siRNA knockdown experiments were independently repeated four times. Results are expressed as mean ± standard deviation (s.d.). The exact P values are shown in the figures.
Figure 6.Pip5k1c loss reduces expression of FA proteins and impairs chondrocyte-ECM adhesion. ( A) IF staining for expressions of activated integrin β1 (9EG7), talin, and vinculin using knee joint sections from control or cKO mice at 13 months after TAM injections. White dashed boxes indicate the higher magnification images in the right panels. White dashed lines indicate the cartilage surfaces. Scale bar: 50 µm. ( B-D) *Quantitative data* of (a). ( E) Western blotting. Protein extracts were isolated from cultured ATDC5 cells which had been transfected with NC siRNA or Pip5k1c siRNA. ( F) Relative protein levels of talin, vinculin, integrin β1, and integrin β3 in ATDC5 cells transfected with NC siRNA or Pip5k1c siRNA. ( g) Representative images of attachment and spreading of ATDC5 cells on type II collagen-coated surfaces after transfection of NC siRNA or Pip5k1c siRNA. ( H) Percentages of attached cells. In vitro experiments were independently repeated at least three times. Results are expressed as mean ± standard deviation (s.d.). The exact P values are shown in the figures.
## Pip5k1c loss reduces the expression of FA proteins and impairs chondrocyte-ECM adhesion
Previous studies have reported a pivotal role of Pip5k1c in controlling the FA formation [10]. Thus, we determined the expressions of FA-related molecules, including activated integrin β1 (9EG7), talin, and vinculin, in articular cartilage by quantitative IF analyses (Fig. 6A). Results showed that the percentages of 9EG7-, talin- and vinculin-positive cells were all dramatically decreased in articular cartilages of cKO mice when compared with those in control mice (Fig. 6A-D) ($P \leq 0.001$ for all indicated parameters, two-tailed unpaired Student’ t test). Consistently, siRNA knockdown of Pip5k1c in ATDC5 cells resulted in decreased protein expressions of talin, vincular, integrin β1, and integrin β3 ($P \leq 0.05$ for all indicated parameters, unpaired nonparametric Mann-Whitney test) (Fig. 6E, F). Furthermore, siRNA knockdown of Pip5k1c drastically impaired the attachment and spreading of ATDC5 cells on collagen-II-coated surfaces in vitro ($P \leq 0.001$ for Pip5kc1 siRNA group vs NC siRNA group, two-way ANOVA test) (Fig. 6G, H).
## DISCUSSION
Although the complex molecular mechanisms underlying the onset and progression of OA remain incompletely understood, cumulating evidence has pointed to the fact that aging itself is the most prominent risk factor contributing to OA development [35, 36]. Results from clinical studies have shown that the incidence and severity of OA are much higher in aged populations when compared with younger populations [37, 38]. OA is the leading cause of disability in the population aged over 65 and is associated with comorbid disorders, higher mortality, and reduced quality of life [39]. In this study, we provide convincing evidence that genetic deletion of Pip5k1c in chondrocytes causes multiple spontaneous OA lesions, including articular cartilage damage, subchondral sclerosis, synovial inflammation, and osteophyte formation in aged mice. We find that Pip5k1c loss inhibits chondrocyte proliferation, and induces chondrocyte hypertrophy, apoptosis, and ECM degradation. Pip5k1c loss reduces the expression of several key FA proteins and impairs the chondrocyte-ECM adhesions. Notably, this is the first demonstration of the crucial role of Pip5k1c in the maintenance of cartilage homeostasis to protect against aging-induced OA.
In healthy articular cartilages, the ECM forms a complex scaffold comprising collagens, proteoglycans, water content, and fibrous proteins, which not only endows the articular cartilage with unique biomechanical properties but also provides chondrocytes a distinctive microenvironment for maintaining their cellular homeostasis [40]. During OA development, articular chondrocytes undergo abnormal hypertrophic differentiation, leading to reduced synthesis of anabolic ECM proteins, excessive production of chondrocyte hypertrophic marker Col10a1, and upregulations of ECM-degrading enzymes, such as Mmp13 and Adamts$\frac{4}{5}$ [1]. Runx2 is a well-known transcriptional factor for its role in promoting chondrocyte hypertrophy and OA development [41-45]. Our recent study has demonstrated that Pip5k1c regulates the expression level of Runx2 protein, but not its mRNA, though mediating calcium/calmodulin-dependent protein kinase 2 (CaMK2) and cytoplasmic Ca2+ levels, in MSCs [25]. Interestingly, in this study, we find that Pip5k1c deletion enhances chondrocyte hypertrophic differentiation and ECM degradation without upregulation of Runx2 expression in articular chondrocytes. Molecular mechanisms whereby Pip5k1c loss induces chondrocyte hypertrophy and ECM degradation in articular cartilage require further investigations.
Unlike other types of arthritis, OA usually develops slowly over many years [46]. Under physiological conditions, the turnover of aggrecan takes up to 25 years, whereas the half-life of type II collagen ranges from several decades to up to 400 years [47]. Interestingly, results from this study suggest that genetic ablation of Pip5k1c in aggrecan-expressing chondrocytes causes OA-like lesions in aged (15-month-old) mice, but not in adult (7-month-old) mice, which highly mimic the pathological features of OA in humans. We find that Pip5k1c loss dramatically decreases the expression levels of anabolic ECM proteins, including Col2a1 and aggrecan, in the articular cartilages, without upregulating the expression of catabolic enzymes Mmp13 and Adamts5. This finding suggests that Pip5k1c loss impairs ECM homeostasis via suppressing the anabolic activities of articular chondrocytes rather than promoting ECM catabolism, which might partially explain the slowly progressive OA-like phenotypes in the cKO mice. Moreover, whether the function of Pip5k1c is compensated by other Pip5k1s, such as Pip5k1a and Pip5k1b, to delay OA progression in adult cKO mice needs to be determined in future studies.
By utilizing genetically modified animal models, several key molecules and signaling pathways responsible for cartilage degradation and OA onset and progression have been identified, which involve Wnt/β-catenin, Runx2, FGF, miRNAs, Ampk, mTOR, and FA signaling pathways [26, 28, 29, 48-55]. For instance, Zhu and coworkers have reported that sustained activation of β-catenin in articular chondrocytes leads to multiple OA-like phenotypes, including cartilage loss, subchondral remodeling, and chondrophyte/osteophyte formation probably by upregulation of Runx2 [48]. Genetic deletion of Runx2 in chondrocytes decelerates the progression of surgically induced OA, whereas overexpression of Runx2 exerts the opposite effects [49, 50]. In addition, our previous study has demonstrated that mechanical loading activates the mTOR signaling pathway and promotes OA development in mouse temporomandibular joints [54]. In this study, we demonstrate that Pip5k1c loss induces dysregulation of several key signaling pathways involved in cartilage homeostasis and survival by downregulating the protein expressions of PI3K, Akt, and Erk in chondrocytes. Moreover, Pip5k1c loss significantly inhibits the proliferative activity of articular chondrocytes while inducing chondrocyte apoptosis and ECM degradation in articular cartilage. Collectively, these findings suggest that Pip5k1c expression may play a critical role in maintaining cartilage homeostasis via regulating cellular signaling transductions.
Results from this study suggest that Pip5k1c loss induces cartilage degradation and OA-like lesions through, at least in part, impairing the chondrocyte-ECM adhesion. We provide several lines of evidence to support this notion. First, Pip5k1c loss significantly decreases the expression levels of key FA-related proteins, including talin and vinculin, and inhibits the activation of integrin β1 in articular chondrocytes in aged mice. Second, siRNA knockdown of Pip5k1c downregulates the protein expressions of talin, vinculin, integrin β1, and integrin β3 in ATDC5 chondrogenic cells. Third, we demonstrate that Pip5k1c loss drastically impairs the adhesion ability of chondrocytes on collagen type II in vitro. These findings, along with results from our previous study that FA-related molecule plays an essential role in preserving the integrity of articular cartilage to protect against OA damages [26], indicate a potential mechanism that involves Pip5k1c and its interactions with the FA signaling pathway in the pathogenesis of OA onset and progression. It is well established that Pip5k1c catalyzes the phosphorylation of Pi4p to synthesize Pip2, the latter can be further phosphorylated by PI3K to form the second messenger phosphatidylinositol 3,4,5-trisphosphate (Pip3) and activate Akt signaling [56]. Moreover, Pip2 interacts with several key FA proteins, such as FA kinase (FAK), talin, and vinculin, to regulate the FA assembly and dynamics [57-60]. Akt can not only phosphorylate Pip5k1c specifically at serine 555 to regulate Pip5k1c-talin interaction and focal adhesion dynamics [61], but can also act downstream of Pip3 to mediate numerous cellular processes [62]. Whether Pip5k1c loss reduces the expression of FA proteins and impairs chondrocyte-ECM adhesion through downregulating PI3K/Akt activity needs to be determined in future studies.
It should be noted that Pip5k1c loss causes synovitis-like changes, including hyperplasia of synovial lining cells and inflammatory infiltration, in the knee joints of aged mice. Our previous study has demonstrated that AggrecanCreERT2 is highly active in articular chondrocytes, but not in cells of the synovium [26]. Thus, the observed alterations in the synovium are indirect results of Pip5k1c deletion in chondrocytes by AggrecanCreERT2. It is well known that OA is a whole joint disease. Loss of Pip5k1c in chondrocytes impairs ECM homeostasis, proliferation, and adhesion, and promotes cell apoptosis in articular cartilages, which may subsequently induce synovial inflammation, for instance, by changing the micro-environment of the joint. The underlying mechanisms need further investigation in future studies.
We acknowledge that this study has several limitations. First, we did not determine the expression level of Pip5k1c in human cartilages. Whether Pip5k1c expression is altered in human OA cartilages needs to be determined. Second, since articular cartilage is a weight-bearing tissue, it will be interesting to determine if and how Pip5k1c loss in chondrocytes affects the OA development in instability-induced OA models, such as the destabilization of the medial meniscus model. Third, while our results clearly show that Pip5k1c loss induces multiple spontaneous osteoarthritic lesions in aged mice, whether overexpression of Pip5k1c in mouse chondrocytes can exert protective effects against aging-induced OA onset and progression remains to be investigated. In conclusion, our study demonstrates a vital role of Pip5k1c expression in aggrecan-expressing chondrocytes in the regulation of the articular cartilage homeostasis in mice.
## References
1. Chen D, Shen J, Zhao W, Wang T, Han L, Hamilton JL. **Osteoarthritis: toward a comprehensive understanding of pathological mechanism**. *Bone Res* (2017) **5** 16044. PMID: 28149655
2. Hunter DJ, Bierma-Zeinstra S. **Osteoarthritis**. *The Lancet* (2019) **393** 1745-1759
3. Glyn-Jones S, Palmer AJR, Agricola R, Price AJ, Vincent TL, Weinans H. **Osteoarthritis**. *The Lancet* (2015) **386** 376-387
4. Cui A, Li H, Wang D, Zhong J, Chen Y, Lu H. **Global, regional prevalence, incidence and risk factors of knee osteoarthritis in population-based studies**. *EClinicalMedicine* (2020) **29-30** 100587. PMID: 34505846
5. Safiri S, Kolahi AA, Smith E, Hill C, Bettampadi D, Mansournia MA. **Global, regional and national burden of osteoarthritis 1990-2017: a systematic analysis of the Global Burden of Disease Study 2017**. *Ann Rheum Dis* (2020) **79** 819-828. PMID: 32398285
6. Long H, Zeng X, Liu Q, Wang H, Vos T, Hou Y. **Burden of osteoarthritis in China, 1990-2017: findings from the Global Burden of Disease Study 2017**. *The Lancet Rheumatology* (2020) **2** e164-e172
7. Heck JN, Mellman DL, Ling K, Sun Y, Wagoner MP, Schill NJ. **A conspicuous connection: structure defines function for the phosphatidylinositol-phosphate kinase family**. *Crit Rev Biochem Mol Biol* (2007) **42** 15-39. PMID: 17364683
8. K Nagata, Nozawa Y. **Role of GTP-binding proteins in phospholipid metabolism in human platelets**. *Nihon Rinsho* (1992) **50** 223-229
9. Funakoshi Y, Hasegawa H, Kanaho Y. **Regulation of PIP5K activity by Arf6 and its physiological significance**. *J Cell Physiol* (2011) **226** 888-895. PMID: 20945365
10. Nader GP, Ezratty EJ, Gundersen GG. **FAK, talin and PIPKIgamma regulate endocytosed integrin activation to polarize focal adhesion assembly**. *Nat Cell Biol* (2016) **18** 491-503. PMID: 27043085
11. Ling K, Doughman RL, Iyer VV, Firestone AJ, Bairstow SF, Mosher DF. **Tyrosine phosphorylation of type Igamma phosphatidylinositol phosphate kinase by Src regulates an integrin-talin switch**. *J Cell Biol* (2003) **163** 1339-1349. PMID: 14691141
12. Di Paolo G, Pellegrini L, Letinic K, Cestra G, Zoncu R, Voronov S. **Recruitment and regulation of phosphatidylinositol phosphate kinase type 1 gamma by the FERM domain of talin**. *Nature* (2002) **420** 85-89. PMID: 12422219
13. Huang G, Yang C, Guo S, Huang M, Deng L, Huang Y. **Adipocyte-specific deletion of PIP5K1c reduces diet-induced obesity and insulin resistance by increasing energy expenditure**. *Lipids Health Dis* (2022) **21** 6. PMID: 34996482
14. Schramp M, Thapa N, Heck J, Anderson R. **PIPKIgamma regulates beta-catenin transcriptional activity downstream of growth factor receptor signaling**. *Cancer Res* (2011) **71** 1282-1291. PMID: 21303971
15. Schill NJ, Hedman AC, Choi S, Anderson RA. **Isoform 5 of PIPKIgamma regulates the endosomal trafficking and degradation of E-cadherin**. *J Cell Sci* (2014) **127** 2189-2203. PMID: 24610942
16. Rodriguez L, Simeonato E, Scimemi P, Anselmi F, Cali B, Crispino G. **Reduced phosphatidylinositol 4,5-bisphosphate synthesis impairs inner ear Ca2+ signaling and high-frequency hearing acquisition**. *Proc Natl Acad Sci U S A* (2012) **109** 14013-14018. PMID: 22891314
17. Xue J, Ge X, Zhao W, Xue L, Dai C, Lin F. **PIPKIgamma Regulates CCL2 Expression in Colorectal Cancer by Activating AKT-STAT3 Signaling**. *J Immunol Res* (2019) **2019** 3690561. PMID: 31781676
18. Peng JM, Lin SH, Yu MC, Hsieh SY. **CLIC1 recruits PIP5K1A/C to induce cell-matrix adhesions for tumor metastasis**. *J Clin Invest* (2021) **131**
19. Wright BD, Loo L, Street SE, Ma A, Taylor-Blake B, Stashko MA. **The lipid kinase PIP5K1C regulates pain signaling and sensitization**. *Neuron* (2014) **82** 836-847. PMID: 24853942
20. Xu W, Wang P, Petri B, Zhang Y, Tang W, Sun L. **Integrin-induced PIP5K1C kinase polarization regulates neutrophil polarization, directionality, and in vivo infiltration**. *Immunity* (2010) **33** 340-350. PMID: 20850356
21. Loo L, Zylka M. **Conditional deletion of Pip5k1c in sensory ganglia and effects on nociception and inflammatory sensitization**. *Mol Pain* (2017) **13** 1744806917737907. PMID: 29020859
22. Volpatti JR, Al-Maawali A, Smith L, Al-Hashim A, Brill JA, Dowling JJ. **The expanding spectrum of neurological disorders of phosphoinositide metabolism**. *Dis Model Mech* (2019) **12**
23. Gilbert Di Paolo, Howard S Moskowitz, Keith Gipson, Markus R Wenk, Sergey Voronov, Masanori Obayashi. **Impaired PtdIns(4,5)P2 synthesis in nerve terminals produces defects in synaptic vesicle trafficking**. *Nature* (2004) **23** 415-422
24. Zhu T, Chappel JC, Hsu FF, Turk J, Aurora R, Hyrc K. **Type I phosphotidylinosotol 4-phosphate 5-kinase gamma regulates osteoclasts in a bifunctional manner**. *J Biol Chem* (2013) **288** 5268-5277. PMID: 23300084
25. Yan Q, Gao H, Yao Q, Ling K, Xiao G. **Loss of phosphatidylinositol-4-phosphate 5-kinase type-1 gamma (Pip5k1c) in mesenchymal stem cells leads to osteopenia by impairing bone remodeling**. *J Biol Chem* (2022) **298** 101639. PMID: 35090892
26. Wu X, Lai Y, Chen S, Zhou C, Tao C, Fu X. **Kindlin-2 preserves integrity of the articular cartilage to protect against osteoarthritis**. *Nature Aging.* (2022)
27. Wu X, Qu M, Gong W, Zhou C, Lai Y, Xiao G. **Kindlin-2 deletion in osteoprogenitors causes severe chondrodysplasia and low-turnover osteopenia in mice**. *Journal of Orthopaedic Translation* (2022) **32** 41-48. PMID: 34934625
28. Li J, Zhang B, Liu WX, Lu K, Pan H, Wang T. **Metformin limits osteoarthritis development and progression through activation of AMPK signalling**. *Ann Rheum Dis* (2020) **79** 635-645. PMID: 32156705
29. Liu J, Wu X, Lu J, Huang G, Dang L, Zhang H. **Exosomal transfer of osteoclast-derived miRNAs to chondrocytes contributes to osteoarthritis progression**. *Nature Aging* (2021) **1** 368-384
30. Chen S, Wu X, Lai Y, Chen D, Bai X, Liu S. **Kindlin-2 inhibits Nlrp3 inflammasome activation in nucleus pulposus to maintain homeostasis of the intervertebral disc**. *Bone Research* (2022) **10**
31. Lei Y, Fu X, Li P, Lin S, Yan Q, Lai Y. **LIM domain proteins Pinch1/2 regulate chondrogenesis and bone mass in mice**. *Bone Res* (2020) **8** 37. PMID: 33083097
32. Gao H, Zhong Y, Ding Z, Lin S, Hou X, Tang W. **Pinch Loss Ameliorates Obesity, Glucose Intolerance, and Fatty Liver by Modulating Adipocyte Apoptosis in Mice**. *Diabetes* (2021) **70** 2492-2505. PMID: 34380695
33. Qin L, Fu X, Ma J, Lin M, Zhang P, Wang Y. **Kindlin-2 mediates mechanotransduction in bone by regulating expression of Sclerostin in osteocytes**. *Communications Biology* (2021) **4**
34. Gao H, Zhou L, Zhong Y, Ding Z, Lin S, Hou X. **Kindlin-2 haploinsufficiency protects against fatty liver by targeting Foxo1 in mice**. *Nat Commun* (2022) **13** 1025. PMID: 35197460
35. Shane Anderson A, Loeser RF. **Why is osteoarthritis an age-related disease?**. *Best Pract Res Clin Rheumatol* (2010) **24** 15-26. PMID: 20129196
36. Sacitharan PK. **Ageing and Osteoarthritis**. *Subcell Biochem* (2019) **91** 123-159. PMID: 30888652
37. Liu Q, Wang S, Lin J, Zhang Y. **The burden for knee osteoarthritis among Chinese elderly: estimates from a nationally representative study**. *Osteoarthritis Cartilage* (2018) **26** 1636-1642. PMID: 30130589
38. Loeser RF. **Age-related changes in the musculoskeletal system and the development of osteoarthritis**. *Clin Geriatr Med* (2010) **26** 371-386. PMID: 20699160
39. Valdes Ana M, Stocks J. **Osteoarthritis and ageing**. *EMJ* (2018) **3** 116-123
40. Rahmati M, Nalesso G, Mobasheri A, Mozafari M. **Aging and osteoarthritis: Central role of the extracellular matrix**. *Ageing Res Rev* (2017) **40** 20-30. PMID: 28774716
41. Li F, Lu Y, Ding M, Napierala D, Abbassi S, Chen Y. **Runx2 contributes to murine Col10a1 gene regulation through direct interaction with its cis-enhancer**. *J Bone Miner Res* (2011) **26** 2899-2910. PMID: 21887706
42. Komori T. **Runx2, an inducer of osteoblast and chondrocyte differentiation**. *Histochem Cell Biol* (2018) **149** 313-323. PMID: 29356961
43. Wang X, Manner PA, Horner A, Shum L, Tuan RS, Nuckolls GH. **Regulation of MMP-13 expression by RUNX2 and FGF2 in osteoarthritic cartilage**. *Osteoarthritis Cartilage* (2004) **12** 963-973. PMID: 15564063
44. Zhao W, Zhang S, Wang B, Huang J, Lu WW, Chen D. **Runx2 and microRNA regulation in bone and cartilage diseases**. *Ann N Y Acad Sci* (2016) **1383** 80-87. PMID: 27526290
45. Chen D, Kim DJ, Shen J, Zou Z, O'Keefe RJ. **Runx2 plays a central role in Osteoarthritis development**. *J Orthop Translat* (2020) **23** 132-139. PMID: 32913706
46. Loeser RF. **Aging processes and the development of osteoarthritis**. *Curr Opin Rheumatol* (2013) **25** 108-113. PMID: 23080227
47. Sophia Fox AJ, Bedi A, Rodeo SA. **The basic science of articular cartilage: structure, composition, and function**. *Sports Health* (2009) **1** 461-468. PMID: 23015907
48. Zhu M, Tang D, Wu Q, Hao S, Chen M, Xie C. **Activation of β-Catenin Signaling in Articular Chondrocytes Leads to Osteoarthritis-Like Phenotype in Adult β-Catenin Conditional Activation Mice**. *Journal of Bone and Mineral Research* (2009) **24** 12-21. PMID: 18767925
49. Catheline SE, Hoak D, Chang M, Ketz JP, Hilton MJ, Zuscik MJ. **Chondrocyte-Specific RUNX2 Overexpression Accelerates Post-traumatic Osteoarthritis Progression in Adult Mice**. *J Bone Miner Res* (2019) **34** 1676-1689. PMID: 31189030
50. Liao L, Zhang S, Gu J, Takarada T, Yoneda Y, Huang J. **Deletion of Runx2 in Articular Chondrocytes Decelerates the Progression of DMM-Induced Osteoarthritis in Adult Mice**. *Sci Rep* (2017) **7** 2371. PMID: 28539595
51. Wang Z, Huang J, Zhou S, Luo F, Tan Q, Sun X. **Loss of Fgfr1 in chondrocytes inhibits osteoarthritis by promoting autophagic activity in temporomandibular joint**. *J Biol Chem* (2018) **293** 8761-8774. PMID: 29691281
52. Kuang L, Wu J, Su N, Qi H, Chen H, Zhou S. **FGFR3 deficiency enhances CXCL12-dependent chemotaxis of macrophages via upregulating CXCR7 and aggravates joint destruction in mice**. *Ann Rheum Dis* (2020) **79** 112-122. PMID: 31662319
53. Lin C, Liu L, Zeng C, Cui ZK, Chen Y, Lai P. **Activation of mTORC1 in subchondral bone preosteoblasts promotes osteoarthritis by stimulating bone sclerosis and secretion of CXCL12**. *Bone Res* (2019) **7** 5. PMID: 30792936
54. Yang H, Wen Y, Zhang M, Liu Q, Zhang H, Zhang J. **MTORC1 coordinates the autophagy and apoptosis signaling in articular chondrocytes in osteoarthritic temporomandibular joint**. *Autophagy* (2019) 1-18
55. Huang J, Zhao L, Fan Y, Liao L, Ma PX, Xiao G. **The microRNAs miR-204 and miR-211 maintain joint homeostasis and protect against osteoarthritis progression**. *Nat Commun* (2019) **10** 2876. PMID: 31253842
56. Huang W, Jiang D, Wang X, Wang K, Sims CE, Allbritton NL. **Kinetic analysis of PI3K reactions with fluorescent PIP2 derivatives**. *Anal Bioanal Chem* (2011) **401** 1881-1888. PMID: 21789487
57. Chinthalapudi K, Rangarajan ES, Patil DN, George EM, Brown DT, Izard T. **Lipid binding promotes oligomerization and focal adhesion activity of vinculin**. *J Cell Biol* (2014) **207** 643-656. PMID: 25488920
58. Thompson PM, Ramachandran S, Case LB, Tolbert CE, Tandon A, Pershad M. **A Structural Model for Vinculin Insertion into PIP2-Containing Membranes and the Effect of Insertion on Vinculin Activation and Localization**. *Structure* (2017) **25** 264-275. PMID: 28089450
59. Orlowski A, Kukkurainen S, Poyry A, Rissanen S, Vattulainen I, Hytonen VP. **PIP2 and Talin Join Forces to Activate Integrin**. *J Phys Chem B* (2015) **119** 12381-12389. PMID: 26309152
60. Mandal K. **Review of PIP2 in Cellular Signaling, Functions and Diseases**. *Int J Mol Sci* (2020) **21**
61. Le OT, Cho OY, Tran MH, Kim JA, Chang S, Jou I. **Phosphorylation of phosphatidylinositol 4-phosphate 5-kinase gamma by Akt regulates its interaction with talin and focal adhesion dynamics**. *Biochim Biophys Acta* (2015) **1853** 2432-2443. PMID: 26149501
62. Carnero A, Paramio JM. **The PTEN/PI3K/AKT Pathway in vivo, Cancer Mouse Models**. *Front Oncol* (2014) **4** 252. PMID: 25295225
|
---
title: 'Impact of Insulin Degludec/Liraglutide Fixed Combination on the Gut Microbiomes
of Elderly Patients With Type 2 Diabetes: Results From A Subanalysis of A Small
Non-Randomised Single Arm Study'
authors:
- Stefano Rizza
- Daniele Pietrucci
- Susanna Longo
- Rossella Menghini
- Adelaide Teofani
- Giacomo Piciucchi
- Martina Montagna
- Massimo Federici
journal: Aging and Disease
year: 2023
pmcid: PMC10017153
doi: 10.14336/AD.2023.0118
license: CC BY 2.0
---
# Impact of Insulin Degludec/Liraglutide Fixed Combination on the Gut Microbiomes of Elderly Patients With Type 2 Diabetes: Results From A Subanalysis of A Small Non-Randomised Single Arm Study
## Abstract
In elderly Type 2 Diabetes (T2D) patients the relationship between the destabilization of gut microbiome and reversal of dysbiosis via glucose lowering drugs has not been explored. We investigated the effect of 6 months therapy with a fixed combination of Liraglutide and Degludec on the composition of the gut microbiome and its relationship with Quality of Life, glucose metabolism, depression, cognitive function, and markers of inflammation in a group of very old T2D subjects ($$n = 24$$, 5 women, 19 men, mean age=82 years). While we observed no significant differences in microbiome biodiversity or community among study participants ($$n = 24$$, 19 men, mean age 82 years) who responded with decreased HbA1c ($$n = 13$$) versus those who did not ($$n = 11$$), our results revealed a significant increase in Gram-negative Alistipes among the former group ($$p \leq 0.013$$). Among the responders, changes in the Alistipes content were associated directly with cognitive improvement ($r = 0.545$, $$p \leq 0.062$$) and inversely with TNFα levels (r=-0.608, $$p \leq 0.036$$). Our results suggest that this combination drug may have a significant impact on both gastrointestinal microbes and cognitive function in elderly T2D individuals.
## INTRODUCTION
Aging is a major non-reversible risk factor for type 2 diabetes mellitus (T2DM) [1]. Because diabetes management in older patients can be complicated by polypharmacy and cognitive impairment, we recently conducted an interventional pilot clinical study which resulted in broadly-perceived improvements in quality of life and cognitive function when we replaced complex anti-diabetic regimens with a single daily dose of fixed combination of insulin degludec and liraglutide (insulin degludec at 100 units/mL and the glucagon-like peptide 1 receptor agonist [GLP-1RA] liraglutide at 3.6 mg/mL) [2]. While results from several recent studies suggested GLP-1RAs might alter the gut microbiome [3], and others examined the impact of the gut microbiome and its role in modulating depression, cognitive impairment, and T2DM [4], no data are available that address the role(s) of these new agents and their specific impact on the gut microbiomes of elderly T2DM patients. Here we report the results of a metagenomics analysis of gut microbiota in elderly patients treated for six months with a fixed combination of insulin degludec and liraglutide.
## MATERIALS AND METHODS
The findings presented in this study are the result of a sub-analysis of an open, single-arm six-month interventional trial conducted in a real-world setting. All study procedures were performed in compliance with ethical standards for human clinical trials (institutional and national) and with the Declaration of Helsinki of 1964 (revised in 2013). This study was approved by the University Hospital Committee of “Tor Vergata” University (protocol number: $\frac{141}{18}$) and registered at ClinicalTrials.gov ID: NCT04190160. Informed consent was obtained before screening and every procedure performed during the protocol. The study parameters were described extensively in a previous publication [2]. The original study included 35 patients (12 women, 23 men, mean age, 81.4 years) who replaced their pre-existing hypoglycemic therapeutic regimen with a flexibly-timed single daily dose of degludeg and liraglutide fixed combination. The clinical protocol included two ambulatory visits, including one at the beginning of the study (V0) and another six months after changing to degludeg and liraglutide combination (V1). Fecal samples collected at each visit were divided into aliquots and stored at -80?C until analysis.
## Assesment of Quality of Life, cognition, depression, and level of independence
We used Control, Autonomy, Self-Realization, and Pleasure-19 (CASP-19) scale to explore factors that affect QoL at an older age and the Diabetes Treatment Satisfaction Questionnaire (DTSQ) to evaluate the self-reported satisfaction related to a change in diabetes therapy. We also used the Geriatric Depression Scale (GDS) to assess depression symptoms whereas Mini Mental State Examination (MMSE) were used to screen the cognitive function in our study population; change in level of independence was assessed by activities of daily living (ADL) and by instrumental activities of daily living (IADL).
## DNA extraction and 16S rRNA gene sequencing
Total DNA was extracted from 200 mg of each frozen stool sample using the PSP Spin Stool DNA Kit (Stratec Molecular, Berlin, Germany), following the manufacturer's protocol. Briefly, stool samples were lysed under denaturing conditions at 95 °C and then treated with proteinase K at 70 °C. DNA was purified through a spin column system, eluted, and quantified using a NanoDrop spectrophotometer ND1000 (ThermoFisher). Sequencing of 16S rRNA amplicons (V3-V4 regions) was performed using Illumina MiSeq 2x300bp.
**Table 1**
| Unnamed: 0 | Degludec and Liraglutide responders (n=13) | Degludec and Liraglutide responders (n=13).1 | Degludec and Liraglutide responders (n=13).2 | Degludec and Liraglutide non-responders (n=11) | Degludec and Liraglutide non-responders (n=11).1 | Degludec and Liraglutide non-responders (n=11).2 | Unnamed: 7 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Variables | V0 | V1 | *p | V0 | V1 | **p | p*** |
| Age (y) | 82 ± 4.6 | | n.p. | 82.3 ± 4.7 | | n.p. | 0.871 |
| Sex (m/f) | 10/3 | | n.p. | 9/2 | | n.p. | 0.768 |
| BMI (kg/m2) | 28.8 ± 3.8 | 28.0 ± 3..4 | 0.070 | 29.1 ± 3.8 | 28.0 ± 3.3 | 0.001 | |
| Systolic BP (mmHg) | 124.0 ± 11.0 | 127.6 ± 22.3 | 0.511 | 136.4 ± 20.1 | 134.8 ± 19.1 | 0.501 | 0.055 |
| Diastolic BP (mmHg) | 76.7 ± 7.2 | 69.5 ± 6.7 | 0.022 | 72.0 ± 8.8 | 72.7 ± 9.4 | 0.775 | 0.130 |
| Fasting glucose (ng/dl) | 193.6 ± 70.2 | 129.8 ± 27.1 | 0.007 | 153.0 ± 34.2 | 127.6 ± 38.4 | 0.028 | 0.059 |
| HbA1c (%, mmol/mol) | 8.07 ± 1.02, 65.1 ± 12 | 6.60 ± 0.5, 49.0 ± 07 | 0.001 | 7.35 ± 0.82, 57.2 ± 10 | 7.72 ± 1.0, 61.4 ± 12 | 0.009 | 0.047 |
| Total cholesterol (mg/dl) | 152.4 ± 32.3 | 138.8 ± 27.1 | 0.044 | 181.1 ± 28.3 | 168.1 ± 30.5 | 0.249 | 0.021 |
| Creatinine (mg/dl) | 1.14 ± 0.29 | 1.09 ± 0.28 | 0.755 | 1.11 ± 0.62 | 1.18 ± 0.34 | 0.519 | 0.812 |
| e-GFR (ml/min/1.73 m2) | 41.32 ± 5.3 | 43.55 ± 5.1 | 0.322 | 39.2 ± 7.3 | 38.71 ± 6.8 | 0.666 | 0.138 |
| IL-1β (ng/ml) | 0.367 ± 0.802 | 1.611 ± 5.16 | 0.684 | 0.165 ± 0.109 | 0.105 ± 0.130 | 0.057 | 0.218 |
| IL-6 (ng/ml) | 4.97 ± 4.58 | 4.80 ± 5.35 | <0.001 | 7.15 ± 11.05 | 8.86 ± 14.17 | 0.229 | 0.441 |
| TNFα (ng/ml) | 15.37 ± 4.46 | 12.98 ± 4.122 | 0.002 | 16.58 ± 4.67 | 15.15 ± 2.91 | 0.227 | 0.347 |
| MMSE | 23.5 ± 2.8 | 24.7 ± 3.3 | 0.102 | 23.7 ± 3.5 | 24.5 ± 2.2 | 0.220 | 0.911 |
| GDS | 5.9 ± 4.4 | 4.5 ± 3.8 | 0.106 | 4.8 ± 2.6 | 3.7 ± 2.2 | 0.124 | 0.442 |
| DTSQs | 21.7 ± 10.0 | 35.9 ± 4.9 | <0.001 | 27.4 ± 8.1 | 34.7 ± 3.9 | 0.009 | 0.107 |
| CASP-19 | 45.4 ± 6.5 | 42.4 ± 5.2 | 0.126 | 39.8 ± 7.4 | 39.4 ± 6.2 | 0.524 | 0.053 |
## Bioinformatics and statistical analyses
Bioinformatics data were generated as previously described [5]. After completion of the quality checks, reads were clustered using DADA2 and taxonomically assigned using the Silva database vr. 138 in the QIIME2 (pipeline) [6, 7]. Metagenomic data were normalized and analyzed using the statistical approaches described by Cerroni et al. [ 8]. We searched for bacterial taxa with abundances that differed between groups using Repeated Measures ANOVA. This model can be used to identify differences in abundance that are potentially related to [1] degludeg and liraglutide combination treatment (i.e., “responder” versus “non-responder” with respect to reductions in HbA1c), [2] the time of the measurement, or [3] the combined effects of both treatment and time.
## RESULTS
Fecal samples were obtained at baseline (V0) and six months after switching to insulin degludec and liraglutide fixed combination (V1) from 24 of the original 35 study patients. Eleven patients were excluded from the metagenomic analysis because they did not collect sufficient feces samples in visit 2.
Patients exhibiting a $0.3\%$ decrease in HbA1c at V1 compared to V0 were classified as responders ($$n = 13$$); those who did not exhibit at least $0.3\%$ reductions in HbA1c were classified as non-responders ($$n = 11$$). Table 1 provides a detailed comparison of clinical and biochemical characteristics of study patients classified as either responders or non-responders as described above. Notably, depression and cognitive function did not show any significant modification in both groups during the six months protocol. Interestingly, DTSQs increased whereas TNFα levels decreased only in responders (Table 1) After the deletion of low-quality and chimeric reads, amplicon sequence variant (ASV) clustering, and removal of low-abundant taxa, 555201 reads were available for statistical analysis. We identified 2731 ASVs representing 254 species, 193 genera, 53 families, 27 orders, 19 classes, and 12 phyla.
Figure 1.Changes in biodiversity (A) and community (B) of the microbiome composition in patients undergoing insulin degludec and liraglutide fixed combination treatment. ( C) Variations in the abundance of Alistipes species in Responders compared to Non-responders to insulin degludec and liraglutide treatment: baseline (V0) to follow-up (V1); p-value=0.013. Correlation between percentage change in Alistipes and TNF-α levels (r = -0.608, $$p \leq 0.036$$) (D) and MMSE score ($r = 0.545$, $$p \leq 0.062$$) (E) among responders to insulin degludec and liraglutide.
We detected no significant differences at biodiversity and community levels in fecal samples from patients classified as responders and non-responders based on the Shannon index (Fig. 1A) or PERMANOVA (Fig. 1B) as measures of alpha and beta diversity, respectively. We then evaluated our findings with a repeated measure ANOVA to identify taxa whose abundances might differ between responders and non-responders over time. We excluded taxa that exhibited the same trend in each of the two groups as well as those that differed in a time-independent manner. Although this secondary analysis yielded no statistically significant findings, we identified one taxon that exhibited a particularly interesting trend. Specifically, our results revealed an increase in the proportion of bacteria of the *Alistipes genus* in samples from the responders, with a mean number of normalized reads at the baseline (V0) of 468 and a mean number of normalized reads after six weeks of degludeg and liraglutide fixed combination (V1) of 618. By contrast, we observed a decrease in the proportion represented by this genus in samples from patients categorized as non-responders, with a mean number of normalized reads of 477 and 438 at V0 and V1, respectively. We interpreted this result as a trend because the p-value was significant only before the correction for false discovery rate (FDR) was applied ($$p \leq 0.013$$, without FDR correction) (Fig. 1C). Of note, we observed no significant modification in cognitive function explored by MMSE in our participants either in non-responders or in responders (Table 1).
Finally, we generated a correlation matrix to identify any associations between changes observed in the prevalence of this genus and clinical parameters. Linear correlation analysis revealed that increased abundances of Alistipes were directly associated with percentage increases in MMSE scores ($r = 0.43$, FDR=0.0027). Only in responders we also observed a statistically significant inverse association between the percentage change in Alistipes abundance and the percentage change in serum TNFα levels (r=-0.608, $$p \leq 0.036$$; Fig. 1D) and a strong, albeit not a statistically significant, direct association between percentage change in the abundance of Alistipes and percentage change in MMSE scores ($r = 0.545$, $$p \leq 0.062$$, Fig. 1E). On contrast, we did not observe any significant correlation between the change in Alistipes abundance and serum IL-1β and IL6 levels.
## CONCLUSIONS
Several publications have described the critical contributions of the intestinal microbiome to the health and well-being of elderly individuals largely based on its capacity to influence metabolic and digestive functions, as well as depression, cognitive impairment, immunity, and resistance to infections [9-10]. However, the mechanisms underlying these positive responses remain unclear.
The results of the present study revealed that six months of fixed combination of insulin degludec and liraglutide did not affect depression and cognitive function assessed by GDS and MMSE, respectively. Similarly, the drug combination had no impact on the composition of the gut microbiomes of elderly T2DM patients that responded with reductions in HbA1c (i.e., responders). This last fact suggested that drug-dependent changes in serum glucose levels had no direct impact on the gut microbiomes of these patients. Interestingly, our analysis revealed an increase in the percentage of bacteria of the *Alistipes genus* among patients identified as responders. Although this finding was interpreted as a trend because statistical significance was lost after the application of an FDR correction, this may be due in part to the low number of samples analyzed in this study. Thus, we believe that this trend could be of clinical interest and might be evaluated further in future studies.
Alistipes are anaerobic Gram-negative bacteria of the phylum, Bacteroidetes. While these bacteria are commonly found in healthy human intestinal microbiota, recent studies have identified alterations in the abundance of Alistipes in patients with non-communicable disorders, including liver diseases, colorectal cancer, atherosclerosis, and mood disorders [11]. Of particular interest, one recent publication [12] reported that Japanese centenarians exhibited a pronounced abundance of fecal Alistipes compared to younger controls; the authors proposed this finding as a potential marker of successful aging. However, we recognize that our study patients were from a Caucasian population; previous studies have yielded conflicting insights into the implications of changes to the gut microbiome-based specifically on ethnicity [13].
Alistipes produce and release acetate in the form of short-chain fatty acids (SFCAs) [14] that may serve to maintain hepatic energy balance by controlling appetite and mediating glucose homeostasis at the systemic level [15]. Moreover, acetate may limit lipopolysaccharide-induced TNFα secretion via stimulation of free fatty acid receptors on mononuclear cells [16]. Although we cannot exclude potential contributions from an intermediary mechanism, the observed increase in the abundance of Alistipes may lead to a parallel increase in SCFA concentrations and ultimately reductions in neuroinflammation. Consistent with this hypothesis, we observed that increases in the proportion of Alistipes in fecal samples were associated directly with cognitive improvement. Moreover, although the was no association between TNFα levels and the increasing of Alistipes abundances, the entanglement between Alistipes modulation and neuro-inflammation needs to be further investigated (Fig. 1D and E).
It has well known that various drugs could impact on gut microbiomes [17]. In this context, metformin may influence the altered gut microbiota affecting specific pathways such as metalloproteins or metal transporters functions [18]. However, in our study the beneficial effects of insulin degludec [19] and liraglutide [20], both associated with a reduced risk of dementia in frail, older adults with T2DM, may have been blunted in non-responders, potentially due to dysbiosis and/or drug resistance. Similarly, while our findings suggest that the fixed combination of Degludec and Liraglutide might have an impact on the relative distribution of microbes in the gastrointestinal tract microbes in elderly T2DM patients, our study involved only a limited number of subjects. Our hypothesis-generating results will need to be validated in a larger study of patients with similar characteristics.
## Ethical approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the Comitato Etico Policlinico Tor Vergata (protocol number: $\frac{141}{18}$)
## References
1. Jessen F, Amariglio RE, Buckley RF, van der Flier WM, Han Y, Molinuevo JL. **The characterization of subjective cognitive decline**. *Lancet Neurol* (2020) **19** 271-278. PMID: 31958406
2. Rizza S, Piciucchi G, Mavilio M, Longo S, Montagna M, Tatonetti R. **Effect of deprescribing in elderly patients with type 2 diabetes: iDegLira might improve quality of life**. *Biomed Pharmacother* (2021) **144** 112341. PMID: 34678725
3. Rotman Y, Sanyal AJ. **Current and upcoming pharmacotherapy for non-alcoholic fatty liver disease**. *Gut* (2017) **66** 180-190. PMID: 27646933
4. Martins LB, Monteze NM, Calarge C, Ferreira AVM, Teixeira AL. **Pathways linking obesity to neuropsychiatric disorders**. *Nutrition* (2019) **66** 16-21. PMID: 31200298
5. Teofani A, Marafini I, Laudisi F, Pietrucci D, Salvatori S, Unida V. **Intestinal Taxa Abundance and Diversity in Inflammatory Bowel Disease Patients: An Analysis including Covariates and Confounders**. *Nutrients* (2022) **14** 260. PMID: 35057440
6. Chen S, Zhou Y, Chen Y, Gu J. **fastp: an ultra-fast all-in-one FASTQ preprocessor**. *Bioinformatics* (2018) **34** i884-i890. PMID: 30423086
7. Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJ, Holmes SP. **DADA2: High-resolution sample inference from Illumina amplicon data**. *Nat Methods* (2016) **13** 581-3. PMID: 27214047
8. Cerroni R, Pietrucci D, Teofani A, Chillemi G, Liguori C, Pierantozzi M. **Not just a Snapshot: An Italian Longitudinal Evaluation of Stability of Gut Microbiota Findings in Parkinson's Disease**. *Brain Sci* (2022) **12** 739. PMID: 35741624
9. Wu L, Zeng T, Zinellu A, Rubino S, Kelvin DJ, Carru C. **A cross-sectional study of compositional and functional profiles of gut microbiota in Sardinian centenarians**. *mSystems* (2019) **4** e00325-19. PMID: 31289141
10. Bárcena C, Valdés-Mas R, Mayoral P, Garabaya C, Durand S, Rodríguez F. **Healthspan and lifespan extension by fecal microbiota transplantation into progeroid mice**. *Nat. Med* (2019) **25** 1234-1242. PMID: 31332389
11. Parker BJ, Wearsch PA, Veloo ACM, Rodriguez-Palacios A. **The Genus Alistipes: Gut Bacteria with Emerging Implications to Inflammation, Cancer, and Mental Health**. *Front Immunol* (2020) **9** 906
12. Sato Y, Atarashi K, Plichta DR, Arai Y, Sasajima S, Kearney SM. **Novel bile acid biosynthetic pathways are enriched in the microbiome of centenarians**. *Nature* (2021) **599** 458-464. PMID: 34325466
13. Zhang X, Ren H, Zhao C, Shi Z, Qiu L, Yang F. **Metagenomic analysis reveals crosstalk between gut microbiota and glucose-lowering drugs targeting the gastrointestinal tract in Chinese patients with type 2 diabetes: a 6 month, two-arm randomised trial**. *Diabetologia* (2022) **65** 1613-1626. PMID: 35930018
14. Rau M, Rehman A, Dittrich M, Groen AK, Hermanns HM, Seyfried F. **Fecal SCFAs and SCFA-producing bacteria in gut microbiome of human NAFLD as a putative link to systemic T-cell activation and advanced disease**. *United Euro. Gastroenterol J* (2018) **6** 1496-1507
15. He J, Zhang P, Shen L, Niu L, Tan Y, Chen L. **Short-Chain Fatty Acids and Their Association with Signalling Pathways in Inflammation, Glucose and Lipid Metabolism**. *Int J Mol Sci* (2020) **21** 6356. PMID: 32887215
16. Masui R, Sasaki M, Funaki Y, Ogasawara N, Mizuno M, Iida A. **G Protein-Coupled Receptor 43 Moderates Gut Inflammation through Cytokine Regulation from Mononuclear Cells**. *Inflamm Bowel Dis* (2013) **19** 2848-2856. PMID: 24141712
17. Forslund K, Hildebrand F, Nielsen T, Falony G, Le Chatelier E, Sunagawa S. **Disentangling type 2 diabetes and metformin treatment signatures in the human gut microbiota**. *Nature* (2015) **10** 262-266. DOI: 10.1038/nature15766
18. Wu H, Esteve E, Tremaroli V, Khan MT, Caesar R, Mannerås-Holm L. **Metformin alters the gut microbiome of individuals with treatment-naive type 2 diabetes, contributing to the therapeutic effects of the drug**. *Nat Med* (2017) **23** 850-858. PMID: 28530702
19. Strain WD, Morgan AR, Evans M. **The Value of Insulin Degludec in Frail Older Adults with Type 2 Diabetes**. *Diabetes Ther* (2021) **12** 2817-2826. PMID: 34608609
20. Bassil F, Fernagut PO, Bezard E, Meissner WG. **Insulin, IGF-1 and GLP-1 signaling in neurodegenerative disorders: targets for disease modification?**. *Prog Neurobiol* (2014) **118** 1-18. PMID: 24582776
|
---
title: CO-Induced TTP Activation Alleviates Cellular Senescence and Age-Dependent
Hepatic Steatosis via Downregulation of PAI-1
authors:
- Jeongmin Park
- Yingqing Chen
- Jeongha Kim
- Eunyeong Hwang
- Gyu Hwan Park
- Chae Ha Yang
- Stefan W Ryter
- Jeong Woo Park
- Hun Taeg Chung
- Yeonsoo Joe
journal: Aging and Disease
year: 2023
pmcid: PMC10017156
doi: 10.14336/AD.2023.0120
license: CC BY 2.0
---
# CO-Induced TTP Activation Alleviates Cellular Senescence and Age-Dependent Hepatic Steatosis via Downregulation of PAI-1
## Abstract
Aging can increase the risk of various hepatic diseases, especially non-alcoholic fatty liver disease (NAFLD). Although the mechanisms underlying the pathogenesis of age-related disorders such as NAFLD remain incompletely understood, recent studies have implicated the accumulation of senescent cells as a contributing factor. Here, we show that tristetraprolin (TTP) deficiency accelerates NAFLD during aging by enhancing the senescence-associated secretory phenotype (SASP) as well as several hallmarks of senescence. The sequestration of plasminogen activator inhibitor (PAI)-1, a mediator of cellular senescence, in stress granules, (SGs) inhibits cellular senescence. In our previous report, we have shown that carbon monoxide (CO), a small gaseous mediator, can induce the assembly of SGs via an integrated stress response. Here, we show that CO treatment promotes the assembly of SGs which can sequester PAI-1, resulting in the inhibition of etoposide (ETO)-induced cellular senescence. Notably, CO-induced TTP activation enhances PAI-1 degradation, leading to protection against ETO-induced cellular senescence. CO-dependent Sirt1 activation promotes the inclusion of TTP into SGs, leading to decreased PAI-1 levels. Therefore, our findings highlight the importance of TTP as a therapeutic target in age-related NAFLD and offer a potential new strategy to reduce the detrimental effects of senescent cells in hepatic disorders.
## INTRODUCTION
Aging is emerging as the major risk factor for chronic diseases such as neurodegenerative diseases [1, 2], cancer [3, 4], diabetes [5], and cardiovascular disease [6, 7]. Notably, aging has a significant impact on the severity of various hepatic diseases including non-alcoholic fatty liver disease (NAFLD), alcoholic liver disease, hepatitis C, and liver transplantation. Indeed, the human population shows an increased prevalence of NAFLD with age [8, 9]. The mechanisms underlying age-associated NAFLD are, however, not yet known.
Aging is promoted by cellular senescence which is a state of irreversible cell cycle arrest caused by a variety of stressors, including telomere shortening [10], DNA damage [11], epigenetic alteration [12], oxidative stress [13], and mitochondrial dysfunction [14]. Importantly, the accumulation of senescent cells during aging can contribute to aging and aging-related diseases. Several studies have reported that the elimination of senescent cells alleviates the symptoms of aging [15] and various age-related diseases [16]. Senescent cells typically appear flattened, enlarged, and show increased cytoplasmic granularity [17]. In addition, senescent cells also display several other characteristics that differ from proliferating cells. These differences include the increase of senescence-associated β-galactosidase (SA-β-gal) activity [18], increase of phosphorylated H2A histone family member X (γ-H2AX) foci [19], increased expression of cyclin-dependent kinase inhibitors (CDKIs) such as p21CIP1 and p16INK4a [20, 21], as well as senescence associated secreted phenotype (SASP) which consists of growth hormones, pro-inflammatory cytokines, chemokines, angiogenic factors and extracellular matrix (ECM)-remodeling proteases [22, 23]. Recent studies suggest that the increased secretion of serine protease inhibitor, plasminogen activator inhibitor 1 (PAI-1), a component of SASP, can accelerate aging in mice. PAI-1 is a marker and critical mediator of cellular senescence [24]. Furthermore, senescence-inducing signals such as the DNA-damage response (DDR) and oxidative stress can enhance the activation of tumor suppressor p53, which triggers the expression and secretion of PAI-1. In turn, PAI-1 prevents cyclin D1-dependent phosphorylation of Rb, resulting in the irreversible cell cycle arrest [25]. Therefore, the inhibition of cellular senescence may be an attractive therapeutic target in age-related diseases.
In response to diverse environmental stresses, including heat, hyperosmolarity and oxidative stresses, eukaryotic cells temporarily cease protein synthesis to control energy expenditure for the repair of stress-induced damage. One of the major underlying mechanisms is the formation of stress granules (SG) in the cytoplasm. These non-membrane-bound SGs can arrest mRNAs and several harmful proteins to protect cells from apoptosis [26, 27]. SG biogenesis is recognized as a conserved stress response, which can be initiated by the oligomerization of Ras GTPase-activating protein-binding protein-1 (G3BP1) and aggregation of RNA binding proteins, including T-cell intracytoplasmic antigen (TIA-1), TIA1 related protein (TIAR) and HuR [28]. Notably, tristetraprolin (TTP), an AU-rich element (ARE)-containing mRNA binding protein, is excluded from SGs through activation of p38 mitogen activated protein kinase (p38 MAPK)/MAPK-activated protein kinase 2 (MK2) cascade [29]. Furthermore, SGs formation can inhibit cellular senescence via sequestration of PAI-1, and subsequently enhance the cyclin D1 pathway to remove cell cycle arrest [30].
Carbon monoxide (CO) is an endogenous gaseous mediator that is produced from heme by the activation of heme oxygenase-1 (HO-1), a stress-inducible response. When applied at low concentration, CO can exert cyto- and tissue- protective effects in various models of cellular and tissue injuries, involving anti-inflammatory, antioxidant, and anti-apoptotic effects [31]. Intriguingly, our recent study has demonstrated that CO can induce the formation of SGs through protein kinase RNA-like endoplasmic reticulum kinase (PERK)-eIF2α signaling pathway, a component of the integrated stress response (ISR) [32]. In addition, CO promotes the increase of TTP levels and its activation by regulation of phosphorylation and acetylation [33]. In this study, we found that CO promotes the sequestration of PAI-1 in SGs and CO-induced TTP activation enhances PAI-1 degradation in SG assembly. Finally, we suggest that TTP may present a new target molecule in age-related NAFLD.
## Reagents
CO-releasing molecule 2 (CORM2) and etoposide (ETO) were from Sigma-Aldrich (St Louis, MO, USA).
## Animals
TTP KO mice (Ttp-/-), in C57BL/6 background, were kindly provided by Dr. Perry J. Blackshear (Laboratory of Signal Transduction, National Institute of Environmental Health Sciences, USA). All mice were bred in the animal facility at the University of Ulsan and were born and housed under specific pathogen-free conditions at 18-24? and 40-70 % humidity, with a 12 h light-dark cycle. Animal studies were approved by the University of Ulsan Animal Care and Use Committee (Reference number HTC-19-020). To study liver aging, at the age of 10, 24, and 96 weeks, mice were anesthetized with intraperitoneal Avertin (250 mg/kg, Sigma-Aldrich), and liver tissues and serum from WT and Ttp-/- male and female mice were collected for various assays.
## Cell culture
The human diploid cell line WI-38 and mouse liver cell line AML12 were cultured in Minimum Essential Medium (MEM, GIBCO, Grand Island, USA) and DMEM/F12 (GIBCO, Grand Island, USA), respectively, with $10\%$ fetal bovine serum (FBS, GIBCO, Melbourne, Australia) and $1\%$ penicillin-streptomycin (GIBCO) solution. Primary MEFs were isolated from E14.5 C57BL/6 embryos [34], the products of the mating of TTP heterozygous mice, and cultured in DMEM (GIBCO) medium with $10\%$ FBS, $1\%$ penicillin-streptomycin, and $1\%$ MEM non-essential amino acid solution (GIBCO). The genotypes from each litter were determined by assessment of genomic DNA from each embryo. Primary hepatocytes were isolated from Ttp-/- mice at the age of 24 and 96 weeks as previously described [35]. The liver tissues were perfused with Ca2+ and Mg2+-free Hanks’ buffered salt solution (HBSS, GIBCO), followed by perfusion with $0.2\%$ collagenase type IV in Williams’ Medium E (GIBCO). The hepatocytes were cultured with DMEM (GIBCO) medium containing $10\%$ FBS and $1\%$ penicillin-streptomycin. Cells were grown at 37? in humidified incubators containing an atmosphere of $5\%$ CO2.
## Transfection with siRNA
To knock down the mRNA expression of PAI-1 and TTP, cells were transfected with scramble siRNA (scRNA) (Ambion, Austin, TX, USA), used as negative control, siRNA against human PAI-1, mouse PAI-1, and mouse TTP (Santa Cruz Biotechnology, CA, USA) by applying Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol.
## RNA isolation and Reverse Transcription-Polymerase Chain Reaction
Total RNA was isolated from cells and liver tissues by utilizing QIAzol Lysis reagent (QIAGEN, Valencia, CA, USA), according to the manufacturer’s instructions. In brief, 2 μg of total RNA was used to generate cDNA using M-MLV reverse transcriptase (Promega, Madison, WI, USA). The synthesized cDNA was subject to PCR-based amplification. The following primers were used: mouse GAPDH (F-AGGCCGGTGCTGAGTATGTC, R-TGCC TGCTTCACCTTCT), mouse TTP (F-CTCTGCCATCT ACGAGAGCC, R-GATGGAGTCCGAGTTTATGTTC C), and mouse PAI-1 (F-GACGTTGTGGAACTGC CCTA, R-GACCTTTTGCAGTGCCTGTG). To perform quantitative real-time PCR (qRT-PCR), the synthesized cDNA was amplified with SYBR Green qPCR Master Mix (Applied Biosystems, Foster City, CA, USA) on ABI 7500 Fast Real-time PCR system (Applied Biosystems). The following primers were mouse GAPDH (F-GGGAAGCCCATCACCATCT, R-CGGCCTCACCCC ATTTG), mouse p21 (F-GTGGCCTTGTCGCTGT CTT, R-GCGCTTGGAGTGATAGAAATCTG), mouse p16 (F-AATCTCCGCGAGGAAAGC, R-GTCTGCAGCGG ACTCCAT), mouse PAI-1 (F-ACTGTCCTATCTCAA GGTCCACTGT, R-TGATCTGTCTATCCGTTG CCC), mouse TNF-α (F-AGACCCTCACACTCAGATCATC TTC, R-TTGCTACGACGTGGGCTACA), mouse IL-6 (F-CCAGAGATACAAAGAAATGATGG, R-ACTCC AGAAGACCAGAGGAAAT), mouse IL-1β (F-TCGC TCAGGGTCACAAGAAA, R-ATCAGAGGCAAGGA GGAAACAC), human GAPDH (F-CAATGACCCCTT CATTGACCTC, R-AGCATCGCCCCAC TTGATT), human p21 (F-CGATGGAACTTCGACTTT GTCA, R-GCACAAGGGTACAAGACAGTG), human PAI-1 (F-TGATGGCTCAGACCAACAAG, R-CAGCAATGAA CATGCTGAGG), human TNF-α (F-GCTGCACTTTG GAGTGATCG, R-GTTTGCTACAACATGGGCTACA G), human IL-6 (F-ACTCACCTCTTCAGAACGAATT G, R-CCATCTTTGGAAGGTTCAGGTTG) human IL-1β (F-TTACAGTGGCAATGAGGATGAC, R-GTCGG AGATTCGTAGCTGGAT).
## Western blotting
Lysates of cells and harvested liver tissues were prepared using RIPA buffer (Thermo Scientific, Waltham, MA, USA) containing protease inhibitor (Sigma-Aldrich), phosphatase inhibitor cocktail 2 (Sigma-Aldrich), and phosphatase inhibitor cocktail 3 (Sigma-Aldrich). Total protein concentration of the lysates was measured using a BCA protein assay kit (Pierce Biotechnology, Rockford, IL, USA). Proteins were resolved by SDS-PAGE, transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, Burlington, MA, USA), and probed with appropriate dilutions of the following antibodies: p53 (sc-6243, 1:1000, Santa Cruz), p21 (ab109199, 1:1000, Abcam, Cambridge, MA, USA), PAI-1 (sc-5297, 1:1000, Santa Cruz), TTP (T5327, 1:2000, Sigma-Aldrich), and α-tubulin (2125S, 1:1000, cell signaling, Danvers, MA, USA). Then, membranes were incubated with secondary antibodies (115-035-003; HRP-Goat Anti-Mouse IgG, 111-035-003; HRP-Goat Anti-Rabbit IgG) at room temperature for 30 min. Antibody binding was visualized with an ECL chemiluminescence system (Pierce Biotechnology) and chemiluminescence signal was read by Azure Biosystems C300 analyzer (Azure Biosystems, Dublin, CA). The relative band density was analyzed by using ImageJ2x software (US National Institutes of Health, Bethesda, USA).
## Enzyme-Linked Immunosorbent Assays (ELISA)
Cultured supernatant and mouse serum were collected, and the concentration of PAI-1 was measured by using a PAI-1 ELISA kit (BD Biosciences, San Jose, CA, USA), according to the manufacturer’s instructions. The concentration of pro-inflammatory cytokines, TNF-α, IL-6, and IL-1β, were analyzed in conditioned medium and measured by BioLegend ELISA kits (BioLegend, San Diego, CA, USA).
## Senescence-associated β-galactosidase staining
To observe the senescent cells, WI-38 and MEF cells were treated with etoposide to construct DNA damage induced premature cellular senescence. Then, senescence-associated (SA)-β-galactosidase (gal) staining was performed by utilizing a cellular senescence cell histochemical stain kit (Sigma-Aldrich) according to the manufacturer’s protocol. Briefly, after treatment, cells were washed with PBS and fixed with $4\%$ paraformaldehyde, and SA-β-gal was stained by treatment with staining mixture. Five images of different sites per each well plate were obtained, and SA-β-gal-stained cells were counted. The percentage of senescent cells were analyzed by dividing the number of stained cells by the total number of cells.
## Immunofluorescence
Liver tissues, WI-38 cells, primary MEFs, and AML12 cells were plated on 4-well Lab-Tek chambered coverglass (Thermo Scientific, Waltham, MA, USA). After treatment, cells were washed in PBS, fixed with $4\%$ (v/v) paraformaldehyde in PBS at room temperature for 15 min and permeabilized with $0.1\%$ (v/v) Triton X-100 in PBS for 5min. Then, cells were washed three times with PBS and blocked with $3\%$ BSA. To observe the formation of SGs and the sequestration of PAI-1, the samples were incubated with anti-TIA-1(ab40693, 1:500, Abcam), anti-G3BP1 (sc-365338, 1:200, Santa Cruz), anti-PAI-1 (sc-5297, 1:200, Santa Cruz) for 2 h at room temperature. To analyze senescence, cells were incubated with anti-γ-H2AX (05-636, 1:200, Millipore) for 2 h at room temperature. Cells were further washed three times with PBS before incubation for 1 h with Alexa Fluor 594 goat anti-rabbit IgG (A-11037, 1:500, Invitrogen) and Alexa Fluor 488 rabbit anti-mouse IgG (A-11059, 1:500, Invitrogen), respectively. Secondary antibodies were diluted in $1\%$ (w/v) BSA in 1 x PBS. After incubation, cells were washed three times with PBS and stained with 1mg/ml DAPI (Sigma-Aldrich) for 15 min. Representative images were obtained using an Olympus FV1200 confocal microscope (Olympus, Tokyo, Japan). Rabbit IgG (ab172730, 1:500, Abcam) and mouse IgG1 (ab280974, 1:500, Abcam) were used as a negative control. The percentage of cells showing co-localization of SGs, PAI-1, and TTP, were determined. Images were analyzed for the number of cells with γ-H2AX foci. Each field contained at least 20 cells and three images per condition were analyzed.
## Luciferase activity
The 3’-UTR cDNA of human PAI-1 was PCR amplified using Taq polymerase (Bioneer, Daejeon, Korea). The following primers were used 5’-CCGCTCGAGC CCTGGGGAAAGACGCCTTCATCT-3’ AND 5’-AT TTGCGGCCGCGCTTCTATTAGATTACATTCATTT-3’. Plasmid psiCHECK2-PAI-1 3’UTR was generated by inserting the PCR product into the Xho-I and Not-I sites of the psiCHECK2 plasmid (Promega). To evaluate TTP-induced degradation of PAI-1 3’-UTR, cells were transfected with psiCHECK2-PAI-1 3’-UTR for 36 h and then treated with 20 and 40 μM CORM2 for 6 h. Luciferase activity was measured using a Dual-Luciferase Reporter Assay System (Promega) and a SpectraMax iD3 (Molecular Devices, Sunnyvale, CA, USA).
## Measurement of triglycerides
Hepatic triglycerides (TGs) were measured using a TG colorimetric assay kit (Cayman Chemical, Ann Arbor, MI, USA). Briefly, 50 mg samples of liver tissue were homogenized in 200 μl diluted standard diluents. After centrifugation, supernatants were obtained and were used for the assay.
## H&E staining
Liver tissues were fixed in $10\%$ neutral-buffered formalin solution (Sigma-Aldrich) and sectioned with a cryostat at 5 μm. Tissue sections were mounted on regular glass slides and deparaffinized in xylene and rehydrated in graded alcohol series (anhydrous ethanol, $85\%$ ethanol, $75\%$ ethanol), then stained with hematoxylin for 3 min and eosin for 30 seconds.
## Hepatic damage assay
Activity of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in serum, as indicators of hepatic injury, were measured using the EnzyChrom ALT assay kit and EnzyChrom AST assay kit from BioAssay Systems (Hayward, CA, USA).
## Statistical analysis
All data were expressed as mean ± SD, which is representative of at least 3 independent experiments with a minimum of 3 biological replicates. The Shapiro-Wilk test was used to normality test of the data. Statistical significance between two groups were assessed by the Student’s t test (passed normality test) or the non-parametric Mann-Whitney U test (did not pass normality test or n < 6). To analyze the three or more groups, one-way analysis of variance (ANOVA) with repeated measures followed by Tukey post hoc test was performed for normally distributed data, and the Kruskal-Wallis test followed by the Dunn post hoc test was used to analyze non-normally distributed data. To analyze differences between WT mice and Ttp-/- mice, data were evaluated by two-way ANOVA with Bonferroni post-tests. All statistical analysis were assessed by GraphPad Prism software version 9.3.1 (San Diego, CA, USA). The statistically significant changes among groups were considered as probability values of p ≤ 0.05.
Figure 1.TTP attenuates aging-related hepatic dysfunction in mice. ( A) Levels of serum ALT and AST were measured in WT (Ttp+/+) and TTP KO (Ttp-/-) mice ($$n = 3$$ mice in each group) at the age of 10, 24, and 96 weeks. ( B) Representative H&E-stained liver sections from Ttp+/+ and Ttp-/- mice at 24 weeks and 96 weeks of age. Scale bar: 100 μm. ( C) Levels of liver triglycerides were measured in Ttp+/+ mice and Ttp-/- mice ($$n = 3$$ mice in each group) at 10, 24, and 96 weeks of age. ( D-H) The mRNA expression of senescence-associated secretory phenotype (SASP)-related genes, (D) TNF-α, (E) IL-1β, and (F) IL-6, and cellular senescence markers, (G) p21 and (H) p16, were assessed by qRT-PCR in liver tissues from Ttp+/+ and Ttp-/- mice at the indicated ages. ( A-H) Data were analyzed using the two-way ANOVA followed by Bonferroni post-test and expressed as the mean ± SD; *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$, and NS, not significant. ( I) Liver sections at the age of 24 and 96 weeks were stained with anti-p21 antibody. Images of p21 immunofluorescence were detected by confocal microscopy, scale bar: 10 μm (left), and quantification of fluorescence intensity was analyzed (right). ( J) γ-H2AX nuclear foci in liver sections were determined by immunofluorescence, scale bar: 5 μm (left). The number of γ-H2AX nuclear foci was counted (right). Rabbit IgG and mouse IgG1 were used as a negative control of anti-p21 antibody and anti-γ-H2AX antibody, respectively. ( I, J) Data were analyzed using the Mann-Whitney U test and expressed as means ± SD; $$n = 5$$ biological replicates; **$p \leq 0.01$ and ***$p \leq 0.001.$
## TTP attenuates aging-related hepatic dysfunction in mice
To explore the potential role of TTP in aging-related hepatic steatosis, we first measured serum ALT and AST levels as markers of liver damage in aging wild type (WT, Ttp+/+) or TTP knockout (KO, Ttp-/-) mice (Fig. 1A). The serum ALT and AST levels in middle-aged (24 weeks) or aged (96 weeks) Ttp-/- mice were higher than in young (10 weeks) Ttp-/- mice. There was no difference between the serum ALT and AST levels middle-aged (24 weeks) vs. young (10 weeks) WT mice. However, as expected, there was a marked increase of hepatic damage markers in extremely old mice (96 weeks), relative to young mice of both strains, with a trend toward higher levels in Ttp-/- mice. We observed severe liver damage including steatosis and inflammation in the liver of aged Ttp-/- mice relative to WT mice, using H&E staining (Fig. 1B). As shown in Fig. 1C, the levels of liver triglycerides (TGs) were increased in aged WT mice relative to young WT mice; but were not elevated in aged Ttp-/- mice relative to younger Ttp-/- mice. These results suggest that severe inflammation in aged Ttp-/- mice may prevent an increase in liver TGs due to inflammation-induced hepatocyte cell death. In addition, we measured several SASPs, including the cytokines TNF-α, IL-1β, and IL-6, and the cell cycle regulators of p21 and p16. Old (96 weeks) Ttp-/- mice displayed an increase in the levels of the SASPs; TNF-α, IL-1β, and IL-6 in the liver (Fig. 1D, 1E, and 1F) relative to younger Ttp-/- mice, and higher values relative to old WT mice. In old mice, p21 and p16 were highly expressed in both WT and Ttp-/- mice, with p16 displaying higher levels in the Ttp-/- mice (Fig. 1G and 1H). To investigate the role of TTP in aging, we measured p21 and the number of γ-H2AX foci per cell in the livers of WT and Ttp-/- mice. Both p21 staining (Fig. 1I) and the number of γ-H2AX foci (Fig. 1J) were significantly increased in the liver of Ttp-/- mice compared to the liver of WT mice, in both middle-aged mice (24 weeks) (left panels) and aged mice (96 weeks) (right panels). Taken together, we suggest that TTP can prevent liver injury and hepatocyte cell senescence during aging.
## TTP deficiency facilitates age-dependent senescence via increasing PAI-1 expression in the liver
To investigate whether TTP is involved in aging and senescence, primary hepatocytes from WT and Ttp-/- mice were stained with SA-β-gal. The percentage of SA-β-gal positive cells significantly increased in aged mice of both strains (96 weeks) compared to middle-aged mice (24 weeks). Aged Ttp-/- mice (96 weeks) displayed more SA-β-gal positive cells compared to aged WT mice (Fig. 2A). Given that the expression of the PAI-1 gene is markedly stimulated in various aging-associated pathologies [36], we investigated whether PAI-1 levels are regulated during aging in a TTP-dependent manner. The levels of PAI-1 protein were elevated with increased age in WT mice; PAI-1 protein levels were increased in Ttp-/- mice relative to WT mice in all age groups (Fig. 2B). We also measured the levels of the senescence marker p21 to determine whether an increase of PAI-1 in Ttp-/- mice was associated with changes in p21 status. The levels of p21 were increased in an aged-dependent manner in Ttp-/- mice (Fig. 2B). Consistent with results observed in male mice, age-dependent increases of PAI-1 and p21 levels were also observed in female Ttp-/- mice (Fig. 2B). In addition, TTP deficiency was associated with an increase in mRNA expression of PAI-1 in all age groups (Fig. 2C). Secreted PAI-1 levels were also increased in Ttp-/- mice compared to WT mice (Fig. 2D). Therefore, TTP may prevent age-associated senescence phenotypes via decreasing PAI-1 levels.
## CO inhibits etoposide-induced cellular senescence in human and murine fibroblasts
Several studies have reported that topoisomerase inhibitors, such as ETO, doxorubicin and topotecan, which are commonly used as chemotherapeutic agents, can induce DNA double strand breaks (DSBs) in tumor cells, and these lesions can be toxic to normal cells [37]. In addition, these drugs are reported as potent inducers of premature senescence in normal human fibroblasts via activating p53 [38].
Figure 2.TTP deficiency facilitates age-dependent senescence via increasing PAI-1 expression in the liver. ( A) Bright field microscopy images of SA-β-gal staining of primary hepatocyte isolated from WT (Ttp+/+) mice and TTP KO (Ttp-/-) mice at the age of 24 and 96 weeks (left). Cells were analyzed to calculate the percentage of SA-β-gal-positive cells (right). Data were analyzed using the two-way ANOVA followed by Bonferroni post-test and expressed as the mean ± SD; $$n = 5$$ biological replicates; **$p \leq 0.01$ and ***$p \leq 0.001.$ ( B) In liver tissues from WT and Ttp-/- male (left) and female (right) mice ($$n = 3$$ mice in each group), the expression levels of PAI-1 and p21 were determined by immunoblot analysis at the indicated ages. ( C, D) The expression levels of (C) hepatic mRNA and (D) serum levels of PAI-1 were analyzed by qRT-PCR and ELISA, respectively, in Ttp+/+ and Ttp-/- mice at the ages of 10 (left), 24 (middle), and 96 (right) weeks. Data were analyzed using the Mann-Whitney U test and expressed as means ± SD; $$n = 3$$ mice in each group; *$p \leq 0.05$ and **$p \leq 0.01.$
Figure 3.CO inhibits ETO-induced cellular senescence in human and murine fibroblasts. ( A-I) WI-38 cells were pretreated with CORM2 (40 μM) for 6 h followed by stimulation with ETO (80 μM) for 24 h, and then cells were changed into fresh media. During the process of senescence, cells were treated with CORM2 (40 μM) for 6 h every two days, and after 7-days incubation, SA-β-gal staining was performed. ( A) Representative images of SA-β-gal staining (left), scale bar: 20 μm. Quantification of the SA-β-gal positive cells is shown in the right panel (mean ± SD; $$n = 5$$ biological replicates; ***$p \leq 0.001$; Kruskal-Wallis test followed by the Dunn post hoc test). ( B) Immunofluorescence for detecting γ-H2AX foci was performed (mean ± SD; $$n = 10$$ biological replicates; ***$p \leq 0.001$; one-way ANOVA followed by Tukey post hoc test; Scale bar: 20 μm). Mouse IgG1 was used as a negative control of anti-γ-H2AX antibody. The mRNA expression levels of (C) p21, (D) IL-6, (E) TNF-α and (F) IL-1β were measured by qRT-PCR. The secretion levels of (G) IL-6, (H) TNF-α, and (I) IL-1β were measured by ELISA in cell culture supernatants. ( C-I) Data were analyzed using Kruskal-Wallis test followed by the Dunn post hoc test and expressed as means ± SD; $$n = 5$$ biological replicates; ***$p \leq 0.001.$ ( J-R) Primary MEFs were pretreated with CORM2 (40 μM) for 6 h, and then cells were treated with ETO (2 μM) for 4 days. During the process of senescence, MEFs were treated with CORM2 (40 μM) for 6 h every two days. After 4-days incubation, cells were subjected with (J) SA-β-gal staining (mean ± SD; $$n = 4$$ biological replicates; ***$p \leq 0.001$; Kruskal-Wallis test followed by the Dunn post hoc test; Scale bar: 20 μm) and (K) stained with anti-γ-H2AX antibody for assessing γ-H2AX foci (mean ± SD; $$n = 10$$ biological replicates; ***$p \leq 0.001$; one-way ANOVA followed by Tukey post hoc test; Scale bar: 20 μm). Mouse IgG1 was used as a negative control of anti-γ-H2AX antibody. ( L-O) The mRNA expression of (L) p21, (M) IL-6, (N) TNF-α, and (O) IL-1β were evaluated by qRT-PCR. ( P-R) The levels of secreted (P) IL-6, (Q) TNF-α, and (R) IL-1β were detected by ELISA. ( L-R) Data were analyzed using Kruskal-Wallis test followed by the Dunn post hoc test and expressed as means ± SD; $$n = 3$$ biological replicates; ***$p \leq 0.001.$
Figure 4.PAI-1 is involved in ETO-induced premature senescence in WI-38 and MEF cells. ( A) WI-38 cells were treated with ETO (80 μM) for 24 h and then cells were refed with fresh media. After 7-day incubation, cells were harvested and the mRNA expression of PAI-1 was measured by qRT-PCR (mean ± SD; $$n = 3$$ biological replicates; ***$p \leq 0.001$; Mann-Whitney U test). ( B-H) WI-38 cells were transfected with scramble siRNA (scRNA) and siRNA against PAI-1 (siPAI-1) for 36 h and then (B) PAI-1 mRNA level was assessed by qRT-PCR (mean ± SD; $$n = 3$$ biological replicates; ***$p \leq 0.001$; Mann-Whitney U test). Transfected cells were treated with ETO (80 μM) for 24 h and then cells were refed with fresh media. ( C) After 7-day incubation, cells were subjected to SA-β-gal staining. Quantification of SA-β-gal-positive cells was shown in the right panel (mean ± SD; $$n = 4$$ biological replicates; ***$p \leq 0.001$; Kruskal-Wallis test followed by the Dunn post hoc test; Scale bar: 20 μm). ( D-G) The mRNA expression of (D) p21, (E) IL-6, (F) TNF-α and (G) IL-1β were measured by qRT-PCR. Data were analyzed using Kruskal-Wallis test followed by the Dunn post hoc test and expressed as means ± SD; $$n = 3$$ biological replicates; ***$p \leq 0.001.$ ( H) The levels of protein expression of PAI-1, p21, and p53 were assessed by immunoblotting.
We investigated whether low dose CO exhibits inhibitory effects on the pro-senescence effects of ETO. We pre-treated the human fibroblast like fetal lung cell line, WI-38, with CORM, a CO-releasing molecule 2. At various concentrations (0, 10, 20, and 40 μM) for 6 h followed by the administration of ETO (80 μM) for 24 h. Then, the cells were cultured in fresh media, and were post-treated with CORM2 for 6 h every two days. After 7-days of incubation, we found that ETO alone significantly increased the number of cells positive for the expression of p21 and several SASPs, including IL-6, TNF-α, and IL-1β. However, treatment of WI-38 cells with low doses of CORM2 (20 and 40 μM) significantly reduced ETO-stimulated levels of p21 and SASPs, indicating that CORM2 may exert an anti-senescent effect on ETO-induced premature senescence (Supplementary Fig. 1A-1D). Given that 40 μM CORM2 was the optimal dose to reduce the levels of p21 and various SASPs, in the following studies, we chose 40 μM CORM2 as the appropriate dose to treat cells. To further assess the anti-senescent effect of CO, we treated WI-38 cells (Fig. 3A-3I) and primary mouse embryonic fibroblasts (MEFs) (Fig. 3J-3R) with or without CORM2 prior to the administration of ETO, and found that the enhanced markers of cellular senescence, including the percentage of senescence associated (SA)-β-gal positive cells (Fig. 3A and 3J), γ-H2AX foci (Fig. 3B and 3K), mRNA levels of p21 (Fig. 3C and 3L), IL-6 (Fig. 3D and 3M), TNF-α (Fig. 3E and 3N), and IL-1β (Fig. 3F and 3O); and secreted protein levels of IL-6 (Fig. 3G and 3P), TNF-α (Fig. 3H and 3Q), and IL-1β (Fig. 3I and 3R), were all significantly decreased by treatment with CORM2. To investigate whether exogenous CO gas can also protect against cellular senescence, we exposed cells to 250 ppm CO for 6 h every two days in the presence or absence of ETO. CO gas dramatically inhibited the corresponding senescence markers increased by ETO treatment (Supplementary Fig. 1E-1J). These results strongly suggested that CO can effectively prevent ETO-induced premature senescence.
Figure 5.TTP is required for the inhibition of senescence by CO through PAI-1 downregulation. ( A-E) WT (Ttp+/+) and TTP KO (Ttp-/-) primary MEFs were pretreated with CORM2 (40 μM) for 6 h and then the cells were treated with ETO (2 μM) for 4 days. During the process of senescence, MEFs were treated with CORM2 (40 μM) for 6 h every two days. ( A) After 4 days of incubation, cells were stained with SA-β-gal. Scale bar: 20 μm (left). Cells were analyzed to calculate the percentage of SA-β-gal-positive cells (right; mean ± SD; $$n = 5$$ biological replicates; ***$p \leq 0.001$ and NS, not significant; two-way ANOVA followed by Bonferroni post-test). ( B) Cells were performed with immunofluorescence for detecting γ-H2AX foci. Scale bar: 20 μm (left). Mouse IgG1 was used as negative control of anti-γ-H2AX antibody. The number of γ-H2AX nuclear foci was counted (right; mean ± SD; $$n = 10$$ biological replicates; ***$p \leq 0.001$ and NS, not significant; two-way ANOVA followed by Bonferroni post-test). The levels of mRNA expression of (C) p21, (D) IL-6, and (E) PAI-1 were detected by qRT-PCR. ( C-E) Data were analyzed using the two-way ANOVA followed by Bonferroni post-test and expressed as the mean ± SD; $$n = 3$$ biological replicates; ***$p \leq 0.001$ and NS, not significant. ( F) mRNA expression of PAI-1 and TTP in AML12 cells treatment with ETO (0, 5, 10, and 20 μM) for 4 days (upper panel) or treatment with 20 μM ETO in the presence or absence 40 μM CORM2 (lower panel). ( G) mRNA levels of PAI-1 and TTP in primary MEFs treatment with ETO (0, 0.5, 1, and 2 μM) for 4 days (upper panel) or treatment with 40 μM CORM2 for 6 h followed by 20 μM ETO treatment (lower panel). ( H) The levels of protein expression of PAI-1 in primary MEFs treatment with ETO (0, 0.5, 1, and 2 μM) for 4 days. ( I) Protein levels of PAI-1 and TTP in pretreated with 40 μM CORM2 for 6 h, followed by treatment with 2 μM ETO for 4 days. ( J) Secretion levels of PAI-1 in WI-38 cells were measured by ELISA in the indicated groups (mean ± SD; $$n = 3$$ biological replicates; ***$p \leq 0.001$; Kruskal-Wallis test followed by the Dunn post-hoc test). ( K) AML12 cells were transfected with scramble siRNA (scRNA) and siRNA against TTP (siTTP) for 36 h and then treated with ETO (20 μM) for 4 days. mRNA levels of TTP and PAI-1 were assessed by RT-PCR. ( L, M) The mRNA levels of PAI-1 (L) and protein levels of PAI-1 and TTP (M) in Ttp+/+ and Ttp-/- primary MEFs treated with 2 μM ETO for 4 days. ( N) Stability of PAI-1 expression at the indicated time points after actinomycin D (5 μg/ml) in Ttp+/+ and Ttp-/- primary MEFs treated with 40 μM CORM2 for 6 h. (O, P) Luciferase activity in (O) Ttp+/+ and Ttp-/- primary MEFs, (P) AML12, and HEK293 cells transfected with a psi-CHECK2-PAI-1 3’-UTR construct, followed by treatment with CORM2 (20 and 40 μM) for 6 h. (O) Data were analyzed using the two-way ANOVA followed by Bonferroni post-test and expressed as the mean ± SD; $$n = 3$$ biological replicates; **$p \leq 0.01$ and NS, not significant. ( P) Data were analyzed using Kruskal-Wallis test followed by the Dunn post hoc test and expressed as means ± SD; $$n = 3$$ biological replicates; *$p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001.$
## PAI-1 mediates ETO-induced premature senescence in WI-38 and MEF cells
PAI-1 is a primary inhibitor of tissue type and urokinase type plasminogen activators, which convert plasminogen into plasmin, a serine proteinase that plays a major role in fibrinolysis [39]. Besides inhibition of fibrinolysis, several lines of evidence suggest that PAI-1 expression is increased in senescent cells and that PAI-1 is not only a marker but also a key mediator of cellular senescence and organismal aging [24]. To determine whether PAI-1 is increased in ETO-treated cells, we first assessed the mRNA expression of PAI-1 in WI-38 cells and primary MEFs. Our results showed that the expression of PAI-1 was significantly increased in both cell types after the administration of ETO (Fig. 4A and Supplementary Fig. 2A). Next, to evaluate whether increased PAI-1 expression is responsible for ETO-induced premature senescence, we transfected cells with siRNA against PAI-1 for 36 h (Fig. 4B and Supplementary Fig. 2B), and then the cells were stimulated with ETO. Silencing PAI-1 dramatically decreased markers of cellular senescence, including the percentage of SA-β-gal-stained cells (Fig. 4C and Fig. 2C), and the mRNA expressions of p21 (Fig. 4D and Supplementary Fig. 2D), IL-6 (Fig. 4E and Supplementary Fig. 2E), TNF-α (Fig. 4F and Supplementary Fig. 2F), and IL-1β?Fig. 4G and Supplementary Fig. 2G) compared to cells transfected with scramble RNA. PAI-1 depletion abolished the ETO-induced increase in the protein levels of p21 and p53 in WI38 cells (Fig. 4H) and primary MEF cells (Supplementary Fig. 2H). These results suggest that ETO-induced cellular senescence is regulated by PAI-1 levels.
## TTP is required for inhibition of senescence by CO via downregulation of PAI-1
We demonstrated that TTP exerts a critical role in the protection against aging-dependent phenotypes in vivo (Fig. 1) and that TTP depletion increased PAI-1 levels (Fig. 2). In addition, CO generated from CORM2 treatment recovered ETO-induced cellular senescence and inhibited ETO-induced SASPs secretion (Fig. 3).
To find the underlying mechanisms by which TTP can regulate age-dependent processes, we analyzed the effects of CORM2 on ETO-induced senescence using primary MEFs isolated from WT and Ttp-/- mice. CORM2 treatment suppressed ETO-induced SA-β gal positive cells in WT murine primary MEFs, but not in Ttp-/- MEFs (Fig. 5A). In addition, cellular senescent phenotypes such as γ-H2AX foci (Fig. 5B), p21 mRNA levels (Fig. 5C), IL-6 mRNA levels (Fig. 5D), and PAI-1 mRNA levels (Fig. 5E) were increased by ETO in both WT murine and Ttp-/- murine primary MEF. CORM2 treatment suppressed the cellular senescent phenotype in WT, but not in Ttp-/- MEFs (Fig. 5B-5E). We also confirmed that ETO increased PAI-1 mRNA levels in a dose-dependent manner in AML-12 cells (Fig. 5F) and in primary MEFs (Fig. 5G). CORM2 treatment inhibited ETO-induced PAI-1 mRNA levels in AML-12 cells (Fig. 5F) and in primary MEFs (Fig. 5G). Also, the levels of PAI-1 protein were increased by ETO treatment in primary MEFs in a dose-dependent manner (Fig. 5H). In addition, we confirmed that the increases in PAI-1 protein levels (Fig. 5I) and PAI-1 secretion (Fig. 5J) by ETO were suppressed by CORM2 treatment. TTP knock-down using siRNA against TTP resulted in increased expression of PAI-1 in response to ETO relative to cells transfected with scrambled (control) siRNA (Fig. 5K). ETO-induced PAI-1 mRNA and protein levels were enhanced in Ttp-/- MEFs relative to WT MEFs (Fig. 5L and 5M).
Figure 6.CO-induced SGs participate in reducing ETO-induced senescence by sequestration of PAI-1. ( A) WI-38 cells were treated with 40 μM CORM2 for 6 h. As a positive control, WI-38 cells were treated with 200 nM thapsigargin (Tg) for 45 min and then an immunofluorescence assay was performed to detect the formation of SGs by visualizing the co-localization of TIA-1(red) and G3BP1 (green). Rabbit IgG and mouse IgG1 were used as a negative control of anti-TIA-1 antibody and anti-G3BP1 antibody, respectively. Scale bar: 20 μm (left). The percentage of cells containing SGs was analyzed and is shown in the right panel. ( B) WI-38 cells were treated with 40 μM CORM2 in the presence or absence of ISRIB (200 nM) for 6 h and then stained with anti-TIA-1 and anti-G3BP1 antibodies. The formation of SGs was detected by visualizing the co-localization of TIA-1(red) and G3BP1 (green). Rabbit IgG and mouse IgG1 were used as negative control of anti-TIA-1 antibody and anti-G3BP1 antibody, respectively. Scale bar: 20 μm (left). Quantification is shown in the bar graphs on the right panel. ( A, B) Data were analyzed using Kruskal-Wallis test followed by the Dunn post hoc test and expressed as means ± SD; $$n = 3$$ biological replicates; **$p \leq 0.01$ and ***$p \leq 0.001.$ ( C, D) WI-38 cells were pretreated with 40 μM CORM2 for 6 h followed by the administration of ETO (80μM) for 24 h and then cells were refed with fresh media. During the process of senescence, cells were post-treated with 40 μM CORM2 for 6 h. (C) After 7 days incubation, an immunofluorescence assay was performed to detect TIA-1 (red) and PAI-1 (green) co-aggregates. Rabbit IgG and mouse IgG1 were used as a negative control of anti-TIA-1 antibody and anti-PAI-1 antibody, respectively. Scale bar: 20 μm. The co-localization of TIA-1 and PAI-1 was quantified (right; mean ± SD; $$n = 6$$ biological replicates; ***$p \leq 0.001$; one-way ANOVA followed by Tukey post hoc test). ( D) The levels of protein expression of PAI-1, p21, and p53 were detected by immunoblotting. ( E, F) WI-38 cells were pretreated with CORM (40 μM) and ISRIB (200 nM) for 6 h followed by treatment with ETO (80 μM) for 24 h, and then cells were refed with fresh media. During the process of senescence, cells were post-treated with CORM2 (40 μM) and ISRIB (200 nM) for 6 h. (E) After 7 days incubation, cells were fixed to perform an immunofluorescence assay of PAI-1 and TIA-1 (left). Rabbit IgG and mouse IgG1 were used as a negative control of anti-TIA-1 antibody and anti-PAI-1 antibody, respectively. Scale bar: 20 μm. Quantification of co-localization of PAI-1 and TIA-1 is shown in the bar graphs to the right panel (mean ± SD; $$n = 6$$ biological replicates; ***$p \leq 0.001$; one-way ANOVA followed by Tukey post hoc test). ( F) The secretion of PAI-1 was measured by ELISA (mean ± SD; $$n = 3$$ biological replicates; ***$p \leq 0.001$; Kruskal-Wallis test followed by the Dunn post hoc test).
To further investigate the effect of TTP on PAI-1 mRNA stabilization, we assayed PAI-1 mRNA stability with the actinomycin D assay, in WT and Ttp-/- MEFs treated with CORM2. In WT MEFs treated with CORM2, PAI-1 mRNA degradation was accelerated compared to WT MEFs without CORM2 treatment and compared to Ttp-/- MEFs (Fig. 5N). In addition, CORM2 reduced PAI-1 3’-UTR stability in WT, but not Ttp-/- MEFs, as detected using a luciferase-based assay (Fig. 5O). Further, CORM2 treatment does-dependently lowered the PAI-1 3’-UTR stability in AML-12 and HEK293 cells (Fig. 5P). Therefore, CORM2-derived CO reversed ETO-induced cellular senescence via PAI-1 degradation, in cells expressing TTP.
## CO-induced SGs reduce ETO-induced senescence by sequestration of PAI-1
Assembly of SGs induced by constitutive stress can decrease the number of senescent cells through the recruitment of PAI-1 [30]. We reported that CO can stimulate the formation of SGs by selective induction of the PERK-eIF2α signaling pathway, a component of the integrated stress response (ISR) [32]. Based on these reports, we hypothesized that the anti-senescence effect of CO is mediated by assembly of SGs. Here, we first confirmed the beneficial effect of CO on SG formation by treating WI-38 cells with CORM2 or exogenous CO gas. Consistent with our earlier work [32], CORM2 (Fig. 6A) and CO gas (Supplementary Fig. 3A) significantly increased the assembly of TIA-1 and G3BP-1 positive SGs in cytoplasm. In addition, we confirmed that an ISR inhibitor (ISRIB) can markedly decrease the formation of SGs in response to CORM2 treatment (Fig. 6B). We also investigated whether CO could stimulate the sequestration of PAI-1 into SGs in ETO-treated WI-38 cells. Notably, the increased number of SGs initiated by CORM2 (Fig. 6C) and CO gas (Supplementary Fig. 3B) treatment significantly sequestrated PAI-1, as detected by co-aggregation of TIA-1 and PAI-1. In addition, ETO-induced PAI-1 protein levels were decreased by CORM2 in WI-38 cells (Fig. 6D). The levels of p21 and p53, as senescence-related proteins, were also enhanced by ETO; and these increases were inhibited by CORM2 (Fig. 6D and Supplementary Fig. 3E). Furthermore, we also observed that ISRIB strongly inhibits the sequestration of PAI-1 into CO-induced SGs (Fig. 6E), and consequently, the ability of CORM2 to decrease the secretion of PAI-1was abolished (Fig. 6F). As expected, co-treatment with ISRIB reversed the protective effects of CORM2 on ETO-induced senescence as measured by activity of SA-β-gal (Supplementary Fig. 3C), γ-H2AX foci (Supplementary Fig. 3D), and the protein levels of PAI-1, p21, and p53 (Supplementary Fig. 3E). Under the same conditions, we also measured the mRNA expression of p21 and several SASPs, such as IL-6, TNF-α, and IL-1β? Supplementary Fig. 3F-3I?. Together, these results suggest that CO-induced SGs prevent cellular senescence by sequestering PAI-1.
## CO-mediated inhibition of PAI-1 requires Sirt1-TTP activation and the assembly of SGs
SG-associated proteins such as TIA-1, TIAR, and HuR bind to ARE-containing mRNAs and control their translation and stability [40]. Under energy deprivation, TTP is also recruited to SGs, which contributes to degrading ARE-containing transcripts [40]. We demonstrated that CORM2 diminished ETO-induced PAI-1 expression in a TTP-dependent manner (Fig. 5) and promoted PAI-1 sequestration in SGs (Fig. 6). Thus, we investigated whether CORM2 can promote the recruitment of TTP to SGs. CORM2 increased TTP migration into SGs, leading to co-localization of G3BP1 and TTP in WT MEFs but not in Ttp-/- MEFs (Fig. 7A and 7B). In our previous report, we demonstrated that CO induces TTP activation via Sirt1 [33]. To demonstrate the effects of Sirt1-dependent TTP activation on the migration of TTP into SGs, we used EX527 as a Sirt1 inhibitor. The increase of the migration of TTP into SGs by CORM2 was inhibited in EX527-treated AML-12 cells, which is due to the suppression of CO-induced TTP activation by EX527 (Fig. 7C). Next, we observed that CO-induced TTP activation facilitated the decrease of PAI-1 in ETO-treated primary MEFs (Fig. 7D and 7E) and AML-12 cells (Fig. 7F). In addition, EX527 inhibited the reduction of PAI-1 by CORM2 (Fig. 7D-7F). Taken together, our results demonstrate that CO-induced Sirt1-dependent TTP activation promoted TTP migration into SGs, leading to increased PAI-1 degradation.
Figure 7.The decrease of PAI-1 by CO requires Sirt1-TTP activation in the assembly of SGs. ( A, B) Ttp+/+ and Ttp-/- primary MEFs were pretreated with CORM2 (40 μM) for 6 h and then cells were treated with ETO (2 μM) for 4 days. During the process of senescence, MEFs were treated with CORM2 (40 μM) for 6 h every two days. ( A) After 4 days incubation, cells were stained with anti-TTP and anti-G3BP1 antibodies for assessing co-localization of SGs and TTP. Rabbit IgG and mouse IgG1 were used as a negative control of anti-TTP and anti-G3BP1 antibody, respectively. Scale bar: 10 μm. ( B) Quantification of co-localization of G3BP1 and TTP is shown in the bar graphs (mean ± SD; $$n = 5$$ biological replicates; ***$p \leq 0.001$ and ND, not determined; two-way ANOVA followed by Bonferroni post-test). ( C) AML12 cells were treated with 20 μM ETO for 4 days in the presence or absence of 40 μM CORM2 and 10 μM EX527 and cells were stained with anti-TTP and anti-G3BP1 antibodies (left). Rabbit IgG and mouse IgG1 were used as negative control of anti-TTP and anti-G3BP1 antibody, respectively. ( D) Primary MEFs were treated with 20μM ETO for 4 days in the presence or absence of 40μM CORM2 and 10μM EX527 and cell were stained with anti-TTP and anti-PAI-1 antibodies (left). Rabbit IgG and mouse IgG1 were used as negative control of anti-TTP and anti-PAI-1 antibody, respectively. Quantification of co-localization of G3BP1 and TTP is shown in the right panel (mean ± SD; $$n = 5$$ biological replicates; ***$p \leq 0.001$; Kruskal-Wallis test followed by the Dunn post hoc test). ( E, F) RT-PCR in (E) primary MEFs and (F) AML12 cells was performed to detect PAI-1 in the indicated groups.
## DISCUSSION
The study of aging is critical to overcoming diseases and maintaining life quality. In aged individuals, PAI-1 expression is elevated in a variety of pathologies associated with the aging process [36], including vascular sclerosis [41], cardiac and lung fibrosis [42], metabolic syndrome [43, 44], cancer [45], and inflammatory and stress responses [46].
In this study, we demonstrated that mice sustain age-dependent increases in PAI-1 expression. We also demonstrate that therapeutic application of CO promotes PAI-1 sequestration by SG assembly and reduced PAI-1 secretion via TTP activation.
Figure 8.TTP can prevent liver injury and hepatocyte cell senescence during aging. TTP may protect against age-associated senescence phenotypes via decreasing PAI-1 levels. CO reversed ETO-induced cellular senescence via PAI-1 degradation, in cells expressing TTP. CO-induced SGs prevent cellular senescence by sequestering PAI-1. CO-induced Sirt1-dependent TTP activation promoted TTP migration into SGs, leading to increased PAI-1 degradation.
The hormesis effects of CO at low doses have been reported as the anti-inflammatory response [33], anti-obese effects [47, 48], and anti-oxidative response [49, 50]. CO, as a protector against cellular stress, activates the PERK-eIF2α signaling pathway through mtROS production, leading to induction of SG assembly [32]. The SG-mediated inhibition of senescence is caused by recruitment of PAI-1, a member of the SASP [30] and a well-known promoter of senescence [51-53]. Although CO has been shown to prevent bleomycin-induced cellular senescence by SG assembly [54], the underlying mechanisms remain unclear. In our previous reports, CO was shown to promote Sirt1 expression and activation, resulting in the deacetylation of p53 [55] and TTP [33]. Here, we demonstrate that the Sirt1 inhibitor prevents the CO-mediated migration of TTP into SGs. In contrast, phosphorylation of TTP induced by the p38-MAPK/MK2 pathway [56, 57] resulted in exclusion of TTP from SGs, leading to TTP:14-3-3 complex formation [29]. Given that SGs form from pools of untranslated mRNA and contain various translation initiation factors, as well as a variety of RNA-binding proteins and non-RNA-binding protein [58], Sirt1-mediated TTP deacetylation permitted TTP to bind to the ARE of PAI-1 mRNA in SGs. Therefore, TTP deacetylation by CO is critical to control aging. Aged Ttp-/- mice showed higher PAI-1 levels than young Ttp-/- mice or corresponding WT mice. Thus, the interaction between TTP and PAI-1 may play a critical role in aging.
Age-mediated NAFLD in Ttp-/- mice represents the increase of inflammatory cytokines, SASP, and liver damage. Thus, we suggest that TTP activation can alleviate age-mediated NAFLD. Given that aging is characterized by cellular senescence, the role of TTP has been studied in cellular senescence. Cellular senescence is a state of stable proliferative arrest triggered by damaging signals such as DNA damage or oncogene-dependent pathways [23].
In this study, we treated human diploid fibroblast (WI-38) cells, and primary mouse embryonic fibroblasts MEFs, with ETO to establish DNA damage induced-cellular senescence. ETO, one of the topoisomerase II poisons, is commonly used as a chemotherapeutic agent, which can cause DNA double strand breaks (DSBs), which are toxic to normal cells [37]. Additionally, the treatment of normal human fibroblasts with ETO can lead to a long-term cell cycle arrest and premature senescence mediated by the activation of p53 and enhancement of CDK inhibitors, including p16INK4a and p21CIP1 [59, 60]. We observed that CO exerts a strong anti-senescent effect on ETO-mediated premature senescence. Our results showed that the administration of CORM2 and exogenous CO gas can significantly decrease multiple hallmarks of cellular senescence, including the percentage of SA-β-gal positive cells, DNA damage associated γ-H2AX foci, CDK inhibitor p21 expression, and the expression of several SASPs, such as IL-6, TNF-α, and IL-1β. Consistent with in vivo aging results, TTP deficiency abolished the ability of CO to prevent ETO-mediated cellular senescence. Moreover, CO can cause the sequestration of PAI-1 into SGs during challenge with ETO, which can dramatically reduce the secretion of PAI-1. Notably, the reduction of PAI-1 by CO under these conditions was dependent on TTP activation. Further studies will be needed to validate these mechanisms in in vivo models. We conclude that CO-dependent TTP activation diminishes PAI-1 levels in SGs, leading to alleviation of age-dependent NAFLD and ETO-induced cellular senescence (Fig. 8). Therefore, we suggest that TTP activation by CO may represent a novel therapeutic strategy to ameliorate cellular senescence and aging-mediated diseases.
## Author contributions
J.P., Y.C., H.T.C., and Y.J. conceived and designed the study. J.P., Y.C., J.K., E.H., and Y.J. performed the experiments. G.H.P., C.H.Y., S.W.R, J.W.P, H.T.C., and Y.J. provided key reagents and revised the paper. S.W.R., H.T.C., and Y.J. analyzed the data and wrote the manuscript. All the authors approved the final version of the manuscript, discussed the results, and approved the final version of the manuscript.
## References
1. Hou Y, Dan X, Babbar M, Wei Y, Hasselbalch SG, Croteau DL. **Ageing as a risk factor for neurodegenerative disease**. *Nat Rev Neurol* (2019) **15** 565-581. PMID: 31501588
2. Wyss-Coray T. **Ageing, neurodegeneration and brain rejuvenation**. *Nature* (2016) **539** 180-186. PMID: 27830812
3. de Magalhaes JP. **How ageing processes influence cancer**. *Nat Rev Cancer* (2013) **13** 357-365. PMID: 23612461
4. Berben L, Floris G, Wildiers H, Hatse S. **Cancer and Aging: Two Tightly Interconnected Biological Processes**. *Cancers(Basel)* (2021) **13**
5. Fazeli PK, Lee H, Steinhauser ML. **Aging Is a Powerful Risk Factor for Type 2 Diabetes Mellitus Independent of Body Mass Index**. *Gerontology* (2020) **66** 209-210. PMID: 31505500
6. North BJ, Sinclair DA. **The intersection between aging and cardiovascular disease**. *Circ Res* (2012) **110** 1097-1108. PMID: 22499900
7. Liberale L, Badimon L, Montecucco F, Luscher TF, Libby P, Camici GG. **Inflammation, Aging, and Cardiovascular Disease: JACC Review Topic of the Week**. *J Am Coll Cardiol* (2022) **79** 837-847. PMID: 35210039
8. Frith J, Day CP, Henderson E, Burt AD, Newton JL. **Non-alcoholic fatty liver disease in older people**. *Gerontology* (2009) **55** 607-613. PMID: 19690397
9. Sheedfar F, Di Biase S, Koonen D, Vinciguerra M. **Liver diseases and aging: friends or foes?**. *Aging Cell* (2013) **12** 950-954. PMID: 23815295
10. Bernadotte A, Mikhelson VM, Spivak IM. **Markers of cellular senescence. Telomere shortening as a marker of cellular senescence**. *Aging (Albany NY)* (2016) **8** 3-11. PMID: 26805432
11. Schumacher B, Pothof J, Vijg J, Hoeijmakers JHJ. **The central role of DNA damage in the ageing process**. *Nature* (2021) **592** 695-703. PMID: 33911272
12. Pagiatakis C, Musolino E, Gornati R, Bernardini G, Papait R. **Epigenetics of aging and disease: a brief overview**. *Aging Clin Exp Res* (2021) **33** 737-745. PMID: 31811572
13. Finkel T, Holbrook NJ. **Oxidants, oxidative stress and the biology of ageing**. *Nature* (2000) **408** 239-247. PMID: 11089981
14. Trifunovic A, Larsson NG. **Mitochondrial dysfunction as a cause of ageing**. *J Intern Med* (2008) **263** 167-178. PMID: 18226094
15. Gasek NS, Kuchel GA, Kirkland JL, Xu M. **Strategies for Targeting Senescent Cells in Human Disease**. *Nat Aging* (2021) **1** 870-879. PMID: 34841261
16. Ogrodnik M, Evans SA, Fielder E, Victorelli S, Kruger P, Salmonowicz H. **Whole-body senescent cell clearance alleviates age-related brain inflammation and cognitive impairment in mice**. *Aging Cell* (2021) **20** e13296. PMID: 33470505
17. Bojko A, Czarnecka-Herok J, Charzynska A, Dabrowski M, Sikora E. **Diversity of the Senescence Phenotype of Cancer Cells Treated with Chemotherapeutic Agents**. *Cells* (2019) **8**
18. Kurz DJ, Decary S, Hong Y, Erusalimsky JD. **Senescence-associated (beta)-galactosidase reflects an increase in lysosomal mass during replicative ageing of human endothelial cells**. *J Cell Sci* (2000) **113** 3613-3622. PMID: 11017877
19. Kumari R, Jat P. **Mechanisms of Cellular Senescence: Cell Cycle Arrest and Senescence Associated Secretory Phenotype**. *Front Cell Dev Biol* (2021) **9** 645593. PMID: 33855023
20. Stein GH, Drullinger LF, Soulard A, Dulic V. **Differential roles for cyclin-dependent kinase inhibitors p21 and p16 in the mechanisms of senescence and differentiation in human fibroblasts**. *Mol Cell Biol* (1999) **19** 2109-2117. PMID: 10022898
21. Sandhu C, Peehl DM, Slingerland J. **p16INK4A mediates cyclin dependent kinase 4 and 6 inhibition in senescent prostatic epithelial cells**. *Cancer Res* (2000) **60** 2616-2622. PMID: 10825132
22. Narita M, Narita M, Krizhanovsky V, Nunez S, Chicas A, Hearn SA. **A novel role for high-mobility group a proteins in cellular senescence and heterochromatin formation**. *Cell* (2006) **126** 503-514. PMID: 16901784
23. Serrano M, Lin AW, McCurrach ME, Beach D, Lowe SW. **Oncogenic ras provokes premature cell senescence associated with accumulation of p53 and p16INK4a**. *Cell* (1997) **88** 593-602. PMID: 9054499
24. Ghosh AK, Rai R, Park KE, Eren M, Miyata T, Wilsbacher LD. **A small molecule inhibitor of PAI-1 protects against doxorubicin-induced cellular senescence**. *Oncotarget* (2016) **7** 72443-72457. PMID: 27736799
25. Kortlever RM, Higgins PJ, Bernards R. **Plasminogen activator inhibitor-1 is a critical downstream target of p53 in the induction of replicative senescence**. *Nat Cell Biol* (2006) **8** 877-884. PMID: 16862142
26. Anderson P, Kedersha N. **RNA granules: post-transcriptional and epigenetic modulators of gene expression**. *Nat Rev Mol Cell Biol* (2009) **10** 430-436. PMID: 19461665
27. Kedersha N, Ivanov P, Anderson P. **Stress granules and cell signaling: more than just a passing phase?**. *Trends Biochem Sci* (2013) **38** 494-506. PMID: 24029419
28. Gilks N, Kedersha N, Ayodele M, Shen L, Stoecklin G, Dember LM. **Stress granule assembly is mediated by prion-like aggregation of TIA-1**. *Mol Biol Cell* (2004) **15** 5383-5398. PMID: 15371533
29. Stoecklin G, Stubbs T, Kedersha N, Wax S, Rigby WF, Blackwell TK. **MK2-induced tristetraprolin:14-3-3 complexes prevent stress granule association and ARE-mRNA decay**. *EMBO J* (2004) **23** 1313-1324. PMID: 15014438
30. Omer A, Patel D, Lian XJ, Sadek J, Di Marco S, Pause A. **Stress granules counteract senescence by sequestration of PAI-1**. *EMBO Rep* (2018) **19**
31. Park J, Chen Y, Zheng M, Ryu J, Cho GJ, Surh YJ. **Pterostilbene 4'-beta-Glucoside Attenuates LPS-Induced Acute Lung Injury via Induction of Heme Oxygenase-1**. *Oxid Med Cell Longev* (2018) **2018** 2747018. PMID: 30425781
32. Chen Y, Joe Y, Park J, Song HC, Kim UH, Chung HT. **Carbon monoxide induces the assembly of stress granule through the integrated stress response**. *Biochem Biophys Res Commun* (2019) **512** 289-294. PMID: 30885431
33. Joe Y, Chen Y, Park J, Kim HJ, Rah SY, Ryu J. **Cross-talk between CD38 and TTP Is Essential for Resolution of Inflammation during Microbial Sepsis**. *Cell Rep* (2020) **30** 1063-1076 e1065. PMID: 31995750
34. Durkin ME, Qian X, Popescu NC, Lowy DR. **Isolation of Mouse Embryo Fibroblasts**. *Bio Protoc* (2013) **3**
35. Galijatovic A, Beaton D, Nguyen N, Chen S, Bonzo J, Johnson R. **The human CYP1A1 gene is regulated in a developmental and tissue-specific fashion in transgenic mice**. *J Biol Chem* (2004) **279** 23969-23976. PMID: 15037607
36. Yamamoto K, Takeshita K, Kojima T, Takamatsu J, Saito H. **Aging and plasminogen activator inhibitor-1 (PAI-1) regulation: implication in the pathogenesis of thrombotic disorders in the elderly**. *Cardiovasc Res* (2005) **66** 276-285. PMID: 15820196
37. Vamvakas S, Vock EH, Lutz WK. **On the role of DNA double-strand breaks in toxicity and carcinogenesis**. *Crit Rev Toxicol* (1997) **27** 155-174. PMID: 9099517
38. Michishita E, Nakabayashi K, Ogino H, Suzuki T, Fujii M, Ayusawa D. **DNA topoisomerase inhibitors induce reversible senescence in normal human fibroblasts**. *Biochem Biophys Res Commun* (1998) **253** 667-671. PMID: 9918785
39. Dellas C, Loskutoff DJ. **Historical analysis of PAI-1 from its discovery to its potential role in cell motility and disease**. *Thromb Haemost* (2005) **93** 631-640. PMID: 15841306
40. Freund A, Orjalo AV, Desprez PY, Campisi J. **Inflammatory networks during cellular senescence: causes and consequences**. *Trends Mol Med* (2010) **16** 238-246. PMID: 20444648
41. Spiess BD. **Ischemia--a coagulation problem?**. *J Cardiovasc Pharmacol* (1996) **27** S38-41
42. Ghosh AK, Vaughan DE. **PAI-1 in tissue fibrosis**. *J Cell Physiol* (2012) **227** 493-507. PMID: 21465481
43. Alessi MC, Juhan-Vague I. **PAI-1 and the metabolic syndrome: links, causes, and consequences**. *Arterioscler Thromb Vasc Biol* (2006) **26** 2200-2207. PMID: 16931789
44. Wang L, Chen L, Liu Z, Liu Y, Luo M, Chen N. **PAI-1 Exacerbates White Adipose Tissue Dysfunction and Metabolic Dysregulation in High Fat Diet-Induced Obesity**. *Front Pharmacol* (2018) **9** 1087. PMID: 30319420
45. Placencio VR, DeClerck YA. **Plasminogen Activator Inhibitor-1 in Cancer: Rationale and Insight for Future Therapeutic Testing**. *Cancer Res* (2015) **75** 2969-2974. PMID: 26180080
46. Wu H, Wang Y, Zhang Y, Xu F, Chen J, Duan L. **Breaking the vicious loop between inflammation, oxidative stress and coagulation, a novel anti-thrombus insight of nattokinase by inhibiting LPS-induced inflammation and oxidative stress**. *Redox Biol* (2020) **32** 101500. PMID: 32193146
47. Joe Y, Kim S, Kim HJ, Park J, Chen Y, Park HJ. **FGF21 induced by carbon monoxide mediates metabolic homeostasis via the PERK/ATF4 pathway**. *FASEB J* (2018) **32** 2630-2643. PMID: 29295856
48. Kim HJ, Joe Y, Surh YJ, Chung HT. **Metabolic signaling functions of the heme oxygenase/CO system in metabolic diseases**. *Cell Mol Immunol* (2018) **15** 1085-1087. PMID: 29807990
49. Kim HJ, Joe Y, Chen Y, Park GH, Kim UH, Chung HT. **Carbon monoxide attenuates amyloidogenesis via down-regulation of NF-kappaB-mediated BACE1 gene expression**. *Aging Cell* (2019) **18** e12864. PMID: 30411846
50. Chen Y, Park HJ, Park J, Song HC, Ryter SW, Surh YJ. **Carbon monoxide ameliorates acetaminophen-induced liver injury by increasing hepatic HO-1 and Parkin expression**. *FASEB J* (2019) **33** 13905-13919. PMID: 31645120
51. Eren M, Boe AE, Murphy SB, Place AT, Nagpal V, Morales-Nebreda L. **PAI-1-regulated extracellular proteolysis governs senescence and survival in Klotho mice**. *Proc Natl Acad Sci U S A* (2014) **111** 7090-7095. PMID: 24778222
52. Boe AE, Eren M, Murphy SB, Kamide CE, Ichimura A, Terry D. **Plasminogen activator inhibitor-1 antagonist TM5441 attenuates Nomega-nitro-L-arginine methyl ester-induced hypertension and vascular senescence**. *Circulation* (2013) **128** 2318-2324. PMID: 24092817
53. Tsuda K. **Letter by Tsuda regarding article, "Plasminogen activator inhibitor-1 antagonist TM5441 attenuates Nomega-nitro-L-arginine methyl ester-induced hypertension and vascular senescence"**. *Circulation* (2014) **130** e83. PMID: 25156921
54. Chen Y, Jiang F, Kong G, Yuan S, Cao Y, Zhang Q. **Gasotransmitter CO Attenuates Bleomycin-Induced Fibroblast Senescence via Induction of Stress Granule Formation**. *Oxid Med Cell Longev* (2021) **2021** 9926284. PMID: 34306316
55. Zheng M, Chen Y, Park J, Song HC, Chen Y, Park JW. **CO ameliorates endothelial senescence induced by 5-fluorouracil through SIRT1 activation**. *Arch Biochem Biophys* (2019) **677** 108185. PMID: 31704100
56. Brook M, Tchen CR, Santalucia T, McIlrath J, Arthur JS, Saklatvala J. **Posttranslational regulation of tristetraprolin subcellular localization and protein stability by p38 mitogen-activated protein kinase and extracellular signal-regulated kinase pathways**. *Mol Cell Biol* (2006) **26** 2408-2418. PMID: 16508015
57. Ronkina N, Kotlyarov A, Dittrich-Breiholz O, Kracht M, Hitti E, Milarski K. **The mitogen-activated protein kinase (MAPK)-activated protein kinases MK2 and MK3 cooperate in stimulation of tumor necrosis factor biosynthesis and stabilization of p38 MAPK**. *Mol Cell Biol* (2007) **27** 170-181. PMID: 17030606
58. Guzikowski AR, Chen YS, Zid BM. **Stress-induced mRNP granules: Form and function of processing bodies and stress granules**. *Wiley Interdiscip Rev RNA* (2019) **10** e1524. PMID: 30793528
59. Di Leonardo A, Linke SP, Clarkin K, Wahl GM. **DNA damage triggers a prolonged p53-dependent G1 arrest and long-term induction of Cip1 in normal human fibroblasts**. *Genes Dev* (1994) **8** 2540-2551. PMID: 7958916
60. Robles SJ, Adami GR. **Agents that cause DNA double strand breaks lead to p16INK4a enrichment and the premature senescence of normal fibroblasts**. *Oncogene* (1998) **16** 1113-1123. PMID: 9528853
|
---
title: 'The Role of Bile Acids in Cardiovascular Diseases: from Mechanisms to Clinical
Implications'
authors:
- Shuwen Zhang
- Junteng Zhou
- Wenchao Wu
- Ye Zhu
- Xiaojing Liu
journal: Aging and Disease
year: 2023
pmcid: PMC10017164
doi: 10.14336/AD.2022.0817
license: CC BY 2.0
---
# The Role of Bile Acids in Cardiovascular Diseases: from Mechanisms to Clinical Implications
## Abstract
Bile acids (BAs), key regulators in the metabolic network, are not only involved in lipid digestion and absorption but also serve as potential therapeutic targets for metabolic disorders. Studies have shown that cardiac dysfunction is associated with abnormal BA metabolic pathways. As ligands for several nuclear receptors and membrane receptors, BAs systematically regulate the homeostasis of metabolism and participate in cardiovascular diseases (CVDs), such as myocardial infarction, diabetic cardiomyopathy, atherosclerosis, arrhythmia, and heart failure. However, the molecular mechanism by which BAs trigger CVDs remains controversial. Therefore, the regulation of BA signal transduction by modulating the synthesis and composition of BAs is an interesting and novel direction for potential therapies for CVDs. Here, we mainly summarized the metabolism of BAs and their role in cardiomyocytes and noncardiomyocytes in CVDs. Moreover, we comprehensively discussed the clinical prospects of BAs in CVDs and analyzed the clinical diagnostic and application value of BAs. The latest development prospects of BAs in the field of new drug development are also prospected. We aimed to elucidate the underlying mechanism of BAs treatment in CVDs, and the relationship between BAs and CVDs may provide new avenues for the prevention and treatment of these diseases.
## 1. Introduction
Cardio-metabolic disease (CMD) is a broad term describing cardiovascular disease (CVD) caused by systemic metabolic changes. Metabolic changes are mechanically involved in almost all forms of cardiovascular disease [1]. CMD is the world’s leading cause of death and encompasses cardiovascular diseases, diabetes, and chronic renal failure [2].
Bile acids (BAs) are mainly synthesized in the liver and are the final product of cholesterol catabolism. Their components are cholesterol derivatives. Since the last century, evidence has shown that pathologically elevated BA circulation in liver disease is harmful to the heart [3, 4]. In the past few years, BAs have been discovered as circulating metabolites that can act as metabolic regulators by binding to multiple BA receptors. In view of their unique biological features, BAs play an important role in regulating multiple metabolic pathways, such as glucose, lipids and amino acids, as well as maintaining homeostasis of gut microbiota metabolism [5]. BAs can be present in most organs, tissues and cells that express its receptors. They are implicated in a variety of metabolic diseases and have evolved from a simple bile component to a complex metabolic integrator according to the researchers' understanding [6, 7].
BAs have not only been shown to play a key role in mediating oxidative stress, reactive oxygen species (ROS), mitochondrial dysfunction, cell membrane disruption and cellular damage [8]; more importantly, BAs and their metabolism are closely related to CVDs and metabolic disorders and help maintain cardiovascular function and health. BAs have two forms of effect on cardiac function: direct and indirect. Direct action requires BAs to interact with muscle cells, affecting myocardial contraction and conduction. These effects may or may not be receptor dependent. The indirect effects include multiple metabolic pathways, such as cardiac function regulation, cholesterol level regulation, and plaque formation in atherosclerosis [9].
In this paper, we summarize the role of different BA synthesis pathways and several regulatory mechanisms on related cells in the development of CVDs, as well as their role in the development and treatment of CVDs and related metabolic diseases. In addition, we introduce the potential therapeutic effects of ursodeoxycholic acid (UDCA) and other BA derivatives.
## 2.1 The synthesis and metabolism of bile acids
BAs are the product of cholesterol metabolism in the liver. BA metabolism is the main pathway for the human body to remove cholesterol, and BAs are also important substances because they affect lipid absorption. Based on their structure, BAs can be divided into primary and secondary BAs. BAs can be further classified into bound and free BAs according to whether they are combined with glycine or taurine [10].
Currently, the major BAs identified in humans include chenodeoxycholic acid (CDCA), cholic acid (CA), and a small amount of lithocholic acid (LCA). UDCA and muricholic acid (MCA) are primary BAs in rodents rather than in humans. According to the different groups in the R1/R2/R3/X position, the lipophilicity of BAs is different. BAs are formed in the liver through a complex process and finally stored in the gallbladder. The formation process includes multiple reaction steps involving at least 17 different enzymes [11].
There are two ways to synthesize BAs, namely, the classical pathway and the alternative pathway [12]. In the classical pathway, cholesterol is catalyzed by cholesterol-7α-hydroxylase (CYP7A1) to first generate 7α-hydroxycholesterol as an intermediate product, which is further catalyzed to generate CA and CDCA [13]. CYP7A1 is the rate-limiting enzyme of the entire pathway, which determines the amount of BA produced. This is the rate-limiting step in the synthesis of BAs. Under normal conditions, at least ¾ of BAs are produced through this pathway [14]. The alternative BA pathway is theoretically present in the mitochondria of all cells or tissues. In this pathway, cholesterol is first catalyzed by sterol-27-hydroxylase (CYP27A1) to generate the intermediate product 27-hydroxycholesterol, which is then hydrogenated by sterol-7α-hydroxylase (CYP7B1) to form CDCA. The alternative approach mainly produces CDCA. Sterol-8α-hydroxylase (CYP8B1) plays a role in the synthesis of CA, and it can determine the ratio of CDCA to CA, the two main Bas [15]. In addition to CA and CDCA, primary BAs in mice also produce MCAs and UDCA [16]. MCAs are generally not detectable in humans. The synthesis of the final product of BAs requires modification by microorganisms in the gut. The 24-position carboxyl group of primary BAs is combined with glycine (in humans) or taurine (in mice), converted into secondary BAs, and excreted into bile. Secondary BAs are generally stored in the gallbladder and transported to the duodenum when needed. The amphiphilic structure of BAs makes them useful in emulsifying and absorbing some lipids and fat-soluble vitamins [17].
The liver contains very few BAs, and approximately $95\%$ of the BAs secreted by the bile ducts are reabsorbed by microorganisms in the gut. BAs are primarily absorbed in the distal ileum in conjugated form by apical sodium-dependent bile acid transporters (ASBTs), recirculated through the portal vein into the liver, and then secreted again. This process occurs in the human body six times a day and is called enterohepatic circulation [14]. BAs form a metabolic axis between the liver and gut microbiota, which contributes to BA metabolic disturbances and significant changes in the composition of the microbiota. Hence, BA metabolism can be used as a new therapeutic strategy for metabolic diseases [18].
Deoxycholic acid (DCA), CDCA, LCA, and CA are crucial for the regulation of the BA pool. The cytotoxicity of BAs depends on their structure, but the hydrophobicity (lipophilicity) of BAs is related to the number and position of hydroxyl groups in their ring structure. The hydrophobicity-based order is as follows: LCA>DCA >CDCA>CA>UDCA>MCA. The most hydrophobic LCA is mostly excreted in the feces, with only a small amount being reabsorbed in enterohepatic circulation. The least toxic and most hydrophilic is UDCA. UDCA is synthesized in the gut by the dehydroxylation of free CDCA with the participation of bacteria. The hydroxyl group of UDCA is located in the β-ring, while CDCA is located in the α-ring. Interestingly, a growing body of research suggests that UDCA can play a protective role in CVDs (Fig. 1) [11].
According to previous studies, increased hydrophobic BA serum levels are associated with various metabolism-related diseases [16].
Figure 1.Synthesis and metabolism of BAs. There are two pathways to synthesize BAs, namely, the classical pathway and the alternative pathway[12]. The formation process includes multiple reaction steps. In the classical pathway, cholesterol is catalyzed by cholesterol-7α-hydroxylase (CYP7A1) to first generate 7α-hydroxycholesterol as an intermediate product, which is further catalyzed to generate CA and CDCA. The BA alternative pathway is theoretically present in the mitochondria of all cells or tissues. In this pathway, cholesterol is first catalyzed by sterol-27-hydroxylase (CYP27A1) to generate the intermediate product 27-hydroxycholesterol, which is then hydrogenated by sterol-7α-hydroxylase (CYP7B1) to form CDCA. The alternative approach mainly produces CDCA.
## 2.2 Bile acid receptors associated with cardiovascular diseases
BAs acting as signaling mediators are thought to bind to various receptors that affect the metabolism and regulation of lipid profiles [19]. These receptors include nuclear receptors, membrane receptors, and Ca2+-activated potassium (K+) (BK) channels [20-22].
These receptors have been recently discovered in endothelial cells, cardiomyocytes, vascular smooth muscle cells, and cardiac fibroblasts [23], indicating that BAs may have an impact on the cardiovascular system.
## 2.2.1 Nuclear receptor
In the past few years, farnesoid X receptor (FXR)-mediated responses have been seen as critical for BA signaling [24]. More recently, other nuclear receptors, such as pregnane X receptor (PXR), liver X receptor (LXR), and vitamin D receptor (VDR), have also been found to play a role in the regulation of glucolipid metabolism [25].
FXR regulates the expression of CYP7A1 through a feedback mechanism by increasing BAs after food intake [19]. FXR serves as the primary target of most BAs. The hydrophobic side of the ligand-binding domain of FXR is able to bind to the hydrophilic side of BAs. CDCA is the most effective endogenous ligand for FXR binding compared with DCA, LCA, and CA. The functional study of FXR was first carried out in the intestine, where CDCA activated the expression of FXR and then mediated cholesterol secretion through intestinal acid-binding protein [22]. OCA is the most extensively studied FXR agonist and has been clinically evaluated [26]. Antagonists of FXR currently include UDCA and its conjugated form, glycine-UDCA [27, 28]. FXR activation is not only associated with maintaining normal cholesterol and triacylglycerol levels but is also expressed in the cardiovascular system. In addition, in the liver and gut, the activation of FXR affects regulators associated with CVD risks, such as glucolipid metabolism and endothelial function. FGF15 expression can be induced after intestinal FXR activation, which further improves glucose metabolism [29, 30]. However, there are studies showing the opposite results. Li et al. found that the improvement of glucose metabolism is achieved by inhibiting intestinal FXR signaling to alleviate the inhibition of L-cell glycolysis and GLP-1 secretion [31, 32]. Desai et al [33]. reported that the role of FXR in regulating BA levels is critical for organs such as the heart, and abnormally increased BA levels lead to cardiac dysfunction and cardiomyopathy in mice. However, when FXR activation is unrestricted, side effects such as pruritus, proatherosclerotic lipid profiles, and hepatotoxicity can also occur [27].
FXR, PXR, and VDR abolish BA-induced toxicity by downregulating the expression of cholesterol 7α hydroxylase, the rate-limiting enzyme in BA synthesis. An early animal study confirmed the significant role of PXR in lipid metabolism. Activation of PXR prevents high-fat diet- and obesity-induced insulin resistance by regulating energy and lipid metabolism [34].
Vitamin D (VD) deficiency may lead to bone and gastrointestinal-related diseases. Recently, CVDs, including heart failure and coronary heart disease, have been found to be associated with VD deficiency [35-38]. As a vital nuclear receptor that regulates calcium homeostasis, immunity, and cell differentiation, VDR is an endocrine nuclear receptor and is expressed in almost all tissues of the human body [39]. It has been reported that activation of VDR is able to participate in BA transport, metabolism, and detoxification by stimulating CYP3A. Furthermore, the natural ligand of LCA, 1α,25-dihydroxyvitamin D3 [1α,25(OH)2D3], can activate VDR. Both ligands activate the VDR signaling pathway via extracellular signal-regulated kinase $\frac{1}{2}$, resulting in VDR phosphorylation and translocation into the nucleus. The selective binding of LCA acetate to VDR was 30-fold higher than that of LCA itself, but the specific binding to FXR and PXR was lower [40].
LXR plays a critical role in the metabolism, transport, and excretion of BAs and maintains cholesterol homeostasis [41]. Endogenous sterols and oxidized derivatives of cholesterol activate LXR. When intra-cellular oxytocin levels increase, activated LXRs are able to protect cells from high levels of cholesterol [42]. Unfortunately, the upregulation of the CYP7A1 gene and ATP-binding cassette caused by the cholesterol/LXR signaling pathway has not been observed in the human liver and has only been confirmed in animal models. Most current research focuses on other nuclear receptor pathways, and further exploration of the LXR activation pathway is needed.
## 2.2.2 Membrane receptor
BAs also bind to three membrane G-protein-coupled receptors (the Takeda G-protein-coupled receptor 5 [TGR5], muscarinic [M] receptor, and S1P receptor), all of which are independent of nuclear hormone receptors and participate in cascades that activate intracellular effectors [40].
Approximately 10 years after the discovery of FXR, TGR5 was the first reported specific membrane receptor for Bas [43]. TGR5 is highly expressed not only in liver, adipose, and other tissues but also in the heart to some extent. TGR5 mRNA has been found in human, mouse, rabbit, and bovine cardiac tissue [44]. Moreover, different types of cells exhibit TGR5 expression, such as muscle, endocrine gland, and immune cells, as well as adipocytes [45].
Both LCA and DCA in secondary BAs are potent ligands for TGR5, and they can affect several important metabolic pathways, such as thermogenesis, glucose homeostasis, and energy metabolism. The activation of TGR5 signaling regulates several metabolic homeostasis pathways in the following ways: [1] improving insulin resistance by inducing type 2 iodase to increase energy expenditure in brown adipose tissue [46] and increase glucagon-like peptide-1 (GLP-1) secretion in enteroendocrine cells [47]; [2] reducing lipid load and inflammation in macrophages to prevent atherosclerosis [11], and [3] reducing blood vessel and liver damage to ameliorate nonalcoholic steatohepatitis [48]. In addition, the immunomodulatory function of TGR5 participates in various pathophysiological processes of multiple systems and exerts an inhibitory effect on inflammatory states [49], such as colitis [50, 51]. Steatohepatitis [52], atherosclerosis [53], sepsis [54], and inflammation related to type 2 diabetes [55].
The muscarinic (M) receptor is also a G-protein-coupled receptor (GPCR), and it is mainly expressed in the intestinal smooth muscle and gastrointestinal tract, including the five receptors M1-M5 [56]. The M2 receptor binds taurocholic acid and can affect transient calcium amplitudes by inhibiting cyclic AMP (cAMP), thereby reducing the contraction of cardiomyocytes [57]. The ligand of the M3 receptor is choline taurine.
Sphingosine-1-phosphate, another GPCR, has also been shown to be sensitive to BAs. As the most efficient substrate for sphingolipids, S1P is produced by sphingolipid kinase, catalyzed by sphingolipid phosphorylation. There are five subtypes of S1P receptors, namely, S1P1R, S1P2R, S1P3R, S1P4R, and S1P5R [58]. S1P1R, S1P2R, and S1P3R are mainly present in the heart, while S1P4R and S1P5R are only found in the immune and nervous systems [22]. S1P1R is the most important expressed isoform in cardiomyocytes, and its activation antagonizes adrenergic receptor-mediated contractility by inhibiting cAMP formation.
Secondary BAs activate S1P2R, affecting cell states by promoting apoptosis or survival signaling. Taurocholic acid promotes cholangiocarcinoma growth by inducing S1P2R expression [59]. S1P2R regulates liver glucose and lipid metabolism through the ERK$\frac{1}{2}$ and AKT signaling pathways [22]. Studies have shown that S1P2R and S1P3R protect against ischemia/reperfusion injury in mice. S1P agonists have a bradycardia effect, which can be mediated by low levels of S1P3R. S1PR is involved in various physiological activities of cardiac fibroblasts, such as proliferation, remodeling, and differentiation. S1PR-mediated pathways are also involved in hepatic fibrogenesis, regulating hepatic myofibroblast motility and vascular cell maturation and angiogenesis. Furthermore, in endothelial cells and smooth muscle cells, S1PR participates in endothelial cell responses and mediates peripheral vascular tone [60].
## 2.2.3 BKCa channels
In addition to the known nuclear and membrane receptors, BAs have been proven to activate nonclassical receptor responses. Among these, the large-conductance calcium-dependent potassium channel (BKCa) has the potential to increase the activity of BKCa in smooth muscle cells. Since this receptor mainly mediates ionic changes, it may play an indispensable role in related functions of cardiac conduction.
The activation of BKCa channels requires higher concentrations of BA than FXR or PXR. According to the findings of Bukiya et al., LCA can enhance the activity of BKCa channels in vascular myocytes [61]. The systemic vasodilation induced by BA in hepatobiliary diseases may be caused by the relaxation of VSMCs through the activation of BKCa. In another study, the activation of BKCa channels by taurine-coupled hydrophobic BAs resulted in the outward expansion of potassium currents, shortened action potential duration, and negative inotropic effects. Additionally, BAs have been shown to increase the risk of cirrhotic cardiomyopathy by activating the BK pathway in cirrhotic patients [62].
Therefore, the contact of BAs with different receptors in different tissues may determine their function and level of regulation. The above studies suggest that BAs function as useful biomarkers in human CVDs. There are reports that the activation of these BA receptors may be dependent on the BA conversion activity of certain gut microbiota, providing a key clue linking cardiovascular diseases with microbiota composition and activity, which warrants further study. Table 1 summarizes the BA receptors and their expression in cardiovascular cells and tissues.
**Table 1**
| Receptor | Organ/Tissue | Cells type | Ligand | Ref. |
| --- | --- | --- | --- | --- |
| FXR | Liver, gut, atherosclerotic blood vessels | Cardiomyocytes/Endothelial cells/Vascular smooth muscle cells | CA, CDCA, LCA, DCA | [22, 62, 63, 78, 82] |
| VDR | Liver | Cardiomyocytes | LCA | [38, 77] |
| PXR | Liver, mesenteric arteries | Cardiomyocytes | LCA | [39] |
| TGR5 | Liver, glands, fat, muscle,immune, endocrine glands, enteric nervous system | Cardiomyocytes/Endothelial cells/Cardiac fibroblasts | CA, DCA, CDCA,LCA, TCDCA | [43, 81] |
| M | Nervous, intestinal,gastrointestinal | Cardiomyocytes/Endothelial cells | TC, LCT, TCA | [55, 71] |
| S1P | Liver, nervous, immune | Cardiac fibroblasts/Endothelial cells/Vascular smooth muscle cells | TCA, UDCA | [19, 58, 80] |
| BKCa | Liver, brain | Cardiomyocytes/Vascular smooth muscle cells | LCA | [9, 59] |
## 3.1 Bile acid metabolism in cardiomyocytes
The heart is composed of noncardiomyocytes ($70\%$) and cardiomyocytes ($30\%$) [63]. BAs have direct and indirect effects on cardiac function. Their indirect effect is to directly affect the contraction and conduction of the myocardium through the interaction between BAs and cardiomyocytes.
In animal experiments, injecting large doses of BAs into animals causes significant bradycardia, indicating the cardiotoxicity of BAs and that BAs have time-varying and dose-dependent effects on cardiomyocytes [29].
In the heart, both cardiomyocytes and fibroblasts express FXR. FXR has a distinct stimulus-dependent effect in regulating cardiomyocyte injury [29]. Pu et al. [ 64] used cultured cardiomyocytes to prove that FXR expressed in cardiomyocytes was activated through mitochondrial death signaling. This finding was validated in a mouse model of myocardial ischemia/reperfusion injury in vivo [65]. In contrast, Xiaoli et al. found that FXR activation ameliorated cardiomyocyte damage induced by oxidative stress [66]. Furthermore, FXR activation reduced cardiomyocyte viability by triggering apoptosis. Therefore, it is speculated that FXR signaling is involved in several cardiac diseases associated with cardiomyocyte growth and apoptosis. FXR also regulates cardiac lipid accumulation in obese and diabetic patients by inducing the expression of β-oxidative genes in cardiomyocytes [67, 68]. As an agonist of FXR, GW4064 can significantly improve insulin resistance and cardiomyocyte disorders [69].
Clinical studies have demonstrated that within a certain concentration range, CDCA and DCA exert either a negative temporal effect by inhibiting the activity of the rat cardiac sinus node or a positive inotropic effect by increasing the concentration of Ca2+ in the cytoplasm of cardiomyocytes [70]. According to previous studies, LCA can reduce the apoptosis rate of cardiomyocytes [71].
The rate and pressure of myocardial contraction are determined by the rate of calcium influx. The t-tubule is the location that regulates intracellular calcium flow and contractility, and VDR is localized in the t-tubule of cardiomyocytes. Thus, loss of cardiomyocyte VDR selectivity results in cardiomyocyte hypertrophy, which affects the systolic and diastolic function of cardiomyocytes [72]. VD supplementation improves the left ventricular structure and restores cardiac function in patients with HF, further indicating that VDR is important in the maintenance of normal function in cardiomyocytes [73].
In addition, when the TGR5 gene is deleted in cardiomyocytes, the ability of the myocardium to adapt to the three stressors (physiological, inotropic, and hemodynamic stress) is significantly impaired. TGR5 can be readily targeted by BAs and their synthetic analogs and then regulate the expression of cardiac PDK4 by activating the Akt signaling pathway to improve cardiac glucose metabolism and play a beneficial role in patients with different types of CVDs [71].
Taurine cholic acid (TCA) acts as a partial agonist of M2 receptors. The binding of the M2 receptor to taurocholic acid inhibits cyclic AMP (cAMP), reduces myocardial cell contraction, and induces arrhythmia in CMs [74]. In cardiomyocytes, the activation of the S1P1 receptor, one of the most important expression subtypes of S1P, can also inhibit the synthesis of cAMP and antagonize adrenergic receptor-mediated contractility. Mohamed et al. reported that the protective effect of UDCA on CMs against hypoxia is partly similar to that of FTY720 (an S1P receptor agonist), which maintains normal intracellular [Ca2+] through S1P1 receptor-mediated hypoxia [75]. UDCA was also able to reverse fetal cardiomyocyte injury in a rat model of ICP [76]. Another study revealed that DCA and CDCA can also induce the production of cyclic adenosine monophosphate and reduce the contraction rate of neonatal mouse ventricular myocytes [77].
In summary, the relationship between BAs and cardiomyocytes involves the regulation of multiple receptors, which is a complex and multifactorial process.
## 3.2 Bile acid metabolism in endothelial cells
Endothelial dysfunction is one of the major drivers of CVDs such as atherosclerosis [78, 79]. FXR ligands in endothelial cells have been found to increase FXR expression, upregulate endothelial nitric oxide synthase (eNOS), reduce endothelin-1, and modulate angiotensin-II receptors, thereby inhibiting VSMC inflammation and migration [80, 81]. Endothelin-1 (ET-1) is the most effective vasoconstrictor currently available. GW4064, a chemical FXR agonist, was proven to increase eNOS expression [82]. Other studies have shown that activation of FXR prevents vasoconstriction mediated by increased eNOS and decreased ET-1 and that FXR can impair the vasorelaxation of endothelial cells under chronic stimulation [83, 84].
After identifying the role of FXR in lung endothelial cells, He et al. confirmed that CDCA activation resulted in a concentration-dependent decrease in endothelin-1 mRNA expression [22]. The expression of TGR5 was also found in aortic endothelial cells, which produce nitric oxide. S1P receptors present in endothelial cells can mediate endothelial cell responses to BAs. Furthermore, the BA-mediated activation of Ca2+-dependent K+ currents has been confirmed in endothelial cells [9]. Consistent with the effect of DCA on the receptors, the muscarinic M2 and M3 receptor responses to cardiac and vascular endothelial cells were attenuated in a model of liver cirrhosis [22].
## 3.3 Bile acid metabolism in vascular smooth muscle cells
Some data suggest that in vascular tissue, FXR not only regulates its own expression but also functions as a transcription factor in vascular smooth muscle cells (VSMCs). FXR regulates vasoconstriction and relaxation by altering the duration of other receptors in blood vessels and the production of active molecules.
In VSMCs, the expression of type II angiotensin receptors increases with FXR ligands. Studies have shown that FXR activation can inhibit the endothelin-1β-mediated induction of eNOS and COX-2 by upregulating ET-1 and eNOS, further inhibiting vascular smooth muscle cell inflammation and migration and finally inducing endothelial vasodilation [85]. However, chronic stimulation of FXR reduces cGMP sensitivity in smooth muscle cells and attenuates NO-dependent vasodilation [81]. Therefore, temporal variables should be considered when exploring BA receptor-related effects.
As another BA-sensitive receptor on VSMCs, S1PR2 mediates NO signaling and participates in peripheral vascular tone and endothelial cell responses. In addition, S1PR2 reduces NO levels in vascular injury by inhibiting the action of inducible NO synthase [86].
BAs also activate Ca2+-dependent K+ channels in VSMCs. In pressurized cerebral resistance arteries, blockers of BKCa channels are able to inhibit LCA-mediated endothelium-dependent vasodilation. In a mouse model, LCA was unable to stimulate arterial vasodilation after knockout of the BK β-1 subunit, indicating that the BK β-1 subunit plays a significant role in activating LCA [9]. Thus, these data suggest that BAs can stimulate vasodilation by activating BKCa channels in VSMCs and indicate a critical role for the BK β-1 subunit in CVDs.
## 3.4 Bile acid metabolism in cardiac fibroblasts
Cardiomyocytes and cardiac fibroblasts are the two most important resident cells in the heart, which participate in various pathophysiological processes of that organ and interact with each other.
The interleukin (IL)-1 family is considered essential in repairing and remodeling infarcted heart damage, and IL-1β is an important effector [87]. In an experiment involving the coculture of fibroblasts and cardiomyocytes, hypoxic stimulation showed that TGR5 mRNA expression was reduced in both types of cells. DCA inhibited the activation and expression of IL-1β in cardiomyocytes and fibroblasts under hypoxic conditions, and IL-1β mRNA expression was decreased in both cell lines [88]. Therefore, controlling BA metabolism by activating the DCA-TGR5 signaling pathway is thought to reduce postinfarction inflammation and improve cardiac function. This strategy may provide new therapeutic avenues for patients with myocardial infarction.
In the gut, FXR activation induces the expression of fibroblast growth Factor 19 (FGF19). In turn, FGF19 activates FGF receptor 4 (FGFR4) in the liver, which reduces BA synthesis by further inhibiting CYP7A1 [89]. Other studies have shown that S1PRs also play a role in the proliferation, remodeling, and differentiation of cardiac fibroblasts [19].
In summary, the relationship between BAs and CVD-related cells involves the regulation of multiple pathways and multisystem interactions.
**Table 2**
| Types of BA or its Derivatives | Receptors | Diseases | Roles in Diseases | REF. |
| --- | --- | --- | --- | --- |
| UDCA | TGR5 | Diabetic cardiomyopathy | Improvement of endoplasmic reticulum stress, blood glucose level and GLP-1 secretion in diabetic cardiomyopathy rats | [108] |
| UDCA | / | Atherosclerosis | Anti-atherosclerotic effects by reducing endoplasmic reticulum (ER) stress and pro-inflammatory responses | [122-124] |
| UDCA | M2/TGR5 | Arrhythmia | Protection of the myocardium by antagonizing other hydrophobic BAs and cardiac wavelengths to mediate antiarrhythmic effects | [57, 130-135] |
| UDCA | TGR5 | Heart failure | Enhancement of the adaptability of the heart to physiological, muscle strength, and hemodynamic stress | [137] |
| UDCA | FXR | Cirrhosis cardiomyopathy | Protection of liver cells by promoting bile flow, reducing liver enzyme levels and replacing hydrophobic BA | [10, 145] |
| DCA | TGR5 | Myocardial infarction | Inhibition of inflammatory responses in cardiomyocytes and fibroblasts through activation of the DCA-TGR5 signaling pathway | [88, 98] |
| CA | PXR | Diabetic cardiomyopathy | Regulation of lipid and energy metabolism to combat high-fat diet-induced obesity and insulin resistance, and increase in insulin secretion in pancreatic B cells for antidiabetic effects | [102, 103] |
| OCA | FXR | Diabetic cardiomyopathy | Improvement of metabolic abnormalities and impaired glucose tolerance, including lowering blood glucose and insulin levels and reducing body weight and heart weight,and protection against diabetic cardiomyopathy by activating the FXR-mediated Nrf2 signaling pathway | [110, 111] |
| INT-747 | | Atherosclerosis | Downregulation of the vasoconstrictor endothelin-1, thereby preventing smooth muscle cell-mediated atherosclerotic effects and migration processes | [95] |
| INT-747 | | Atherosclerosis | Inhibition of the accumulation of triglyceride- and phosphate-induced mineralization | [9] |
| INT-777 | TGR5 | Atherosclerosis | Improvement of metabolic syndrome and atherosclerosis | [71] |
| GUDCA | FXR | Atherosclerosis | 1) Improvement of cholesterol homeostasis by modulating gut microbiota, and by inhibiting foam cell formation. 2) improvement of local chronic inflammation, lipid deposition, plaque area, and plaque stability to slow the progression of atherosclerosis | [120] |
| CDCA | FXR | Atherosclerosis | Reduces hepatic lipolysis, cholesterol levels and bile acid efflux, activates hepatic FXR-BSEP signaling and reduces atherosclerotic damage | [119] |
| CDCA | LXR | Diabetic cardiomyopathy | Promotion of glucose metabolism by upregulating the expression of LXRs and increasing the secretion of GLP-1 and glucagon | [106, 107] |
## 4. The Role of Bile Acid Metabolism in Cardio-vascular Diseases
Previous studies have shown that elevated concentrations of BAs can reduce the heart rate and cardiac contractility in rats. Taurine deoxycholic acid may improve cardiac contractility by inhibiting endoplasmic reticulum (ER) stress, apoptosis, inflammation, and fibrosis [90]. Additionally, elevated serum BA levels are related to adult arrhythmia, poor contractility of cardiomyocytes, and poor fetal outcomes in pregnant women with obstetric cholestasis [91]. In contrast, UDCA, the most hydrophilic BA, has been proven to help improve chronic heart failure and to play a protective role in cardiac ischemia-reperfusion injury and myocardial infarction [10].
Furthermore, the composition of BA pools has been altered in patients with chronic heart failure [92]. In patients with liver cirrhosis, cardiac dysfunction is closely related to the increase in serum BA concentrations [10]. Based on relevant studies, we know that different BAs may have different effects on cardiac function. Some recent evidence indicates that BAs not only affect the pathogenesis of metabolic diseases but may also serve as markers of these diseases [93]. Thus, BAs may be potential biomarkers of metabolic health and diseases. In clinical conditions, tracking BAs through advanced analytical techniques may provide a potential and effective avenue for identifying new treatments for cardiovascular diseases (Table 2, Fig. 2).
## 4.1 Bile acid metabolism in myocardial infarction
The pathogenesis of myocardial infarction (MI) and its complications involve a variety of metabolic disorders. Various inflammatory responses during heart remodeling after MI are critical for cardiac repair. Metabolic changes also affect systemic inflammatory activation status. Therefore, increased attention is directed to other pathways that modulate the inflammatory response by modulating metabolic pathways. There are an increasing number of studies on BA metabolites as signaling molecules that affect various cardiovascular functions [94]. In recent years, studies have suggested that BAs can directly regulate a variety of pathophysiological processes. Research has also confirmed that myocardial infarction is intrinsically linked to cholesterol metabolism regulated by Bas [95].
FXR agonists can improve cardiac dysfunction after myocardial infarction by stimulating adiponectin secretion [96]. Moreover, FXR knockout maintains cardiac function after myocardial infarction by reducing cellular fibrosis and chronic apoptosis [97]. DCA is one of the most potent activators of TGR5. The protective role of DCA in cardiac repair after myocardial infarction is based on its anti-inflammatory effect. The activation of DCA-TGR5 signaling can inhibit the inflammatory response of cardiomyocytes and fibroblasts, which helps DCA play a protective role in myocardial infarction [88, 98]. The levels of DCA in patients with AMI were significantly lower than those in controls. Interestingly, after DCA supplementation, the area of myocardial infarction was reduced, and heart function was also improved [99]. The authors of another study found that TGR5 regulates the function and subpopulation distribution of CD4+ T cells in the heart, thus playing a protective role in myocardial infarction [100]. UDCA and its conjugated metabolite GUDC have been reported to be decreased in patients with acute myocardial infarction [88].
Figure 2.BAs affect cardiovascular disease by binding to various receptors that affect metabolism and regulation of lipid profiles. These receptors include nuclear receptors (FXR, PXR, VDR, LXR), G protein-coupled receptor (TGR5, muscarinic receptor, S1PR) and Ca2+-activated potassium (K+) (BK) channels. These receptors are highly expressed in cells associated with the cardiovascular system, such as cardiomyocytes, endothelial cells, cardiac fibroblasts and vascular smooth muscle cells. By binding to receptors on cells, BAs further influence intracellular regulators related to CVD risk. They are ultimately involved in the occurrence and development of cardiovascular system diseases such as cardiomyopathy, atherosclerosis, arrhythmia, and heart failure by affecting the relevant regulatory factors of cardiovascular disease risk.
In conclusion, DCA, as one of the strongest ligands of BAs, mainly mediates its biological function via TGR5, which plays a mitigating role in the process of myocardial infarction. Therefore, strategies aimed at regulating BA metabolism and related signal transduction to improve the inflammatory response may be helpful for patients with myocardial infarction.
## 4.2 Bile acid metabolism in diabetic cardiomyopathy
Diabetic cardiomyopathy (DCM) is considered the leading cause of high mortality from heart disease, and it is also one of the common cardiac complications in diabetic patients. Hyperglycemia and insulin resistance are key factors in the pathogenesis of DCM and are associated with inflammation, oxidative stress, and mitochondrial dysfunction. The pathological mechanisms underlying DCM include apoptosis, hyperglycemia and hyperlipidemia, the accumulation of extracellular matrix, the disturbance of calcium homeostasis in cardiomyocytes, and diastolic dysfunction [101].
In one animal study, researchers found that PXR activation mitigated obesity and insulin resistance caused by high-fat diets by regulating lipid and energy metabolism, indicating that PXR plays an important antidiabetic role [102]. Studies have shown that CA and CDCA are associated with a reduced risk of diabetes. CA exerts anti-diabetic effects by increasing insulin production in pancreatic B cells [103]. As an intestinal hormone, incretin GLP-1 stimulates insulin secretion and sensitivity, glucose production and lipolysis, and it also increases satiety, which is beneficial to the human body [104]. In previous studies, GLP-1 was shown to have cardioprotective effects, such as regulating cardio-myocyte function and reducing atherogenic plaque inflammation [105]. CDCA promotes glucose metabolism by upregulating the expression of LXRs and increasing the secretion of GLP-1 and glucagon [106, 107]. Additionally, GLP-1 secretion can be regulated by TGR5. In bariatric surgery, TGR5 improves glucose homeostasis [108].
Another study demonstrated that UDCA reduces ER stress in rats with diabetic cardiomyopathy, dissolves cholesterol formed in the gallbladder, and then reduces the absorption of cholesterol. Furthermore, UDCA has been shown to reduce oxidative damage induced by hydrophobic Bas [71]. A study by Basso et al. showed that an increase in UDCA and its conjugates increased insulin sensitivity during bariatric surgery [109]. Rat UDCA levels are elevated after partial gastrectomy, which correlates with the distribution of fat and the enhancement of insulin sensitivity. Ingestion of UDCA particles in healthy individuals improves blood glucose levels and GLP-1 secretion, promotes gastric emptying, and modulates glucose-induced insulin excretion [109]. Therefore, BAs may play a significant role in diabetic cardiomyopathy through TGR5-mediated GLP-1 secretion.
Because of the limited effect of UDCA, obeticholic acid (6-ethyl goose deoxycholic acid [OCA]) has been synthesized with CDCA. OCA is currently the most clinically advanced BA derivative. It is anticipated that OCA can be used as a new therapeutic drug to replace UDCA. OCA has a variety of biological and pharmacological applications. As a semisynthetic BA analog, it has a strong binding affinity to FXR [110]. Wu et al.[101]. confirmed that OCA improves metabolic abnormalities and reduces impaired glucose tolerance, including lowering blood glucose and insulin levels and reducing body weight and heart weight. Data have shown that in diabetic mice, OCA exhibits antioxidant activity and protects against diabetic cardiomyopathy by activating the FXR-mediated Nrf2 signaling pathway [111].
The metabolic effects of several drugs commonly used in the clinical treatment of diabetes are generally dependent on the regulation of BA metabolism. Among them, metformin is thought to have a hypoglycemic effect by reducing the intestinal absorption of BA [112]. The synthesis and metabolism of BAs may change with the course of diabetic cardiomyopathy. Therefore, more research is needed to determine whether the risk of diabetic cardiomyopathy can be reduced by interfering with the content of BA in the serum. These studies may provide new insights into the diagnosis and treatment of diabetic cardiomyopathy.
## 4.3 Bile acid metabolism in atherosclerosis
Several studies have shown that hydrophobic BAs can be blocked by inhibiting ER stress, free cholesterol-induced cell death in macrophages, and the presentation of major histocompatibility complex (MHC)-related antigens. Fat production increases the risk of developing athero-sclerosis and other cardiovascular diseases [113]. It has been reported that BA sequestrants regulate blood cholesterol levels, thereby affecting the formation of atherosclerotic plaques [114].
FXR has a tissue-specific role in the development and progression of atherosclerosis, and its prevention depends not only on hepatic FXR activation but also on global and intestinal depletion of FXR [115-117]. Activation of FXR is associated with regulators in the liver and gut that affect CVD, such as endothelial function, lipid and glucose homeostasis, and athero-sclerosis. It is associated with maintaining normal cholesterol triacylglycerol levels [19]. FXR also regulates inflammation in blood vessels. Synthetic FXR ligands were able to inhibit the inflammatory response of rat smooth muscle cells, which is induced by interleukin-1β, suggesting that FXR agonists have antiatherosclerotic potential [118]. Calcification is a feature of athero-sclerosis. The CDCA derivative INT-747 can inhibit the accumulation of triglyceride- and phosphate-induced mineralization. The anti-calcium effect of INT-747 is regulated by FXR, and when FXR is inhibited, the mineralization of CVCs is increased [9]. In ovariectomized mice, increased levels of CDCA in the liver activate hepatic FXR-BSEP (bile salt export pump) signaling and reduce atherosclerotic damage by reducing hepatic lipolysis, cholesterol levels, and bile acid efflux [119]. As a gut FXR antagonist, GUDCA may improve cholesterol homeostasis by modulating gut microbiota. In addition, by inhibiting foam cell formation, GUDCA improves local chronic inflammation, reduces lipid deposition and plaque area, and improves plaque stability to slow the progression of atherosclerosis [120].
OCA, another potent FXR agonist, has shown treatment efficacy in preventing high-fat diet-induced atherosclerosis. There are some data supporting the ability of FXR agonists to downregulate the vasoconstrictor endothelin-1, thereby preventing smooth muscle cell-mediated atherosclerotic effects and migration processes [95]. As shown, the role of FXR in atherosclerosis is complicated, and more research is needed to more fully evaluate the effects of long-term FXR stimulation on atherosclerosis, as well as more in vivo experiments to determine the BA-FXR interaction. LXR has also been reported to regulate CVDs such as atherosclerosis. Bradley et al. showed that activation of LXR reduces the formation of atherosclerotic lesions [10]. However, the BA/LXR signaling pathway mainly functions in animal models. Therefore, further research is needed on the BA/LXR signaling pathway.
One study claimed that in the heart, TGR5 can inhibit inflammation and the formation of atherosclerotic plaques, thereby improving atherosclerosis [121]. A semisynthetic derivative of CA, 6α-ethyl-23(S)-methylcholic acid (S-MECA, INT777), which acts as a TGR5 agonist, has been shown to improve metabolic syndrome and reduce atherosclerosis in mice. It is suggested that TGR5 plays a potential role in atherosclerosis prevention [71]. In addition, in bovine aortic endothelial cells, activation of TGR5 also inhibited NF-κB activity and induced NO production, inhibiting monocyte adhesion, macrophage lipid load and intraplaque inflammation and thereby preventing the accumulation of atherosclerotic plaque in the arteries [121].
Encouragingly, it has been reported that in a mouse model of diabetic atherosclerosis, hydrophilic BA-UDCA was able to exert antiatherosclerotic effects by alleviating ER stress and proinflammatory responses [122, 123]. An experimental study by Hanafi et al. reported that UDCA mediated the direct protection of the heart by regulating the ERK/Akt pathway [124].
Although BAs play a crucial role in the progression of atherosclerosis, the potential value of BA metabolism in the early stages of atherosclerosis remains unclear. Therefore, it is necessary to further explore the biological mechanism of BAs in the occurrence and progression of atherosclerosis.
## 4.4 Bile acid metabolism in arrhythmia
It has been gradually discovered and confirmed that high levels of BAs can cause various types of arrhythmias through various mechanisms. Furthermore, BA-induced arrhythmias are more common in fetuses than adults [125].
Data show that changes in the composition of BA pools in patients’ serum can induce atrial arrhythmias. In some in vivo experiments, the researchers found a significant increase in the proportion of BAs other than UDCA in the plasma of the atrial fibrillation group. Therefore, serum UDCA concentrations and the non-UDCA ratio may serve as independent predictors of atrial fibrillation [95]. Furthermore, the most hydrophilic BA, UDCA, has been proven to be cardioprotective against BA-induced arrhythmias in a cholestatic fetal heart model [126].
BA concentrations in patients with intrahepatic cholestasis (ICP) are associated with ventricular arrhythmias. Patients with PBC have a significantly prolonged corrected QT interval, which causes ventricular arrhythmias and further increases the risk of sudden death [127]. The increase in the concentration of BA in pregnant women with ICP could lead to the accumulation of BAs in fetal serum and fetal arrhythmia [128]. Sheikh et al.[129] stimulated the heart with the M2 receptor agonist carbachol and found the onset of bradycardia in mice. Moreover, abrogation of the M2 receptor improved TCA-induced cardiac arrhythmias in a fetal heart model. Therefore, it is suggested that TCA-induced arrhythmias are mediated by partial agonism of M2 receptors.
In another study, Ibrahim et al. described other possible mechanisms by which elevated serum BA levels could affect fetal arrhythmias [57]. Elevated concentrations of secondary BAs are known to cause TGR5-mediated cAMP release in cardiomyocytes without altering contractility. However, secondary BAs act as partial agonists of M2 receptors with a concomitant reduction in contraction rate. Therefore, partial agonism of M2 receptors may serve as a novel mechanism by which BAs induce arrhythmias. This mechanism is expected to be a new target for the treatment of adult and fetal cardiac arrhythmias.
In animal experiments, elevated concentrations of BA tended to lead to arrhythmias and cardiac dysfunction, but UDCA was able to protect the myocardium by antagonizing other hydrophobic Bas [130]. The antiarrhythmic protective effect of UDCA has been validated in a rat fetal cardiac cholestasis model in vitro [131]. Another study has shown that UDCA may mediate the antiarrhythmic effect through the increase in cardiac wavelength, which suggested that the treatment of UDCA for arrhythmias has potential value [132]. In a coculture model of neonatal rat CM-myofibroblasts, UDCA can also depolarize myofibroblasts to prevent ventricular conduction slowing and arrhythmias [133]. More recently, data from Ferraro and associates have shown that the effects of UDCA on arrhythmias are not limited to fetal myocardium [134].
UDCA was shown to protect CMs against arrhythmias mediated by adenosine triphosphate-gated K+ channels and [Ca2+]I [135]. Other studies indicated that UDCA could alter the expression of BA transporters and metabolism-related genes in cardiomyocytes. Hence, it is speculated that the protective effect of UDCA on the heart may be similar to that of dexamethasone in that it has a protective effect on the contractility of cardiomyocytes during arrhythmias.
## 4.5 Bile acid metabolism in heart failure
In patients with chronic heart failure, the serum concentration of secondary BAs was found to have increased, thus resulting in a larger proportion of secondary BAs in the BA pool [136]. BAs act as polar amphiphiles to affect the exchange of sodium and calcium ions on the myocardial cell membrane, inducing backward depolarization of the cell. Subsequent depolarization is one of the initiating mechanisms of heart failure. During the treatment of chronic heart failure, UDCA has been shown to improve peripheral blood flow and liver function in patients by improving vasodilation (both endothelium-dependent and -independent), thereby ensuring NO production in impaired arterial blood flow [137]. In another clinical study, peripheral blood flow improved after extremity ischemia in patients with chronic heart failure who were treated with 500 mg of UDCA twice daily for four weeks [92].
It has been reported that TGR5 agonists enhance the adaptability of the heart to physiological, muscle strength, and hemodynamic stress, further inducing changes in its protective mechanisms [71]. Therefore, TGR5 may be a potential therapeutic target for heart failure. Studies have shown that FXR is downregulated in the left ventricle of spontaneously hypertensive rats with end-stage heart failure [95], but more evidence is needed.
## 4.6 Bile acid metabolism in cirrhosis cardiomyopathy
The deterioration of cholestatic disease leads to liver cirrhosis [138]. The course of patients with cirrhosis is based on the severity of complications caused by changes in the internal structure and overall metabolism of the liver. Patients with visceral and arterial vasodilation develop an abnormal heart rate. Cirrhosis is often accompanied by worsening cardiac output and cardiac insufficiency, as well as changes in cardiac structure and size and impaired function [139, 140]. When the cardiac output increases, arterial blood pressure and systemic vascular resistance decrease, forming a “hyperdynamic circulatory state.” This chronotropic and inotropic cardiac insufficiency is known as “cirrhotic cardiomyopathy,” a type of severe cardiovascular disease characterized by advanced cardiac fibrosis and remodeling [22]. As endogenous amphiphilic products of cholesterol metabolism, BA may be a source of liver cirrhosis and heart disease and is associated with cardiac hypertrophy and atherosclerotic lesions [70].
In models of liver cirrhosis, several mechanisms associated with BAs have been proposed to induce vasodilation of splanchnic and systemic vessels, resulting in hyperdynamic circulation. Therefore, research on the relationship between the pathophysiological characteristics of cirrhosis cardiomyopathy and the abnormal metabolism of BA has sparked interest among researchers [141]. According to previous research results, in the fasting state, the normal level of serum BAs in adults was less than 15 μmol/L, and the serum BA concentration in patients with liver cirrhosis was greater than 100 μmol/L, indicating the development of cirrhotic cardiomyopathy [142]. Therefore, elevated serum BA may be associated with the occurrence and development of cirrhotic cardiomyopathy.
TGR5 can regulate metabolic homeostasis in the heart. Activation of the TGR5 signaling pathway can prevent nonalcoholic steatohepatitis by reducing vascular and liver damage [143]. S1P1R and S1P2R, which are also highly expressed in hepatocytes, are involved in the activation of protein kinase B and extracellular signal-regulated kinase $\frac{1}{2}$, regulating vasodilation and increasing blood flow. Therefore, BAs are able to mediate the hemodynamic complications involved in cirrhosis through S1PR. BAs have also been proven to activate the BK pathway and increase the probability of cardiomyopathy in patients with liver cirrhosis [40]. Increased hydrophobic BAs in patients with liver cirrhosis can lead to QT interval prolongation and arrhythmias [144]. BK channels may play an important role in cardiac conduction, and BAs may increase the risk of cirrhotic cardiomyopathy by activating the BK pathway in patients with liver cirrhosis.
In recent years, UDCA, a highly hydrophilic secondary BA, has been shown to act as an alternative drug to protect hepatocytes by promoting bile flow and reducing liver enzyme levels [145]. Furthermore, the replacement of hydrophobic BAs with UDCA reduced cardiac injury in both cirrhotic and noncirrhotic portal stenosis models, suggesting that BAs themselves are important factors in the development of cirrhotic cardiomyopathy [10].
Much of the abovementioned evidence suggests that BAs can affect or regulate the function of the heart in cirrhotic cardiomyopathy. However, to date, there have been few clinical studies specifically targeting the interaction of BAs and cardiovascular function in patients with liver cirrhosis, and more trials are needed in the future.
## 4. Clinical Diagnosis and Application Value of Bile Acids
As the study progressed, researchers have begun to explore the diagnostic and prognostic value, as well as pharmacological application, of BA metabolism and related signaling pathways in cardiovascular and metabolic diseases.
In a previous study by our team, Liao Y et al [146] used metabolomics analysis to explore the changes in systemic and cardiac metabolites in patients with aortic stenosis (AS) before and after transcatheter aortic valve replacement (TAVR) surgery. It was found for the first time that TAVR surgery contributed to a significant increase in primary BA synthesis in patients with AS. Furthermore, based on the accession number GSE141910 from the NCBI GEO database (www.ncbi.nlm.nih.gov/geo), we analyzed the expression level of the regulatory genes and receptors related to BA synthesis and metabolism in dilated cardiomyopathy (DCM) patients, hypertrophic cardiomyopathy (HCM) patients and nonfailing healthy donors. According to the obtained table and heatmap, the expression of most of the BA metabolism-related genes was disturbed compared with that in the nonfailing healthy donors (Table 3, Fig. 3).
In another study, Li W et al.[147] investigated the relationship between serum total BAs (TBAs) and coronary artery disease (CAD). Fasting TBA levels were measured in 7,438 participants who underwent coronary angiography, and the results showed that fasting serum TBA levels were positively correlated with the severity of coronary lesions, coronary artery disease, and MI.
As a second-generation bile acid sequestrant (BAS), colesevelam is approved for the treatment of type 2 diabetes mellitus (T2DM) and hyperlipidemia [148]. Colesevelam improves glycemic control in patients with T2DM, but its mechanism underlying the glucose-lowering effect is not fully understood [149, 150]. In clinical trials, colesevelam was able to lower blood glucose levels [151]. Clinical data indicate that colesevelam reduces total plasma cholesterol levels by $10\%$ and LDL-C levels by $15\%$ [152]. Furthermore, colesevelam reduces the risk of CVD by lowering the level of LDL-C in the plasma during the process of cholesterol to bile conversion.
Figure 3.A heatmap analysis of the expression of regulated genes and associated receptors during BA anabolism in DCM and HCM patients based on the accession number GSE141910 from the NCBI GEO database (www.ncbi.nlm.nih.gov/geo).
Chevli PA et al. [ 153] investigated the relationship between plasma metabolites and subclinical atherosclerosis in 700 patients with type 2 diabetes and found that coronary artery calcium (CAC) was positively correlated with the BA metabolic subpathways. However, another clinical study initiated by Feng X et al. [ 154] showed that in postmenopausal women with type 2 diabetes, TBA was inversely associated with the occurrence of coronary artery disease and myocardial infarction. This result suggests that sex may influence the association of BA with CVD. Moreover, BAs are inherently sexually dimorphic in humans. Due to the higher activity of 12 α-hydroxylase in women, among the types of BAs, chenodeoxycholic acid levels are higher in women, and cholic acid levels are higher in men [155]. However, the total BA pool in men is larger than that in women [156]. These findings highlight that biological sex is also an important factor to consider when investigating potential treatment strategies.
In a tissue Doppler imaging study, the authors investigated the association of TBA levels and fetal cardiac function in women with intrahepatic cholestasis (ICP). Elevated maternal and fetal serum BA levels in severe ICP have been found to be associated with abnormal fetal cardiac phenotype and fetal cardiac insufficiency compared with those in healthy women with normal TBA levels and in women with mild ICP. In particular, when the maternal TBA level was greater than 440 mmol/L, the incidence of fetal complications, such as spontaneous preterm deliveries, asphyxia events, and meconium staining, was significantly higher [157]. However, the deteriorated fetal cardiac phenotype was partially attenuated by UDCA treatment [158].
In a prospective, single-center study, elevated levels of specific secondary BAs and decreased levels of primary BAs were found in patients with chronic HF. Specifically, Mayerhofer CCK et al.[159] measured the plasma levels of primary, secondary and total BAs in 142 chronic heart failure patients and 20 sex- and age-matched healthy controls to explore the association of BAs with clinically relevant variables and the long-term survival rate. It was found that plasma levels of primary BAs decreased, secondary BAs increased and the ratio of secondary BAs to primary BAs increased in HF patients compared with healthy controls. After a median follow-up time of 5.6 years, the patients in the highest tertile (T3) of the ratio of secondary to primary BAs had an approximately twofold mortality rate compared with the patients in the lowest tertile (T1), although this association was weakened after correcting for other confounders. In another single-center study, Voiosu AM et al [160] showed that total BA levels correlated with cardiac output and left atrial volume in patients with cirrhosis. The authors evaluated 58 patients with cirrhosis according to the Child classification, 49 of whom had decompensated cirrhosis. Patients' total BA levels (median, 45 µmol/L) were associated with increased left atrial volume in multivariate analysis and several echocardiographic parameters of hyperdynamic syndrome in univariate analysis.
**Table 3**
| Gene symbol | Gene name | DCM | HCM |
| --- | --- | --- | --- |
| CYP7A1 | / | Down | Down |
| CYP7B1 | / | Satble | Satble |
| CYP8B1 | / | Up | Up |
| CYP27A1 | / | Up | Up |
| FXR1 | / | Satble | Satble |
| FXR2 | / | Down | Down |
| PXR | NR1I2 | Satble | Satble |
| LXR | NR1H3 | Satble | Satble |
| VDR | / | Down | Down |
| TGR5 | GPBAR1 | Satble | Satble |
| S1PR1 | / | Down | Down |
| S1PR2 | / | Down | Satble |
| S1PR3 | / | Down | Down |
| S1PR4 | / | Up | Satble |
| S1PR5 | / | Up | Up |
| BKCA alpha | KCNMA1 | Up | Up |
| BKCA beta1 | KCNMB1 | Up | Up |
| BKCA beta2 | KCNMB2 | Up | Up |
| BKCA beta3 | KCNMB3 | Satble | Satble |
| BKCA beta4 | KCNMB4 | Satble | Up |
Furthermore, BA homeostasis is jointly maintained by hepatic and intestinal BA signaling pathways. BA induces enterohepatic feedback signals by releasing intestinal hormones and regulates enterohepatic circulation [161]. The role of the gut microbiota as a regulator of intestinal BA metabolism is gradually being implicated in the development of human cardiometabolic diseases by increasing evidence [162, 163]. In addition to BA, trimethylamine nitroxide (TMAO), a gut microbiota-derived metabolite, has recently been implicated in the pathogenesis of CVDs [164, 165]. Both in vitro and in vivo studies in humans have shown that TMAO has pleiotropic negative effects on the cardiovascular system [166-168]. TMAO promotes atherosclerosis and ventricular remodeling by regulating BA metabolism, leading to vascular dysfunction [169, 170]. Several nonantibiotic small-molecule inhibitors targeting gut microbial choline-TMA lyase are already available [171, 172]. Several preclinical animal model studies have demonstrated that these drugs have great therapeutic potential for various cardiometabolic diseases. They can effectively exert antiatherosclerotic, antiobesity and antithrombotic effects [173, 174]. This also confirms that the gut microbial TMAO pathway is closely related to host BA metabolism and provides another new possible avenue for developing drugs for the treatment of human cardiometabolic diseases.
In addition, in recent years, an increasing number of studies have confirmed the pharmacological applications of bile acid derivatives. BAs are considered to be very helpful for the preparation of novel drugs due to their rigid backbone and potential for surface amphiphilicity [175]. The broad availability, inherent chemical and biological properties and facile derivatization methods of BAs render them useful as scaffolds in drug, supramolecular, and materials chemistry [176].
BAs serve as an attractive cornerstone for designing novel hydrogel systems for the delivery of biomolecules, drugs and vaccines [177, 178]. This has attracted the attention of many researchers, making it a new area of research that warrants increasing attention. BAs may open a new avenue for drug therapy of cardiovascular disease.
## 6. Conclusion and Perspectives
Recently, evidence has accumulated indicating that the relationship between BA metabolism disturbances and CVDs is closely related. When BA metabolism is disordered, a series of cardiac dysfunction and CVDs may also be present. In this paper, we have clarified the metabolic mechanism of BAs and their pharmacological potential in regulating cardiovascular function. BA signaling plays an important role in different cell types through receptor-dependent or channel-mediated mechanisms. Future work should aim to further elucidate the deeper interactions between BAs and their receptors to facilitate the development of new treatments for CVDs.
Current clinical studies as well as our previous metabolomic and bioinformatics analyses have revealed that TBA levels, BA pool composition ratios, and BA-related receptors are partially disturbed in human CVD. However, whether altered BA in humans can serve as a potential biomarker in the pathogenesis of CVD remains unclear and warrants further study.
Fortunately, UDCA has now been found to play a protective role in CVD, although its specific protective mechanism has not been fully elucidated. Currently, some drugs targeting UDCA and its alternatives, some synthetic BA analogs such as OCA and Colesevelam, have been used in clinical practice. However, to further confirm their importance in cardioprotection, more information on their application in preclinical and clinical studies should be provided in the future.
Recent reports have focused on the application of BAs in the preparation of new drugs. BAs have become the main molecules in drug carrier systems due to their good compatibility with different biologically active compounds, showing great potential in medical and biological applications. However, there is no further research on the development of drugs for the treatment of CVDs. Furthermore, the safety and effectiveness of drugs targeting BAs should be evaluated in the treatment of patients with diverse CVDs. Therefore, it is expected that the uniqueness of BAs as drug carriers can be fully utilized to complete further development in the future.
In addition, it is necessary to conduct studies to investigate the effects of drugs that modulate BA metabolism or signaling pathways on lipid metabolism and other related proteins in patients with CVD in the future. This can form a better pharmacological basis for the clinical treatment of CVDs such as atherosclerosis, coronary heart disease and heart failure.
Another potentially interesting area of research is the possible role of VD as a therapeutic target for cardio-vascular diseases. There is evidence that VD may affect cardiovascular function through multiple pathways and that VDR plays an important role in BA transport, metabolism, and detoxification. However, current research on the association between BA and its receptor, VDR, is still insufficient. More research in this area will be needed in the future to further elucidate whether altered VD may serve as a potential biomarker in cardiovascular pathogenesis.
It is worth noting that current research on the mechanism of BA metabolism and CVD progression is mainly carried out in rodents. Nevertheless, their BA metabolism is fundamentally different from that of humans. Some animals do not have gallbladders, such as rats. Therefore, some preclinical findings may vary by species, and the use of animal models to study BA metabolism still has some limitations. Translating findings from animal models into humans is challenging. Future studies need to consider important issues related to species limitations in clinical trial design and seek more efficient ways to explore the potential roles of BA metabolism and BA pool components in human CVD clinically.
Furthermore, the manipulation of factors affecting BA metabolism is unclear. Further research needs to consider whether some factors, including gender dimorphism of BAs and potential signaling crosstalk between gut microbiota and BA signaling pathways, can alter clinical outcomes. Further studies have the potential to open a new era for the application of BA in clinical practice to further prevent the risk of CVD.
## References
1. de Waard AM, Hollander M, Korevaar JC, Nielen M, Carlsson AC, Lionis C. **Selective prevention of cardiometabolic diseases: activities and attitudes of general practitioners across Europe**. *Eur J Public Health* (2019) **29** 88-93. PMID: 30016426
2. Murphy AJ, Febbraio MA. **Immune-based therapies in cardiovascular and metabolic diseases: past, present and future**. *Nat Rev Immunol* (2021) **21** 669-679. PMID: 34285393
3. Desai MS, Mathur B, Eblimit Z, Vasquez H, Taegtmeyer H, Karpen SJ. **Bile acid excess induces cardiomyopathy and metabolic dysfunctions in the heart**. *Hepatology* (2017) **65** 189-201. PMID: 27774647
4. Desai MS, Eblimit Z, Thevananther S, Kosters A, Karpen SJ. **Cardiomyopathy reverses with recovery of liver injury, cholestasis and cholanemia in mouse model of biliary fibrosis**. *Liver International Official Journal of the International Association for the Study of the Liver* (2015) **35** 1464-1477. PMID: 24330504
5. Guan B, Tong J, Hao H, Yang Z, Chen K, Xu H. **Bile acid coordinates microbiota homeostasis and systemic immunometabolism in cardiometabolic diseases**. *Acta Pharm Sin B* (2022) **12** 2129-2149. PMID: 35646540
6. Lazarevic S, Danic M, Golocorbin-Kon S, Al-Salami H, Mikov M. **Semisynthetic bile acids: a new therapeutic option for metabolic syndrome**. *Pharmacol Res* (2019) **146** 104333. PMID: 31254667
7. Wang Z, Zhao Y. **Gut microbiota derived metabolites in cardiovascular health and disease**. *Protein Cell* (2018) **9** 416-431. PMID: 29725935
8. Orozco-Aguilar J, Simon F, Cabello-Verrugio C. **Redox-Dependent Effects in the Physiopathological Role of Bile Acids**. *Oxid Med Cell Longev* (2021) **4847941**
9. Khurana S, Raufman J, Pallone TL. **Bile Acids Regulate Cardiovascular Function**. *Clinical and Translational Science* (2011) **4** 210-218. PMID: 21707953
10. Hanafi NI, Mohamed AS, Sheikh Abdul Kadir SH, Othman MHD. **Overview of Bile Acids Signaling and Perspective on the Signal of Ursodeoxycholic Acid, the Most Hydrophilic Bile Acid, in the Heart**. *Biomolecules* (2018) **8** 159. PMID: 30486474
11. Chiang J, Ferrell JM, Wu Y, Boehme S. **Bile Acid and Cholesterol Metabolism in Atherosclerotic Cardiovascular Disease and Therapy**. *Cardiol Plus* (2020) **5** 159-170. PMID: 34350368
12. Li T, Chiang JY. **Bile acid signaling in metabolic disease and drug therapy**. *Pharmacol Rev* (2014) **66** 948-983. PMID: 25073467
13. Rizzolo D, Kong B, Taylor RE, Brinker A, Goedken M, Buckley B. **Bile acid homeostasis in female mice deficient in Cyp7a1 and Cyp27a1**. *Acta Pharm Sin B* (2021) **11** 3847-3856. PMID: 35024311
14. Wahlstrom A, Sayin SI, Marschall HU, Backhed F. **Intestinal Crosstalk between Bile Acids and Microbiota and Its Impact on Host Metabolism**. *Cell Metab* (2016) **24** 41-50. PMID: 27320064
15. Hoogerland JA, Yu L, Wolters JC, Boer J, Oosterveer MH. **Glucose-6-Phosphate Regulates Hepatic Bile Acid Synthesis in Mice**. *Hepatology* (2019) **70**
16. Jia W, Wei M, Rajani C, Zheng X. **Targeting the alternative bile acid synthetic pathway for metabolic diseases**. *Protein & Cell* (2021) **12** 411-425. PMID: 33252713
17. Simmermacher J, Sinz M. **Evaluation of Farnesoid X Receptor Target Gene Induction in Human Hepatocytes: Amino Acid Conjugation**. *Drug Metab Lett* (2017) **11** 138-143. PMID: 29283075
18. Zheng X, Huang F, Zhao A, Lei S, Zhang Y, Xie G. **Bile acid is a significant host factor shaping the gut microbiome of diet-induced obese mice**. *BMC Biol* (2017) **15** 120. PMID: 29241453
19. Pushpass RG, Alzoufairi S, Jackson KG, Lovegrove JA. **Circulating bile acids as a link between the gut microbiota and cardiovascular health: impact of prebiotics, probiotics and polyphenol-rich foods**. *Nutrition Research Reviews:* (2021) 1-20
20. Sukocheva OA, Furuya H, Ng ML, Friedemann M, Menschikowski M, Tarasov VV. **Sphingosine kinase and sphingosine-1-phosphate receptor signaling pathway in inflammatory gastrointestinal disease and cancers: A novel therapeutic target**. *Pharmacol Ther* (2020) **207** 107464. PMID: 31863815
21. Yanagida K, Engelbrecht E, Niaudet C, Jung B, Gaengel K, Holton K. **Sphingosine 1-Phosphate Receptor Signaling Establishes AP-1 Gradients to Allow for Retinal Endothelial Cell Specialization**. *Dev Cell* (2020) **52** 779-793. PMID: 32059774
22. Voiosu A, Wiese S, Voiosu T, Bendtsen F, Møller S. **Bile acids and cardiovascular function in cirrhosis**. *Liver International* (2017) **37** 1420-1430. PMID: 28222247
23. Swales KE, Moore R, Truss NJ, Tucker A, Warner TD, Negishi M. **Pregnane X receptor regulates drug metabolism and transport in the vasculature and protects from oxidative stress**. *Cardiovasc Res* (2012) **93** 674-681. PMID: 22166712
24. Fiorucci S, Distrutti E, Carino A, Zampella A, Biagioli M. **Bile acids and their receptors in metabolic disorders**. *Prog Lipid Res* (2021) **82** 101094. PMID: 33636214
25. Ding L, Yang L, Wang Z, Huang W. **Bile acid nuclear receptor FXR and digestive system diseases**. *Acta Pharm Sin B* (2015) **5** 135-144. PMID: 26579439
26. Zhang B, Kuipers F, de Boer JF, Kuivenhoven JA. **Modulation of Bile Acid Metabolism to Improve Plasma Lipid and Lipoprotein Profiles**. *J Clin Med* (2021) **11**
27. Panzitt K, Zollner G, Marschall HU, Wagner M. **Recent advances on FXR-targeting therapeutics**. *Mol Cell Endocrinol* (2022) **552** 111678. PMID: 35605722
28. Sun L, Xie C, Wang G, Wu Y, Wu Q, Wang X. **Gut microbiota and intestinal FXR mediate the clinical benefits of metformin**. *Nat Med* (2018) **24** 1919-1929. PMID: 30397356
29. Zhou W, Anakk S. **Enterohepatic and non-canonical roles of farnesoid X receptor in controlling lipid and glucose metabolism**. *Mol Cell Endocrinol* (2022) **549** 111616. PMID: 35304191
30. Fang S, Suh JM, Reilly SM, Yu E, Osborn O, Lackey D. **Intestinal FXR agonism promotes adipose tissue browning and reduces obesity and insulin resistance**. *Nat Med* (2015) **21** 159-165. PMID: 25559344
31. Li P, Zhu L, Yang X, Li W, Sun X, Yi B. **Farnesoid X receptor interacts with cAMP response element binding protein to modulate glucagon-like peptide-1 (7-36) amide secretion by intestinal L cell**. *J Cell Physiol* (2019) **234** 12839-12846. PMID: 30536761
32. Trabelsi MS, Daoudi M, Prawitt J, Ducastel S, Touche V, Sayin SI. **Farnesoid X receptor inhibits glucagon-like peptide-1 production by enteroendocrine L cells**. *Nat Commun* (2015) **6** 7629. PMID: 26134028
33. Desai MS, Mathur B, Eblimit Z, Vasquez H, Taegtmeyer H, Karpen SJ. **Bile acid excess induces cardiomyopathy and metabolic dysfunctions in the heart**. *Hepatology* (2017) **65** 189-201. PMID: 27774647
34. Ma Y, Liu D. **Activation of pregnane X receptor by pregnenolone 16 alpha-carbonitrile prevents high-fat diet-induced obesity in AKR/J mice**. *PLoS One* (2012) **7** e38734. PMID: 22723881
35. Polly P, Tan TC. **The role of vitamin D in skeletal and cardiac muscle function**. *Front Physiol* (2014) **5** 145. PMID: 24782788
36. Aljack HA, Abdalla MK, Idris OF, Ismail AM. **Vitamin D deficiency increases risk of nephropathy and cardiovascular diseases in Type 2 diabetes mellitus patients**. *J Res Med Sci* (2019) **24** 47. PMID: 31160914
37. Mozos I, Marginean O. **Links between Vitamin D Deficiency and Cardiovascular Diseases**. *Biomed Res Int* (2015) **2015** 109275. PMID: 26000280
38. Mozos I, Stoian D, Luca CT. **Crosstalk between Vitamins A, B12, D, K, C, and E Status and Arterial Stiffness**. *Disease Markers* (2017) **2017** 8784971. PMID: 28167849
39. Michiyasu I, Daisuke A, Makoto M. **Lithocholic Acid Is a Vitamin D Receptor Ligand That Acts Preferentially in the Ileum**. *International Journal of Molecular Sciences* (2018) **19** 1975. PMID: 29986424
40. Zhang R, Ma W, Fu M, Li J, Hu C, Chen Y. **Overview of bile acid signaling in the cardiovascular system**. *World Journal of Clinical Cases* (2021) **9** 308-320. PMID: 33521099
41. Li T, Matozel M, Boehme S, Kong B, Nilsson LM, Guo G. **Overexpression of cholesterol 7alpha-hydroxylase promotes hepatic bile acid synthesis and secretion and maintains cholesterol homeostasis**. *Hepatology* (2011) **53** 996-1006. PMID: 21319191
42. Chiang J, Ferrell JM. **Up to date on cholesterol 7 alpha-hydroxylase (CYP7A1) in bile acid synthesis**. *Liver Res* (2020) **4** 47-63. PMID: 34290896
43. Duboc H, Tache Y, Hofmann AF. **The bile acid TGR5 membrane receptor: from basic research to clinical application**. *Dig Liver Dis* (2014) **46** 302-312. PMID: 24411485
44. Donepudi AC, Boehme S, Li F, Chiang JYL. **G protein-coupled bile acid receptor plays a key role in bile acid metabolism and fasting-induced hepatic steatosis**. *Hepatology* (2017)
45. Kumar DP, Asgharpour A, Mirshahi F, Park SH, Liu S, Imai Y. **Activation of Transmembrane Bile Acid Receptor TGR5 Modulates Pancreatic Islet alpha Cells to Promote Glucose Homeostasis**. *J Biol Chem* (2016) **291** 6626-6640. PMID: 26757816
46. Broeders E, Nascimento E, Havekes B, Brans B, Schrauwen P. **The Bile Acid Chenodeoxycholic Acid Increases Human Brown Adipose Tissue Activity**. *Cell Metabolism* (2015) **22** 418-426. PMID: 26235421
47. Qi Y, Jiang C, Cheng J, Krausz KW, Li T, Ferrell JM. **Bile acid signaling in lipid metabolism: metabolomic and lipidomic analysis of lipid and bile acid markers linked to anti-obesity and anti-diabetes in mice**. *Biochim Biophys Acta* (2015) **1851** 19-29. PMID: 24796972
48. Carino A, Marchiano S, Biagioli M, Bucci M, Vellecco V, Brancaleone V. **Agonism for the bile acid receptor GPBAR1 reverses liver and vascular damage in a mouse model of steatohepatitis**. *Faseb J* (2019) **33** 2809-2822. PMID: 30303744
49. Guo C, Xie S, Chi Z, Zhang J, Liu Y, Zhang L. **Bile Acids Control Inflammation and Metabolic Disorder through Inhibition of NLRP3 Inflammasome**. *Immunity* (2016) **45** 802-816. PMID: 27692610
50. Biagioli M, Carino A, Cipriani S, Francisci D, Marchiano S, Scarpelli P. **The Bile Acid Receptor GPBAR1 Regulates the M1/M2 Phenotype of Intestinal Macrophages and Activation of GPBAR1 Rescues Mice from Murine Colitis**. *J Immunol* (2017) **199** 718-733. PMID: 28607110
51. Sorrentino G, Perino A, Yildiz E, El AG, Bou SM, Gioiello A. **Bile Acids Signal via TGR5 to Activate Intestinal Stem Cells and Epithelial Regeneration**. *Gastroenterology* (2020) **159** 956-968. PMID: 32485177
52. Merlen G, Bidault-Jourdainne V, Kahale N, Glenisson M, Ursic-Bedoya J, Doignon I. **Hepatoprotective impact of the bile acid receptor TGR5**. *Liver Int* (2020) **40** 1005-1015. PMID: 32145703
53. Miyazaki-Anzai S, Masuda M, Kohno S, Levi M, Shiozaki Y, Keenan AL. **Simultaneous inhibition of FXR and TGR5 exacerbates atherosclerotic formation**. *J Lipid Res* (2018) **59** 1709-1713. PMID: 29976576
54. Chang S, Kim YH, Kim YJ, Kim YW, Moon S, Lee YY. **Taurodeoxycholate Increases the Number of Myeloid-Derived Suppressor Cells That Ameliorate Sepsis in Mice**. *Front Immunol* (2018) **9** 1984. PMID: 30279688
55. Duan H, Ning M, Zou Q, Ye Y, Feng Y, Zhang L. **Discovery of Intestinal Targeted TGR5 Agonists for the Treatment of Type 2 Diabetes**. *J Med Chem* (2015) **58** 3315-3328. PMID: 25710631
56. Schledwitz A, Sundel MH, Alizadeh M, Hu S, Xie G, Raufman JP. **Differential Actions of Muscarinic Receptor Subtypes in Gastric, Pancreatic, and Colon Cancer**. *Int J Mol Sci* (2021) **22**
57. Ibrahim E, Diakonov I, Arunthavarajah D, Swift T, Goodwin M, McIlvride S. **Bile acids and their respective conjugates elicit different responses in neonatal cardiomyocytes: role of Gi protein, muscarinic receptors and TGR5**. *Scientific Reports* (2018) **8**
58. Kurano M, Yatomi Y. **Sphingosine 1-Phosphate and Atherosclerosis**. *J Atheroscler Thromb* (2018) **25** 16-26. PMID: 28724841
59. Runping Liu, Xiaojiaoyang Li, Xiaoyan Qiang. **Taurocholate Induces Cyclooxygenase-2 Expression via the Sphingosine 1-phosphate Receptor 2 in a Human Cholangiocarcinoma Cell Line**. *The Journal of Biological Chemistry* (2015)
60. Ruiz M, Frej C, Holmer A, Guo LJ, Tran S, Dahlback B. **High-Density Lipoprotein-Associated Apolipoprotein M Limits Endothelial Inflammation by Delivering Sphingosine-1-Phosphate to the Sphingosine-1-Phosphate Receptor 1**. *Arterioscler Thromb Vasc Biol* (2017) **37** 118-129. PMID: 27879252
61. Bukiya AN, McMillan J, Liu J, Shivakumar B, Parrill AL, Dopico AM. **Activation of calcium- and voltage-gated potassium channels of large conductance by leukotriene B4**. *J Biol Chem* (2014) **289** 35314-35325. PMID: 25371198
62. Binah O, Rubinstein I, Bomzon A, Better OS. **Effects of bile acids on ventricular muscle contraction and electrophysiological properties: studies in rat papillary muscle and isolated ventricular myocytes**. *Naunyn-Schmiedeberg's Archives of Pharmacology* (1987) **335** 160-165. PMID: 3561530
63. Bulluck H, Rosmini S, Abdel-Gadir A, White SK, Bhuva AN, Treibel TA. **Residual Myocardial Iron Following Intramyocardial Hemorrhage During the Convalescent Phase of Reperfused ST-Segment-Elevation Myocardial Infarction and Adverse Left Ventricular Remodeling**. *Circ Cardiovasc Imaging* (2016) **9**
64. Pu J, Yuan A, Shan P, Gao E, Wang X, Wang Y. **Cardiomyocyte-expressed farnesoid-X-receptor is a novel apoptosis mediator and contributes to myocardial ischaemia/reperfusion injury**. *Eur Heart J* (2013) **34** 1834-1845. PMID: 22307460
65. Gao Y, Zhao Y, Yuan A, Xu L, Huang X, Su Y. **Effects of farnesoid-X-receptor SUMOylation mutation on myocardial ischemia/reperfusion injury in mice**. *Exp Cell Res* (2018) **371** 301-310. PMID: 30098335
66. Xiaoli L, Zhen Z, Jiuchang Z, Hongjiang W, Xinchun Y. **Activation of FXR receptor reduces damage of ET-1 on H9C2 cardiomyocytes by activating AMPK signaling pathway**. *Panminerva Med* (2020)
67. Qiang S, Tao L, Zhou J, Wang Q, Wang K, Lu M. **Knockout of farnesoid X receptor aggravates process of diabetic cardiomyopathy**. *Diabetes Res Clin Pract* (2020) **161** 108033. PMID: 32006644
68. Mencarelli A, Cipriani S, Renga B, D'Amore C, Palladino G, Distrutti E. **FXR activation improves myocardial fatty acid metabolism in a rodent model of obesity-driven cardiotoxicity**. *Nutr Metab Cardiovasc Dis* (2013) **23** 94-101. PMID: 21924881
69. Han SY, Song HK, Cha JJ, Han JY, Kang YS, Cha DR. **Farnesoid X receptor (FXR) agonist ameliorates systemic insulin resistance, dysregulation of lipid metabolism, and alterations of various organs in a type 2 diabetic kidney animal model**. *Acta Diabetol* (2021) **58** 495-503. PMID: 33399988
70. Gao J, Yuan G, Xu Z, Lan L, Xin W. **Chenodeoxycholic and deoxycholic acids induced positive inotropic and negative chronotropic effects on rat heart**. *Naunyn-Schmiedeberg's Archives of Pharmacology* (2021) **394** 765-773. PMID: 32808070
71. Eblimit Z, Thevananther S, Karpen SJ, Taegtmeyer H, Moore DD, Adorini L. **TGR5 activation induces cytoprotective changes in the heart and improves myocardial adaptability to physiologic, inotropic, and pressure-induced stress in mice**. *Cardiovascular Therapeutics* (2018) **36** e12462. PMID: 30070769
72. Chen S, Law CS, Grigsby CL, Olsen K, Hong TT, Zhang Y. **Cardiomyocyte-specific deletion of the vitamin D receptor gene results in cardiac hypertrophy**. *Circulation* (2011) **124** 1838-1847. PMID: 21947295
73. Rodriguez AJ, Mousa A, Ebeling PR, Scott D, de Courten B. **Effects of vitamin D supplementation on inflammatory markers in heart failure: a systematic review and meta-analysis of randomized controlled trials**. *Sci Rep* (2018) **8** 1169. PMID: 29348609
74. Kida T, Tsubosaka Y, Hori M, Ozaki H, Murata T. **Bile acid receptor TGR5 agonism induces NO production and reduces monocyte adhesion in vascular endothelial cells**. *Arteriosclerosis Thrombosis & Vascular Biology* (2013) **33** 1663-1669. PMID: 23619297
75. Mohamed AS, Hanafi NI, Hamimah S, Noor JM, Narimah A, Rahim SA. **Ursodeoxycholic acid protects cardiomyocytes against cobalt chloride induced hypoxia by regulating transcriptional mediator of cells stress hypoxia inducible factor 1α and p53 protein**. *Cell Biochemistry & Function* (2017)
76. Song Z, Tian X, Shi Q. **Fas, Caspase-8, and Caspase-9 pathway-mediated bile acid-induced fetal cardiomyocyte apoptosis in intrahepatic cholestasis pregnant rat models**. *J Obstet Gynaecol Res* (2021) **47** 2298-2306. PMID: 33847039
77. Gao J, Yuan G, Xu Z, Lan L, Xin W. **Chenodeoxycholic and deoxycholic acids induced positive inotropic and negative chronotropic effects on rat heart**. *Naunyn Schmiedebergs Arch Pharmacol* (2021) **394** 765-773. PMID: 32808070
78. Fiorucci S, Zampella A, Cirino G, Bucci M, Distrutti E. **Decoding the vasoregulatory activities of bile acid-activated receptors in systemic and portal circulation: role of gaseous mediators**. *Am J Physiol Heart Circ Physiol* (2017) **312** H21-H32. PMID: 27765751
79. Guizoni DM, Vettorazzi JF, Carneiro EM, Davel AP. **Modulation of endothelium-derived nitric oxide production and activity by taurine and taurine-conjugated bile acids**. *Nitric Oxide* (2020) **94** 48-53. PMID: 31669041
80. Li J, Wilson A, Kuruba R, Zhang Q, Gao X, He F. **FXR-mediated regulation of eNOS expression in vascular endothelial cells**. *Cardiovasc Res* (2008) **77** 169-177. PMID: 18006476
81. Kida T, Murata T, Hori M, Ozaki H. **Chronic stimulation of farnesoid X receptor impairs nitric oxide sensitivity of vascular smooth muscle**. *Am J Physiol Heart Circ Physiol* (2009) **296** H195-H201. PMID: 19011043
82. Moraes LA, Unsworth AJ, Vaiyapuri S, Ali MS, Sasikumar P, Sage T. **Farnesoid X Receptor and Its Ligands Inhibit the Function of Platelets**. *Arterioscler Thromb Vasc Biol* (2016) **36** 2324-2333. PMID: 27758768
83. Zhang R, Ran HH, Zhang YX, Liu P, Lu CY, Xu Q. **Farnesoid X receptor regulates vascular reactivity through nitric oxide mechanism**. *J Physiol Pharmacol* (2012) **63** 367-372. PMID: 23070085
84. Ting TC, Miyazaki-Anzai S, Masuda M, Levi M, Demer LL, Tintut Y. **Increased lipogenesis and stearate accelerate vascular calcification in calcifying vascular cells**. *J Biol Chem* (2011) **286** 23938-23949. PMID: 21596756
85. Nadro B, Juhasz L, Szentpeteri A, Pall D, Paragh G, Harangi M. **The role of apolipoprotein M and sphingosine 1-phosphate axis in the prevention of atherosclerosis]**. *Orv Hetil* (2018) **159** 168-175. PMID: 29376427
86. Machida T, Matamura R, Iizuka K, Hirafuji M. **Cellular function and signaling pathways of vascular smooth muscle cells modulated by sphingosine 1-phosphate**. *J Pharmacol Sci* (2016) **132** 211-217. PMID: 27581589
87. Abbate A, Toldo S, Marchetti C, Kron J, Van Tassell BW, Dinarello CA. **Interleukin-1 and the Inflammasome as Therapeutic Targets in Cardiovascular Disease**. *Circ Res* (2020) **126** 1260-1280. PMID: 32324502
88. Wang J, Zhang J, Lin X, Wang Y, Wu X, Yang F. **DCA-TGR5 signaling activation alleviates inflammatory response and improves cardiac function in myocardial infarction**. *Journal of Molecular and Cellular Cardiology* (2021) **151** 3-14. PMID: 33130149
89. Chiang JYL, Ferrell JM. **Bile Acids as Metabolic Regulators and Nutrient Sensors**. *Annual Review of Nutrition* (2019) **39** 175-200
90. Bal NB, Han S, Kiremitci S, Sadi G, Uludag O, Demirel Yilmaz E. **Hypertension-induced cardiac impairment is reversed by the inhibition of endoplasmic reticulum stress**. *Journal of Pharmacy and Pharmacology* (2019)
91. Rainer PP, Primessnig U, Harenkamp S, Doleschal B, Wallner M, Fauler G. **Bile acids induce arrhythmias in human atrial myocardium--implications for altered serum bile acid composition in patients with atrial fibrillation**. *Heart* (2013) **99** 1685-1692. PMID: 23894089
92. von Haehling S, Schefold JC, Jankowska EA, Springer J, Vazir A, Kalra PR. **Ursodeoxycholic acid in patients with chronic heart failure: a double-blind, randomized, placebo-controlled, crossover trial**. *J Am Coll Cardiol* (2012) **59** 585-592. PMID: 22300693
93. Vítek L. **Bile Acids in the Treatment of Cardiometabolic Diseases**. *Annals of Hepatology* (2017) **16** S43-S52
94. Prabhu SD, Frangogiannis NG. **The Biological Basis for Cardiac Repair After Myocardial Infarction: From Inflammation to Fibrosis**. *Circulation Research* (2016) **119** 91-112. PMID: 27340270
95. Ryan PM, Stanton C, Caplice NM. **Bile acids at the cross-roads of gut microbiome-host cardiometabolic interactions**. *Diabetology & Metabolic Syndrome* (2017) **9** 102. PMID: 29299069
96. Xia Y, Zhang F, Zhao S, Li Y, Chen X, Gao E. **Adiponectin determines farnesoid X receptor agonism-mediated cardioprotection against post-infarction remodelling and dysfunction**. *Cardiovasc Res* (2018) **114** 1335-1349. PMID: 29668847
97. Gao J, Liu X, Wang B, Xu H, Xia Q, Lu T. **Farnesoid X receptor deletion improves cardiac function, structure and remodeling following myocardial infarction in mice**. *Mol Med Rep* (2017) **16** 673-679. PMID: 28560412
98. Irak K, Bayram M, Cifci S, Acar Z, Kazezoglu C, Ogutmen KD. **Evaluation of G-Protein-Coupled Bile Acid Receptor 1 (TGR5) Levels in Intrahepatic Cholestasis of Pregnancy**. *Cureus* (2021) **13** e19654. PMID: 34976450
99. Alemi F, Kwon E, Poole DP, Lieu T, Lyo V, Cattaruzza F. **The TGR5 receptor mediates bile acid-induced itch and analgesia**. *J Clin Invest* (2013) **123** 1513-1530. PMID: 23524965
100. Wang J, Xu T, Xu M. **Roles and Mechanisms of TGR5 in the Modulation of CD4(+) T Cell Functions in Myocardial Infarction**. *J Cardiovasc Transl Res* (2022) **15** 350-359. PMID: 34402028
101. Wu H, Liu G, He Y, Da J, Xie B. **Obeticholic acid protects against diabetic cardiomyopathy by activation of FXR/Nrf2 signaling in db/db mice**. *European Journal of Pharmacology* (2019) **858** 172393. PMID: 31085240
102. Ma Y, Liu D. **Activation of Pregnane X Receptor by Pregnenolone 16 α-carbonitrile Prevents High-Fat Diet-Induced Obesity in AKR/J Mice**. *Plos One* (2012) **7** e38734. PMID: 22723881
103. Lu J, Wang S, Li M, Gao Z, Xu Y, Zhao X. **Association of Serum Bile Acids Profile and Pathway Dysregulation With the Risk of Developing Diabetes Among Normoglycemic Chinese Adults: Findings From the 4C Study**. *Diabetes Care* (2021) **44** 499-510. PMID: 33355246
104. Drucker DJ. **The Cardiovascular Biology of Glucagon-like Peptide-1**. *Cell Metab* (2016) **24** 15-30. PMID: 27345422
105. Kristensen SL, Rorth R, Jhund PS, Docherty KF, Sattar N, Preiss D. **Cardiovascular, mortality, and kidney outcomes with GLP-1 receptor agonists in patients with type 2 diabetes: a systematic review and meta-analysis of cardiovascular outcome trials**. *Lancet Diabetes Endocrinol* (2019) **7** 776-785. PMID: 31422062
106. Hansen M, Scheltema MJ, Sonne DP, Hansen JS, Sperling M, Rehfeld JF. **Effect of chenodeoxycholic acid and the bile acid sequestrant colesevelam on glucagon-like peptide-1 secretion**. *Diabetes Obes Metab* (2016) **18** 571-580. PMID: 26888164
107. Zhang H, Wang X, Wu Z, Liu H, Chen W, Zhang Z. **Beneficial effect of farnesoid X receptor activation on metabolism in a diabetic rat model**. *Molecular Medicine Reports* (2016)
108. Mcgavigan AK, Garibay D, Henseler ZM, Chen J, Bettaieb A, Haj FG. **TGR5 contributes to glucoregulatory improvements after vertical sleeve gastrectomy in mice**. *Gut* (2017)
109. Basso N, Soricelli E, Castagneto-Gissey L, Casella G, Albanese D, Fava F. **Insulin Resistance, Microbiota, and Fat Distribution Changes by a New Model of Vertical Sleeve Gastrectomy in Obese Rats**. *Diabetes* (2016) **65** 2990-3001. PMID: 27431457
110. Singh S, Khera R, Allen AM, Murad MH, Loomba R. **Comparative effectiveness of pharmacological interventions for nonalcoholic steatohepatitis: A systematic review and network meta-analysis**. *Hepatology* (2015) **62** 1417-1432. PMID: 26189925
111. Wang XX, Wang D, Luo Y, Myakala K, Dobrinskikh E, Rosenberg AZ. **FXR/TGR5 Dual Agonist Prevents Progression of Nephropathy in Diabetes and Obesity**. *J Am Soc Nephrol* (2018) **29** 118-137. PMID: 29089371
112. Napolitano A, Miller S, Nicholls AW, Baker D, Van Horn S, Thomas E. **Novel gut-based pharmacology of metformin in patients with type 2 diabetes mellitus**. *PLoS One* (2014) **9** e100778. PMID: 24988476
113. Vasavan T, Ferraro E, Ibrahim E, Dixon P, Gorelik J, Williamson C. **Heart and bile acids - Clinical consequences of altered bile acid metabolism**. *Biochim Biophys Acta* (2018) 1345-1355
114. Schmid A, Schlegel J, Thomalla M, Karrasch T, Schffler A. **Evidence of functional bile acid signaling pathways in adipocytes**. *Molecular and Cellular Endocrinology* (2019) **483** 1-10. PMID: 30543876
115. Byun S, Jung H, Chen J, Kim YC, Kim DH, Kong B. **Phosphorylation of hepatic farnesoid X receptor by FGF19 signaling-activated Src maintains cholesterol levels and protects from atherosclerosis**. *J Biol Chem* (2019) **294** 8732-8744. PMID: 30996006
116. Xu Y, Li F, Zalzala M, Xu J, Gonzalez FJ, Adorini L. **Farnesoid X receptor activation increases reverse cholesterol transport by modulating bile acid composition and cholesterol absorption in mice**. *Hepatology* (2016) **64** 1072-1085. PMID: 27359351
117. Wu Q, Sun L, Hu X, Wang X, Xu F, Chen B. **Suppressing the intestinal farnesoid X receptor/sphingomyelin phosphodiesterase 3 axis decreases atherosclerosis**. *J Clin Invest* (2021) **131**
118. Ghosh S, Dass J. **Study of pathway cross-talk interactions with NF-kappaB leading to its activation via ubiquitination or phosphorylation: A brief review**. *Gene* (2016) **584** 97-109. PMID: 26968890
119. Fu Y, Feng H, Ding X, Meng QH, Zhang SR, Li J. **Alisol B 23-acetate adjusts bile acid metabolisim via hepatic FXR-BSEP signaling activation to alleviate atherosclerosis**. *Phytomedicine* (2022) **101** 154120. PMID: 35523117
120. Huang K, Liu C, Peng M, Su Q, Liu R, Guo Z. **Glycoursodeoxycholic Acid Ameliorates Atherosclerosis and Alters Gut Microbiota in Apolipoprotein E-Deficient Mice**. *J Am Heart Assoc* (2021) **10** e19820
121. Kida T, Tsubosaka Y, Hori M, Ozaki H, Murata T. **Bile acid receptor TGR5 agonism induces NO production and reduces monocyte adhesion in vascular endothelial cells**. *Arteriosclerosis Thrombosis & Vascular Biology* (2013) **33** 1663-1669. PMID: 23619297
122. Chung J, An SH, Kang SW, Kwon K. **Ursodeoxycholic Acid (UDCA) Exerts Anti-Atherogenic Effects by Inhibiting RAGE Signaling in Diabetic Atherosclerosis**. *PLoS One* (2016) **11** e147839
123. Chung J, Kim KH, Lee SC, An SH, Kwon K. **Ursodeoxycholic Acid (UDCA) Exerts Anti-Atherogenic Effects by Inhibiting Endoplasmic Reticulum (ER) Stress Induced by Disturbed Flow**. *Mol Cells* (2015) **38** 851-858. PMID: 26442866
124. Hanafi NI, Mohamed AS, Md NJ, Abdu N, Hasani H, Siran R. **Ursodeoxycholic acid upregulates ERK and Akt in the protection of cardiomyocytes against CoCl2**. *Genet Mol Res* (2016) 15
125. Yan JJ, Fan HQ, Yang L. **Bile acids in arrhythmia]**. *Zhonghua Gan Zang Bing Za Zhi* (2020) **28** 361-364. PMID: 32403891
126. Borges ML, Papacleovoulou G, Flaviani F, Pataia V, Qadri A, Abu-Hayyeh S. **Ursodeoxycholic acid improves feto-placental and offspring metabolic outcomes in hypercholanemic pregnancy**. *Sci Rep* (2020) **10** 10361. PMID: 32587408
127. Kirbas O, Biberoglu EH, Kirbas A, Daglar K, Kurmus O, Danisman N. **Evaluation of ventricular repolarization in pregnant women with intrahepatic cholestasis**. *Int J Cardiol* (2015) **189** 25-29. PMID: 25885869
128. Geenes V, Lovgren-Sandblom A, Benthin L, Lawrance D, Chambers J, Gurung V. **The reversed feto-maternal bile acid gradient in intrahepatic cholestasis of pregnancy is corrected by ursodeoxycholic acid**. *PLoS One* (2014) **9** e83828. PMID: 24421907
129. Sheikh AKS, Miragoli M, Abu-Hayyeh S, Moshkov AV, Xie Q, Keitel V. **Bile acid-induced arrhythmia is mediated by muscarinic M2 receptors in neonatal rat cardiomyocytes**. *PLoS One* (2010) **5** e9689. PMID: 20300620
130. Murakami M, Une N, Nishizawa M, Suzuki S, Ito H, Horiuchi T. **Incretin secretion stimulated by ursodeoxycholic acid in healthy subjects**. *Springerplus* (2013) **2** 20. PMID: 23450079
131. Miragoli M, Kadir SH, Sheppard MN, Salvarani N, Virta M, Wells S. **A protective antiarrhythmic role of ursodeoxycholic acid in an in vitro rat model of the cholestatic fetal heart**. *Hepatology* (2011) **54** 1282-1292. PMID: 21809354
132. Gorelik J, Shevchuk AI, Diakonov I, de Swiet M, Lab M, Korchev Y. **Dexamethasone and ursodeoxycholic acid protect against the arrhythmogenic effect of taurocholate in an in vitro study of rat cardiomyocytes**. *Bjog* (2003) **110** 467-474. PMID: 12742331
133. Schultz F, Hasan A, Alvarez-Laviada A, Miragoli M, Bhogal N, Wells S. **The protective effect of ursodeoxycholic acid in an in vitro model of the human fetal heart occurs via targeting cardiac fibroblasts**. *Prog Biophys Mol Biol* (2016) **120** 149-163. PMID: 26777584
134. Ferraro E, Pozhidaeva L, Pitcher DS, Mansfield C, Koh J, Williamson C. **Prolonged ursodeoxycholic acid administration reduces acute ischaemia-induced arrhythmias in adult rat hearts**. *Sci Rep* (2020) **10** 15284. PMID: 32943714
135. Adeyemi O, Alvarez-Laviada A, Schultz F, Ibrahim E, Trauner M, Williamson C. **Ursodeoxycholic acid prevents ventricular conduction slowing and arrhythmia by restoring T-type calcium current in fetuses during cholestasis**. *PLoS One* (2017) **12** e183167
136. Mayerhofer C, Ueland T, Broch K, Vincent RP, Cross GF, Dahl CP. **Increased Secondary/Primary Bile Acid Ratio in Chronic Heart Failure**. *J Card Fail* (2017) **23** 666-671. PMID: 28688889
137. Goossens JF, Bailly C. **Ursodeoxycholic acid and cancer: From chemoprevention to chemotherapy**. *Pharmacol Ther* (2019) **203** 107396. PMID: 31356908
138. Woolbright BL. **Inflammation: Cause or consequence of chronic cholestatic liver injury**. *Food Chem Toxicol* (2020) **137** 111133. PMID: 31972189
139. Zardi EM, Abbate A, Zardi DM, Dobrina A, Margiotta D, Van Tassell BW. **Cirrhotic cardiomyopathy**. *J Am Coll Cardiol* (2010) **56** 539-549. PMID: 20688208
140. Kim SY, Kim KH, Schilling JM, Leem J, Dhanani M, Head BP. **Protective role of cardiac-specific overexpression of caveolin-3 in cirrhotic cardiomyopathy**. *Am J Physiol Gastrointest Liver Physiol* (2020) **318** G531-G541. PMID: 31961720
141. Desai MS, Mathur B, Eblimit Z, Vasquez H, Taegtmeyer H, Karpen SJ. **Bile acid excess induces cardiomyopathy and metabolic dysfunctions in the heart**. *Hepatology* (2017) **65** 189-201. PMID: 27774647
142. Frommherz L, Bub A, Hummel E, Rist MJ, Roth A, Watzl B. **Age-Related Changes of Plasma Bile Acid Concentrations in Healthy Adults--Results from the Cross-Sectional KarMeN Study**. *PLoS One* (2016) **11** e153959
143. Donepudi AC, Boehme S, Li F, Chiang JYL. **G protein-coupled bile acid receptor plays a key role in bile acid metabolism and fasting-induced hepatic steatosis**. *Hepatology* (2017)
144. Mozos I. **Arrhythmia risk in liver cirrhosis**. *World J Hepatol* (2015) **7** 662-672. PMID: 25866603
145. Zhang Y, Jiang R, Zheng X, Lei S, Huang F, Xie G. **Ursodeoxycholic acid accelerates bile acid enterohepatic circulation**. *Br J Pharmacol* (2019) **176** 2848-2863. PMID: 31077342
146. Liao Y, Liu C, Xiong T, Zhao M, Zheng W, Feng Y. **Metabolic Modulation and Potential Biomarkers of the Prognosis Identification for Severe Aortic Stenosis after TAVR by a Metabolomics Study**. *Cardiol Res Pract* (2020) **2020** 3946913. PMID: 33204525
147. Li W, Shu S, Cheng L, Hao X, Wang L, Wu Y. **Fasting serum total bile acid level is associated with coronary artery disease, myocardial infarction and severity of coronary lesions**. *Atherosclerosis* (2020) **292** 193-200. PMID: 31811964
148. Ferrell JM, Chiang J. **Understanding Bile Acid Signaling in Diabetes: From Pathophysiology to Therapeutic Targets**. *Diabetes Metab J* (2019) **43** 257-272. PMID: 31210034
149. Smushkin G, Sathananthan M, Piccinini F, Dalla MC, Law JH, Cobelli C. **The effect of a bile acid sequestrant on glucose metabolism in subjects with type 2 diabetes**. *Diabetes* (2013) **62** 1094-1101. PMID: 23250357
150. Hansen M, Sonne DP, Mikkelsen KH, Gluud LL, Vilsboll T, Knop FK. **Bile acid sequestrants for glycemic control in patients with type 2 diabetes: A systematic review with meta-analysis of randomized controlled trials**. *J Diabetes Complications* (2017) **31** 918-927. PMID: 28238556
151. Nwose OM, Jones MR. **Atypical mechanism of glucose modulation by colesevelam in patients with type 2 diabetes**. *Clin Med Insights Endocrinol Diabetes* (2013) **6** 75-79. PMID: 24348081
152. Sedgeman LR, Beysen C, Allen RM, Ramirez SM, Turner SM, Vickers KC. **Intestinal bile acid sequestration improves glucose control by stimulating hepatic miR-182-5p in type 2 diabetes**. *Am J Physiol Gastrointest Liver Physiol* (2018) **315** G810-G823. PMID: 30160993
153. Chevli PA, Freedman BI, Hsu FC, Xu J, Rudock ME, Ma L. **Plasma metabolomic profiling in subclinical atherosclerosis: the Diabetes Heart Study**. *Cardiovasc Diabetol* (2021) **20** 231. PMID: 34876126
154. Feng X, Zhai G, Yang J, Liu Y, Zhou Y, Guo Q. **Myocardial Infarction and Coronary Artery Disease in Menopausal Women With Type 2 Diabetes Mellitus Negatively Correlate With Total Serum Bile Acids**. *Front Endocrinol (Lausanne)* (2021) **12** 754006. PMID: 34675887
155. Baars A, Oosting A, Lohuis M, Koehorst M, El AS, Hugenholtz F. **Sex differences in lipid metabolism are affected by presence of the gut microbiota**. *Sci Rep* (2018) **8** 13426. PMID: 30194317
156. Ishimwe JA, Dola T, Ertuglu LA, Kirabo A. **Bile acids and salt-sensitive hypertension: a role of the gut-liver axis**. *Am J Physiol Heart Circ Physiol* (2022) **322** H636-H646. PMID: 35245132
157. Ataalla WM, Ziada DH, Gaber R, Ossman A, Bayomy S, Elemary BR. **The impact of total bile acid levels on fetal cardiac function in intrahepatic cholestasis of pregnancy using fetal echocardiography: a tissue Doppler imaging study**. *J Matern Fetal Neonatal Med* (2016) **29** 1445-1450. PMID: 26067266
158. Vasavan T, Deepak S, Jayawardane IA, Lucchini M, Martin C, Geenes V. **Fetal cardiac dysfunction in intrahepatic cholestasis of pregnancy is associated with elevated serum bile acid concentrations**. *J Hepatol* (2021) **74** 1087-1096. PMID: 33276032
159. Mayerhofer C, Ueland T, Broch K, Vincent RP, Cross GF, Dahl CP. **Increased Secondary/Primary Bile Acid Ratio in Chronic Heart Failure**. *J Card Fail* (2017) **23** 666-671. PMID: 28688889
160. Voiosu AM, Wiese S, Voiosu TA, Hove J, Bendtsen F, Moller S. **Total bile acid levels are associated with left atrial volume and cardiac output in patients with cirrhosis**. *Eur J Gastroenterol Hepatol* (2018) **30** 392-397. PMID: 29227330
161. Simbrunner B, Trauner M, Reiberger T. **Review article: therapeutic aspects of bile acid signalling in the gut-liver axis**. *Aliment Pharmacol Ther* (2021) **54** 1243-1262. PMID: 34555862
162. Witkowski M, Weeks TL, Hazen SL. **Gut Microbiota and Cardiovascular Disease**. *Circ Res* (2020) **127** 553-570. PMID: 32762536
163. Zhou W, Cheng Y, Zhu P, Nasser MI, Zhang X, Zhao M. **Implication of Gut Microbiota in Cardiovascular Diseases**. *Oxid Med Cell Longev* (2020) **2020** 5394096. PMID: 33062141
164. Karlsson FH, Tremaroli V, Nookaew I, Bergstrom G, Behre CJ, Fagerberg B. **Gut metagenome in European women with normal, impaired and diabetic glucose control**. *Nature* (2013) **498** 99-103. PMID: 23719380
165. Koopen AM, Groen AK, Nieuwdorp M. **Human microbiome as therapeutic intervention target to reduce cardiovascular disease risk**. *Curr Opin Lipidol* (2016) **27** 615-622. PMID: 27676197
166. Mamic P, Chaikijurajai T, Tang W. **Gut microbiome - A potential mediator of pathogenesis in heart failure and its comorbidities: State-of-the-art review**. *J Mol Cell Cardiol* (2021) **152** 105-117. PMID: 33307092
167. Zhu W, Wang Z, Tang W, Hazen SL. **Gut Microbe-Generated Trimethylamine N-Oxide From Dietary Choline Is Prothrombotic in Subjects**. *Circulation* (2017) **135** 1671-1673. PMID: 28438808
168. Zhu W, Gregory JC, Org E, Buffa JA, Gupta N, Wang Z. **Gut Microbial Metabolite TMAO Enhances Platelet Hyperreactivity and Thrombosis Risk**. *Cell* (2016) **165** 111-124. PMID: 26972052
169. Wilson A, McLean C, Kim RB. **Trimethylamine-N-oxide: a link between the gut microbiome, bile acid metabolism, and atherosclerosis**. *Curr Opin Lipidol* (2016) **27** 148-154. PMID: 26959704
170. Chen K, Zheng X, Feng M, Li D, Zhang H. **Gut Microbiota-Dependent Metabolite Trimethylamine N-Oxide Contributes to Cardiac Dysfunction in Western Diet-Induced Obese Mice**. *Front Physiol* (2017) **8** 139. PMID: 28377725
171. Roberts AB, Gu X, Buffa JA, Hurd AG, Wang Z, Zhu W. **Development of a gut microbe-targeted nonlethal therapeutic to inhibit thrombosis potential**. *Nat Med* (2018) **24** 1407-1417. PMID: 30082863
172. Pathak P, Helsley RN, Brown AL, Buffa JA, Choucair I, Nemet I. **Small molecule inhibition of gut microbial choline trimethylamine lyase activity alters host cholesterol and bile acid metabolism**. *Am J Physiol Heart Circ Physiol* (2020) **318** H1474-H1486. PMID: 32330092
173. Knauf F, Brewer JR, Flavell RA. **Immunity, microbiota and kidney disease**. *Nat Rev Nephrol* (2019) **15** 263-274. PMID: 30796361
174. Schugar RC, Gliniak CM, Osborn LJ, Massey W, Sangwan N, Horak A. **Gut microbe-targeted choline trimethylamine lyase inhibition improves obesity via rewiring of host circadian rhythms**. *Elife* (2022) **11**
175. Bariya D, Anand V, Mishra S. **Recent advances in the bile acid based conjugates/derivatives towards their gelation applications**. *Steroids* (2021) **165** 108769. PMID: 33207227
176. Kovacevic B, Jones M, Ionescu C, Walker D, Wagle S, Chester J. **The emerging role of bile acids as critical components in nanotechnology and bioengineering: Pharmacology, formulation optimizers and hydrogel-biomaterial applications**. *Biomaterials* (2022) **283** 121459. PMID: 35303546
177. Faustino C, Serafim C, Rijo P, Reis CP. **Bile acids and bile acid derivatives: use in drug delivery systems and as therapeutic agents**. *Expert Opin Drug Deliv* (2016) **13** 1133-1148. PMID: 27102882
178. Kecman S, Skrbic R, Badnjevic CA, Mooranian A, Al-Salami H, Mikov M. **Potentials of human bile acids and their salts in pharmaceutical nano delivery and formulations adjuvants**. *Technol Health Care* (2020) **28** 325-335. PMID: 31594273
|
---
title: Crude extract of Ficus deltoidea Jack (FD) as a natural biological therapy
authors:
- Mahmoud Dogara Abdulrahman
journal: Exploration of Targeted Anti-tumor Therapy
year: 2023
pmcid: PMC10017191
doi: 10.37349/etat.2023.00123
license: CC BY 4.0
---
# Crude extract of Ficus deltoidea Jack (FD) as a natural biological therapy
## Abstract
### Aim:
This study shows how important it is to coordinate research on *Ficus deltoidea* Jack (FD) so that results from different sources can be compared directly and a scientific conclusion can be made.
### Methods:
The author looked for research papers on Ficus (F.) deltoidea on Google Scholar, Science Direct, Google.com, Wiley, PubMed, Hindawi, Springer, and other related databases. This analysis excludes data that cannot be trusted, thesis papers, and review articles about F. deltoidea.
### Results:
In traditional medicine, the plant’s leaves and syconia are used to cure a wide variety of ailments, including itchiness, diarrhoea, cancer, sexual dysfunction, age-related issues, malaria, cancer, anxiety, pain, constipation, fever, diabetes, tooth pain, and tooth decay. In vitro and in vivo studies showed the effectiveness of the leaves against cancer cell lines.
### Conclusions:
Based on the existing research on the health benefits of FD, it is critical to focus on its more active constituents and their identification, determination, further development, and, most importantly, standardization of the leaves for the management and treatment of cancer and its related cases. More research is needed before it can be considered a promising herbal source of novel medication candidates for treating various disorders.
## Introduction
Nature provides numerous plants that serve as the primary source of traditional medicines that can treat a wide range of illnesses [1]. Humans have used medicinal plants for thousands of years as a source of antimicrobial, antifungal, and anticancer agents, and for many other uses [2]. People have been very interested in biological products for a long time. Discovering new compounds with potential future applications is one of the main motivations for researching these priceless by-products [3]. Plant-based remedies for health issues have been on the rise recently. The need for new drugs derived from numerous species of medicinal plants is continually growing today [1]. Investigating potent natural compounds from plants with high biological activity is still ongoing. Ficus deltoidea Jack (FD) is one of the most well-known and widely appreciated plants. Many studies have been published on the plant’s biological properties. The current literature [4, 5] on its potential for managing and treating diseases, especially cancer, needs to be reviewed, analysed, and brought up to date. This study combines the scattered data on the biological impacts of Ficus (F.) deltoidea and synthesizes the data into a cohesive whole, paving the way for a more thorough understanding of the plant and for a clearer guidance on how to make the best use of its components.
## Materials and methods
Inclusion criteria: A comprehensive search of online resources like Web of Science, Taylor and Francis’ Science Direct, Google Scholar, Scopus, Springer Link, PubMed Central, SciELO, and Elsevier databases. Keywords such as Ficus deltoidea, *Ficus deltoidea* Jack, F. deltoidea in combination with antimicrobial, antioxidants, anticancer, anti-inflammation, anti-inflammatory, and other related and relevant phrases were used to search in the above databases. No time constraints were imposed, and all relevant databases were considered (Figure 1). Exclusion criteria: Only published research articles were considered; reviews, thesis abstracts, and unpublished papers were excluded (Figure 1).
**Figure 1.:** *Methodological flow diagram showing inclusion and exclusion criteria*
## Origin, distribution and taxonomic distribution of F. deltoidea
About 1,000 species from all over the tropics and subtropics make up the genus F. [6]. F. deltoidea is a species of shrub that is indigenous to Southeast Asia. It is known in Malay as the mistletoe fig or Mas cotek [7]. It is also called Sempit-sempit or agolaran by southern Malays [8]. In Central Africa, people call it Kangkaliban, but people call it Tabat Barito in Indonesia. F. deltoidea is indigenous to several Southeast Asian countries and may be seen growing widely throughout the region [9]. However, this plant can also be found in Africa [10]. This type of plant is usually found in Malaysia, Indonesia, and the southern Philippines, all of which are in Southeast Asia [8]. FD may be found in tropical and subtropical regions and comes in several different types [11]. FD is a natural shrub from the Moraceae family [8]. The form of the F. deltoidea leaves led to the separation of this species into two subspecies: F. deltoidea subsp. motleyana and F. deltoidea subsp. deltoidea [12]. There are two kinds of F. deltoidea plants: male and female. The difference between a male plant and a female plant is that the male plant has longer leaves, while the female plant has big, round, and long leaves [13]. The evergreen little tree or shrub can grow to 7–10 meters in its natural habitat [8]. The local people cultivate FD as a houseplant for aesthetic and medical advantages [14].
## Traditional uses
The plant is well-known among the Malay people and is utilised in treating diabetes, headaches, sore throats, and colds [7]. Traditional medicine uses various sections of the plant to cure various conditions [15]. In traditional medicine, hyperlipidemia, hypertension, and diabetes are all treated with F. deltoidea [7, 12]. This plant plays a significant role in traditional medicine, with its fruit used to treat a wide range of ailments, from headaches and toothaches to wounds (roots and leaves) [8, 15]. The consumption of fruit is a common method for alleviating pain associated with toothache and migraine, root and leaf remedies for cuts and scrapes [8]. After giving birth, women drank a decoction of the leaves to help tighten the uterine and vaginal muscles [8, 15]. It has been theorized that drinking a concoction made from the leaves can increase blood flow, have aphrodisiac effects, and even can fight diabetes [13]. Traditional uses for the extract include treating wounds, rheumatism, and ulcers; it is also effective as an antidiabetic medicine and a tonic for usage after giving birth [16].
## Antioxidants activity
An antioxidant defence system is in place to counteract the oxidative stress caused by the body’s normal physiological process of radical and reactive oxygen species (ROS) formation [17]. ROS are made when oxidative stress and the antioxidative defence system are out of balance. ROS can damage lipids, carbohydrates, proteins, and DNA, leading to many diseases [17]. Because of antioxidants’ ability to protect the body from harmful free radicals and ROS, many chronic diseases can be avoided and even reversed. Different parts of F. deltoidea were evaluated for antioxidant potential (Table 1). Antioxidant activities of F. deltoidea have been documented in several investigations, but the portion of the plant utilized in the vast majority of the studies was the leaf [18]. The removal of ROS by the hydrolysed protein fractions was superior to that by the unhydrolyzed protein fractions [19]. Based on a one-way analysis of variance, only the protein hydrolysates of 30 and 100 kDa indicated significant differences in radical scavenging capacities [19]. Methanolic leaf extract had the highest antioxidants for ferric reducing antioxidant power (FRAP) (6–9 mmol Fe2+/g), 2,2’-azino-bis (3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) (2–3 mmol TE/g), and 2,2-diphenyl-1-picrylhydrazyl (DPPH) (EC50:200–410 μg/mL) [12]. The methanol extract of F. deltoidea was the most effective at scavenging free radicals at 400 μg/mL ($85.41\%$). Lacklustre radical scavenging activity was observed in butanol extract [1]. The findings of this study revealed that solvent extracts play a critical role in demonstrating biological activities. It was discovered that antioxidant and total phenolic content (TPC) depend on the polarity of the solvent in the case of antioxidant activity [1]. Eighty-five per cent of the antioxidant activity of the FD extract was attributed to flavan-3-ol monomers and proanthocyanidins [20]. Based on these results, it is plausible that the leaves of F. deltoidea could be employed as a natural antioxidant. These enzymes’ activity and protein levels were elevated after exposure to F. deltoidea extract, suggesting that this compound may be responsible for reducing ROS production by acting on these enzymes. As a result, the extract directly scavenges ROS and indirectly stimulates the production of antioxidant enzymes (Figure 2). The main antioxidant defence system consists of antioxidant enzymes like superoxide dismutase (SOD), catalase (CAT), and glutathione peroxidase (GPx). Anti-ageing is partially achieved by SOD’s ability to scavenge the superoxide anion formed in the early stages of oxidative stress. With the help of SOD, superoxide can be converted into harmless hydrogen peroxide and dioxygen. To quickly catalyse the decomposition of hydrogen peroxide, cells frequently use CAT and GPx (Figure 2).
## Anti-inflammation activity
The process through which the body reacts to cellular damage is called inflammation [42]. It is a chain of events that can be set off by various stimuli, resulting in a predictable behavioural signature [42]. Results showed that FD aqueous extract (FDA) had significant anti-inflammatory effects in all assays at ($P \leq 0.05$) [11], and the paw oedema and formalin tests showed dose-response effects. In conclusion, the F. deltoidea leaf can reduce short-term and long-term inflammation and pain-related inflammation [11]. The findings, therefore, demonstrated the presence of pharmacologically active compounds with antinociceptive activity in the aqueous extract of F. deltoidea leaves [15]. Because of this, it is frequently applied in medicine to treat ailments that cause pain [15]. The fact that the FDA suppressed carrageenan-induced rat paw oedema for 5 h beginning 0.5 h after administration of the phlogistic drug implies that the extract’s mode of action entailed suppressing the cyclooxygenase (COX)-dependent response [42]. The lipopolysaccharides (LPS) stimulated microglial cells at a maximal dose (100 μg/mL), and the extract significantly decreased the production of ROS, nitric oxide (NO), tumour necrosis factor-α (TNF-α), interleukin-1 (IL-1), and IL-6 (Table 1). The extract of F. deltoidea considerably decreased the ultraviolet (UV)-induced production of TNF-α, IL-1, IL-6, and COX-2. The F. deltoidea extract may block proinflammatory cytokines, making it a potent remedy for skin conditions (Table 2). Numerous compounds that inhibit the immune response have been identified in plants. The first large class of plant chemicals, phenolic compounds, are crucial to many plant functions. Depending on environmental conditions, phenolic chemicals can accumulate in various plant tissues and cells during ontogenesis. It has been shown through research that, phenolic chemicals, many of which are found in the cell walls, vacuoles, and nuclei of cells, have anti-inflammatory and anti-septic characteristics (Table 2). Action mechanisms because of exposure to inflammatory substances, cells secrete arachidonic acid and inflammatory mediators like cytokines, serotonin, histamine, prostaglandins, leukotrienes, and vascular permeability and leukocyte recruitment are increased (Figure 3).
## Effects on microorganisms
By measuring the minimum inhibitory concentrations (MICs) and the diameter of the zone of inhibition, the antimicrobial activity against bacteria and fungi was tested (Table 3). The utility of F. deltoidea extracts against Gram-positive and Gram-negative bacteria are extensively known (Table 3). According to scientific research on F. deltoidea, these plants have garnered increasing interest in recent years. At a concentration of 31.26 mg/L, the plant extract did not stimulate the growth of the bacteria Edwardsiella tarda, Escherichia (E.) coli, Flavobacterium sp., Pseudomonas aeruginosa, or *Vibrio cholera* (Table 3). An extract from the plant prevented the growth of Aeromonas hydrophila, Klebsiella sp., Salmonella sp., and *Vibrio alginolyticus* when administered at a dosage of 62.5 mg/L. At a concentration of 125 mg/L, the plant extract inhibits the expansion of the pathogen *Vibrio parahaemolyticus* [10]. A 10–12 mm inhibition was found against the tested bacterial strain [50]. All bacteria tested were inhibited by the extract; however, Bacillus (B.) subtilis showed the greatest inhibition at 12 mm [51]. When tested against Staphylococcus (S.) aureus, the plant extract exhibited an inhibitory zone of 15.67 mm and a MIC of 3.125 mg/mL. The smallest reported sensitivity to chloroform extract was 6.33 mm, while the largest MIC was 25 mg/mL, both for B. subtilis [52]. Except for chloroform and aqueous extracts of B. subtilis, E. coli, and P. aeroginosa, all extracts demonstrated inhibitory effects on the fungi, Gram-positive and Gram-negative bacteria [5]. Results showed that the methanol extract was effective against the bacteria and the fungi used in the tests. The methanol extract showed the lowest MIC value (3.125 mg/mL) and the widest inhibition zone (15.67 mm) against the growth of S. aureus. B. subtilis had the highest MIC value (25 mg/mL) and the lowest sensitivity (6.33 mm) to the chloroform extract [5]. At 50 and 100 mg/mL, the MIC and minimum fungicidal concentration (MFC) for Candida albicans were both achieved with the studied extracts. The extract had a $69.5\%$ inhibitory effect on biofilm formation by Candida [53]. All test organisms showed that the extracts had a strong antimicrobial effect (Table 3). The qualitative and quantitative variability in the antifungal characteristics of the extracts is the root cause of the diversity in the inhibitory impact of plant extracts. The antimicrobial properties of these species may have come from the alkaloids, flavonoids, and cardiac glycosides found in these species’ leaves. Based on our findings, F. deltoidea extracts may be a viable alternative to antibiotics for managing drug-resistant bacterial and fungal strains. Apparently, the secondary metabolites in this plant are responsible for the extract’s extensive antibacterial activity. Various phytochemical substances have been reported to give F. deltoidea its medicinal benefits [1]. Species of the fig tree, F., are excellent resources for polyphenolic chemicals. The enhanced efficacy of the extracts is considered due to the synergistic effects of the various bioactive compounds present in F. deltoidea. In particular, the released chemicals attach easily to the negatively charged cell wall and break it, causing protein denaturation and cell death in microorganisms (Figure 4), ultimately leading to its rupture and denaturation of its proteins and the death of the cell.
## Effects on the endocrine system
About 1.9 billion persons globally are overweight, and about 600 million are clinically obese [65]. This makes obesity the largest public health problem in the world today [65]. Fat accumulation in the cytoplasm of adipocytes defines the increased adipose cell size characteristic of obesity [66]. Several enzymes, including fatty acid synthase, lipoprotein lipase, and adipocyte fatty acid-binding protein, control this metabolic shift in adipocytes [66]. People have used traditional medicinal plants and their active phyto constituents to treat obesity and the problems that come with it. There is much-untapped potential in natural products for treating obesity, and they could be a great substitute for developing safe and effective anti-obesity medications. After therapy with F. deltoidea at 500 and 1,000 mg/kg/day, insulin resistance, obesity index, TC, triglycerides, low-density lipoprotein (LDL) cholesterol, MDA, testosterone, and follicle-stimulating hormone (FSH) were decreased to nearly normal levels in polycystic ovary syndrome (PCOS) rats (Table 4). The capacity of F. deltoidea leaf extracts to prevent the development of mature adipocytes suggests that they may have anti-obesity capabilities [67]. The findings showed that F. deltoidea is a viable medicinal plant for creating novel functional foods, herbal medicines, and contemporary drugs with enormous potential for treating obesity. It has been demonstrated through scientific research that F. deltoidea can lower hyperglycaemia in a variety of prandial situations (Table 4). Different studies have claimed that F. deltoidea has antidiabetic and antioxidant properties, but most of these investigations have only employed the leaf. Researchers have found a link between the phenolic content of plants and their ability to combat diabetes [18]. Increased protein content and lower glucosidase activity in treated F. deltoidea samples provide compelling evidence for the critical role of proteins in demonstrating the beneficial antidiabetic effect [18]. Some research has suggested that F. deltoidea antihyperglycemic effects are mediated by the plant’s ability to increase insulin secretion from pancreatic cells, boost adipocyte glucose uptake, and boost adiponectin release from adipocytes [68]. Particularly, flavonoids and isoflavonoids are because of the high levels of antioxidant activity found in the extract, which benefits in awarding against illnesses caused by oxidative damage [25]. The findings suggested that the extracts may be a viable antibiotic option for regulating the growth of various bacterial strains. Pure substances or crude extracts may work through the production of inflammatory cytokines, leading to the death of microphages in the circulation and the release of secretory insulin. On the other hand, stimulating dendritic cells in the brain, where *Hypoglycemia is* present, will indicate insulin release. By inhibiting the connection between the insulin receptor in the cells and insulin release, the presence of fat that covers the insulin receptor likely contributes to insulin resistance and diabetes mellitus. The condition depicted in Figure 5 highlights complications associated with diabetes mellitus: The neurological system is harmed by nephropathy, retinopathy, and neuropathy (damage to nephrons).
## Antihypertensive activity
There is a strong association between hypertension and the development of cardiovascular disease [92]. Over 1 billion people throughout the world are afflicted with it. Hypertension is more common in the elderly than in the young [92]. People over 60 have a prevalence of $65.5\%$ of hypertension. Insulin resistance, obesity, glucose intolerance, concomitance, haemagglutinin, excessive uric acidemia, atherosclerosis, and cardiovascular illnesses are just a few of the chronic diseases and difficulties connected to hypertension [92].
High blood pressure in spontaneously hypertensive rats (SHR) can be reduced by administering an ethanol and water extract of FD. The renin-angiotensin-aldosterone system (RAAS) pathway, antioxidants, and the endothelial system may all play a role in this [93]. Rats treated with FD extract and losartan had significantly reduced blood pressure, heart rate, and heart weight compared to the controls (Table 5). The systolic blood pressure (SBP) of rats treated with FD plus losartan for four weeks was significantly lower than that of rats not treated with FD. Urine spectral analysis revealed 24 putative biomarkers with significance estimates greater than 0.5 (Table 1). Results from this study show that F. deltoidea has strong antihypertensive efficacy and shows promise for further research and development as an antihypertensive drug.
**Table 5.**
| S/N | Methods | Solvent | Concentrations | Major findings | Reference |
| --- | --- | --- | --- | --- | --- |
| 1 | In vivo | Ethanol and aqueous | 500, 800, 1,000, and 1,300 mg/kg | The ethanol and water extract of FDK lowers blood pressure in SHR. This could be due to RAAS, antioxidant, and endothelial systems | [93] |
| 2 | In vivo | Ethanol-water | 800 or 1,000 mg/kg/day | When FDA and losartan were used to treat rats, the rats’ blood pressure, heart rate, and heart weight were all much lower than the controls | [94] |
| 3 | In vivo | | 1,000 mg/kg/day | After four weeks of treatment with FD, the SBP of rats treated with FD and losartan was much lower than that of rats that were not treated. An analysis of the spectra of urine showed that there were 24 possible biomarkers with variable importance projections above 0.5 | [95] |
| 4 | | | 1,000 mg/kg | Compared to the control group, rats treated with 1,000 mg/kg of FD and losartan demonstrated a significant decrease in blood pressure. Rats treated with FDK had decreased serum concentrations of angiotensin II and aldosterone compared to controls and rats treated with losartan. There were no variations between serum and urine electrolytes | [96] |
## Aphrodisiac activity
Disturbances in a man’s sex drive might manifest as erectile dysfunction, ejaculatory dysfunction, or even an orgasmic condition [97]. Sexual activity is necessary for all humans and has been shown to increase the link between husband and wife, making it a good indicator of marital satisfaction. Psychological discomfort, infertility, and even suicide have all been linked to sexual dysfunction. Male mice benefit from an ethanolic extract from the leaves because it increases fertility and reproductive hormones (Table 1). Testosterone levels, sperm counts, and rat mobility improved after administering an aqueous and an ethanolic extract of FD leaves (Table 6). These treatments also substantially impacted the lowering of blood glucose levels, the number of abnormal sperm, and the frequency with which blood clots formed. Alloxan monohydrate’s negative effects on blood clotting, sperm quality, and testosterone level in male rats can be mitigated by administering an oral dose of an aqueous and ethanolic extract of FD leaves [98]. The sperm count, LDH-C 4 activity, and testosterone concentration of rats with diabetes were all enhanced after oral administration of an aqueous and ethanolic extract of FD leaves. These treatments had a major effect on blood glucose levels and sperm abnormalities [99]. Studies have demonstrated that plant polysaccharides help prevent testicular injury and encourage the renewal of testicular architecture. This study’s findings suggest that F. deltoidea extracts enhance general sexual performance and may also be useful in treating erectile dysfunction. The results support traditional medical claims that these plants have aphrodisiac properties and could improve sexual performance. These findings support the legitimacy of traditional Indian medicine’s recommendation for these plants.
**Table 6.**
| S/N | Methods | Solvent | Plant parts | Concentrations | Major findings | Reference |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | In vivo | Ethanol | Leaves | 125, 250, 500, and 1,000 mg/kg BW for 28 days | By enhancing fertility, reproductive hormones, and antioxidant activity, the ethanolic extract of the leaves has a positive impact on the reproduction of male mice | [100] |
| 2 | In vivo | Hot aqueous | Leaves | 0.125–4.0 mg/mL | The uterus of the isolated rat was subjected to a dose-dependent constriction by the FDA. As a result, maximum force of contraction (Emax) was lowered by all three of the drugs studied, with atosiban having a greater effect. The Emax was likewise lowered after oxodipine and ethylenediamine tetraacetic acid (EDTA) treatment. After the 2-aminoethoxydiphenyl borate (2-APB) administration, there was no noticeable difference. Thapsigargin, on the other hand, increased Emax | [101] |
| 3 | In vivo | Aqueous and ethanolic | Leaves | 800 mg/kg | Oral administration of F. deltoidea leaf aqueous and ethanolic extract improved sperm count, LDH-C 4 activity, and testosterone concentration in diabetic rats. In addition to lowering blood glucose and sperm abnormalities, these treatments had a considerable impact | [99] |
| 4 | In vivo | Methanol | Fruits | 50 mg/kg | Taking F. deltoidea dramatically improved male fertility | [102] |
| 5 | In vivo | Aqueous and ethanolic | Leaves | | When rats were given an aqueous, and an ethanolic extract of F. deltoidea leaves, their testosterone level, sperm count, and mobility improved. Moreover, these treatments greatly reduced the blood glucose levels, the number of abnormal sperm, and the rate at which the blood clots. In conclusion, giving male rats an oral dose of the aqueous and ethanolic extract of F. deltoidea leaves may reverse the effects of alloxan monohydrate on blood clotting, sperm quality, and testosterone levels | [98] |
## Wound healing activity
Major causes of physical incapacity include wounds [103]. A wound is a tissue disturbance brought on by physical, chemical, microbial, or functional losses [103]. Numerous factors, including bacterial infection, necrotic tissue, obstruction of the blood supply, lymphatic blockage, and diabetes mellitus, cause the wound healing process to be delayed (or reduced). Generally speaking, if any agent could change the aforementioned factors, the healing rate would be increased [104]. In Ayurveda, many plants play a crucial part in the recovery from injury. Plants are superior medicines because they work by stimulating the body’s natural mending processes [105]. Healing time is reduced, and aesthetics are preserved with plant-based therapy [106]. Animal products comprise less than $10\%$ of wound-healing pharmaceuticals, whereas plants account for over $70\%$. Antiseptic coagulants and wound washes made from plant-based ingredients are utilized in emergencies [106]. Compared to wounds treated with sterile deionized water or dressed with a blank placebo, wounds treated with a placebo containing $5\%$, $10\%$ F. deltoidea extract, or intrasite gel dramatically accelerated the healing rate [107]. A dose-dependent increase in cell proliferation can be achieved with leaf extract. In scratch testing, F. deltoidea leaf extract sped up wound healing compared to ascorbic acid-treated and untreated cells [108]. The $20\%$ methanolic extract of leaves has been shown to speed up the healing process of wounds (Table 7). In terms of DNA and hydroxyl proline content, mice given an extract concentration of $80\%$ showed the highest levels (Table 7). The extract’s wound-healing efficacy is proportional to its concentration (Table 7). FD extract is an effective treatment for wound healing since it can activate the body’s natural repair processes. The mechanism of F. deltoidea extract’s activity on the healing of wounds still must be understood.
**Table 7.**
| S/N | Methods | Solvent | Plant parts | Concentration | Major findings | Reference |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | In vivo | Methanol | Leaves | 20, 40, 60, and 80% | Methanolic extract of leaves at 20% concentration can heal wounds. The mice administered with an extract concentration of 80% exhibited the highest quantities of DNA and hydroxyl proline. The concentration of the extract influences how well it heals wounds | [109] |
| 2 | Scratch assays | Hot aqueous | Leaves | | Leaf extract can stimulate cell growth in a dose-dependent way. Compared to cells treated with ascorbic acid and untreated cells, F. deltoidea leaf extract accelerated wound closure in scratch assays | [108] |
| 3 | In vivo | Aqueous | Whole plants | Placebo containing 5% and 10% | Wounds treated with F. deltoidea extract containing 5 or 10% of the total extract considerably expedited the healing process compared to wounds treated with sterile deionized water | [107] |
| 4 | | Aqueous, methanol, and ethanol | Leaves | 10–1,000 μg/mL | Inhibition of human liver glucuronidation activity was found in the range from 34.69 μg/mL to 398.10 μg/mL | [110] |
| 5 | In vivo | Hot aqueous | Leaves | | The liver and kidneys were unaffected by the extract. Rats treated with the extract gained weight, improved depressed behaviour, and had fewer pyknotic and dark-stained neurons in their hippocampus | [111] |
## Anticancer activity
There has been much focus on plant-based biological products for quite some time. The potential for finding novel biomolecules with future applications motivates the investigation of these priceless by-products. Natural plant products are becoming increasingly popular for use in both the prevention and treatment of disease. The traditional use of medicinal plants as a treatment method is the basis for contemporary medicine [112]. The success rate of treating cancer with allopathic medicine or chemotherapy drugs like cisplatin has increased over time [16]. However, this course of treatment is commonly cited as having dangerous side effects because of the toxicity of chemotherapeutic drugs. Additionally, chemoresistance is to blame for $90\%$ of drug failures in patients with metastatic cancer [16]. Researchers have tried to find alternative treatment approaches to treat cancer, some of which involve using natural products. Drugs used in chemotherapy to treat cancer are typically based on chemicals first identified in plants or synthetic versions of these molecules [112]. Unfortunately, despite many efforts, cancer is still a major cause of death worldwide. Because of this, researchers are constantly looking for new, cost-effective treatments for cancer. Growing evidence suggests that compounds produced from plants may be able to inhibit several steps in carcinogenic and inflammation-related processes, making these products increasingly important in cancer prevention and treatment. Both 48.2 and 62.7 g/mL had IC50 values that suppressed microvascular proliferation (Table 8). Mice infected with azoxymethane/dextran sodium sulfate (AOM/DSS) had lower levels of alpha-catenin in their colons, which was inhibited by the FD ethanol extract. Human colon cancer (HCT 116) was also inhibited by the FD ethanol extract [113]. Nuclear DNA fragmentation showed that the extracts produced apoptosis, a form of cell death ($P \leq 0.05$). In PC3 and L ymph N ode *Ca rcinoma* of the P rostate (LNCaP) cell lines, there was also a substantial increase in mitochondrial membrane potential (MMP) depolarization ($P \leq 0.05$) and caspase 3 and 7 activations ($P \leq 0.05$) (Table 8). The IC50 values were calculated to be 224.39 μg/mL for the aqueous extract and 143.03 g/mL for the ethanolic extract. However, only the ethanolic extract (1,000 μg/mL) caused DNA fragmentation, while the water-based extract had no effect. The breaking caused a loss of about 200 kbp of DNA. Morphological testing revealed apoptotic bodies appeared in both extracts at concentrations of 1,000 μg/mL [114]. When tested for cytotoxic effects on the HL-60 cell line, it was discovered that the FD leaf extract was far more potent than the fruit extract [115]. The FD extract positively impacted tumour development. When the FD extract was used, the incidence of oral squamous cell cancer (OSCC) decreased from $100\%$ to $14.3\%$ in the high-dose groups [116]. At the end of the treatment period, there was a significant decrease in testosterone, FSH, and luteinizing hormone (LH) levels at $P \leq 0.05$ but a significant increase in progesterone and estrogen levels at $P \leq 0.05$ in extract treated groups compared to the control group [117]. We found that F. deltoidea extract significantly slowed the growth of established tumours, indicating that it possessed potent anticancer properties (Table 8). Flavonoids abundant in F. deltoidea include isovitexin, gallocatechin, ellagic acid, coumaroylquinic acid, catechin, gallic acid, quercetin (Figure 6), and naringenin. The anticancer benefits of the plant are due to these compounds [8]. F. deltoidea has high levels of polyphenolics, flavonoids (such as genistin), alkaloids (such as antofine), and tannins [118, 119]. Flavonoids like epigallocatechin have been proven to inhibit the growth of prostate cancer cells in vitro [120]. Vitexin has a cytotoxic effect on breast, ovary, and prostate cancer cells by upregulating BCL2-associated X protein (Bax) and downregulating BCL2 and causing the breakage of the poly[adenosine diphosphate (ADP)-ribose] polymerase (PARP) protein [121]. By decreasing the BCL2/Bax ratio and activating caspases, vitexin inhibits tumour growth and spread by killing cancer cells [122]. Naturally occurring antioxidant ellagic acid has been demonstrated to have antiproliferative and pro differentiation actions on prostate cancer cells via suppressing eicosanoid synthesis and the heme oxygenase system [123]. Murine leukaemia cells and the human lung cancer cell line have both been shown to undergo apoptosis when treated with the antioxidants rutin and quercetin, respectively, which have been linked to having anticancer characteristics. Plant polyphenols have long been recognized for their antioxidative effects against oxidative stress, which has been associated with cancer [119]. The flavonoids in F. deltoidea have therapeutic potential as a treatment for prostate cancer [122]. The possible mechanism of action of F. deltoidea as a tumour suppressor and its crude extract or pure components is shown in Figure 7. It also depicts the anti-tumorigenic activities induced by signal transducing components by crude extracts or pure chemicals and the tumour cascade pathways initiated in cancerous cells by various growth factors in Figure 7. The expression of the tumour-inducing pathways phosphoinositide 3-kinase (PI3K), protein kinase B (Akt), natriuretic peptide type B (NP-B), mitogen-activated protein kinase (MAPK), and ROS is downregulated at the infection site by crude extract or pure chemical in a conjugated form. Pure substance or crude extract interrupts the cycle, prevents the synthesis of p21 and p27 cyclin-dependent kinase inhibitors, and prevents the mitotic effects. These activities are all connected to cancer cell development, dissemination, and proliferation. F. deltoidea killed multiple tumour cell lines; however, the effectiveness was dose- and time-dependent. This analysis verified the results of ethnobotanical studies that reveal the medicinal potential of F. deltoidea used in traditional medicine. Based on our findings, F. deltoidea extracts or pure compounds have great potential as an anticancer drug.
## Toxicity evaluation
Despite their efficacy in treating specific diseases, the widespread use of some medicinal plant species is associated with serious adverse effects. Many pharmaceuticals owe their existence to the discovery of a chemical in a plant that has biological activity and has subsequently been used to treat medical conditions. The natural chemical compounds in plants give them pharmacological and therapeutic properties, and their potential toxicity must be evaluated to ensure that the product is safe for human consumption (Table 9). Compared to the control group, no appreciable variations in the number of micro-nucleated cells were seen. At concentrations up to 5,000 g/plate, the extract was not found to increase the number of revertant colonies in any strains tested. In conclusion, more studies using animal models are necessary to confirm non-geno-harmful effects [128]. Uterine abnormalities in bisphenol A (BPA)-exposed rats improved significantly after six weeks of concurrent therapy with F. deltoidea. The histology of the myometrium and glandular epithelium appeared normal, and mitotic patterns were visible in the interstitial gaps between the stromal cells [129]. In an acute toxicity assay, the extract’s median lethal dose (LD50) was greater than at the concentration of 5,000 mg/kg. The sub-chronic toxicity study findings were shown to have no impact on food consumption, BW, organ weight, mortality, clinical chemistry, haematological, gross pathology, or histology (Table 1). All extracts had a higher than 2,000 mg/kg BW and LD50, and the acute toxicity test showed no signs of morbidity or mortality. Histopathological analyses of the kidneys and liver showed no abnormalities [84] despite the non-reportage of any toxic part of F. deltoidea. Additional testing is required using various cell lines in a range of sample dilutions. The medication is then tested in rodents and other animals, particularly mice and rats, before being administered to human patients.
**Table 9.**
| S/N | Methods | Solvent | Plant parts | Concentrations | Major findings | Reference |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | | Methanol, chloroform, ethyl acetate, and butanol | Leaves | 0.01–100 mg/mL | The extracts also had an anti-proliferative activity that was dosage dependant. All three cell lines tested were practically ineffective against both butanol and ethyl acetate extracts (IC50 values > 1,000 g/mL). All three cell lines responded well after 48 h of treatment with methanol extract | [1] |
| 2 | Alkaline comet assay | Aqueous | Leaves | 5, 2.5, 1.25, 0.625, 0.3125, and 0.15625 mg/mL | No significant differences were identified in the number of micro-nucleated cells compared to the control group. The extract did not enhance the number of revertant colonies in any strains tested at levels up to 5,000 μg/plate. To summarize, additional research in animal models is required to verify FDA’s non-geno toxic actions | [128] |
| 3 | In vivo | Methanol, n-hexane, chloroform, and n-butanol | Leaves | 100, 200, and 400 mg/kg | Unlike hazardous chloroform and hexane sub extracts, hydrophilic methanol extract resulted in zero per cent mortality up to 6,400 mg/kg in 14 days. After four weeks of administration of 200 mg/kg, it did not generate liver or renal damage. The methanol extract revealed a low level of oral toxicity and diverse antidiabetic effects | [81] |
| 4 | MTT | Aqueous | Leaves and fruits | 1 ng/mL to 1 mg/mL | Lowering the F. deltoidea leaf extract concentration from 1 mg/mL to 1 ng/mL increased cell viability | [21] |
| 5 | MTT | Aqueous | | 0.1, 1, 10, and 100 μg/mL | The extract was not harmful at any concentrations tested because microglial cell viability was consistently more than 80% | [43] |
| 6 | In vivo | | Leaves | 2,000 mg/kg | At oral doses of 2 g/kg, neither vitexin (1) nor isovitexin (2) showed any symptoms of toxicity in normoglycemic mice or diabetic rats | [82] |
| 7 | In vivo | Petroleum ether, chloroform, and methanol | Leaves | 50–5,000 mg/kg BW | All the doses examined resulted in no treatment-related deaths. After 14 days, there were no significant changes in the animals’ behaviour, such as apathy and hyperactivity, as well as illness and mortality | [83] |
| 8 | MTT | Hot aqueous | Leaves | | The maximum extract concentration that did not affect cell viability was 0.1% (w/v) | [125] |
| 9 | | | Leaves | 1,000 mg/BW | This group did not affect glycaemia variables, although total and LDL cholesterol values were dramatically reduced. Vital signs and safety lab tests were within normal ranges at baseline and after 8 weeks of intervention, there were no significant differences between groups or attributable to the intervention | [130] |
| 10 | Brine shrimp lethality assay and in vivo | Aqueous | Leaves | | According to the research, the extracts had no harmful effects on brine shrimp (up to 4,000 μg/mL) or rats (up to 0.2 per cent BW) | [89] |
| 11 | In vivo | Ethanolic | Leaves | | The LD50 of the extract was found to be more than 5,000 mg/kg in an acute toxicity assay. Food consumption, BW, organ weight, mortality, clinical chemistry, haematological, gross pathology, and histopathology were all unaffected by the sub-chronic toxicity study results | [131] |
| 12 | In vivo | | Leaves | 1,000 mg/kg | It was shown that up to 1,000 mg/kg of F. deltoidea leaf extract was not harmful | [124] |
| 13 | In vivo | Aqueous | Leaves | 100 mg/kg/day | Uterine abnormalities in the BPA-exposed rats significantly improved after six weeks of concomitant therapy with F. deltoidea. The myometrium and glandular epithelium histology seemed normal, and mitotic patterns were present in the interstitial gaps between the stromal cells | [129] |
| 14 | In vivo | Ethanol | Leaves | 125, 250, 500, and 1,000 mg/kg BW for 28 days | The leaves’ ethanolic extract has no harmful effects and does not alter the histological structure of the testes | [100] |
| 15 | In vivo | Aqueous | Leaves | 100 mg/kg/BW | The data demonstrated that F. deltoidea had a protective effect against BPA-induced ovarian damage. Normalization of FSH and sexual steroid hormone (progesterone) levels supported this conclusion | [132] |
| 16 | In vivo | Ethanolic aqueous | Leaves | 5, 50, 300, and 2,000 mg/kg | The LD50 of the extracts for all kinds was higher than 2,000 mg/kg BW, and the acute toxicity test revealed no symptoms of morbidity or mortality. The kidneys and liver’s histopathological evaluation revealed no abnormalities | [84] |
| 17 | In vivo | | Fruits | | According to the testing data, the tensile strength of carbon nanotube (CNT)-filled composites increased by 7.73% compared to the control unfilled hybrid composites. For the CNT-filled composites, the flexural characteristics decreased by 49.37% compared to the control, which had no CNTs | [133] |
| 18 | In vivo | Methanol: distilled water (60:40 % v/v) | Leaves | 300, 2,000, and 4,000 mg/kg | Some important organs underwent haematological and histological examination. Mortality was not recorded at any point during the study in either the acute or sub chronic toxicity groups Encapsulated plant extracts (600 and 1,000 mg/kg) increased serum glutamic oxaloacetic transaminase (SGOT) and serum glutamic pyruvic transaminase (SGPT) levels significantly, and histological assessment of the liver, kidneys, and spleen showed normal tissue limits | [134] |
| 19 | Viability assay | Methanol | | 100 μL | Viability was only shown to be hazardous at 500 and 1,000 μg/mL (P < 0.001) | [76] |
| 20 | In vivo | Ethanol | Leaves | 125, 250, 500, and 1,000 mg/kg BW | The absence of toxic symptoms and death at a 2,000 mg/kg BW dose suggests that the LD50 was higher. Throughout this time, no changes in the mouse’s behaviour, substantial weight changes, haematological parameters, or serum biochemistry were noticed | [135] |
## Discussion
The therapeutic properties of FD have been recognized for centuries, and the elderly have found several uses. Scientific research was conducted to confirm its effects, particularly in pharmaceutical applications, as it gained increasing attention. Its biological efficacy was documented in the current study. These findings provide solid evidence for the considerable and positive benefits of F. deltoidea extract on the rate of wound healing, cancer, fever, diabetes, blood pressure, bacterial infection, fungal infection, and many other diseases due to the presence of phenolic and flavonoid bioactive compounds. However, additional research into the bioactive components that may be responsible for its anticancer activity is necessary. Further studies must determine the appropriate dosage for treating and controlling cancer and related disorders globally. This study may serve as a solid foundation for creating herbal medicines or active compounds with tremendous potential for use in the treatment and prevention of cancer and its related future.
## Author contributions
The author contributed solely to the work.
## Conflicts of interest
The author declares that he has no conflicts of interest.
## Ethical approval
Not applicable.
## Consent to participate
Not applicable.
## Consent to publication
Not applicable.
## Availability of data and materials
Not applicable.
## Funding
Not applicable.
## Copyright
© The Author(s) 2023.
## References
1. Ashraf K, Halim H, Lim SM, Ramasamy K, Sultan S.. *Saudi J Biol Sci* (2020.0) **27** 417-32. DOI: 10.1016/j.sjbs.2019.11.003
2. Abdulrahman MD, Hamad SW, Hama HA, Bradosty SW, Kayfi S, Al-Rawi SS. **Biological evaluation of**. *Adv Pharmacol Pharm Sci* (2022.0) **2022** 3837965. DOI: 10.1155/2022/3837965
3. Abdulrahman MD.. **Biological activity and chemical composition of**. *Adv Pharmacol Pharm Sci* (2022.0) **2022** 7219401. DOI: 10.1155/2022/7219401
4. Bunawan H, Amin NM, Bunawan SN, Baharum SN, Mohd Noor N.. *Evid Based Complement Alternat Med* (2014.0) **2014** 902734. DOI: 10.1155/2014/902734
5. Ashraf K, Haque MR, Amir M, Ahmad N, Ahmad W, Sultan S. **An overview of phytochemical and biological activities:**. *J Pharm Bioallied Sci* (2021.0) **13** 11-25. DOI: 10.4103/jpbs.JPBS_232_19
6. Berg CC, Corner EJH, Jarrett FM.. **Moraceae genera other than Ficus**. *Flora Malesiana, Series I: Spermatophyta* (2006.0) **17** 1-146
7. Abrahim NN, Abdul-Rahman PS, Aminudin N.. **The antioxidant activities, cytotoxic properties, and identification of water-soluble compounds of**. *PeerJ* (2018.0) **6** e5694. DOI: 10.7717/peerj.5694
8. Hanafi MMM, Afzan A, Yaakob H, Aziz R, Sarmidi MR, Wolfender JL. *Front Pharmacol* (2017.0) **8** 895. DOI: 10.3389/fphar.2017.00895
9. Suryati S, Nurdin H, Hamidi D, Lajis MN.. **Structure elucidation of antibacterial compound from**. *Indones J Chem* (2011.0) **11** 67-70. DOI: 10.22146/ijc.21422
10. Wei LS, Wee W, Siong JYF, Syamsumir DF.. **Characterization of antioxidant, antimicrobial, anticancer property and chemical composition of**. *J Biol Act Prod Nat* (2011.0) **1** 1-6. DOI: 10.1080/22311866.2011.10719067
11. Abdullah Z, Hussain K, Ismail Z, Ali RM.. **Anti-inflammatory activity of standardised extracts of leaves of three varieties of**. *Int J Pharm Clin Res* (2009.0) **1** 100-5
12. Abolmaesoomi M, Abdul Aziz A, Mat Junit S, Ali JM.. *Eur J Integr Med* (2019.0) **28** 57-67. DOI: 10.1016/j.eujim.2019.05.002
13. Fatihah HNN, Mat N, Zaimah ARN, Khairil M, Ali AM.. **Leaf morphology and anatomy of 7 varieties of**. *Turk J Bot* (2014.0) **38** 677-85. DOI: 10.3906/bot-1301-7
14. Fatihah HNN, Mat N, Zaimah ARN, Zuhailah MN, Norhaslinda H, Khairil M. **Correction: morphological phylogenetic analysis of seven varieties of**. *PLoS One* (2013.0) 8. DOI: 10.1371/annotation/92e7b830-9db4-47cf-9851-8fa939b1dc2c
15. Sulaiman MR, Hussain MK, Zakaria ZA, Somchit MN, Moin S, Mohamad AS. **Evaluation of the antinociceptive activity of**. *Fitoterapia* (2008.0) **79** 557-61. DOI: 10.1016/j.fitote.2008.06.005
16. Al-Koshab M, Alabsi AM, Mohd Bakri M, Ali-Saeed R, Selvi Naicker M.. **Antitumor activity of**. *J Oncol* (2020.0) **2020** 5490468. DOI: 10.1155/2020/5490468
17. Zunoliza A, Khalid H, Zhari I, Rasadah MA, Mazura P, Fadzureena J. **Evaluation of extracts of leaf of three**. *Pharmacogn Res* (2009.0) **1** 216-23
18. Misbah H, Aziz AA, Aminudin N.. **Antidiabetic and antioxidant properties of**. *BMC Complement Altern Med* (2013.0) **13** 118. DOI: 10.1186/1472-6882-13-118
19. Abdullah FI, Chua LS, Rahmat Z, Soontorngun N, Somboon P.. **Trypsin hydrolysed protein fractions as radical scavengers and anti-bacterial agents from**. *Int J Pept Res Ther* (2018.0) **24** 279-90. DOI: 10.1007/s10989-017-9613-5
20. Omar MH, Mullen W, Crozier A.. **Identification of proanthocyanidin dimers and trimers, flavone**. *J Agric Food Chem* (2011.0) **59** 1363-9. DOI: 10.1021/jf1032729
21. Dzolin S, Aris SRS, Ahmad R, Zain MM.. **Radical scavenging and neurotoxicity of four varieties of**. 11-5. DOI: 10.1109/CSSR.2010.5773717
22. Mohd Dom NS, Yahaya N, Adam Z, Nik Abd Rahman NMA, Hamid M.. **Antiglycation and antioxidant properties of**. *Evid Based Complement Alternat Med* (2020.0) **2020** 6374632. DOI: 10.1155/2020/6374632
23. Manurung H, Kustiawan W, Kusuma IW, Marjenah M, Nugroho RA.. **Growth, phytochemical profile, and antioxidant activity of cultivated tabat barito (**. *Int J Biosci* (2019.0) **14** 366-78
24. Ahmed RK, Ahmed MS, Mamat AS, Dahham SS.. **A comparison of antioxidant potential, the total phenolicand flavonoid content of male and female (Ficus deltoidea)**. *Int J Appl Chem* (2016.0) **12** 120-4
25. Yunusa AK, Rashid ZM, Mat N, Bakar CAA, Ali AM.. **Chemicals and bioactivity discrimination of syconia of seven varieties of**. *Pharmacog J* (2018.0) **10** s147-51. DOI: 10.5530/pj.2018.6s.27
26. Hakiman M, Syed MA, Ahmad S, Mahmood M.. **Total antioxidant, polyphenol, phenolic acid, and flavonoid content in**. *J Med Plants Res* (2012.0) **6** 4776-84. DOI: 10.5897/JMPR11.1027
27. Wahid S, Mahmud TMM, Maziah M, Yahya A, Rahim MA.. **Total phenolics content and antioxidant activity of hot water extracts from dried**. *J Trop Agric Food Sci* (2010.0) **38** 115-22
28. Farsi E, Shafaei A, Hor SY, Ahamed MBK, Yam MF, Attitalla IH. **Correlation between enzymes inhibitory effects and antioxidant activities of standardized fractions of methanolic extract obtained from**. *Afri J Biotechnol* (2011.0) **10** 15184-94. DOI: 10.5897/AJB11.1365
29. Hakiman M, Ariff SM, Ahmad S, Zulperi D, Mahmood M.. **Estimation of total phenolic acids, flavonoid compounds and antioxidant activity of**. *J Agrobiotechnol* (2018.0) **9** 58-69
30. Soib HH, Ware I, Yaakob H, Mukrish H, Sarmidi MR.. **Antioxidant and anti-cancer activity of standardized extracts of three varieties of**. *J Teknol* (2015.0) **77** 19-25. DOI: 10.11113/jt.v77.6000
31. Sin MH, Mamat AS, Aslam MS, Ahmad MS.. **Total phenolic content and anti-oxidant potential of**. *J Pharm Negat Results* (2017.0) **8** 15-9. DOI: 10.4103/0976-9234.204913
32. Manurung H, Kustiawan W, Kusuma IW, Marjenah M.. **Total flavonoid content and antioxidant activity in leaves and stems extract of cultivated and wild tabat barito (**. *AIP Conf Proc* (2017.0) **1813** 020007. DOI: 10.1063/1.4975945
33. Aris SRS, Mustafa S, Ahmat N, Jaafar FM, Ahmad R.. **Phenolic content and antioxidant activity of fruits of**. *Malaysian J Anal Sci* (2009.0) **13** 146-50
34. Ibrahim FW, Abdullah AF, Chan YL, Jufri NF, Mohammad N, Rajab NF.. *J Appl Pharm Sci* (2021.0) **11** 147-53. DOI: 10.7324/JAPS.2021.110820
35. Jani NA, Azman Shah NN, Preshahdin NA, Rokman FA, Shamsuri NN.. *International Jasin Multimedia & Computer Science Invention & Innovation Exhibition* (2020.0) 89-92
36. Hasham R, Choi HK, Sarmidi MR, Park CS.. **Protective effects of a**. *Biotechnol Bioprocess Eng* (2013.0) **18** 185-93. DOI: 10.1007/s12257-012-0353-2
37. Dzolin S, Ahmad R, Mazatulikhma MZ, Aris SRS.. **Inhibition of free radical and neuroprotective effect of four varieties of**. *Adv Mat Res* (2012.0) **554–556** 1371-80. DOI: 10.4028/www.scientific.net/AMR.554-556.1371
38. Lee SY, Sew JY, Chin KL, Tee EC, Tang MW, Chee CF. **Isolation and identification of antioxidant compounds in methanolic extract of both female and male plants of**. *Malays J Chem* (2013.0) **15** 55-66
39. Abu Bakar AR, Manaharan T, Merican AF, Mohamad SB.. **Experimental and computational approaches to reveal the potential of**. *Nat Prod Res* (2018.0) **32** 473-6. DOI: 10.1080/14786419.2017.1312393
40. Alwi A, Rosli AS, Mohd K, Azemin A, Mat N, Ali AM.. **Evaluation of antioxidant and anti-diabetic potentials of eight varieties of medicinal plant**. 2014
41. Manurung H, Kustiawan W, Kusuma IW, Marjenah M.. **Total flavonoid content and antioxidant activity of tabat Barito (**. *J Med Plants Stud* (2017.0) **5** 120-5
42. Zakaria ZA, Hussain MK, Mohamad AS, Abdullah FC, Sulaiman MR.. **Anti-inflammatory activity of the aqueous extract of**. *Biol Res Nurs* (2012.0) **14** 90-7. DOI: 10.1177/1099800410395378
43. Zolkiffly SZI, Stanslas J, Abdul Hamid H, Mehat MZ.. *J Ethnopharmacol* (2021.0) **279** 114309. DOI: 10.1016/j.jep.2021.114309
44. Omar NI, Baharin B, Lau SF, Ibrahim N, Mohd N, Ahmad Fauzi A. **The influence of**. *Vet Med Int* (2020.0) **2020** 8862489. DOI: 10.1155/2020/8862489
45. Hasham R, Choi HK, Park CS.. *Int J Biol Vet Agric Food Eng* (2014.0) **8** 1096-100. DOI: 10.5281/zenodo.1096717
46. Che Ahmad Tantowi NA, Lau SF, Mohamed S.. **Ficus deltoidea prevented bone loss in preclinical osteoporosis/osteoarthritis model by suppressing inflammation**. *Calcif Tissue Int* (2018.0) **103** 388-99. DOI: 10.1007/s00223-018-0433-1
47. Mohd Ariff A, Abu Bakar NA, Abd Muid S, Omar E, Ismail NH, Ali AM. *BMC Complement Med Ther* (2020.0) **20** 56. DOI: 10.1186/s12906-020-2844-6
48. Che Ahmad Tantowi NA, Hussin P, Lau SF, Mohamed S.. **Mistletoe fig (**. *Menopause* (2017.0) **24** 1071-80. DOI: 10.1097/GME.0000000000000882
49. Santhanam R, Gothai S, Karunakaran T, Muniandy K, Kandasamy S.. **Identification of chemical constituents and inhibitory effect of**. *Pharmacogn Mag* (2021.0) **17** 236. DOI: 10.4103/pm.pm_433_20
50. Tkachenko H, Buyun L, Terech-Majewska E, Osadowski Z.. *Fish Aquat Life* (2017.0) **24** 219-30. DOI: 10.1515/aopf-2016-0019
51. Jamal P, Ismail AK, Abdullah E, Ahmad Raus R, Hashim YZH.. **Phytochemical screening for antibacterial activity of potential Malaysian medicinal plants**. *Afr J Biotechnol* (2011.0) **10** 18795-9. DOI: 10.5897/AJB11.2755
52. Abdsamah O, Zaidi NT, Sule AB.. **Antimicrobial activity of**. *Pak J Pharm Sci* (2012.0) **25** 675-8. PMID: 22713960
53. Ahmad VN, Muhammad Zain N, Mohd Amin I, Zulzaidi F, Ahmad Jafri MS., Yusof MYPM, Bakri NN. *9th Dental Students’ Scientific Symposium – Proceeding Book* (2019.0) 26-9
54. Janatiningrum I, Lestari Y.. **Enzyme production, antibacterial and antifungal activities of actinobacteria isolated from**. *Biodiversitas* (2022.0) **23** 1950-7. DOI: 10.13057/biodiv/d230429
55. Uyub AM, Nwachukwu IN, Ahmad AL, Fariza SS.. (2010.0) **8** 95-106. DOI: 10.17348/era.8.0.95-106
56. Suryati H, Nurdin MN.. **Characterization antibacterial constituent from**. *Indones J Pharm* (2010.0) **21** 134-8. DOI: 10.14499/indonesianjpharm0iss0pp134-138
57. Azizan N, Mohd Said S, Zainal Abidin Z, Jantan I., Kumolosasi E, Md Ridzuan F, Islahudin FH, Mohd Shah N, Mohd Makmor B, Md Ali A, et al.,. **Antibacterial activity of**. *Proceedings of pharmaceutical sciences research day 2015* (2015.0) 131-2
58. Azizan N, Mohd Said S, Zainal Abidin Z, Jantan I.. **Composition and antibacterial activity of the essential oils of**. *Molecules* (2017.0) **22** 2135. DOI: 10.3390/molecules22122135
59. Rahayu S, Amaliah N, Patimah R.. **Uji aktivitas antibakteri ekstrak daun Tabat Barito (**. *Jurnal Riset Kefarmasian Indonesia* (2022.0) **4** 34-45. DOI: 10.33759/jrki.v4i1.229
60. Krisyanella K, Ardianti M, Dachriyanus D.. **Uji aktivitas antimikroba ekstrak metanol daun Tabat Barito (Ficus deltoidea Jack)**. *Jurnal Farmasi Higea* (2009.0) **1** 29-35. DOI: 10.52689/higea.v1i1.6
61. Tkachenko H, Buyun L, Terech-Majewska E, Osadowski Z.. **Antibacterial activity of ethanolic leaf extracts obtained from various**. *J Ecol Prot Coastline* (2016.0) **20** 117-36
62. Alimon H, Safiai SN, Noor NNM, Daud N, Abdul Azziz SSS.. **Comparisons of antibacterial activity of crude extracts between two varieties of**. *Open Conf Proc J* (2013.0) **4** 117. DOI: 10.2174/2210289201304010117
63. Wong MY, Hamid S, Iskandar Shah NA, Rahim NM, Zainal Kasim Z, Ab Razak NH.. **Antifungal potential of six herbal plants against selected plant fungal pathogens**. *Pertanika J Trop Agric Sci* (2020.0) **43** 107-17. DOI: 10.47836/pjtas.43.4.03
64. Mukhtar HM, Singh A, Kaur H.. **Bioassay guided fractionation and**. *Pharmacogn J* (2018.0) **10** 235-40. DOI: 10.5530/pj.2018.2.41
65. 65.Guideline: sugars intake for adults and children. Geneva: World Health Organization. 2015.. *Guideline: sugars intake for adults and children* (2015.0)
66. Devlin MJ, Yanovski SZ, Wilson GT.. **Obesity: what mental health professionals need to know**. *Am J Psychiatry* (2000.0) **157** 854-66. DOI: 10.1176/appi.ajp.157.6.854
67. Woon SM, Seng YW, Ling AP, Chye SM, Koh RY.. **Anti-adipogenic effects of extracts of**. *J Zhejiang Univ Sci B* (2014.0) **15** 295-302. DOI: 10.1631/jzus.B1300123
68. Adam Z, Khamis S, Ismail A, Hamid M.. *Evid Based Complement Alternat Med* (2012.0) **2012** 632763. DOI: 10.1155/2012/632763
69. Haslan MA, Samsulrizal N, Hashim N, Zin NSNM, Shirazi FH, Goh YM.. *BMC Complement Med Ther* (2021.0) **21** 291. DOI: 10.1186/s12906-021-03452-6
70. Adam Z, Khamis S, Ismail A, Hamid M.. **Inhibitory properties of**. *Res J Med Plant* (2010.0) **4** 61-75. DOI: 10.3923/rjmp.2010.61.75
71. Adam Z, Hamid M, Ismail A, Khamis S.. **Effect of**. *J Biol Sci* (2009.0) **9** 796-803. DOI: 10.3923/jbs.2009.796.803
72. Janatiningrum I, Solihin DD, Meryandini A, Lestari Y.. **Rat alpha glucosidase inhibitor and phytochemicals activities of endophytic actinobacteria from**. *Pak J Pharm Sci* (2020.0) **33** 969-75. PMID: 33191220
73. Adam Z, Razali R, Arapoc DJ, Hanif A, Marsidi N.. **The enhancement effect of**. (2021.0) 1-9
74. Adam Z, Khamis S, Ismail A, Hamid M.. *J Nucl Relat Technol* (2015.0) **12** 54-65
75. Dramant S, Md Aris MA, Rus RM, Akter SFU, Azlina H, Norazlina AR. **Mas Cotek (**. *Pertanika J Trop Agric Sci* (2012.0) **35** 93-102
76. Yahaya N, Mohd Dom NS, Adam Z, Hamid M.. **Insulinotropic activity of standardized methanolic extracts of**. *Evid Based Complement Alternat Med* (2018.0) **2018** 3769874. DOI: 10.1155/2018/3769874
77. Adam Z, Hamid M, Ismail A, Khamis S.. **Insulin secreting and insulin-like activity of**. *Seminar R&D 2008* (2008.0) 1-11
78. Abdel-Rahman RF, Ezzat SM, Ogaly HA, Abd-Elsalam RM, Hessin AF, Fekry MI. *J Nutr Sci* (2020.0) **9** e2. DOI: 10.1017/jns.2019.40
79. Adam Z, Ismail A, Hamid M, Khamis S.. **Ficus deltoidea: a potential source for new oral antidiabetic agent**. *Seminar R&D Nuklear Malaysia* (2012.0)
80. Nurdiana S, Goh YM, Hafandi A, Dom SM, Nur Syimal’ain A, Noor Syaffinaz NM. **Improvement of spatial learning and memory, cortical gyrification patterns and brain oxidative stress markers in diabetic rats treated with**. *J Tradit Complement Med* (2018.0) **8** 190-202. DOI: 10.1016/j.jtcme.2017.05.006
81. Ilyanie Y, Wong TW, Choo CY.. **Evaluation of hypoglycemic activity and toxicity profiles of the leaves of Ficus deltoidea in rodents**. *J Complement Integr Med* (2011.0) **8** 1-16. DOI: 10.2202/1553-3840.1469
82. Choo CY, Sulong NY, Man F, Wong TW.. **Vitexin and isovitexin from the leaves of**. *J Ethnopharmacol* (2012.0) **142** 776-81. DOI: 10.1016/j.jep.2012.05.062
83. Farsi E, Ahmad M, Hor SY, Ahamed MB, Yam MF, Asmawi MZ. **Standardized extract of**. *BMC Complement Altern Med* (2014.0) **14** 220. DOI: 10.1186/1472-6882-14-220
84. Noor HSM, Ismail NH, Kasim N, Mohd Zohdi R, Ali AM.. **Hypoglycemic and glucose tolerance activity of standardized extracts**. *J Med Plants Stud* (2016.0) **4** 275-9
85. Adam Z, Hamid M, Ismail A, Khamis S.. **Effect of**. *Malaysian J Health Sci* (2007.0) **5** 9-16
86. Mohammad Noor HS, Ismail NH, Kasim N, Mediani A, Mohd Zohdi R, Ali AM. **Urinary metabolomics and biochemical analysis of antihyperglycemic effect of**. *Appl Biochem Biotechnol* (2020.0) **192** 1-21. DOI: 10.1007/s12010-020-03304-y
87. Adam Z, Ismail A, Khamis S, Mohamad Mokhtar MH, Hamid M.. **Antihyperglycemic activity of**. *Sains Malays* (2011.0) **40** 489-95
88. Papitha R, Renu K, Immanuel Selvaraj C, Valsala Gopalakrishnan A., Roopan SM, Madhumitha G. **Anti-diabetic effect of fruits on different animal model system**. *Bioorganic phase in natural food: an overview* (2018.0) 157-85. DOI: 10.1007/978-3-319-74210-6_9
89. Khamis S, Ibrahim SH, Jamaludin NA, Mohamad Mokhtar MH, Nor Azizah M.. **Evaluation for antidiabetic activity in selected medicinal plants used in Malaysian traditional medicine for the treatment of diabetes**. 158-61
90. Nurdiana S, Goh YM, Ahmad H, Dom SM, Syimal’ain Azmi N, NoorMohamad Zin NS. **Changes in pancreatic histology, insulin secretion and oxidative status in diabetic rats following treatment with**. *BMC Complement Altern Med* (2017.0) **17** 290. DOI: 10.1186/s12906-017-1762-8
91. Nurdiana S, Goh YM, Ahmad H, Dom SM, Syimal’ain Azmi N, NoorMohamad Zin NS. *Pharm Biol* (2021.0) **59** 66-73. DOI: 10.1080/13880209.2020.1865411
92. Guerrero MF, Puebla P, Carrón R, Martín ML, Arteaga L, Román LS.. **Assessment of the antihypertensive and vasodilator effects of ethanolic extracts of some Colombian medicinal plants**. *J Ethnopharmacol* (2002.0) **80** 37-42. DOI: 10.1016/S0378-8741(01)00420-2
93. Azis NA, Agarwal R, Ismail NM, Ismail NH, Kamal MSA, Radjeni Z. **Blood pressure lowering effect of**. *Mol Biol Rep* (2019.0) **46** 2841-9. DOI: 10.1007/s11033-019-04730-w
94. Kamal MSA, Ismail NH, Satar NA, Azis NA, Radjeni Z, Mohammad Noor HS. **Standardized ethanol-water extract of**. *Clin Exp Hypertens* (2019.0) **41** 444-51. DOI: 10.1080/10641963.2018.1506467
95. Kamal MSA, Mediani A, Kasim N, Ismail NH, Satar NA, Azis NA. **Blood pressure and urine metabolite changes in spontaneously hypertensive rats treated with leaf extract of**. *J Pharm Biomed Anal* (2022.0) **210** 114579. DOI: 10.1016/j.jpba.2021.114579
96. Azis N, Agarwal R, Mohd Ismail N, Ismail NH, Ahmad Kamal MS, Radjeni Z. **Possible involvement of RAAS in the mechanism of anti-hypertensive effect of standardized aqueous ethanolic extract of Ficus deltoidea Kunstleri in spontaneously hypertensive rats**. *Int J Cardiol* (2017.0) **249** S6. DOI: 10.1016/j.ijcard.2017.09.042
97. Fauzi F, Widodo H.. **Short communication: plants used as aphrodisiacs by the Dayak ethnic groups in Central Kalimantan, Indonesia**. *Biodiversitas* (2019.0) **20** 1859-65. DOI: 10.13057/biodiv/d200710
98. Nurdiana S, Mohd Idzham AZ, Zanariah A, Hakim MML.. **Effect of Ficus deltoidea leaves extracts on blood clotting, sperm quality and testosterone level in alloxan-induced male diabetic rats**. *Int J Pharm Sci Rev Res* (2012.0) **13** 111-4
99. Samsulrizal N, Awang Z, Mohd Najib MLH, Idzham M, Zarin A.. **Effect of Ficus deltoidea leaves extracts on sperm quality, LDH-C 4 activity and testosterone level in alloxan-induced male diabetic rats**. 888-91. DOI: 10.1109/CHUSER.2011.6163864
100. Nugroho RA, Aryani R, Manurung H, Anindita DF, Hidayati FSN, Prahastika W. **Effects of the ethanol extracts of**. *Open Access Maced J Med Sci* (2022.0) **10** 146-52. DOI: 10.3889/oamjms.2022.8068
101. Salleh N, Ahmad VN.. *BMC Complement Altern Med* (2013.0) **13** 359. DOI: 10.1186/1472-6882-13-359
102. Daud D, Awe WA, Kamal NA, Tawang A.. **Comparative effects of**. *J Pharm Adv Res* (2019.0) **2** 723-7
103. Nagori BP, Solanki R.. **Role of medicinal plants in wound healing**. *Res J Medicinal Plant* (2011.0) **5** 392-405. DOI: 10.3923/rjmp.2011.392.405
104. Chithra P, Sajithlal GB, Chandrakasan G.. **Influence of aloe vera on collagen turnover in healing of dermal wounds in rats**. *Indian J Exp Biol* (1998.0) **36** 896-901. PMID: 9854430
105. Shenoy C, Patil MB, Kumar R, Patil S.. **Preliminary phytochemical investigation and wound healing activity of**. *Int J Pharm Pharm Sci* (2009.0) **2** 167-75
106. Kumar B, Vijayakumar M, Govindarajan R, Pushpangadan P.. **Ethnopharmacological approaches to wound healing—exploring medicinal plants of India**. *J Ethnopharmacol* (2007.0) **114** 103-13. DOI: 10.1016/j.jep.2007.08.010
107. Abdulla MA, Ahmed KA, Faisal MA, Mazin M.. **Role of**. *Biomed Res* (2010.0) **21** 241-5
108. Mustaffa NAAW, Hasham R, Sarmidi MR.. **An**. *J. Teknol* (2015.0) 77. DOI: 10.11113/jt.v77.6008
109. Aryani R, Nugroho RA, Manurung H, Rani M, Rudianto R, Prahastika W. *Eurasia J Biosci* (2020.0) **14** 85-91
110. Husni Z, Ismail S, Zulkiffli MH, Afandi A, Haron M.. *Pharmacogn Mag* (2017.0) **13** S236-43. DOI: 10.4103/pm.pm_299_16
111. Wang H, Mohd Moklas MA, Vidyadaran S, Hidayat Baharuldin MT.. **Antidepressant-like effects of**. *Aust J Sci Technol* (2017.0) **1** 21-7
112. Solowey E, Lichtenstein M, Sallon S, Paavilainen H, Solowey E, Lorberboum-Galski H.. **Evaluating medicinal plants for anticancer activity**. *ScientificWorldJournal* (2014.0) **2014** 721402. DOI: 10.1155/2014/721402
113. Kusmardi K, Aryo T, Eka WP, Fadilah F, Pontjo PB, Wilzar F.. *Pharmacogn J* (2018.0) **10** 808-13. DOI: 10.5530/pj.2018.4.137
114. Akhir NAM, Chua LS, Majid FAA, Sarmidi MR.. **Cytotoxicity of aqueous and ethanolic extracts of**. *Br J Med Med Res* (2011.0) **1** 397-409. DOI: 10.9734/BJMMR/2011/507
115. Norrizah JS, Norizan A, Sharipah Ruzaina SA, Dzulsuhaimi D, Nurul Hidayah MS.. **Cytotoxicity activity and reproductive profiles of male rats treated with methanolic extracts of**. *Res J Med Plant* (2012.0) **6** 197-202. DOI: 10.3923/rjmp.2012.197.202
116. Hanafi MMM, Alqahtani OSO, Gui L, Yaakob H, Prieto JM.. **Phytochemical fingerprint and biological activities of three Malaysian**. *J Nat Prod Discovery* (2022.0) **1** 668. DOI: 10.24377/jnpd.article668
117. Suhaimi NA, Hashim N, Samsulrizal N.. *Malays Appl Biol* (2017.0) **46** 147-52
118. Manach C, Scalbert A, Morand C, Rémésy C, Jiménez L.. **Polyphenols: food sources and bioavailability**. *Am J Clin Nutr* (2004.0) **79** 727-47. DOI: 10.1093/ajcn/79.5.727
119. Rice-Evans CA, Miller NJ, Paganga G.. **Structure-antioxidant activity relationships of flavonoids and phenolic acids**. *Free Radic Biol Med* (1996.0) **20** 933-56. DOI: 10.1016/0891-5849(95)02227-9
120. Albrecht DS, Clubbs EA, Ferruzzi M, Bomser JA.. **Epigallocatechin-3-gallate (EGCG) inhibits PC-3 prostate cancer cell proliferation via MEK-independent ERK1/2 activation**. *Chem Biol Interact* (2008.0) **171** 89-95. DOI: 10.1016/j.cbi.2007.09.001
121. Zhou Y, Liu YE, Cao J, Zeng G, Shen C, Li Y. **Vitexins, nature-derived lignan compounds, induce apoptosis and suppress tumor growth**. *Clin Cancer Res* (2009.0) **15** 5161-9. DOI: 10.1158/1078-0432.CCR-09-0661
122. Soib HH, Yaakob H, Sarmidi MR, Mohamad Rosdi MN.. **Fractionation of aqueous extract of**. *Malaysian J Anal Sci* (2019.0) **23** 534-47. DOI: 10.17576/mjas-2019-2303-18
123. Vanella L, Di Giacomo C, Acquaviva R, Barbagallo I, Cardile V, Kim DH. **Apoptotic markers in a prostate cancer cell line: effect of ellagic acid**. *Oncol Rep* (2013.0) **30** 2804-10. DOI: 10.3892/or.2013.2757
124. Ahmad VN, Mohd Amin I.. **Anti-oral ulcer activity of**. *Pertanika J Soc Sci Humanit* (2017.0) 25
125. Oh MJ, Hamid MA, Ngadiran S, Seo YK, Sarmidi MR, Park CS.. *Arch Dermatol Res* (2011.0) **303** 161-70. DOI: 10.1007/s00403-010-1089-5
126. Shafaei A, Muslim NS, Nassar ZD, Aisha AF, Majid AMSA, Ismail Z.. **Antiangiogenic effect of Ficus deltoidea Jack standardised leaf extracts**. *Trop J Pharm Res* (2014.0) **13** 761-8. DOI: 10.4314/tjpr.v13i5.16
127. Khan AA, Omer KA, Talib A, Ahmad H, Javed MA, Sarmidi RM.. **Green tropical phytoextracts-promising anticancer alternative**. *Braz Arch Biol Technol* (2016.0) 59. DOI: 10.1590/1678-4324-2016160062
128. Muhammad H, Omar MH, Rasid ENI, Suhaimi SN, Mohkiar FH, Siu LM. **Phytochemical and in vitro genotoxicity studies of standardized**. *Plants (Basel)* (2021.0) **10** 343. DOI: 10.3390/plants10020343
129. Zaid SSM, Othman S, Kassim NM.. **Protective role of Mas Cotek (**. *Biomed Pharmacother* (2021.0) **140** 111757. DOI: 10.1016/j.biopha.2021.111757
130. Kalman DS, Schwartz HI, Feldman S, Krieger DR.. **Efficacy and safety of Elaeis guineensis and Ficus deltoidea leaf extracts in adults with pre-diabetes**. *Nutr J* (2013.0) **12** 36. DOI: 10.1186/1475-2891-12-36
131. Farsi E, Shafaei A, Hor SY, Ahamed MB, Yam MF, Asmawi MZ. **Genotoxicity and acute and subchronic toxicity studies of a standardized methanolic extract of**. *Clinics (Sao Paulo)* (2013.0) **68** 865-75. DOI: 10.6061/clinics/2013(06)23
132. Zaid SSM, Othman S, Kassim NM.. **Protective role of**. *J Ovarian Res* (2018.0) **11** 99. DOI: 10.1186/s13048-018-0466-0
133. Aliff MH, Nooraain H, Nurdiana S.. **Ameliorative effects of**. *AIP Conference Proceedings* (2018.0) **2030** 020271. DOI: 10.1063/1.5066912
134. Ahmed RK, Syarhabilahmad M, Nabil M, Al–Suede FSR.. **Acute and sub-chronic toxicity study of encap-sulation of combine plants extract of**. *J Angiothe* (2019.0) **3** 147-55
135. Nugroho RA, Aryani R, Manurung H, Rudianto R, Prahastika W, Juwita A. **Acute and subchronic toxicity study of the ethanol extracts from**. *Open Access Maced J Med Sci* (2020.0) **8** 76-83. DOI: 10.3889/oamjms.2020.3989
|
---
title: 'The Effectiveness of Zinc Supplementation in Taste Disorder Treatment: A Systematic
Review and Meta-Analysis of Randomized Controlled Trials'
authors:
- Boshra Mozaffar
- Arash Ardavani
- Hisham Muzafar
- Iskandar Idris
journal: Journal of Nutrition and Metabolism
year: 2023
pmcid: PMC10017214
doi: 10.1155/2023/6711071
license: CC BY 4.0
---
# The Effectiveness of Zinc Supplementation in Taste Disorder Treatment: A Systematic Review and Meta-Analysis of Randomized Controlled Trials
## Abstract
### Introduction
Food taste and flavour affect food choice and acceptance, which are essential to maintain good health and quality of life. Reduced circulating zinc levels have been shown to adversely affect the taste, but the efficacy of zinc supplementation to treat disorders of taste remains unclear. In this systematic review and meta-analysis, we aimed to examine the efficacy of zinc supplementation in the treatment of taste disorders.
### Methods
We searched four electronic bibliographical databases: Ovid MEDLINE, Ovid Embase, Ovid AMAD, and PubMed. Article bibliographies were also searched, which yielded additional relevant studies. There were no restrictions on the publication date to facilitate the collection and identification of all available and relevant articles published before 7 February 2021. We performed a systematic review and meta-analysis according to the PRISMA Statement. This review was registered at PROSPERO and given the identification number CRD42021228461.
### Results
In total, we included 12 randomized controlled trials with 938 subjects. The intervention includes zinc (sulfate, gluconate, picolinate, polaprezinc, and acetate), and the pooled results of the meta-analysis of subjects with idiopathic and zinc-deficient taste disorder indicate that improvements in taste disorder occurred more frequently in the experimental group compared to the control group (RR = 1.38; $95\%$ CI: 1.16, 1.64, $$p \leq 0.0002$$). Zinc supplementation appears to confer a greater improvement in taste perception amongst those with chronic renal disease using zinc acetate (overall RR = 26.69, $95\%$ CI = 5.52–129.06, $p \leq 0.0001$). The doses are equivalent to 17 mg–86.7 mg of elemental zinc for three to six months.
### Conclusion
Zinc supplementation is an effective treatment for taste disorders in patients with zinc deficiency, idiopathic taste disorders, and in patients with taste disorders induced by chronic renal failure when given in high doses ranging from 68 to 86.7 mg/d for up to six months.
## 1. Introduction
Food taste and flavour are important elements that affect food choice and acceptance [1]. Disorders of taste can adversely affect patients' health and quality of life [2], resulting in loss of food enjoyment, poor appetite, unintended weight loss, malnutrition, and other psychological and physiological complications [3–5]. Taste disorder is characterised by unpleasant tastes, where patients can experience hypogeusia (a condition of reduced ability to taste sweet, sour, bitter, salty, and umami tastes) or ageusia (a total loss of the ability to detect tastes) or dysgeusia (persistent foul, salty, rancid, or metallic taste sensation in the mouth) [6]. Around 200,000 patients visit doctors each year in the US complaining of a change in either taste or smell [1]. In 2003, about 240,000 patients were diagnosed with taste disorders in Japan [2]. A recent US survey using the Chemical Senses Questionnaire (CSQ) reported that the prevalence of taste alteration was $19\%$ in the adult population, with $5\%$ reporting dysgeusia, reaching $27\%$ in elderly populations [7]. More than half of patients ($56.9\%$) in Italy with COVID-19 have reported a reduction of taste and/or smell; a severe reduction of taste was present in $39.7\%$ of patients [8]. Taste alteration is also observed in $66\%$ of chemotherapy patients [9]. The most common causes of taste disorder are medications ($21.7\%$), followed by zinc deficiency ($14.5\%$), oral and perioral infections, Bell's palsy, oral appliances and age while less common causes include nutritional factors, tumours or lesions associated with taste pathways, head trauma, exposure to toxic chemicals and radiation treatment of the head and neck [10].
Zinc is an important element that supports many functions in humans including the immune system, growth, and development [11]. In addition, zinc is important for the functioning of taste buds [12]. Disturbance of salivary zinc levels has been found to be associated with a decreased level of gustin [13]. Gustin is the major zinc-containing protein in the human parotid saliva [12]; decreases in the secretion of gustin have been linked with abnormalities of the growth and development of the taste buds and the resultant loss of taste [14]. This mechanism is supported by numerous studies finding that patients with hypogeusia had low levels of gustin and salivary zinc [14–16] as well as significant alterations in the shape of taste buds [15]. The association between zinc deficiency and taste disorders has been well known for years [17–19], but evidence for efficacious treatment for taste disorders in clinical practice remains lacking. Although taste disorder has not been given sufficient attention by the medical community and researchers, in recent years, increased interest has emerged in evaluating potential treatments for disorders of taste due to the increasingly recognised adverse effects affecting taste due to bariatric surgery [20] and most recently due to COVID-19 infections [21]. We, therefore, aim to perform a systematic literature review and meta-analysis for available randomized controlled trials to investigate the efficacy of zinc supplementation in the treatment of taste disorders in the adult population.
## 2. Methods
We performed our systematic review and meta-analysis according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [22] to identify the effectiveness of zinc supplementation to prevent and treat taste disorder in patients who had been diagnosed with zinc deficiency, idiopathic taste disorder, or taste disorder secondary to chronic renal failure. Included and excluded studies were assessed based on outcomes, participants, intervention types, and study types.
## 2.1.1. Study Types
We only included randomized control trials; all other study designs were excluded.
## 2.1.2. Participants
All included participants consisted of human populations, and animal studies were excluded. Participant groups consisting of adults ≥18 years were included. We excluded patients who received chemotherapy and radiation, children, and pregnant women. We also excluded patients with taste disorders induced by drug use or taste disorders induced by the common cold.
## 2.1.3. Intervention
The participants received zinc-based therapy for the prevention and treatment of taste disorders compared to controls who received a placebo.
## 2.1.4. Outcomes
Improvement of taste disorder in response to zinc treatment was observed in intervention groups compared to the control group at the baseline and during a follow-up period. Zinc levels were also compared before and after treatment. Papers that did not include zinc or taste change outcomes were excluded.
## 2.2. Search Strategy
A literature search was conducted to describe the effects of zinc supplementation to improve subjective and objective symptoms of taste disorder induced by zinc deficiency, idiopathic conditions, or chronic renal failure. Two authors conducted the systematic search in the following electronic bibliographical databases: Ovid MEDLINE, Ovid Embase, Ovid AMAD, and PubMed. Article bibliographies were also searched and yielded additional relevant studies. There were no restrictions on publication date, facilitating the collection and identification of all available and relevant articles published before 7 February 2021. The following keywords were used: “taste change,” “taste disorder,” “taste dysfunction,” “dysgeusia,” “zinc,” “zinc sulphates,” and “deficiency.” The systematic review was registered at PROSPERO (https://www.crd.york.ac.uk) and given the identification number CRD42021228461.
PubMed search strategies are as follows. (“ taste disorders” (MeSH Terms) OR taste disorder (Text Word)) OR (“taste” (MeSH Terms) OR taste (Text Word)) AND change (All Fields)) OR (“taste” (MeSH Terms) OR taste (Text Word)) AND disfunction (All Fields)) OR (“dyspepsia” (MeSH Terms) OR dyspepsia (Text Word) AND (“zinc” (MeSH Terms) OR zinc (Text Word)) OR (“zinc” (MeSH Terms) OR zinc (Text Word)).
## 2.3. Data Extraction
We reviewed the articles according to the inclusion and exclusion criteria and summarised the main findings. Data regarding study duration, sample size, methods of detection of taste disorder, zinc dose, treatment period, and outcomes were extracted and are summarised in Tables 1 and 2. All the data those were utilised for the meta-analysis component were dichotomous data to find out the number of events in both the intervention and placebo groups. Additionally, all zinc supplement doses were considered for meta-analysis implementation.
## 2.4. Assessment of the Risk of Bias in Selected Studies
We used the Cochrane quality assessment tool to the assessed risk of bias for randomized controlled trials. The Cochrane tool, as described in the Handbook for Systematic Reviews of Interventions, evaluates the following attributes: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other forms of bias. Rating criteria include low risk of bias, high risk of bias, or unclear risk of bias [35]. The Cochrane risk-of-bias tool for randomized trials (RoB) was independently performed by two investigators (BM and HM).
## 2.5. Statistical Procedures
The meta-analysis was conducted using Review Manager 5. The Mantel–Haenszel (M–H) statistical method was selected with the random effect method for dichotomous data and established the outcome measure as a total and event based on Cochrane recommendation. All pooled results were reported as relative risk (RR) and $95\%$ confidence intervals (CI) for all individual studies, in addition to an effect size estimate (Z-statistic) and a measure of statistical significance ($p \leq 0.05$). To distinguish between the observed effects of zinc supplementation in iatrogenic or primary zinc deficiency versus chronic renal disease, two separate forest plots were generated for each. Further, data points from all studies at the synthesis stage were included, where data pertaining to event and total count, the equivalent quantity of elemental zinc, and the pharmaceutical name of the zinc supplement are stated. Finally, subanalysis was performed, based on the pharmaceutical name of the zinc supplement (s) included at the quantitative synthesis stage.
## 2.6. Assessment of Heterogeneity
We followed the Cochrane Handbook for Systematic Review of Interventions guidelines to assess the heterogeneity of the studies that were generated through the associated forest plots using Review Manager 5. Using the chi-squared test, we interpreted the heterogeneity according to I2 statistics: 75–$100\%$ indicates considerable heterogeneity, 50–$90\%$ represents substantial heterogeneity, 30–$60\%$ represents moderate heterogeneity, and 0–$40\%$ represents insignificant heterogeneity [35].
## 2.7. Summarizing and Interpreting Results
Review Manager 5 was used to conduct the meta-analysis, the risk-of-bias assessment, and the summary of the findings in Table 3 for each outcome included in this review. We imported the data to GRADEpro software to assess the evidence for each outcome. GRADE was also used to assess the quality of reported results in Table 4. We did not perform an analysis for publication bias via funnel plot as there were less than 10 studies included in the meta-analysis. This is because when there are fewer studies the power of the tests is too low to distinguish the chance from real asymmetry and in this study the largest forest plot only had seven data points across four studies.
## 3.1. Study Selection
A flow diagram of our literature search is shown in Figure 1. Following exclusions and removals, complete data extraction was performed on a total of 12 articles that met the inclusion criteria. Of these studies, four were included in a qualitative synthesis, and eight were included in a quantitative synthesis (meta-analysis) [36]. The characteristics of these 12 articles are shown in Table 1.
## 3.2.1. Trial Settings
Twelve randomized controlled trials (RCTs) are included in this review; all but one was written in English. One was in Japanese but was translated to English (Ikeda et al. [ 23]) The most common countries of origin of these studies were Japan and the US; one was from the UK, and one was from Germany. Out of 12 trials, 2 were crossover trials.
## 3.3. Study Populations
A total of 938 subjects were included in this study, all adults. The minimum age included in the trials was 18 years or older and the highest age observed was 84 years old; the lowest sample size was 22 and the highest sample size was 219. Eight studies included both genders in their trials; one study included only males and three trials did not report gender distribution.
Four studies were on idiopathic taste disorder, three concerned idiopathic and zinc-deficient taste disorder, and five were on renal failure-induced taste disorder.
## 3.4. Risk of Bias in Included Studies
Most studies were found to have an unclear risk of bias. However, four studies have a high risk of bias and three studies have a low risk of bias.
## 3.5.1. Idiopathic and Zinc-Deficient Taste Disorder
[1] Polaprezinc. First, we evaluated the efficacy of polaprezinc supplementation in idiopathic and zinc-deficient taste disorders. The efficacy of polaprezinc was examined in two studies, using different dosages. Sakagami et al. [ 24] introduced three different dosages to the intervention group: 75 mg, 150 mg, and 300 mg, which are equivalent to 17 mg, 34 mg, and 68 mg of elemental zinc. Despite the utilisation of identical doses (17 mg), Ikeda et al. [ 23] and Sakagami et al. [ 24] presented with differing outcomes (RR = 1.54, $95\%$ CI = 1.12–2.12, and RR = 0.81, $95\%$ = 0.51–1.27, respectively) (Figure 2). Nonetheless, across the Polaprezinc subgroup data points from Sakagami et al. [ 24], an increase in effect size is observed (Figure 2). Although an overall supplement-specific RR is positive (RR = 1.26, $95\%$ CI = 1.00–1.60), statistical significance was found to be borderline ($$p \leq 0.05$$) (Figure 2).
[2] Zinc Gluconate. Three trials studied the efficacy of zinc gluconate supplementation in idiopathic and zinc-deficient taste disorders. Yoshida et al. [ 29] administered 158 mg of zinc gluconate (equivalent to 22.59 mg/d of elemental zinc) for four months at a high risk of bias. Heckmann et al. [ 26] administered 140 mg (equivalent to 20 mg of elemental zinc) for three months at low risk of bias. An improvement in taste disorder was observed for the zinc supplement groups (RR 1.61, $95\%$ CI: 1.12–2.31, $$p \leq 0.01$$) among 102 participants (Figure 2).
Stewart-Knox et al. [ 25] administered zinc gluconate equivalent to 15 or 30 mg of elemental zinc per day over six months and were at high risk of bias. The study showed that zinc level increased postintervention in both groups and were greater in the 30 mg supplemented group; acuity for salt taste was greater in the 30 mg supplemented group ($$p \leq 0.031$$) while 15 and 30 mg Zn groups did not improve any tastes acuity. However, we could not conduct a meta-analysis of the results because the study did not report the number of events in the placebo group.
[3] Zinc Picolinate. Of the studies included, only one [28] was found to examine the efficacy of zinc picolinate on taste disorder patients at a high risk of bias. An improvement in taste disorder at a dosage of 28.9 mg three times/d for three months (RR 1.70, $95\%$ CI: 1.13–2.56, $$p \leq 0.01$$) (Figure 2), with 73 participants.
[4] Zinc Sulphate. In 1976, Henkin et al. [ 34] examined the effectiveness of four doses of 100 mg of zinc ion, with an unclear risk of bias. The results from this study indicated that both placebo and treatments groups with zinc sulfate showed equivalent improvements. We excluded this study from the meta-analysis because number of events in both the intervention and placebo groups was unclear.
## 3.5.2. Zinc Disorder Secondary to Chronic Renal Failure
[1] Zinc Acetate. Zinc acetate was used as a treatment for taste disorder induced by chronic renal failure in three studies [30–32]. Each study provided a single data point each, with the overall RR for zinc acetate found to be 26.69 ($95\%$ CI = 5.52–129.06, $p \leq 0.0001$) (Figure 3). The total number of participants in the three studies was 77 patients. A heterogeneity assessment was inconclusive (I2 = $0\%$, $$p \leq 0.98$$) (Figure 3).
[2] Zinc Sulphate. Two studies, Atkin-Thor et al. [ 33] and Matson et al. [ 27], examined the efficacy of zinc sulfate in taste disorder induced by chronic renal failure for up to a six-week intervention period. In a double-blind crossover trial, Atkin-Thor et al. [ 33] introduced 440 mg of zinc sulfate three times per week at a high risk of bias, the results of this study showed a significant improvement in taste acuity in the supplemented group. Whereas Matson introduced 220 mg of zinc sulphate per day at an unclear risk of bias, the results from this study showed no improvements in both the intervention and placebo groups. These two trials did not provide sufficient details about the placebo groups. We have therefore excluded them from the meta-analysis.
## 4. Discussion
This systematic review assessed the efficacy of zinc supplementation to improve taste disorders. We focused on the outcomes of intervention groups compared to placebo among patients with zinc deficiency and idiopathic taste disorder or taste disorder induced by chronic renal failure. We included 12 randomized controlled trials: four were included in a qualitative synthesis and eight in a meta-analysis. We assessed five studies as having an unclear risk of bias [23, 24, 27, 32, 34], four studies at a high risk of bias [25, 28, 29, 33], and three studies at low risk of bias [26, 30, 31]. Seven included studies examined the effectiveness of different zinc supplementations (polapre zinc, picolinate, zinc gluconate, and zinc sulphate) among patients with zinc deficiency and idiopathic taste disorder. We did not include two studies such as the study by Henkin et al. [ 34] and Stewart-Knox et al. [ 25] in the meta-analysis because of their unclear methodologies and unreported data for the placebo groups. Out of seven studies that examined the efficacy of zinc supplementation in taste disorders induced by chronic renal failure, we did not include Atkin-Thor et al. [ 33] and Matson et al. [ 27] in the meta-analysis because they did not report data about the placebo groups.
## 4.1. Summary of Main Results
The pooled results of this meta-analysis indicated that improvement in taste disorder occurred significantly more frequently in the intervention group compared to the control group. There was a significant improvement in taste following zinc supplementation at the study level except in three studies [24, 26, 29]. The improvement in taste following zinc supplementation was observed at the meta-analysis level. We found that zinc supplements reduced the risk of taste disorder by $51\%$. Moreover, the pooled results of the largest studies [23, 24, 28] indicated that zinc supplementation is an effective treatment for taste disorders in patients with zinc deficiency or idiopathic taste disorders when given in high doses ranging from 68 to 86.7 mg/d for up to three months. This results in agreement with Yagi et al. 's [37] review which indicated that zinc supplementation contributes to the treatment of taste disorders caused by zinc deficiency. In contrast, Kumbargere Nagraj et al. [ 38] did not find sufficient trials to support the effectiveness of zinc in taste disorder improvement. The level of included studies ranged from moderate to high using The Grading of Recommendations Assessment, Development and Evaluation (GRADE). Heckmann et al. [ 26] and Yoshida et al. [ 29] introduced a low dose of elemental zinc, around 20–22.59 mg/d, for up to three to four months to patients with taste disorders induced by zinc deficiency or idiopathic disease and our meta-analysis showed insignificant improvement of taste disorders, however, the results for these two trials should be viewed with caution due the quality of evidence was rated as low, and high risk of bias for one study Yoshida et al. [ 29].
In the three studies concerning taste disorder induced by chronic renal failure, we found the level of evidence and its quality to be low. This was driven by the fact that the studies mainly had small sample size and the absence of event numbers in the placebo group, which resulted in a high upper limit of the CI [30–32] in the meta-analysis. Overall, per the available data, zinc supplementation appears to confer a greater improvement in taste perception amongst those with chronic renal disease using zinc acetate (overall RR = 26.69, $95\%$ CI = 5.52–129.06, $p \leq 0.0001$) (Figure 3) in comparison to the extent of improvement using alternative supplements in the iatrogenic or zinc deficiency disease groups (Figure 2). Unfortunately, a direct comparison in the response to zinc acetate between the chronic renal disease and iatrogenic or zinc deficiency cohorts was not possible due to missing data. Furthermore, zinc picolinate was represented by a single data point [28]. In all studies included in this meta-analysis, we did not find considerable statistical heterogeneity. Nevertheless, there is substantial heterogeneity based on elemental zinc-equivalent dose, supplement chemical structure, follow-up time, and disease state exists, as inferred based on the study characteristics as we aimed to collect all available RCTs to examine the effectiveness of zinc supplementation in taste disorder treatment. We suggest that zinc supplementation may improve specific tastes more than others depending on the case or the disease-induced taste disorder. We suggest a high dose of elemental zinc 68–86.7 mg/d for up to six months to improve taste disorders. However, the results of this meta-analysis should be interpreted with caution as excessive zinc supplementation might have serious health outcomes and toxicity when taken at a significantly higher than the Recommended Dietary Allowance (RDA) (100–300 mg/day vs. 15 mg daily). It has been proposed that even smaller doses of zinc supplementation, closer to the RDA, interfere with the utilisation of copper and iron and negatively impact HDL cholesterol levels. Zinc supplement users should be informed of any potential risks associated with its usage [39].
## 4.2. Strengths and Limitations of This Study
Unlike other reviews in this area, our systematic review provided additional evidence and clarification of zinc supplementation's efficacy in improving taste disorder in adult populations by stratifying according to zinc dose, formulation type, and treatment duration. However, one aspect that can limit the analysis and discussion of the results is the heterogeneity of the methods used. The studies assessed combined objective outcomes (e.g., filter paper disk; detection and recognition thresholds for sweet, sour, salty, bitter, and umami tastes) and subjective outcomes (e.g., questionnaires results). However, whether the difference between subjective and objective methods could significantly affect the results of improvement is unclear. In another review, the author examined the overall improvement in taste acuity using both subjective and objective methods; however, the author could not conclude the overall effect because of the very low level of evidence. High-quality research is required to compare different objective and subjective methods [38]. We observed that some studies detected taste improvement in only one type of taste, so a further limitation of our meta-analysis is that we defined “improvement” as an improvement of any of the five basic tastes: sweet, sour, bitter, salty, and umami tastes.
## 5. Conclusion
High-dose zinc supplementation is an effective treatment for taste disorders in patients with zinc deficiency or idiopathic taste disorder and in patients with taste disorders induced by chronic renal failure.
## Data Availability
The data used to support the findings of this study are available from the corresponding author upon reasonable request.
## Disclosure
The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. A previous version of this study has been submitted as an abstract as a conference paper https://publications.waset.org/abstracts/148658/pdf.
## Conflicts of Interest
The authors declare that there are no conflicts of interest.
## Authors' Contributions
BM is the first author for this study and undertook data collection, analysis, and wrote the first draft. BM and II conceptualized the study; BM and II lead and designed the study; BM and II performed logistical planning, allocation, and implementation; BM and HM performed literature search; BM and HM performed data extraction; BM and HM performed risk of bias assessment; BM performed initial statistics execution, BM composed the manuscript; AA reviewed statistical approach and executed the study; BM, AA, and II preliminarily reviewed the draft; and II reviewed and revised the study and approved the final manuscript for submission.
## References
1. Clark J. E.. **Taste and flavour: their importance in food choice and acceptance**. (1998) **57** 639-643. DOI: 10.1079/pns19980093
2. Alvarez-Camacho M., Gonella S., Ghosh S.. **The impact of taste and smell alterations on quality of life in head and neck cancer patients**. (2016) **25** 1495-1504. DOI: 10.1007/s11136-015-1185-2
3. Malaty J., Malaty I. A. C.. **Smell and taste disorders in primary care**. (2013) **88** 852-859. PMID: 24364550
4. Hummel T., Basile N. L., Karl-Bernd H.. **The impact of smell and taste disorders**. (2018) **10**. DOI: 10.3205/cto000077
5. Hur K., Choi J. S., Zheng M., Shen J., Wrobel B.. **Association of alterations in smell and taste with depression in older adults**. (2018) **3** 94-99. DOI: 10.1002/lio2.142
6. **Taste disorders**. (2021)
7. Rawal S., Hoffman H. J., Bainbridge K. E., Huedo-Medina T. B., Duffy V. B.. **Prevalence and risk factors of self-reported smell and taste alterations: results from the 2011–2012 US national health and nutrition examination survey (NHANES)**. (2016) **41** 69-76. DOI: 10.1093/chemse/bjv057
8. Mercante G., Ferreli F., De Virgilio A.. **Prevalence of taste and smell dysfunction in coronavirus disease 2019**. (2020) **146** 723-726. DOI: 10.1001/jamaoto.2020.1155
9. Campagna S., Gonella S., Sperlinga R.. **Prevalence, severity, and self-reported characteristics of taste alterations in patients receiving chemotherapy**. (2018) **45** 342-353. DOI: 10.1188/18.onf.342-353
10. Imoscopi A., Inelmen E. M., Sergi G., Miotto F., Manzato E.. **Taste loss in the elderly: epidemiology, causes and consequences**. (2012) **24** 570-579. DOI: 10.3275/8520
11. Roohani N., Hurrell R., Kelishadi R., Schulin R.. **Zinc and its importance for human health: an integrative review**. (2013) **18** 144-157. PMID: 23914218
12. Henkin R. I.. **Zinc in taste function: a critical review**. (1984) **6** 263-280. DOI: 10.1007/bf02917511
13. Brennan F., Stevenson J., Brown M.. **The pathophysiology and management of taste changes in chronic kidney disease: a review**. (2020) **30** 368-379. DOI: 10.1053/j.jrn.2019.11.004
14. Henkin R. I., Martin B. M., Agarwal R. P.. **Decreased parotid saliva gustin/carbonic anhydrase VI secretion: an enzyme disorder manifested by gustatory and olfactory dysfunction**. (1999) **318** 380-391. DOI: 10.1097/00000441-199912000-00005
15. Shatzman A. R., Henkin R. I.. **Gustin concentration changes relative to salivary zinc and taste in humans**. (1981) **78** 3867-3871. DOI: 10.1073/pnas.78.6.3867
16. Zhu Y., Feron G., Von Koskull D., Neiers F., Brignot H., Hummel T.. **The association between changes of gustatory function and changes of salivary parameters: a pilot study**. (2021) **46** 538-545. DOI: 10.1111/coa.13705
17. Ikeda M. I., Ikui A., Komiyama A., Kobayashi D., Tanaka M.. **Causative factors of taste disorders in the elderly, and therapeutic effects of zinc**. (2008) **122** 155-160. DOI: 10.1017/s0022215107008833
18. Matsugasumi M., Hashimoto Y., Okada H.. **The association between taste impairment and serum zinc concentration in adult patients with type 2 diabetes**. (2018) **42** 520-524. DOI: 10.1016/j.jcjd.2018.01.002
19. Yagi T.. **The role of zinc in the treatment of taste disorders**. (2013) **5** 44-51
20. Ahmed K., Penney N., Darzi A., Purkayastha S.. **Taste changes after bariatric surgery: a systematic review**. (2018) **28** 3321-3332. DOI: 10.1007/s11695-018-3420-8
21. Cazzolla A. P., Lovero R., Lo Muzio L.. **Taste and smell disorders in COVID-19 patients: role of interleukin-6**. (2020) **11** 2774-2781. DOI: 10.1021/acschemneuro.0c00447
22. Page Mj M. J., Bossuyt P. M., Boutron I., Hoffmann T. C., Mulrow C. D.. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews**. (2021) **88**. DOI: 10.1016/j.ijsu.2021.105906
23. Ikeda M., Kurono Y., Inokuchi A.. **The effect of zinc agent in 219 patients with zinc deficiency-inductive/idiopathic taste disorder: a placebo controlled randomized study**. (2013) **116** 17-26. DOI: 10.3950/jibiinkoka.116.17
24. Sakagami M. I., Ikeda M., Tomita H.. **A zinc-containing compound, Polaprezinc, is effective for patients with taste disorders: randomized, double-blind, placebo-controlled, multi-center study**. (2009) **129** 1115-1120. DOI: 10.1080/00016480802552550
25. Stewart-Knox B. J. S., Simpson E. E., Parr H.. **Taste acuity in response to zinc supplementation in older Europeans**. (2008) **99** 129-136. DOI: 10.1017/s0007114507781485
26. Heckmann S. M., Hujoel P., Habiger S.. **Zinc gluconate in the treatment of dysgeusia--a randomized clinical trial**. (2005) **84** 35-38. DOI: 10.1177/154405910508400105
27. Matson A. W., Wright M., Oliver A.. **Zinc supplementation at conventional doses does not improve the disturbance of taste perception in hemodialysis patients**. (2003) **13** 224-228. DOI: 10.1016/s1051-2276(03)00072-4
28. Sakai F. Y., Yoshida S., Endo S., Tomita H.. **Double-blind, placebo-controlled trial of zinc picolinate for taste disorders**. (2002) **122** 129-133. DOI: 10.1080/00016480260046517
29. Yoshida S. E., Endo S., Tomita H.. **A double-blind study of the therapeutic efficacy of zinc gluconate on taste disorder**. (1991) **18** 153-161. DOI: 10.1016/s0385-8146(12)80219-7
30. Mahajan S. K., Prasad A. S., Rabbani P., Briggs W. A., McDonald F. D.. **Zinc deficiency: a reversible complication of uremia**. (1982) **36** 1177-1183. DOI: 10.1093/ajcn/36.6.1177
31. Mahajan S. K., Prasad A. S., Lambujon J., Abbasi A. A., Briggs W. A., McDonald F. D.. **Improvement of uremic hypogeusia by zinc: a double-blind study**. (1980) **33** 1517-1521. DOI: 10.1093/ajcn/33.7.1517
32. Mahajan S. K., Prasad A. S., Lambujon J., Abbasi A. A., Briggs W. A., McDonald F. D.. **Improvement of uremic hypogeusia by zinc**. (1979) **25** 443-448. DOI: 10.1097/00002480-197902500-00085
33. Atkin-Thor E., Goddard B. W., O’Nion J., Stephen R. L., Kolff W. J.. **Hypogeusia and zinc depletion in chronic dialysis patients**. (1978) **31** 1948-1951. DOI: 10.1093/ajcn/31.10.1948
34. Henkin R. I., Schecter P. J., Friedewald W. T., Demets D. L., Raff M.. **A double blind study of the effects of zinc sulfate on taste and smell dysfunction**. (1976) **272** 285-299. DOI: 10.1097/00000441-197611000-00006
35. Higgins T. J., Chandler J., Cumpston M., Li T., Page M. J., Welch V. A.. (2019)
36. Page M. J., McKenzie J. E., Bossuyt P. M.. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews**. (2021) **88**. DOI: 10.1016/j.ijsu.2021.105906
37. Yagi T., Asakawa A., Ueda H., Ikeda S., Miyawaki S., Inui A.. **The role of zinc in the treatment of taste disorders**. (2013) **5** 44-51. DOI: 10.2174/2212798411305010007
38. Kumbargere Nagraj S., George R. P., Shetty N., Levenson D., Ferraiolo D. M., Shrestha A.. **Interventions for managing taste disturbances**. (2017) **12**. DOI: 10.1002/14651858.CD010470.pub3
39. Fosmire G. J.. **Zinc toxicity**. (1990) **51** 225-257. DOI: 10.1093/ajcn/51.2.225
|
---
title: Prognosis and Characterization of Microenvironment in Cervical Cancer Influenced
by Fatty Acid Metabolism-Related Genes
authors:
- Yanjun Zhou
- Jiahao Zhu
- Mengxuan Gu
- Ke Gu
journal: Journal of Oncology
year: 2023
pmcid: PMC10017219
doi: 10.1155/2023/6851036
license: CC BY 4.0
---
# Prognosis and Characterization of Microenvironment in Cervical Cancer Influenced by Fatty Acid Metabolism-Related Genes
## Abstract
Increasing evidence suggests that diverse activation patterns of metabolic signalling pathways may lead to molecular diversity of cervical cancer (CC). But rare research focuses on the alternation of fatty acid metabolism (FAM) in CC. Therefore, we constructed and compared models based on the expression of FAM-related genes from the Cancer Genome Atlas by different machine learning algorithms. The most reliable model was built with 14 significant genes by LASSO-Cox regression, and the CC cohort was divided into low-/high-risk groups by the median of risk score. Then, a feasible nomogram was established and validated by C-index, calibration curve, net benefit, and decision curve analysis. Furthermore, the hub genes among differential expression genes were identified and the post-transcriptional and translational regulation networks were characterized. Moreover, the somatic mutation and copy number variation landscapes were depicted. Importantly, the specific mutation drivers and signatures of the FAM phenotypes were excavated. As a result, the high-risk samples were featured by activated de novo fatty acid synthesis, epithelial to mesenchymal transition, angiogenesis, and chronic inflammation response, which might be caused by mutations of oncogenic driver genes in RTK/RAS, PI3K, and NOTCH signalling pathways. Besides the hyperactivity of cytidine deaminase and deficiency of mismatch repair, the mutations of POLE might be partially responsible for the mutations in the high-risk group. Next, the antigenome including the neoantigen and cancer germline antigens was estimated. The decreasing expression of a series of cancer germline antigens was identified to be related to reduction of CD8 T cell infiltration in the high-risk group. Then, the comprehensive evaluation of connotations between the tumour microenvironment and FAM phenotypes demonstrated that the increasing risk score was related to the suppressive immune microenvironment. Finally, the prediction of therapy targets revealed that the patients with high risk might be sensitive to the RAF inhibitor AZ628. Our findings provide a novel insight for personalized treatment in CC.
## 1. Introduction
Even with the implementation of HPV vaccination and screening programs, cervical cancer (CC) remains a major public health problem among women in high-development index countries and poverty areas [1]. CC patients often progress into the advanced stage, and recurrence leads to a poor prognosis [2]. How to identify the CC patients with high risk at the time of diagnosis still needs to be addressed in clinical practice.
Fatty acids (FAs) serve as important components of the membrane structure, secondary messengers, and fuels of energy production in cells [3]. To keep rapid and uncontrolled growth, the cancer cells consume a huge amount of nutrients, such as FAs and glucose, while excreting wastes which lead to a nutrient-deficient, acidic, and hypoxic tumour microenvironment (TME) [4]. Such hostile TME impairs the normal metabolic requirements of other intertumoural cells [5]. Among the genetically driven metabolic reprogramming of cancer cells, the FA metabolism (FAM) has been demonstrated to influence the growth and metastasis of tumour cells and modulate the recruitment and differentiation of tumour infiltrating cells in the TME [6]. Memory T cells fail to develop without FAs in culture [7]. Dendritic cells accumulating large amounts of lipids have been found to lose their antigen-presenting function in a variety of cancers [8]. Cancer-associated fibroblasts in the TME enhance FAO to boost colorectal cancer metastasis resulting in a poor prognosis [9]. The enhanced lipolysis and de novo FA synthesis also lead to lymphangiogenesis with endothelial cells [10]. However, rare research studies focus on the correlation between the prognosis of CC and FAM, which needs to be elucidated. Furthermore, the FAM can be regulated by oncogenic signalling pathway directly, namely, growth-factor receptor tyrosine kinases (RTKs)/RAS [11], phosphoinositide 3-kinase/protein kinase B (PI3K/AKT), and the mitogen-activated protein kinase (MAPK) signalling pathway [12]. However, the mechanisms driving particular FAM phenotypes in CC are still unclear. Moreover, increasing pieces of evidences indicate that the FAs as pivotal mediators can rewrite the TME and enhance cancer immune evasion and spread [5]. How FAM phenotypes affect the infiltration of immune and stromal cells is also unknown in the TME of CC.
To clarify the questions mentioned above, our study aims to construct a reliable and feasible prognostic model according to the FAM to stratify the CC patients. Furthermore, the specific mutation drivers of the FAM phenotypes and the connotations between the TME landscape and FAM phenotypes were evaluated systematically. Finally, due to the essential role of FAM reprogramming in cancer progression, potential therapy targets were predicted in CC patients which may provide a novel insight for personalized treatment.
## 2.1. Data Resource and Collection of FAM-Associated Genes
The transcriptome data and clinicopathologic information were downloaded from the Cancer Genome Atlas (TCGA) database (https://tcga-data.nci.nih.gov/tcga/ and https://portal.gdc.cancer.gov/), and mRNA expression was extracted from TCGA RNA-seq data for 306 CCs and 3 surrounding non-cancer tissues [13]. *The* genes were annotated by gencode.gene.info.v22. After removing patients without detailed clinicopathologic and overall survival (OS) information, we obtained 274 patients with CC in TCGA. The expression profiles of GSE44001 [14], which contained 300 early CC cases with disease-free survival (DFS) information, were obtained from the GEO (https://www.ncbi.nlm.nih.gov/geo/) database. The intersection was made from the genes related with FAM from GeneCards (https://www.genecards.org) and gene sets concerning FAM from the Molecular Signature Database (MSigDB) v7.4. Finally, 309 FAMs were selected (Supplementary Table 1).
## 2.2. Construction and Validation of FAM-Relevant Prognostic Signature
Randomly drawn $55\%$ of samples (151 samples) were used for model training, and the remaining $45\%$ (123 samples) were used for validation in the following analysis. The least absolute shrinkage and selection operator (LASSO) Cox regression analysis was employed in the training set to build a prognostic model for OS, using the R package “glmnet” [15]. LASSO-Cox regression analysis primarily selected useful predictive features to reduce the model complexity and multicollinearity and avoided overfitting to some extent. The proportional hazards (PH) assumption was conducted on the FAM-relevant prognostic genes by the R package “survival” and “survminer.” According to the prognostic model, the risk score was exported for each CC patient:[1]risk scoreRS=∑inExpi∗Coefi.
Expi means the expression level of each FAM gene, and Coefi stands for the corresponding regression coefficient. To make the prognostic model as concise as possible, the patients were divided into high-risk and low-risk groups by the median of RS. The risk curve was plotted according to the RS and risk group, and the survival status and RS were evaluated according to the curve.
To evaluate the feasibility of the prognostic model in predicting survival in CC patients, we conducted a Kaplan–*Meier analysis* of overall survival (OS) by the R package “survival,” operating characteristic curve (ROC), and area under the curve (AUC) by the means of the R package “timeROC” in both the training dataset and testing dataset. Kaplan–Meier survival curves were plotted and p values were calculated using the log-rank test to explore the survival difference between risk groups [16]. The AUC ranges from 0 to 1. When AUC lies between 0.5 and 0.6, 0.6 and 0.7, or is >0.7, the performance of the model is considered poor, fair, or good, respectively.
## 2.3. Establishment and Validation of a Nomogram
The PH assumption was conducted on the RS and risk grouping by the R package “survival” and “survminer.” To compare the predictive value of risk grouping or RS in survival analysis with traditional clinical-pathologic parameters, the univariate Cox regression and multivariate Cox regression were employed to calculate hazard ratios (HRs) and $95\%$ confidence intervals (CIs) by the R package “survival” and “survminer.” To further improve the predictive accuracy of our FAM gene signature by combining it with other clinical-pathologic features (e.g., body mass index (BMI), pathology type, pathology grade, and stage), we built an easy-to-use and clinically adaptable risk nomogram for predicting the OS probability in CC patients using the “rms” package in R [17]. The OS probabilities were predicted for 1-, 2-, 3-, and 5-year survival.
The validation of the nomogram-based prediction model was accessed via bootstrapped calibration curves using “rms” in R and quantified as a Harrell's concordance index (C-index) by the function “rcorrcens” in R package “Hmisc.” C-index was utilized to evaluate the discriminative capabilities of the nomograms. Calibration curves (1000 bootstrap resamples) were generated to compare the consistency between the predicted and observed OS for 1, 2, 3, and 5 years [18]. The net reclassification index (NRI) was employed to evaluate the added value of new risk group or RS to existing prognostic models. Decision curve analysis (DCA) was applied to evaluate the impact on decision making in clinical practice of the nomograms using the “stdca” function in R [19].
## 2.4. Comparison of Models Built by Other Machine Learning Methods
Furthermore, to choose the best prognostic model, the support vector machine (SVM) and the random forest method were performed to classify the vital status in the CC cohort using “e1071” and “randomForest” packages in R. The 274 CC cohort was randomly divided into a training cohort and a testing cohort as described before. The Wilcoxon test assessed the performance of the SVM model and random forest model. The discriminatory power of the SVM model and random forest model on vital status was assessed by the AUC in training and testing datasets. To obtain the best SVM model, the AUC and prediction accuracy of the linear, polynomial, radial, and sigmoid models were compared. The variables were derived from the best polynomial model. Then, to further compare the discriminative ability between the FAM genes and our signature genes, principal component analysis (PCA) was carried out using the “pca” function.
## 2.5. Identification of Differential Expression Genes and Functional Enrichment Analysis
Differential expression genes (DEGs) for low/high-risk groups were calculated by the R “limma” package. The threshold (adjusted p value <0.05 and |Log2 fold change (FC)| > 1.2) was used as a selection criterion for the DEGs. A volcano plot and a heat map of the DEGs were depictured. Gene ontology (GO) enrichment [20] and gene set enrichment analysis (GSEA) [21] were employed to decipher the enrichment of signalling transductions and biological functions in the DEGs in CC patients using the functions of gseGO and gseKEGG in “GSEA” package. The enrichments according to the MSigDB and Reactome were analysed. Then, gene set variation analysis (GSVA) was further carried out by the “GSVA” in R [22]. *The* gene sets of “h.hallmark.v7.4.symbols.gmt” (HALLMARK) and “c2.cp.kegg.v7.4.symbols.gmt” (KEGG) were used as the reference molecular signature databases, and adjusted p value <0.05 and |Log2 (FC)| > 0.1) were considered statistically significant.
## 2.6. Identification of Hub Genes and Regulation Network
To obtain the hub genes among the DEGs, the “GOSemSim” package in R was employed [23]. Meanwhile, the likelihood of protein-protein interactions (PPIs) among the DEGs was identified in our study from STRING database, which is based on either literature of direct interaction experiments or prediction from co-expression and gene arrangement in the genome. Moreover, the list of 318 transcription factors was acquired from https://www.cistrome.org/. The correlation between the DEGs and transcription factors was defined as the correlation coefficient (R) = 0.5, p value = 0.001. Additionally, the miRNAs that interact with the DEGs, validated by luciferase reporter assay, were obtained by R packages “multiMiR” and “mirtarbase” [24]. The long non-coding RNAs (lncRNAs) interacting with miRNAs were obtained from “starbase” [25]. The correlation among the DEGs, miRNAs, and lncRNAs was illustrated by “ggalluvial” package.
## 2.7. Somatic Genomic Alternation Analysis
To identify the gene mutation characteristics in CC patients, we analysed somatic mutation data by the R package “maftools” [26]. The summary oncoplots were based on MutSigCV algorithm by maftools. The mutation pattern of specific genes was represented by oncoplot function in maftools. Transitions and transversions were calculated using the titv function in maftools. The changes in the amino acid of a certain protein were depictured by the lollipopPlot in maftools. The tumour mutational burden (TMB) values were calculated in units of mutations per megabase (MB) and characterized as low (TMB < 6), intermediate (6 ≤ TMB < 20), or high (TMB ≥ 20) [27].
## 2.8. Identification of Mutation Driver and Affected Signalling Pathway
The function oncodrive in maftools [26] was employed to identify driver genes, based on OncodriveCLUST algorithm [28]. The OncogenicPathways function was used to check the enrichment of oncogenic signalling pathways. The effects of a specific gene mutation on OS were manifested by mafSurvival in maftools [26]. The comparison of the two risk groups to detect differentially mutated genes was achieved by mafCompare in maftools and then the result was visualized by forestPlot in maftools. The drug-gene interactions were checked by the drugInteractions in maftools.
## 2.9. De Novo Mutational Signature Analysis and APOBEC Enrichment Estimation
The signature analysis was performed by a series of functions in maftools. The mutational matrix was first decomposed into signatures by negative matrix factorization. The extracted signatures then were compared against the Catalogue of Somatic Mutations in Cancer (COSMIC) mutational signatures v2 and updated v3 [29]. The different mutation patterns between apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like (APOBEC) enriched and non-APOBEC enriched samples were achieved by the function plotApobecDiff of maftools.
## 2.10. Copy Number Variation Analysis
Because the copy number variations (CNVs) can contribute to cancer susceptibility, we further detected the common CNV regions by the GenePattern website (https://cloud.genepattern.org/gp/pages/index.jsf) with corresponding GISTIC 2.0 module [30].
## 2.11. Identification of Neoantigens
We sought to explore neoantigen in the both groups of CC patients and responsiveness to therapies. The neoantigen data from the CC cohort were downloaded from https://biopharm.zju.edu.cn/tsnadb [31] and https://tcia.at/home [32]. The neoantigen burden of a certain patient was predicted bioinformatically, as following standards. The half maximal inhibitory concentration (IC50) < 500 nM was considered a predicted binder. Patient-specific neoantigens were defined as any unique combination of peptide sequence: human leukocyte antigen (HLA)-allele with mutant peptide-binding affinity IC50 < 500 nM, and corresponding wild-type peptide IC50 > 500 nM. Expressed neoantigens were defined as neoantigens with RNA-sequencing counts ≥1 [33].
## 2.12. Correlation between Cancer Germline Antigens and Immune Cell Infiltration
To chart the antigenome for each sample, we used RNA-sequencing data to derive expression levels of cancer germline antigens (CGAs). Due to the low tumoural specificity of CGAs, we used the CGA gene list retrieved from https://tcia.at [32]. The expression levels of CGA genes were compared between both of the groups, and we obtained 37 differentially expressed CGAs according to the risk levels. Furthermore, we explored the relationship between the 37 genes and CD8 T and regulatory T cell enrichment by the “corrplot” in R.
## 2.13. Evaluation of the Cellular Composition in Tumour Microenvironment
To provide a comprehensive view of the cellular composition of the intratumoural immune infiltrates, we carried out the immunogenomic characterization of the CC patients by the “IOBR” package [34], which includes 8 algorithms to estimate the immune infiltrating cells. To further identify the significantly enriched cells in the TME, the correlations between the RS and each type of infiltrating cells were calculated by “corrplot” in R and the importance of each infiltrating cells in survival was calculated as log10 transformed p value by Cox regression. The infiltrating cell types with a p value of correlation under 0.01 were selected and depictured. Potential implications for immunotherapy were calculated by the website https://tide.dfci.harvard.edu/ [35].
## 2.14. Exploration of Potential Therapeutic Drugs concerning Prognostic Models
To explore potential clinical drugs for the treatment of high-risk CC patients, we used the R package “pRRophetic” to predict the sensitivity to the compounds obtained from the Genomics of Drug Sensitivity in Cancer (GDSC) website according to the CC dataset in TCGA database [36].
## 2.15. Statistical Analysis
Continuous variables were compared by the Wilcox test, while categorized variables were compared by ANOVA. All the analyses were performed by R software (Version 4.1.3, the R foundation for statistical computing). P values lower than 0.05 were considered to be significant unless special instruction was given.
## 3.1. Construction and Validation of FAM-Relevant Prognostic Model
We utilized the LASSO-penalized Cox regression to determine the LASSO tuning parameter λ, resulting in the minimum squared error. The results showed that when specific 14 genes were included in the prognostic model, the model contraction was stable, the partial likelihood deviance was minimal, and the optimal λ was 0.01961 (Figures 1(a) and 1(b)). Finally, 14-gene signature based on FAM, including CD1d molecule (CD1D), carboxyl ester lipase (CEL), non-SMC condensin II complex subunit H2 (NCAPH2), succinate dehydrogenase complex subunit D (SDHD), alcohol dehydrogenase class II Pi chain (ADH4), holocytochrome C synthase (HCCS), thyroid hormone-responsive (THRSP), glutaryl-CoA dehydrogenase (GCDH), nudix hydrolase 7 (NUDT7), dipeptidase 2 (DPEP2), serine incorporator 1 (SERINC1), macrophage migration inhibitory factor (MIF), ELOVL fatty acid elongase 7 (ELOVL7), and cytochrome P450 family 1 subfamily A member 1 (CYP1A1), was identified to construct the prognostic model. The coefficient of each gene is summarized in Supplementary Table 2. The results of the PH assumption of each gene are listed in Supplementary Table 3 and Supplementary Figures 3a–3n and 3q–3t. The CC cohort was divided into a low-risk group and a high-risk group by the median of RS (Figure 1(d)). The expression levels of these 14 genes were represented in the different risk groups (Figure 1(c)). To verify the utility of our prognostic model, the associations among vital status, time, and RS of each group were determined. With the increasing RS, the death events tended to increase in CC patients (Figure 1(e)). Furthermore, the results showed that the patients in the high-risk group had a worse OS than those in the low-risk group in both the training cohort and testing cohort ($p \leq 0.001$ and $$p \leq 0.035$$, respectively, Figures 1(f) and 1(g)). Furthermore, ROC was employed to confirm the predictive value of the model. We observed that the AUC values were 0.891, 0.851, and 0.870 at 1 year, 3 years, and 5 years, respectively, which may suggest that the performance of the model was good at all three time points in the training dataset (Figure 1(i)). In the testing dataset, the performance of the model decreased a bit at 1 year to a fair level (AUC: 0.674), but it came back to a good level at 3 years and 5 years (AUC: 0.724 and 0.730, Figure 1(j)). As there is no available CC dataset with OS, we utilized the early CC dataset GSE44001 with DFS to validate the predictive validity of our model. We revealed a trend that the patients had shorter DFS in the high-risk group than those in the low-risk group ($$p \leq 0.061$$, Figures 1(h)). In the early CC cohort, the AUC was 0.631, 0.627, and 0.544 at 1, 3, and 5 years, respectively (Figure 1(k)), which indicates a fair performance to predict DFS in early CC patients. All the results above suggest that our FAM model can be used in the prediction of the survival status in CC patients.
## 3.2. Establishment and Validation of a Nomogram
In the univariate and multivariate analysis, the risk group ($p \leq 0.001$, HR 3.858, $95\%$ CI: 2.025–7.348; $p \leq 0.001$, HR: 3.963, $95\%$ CI: 2.064–7.612, respectively) and advanced stage ($p \leq 0.001$, HR: 2.900, $95\%$ CI: 1.606–5.236; $p \leq 0.001$, HR: 3.091, $95\%$ CI: 1.585–6.107, respectively) emerged as significant risk factors for worse OS (Figures 2(a) and 2(b)). We also established the univariate and multivariate Cox regression model for the RS and revealed that the RS was the independent predictor for predicting worse OS in both univariate and multivariate analysis in CC patients ($p \leq 0.001$, HR: 5.746, $95\%$ CI: 3.281–10.061; $p \leq 0.001$, HR: 5.210, $95\%$ CI: 2.695–10.072; Supplementary Figures 1a and 1b), while only the advanced stage showed significance in univariate analysis ($p \leq 0.001$, HR: 1.779, $95\%$ CI: 1.353–2.339; Supplementary Figures 1a and 1b). In addition, the results of the PH assumption of risk score and risk group are shown in Supplementary Figures 3o, 3p, 3u, and 3v and Supplementary Table 3, and no statistically significant results were found. Taken together, our results suggest that our FAM gene signature is not inferior to traditional clinicopathological variables, such as stage, and even superior to the pathology grade and type in the clinical practice and can serve as an independent predictor of survival in CC patients. As depicted in Figure 2(c) and Supplementary Figure 1c, a higher total score according to the sum of the assigned numbers for each parameter in the nomogram was correlated with worse 1-, 2-, 3-, and 5-year OS probabilities. For instance, a patient with an advanced stage and a higher risk score would yield a total of 180 points (80 points for stage 4, and 100 points for the high-risk group), with predicted 3-year OS rates of less than $90\%$ (Figure 2(c)).
To validate the risk nomogram model, the predictive performance of the nomogram was assessed by computing the discrimination index and the calibration plot of the model for the 1-, 2-, 3-, and 5-year survival. The C-index was 0.77 or 0.79 for our nomogram with the risk group or RS, respectively, which suggests a good discriminative ability of the nomogram. Calibration plots measure the coherence between the outcomes predicted by the nomogram models and the actual outcomes in the CC cohort. The predictions made by the nomogram model were close to the observed outcomes (1-, 2-, 3-, and 5-year survival) (Figures 3(a)–3(d) for the nomogram with risk group; Supplementary Figures 2a–2d for the nomogram with RS). In addition, to access the accuracy of movement in risk classification, we calculated the NRI for our new prognostic model with the risk group. As a result, when our new model with risk group was compared with the previous standard, NRI displayed an improved reclassification with $72.14\%$ improvement in the prediction accuracy of 3-year survival probability and $49.98\%$ improvement in the prediction accuracy of 5-year survival probability (Figures 3(e)–3(h) for the nomogram with risk group). For the nomogram model with RS, a $45.72\%$ improvement in the prediction accuracy of 3-year survival probability and a $33.75\%$ improvement in the prediction accuracy of 5-year survival probability were observed in the new model with RS (Supplementary Figures 2e–2h). Finally, DCA plots revealed the clinical utility of the nomogram model with or without the risk group and the net benefit of using both models to stratify patients relative to none (assuming that no patient will have an event). The nomogram with risk group displayed a larger net benefit across the range of risk thresholds (≥0.15 for 1-year survival, 0 to around 0.22 for 2-year survival, and 0 to around 0.4 for 3-year and 5-year survival) compared to the model with clinical variables only (Figures 3(i)–3(l)). Better net benefits were observed in the nomogram with RS in comparison to the traditional model of 1-, 2-, 3-, and 5-year survival (Supplementary Figures 2i–2l).
## 3.3. Comparison of Models Built by Other Machine Learning Methods
Next, to compare the predictive value among different models, we used the expression levels of the 309 FAM genes to construct the SVM model. The polynomial SVM model performed rather impressively with the best degree of 4 and coefficient of 0.1 in the training dataset ($p \leq 0.001$ by Wilcoxon test, AUC = 1, Figures 4(a) and 4(c)), but it failed in the testing dataset ($$p \leq 0.52$$ by Wilcoxon test, AUC = 0.54, Figures 4(b) and 4(d)). The significant variable list in SVM is summarized in Supplementary Table 4 (Figure 4(q)). Similar classification results were observed using the random forest algorithm, a significant result in the training cohort ($p \leq 0.001$ by Wilcoxon test, AUC = 1, Figures 4(i) and 4(k)), but an underwhelming result in test dataset ($$p \leq 0.048$$ by Wilcoxon test, AUC = 0.62, Figures 4(j) and 4(l)). The selection process and the top 40 important genes in the random forest are shown in Figures 4(r) and 4(s). Besides, we performed the univariate Cox regression to select the genes correlated with OS (Supplementary Table 5). The intersection of significant genes in each model is displayed in Figure 4(t) and Supplementary Table 6. Then, we employed the 14 specific genes to construct the classification by SVM and random forest to check whether the predictive value was improved. In the training cohort, the performance of the SVM and random forest model was impressive ($p \leq 0.001$ by Wilcoxon test, AUC = 0.91 in the SVM model; $p \leq 0.001$ by the Wilcoxon test, AUC = 0.1 in the random forest model; Figures 4(e), 4(g), 4(m) and 4(o)). The predictive value of the SVM model with signature genes ($$p \leq 0.039$$ by the Wilcoxon test, AUC = 0.63, Figures 4(f) and 4(h)) was slightly improved as compared to the SVM model with the FAM genes in the testing dataset. The discriminative ability of the random forest model with signature genes did not improve compared to the model with FAM genes ($$p \leq 0.052$$ by Wilcoxon test, AUC = 0.62, Figures 4(n) and 4(p)) in the testing cohort. In addition, a separation was observed using a PCA analysis using the 14-gene signature, but there was no separation using the FAM genes (Figures 4(u) and 4(v)). Taken together, the predictive ability of the LASSO-Cox model was the best among all the models we established, which was superior to the SVM and random forest models according to the AUC.
## 3.4. Identification of the FAM Phenotype concerning the Risk Grouping
The FAM remodeling in cancer contains aberrant changes in endogenous FA uptake, de novo synthesis, and β-oxidation to produce energy and store FA. Therefore, we further explored the FAM alteration in high-risk CC patients. The transporters of FA on the plasma membranes contain the FA transport protein family, FA binding proteins, and FA translocase [37]. We found that the FA translocase, CD36, tended to be increased ($$p \leq 0.067$$, Supplementary Figure 4e), while the solute carrier protein family 27 (SLC27) was decreased ($$p \leq 0.029$$ for SLC27A1, $$p \leq 0.0069$$ for SLC27A2, $$p \leq 0.00044$$ for SLC27A3, $$p \leq 0.015$$ for SLC27A5, Supplementary Figures 4ag, 4ah, 4ad, and 4ak) in the high-risk group, which may suggest that those cancer cells do not rely on the exogenous uptake of FA much. Since the CC cells in the high-risk group do not rely on the exogenous uptake of FA, the biosynthesis of FA from glucose, acetate, or glutamine is apparent to be important. Notably, we revealed that the de novo synthesis of FA was significantly upregulated in CC patients with high risk. The enzymes involved in the synthesis of glutamine or acetate to citrate were enhanced, including glutaminase (GLS, $$p \leq 0.0013$$) (Supplementary Figure 4ab). Moreover, the production of palmitate from citrate was promoted through the upregulated expression of ATP-citrate lyase (ACLY, $$p \leq 0.034$$, Supplementary Figure 4a), FA synthase (FASN, $$p \leq 0.042$$, Supplementary Figure 4aa), and long-chainacyl-CoA synthetase 3 (ACSL3, $$p \leq 0.041$$, Supplementary Figure 4b). Then, the saturation of FA was promoted by high expression of stearoyl-CoA desaturase (SCD, $$p \leq 0.018$$, Supplementary Figure 3af). However, the alternation of β oxidation was complex. Some enzymes of β oxidation were downregulated, such as carnitine palmitoyltransferase 1A (CPT1A, $$p \leq 1.9$$ × 10−5) and CPT1B ($$p \leq 0.00024$$) while the CUB domain-containing protein 1 (CDCP1) was increased ($$p \leq 2.6$$ × 10−5, Supplementary Figure 4g). Similarly, the elongation of FA has to be checked comprehensively, as the expression of ELOVL FA elongase 2 (ELOVL2, $$p \leq 0.0088$$) was increased while ELOVL7 ($p \leq 0.001$) was decreased in the high-risk group (Supplementary Figures 4k and 4p). So, we may speculate that the CC patients with high risk were featured by enhanced de novo synthesis of FA in our study.
## 3.5. Identification of DEGs and Functional Enrichment Analysis
To search for the regulation factors and effectors between the two groups, differential gene expression analysis was first performed. Differential expression analysis identified 51 DEGs between the two groups, in which 27 genes were upregulated, whereas 24 genes were downregulated (Figure 5(a)). The top 5 upregulated and downregulated genes are highlighted in Figure 5(b). Then, the GO and KEGG enrichment analyses were performed among the DEGs. The results of the GO analysis are demonstrated in Supplementary Figure 5a. In the KEGG analysis, the metabolic pathways were significantly enriched ($$p \leq 0.0054$$, Supplementary Figures 5b and 5c). In addition, the housekeeping genes were activated in the Msigdb enrichment, among which COX6A1 and COX8A were involved in ATP synthesis and mitochondrial energy metabolism (Supplementary Figure 6a), and the metabolic genes regulated by TP53 were activated in the *Reactome analysis* (Supplementary Figure 6b). The results above indicate that the metabolism was enhanced in the high-risk group.
In the GSVA analysis, according to the HALLMARK gene set, we found that the coagulation, Kirsten rat sarcoma viral oncogene homolog (KRAS) signalling, tumour necrosis factor (TNF) signalling via nuclear factor κB (NFκB), complement and inflammatory response, interleukin 6(IL6)-Janus kinase (JAK)-signal transducer and activator of transcription 3 (STAT3) signalling, transforming growth factor β (TGFβ) signalling, apical junction, angiogenesis, and epithelial-mesenchymal transition were activated in the high-risk group, whereas the E2F targets, G2M checkpoint, DNA repair, and oxidative phosphorylation were inhibited in the high-risk group (Figure 5(c)). Similar results were obtained in KEGG analysis, for example, complement and coagulation cascades were activated in the high-risk group, and oxidative phosphorylation, homologous recombination, base excision repair nucleotide excision repair, DNA replication, mismatch repair, and cell cycle were downregulated in the high-risk group (Figure 5(d)). Those results may indicate that special inflammatory mediators, TNFα, IL6, TGFβ, and complements, might create an immunosuppressive microenvironment with chronic inflammation in high-risk CC patients, which support tumour progression and metastasis by activating several signalling pathways, namely, NFκB, JAK-STAT3, TGFβ, and KRAS signalling.
## 3.6. Identification of Hub Genes and Regulation Network
The hub genes among the DEGs obtained by “GOSemSim” analysis are listed in Figure 5(e). The likelihood of PPI is identified in Figure 6(a). Then, the proteins in the PPI network were further analysed by Cytoscape CytoHubba. The top 40 genes were filtered by the algorithm “closeness” as in Figure 5(f). The intersection of the hub genes derived from “friends” and “closeness” is listed in Supplementary Table 7. Besides the direct interaction of proteins among DEGs (Figure 6(a)), we also explored the transcription regulation and identified 32 transcription factors based on the DEGs as shown in Figure 6(b) (the transcription factors are listed in the Supplementary Table 8). Furthermore, the post-transcriptional regulations by miRNA and lncRNA were inferred (Figure 6(c)).
Notably, the SRY-box transcription factor 2, SOX2, was rather active. SOX2 is the centre of the transcriptional network influencing pluripotency and is essential in formation of cancer stem cells and resistance to treatment [38, 39].
## 3.7. Somatic Genomic Alternation Analysis
First, the somatic mutation landscapes were summarized according to risk grouping (Figures 7(a) and 7(b)). The somatic variants contain single-nucleotide variants (SNVs) and small insertions/deletions (indels). In both the risk groups, the top 3 variant classifications were missense mutation, nonsense mutation, and frameshift deletion and the most frequent variant type was single-nucleotide polymorphism (SNP) (Figures 7(a) and 7(b)). SNV with C > T occurred predominantly in both groups. There were 17606 C > Tbase substitutions in the high risk group and 30151 C > T base substitutions in the high risk group (Figures 7(a) and 7(b)). The median of variants per sample was 69.5 in the low-risk group and 64.5 in the high-risk group (Figures 7(a) and 7(b)). Similar to the results of variants per sample, the TMB was 1.39/MB in the low-risk group and 1.29/MB in the high-risk group, suggesting low TMB in CC patients (Supplementary Figures 7c and 7d). The top three mutated genes were tinin (TTN), mucin 4 (MUC4), and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) in the low-risk group and TTN, mucin 16 (MUC16), and PIK3CA in the high-risk group (Figures 7(a)–7(c)). In addition, 210 samples ($84.68\%$) were detected to have somatic mutations in the whole CC cohort (Figure 7(c)). Among them, 114 samples ($90.48\%$) had somatic mutations in the low-risk group, and 107 ($87.7\%$) had somatic mutations in the high-risk group (Supplementary Figures 8a and 8b). Next, we found that the mutation frequency of the signature genes differed in different risk groups (Figure 7(d)). For example, the mutation rate of NCAPH2 was $6\%$ in the high-risk group, whereas only $1\%$ was in the low-risk group (Figure 7(d)). The mutation pattern of the signature genes was distinguished in the respective group (Figure 7(e)), especially NCAPH2, which was identified as a potential driver gene in CC [40].
As SNPs are classified into two conversions of transitions (A > G/G > A and T > C/C > T) and four conversions of transversions (C > A/A > C, C > G/G > C, T > A/A > T, and T > G/G > T) according to the types of base substitution. Supplementary Figure 9 shows the fraction of conversions in each sample. The C > T transversion accounted for the highest incidence among the six conversions in both groups.
## 3.8. Identification of Mutation Driver and Affected Signalling Pathway
During the progression of cancer, initiation and promotion of tumour development are considered by driver mutations [41]. The comparison revealed 28 significant genes with differential mutation patterns concerning the risk grouping ($p \leq 0.05$, Figure 8(a)). Among them, 25 genes were significantly enriched in the high-risk group, and the other four were enriched in the low-risk group (Figures 8(a) and 8(b)). Furthermore, the top four driver genes, KRAS, PIK3CA, F-box and WD repeat domain containing 7 (FBXW7), and ERBB3, were identified in the low-risk group, enriched in RTK-RAS and PI3K-AKT signalling pathway (Figure 8(c)). Interestingly, the mutation of FBXW7, involved in NOTCH signalling, was detected in $16\%$ of low-risk patients (Figure 8(b)); however, the overall mutation frequency of FBXW7 in CC was around $6\%$ [42]. Meanwhile, neuroblastoma breakpoint family member 14 (NBPF14), ERBB2, MAPK1, and KRAS were identified as driver genes in the high-risk group, mainly enriched in the RTK-RAS-MAPK signalling pathway (Figure 8(d)). The mutation hotspots of the top four driver genes in the two cohorts are shown in Supplementary Figure 10. G12V and G12D mutations of KRAS were observed in the low-risk groups, while G12C was observed in the high-risk group and G13D was observed in both groups (Supplementary Figure 10a). The E542K and E545K mutations of PIK3CA were present in both groups, (Supplementary Figure 10b). This result was consistent with the previous result that PIK3CA is the third mutated gene in both groups (Figures 7(a) and 7(b)). The E322K of MAPK1 was found in both groups, and D321N and R135K were exclusively in the high-risk group (Supplementary Figure 10c). The R505, R479, and R465 mutations of FBXW7 mostly occurred in the low-risk group (Supplementary Figure 10d). The E872G mutation of NBPF14 was only found in the high-risk group (Supplementary Figure 10e).
Notably, the mutation patterns of RTKs, ERBB3 and ERBB2, differed in both groups (Figures 9(b) and 9(d)). The S310F mutation of ERBB2 was only present in the high-risk group, known as oncogenic driver mutation (Figure 9(b)) [43]. The mutated ERBB2 led to worse OS in CC patients ($$p \leq 1$$ × 10−4, Figure 9(a)), whereas the mutated ERBB3 seemed to not affect OS ($$p \leq 0.363$$, Figure 9(c)). The results may partially explain the high-risk group with the ERBB2 as the driver gene had worse OS. Additionally, mutations occurred in a mutual co-occurrence manner in both groups (Supplementary Figures 7a and 7b).
Consistent with the enriched signalling pathway by driver mutation genes, the top three frequently mutated oncogenic signalling pathways were RTK/RAS, NOTCH, and PI3K in the two groups (Figures 8(e) and 8(f)). The detailed mutation patterns of RTK/RAS, PI3K, and NOTCH signalling pathways are shown in Supplementary Figures 11–13.
## 3.9. De Novo Mutational Signature Analysis
The progression of cancer leaves behind a distinctive mutational pattern that can display its mutagenic processes [29]. In the mutational processes analysis, we obtained three signatures as compared against the COSMIC signatures v2 in the low-risk group, while five signatures were in the high-risk group (Supplementary Figures 14a and 14b). Those signatures also were compared against the updated COSMIC signatures v3, and the results are demonstrated in Figures 10(a) and 10(b). The matched COSMIC signatures and corresponding aetiology are summarized in Table 1. Notably, the SBS10, only observed in the high-risk group, is related to defective DNA polymerase ε which is responsible for the exonuclease proofreading and prevention of the accumulation of mutations. The POLE gene encodes the catalytic subunit for 5′-3′ DNA polymerase and 3′-5′ exonuclease, which is important for genome stability. The incidence of POLE somatic mutations was $2.79\%$; however, it is $4.92\%$ in the high-risk group while $2.38\%$ in the low-risk group (Supplementary Figure 14c). We further identified the P254L, S297F, and V411L mutations of POLE only presented in the high-risk group (Supplementary Figure 14c).
Oncogenes are clustered around mutational hotspots [28]. Hypermutated genomic regions, named “kataegis,” are defined as those genomic segments containing six or more consecutive mutations with an average inter-mutation distance of less than or equal to 100 base pairs [44]. The formation of kataegis is hypothesized to result from multiple cytosine deamination and enrichment in C > T and C > G substitutions, which is caused by the unleashed activity of apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like (APOBEC), a family of cytidine deaminases [44]. Figures 10(c) and 10(d) demonstrate the samples with the most kataegis region, TCGA-JW-A5VL in the low-risk group and TCGA-2W-A8YY in the high-risk group. Furthermore, we explored the status of APOBEC-associated mutations in both the risk groups in CC patients (Figures 10(e) and 10(f)). As a result, $73.81\%$ (93 of 126 samples) of patients in the low-risk group were enriched for APOBEC-associated mutations (APOBEC enrichment score >2, Figure 10(e)), while $73.98\%$ (91 of 123 samples) of patients in the high-risk group (Figure 10(f)). Furthermore, in the low-risk group, increased mutation rates within FLG, TNN, and NAV3 genes were detected in APOBEC-enriched samples, while FOLH1, ADGRG4, MAP3K15, MEGF8, and ADAMTS18 with higher mutation rates were detected in non-APOBEC-enriched samples (Figure 10(e)). However, in the high-risk group, top genes with increased mutation rates were found in non-APOBEC-enriched samples, such as PTEN and ARID1B (Figure 10(f)). Since the most frequent mutations were in the non-APOBEC-enriched samples in the high-risk group, we speculate that the mutations might be related to the SBS10a exonuclease domain mutations of POLE in the high-risk group. Interestingly, R793C, R616C, V411L, and P254L of POLE occurred in the same patient with the most kataegis region in the high-risk group, TCGA-2W-A8YY (Figure 10(d)).
## 3.10. Copy Number Variation Analysis
We identified several CNVs in the low-risk group (Supplementary Figure 15), including the deletions on 2q33.2 (NBEAL1, CD28, CTLA4, CYP20A1), 2q22.1 (THSD7B, CXCR4), 2q37.2 (SH3BP4), 10q23.31 (FAS, PTEN), and 13q14.2 (RCBTB1) (Figures 11(a) and 11(c)). In the high-risk group, the most prevalent duplications were 11q22.2 (MMP1) and 11q22.1 (YAP1, BIRC$\frac{2}{3}$), while the most prevalent deletions were 2q37.1 (UGT1A1), 2q22.1 (THSD7B, CXCR4), and 19p13.3 (granzyme M) (Figures 11(b) and 11(d)). In the high-risk group, the deletion of granzyme M might lead to deficiency of cytotoxic lymphocytes [45] and amplification of YAP1 and MMP1 may result in enhanced angiogenesis and EMT [46, 47].
## 3.11. Identification of Neoantigens
The correlation between the mutation burden and predicted neoantigen load revealed a positive linear relationship ($r = 0.89$, $$p \leq 1.36$$ × 10−53, Supplementary Figure 16). Sparse predicted neoantigens were shared across the population. The most common neoantigen, PIK3CA-STRDPISEITK-HLA A∗03: 01, was present in 8 patients (Figure 12(a)), which might be generated as off-shelf products. The most frequent neoantigens were derived from the mutation of PIK3CA, E1A binding protein P300 (EP300), and ERBB3in the low-risk group (Figure 12(b)) and PI3KCA, MAPK1, and ERBB2 in the high-risk group (Figure 12(c)), which is coherent with the driver mutation in respective risk groups (Figures 8(c) and 8(d)). More types of neoantigen were identified in the low-risk group than in the high-risk group, and this result is partially in agreement with the previous results that higher TMB in the low-risk group (Supplementary Figures 7c and 7d).
## 3.12. Correlation between Cancer Germline Antigens and Immune Cell Infiltration
Besides neoantigens, which result from somatic mutations, the cancer antigenome also contains CGAs. CGAs are proteins normally expressed in germline cells and aberrantly expressed in tumour tissue. We found a distinct expression pattern of CGAs between low/high-risk groups (Supplementary Figure 17). There are 36 significant genes and differentially expressed CGA genes between the low- and high-risk groups (Supplementary Figure 18). The expression of a number of CGA genes, including PBK, SPAG8, TSGA10, LDHC, TAF7L, PRSS55, ODF2, TPPP2, OIP5, NUF2, TSSK6, CEP55, IGSF11, and CASC5, was significantly downregulated in the high-risk group (Supplementary Figure 18). The expression levels of MPP1 andKDM5B were enhanced in the high-risk group (Supplementary Figure 18). Next, we further explored whether the expression of CGAs is associated with the immune infiltration cells. Notably, we identified that several CGAs were positively correlated with the CD8 T cell enrichment, namely, PBK (Figures 13(b), 13(d), 13(k), and 13(v)), SPAG8 (Figures 13(c), 13(z), and 13(w)), TSGA10 (Figures 13(f) and 13(y)), LDHC (Figures 13(h) and 13(m)), TAF7L (Figure 13), ODF2 (Figure 13(l)), TPPP2 (Figure 13(n)), OIP5 (Figure 13(o)), CASC5 (Figure 13(q)), and PRSS55 (Figure 13(s) and 13(x)), which were remarkably suppressed in the high-risk group (Supplementary Figure 17), indicating an immune inhibitory environment in the high-risk group, while KDM5B, enhanced expression in the high-risk group, was negatively associated with CD8 T cell enrichment, also suggesting an inhibitory immune environment in the high-risk group (Figures 13(a), 13(g), 13(j), 13(r), and 13(t)). However, there were some CGA genes negatively correlated with CD8 T cell enrichment and most of them were in a relatively low expression level such as MAFEA9B (Supplementary Figure 19). Our results are consistent with the findings that some CGAs are significantly correlated with activated CD8T cells [32]. Moreover, we also identified a negative correlation between the enrichment of Treg cells and the expression of PBK, NUF2, TSGA10, TSSK6, ODF2, OIP5, LDHC, and CEP55 (Supplementary Figures 20 and 21). *Those* genes were downregulated in the high-risk group (Supplementary Figure 18), also suggesting a suppressive immune environment in the high-risk group. Interestingly, KDM5B, associated with negative T cell enrichment in the high-risk group in our analysis, was reported to promote immune evasion and reprogramming lipid metabolism [48, 49]. Our results might suggest that the FAM phenotype in the high-risk group may be related to the inhibitory immune environment.
## 3.13. Evaluation of the Cellular Composition in Tumour Microenvironment
To explore the landscape of TME, an analysis of immune infiltrating cells and other cells in the TME of CC was performed by the “IOBR” package in R [34]. The influence of infiltrating cells and risk score on survival by Cox regression are summarized in Supplementary Table 9.
As shown in Supplementary Figures 22 and 14(a), CD8 T cells, B naïve, plasma cells, and resting mast cells by CIBERSORT [50] were negatively associated with the RS which suggests an inhibition of adaptive immune responses in the high-risk group. The CD8 T cell by CIBERSORT was associated with an improved prognosis (HR = 0.86, $95\%$ CIs: 0.52–0.89, $$p \leq 0.0053$$, Supplementary Table 9 and Figure 14(a)). However, activated mast cells by CIBERSORT were positively correlated with RS and may lead to a worse prognosis (HR = 2.31, $95\%$ CIs: 1.76–3.04, $$p \leq 1.93$$ × 10−9, Supplementary Table 9 and Figure 14(a)).
Moreover, endothelial cells by both MCPcounter and xCell [51] were positively correlated with the RS, which coincided with our result that the genes of hallmarks of angiogenesis were significantly enriched in the high-risk group in the GSVA analysis (Figures 4(c) and 4(d)). The endothelial cells in the MCPcounter resulted in worse survival (HR = 1.37, $95\%$CIs: 1.02–1.84, $$p \leq 0.0032$$, Supplementary Table 9 and Figure 14(a)).
Furthermore, fibroblasts by MCPcounter [52], cancer-associated fibroblasts (CAFs) by EPIC [53], and stromal score by estimate [54] were positively related to the RS, indicating that abnormal FAM may relate to enhance fibroblast activity. Meanwhile, M0 macrophages were inhibited, but the M1 macrophages were enhanced with the increasing RS, indicating a chronic inflammation featured by macrophage and lymphocyte infiltration [55].
Adipocytes were notably accumulated in the TME of the high-risk group and had a negative influence on survival (HR = 4.89, $95\%$ CIs: 2.01–10.94, $$p \leq 0.00035$$, Supplementary Table 9 and Figure 14(a)). Megakaryocyte-erythroid progenitors (MEPs) by xCell and AZ by IPS [32] were abrogated in the high-risk group (Figure 14). It has been reported that the activated mast cells, macrophages, and neutrophils can secrete the pro-inflammatory cytokines, IL-6 and TNF α, which may activate the IL-6-JAK-STAT3 signalling and TNF-NFκB signalling in the high-risk group as the results in HALLMARK enrichment (Figures 4(c) and 4(d)). The immunosuppressive and chronic inflammatory TME may be the reason for worse OS in the high-risk CC patients in our study.
## 3.14. Exploration of Potential Therapeutic Drugs concerning Prognostic Models
According to the prediction, the RS was positively correlated with the predicted IC50 of crizotinib ($R = 0.14$, $$p \leq 0.02$$, Figure 14(f)), FK866 ($R = 0.15$, $$p \leq 0.012$$, Figure 14(g)), and rapamycin, ($R = 0.21$, $$p \leq 0.00061$$, Figure 14(h)) but negatively correlated with the predicted IC50 of AZ628 (R = −0.22, $$p \leq 0.00039$$, Figure 14(e)). Furthermore, the CC patients with high risk were more resistant to crizotinib, FK866, and rapamycin ($$p \leq 0.027$$, $$p \leq 0.0064$$, and $$p \leq 0.0099$$, respectively, Figures 14(j)–14(l)). Meanwhile, the CC patients with high risk were predicted to be more sensitive to the irreversible rapidly accelerated fibrosarcoma (RAF) inhibitor AZ628 than the CC patients with low risk (Figures 14(e) and 14(i)). A further prediction of the immune therapy response revealed no difference between the two groups ($$p \leq 0.48$$, Figure 14(b)). Since POLE mutations are reported to be related with good response to the immunotherapy [56], we compared the prognosis and predicted immune therapy benefits concerning the POLE mutation status. However, concerning the POLE mutation status, we did not find the improved prognosis (Supplementary Figure 14d) and predicted immunotherapy benefits either in all samples or in the high-risk group ($$p \leq 0.41$$, $$p \leq 0.15$$, respectively, Figures 14(c) and 14(d)). Additionally, the potential drug-target categories based on the risk grouping are summarized in Supplementary Figure 23.
## 4. Discussion
Fatty acids (FAs) are the major components of phospholipids, sphingolipids, and triglycerides and have significant roles in the production and storage of energy, synthesis of the membrane, regulation of membrane fluidity, and secondary messengers [3]. Remodeling of FAM broadly contains alterations in the transportation of FA, de novo FA synthesis in the cytosol, and β-oxidation of FA to generate ATP in the mitochondria in cancer [57]. Enhanced de novo FA synthesis is necessary for cancer cells to produce phospholipids for membranes and lipid rafts [58]. β-Oxidation of FA supplies the tumour cells with tremendous energy for aggressiveness [59].
With Pap smear-based screening and HPV vaccination, the incidence of CC decline significantly in high-income countries [1]. However, CC is still the fourth most commonly diagnosed cancer and the fourth leading cause of cancer deaths in women worldwide [1]. Especially in transitioning countries, patients diagnosed at advanced stages lack efficient therapy [43]. On the other hand, as mentioned in previous research, diverse activation patterns of metabolic signalling pathways may be the reason for the molecular diversity of CC [60]. Therefore, we constructed a valid prognostic model based on FAM genes to distinguish CC patients at different risks. By LASSO-Cox regression, we obtained a prognostic model with good to fair performance. Other models built by the SVM and the random forest did not reach a good performance in the testing cohort. The nomogram for clinical application also achieved a good performance in calibration curves, NRI, and DCA analysis. Therefore, our model is easy to use and robust. In our FAM signature, SDHD is vital for cell growth and metabolism [61]. HCCS participates in oxidative phosphorylation and apoptosis [62]. SERINC1 is involved in serine-derived lipids [63]. THRSP can maintain mitochondrial function and regulate sphingolipid metabolism in human adipocytes [64]. Importantly, we identified that the high-risk CC patients were featured by enhanced de novo synthesis of FA.
Then, we set out to find out the underlying mechanisms leading to different FAM phenotypes and OS in our model. Besides the activation of metabolic pathways in the high-risk CC patients, our FAM phenotypes are also related to increased inflammatory responses. The inflammatory factors, such as complements, IL6, TNFα, and TGFβ, and the inflammatory responses were enriched in the high-risk group. The inflammatory TME is also verified from cell levels. With the increase of RS and the infiltration of CD8 T cells, B naïve, plasma cells, and resting mast cells, M0 macrophages were suppressed, while neutrophils, activated mast cells, and M1 macrophages were boosted. The impaired infiltration of CD8 T cells may lead to immunosuppressive TME and a worse prognosis in CC patients with high RS. Next, we identified several CGAs which were associated with CD8 T cell enrichment. However, those genes were downregulated in the high-risk group, which might be one reason for the inhibitory immune TME. Notably, enriched adipocytes in the high-risk group are reported to secrete a variety of inflammatory cytokines and adipokines, such as TNFα and IL-6, recruiting lymphocytes, and macrophages [65].
Moreover, in the GSVA analysis, the angiogenesis and EMT were also enhanced in the high-risk group, which is in agreement with the results that endothelial cells, CAFs, and stromal score enrichment are positively associated with RS. CAFs can be derived from the migration of adipocytes [66], and endothelial and epithelial cells, through endothelial or epithelial to mesenchymal transition [67, 68]. CAFs have been reported to induce EMT and enhance angiogenesis and immunosuppression in TME [69, 70]. Besides the production of free FAs to support metastatic cancer cell survival [71], enriched adipocytes can promote the EMT of tumour cells and stabilize vascularization [72, 73].
Above all, we might confer that aberrant FAM may trigger tumour-extrinsic inflammation, which leads to an immunosuppressive, proangiogenic, and pro-tumoural microenvironment [74].
Oncogenic signalling pathways can directly regulate FAM enzymes to shape tumour lipidome. We identified PIK3CA as the frequently mutated gene in both risk groups, as in the literature [75]. E542K and E545K mutations in CC patients, activating mutations of the PIK3A helical domain, are considered to be correlated with APOBEC mutagenesis [60]. In agreement with the finding that PI3K/AKT pathway increases FA synthesis while suppressing the β-oxidation in diabetes [76], we also found that the CC patients were featured by enrichment of PI3K/AKT signalling and enhanced de novo FA synthesis.
Besides oncogenic mutations in PICK3CA, aggravated stimulation from RTKs can activate the PI3K-AKT signalling. We identified ERBB3 in the low-risk group and ERBB2 in the high-risk group as driver mutations. Especially the S310F in the high-risk group is most frequent among HER2 extracellular domain mutations, which can form an active heterodimer with the EGFR [77]. The activity of HER2 is stabilized and activated by MUC4, the second mutated gene in the low-risk group and the fourth mutated gene in the high-risk group [78]. Furthermore, the recruitment of PI3K to activated ERBB3 can be promoted by the interaction between MUC4 and ERBB2 [79]. HER2-positive tumours are featured by sustained de novo synthesis of lipids [80].
Besides the PI3K-AKT signalling pathway, the RTKs also activate the RAS/MAPK signalling transduction [81]. In the RAS/MAPK signalling, KRAS mutation in both groups and MAPK1 mutation in the high-risk group were also identified as driver mutations. KRAS is expressed by the uterus at a high level [82]. G12V and G12D were observed in the low-risk group, which are the most frequent mutations across tumour types [82]. G12V and G12D are weak drivers in colorectal cancer and lung adenocarcinoma with smoking; however, in endometrial cancer, they are considered major drivers [82, 83]. G12C was observed in the high-risk group, which is coherent with the KRAS-G12C as a major driver in lung adenocarcinoma with smoking [29]. G13D of KRAS, presented in both CC groups with the signature of mismatch repair deficiency, is reported to be associated with mismatch repair deficiency signatures in gastric and endometrial tumours [82]. MAPK1 E322K can hyperactivate EGFR and serve as a biomarker for erlotinib sensitivity in head and neck squamous cell cancer [84]. MAPK1 E322K was observed in both CC groups, coherent with the previous study [43]. D321N and R135K of MAPK1 were observed only in the high-risk group, which can enhance EGFR activation [85]. D321N has been reported to contribute high sensitivity to erlotinib, and R135K confers moderate sensitivity to erlotinib in head and neck squamous cell carcinoma [85], as potential therapeutic targets in high-risk CC patients. KRAS/ERK (MAPK1) signalling can increase the biosynthesis of acetyl-CoA from acetate and the expression of FASN and SCD in the de novo FA synthesis [86, 87].
In addition, the NOTCH signalling pathway was found to be activated in CC patients. The mutations of FBXW7, involved in NOTCH signal transduction, were detected in the low-risk group [88]. FBXW7, a tumour suppressor, can recognize the substrates, namely, Cyclin E, c-Myc, Mcl, mTOR, Jun, and NOTCH, for the component of the SCF E3 ubiquitin ligase [89]. The mutational hotspots of FBXW7 R505, R479, and R465 in the substrate binding domain, WD40 motif, were observed exclusively in the low-risk group, which may impair the ubiquitylation and degradation of specific substrates [90]. Previous studies have suggested that FBXW7 mutations strengthen the interaction among cancer-initiating cells via the NOTCH signalling pathway [42].
In the mutational process analysis, three signatures, SBS13, SBS2, and SBS6, were present in all CC patients. SBS13 and SBS2 mutations often occur in the kataegis in the same samples [91]. Both SBS2 and SBS13 are mainly associated with APOBEC hyperactivation, which may reflect the innate immune response to the virus, retrotransposon jumping, or tissue inflammation in cancer [91]. Therefore, we can infer that SBS2 and SBS13 may represent the damage to the genome in the context of HPV infection and persistent inflammation caused by aberrant FAM in CC patients. SBS6 is featured predominantly by C > T at NpCpG mutations leading to substitution and small indels termed as microsatellite instability caused by defective DNA mismatch repair, which is coherent with the findings that G13D mutation of KRAS, related to defective mismatch repair, was present in CC patients [91]. Acquisition of SBS1 mutations in the high-risk group referred to as a cell division/mitotic clock correlates with time or age and the rates of stem cell division which reflects the endogenous mutation process [91]. The SBS10a, only present in the high-risk group, may be generated by POLE exonuclease domain mutations. The exonuclease domain of POLE recognizes and removes wrong bases generated during replication [92]. The mutations in the exonuclease of POLE, referred to as hypermutators, cause a 10-to-100-fold increase in the mutation rate during replication [92], which is in accordance with the interesting finding that CC patients with several hypermutators had the most kataegis in the high-risk group. The mutations of S297F and V411L of POLE were reported to be hotspot mutations and associated with high TMB in endometrial carcinoma, which was present exclusively in the high-risk group in our study [93]. There are also some shreds of evidence that the non-exonuclease domain mutations of POLE have pathogenicity. The patients with POLE mutations had a high immune response and good prognoses in endometrial carcinoma [94]. However, we did not observe either a good prognosis or an improved prediction of immune therapy response according to the mutation status of POLE. According to the instruction, a high TIDE score indicates a potential immune escape phenotype and resistance to cancer immunotherapies. We might infer cautiously that the POLE mutation of exonuclease might lead to a good immune therapy response since the TIDE score had a decreasing trend in the high-risk group with mutated POLE [35]. Due to the insufficient sample size with POLE mutation in the CC cohort and follow-up information, further verification is needed.
In the following exploration of potential therapies for high-risk CC patients, we found that the CC patients with high risk might be more resistant to rapamycin (an allosteric inhibitor of mTOR), crizotinib (an adenosine triphosphate inhibitor of receptor tyrosine kinases), and FK866 (nicotinamide phosphoribosyltransferase inhibitor). However, AZ628 might be the potential therapeutic option for CC patients with high risk, which is reported to cause suppression of RAF/ERK signalling in KRAS mutant lung cancer [95]. We also identified that the neoantigen PIK3CA-STRDPISEITK-HLA A∗03: 01 with relatively high frequency in CC patients might also be a treatment option. Hence, our results might improve current treatment strategies to defeat CC.
Concerning the limitation of our study, the gene expression analyses may not provide a direct reflection of enzyme activity or dependencies on specific metabolic pathways, and further experiments are needed to verify the FAM phenotypes, the altered signalling pathways, and the efficiency of those potential drug targets. Next, presently very limited open data on CC are available, additional studies are required to validate our prognostic model. Moreover, we did not consider HPV infection status in our model which is proved as an important factor in the progression of CC [96]. Disturbed FAM can influence chronic inflammation, persistent HPV infection, and carcinogenesis [97, 98].
According to our aforementioned results, we propose that distinct mutated driver genes may lead to different FAM features, and aberrant FAM then results in the differential infiltration and function of cells in TMEs, ultimately leading to different prognoses in CC.
## 5. Conclusions
In this study, we constructed a reliable model with 14 FAM-related genes by LASSO-Cox regression, by which we achieved a good risk stratification in cervical cancer patients. With the risk grouping, a feasible nomogram was established and validated. To understand the underlying mechanism, we found the high-risk samples featured by activated de novo fatty acid synthesis, epithelial to mesenchymal transition, angiogenesis, and inflammation response, which might be caused by mutations of oncogenic driver genes in RTK/RAS, PI3K, and NOTCH signalling pathways. Especially, the oncogenic mutations of ERBB2, only present in the high-risk group, led to worse survival. Besides the hyperactivity of cytidine deaminase and deficiency of mismatch repair, the mutations of POLE might be partially responsible for the gene mutation in patients with high risk. Moreover, increasing RS was found to be related to chronic inflammatory and suppressive immune microenvironment. The reduced expression of CGAs might result in the reduction of CD8 T cell infiltration in the high-risk group. Finally, the RAF inhibitor AZ628 was predicted to be sensitive in patients with high risk. Our findings provide a novel insight for personalized treatment in CC.
## Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## Authors' Contributions
Yanjun Zhou and Jiahao Zhu contributed equally to this work.
## References
1. Sung H., Ferlay J., Siegel R. L.. **Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. (2021) **71** 209-249. DOI: 10.3322/caac.21660
2. Boussios S., Seraj E., Zarkavelis G.. **Management of patients with recurrent/advanced cervical cancer beyond first line platinum regimens: where do we stand? A literature review**. (2016) **108** 164-174. DOI: 10.1016/j.critrevonc.2016.11.006
3. Koundouros N., Poulogiannis G.. **Reprogramming of fatty acid metabolism in cancer**. (2020) **122** 4-22. DOI: 10.1038/s41416-019-0650-z
4. Sugiura A., Rathmell J. C.. **Metabolic barriers to T cell function in tumors**. (2018) **200** 400-407. DOI: 10.4049/jimmunol.1701041
5. Ringel A. E., Drijvers J. M., Baker G. J.. **Obesity shapes metabolism in the tumor microenvironment to suppress anti-tumor immunity**. (2020) **183** 1848-1866 e26. DOI: 10.1016/j.cell.2020.11.009
6. Li Z., Zhang H.. **Reprogramming of glucose, fatty acid and amino acid metabolism for cancer progression**. (2016) **73** 377-392. DOI: 10.1007/s00018-015-2070-4
7. Pearce E. L., Walsh M. C., Cejas P. J.. **Enhancing CD8 T-cell memory by modulating fatty acid metabolism**. (2009) **460** 103-107. DOI: 10.1038/nature08097
8. Herber D. L., Cao W., Nefedova Y.. **Lipid accumulation and dendritic cell dysfunction in cancer**. (2010) **16** 880-886. DOI: 10.1038/nm.2172
9. Peng S., Chen D., Cai J.. **Enhancing cancer-associated fibroblast fatty acid catabolism within a metabolically challenging tumor microenvironment drives colon cancer peritoneal metastasis**. (2021) **15** 1391-1411. DOI: 10.1002/1878-0261.12917
10. Shang C., Wang W., Liao Y.. **LNMICC promotes nodal metastasis of cervical cancer by reprogramming fatty acid metabolism**. (2018) **78** 877-890. DOI: 10.1158/0008-5472.can-17-2356
11. Menendez J. A., Vellon L., Mehmi I.. **Inhibition of fatty acid synthase (FAS) suppresses HER2/neu (erbB-2) oncogene overexpression in cancer cells**. (2004) **101** 10715-10720. DOI: 10.1073/pnas.0403390101
12. Augustine D., Khan W., Rao R.. **Lipid metabolism in cancer: a systematic review**. (2021) **20** p. 4. DOI: 10.4103/jcar.jcar_15_20
13. Wang Z., Jensen M. A., Zenklusen J. C.. **A practical guide to the cancer genome atlas (TCGA)**. (2016) **1418** 111-141. DOI: 10.1007/978-1-4939-3578-9_6
14. Lee Y. Y., Kim T. J., Kim J. Y.. **Genetic profiling to predict recurrence of early cervical cancer**. (2013) **131** 650-654. DOI: 10.1016/j.ygyno.2013.10.003
15. Friedman J., Hastie T., Tibshirani R.. **Regularization paths for generalized linear models via coordinate descent**. (2010) **33** 1-22. DOI: 10.18637/jss.v033.i01
16. Blanche P., Dartigues J. F., Jacqmin-Gadda H.. **Estimating and comparing time-dependent areas under receiver operating characteristic curves for censored event times with competing risks**. (2013) **32** 5381-5397. DOI: 10.1002/sim.5958
17. Zhang Z., Kattan M. W.. **Drawing Nomograms with R: applications to categorical outcome and survival data**. (2017) **5** p. 211. DOI: 10.21037/atm.2017.04.01
18. Alba A. C., Agoritsas T., Walsh M.. **Discrimination and calibration of clinical prediction models: users’ guides to the medical literature**. (2017) **318** 1377-1384. DOI: 10.1001/jama.2017.12126
19. Vickers A. J., Elkin E. B.. **Decision curve analysis: a novel method for evaluating prediction models**. (2006) **26** 565-574. DOI: 10.1177/0272989x06295361
20. Ashburner M., Ball C. A., Blake J. A.. **Gene Ontology: tool for the unification of biology**. (2000) **25** 25-29. DOI: 10.1038/75556
21. Subramanian A., Tamayo P., Mootha V. K.. **Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles**. (2005) **102** 15545-15550. DOI: 10.1073/pnas.0506580102
22. Hanzelmann S., Castelo R., Guinney J.. **GSVA: gene set variation analysis for microarray and RNA-seq data**. (2013) **14** p. 7. DOI: 10.1186/1471-2105-14-7
23. Yu G., Li F., Qin Y., Bo X., Wu Y., Wang S.. **GOSemSim: an R package for measuring semantic similarity among GO terms and gene products**. (2010) **26** 976-978. DOI: 10.1093/bioinformatics/btq064
24. Huang H. Y., Lin Y. C. D., Li J.. **miRTarBase 2020: updates to the experimentally validated microRNA-target interaction database**. (2020) **48** D148-D154. DOI: 10.1093/nar/gkz896
25. Li J. H., Liu S., Zhou H., Qu L. H., Yang J. H.. **starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction networks from large-scaleCLIP-Seq data**. (2014) **42** D92-D97. DOI: 10.1093/nar/gkt1248
26. Mayakonda A., Lin D. C., Assenov Y., Plass C., Koeffler H. P.. **Maftools: efficient and comprehensive analysis of somatic variants in cancer**. (2018) **28** 1747-1756. DOI: 10.1101/gr.239244.118
27. Skoulidis F., Goldberg M. E., Greenawalt D. M.. **STK11/LKB1 mutations and PD-1 inhibitor resistance in KRAS-mutant lung adenocarcinoma**. (2018) **8** 822-835. DOI: 10.1158/2159-8290.CD-18-0099
28. Tamborero D., Gonzalez-Perez A., Lopez-Bigas N.. **OncodriveCLUST: exploiting the positional clustering of somatic mutations to identify cancer genes**. (2013) **29** 2238-2244. DOI: 10.1093/bioinformatics/btt395
29. Alexandrov L. B., Nik-Zainal S., Wedge D. C.. **Signatures of mutational processes in human cancer**. (2013) **500** 415-421. DOI: 10.1038/nature12477
30. Mermel C. H., Schumacher S. E., Hill B., Meyerson M. L., Beroukhim R., Getz G.. **GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers**. (2011) **12** p. R41. DOI: 10.1186/gb-2011-12-4-r41
31. Wu J., Zhao W., Zhou B.. **TSNAdb: a database for tumor-specific neoantigens from immunogenomics data analysis**. (2018) **16** 276-282. DOI: 10.1016/j.gpb.2018.06.003
32. Charoentong P., Finotello F., Angelova M.. **Pan-cancer immunogenomic analyses reveal genotype-immunophenotype relationships and predictors of response to checkpoint blockade**. (2017) **18** 248-262. DOI: 10.1016/j.celrep.2016.12.019
33. Miller A., Asmann Y., Cattaneo L.. **High somatic mutation and neoantigen burden are correlated with decreased progression-free survival in multiple myeloma**. (2017) **7** p. e612. DOI: 10.1038/bcj.2017.94
34. Zeng D., Ye Z., Shen R.. **IOBR: multi-omicsimmuno-oncology biological research to decode tumor microenvironment and signatures**. (2021) **12**. DOI: 10.3389/fimmu.2021.687975
35. Jiang P., Gu S., Pan D.. **Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response**. (2018) **24** 1550-1558. DOI: 10.1038/s41591-018-0136-1
36. Geeleher P., Cox N., Huang R. S.. **pRRophetic: an R package for prediction of clinical chemotherapeutic response from tumor gene expression levels**. (2014) **9**. DOI: 10.1371/journal.pone.0107468
37. Su X., Abumrad N. A.. **Cellular fatty acid uptake: a pathway under construction**. (2009) **20** 72-77. DOI: 10.1016/j.tem.2008.11.001
38. Boyer L. A., Lee T. I., Cole M. F.. **Core transcriptional regulatory circuitry in human embryonic stem cells**. (2005) **122** 947-956. DOI: 10.1016/j.cell.2005.08.020
39. Novak D., Huser L., Elton J. J., Umansky V., Altevogt P., Utikal J.. **SOX2 in development and cancer biology**. (2020) **67** 74-82. DOI: 10.1016/j.semcancer.2019.08.007
40. Xu Y., Luo H., Hu Q., Zhu H.. **Identification of potential driver genes based on multi-genomic data in cervical cancer**. (2021) **12**. DOI: 10.3389/fgene.2021.598304
41. Nehrt N. L., Peterson T. A., Park D., Kann M. G.. **Domain landscapes of somatic mutations in cancer**. (2012) **13** p. S9. DOI: 10.1186/1471-2164-13-s4-s9
42. Akhoondi S., Sun D., von der Lehr N.. **FBXW7/hCDC4 is a general tumor suppressor in human cancer**. (2007) **67** 9006-9012. DOI: 10.1158/0008-5472.can-07-1320
43. Ojesina A. I., Lichtenstein L., Freeman S. S.. **Landscape of genomic alterations in cervical carcinomas**. (2014) **506** 371-375. DOI: 10.1038/nature12881
44. Lada A. G., Dhar A., Boissy R. J.. **AID/APOBEC cytosine deaminase induces genome-wide kataegis**. (2012) **7** p. 47. DOI: 10.1186/1745-6150-7-47
45. de Poot S. A. H., Bovenschen N., Granzyme M.. **Granzyme M: behind enemy lines**. (2014) **21** 359-368. DOI: 10.1038/cdd.2013.189
46. Shao D. D., Xue W., Krall E.. **KRAS and YAP1 converge to regulate EMT and tumor survival**. (2014) **158** 171-184. DOI: 10.1016/j.cell.2014.06.004
47. Scheau C., Badarau I. A., Costache R.. **The role of matrix metalloproteinases in the epithelial-mesenchymal transition of hepatocellular carcinoma**. (2019) **2019** 10. DOI: 10.1155/2019/9423907
48. Zhang S. M., Cai W. L., Liu X.. **KDM5B promotes immune evasion by recruiting SETDB1 to silence retroelements**. (2021) **598** 682-687. DOI: 10.1038/s41586-021-03994-2
49. Zhang Z. G., Zhang H. S., Sun H. L., Liu H. Y., Liu M. Y., Zhou Z.. **KDM5B promotes breast cancer cell proliferation and migration via AMPK-mediated lipid metabolism reprogramming**. (2019) **379** 182-190. DOI: 10.1016/j.yexcr.2019.04.006
50. Newman A. M., Liu C. L., Green M. R.. **Robust enumeration of cell subsets from tissue expression profiles**. (2015) **12** 453-457. DOI: 10.1038/nmeth.3337
51. Aran D., Hu Z., Butte A. J.. **xCell: digitally portraying the tissue cellular heterogeneity landscape**. (2017) **18** p. 220. DOI: 10.1186/s13059-017-1349-1
52. Becht E., Giraldo N. A., Lacroix L.. **Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression**. (2016) **17** p. 218. DOI: 10.1186/s13059-016-1070-5
53. Racle J., de Jonge K., Baumgaertner P., Speiser D. E., Gfeller D.. **Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data**. (2017) **6**. DOI: 10.7554/elife.26476
54. Yoshihara K., Shahmoradgoli M., Martinez E.. **Inferring tumour purity and stromal and immune cell admixture from expression data**. (2013) **4** p. 2612. DOI: 10.1038/ncomms3612
55. Zhao H., Wu L., Yan G.. **Inflammation and tumor progression: signaling pathways and targeted intervention**. (2021) **6** p. 263. DOI: 10.1038/s41392-021-00658-5
56. Wang F., Zhao Q., Wang Y. N.. **Evaluation of POLE and POLD1 mutations as biomarkers for immunotherapy outcomes across multiple cancer types**. (2019) **5** 1504-1506. DOI: 10.1001/jamaoncol.2019.2963
57. Luo Y., Wang H., Liu B., Wei J.. **Fatty acid metabolism and cancer immunotherapy**. (2022) **24** 659-670. DOI: 10.1007/s11912-022-01223-1
58. Swinnen J. V., Van Veldhoven P. P., Timmermans L.. **Fatty acid synthase drives the synthesis of phospholipids partitioning into detergent-resistant membrane microdomains**. (2003) **302** 898-903. DOI: 10.1016/s0006-291x(03)00265-1
59. Camarda R., Zhou A. Y., Kohnz R. A.. **Inhibition of fatty acid oxidation as a therapy for MYC-overexpressingtriple-negative breast cancer**. (2016) **22** 427-432. DOI: 10.1038/nm.4055
60. **Integrated genomic and molecular characterization of cervical cancer**. (2017) **543** 378-384. DOI: 10.1038/nature21386
61. Bandara A. B., Drake J. C., Brown D. A.. **Complex II subunit SDHD is critical for cell growth and metabolism, which can be partially restored with a synthetic ubiquinone analog**. (2021) **22** p. 35. DOI: 10.1186/s12860-021-00370-w
62. Kim T. E., Kim Y. W., Hwang S. Y.. **Candidate tumor suppressor, HCCS-1, is downregulated in human cancers and induces apoptosis in cervical cancer**. (2002) **97** 780-786. DOI: 10.1002/ijc.10124
63. Chu E. P., Elso C. M., Pollock A. H.. **Disruption of Serinc1, which facilitates serine-derived lipid synthesis, fails to alter macrophage function, lymphocyte proliferation or autoimmune disease susceptibility**. (2017) **82** 19-33. DOI: 10.1016/j.molimm.2016.12.007
64. Ahonen M. A., Horing M., Nguyen V. D.. **Insulin-inducible THRSP maintains mitochondrial function and regulates sphingolipid metabolism in human adipocytes**. (2022) **28** p. 68. DOI: 10.1186/s10020-022-00496-3
65. Howe L. R., Subbaramaiah K., Hudis C. A., Dannenberg A. J.. **Molecular pathways: adipose inflammation as a mediator of obesity-associated cancer**. (2013) **19** 6074-6083. DOI: 10.1158/1078-0432.ccr-12-2603
66. Kidd S., Spaeth E., Watson K.. **Origins of the tumor microenvironment: quantitative assessment of adipose-derived and bone marrow-derived stroma**. (2012) **7**. DOI: 10.1371/journal.pone.0030563
67. Iwano M., Plieth D., Danoff T. M., Xue C., Okada H., Neilson E. G.. **Evidence that fibroblasts derive from epithelium during tissue fibrosis**. (2002) **110** 341-350. DOI: 10.1172/jci0215518
68. Zeisberg E. M., Potenta S., Xie L., Zeisberg M., Kalluri R.. **Discovery of endothelial to mesenchymal transition as a source for carcinoma-associated fibroblasts**. (2007) **67** 10123-10128. DOI: 10.1158/0008-5472.can-07-3127
69. Nagasaki T., Hara M., Nakanishi H., Takahashi H., Sato M., Takeyama H.. **Interleukin-6 released by colon cancer-associated fibroblasts is critical for tumour angiogenesis: anti-interleukin-6 receptor antibody suppressed angiogenesis and inhibited tumour-stroma interaction**. (2014) **110** 469-478. DOI: 10.1038/bjc.2013.748
70. Yu Y., Xiao C. H., Tan L. D., Wang Q. S., Li X. Q., Feng Y. M.. **Cancer-associated fibroblasts induce epithelial-mesenchymal transition of breast cancer cells through paracrine TGF-beta signalling**. (2014) **110** 724-732. DOI: 10.1038/bjc.2013.768
71. Nieman K. M., Kenny H. A., Penicka C. V.. **Adipocytes promote ovarian cancer metastasis and provide energy for rapid tumor growth**. (2011) **17** 1498-1503. DOI: 10.1038/nm.2492
72. Takehara M., Sato Y., Kimura T.. **Cancer-associated adipocytes promote pancreatic cancer progression through SAA1 expression**. (2020) **111** 2883-2894. DOI: 10.1111/cas.14527
73. Cao Y.. **Angiogenesis and vascular functions in modulation of obesity, adipose metabolism, and insulin sensitivity**. (2013) **18** 478-489. DOI: 10.1016/j.cmet.2013.08.008
74. Avgerinos K. I., Spyrou N., Mantzoros C. S., Dalamaga M.. **Obesity and cancer risk: emerging biological mechanisms and perspectives**. (2019) **92** 121-135. DOI: 10.1016/j.metabol.2018.11.001
75. Campbell I. G., Russell S. E., Choong D. Y. H.. **Mutation of the PIK3CA gene in ovarian and breast cancer**. (2004) **64** 7678-7681. DOI: 10.1158/0008-5472.can-04-2933
76. Huang X., Liu G., Guo J., Su Z.. **The PI3K/AKT pathway in obesity and type 2 diabetes**. (2018) **14** 1483-1496. DOI: 10.7150/ijbs.27173
77. Shin J. W., Kim S., Ha S.. **The HER2 S310F mutant can form an active heterodimer with the EGFR, which can Be inhibited by cetuximab but not by trastuzumab as well as pertuzumab**. (2019) **9** p. 629. DOI: 10.3390/biom9100629
78. Bafna S., Kaur S., Batra S. K.. **Membrane-bound mucins: the mechanistic basis for alterations in the growth and survival of cancer cells**. (2010) **29** 2893-2904. DOI: 10.1038/onc.2010.87
79. Carraway K. L., Perez A., Idris N.. **Muc4/sialomucin complex, the intramembrane ErbB2 ligand, in cancer and epithelia: to protect and to survive**. (2002) **71** 149-185. DOI: 10.1016/s0079-6603(02)71043-x
80. Menendez J. A.. **Fine-tuning the lipogenic/lipolytic balance to optimize the metabolic requirements of cancer cell growth: molecular mechanisms and therapeutic perspectives**. (2010) **1801** 381-391. DOI: 10.1016/j.bbalip.2009.09.005
81. Chong C. R., Janne P. A.. **The quest to overcome resistance to EGFR-targeted therapies in cancer**. (2013) **19** 1389-1400. DOI: 10.1038/nm.3388
82. Timar J., Kashofer K.. **Molecular epidemiology and diagnostics of KRAS mutations in human cancer**. (2020) **39** 1029-1038. DOI: 10.1007/s10555-020-09915-5
83. Temko D., Tomlinson I. P. M., Severini S., Schuster-Bockler B., Graham T. A.. **The effects of mutational processes and selection on driver mutations across cancer types**. (2018) **9** p. 1857. DOI: 10.1038/s41467-018-04208-6
84. Van Allen E. M., Lui V. W. Y., Egloff A. M.. **Genomic correlate of exceptional erlotinib response in head and neck squamous cell carcinoma**. (2015) **1** 238-244. DOI: 10.1001/jamaoncol.2015.34
85. Ngan H. L., Poon P. H. Y., Su Y. X.. **Erlotinib sensitivity of MAPK1p.D321N mutation in head and neck squamous cell carcinoma**. (2020) **5** p. 17. DOI: 10.1038/s41525-020-0124-5
86. Ricoult S. J. H., Yecies J. L., Ben-Sahra I., Manning B. D.. **Oncogenic PI3K and K-Ras stimulate de novo lipid synthesis through mTORC1 and SREBP**. (2016) **35** 1250-1260. DOI: 10.1038/onc.2015.179
87. Gouw A. M., Eberlin L. S., Margulis K.. **Oncogene KRAS activates fatty acid synthase, resulting in specific ERK and lipid signatures associated with lung adenocarcinoma**. (2017) **114** 4300-4305. DOI: 10.1073/pnas.1617709114
88. Liu F., Zou Y., Wang F.. **FBXW7 mutations promote cell proliferation, migration, and invasion in cervical cancer**. (2019) **23** 409-417. DOI: 10.1089/gtmb.2018.0278
89. Yeh C. H., Bellon M., Nicot C.. **FBXW7: a critical tumor suppressor of human cancers**. (2018) **17** p. 115. DOI: 10.1186/s12943-018-0857-2
90. Forbes S. A., Beare D., Boutselakis H.. **COSMIC: somatic cancer genetics at high-resolution**. (2017) **45** D777-D783. DOI: 10.1093/nar/gkw1121
91. Helleday T., Eshtad S., Nik-Zainal S.. **Mechanisms underlying mutational signatures in human cancers**. (2014) **15** 585-598. DOI: 10.1038/nrg3729
92. Ma X., Dong L., Liu X., Ou K., Yang L.. **POLE/POLD1 mutation and tumor immunotherapy**. (2022) **41** p. 216. DOI: 10.1186/s13046-022-02422-1
93. Leon-Castillo A., Britton H., McConechy M. K.. **Interpretation of somatic POLE mutations in endometrial carcinoma**. (2020) **250** 323-335. DOI: 10.1002/path.5372
94. Levine D. A.. **Integrated genomic characterization of endometrial carcinoma**. (2013) **497** 67-73. DOI: 10.1038/nature12113
95. Wang Z., Yin M., Chu P., Lou M.. **STAT3 inhibitor sensitized KRAS-mutant lung cancers to RAF inhibitor by activating MEK/ERK signaling pathway**. (2019) **11** 7187-7196. DOI: 10.18632/aging.102244
96. Huang X., Zhao Q., Yang P.. **Metabolic syndrome and risk of cervical human papillomavirus incident and persistent infection**. (2016) **95**. DOI: 10.1097/md.0000000000002905
97. Baker R., Dauner J. G., Rodriguez A. C.. **Increased plasma levels of adipokines and inflammatory markers in older women with persistent HPV infection**. (2011) **53** 282-285. DOI: 10.1016/j.cyto.2010.11.014
98. Kemp T. J., Hildesheim A., Garcia-Pineres A.. **Elevated systemic levels of inflammatory cytokines in older women with persistent cervical human papillomavirus infection**. (2010) **19** 1954-1959. DOI: 10.1158/1055-9965.epi-10-0184
|
---
title: 'Effect of Self-Directed Home Therapy Adherence Combined with TheraBracelet
on Poststroke Hand Recovery: A Pilot Study'
authors:
- Gabrielle Scronce
- Viswanathan Ramakrishnan
- Amanda A. Vatinno
- Na Jin Seo
journal: Stroke Research and Treatment
year: 2023
pmcid: PMC10017223
doi: 10.1155/2023/3682898
license: CC BY 4.0
---
# Effect of Self-Directed Home Therapy Adherence Combined with TheraBracelet on Poststroke Hand Recovery: A Pilot Study
## Abstract
Hand impairment is a common consequence of stroke, resulting in long-term disability and reduced quality of life. Recovery may be augmented through self-directed therapy activities at home, complemented by the use of rehabilitation devices such as peripheral sensory stimulation. The objective of this study was to determine the effect of adherence to self-directed therapy and the use of TheraBracelet (subsensory random-frequency vibratory stimulation) on hand function for stroke survivors. In a double-blind, randomized controlled pilot trial, 12 chronic stroke survivors were assigned to a treatment or control group ($$n = 6$$/group). All participants were instructed to perform 200 repetitions of therapeutic hand tasks 5 days/week while wearing a wrist-worn device 8 hours/day for 4 weeks. The treatment group received TheraBracelet vibration from the device, while the control group received no vibration. Home task repetition adherence and device wear logs, as well as hand function assessment (Stroke Impact Scale Hand domain), were obtained weekly. Repetition adherence was comparable between groups but varied among participants. Participants wore the device to a greater extent than adhering to completing repetitions. A linear mixed model analysis showed a significant interaction between repetition and group ($$p \leq 0.01$$), with greater adherence resulting in greater hand function change for the treatment group ($r = 0.94$; R2 = 0.88), but not for the control group. Secondary analysis revealed that repetition adherence was greater for those with lower motor capacity and greater self-efficacy at baseline. This pilot study suggests that adherence to self-directed therapy at home combined with subsensory stimulation may affect recovery outcomes in stroke survivors. This trial is registered with NCT04026399.
## 1. Introduction
Stroke is a major medical event that occurs in nearly 800,000 people in the United States each year [1]. Upper extremity (UE) sensorimotor impairment is a common consequence of stroke, affecting $77\%$ of stroke survivors [2]. UE sensorimotor impairment decreases individuals' ability to perform functional activities for self-care, hygiene, employment, and recreation, thereby diminishing their independence and quality of life [3, 4].
Research shows that extensive practice of task-specific activities results in improved functional recovery of the UE poststroke [5–9]. However, the high amount of UE activity necessary for functional recovery [5] cannot be achieved within typical therapy sessions [7, 10–12]. To circumvent the limited time available with a therapist, a home exercise program (HEP) consisting of self-directed therapeutic activity is commonly prescribed as part of the standard therapy [13]. However, adherence to HEP varies substantially among patients [13–16]. Varying adherence levels have been shown to explain the variability in recovery of overall physical mobility post-stroke [17–19]. However, a relationship has not been studied between UE HEP adherence and UE functional outcome.
HEP can be complemented by rehabilitation devices, such as a peripheral sensory stimulation device [20]. Meta-analysis has shown that the use of peripheral sensory stimulation along with UE therapy can the increase functional recovery of the UE [21]. Previously used peripheral sensory stimulation is typically at a suprathreshold level. Suprathreshold stimulation applied before each UE therapy session [21] lengthens the treatment durations and can lower patient adherence [20, 22]. Suprathreshold stimulation applied during therapy sessions can interfere with sensory feedback required for manipulation of objects. Therefore, a new stimulation device named TheraBracelet [23] (Figure 1) has been proposed. TheraBracelet is imperceptible random-frequency vibration applied to the wrist skin [23]. TheraBracelet does not interfere with UE hand tasks since the stimulation is imperceptible and delivered via a small device worn on the wrist like a watch [24–26]. TheraBracelet vibration stimulates mechanoreceptors in the skin and subsequently afferent neurons [27, 28], thereby adding small random currents to neurons in the sensorimotor cortex [29]. These small random currents trigger coherent [30] neuronal firing [29, 31, 32] during the performance of hand tasks and enhance neural communication [33, 34] required for hand tasks [35–39] via stochastic facilitation [35]. As a result, TheraBracelet has demonstrated the potential to improve finger touch sensation [26, 36, 38] and dexterity [23, 37, 40, 41], as well as functional recovery [23, 40].
Preliminary efficacy of TheraBracelet has been examined in the laboratory setting [23, 40]. However, the efficacy of using TheraBracelet in conjunction with HEP in stroke survivors' homes has not been examined. As the logical next step, the objective of this pilot study was to determine the effect of adherence to UE HEP combined with TheraBracelet on hand function for stroke survivors. Our hypothesis was that the combination of greater adherence to UE HEP and receiving TheraBracelet stimulation would result in greater improvement in hand function [42].
## 2.1. Participants
Inclusion criteria were adult (age ≥18 years), chronic stroke survivors (≥6 months post-stroke) with tactile sensory deficits of the fingertips (Monofilament [43] score>2.83, 2-Point Discrimination Test [44] score>5 mm, or sense of numbness based on verbal report). Additional inclusion criteria were the ability to put on a watch daily (with or without caregiver help) and ability to move objects with the paretic hand as necessary to perform HEP. Exclusion criteria included complete upper limb deafferentation, rigidity (Modified Ashworth Scale [45] = 4), UE botulinum toxin injection within 3 months prior to enrollment or during enrollment, brainstem stroke, comorbidity (peripheral neuropathy, orthopedic conditions in the hand that limit motion [46], premorbid neurologic conditions, compromised skin integrity of the hand/wrist unrelated to stroke, such as from long term use of blood thinners), change in neurological disorder medications during enrollment, concurrent upper extremity rehabilitation therapy, and language barrier or cognitive impairment that precluded following instructions or providing consent. All participants signed a consent form that was approved by the Medical University of South Carolina (MUSC) Institutional Review Board before participation in the study.
## 2.2. Study Design
This was a pilot, double-blind, randomized controlled trial in which chronic stroke survivors were randomly assigned to a treatment or control group. All participants were instructed to perform HEP consisting of 200 repetitions of therapeutic UE tasks 5 days/week for 4 weeks. In addition, participants were provided a wearable prototype device for TheraBracelet, which has been shown to be successfully worn by stroke survivors in their homes every day during daily living without safety issues [25]. All participants were instructed to wear the device at least 8 hours/day every day for 4 weeks, consistent with the previous study [25]. While the control group received no vibration from the device, the treatment group received subsensory TheraBracelet vibration from the device. The device, which is further described in Seo et al. [ 25], provided continuous vibration at $60\%$ of the participant's sensory threshold, determined at each visit. To ensure blinding of researchers, different research personnel administered the device to the participant, who was not the research therapist who administered HEP and assessments.
## 2.3. Home Exercise Program (HEP)
Each week, beginning with the baseline visit, participants met individually with an occupational therapist and were administered HEP with specific tasks to practice at home for the following week. Tasks were selected from a menu of task practice activities that was developed by two experienced occupational therapists based on the EXCITE trial [47] manual and the task-specific training text by Lang and Birkenmeier [48]. Tasks included self-care, household, leisure, and vocational tasks and were separated into two types: tasks requiring [1] primarily in-hand manipulation and [2] reaching. The participant and therapist collaboratively selected 2 tasks of each type that the participant considered meaningful to perform during the week. The selection from a menu ensured consistency while allowing saliency of selected tasks to increase motivation for the participant to complete the task [47]. When possible, tasks were selected to be more challenging than previous weeks' tasks so that the intervention was both individualized and progressive [11, 48]. In addition, the therapist provided written options to grade each task to make it easier or harder at the participant's discretion to allow the participant to be challenged but not overwhelmed by the task for optimal neural plasticity [48]. See Table 1 for task and grading examples.
Participants were instructed to complete each of the 4 selected tasks 50 times per day, 5 days per week so that participants would complete 4000 repetitions of task-specific practice over the 4-week intervention. This dose was selected because it was considered feasible for home-based, self-directed practice within a 1-2 hour timeframe [11, 49, 50] while corresponding to the lower end of repetitions that have been shown to promote neural plasticity and functional recovery in animal and motor learning studies [11, 51, 52].
To facilitate adherence, a transfer package [53] was implemented that included a contract for adherence [54], a written log to track HEP adherence, and problem solving to overcome barriers to completing HEP [55–57]. In the contract, the participant agreed to adhere to the intervention including completing all HEP assignments and using the paretic hand on specific activities of daily living as much as possible outside the lab. The contract was signed by the participant and therapist to emphasize its importance [54]. To track HEP adherence, participants were provided a paper log to record the number of repetitions completed for each of their prescribed HEP tasks for each day. At the weekly meetings, if HEP adherence according to the written log was less than $100\%$, the therapist facilitated a discussion with the participant to help them think through barriers to completing HEP and ways to overcome them [55–57].
## 2.4. Assessments
At the weekly meetings, HEP adherence, device wear logs, and hand function assessment were obtained. HEP adherence and device wear information was obtained from the paper log in which participants recorded the number of repetitions for each task they completed as well as the time they put on and took off the device for each day. Average percent HEP adherence was defined as the percentage of HEP repetitions completed out of the number prescribed.
Hand function, the primary outcome, was assessed by the Stroke Impact Scale (SIS) Hand domain [58, 59]. The SIS was used because it is a stroke-specific, self-report measure with high test-retest reliability, concurrent validity, and responsiveness to change [59, 60] and because this assessment could be administered by phone during COVID-19 quarantine when in-person visits were restricted.
To characterize the participant pool, demographic information was obtained at baseline. Baseline assessment also included motor function and self-efficacy, as they may affect HEP adherence and functional recovery [61, 62]. In particular, baseline motor capacity was measured by Box and Blocks Test (BBT), a functional performance test of upper limb motor capacity with high validity, test-retest reliability, ability to detect change, and clinical utility [63–65]. The BBT score represents the number of blocks moved in one minute [63] with the affected hand. For self-efficacy, we implemented a 4-item, self-report measure tailored to the language of UE rehabilitation therapy (see appendix) [66–68]. Specifically, participant's knowledge and confidence in taking responsibility for their UE treatment were scored on a Likert scale from 1 (disagree strongly) to 4 (agree strongly).
Adverse events were explicitly asked and recorded at each weekly meeting. To assess maintenance of blinding of participants, a questionnaire was administered at the end of the study. It asked whether participants had felt the device vibrating during the past month, and if they did how long they felt the vibration.
## 2.5. Analysis
Baseline characteristics were summarized using means and standard deviations (SD) for continuous variables and numbers and percentages for categorical variables. As preliminary analysis, we examined the group difference in adherence using t-test for continuous variables and Fisher's exact test for categorical variables. Participants' adherence to HEP and device wear level were also summarized using means and SD. For adherence, average percent HEP adherence (the percentage of prescribed repetitions completed each day averaged over the total duration of participation) was used to represent the person's mean adherence level that is not influenced by dropouts. We also compared individuals' HEP adherence level with device wear level using paired t-test. Similarly, average percent device wear was used for this summary.
For the primary analysis, the SIS change from baseline at each week was the dependent variable. A linear mixed model with group, adherence to HEP, and time (week) along with their interaction as independent variables was performed. Adherence to HEP was quantified as the cumulative number of repetitions completed by each week. To account for within subject repeated measures, a subject level random intercept term was included. Other structures for the within subject correlation were also examined. PROC GLIMMIX in SAS was used for the analysis. Model diagnostics were used to verify normality and model adequacy. In addition, as secondary analysis to explore factors affecting adherence, we examined Pearson correlations between average percent HEP adherence and baseline motor function and self-efficacy.
## 3.1. Participants
Twelve chronic stroke survivors with mean ± SD age of 61 ± 10 years participated in the study (see Figure 2 for CONSORT diagram). Baseline characteristics were similar between the two groups (Table 2). During the study period, there were no adverse events reported by participants in the treatment group. However, one participant in the control group reported increased pain, tone, and stiffness with HEP. As for blinding, three participants reported feeling vibration briefly. Two were in the treatment group, and one was in the control group.
## 3.2. HEP Adherence
The average percent adherence to HEP was similar between the two groups (71 ± $39\%$ and 77 ± $40\%$ for the treatment and control group, respectively). However, the average percent adherence to HEP varied across participants, ranging from $7\%$ to $119\%$. Eight of the 12 participants did not meet the prescribed HEP of 200 repetitions per day, 5 days per week.
## 3.3. Device Wear
Device wear was similar between groups (treatment 108 ± $36\%$, control 129 ± $34\%$). Ten participants wore the device as instructed, for at least 8 hours per day. The other two participants, one in each group, still wore the device on average 6 and 9 hours per day on the days that s/he performed HEP. Participants wore the device as instructed more than they adhered to HEP (119 ± $35\%$ for device wear vs. 74 ± $38\%$ for HEP adherence). Since the device was worn for longer durations than HEP, device wear was not included as a covariate in the primary analysis for the hand function outcome below.
## 3.4. Hand Function Outcome
SIS Hand domain scores were comparable between the two groups at baseline (mean ± SD for the treatment group = 69.2 ± 33.4, control group = 61.7 ± 35.0; $$p \leq 0.71$$). The primary linear mixed model analysis showed that the change in SIS Hand was affected by the HEP adherence differently for the two groups ($$p \leq 0.01$$ for the interaction effect). Specifically, greater weekly cumulative repetitions in HEP resulted in greater improvement of SIS Hand in the treatment group (Pearson $r = 0.94$; R2 = 0.88; $p \leq .001$), while there was no improvement observed in the control group (Pearson r = −0.18; R2 = 0.03; $$p \leq .45$$) (Figure 3). Final SIS Hand domain scores were 76.7 ± 28.9 for the treatment group and 52.9 ± 36.4 for the control group.
## 3.5. Factors for Adherence
While adherence was similar between groups, HEP adherence was greater for participants with lower motor capacity measured by BBT (Figure 4(a)). In addition, HEP adherence was greater for those with greater self-efficacy (Figure 4(b)).
## 4. Discussion
The aim of this pilot study was to determine the effects of adherence to HEP combined with the use of TheraBracelet on hand function for stroke survivors. There was a statistically significant interaction between groups and HEP adherence for hand function measured by the SIS Hand. Greater HEP adherence combined with TheraBracelet treatment resulted in increased perceived hand function (Figure 3). This interaction effect was statistically significant in the weekly analysis. In a posthoc analysis examining only the baseline to post changes, similar trends were observed. This trend supports the need for a future study to investigate the effect among a larger sample and over a longer intervention duration or greater dosage as discussed below.
Findings from this study are in concordance with findings from previous studies supporting the use of subsensory vibration to improve upper extremity motor recovery after stroke [21, 23, 36, 37]. This study expands upon previous knowledge [25] by demonstrating that not only wearing a device delivering TheraBracelet stimulation daily is feasible for stroke survivors [24, 25] but also the addition of HEP to TheraBracelet is an important component of improving hand function. Furthermore, while the previous research showed efficacy of TheraBracelet with laboratory-based task practice therapy sessions [23], this study suggests that an independently performed, home-based exercise program combined with TheraBracelet could improve the hand function. Results from the present study encourage a larger study adequately powered to determine the efficacy of TheraBracelet combined with a HEP to improve hand function.
The mean change in hand function measured by SIS Hand domain did not exceed the minimal detectable change (MDC = 17.8) or minimal clinically important difference (MCID = 25.9) [69]. Previous research showed that lab-based therapy with TheraBracelet showed progressive improvement in hand function week by week over the 6-week period, resulting in clinically meaningful improvement in the SIS Hand and Activities of Daily Living domains [23]. With the overall literature supporting greater treatment dose leading to greater improvement [5–9], a longer intervention duration, greater dosage, or higher adherence may be necessary to create a change that is clinically significant.
Previous research suggests adherence is affected by psychosocial factors, including self-efficacy, as was found in this study [62, 70]. Specifically, interviews showed stroke survivors perceived self-efficacy as an important factor for participating in daily physical activity [70], and a prospective study of older adults with recent stroke demonstrated that those with high self-efficacy had greater improvements in balance than those with low self-efficacy [62]. These findings indicate that self-efficacy influences adherence to activity, and the current study supports the importance of self-efficacy in adherence to UE HEP among stroke survivors. As a result, self-efficacy and other psychosocial factors should be investigated to include in the development of behavioral interventions to increase HEP adherence and improve motor recovery after stroke [71, 72].
In this study, we found that lower motor capacity at baseline, measured by the BBT, was associated with greater HEP adherence. Previous literature showed that improving health conditions and functional abilities are strong motivators to exercise [73, 74]. Individuals with lower motor capacity may be more motivated to adhere to HEP because of the desire to improve. Clinicians helping patients increase adherence to HEP may emphasize the importance of HEP adherence for patients to improve functional recovery.
## 4.1. Limitations
A major limitation of this study was the necessity of using self-reported measures for both HEP adherence and hand function. Measuring HEP adherence by participant self-report is known to introduce inaccuracy [14–16]. Accuracy in the measurement of adherence could be improved through the development of technology to objectively measure the UE activity of patients during their activities of daily living [75]. Additionally, hand function was measured by the self-reported SIS Hand domain instead of in-person objective assessments due to COVID-19 lockdown. While the SIS Hand domain provides insight into participants' perceived hand function, objective physical performance outcomes would provide more objective and clinically meaningful data. Additionally, the sample size for this study was small. Future studies will require a larger sample to adequately determine effects in objective functional performance measures.
## 5. Conclusion
This pilot study suggests that adherence to self-directed therapy at home combined with subsensory TheraBracelet stimulation may improve upper extremity recovery outcomes in stroke survivors. The clinical implication of these findings is increased need to effectively promote adherence to prescribed HEP. Additionally, more research is indicated to investigate the effectiveness of TheraBracelet in facilitating recovery among a larger sample of stroke survivors.
## Data Availability
Data is available upon request to the corresponding author, Dr. Gabrielle Scronce, at [email protected].
## Ethical Approval
The study protocol was approved by the Institutional Review Board at the Medical University of South Carolina (approval id. Pro00086270).
## Conflicts of Interest
The authors declare the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: N. J. Seo is an inventor of a patent regarding the investigated sensory stimulation. The other authors report no conflicts of interest.
## References
1. Virani S. S., Alonso A., Benjamin E. J.. **Heart Disease and Stroke Statistics—2020 Update: A Report From the American Heart Association**. (2020) **141**. DOI: 10.1161/CIR.0000000000000757
2. Lawrence E. S., Coshall C., Dundas R.. **Estimates of the prevalence of acute stroke impairments and disability in a multiethnic population**. (2001) **32** 1279-1284. DOI: 10.1161/01.STR.32.6.1279
3. Raghavan P.. **Upper Limb Motor Impairment After Stroke**. (2015) **26** 599-610. DOI: 10.1016/j.pmr.2015.06.008
4. Dobkin B. H.. **Rehabilitation after stroke**. (2014) **352** 1677-1684. DOI: 10.1093/med/9780199641208.003.0021
5. Schaefer S. Y., Patterson C. B., Lang C. E.. **Transfer of training between distinct motor tasks after stroke**. (2013) **27** 602-612. DOI: 10.1177/1545968313481279
6. Lohse K. R., Lang C. E., Boyd L. A.. **Is more better? Using metadata to explore dose–response relationships in stroke rehabilitation**. (2014) **45** 2053-2058. DOI: 10.1161/STROKEAHA.114.004695
7. Winstein C., Kim B., Kim S., Martinez C., Schweighofer N.. **Dosage matters**. (2019) **50** 1831-1837. DOI: 10.1161/STROKEAHA.118.023603
8. Langhorne P., Coupar F., Pollock A.. **Motor recovery after stroke: a systematic review**. (2009) **8** 741-754. DOI: 10.1016/S1474-4422(09)70150-4
9. Veerbeek J. M., van Wegen E., van Peppen R.. **What is the evidence for physical therapy poststroke? A systematic review and meta-analysis**. (2014) **9**. DOI: 10.1371/journal.pone.0087987
10. Lang C. E., Macdonald J. R., Reisman D. S.. **Observation of amounts of movement practice provided during stroke rehabilitation**. (2009) **90** 1692-1698. DOI: 10.1016/j.apmr.2009.04.005
11. Lang C. E., Strube M. J., Bland M. D.. **Dose response of task-specific upper limb training in people at least 6 months poststroke: a phase II, single-blind, randomized, controlled trial**. (2016) **80** 342-354. DOI: 10.1002/ana.24734
12. Schneider E. J., Lannin N. A., Ada L., Schmidt J.. **Increasing the amount of usual rehabilitation improves activity after stroke: a systematic review**. (2016) **62** 182-187. DOI: 10.1016/j.jphys.2016.08.006
13. Miller K. K., Porter R. E., DeBaun-Sprague E., Van Puymbroeck M., Schmid A. A.. **Exercise after stroke: patient adherence and beliefs after discharge from rehabilitation**. (2017) **24** 142-148. DOI: 10.1080/10749357.2016.1200292
14. Waddell K. J., Lang C. E.. **Comparison of self-report versus sensor-based methods for measuring the amount of upper limb activity outside the clinic**. (2018) **99** 1913-1916. DOI: 10.1016/j.apmr.2017.12.025
15. Mahmood A., Solomon J. M., English C., Bhaskaran U., Menon G., Manikandan N.. **Measurement of adherence to home-based exercises among community-dwelling stroke survivors in India**. (2020) **25**. DOI: 10.1002/pri.1827
16. Levy T., Laver K., Killington M., Lannin N., Crotty M.. **A systematic review of measures of adherence to physical exercise recommendations in people with stroke**. (2019) **33** 535-545. DOI: 10.1177/0269215518811903
17. Duncan P. W., Horner R. D., Reker D. M.. **Adherence to postacute rehabilitation guidelines is associated with functional recovery in stroke**. (2002) **33** 167-178. DOI: 10.1161/hs0102.101014
18. Gunnes M., Indredavik B., Langhammer B.. **Associations between adherence to the physical activity and exercise program applied in the LAST study and functional recovery after stroke**. (2019) **100** 2251-2259. DOI: 10.1016/j.apmr.2019.04.023
19. Van De Port I. G. L., Kwakkel G., Van Wijk I., Lindeman E.. **Susceptibility to deterioration of mobility long-term after stroke: a prospective cohort study**. (2006) **37** 167-171. DOI: 10.1161/01.STR.0000195180.69904.f2
20. Morrow C. M., Johnson E., Simpson K. N., Seo N. J.. **Determining factors that influence adoption of new post-stroke sensorimotor rehabilitation devices in the USA**. (2021) **29** 1213-1222. DOI: 10.1109/TNSRE.2021.3090571
21. Conforto A. B., dos Anjos S. M., Bernardo W. M.. **Repetitive peripheral sensory stimulation and upper limb performance in stroke: a systematic review and meta-analysis**. (2018) **32** 863-871. DOI: 10.1177/1545968318798943
22. Jin J., Sklar G. E., VMS O., Li S. C.. **Factors affecting therapeutic compliance: a review from the patient’s perspective**. (2008) **4** 269-286. DOI: 10.2147/tcrm.s1458
23. Seo N. J., Woodbury M. L., Bonilha L.. **TheraBracelet stimulation during task-practice therapy to improve upper extremity function after stroke: a pilot randomized controlled study**. (2019) **99** 319-328. DOI: 10.1093/ptj/pzy143
24. Lakshminarayanan K., Wang F., Webster J. G., Seo N. J.. **Feasibility and usability of a wearable orthotic for stroke survivors with hand impairment**. (2017) **12** 175-183. DOI: 10.3109/17483107.2015.1111945
25. Seo N. J., Enders L. R., Fortune A.. **Phase I safety trial: extended daily peripheral sensory stimulation using a wrist-worn vibrator in stroke survivors**. (2020) **11** 204-213. DOI: 10.1007/s12975-019-00724-9
26. Wang F., Lakshminarayanan K., Slota G. P., Seo N. J., Webster J. G.. **An MRI-compatible hand sensory vibrotactile system**. (2015) **36** N15-N21. DOI: 10.1088/0967-3334/36/1/N15
27. Vallbo A. B., Johansson R. S.. **Properties of cutaneous mechanoreceptors in the human hand related to touch sensation**. (1984) **3** 3-14. PMID: 6330008
28. Vallbo Å. B.. **Microneurography: how it started and how it works**. (2018) **120** 1415-1427. DOI: 10.1152/jn.00933.2017
29. Seo N. J., Lakshminarayanan K., Lauer A. W.. **Use of imperceptible wrist vibration to modulate sensorimotor cortical activity**. (2019) **237** 805-816. DOI: 10.1007/s00221-018-05465-z
30. Ward L. M.. **Physics of neural synchronisation mediated by stochastic resonance**. (2009) **50** 563-574. DOI: 10.1080/00107510902879246
31. Seo N. J., Lakshminarayanan K., Bonilha L., Lauer A. W., Schmit B. D.. **Effect of imperceptible vibratory noise applied to wrist skin on fingertip touch evoked potentials - an EEG study**. (2015) **3**. DOI: 10.14814/phy2.12624
32. Moss F., Ward L. M., Sannita W. G.. **Stochastic resonance and sensory information processing: a tutorial and review of application**. (2004) **115** 267-281. DOI: 10.1016/j.clinph.2003.09.014
33. Ward L. M., MacLean S. E., Kirschner A.. **Stochastic resonance modulates neural synchronization within and between cortical sources**. (2010) **5**. DOI: 10.1371/journal.pone.0014371
34. Fries P.. **Rhythms for cognition: communication through coherence**. (2015) **88** 220-235. DOI: 10.1016/j.neuron.2015.09.034
35. Collins J. J., Imhoff T. T., Grigg P.. **Noise-enhanced tactile sensation**. (1996) **383** 770-770. DOI: 10.1038/383770a0
36. Enders L. R., Hur P., Johnson M. J., Seo N. J.. **Remote vibrotactile noise improves light touch sensation in stroke survivors’ fingertips via stochastic resonance**. (2013) **10** p. 105. DOI: 10.1186/1743-0003-10-105
37. Seo N. J., Kosmopoulos M. L., Enders L. R., Hur P.. **Effect of remote sensory noise on hand function post stroke**. (2014) **8** 1-9. DOI: 10.3389/fnhum.2014.00934
38. Lakshminarayanan K., Lauer A. W., Ramakrishnan V., Webster J. G., Seo N. J.. **Application of vibration to wrist and hand skin affects fingertip tactile sensation**. (2015) **3**. DOI: 10.14814/phy2.12465
39. Schranz C., Vatinno A., Ramakrishnan V., Seo N. J.. **Neuroplasticity after upper-extremity rehabilitation therapy with sensory stimulation in chronic stroke survivors**. (2022) **4**. DOI: 10.1093/braincomms/fcac191
40. Vatinno A. A., Hall L., Cox H.. **Using subthreshold vibratory stimulation during poststroke rehabilitation therapy: a case series**. (2022) **42** 30-39. DOI: 10.1177/15394492211042275
41. Hur P., Wan Y. H., Seo N. J.. **Investigating the role of vibrotactile noise in early response to perturbation**. (2014) **61** 1628-1633. DOI: 10.1109/TBME.2013.2294672
42. Scronce G., Seo N. J., Ramakrishnan V.. (2023)
43. Bell-Krotoski J., Tomancik E.. **The repeatability of testing with Semmes-Weinstein monofilaments**. (1987) **12** 155-161. DOI: 10.1016/S0363-5023(87)80189-2
44. Bell-Krotoski J., Weinstein S., Weinstein C.. **Testing sensibility, including touch-pressure, two-point discrimination, point localization, and vibration**. (1993) **6** 114-123. DOI: 10.1016/S0894-1130(12)80292-4
45. Bohannon R. W., Smith M. B.. **Interrater reliability of a modified Ashworth scale of muscle spasticity**. (1987) **67** 206-207. DOI: 10.1093/ptj/67.2.206
46. Seo N. J., Sindhu B. S., Shechtman O.. **Influence of pain associated with musculoskeletal disorders on grip force timing**. (2011) **24** 335-344. DOI: 10.1016/j.jht.2011.06.004
47. Wolf S. L., Winstein C. J., Miller J. P.. **Effect of constraint-induced movement therapy on upper extremity function 3 to 9 months after stroke**. (2006) **296** 2095-2104. DOI: 10.1001/jama.296.17.2095
48. Lang C. E., Birkenmeier R. L.. (2013). DOI: 10.7139/2017.978-1-56900-440-1
49. Waddell K. J., Birkenmeier R. L., Moore J. L., Hornby T. G., Lang C. E.. **Feasibility of high-repetition, task-specific training for individuals with upper-extremity paresis**. (2014) **68** 444-453. DOI: 10.5014/ajot.2014.011619
50. Birkenmeier R. L., Prager E. M., Lang C. E.. **Translating animal doses of task-specific training to people with chronic stroke in 1-hour therapy sessions: a proof-of-concept study**. (2010) **24** 620-635. DOI: 10.1177/1545968310361957
51. Nudo R. J., Milliken G. W., Jenkins W. M., Merzenich M. M.. **Use-dependent alterations of movement representations in primary motor cortex of adult squirrel monkeys**. (1996) **16** 785-807. DOI: 10.1523/jneurosci.16-02-00785.1996
52. Carey J. R., Kimberley T. J., Lewis S. M.. **Analysis of fMRI and finger tracking training in subjects with chronic stroke**. (2002) **125** 773-788. DOI: 10.1093/brain/awf091
53. Taub E., Uswatte G., Mark V. W.. **Method for enhancing real-world use of a more affected arm in chronic stroke**. (2013) **44** 1383-1388. DOI: 10.1161/STROKEAHA.111.000559
54. Eyberg S. M., Johnson S. M.. **Multiple assessment of behavior modification with families: effects of contingency contracting and order of treated problems**. (1974) **42** 594-606. DOI: 10.1037/h0036723
55. Wade S. L., Michaud L., Brown T. M.. **Putting the pieces together**. (2006) **21** 57-67. DOI: 10.1097/00001199-200601000-00006
56. Devellis B. M., Blalock S. J., Hahn P. M., Devellis R. F., Hochbaum G. M.. **Evaluation of a problem-solving intervention for patients with arthritis**. (1988) **11** 29-42. DOI: 10.1016/0738-3991(88)90074-2
57. Glasgow R. E., Toobert D. J., Barrera M., Strycker L. A.. **Assessment of problem-solving: a key to successful diabetes self-management**. (2004) **27** 477-490. DOI: 10.1023/B:JOBM.0000047611.81027.71
58. Duncan P. W., Bode R. K., Min Lai S., Perera S.. **Rasch analysis of a new stroke-specific outcome scale: the stroke impact scale**. (2003) **84** 950-963. DOI: 10.1016/S0003-9993(03)00035-2
59. Duncan P. W., Wallace D., Lai S. M., Johnson D., Embretson S., Laster L. J.. **The stroke impact scale version 2.0**. (1999) **30** 2131-2140. DOI: 10.1161/01.str.30.10.2131
60. Lin K. C., Fu T., Wu C. Y., Hsieh Y. W., Chen C. L., Lee P. C.. **Psychometric comparisons of the stroke impact scale 3.0 and stroke-specific quality of life scale**. (2010) **19** 435-443. DOI: 10.1007/s11136-010-9597-5
61. Bailey R. R.. **Self-efficacy, self-regulation, social support, and outcomes expectations for daily physical activity in adults with chronic stroke: a descriptive, exploratory study**. (2019) **33** 129-141. DOI: 10.1080/07380577.2018.1558326
62. Hellström K., Lindmark B., Wahlberg B., Fugl-Meyer A. R.. **Self-efficacy in relation to impairments and activities of daily living disability in elderly patients with stroke: a prospective investigation**. (2003) **35** 202-207. DOI: 10.1080/16501970310000836
63. Mathiowetz V., Weber K.. **Adult norms for the box and block**. (1985) **39** 387-391
64. Connell L. A., Tyson S. F.. **Clinical reality of measuring upper-limb ability in neurologic conditions: a systematic review**. (2012) **93** 221-228. DOI: 10.1016/j.apmr.2011.09.015
65. Slota G. P., Enders L. R., Seo N. J.. **Improvement of hand function using different surfaces and identification of difficult movement post stroke in the box and block test**. (2014) **45** 833-838. DOI: 10.1016/j.apergo.2013.10.014
66. Kanter J. W., Mulick P. S., Busch A. M., Berlin K. S., Martell C. R.. **The behavioral activation for depression scale (BADS): psychometric properties and factor structure**. (2007) **29** 191-202. DOI: 10.1007/s10862-006-9038-5
67. Hibbard J. H., Stockard J., Mahoney E. R., Tusler M.. **Development of the patient activation measure (PAM): conceptualizing and measuring activation in patients and consumers**. (2004) **39, 4, Part 1** 1005-1026. DOI: 10.1111/j.1475-6773.2004.00269.x
68. Do V., Young L., Barnason S., Tran H.. **Relationships between activation level, knowledge, self-efficacy, and self-management behavior in heart failure patients discharged from rural hospitals**. (2015) **4** p. 150. DOI: 10.12688/f1000research.6557.1
69. Lin K. C., Fu T., Wu C. Y.. **Minimal detectable change and clinically important difference of the stroke impact scale in stroke patients**. (2010) **24** 486-492. DOI: 10.1177/1545968309356295
70. Bailey R.. **Examining daily physical activity in community-dwelling adults with stroke using social cognitive theory: an exploratory, qualitative study**. (2020) **42** 2631-2639. DOI: 10.1080/09638288.2019.1568591
71. Dobkin B. H.. **Behavioral self-management strategies for practice and exercise should be included in neurologic rehabilitation trials and care**. (2016) **29** 693-699. DOI: 10.1097/WCO.0000000000000380
72. Winstein C. J., Stein J., Arena R.. **Guidelines for Adult Stroke Rehabilitation and Recovery: A Guideline for Healthcare Professionals from the American Heart Association/American Stroke Association**. (2016) **47**. DOI: 10.1161/STR.0000000000000098
73. Newson R. S., Kemps E. B.. **Factors that promote and prevent exercise engagement in older adults**. (2007) **19** 470-481. DOI: 10.1177/0898264307300169
74. Jurkiewicz M. T., Marzolini S., Oh P.. **Adherence to a home-based exercise program for individuals after stroke**. (2011) **18** 277-284. DOI: 10.1310/tsr1803-277
75. Lee S. I., Adans-Dester C. P., Grimaldi M.. **Enabling stroke rehabilitation in home and community settings: a wearable sensor-based approach for upper-limb motor training**. (2018) **6** 1-11. DOI: 10.1109/JTEHM.2018.2829208
|
---
title: Dysregulation of miR-155 Expression in Professional Mixed Martial Arts (MMA)
Fighters
journal: Cureus
year: 2023
pmcid: PMC10017279
doi: 10.7759/cureus.34944
license: CC BY 3.0
---
# Dysregulation of miR-155 Expression in Professional Mixed Martial Arts (MMA) Fighters
## Abstract
Psychological and physical stress can induce dysregulation of gene expression via changes in DNA methylation and microRNA (miRNA) expression. Such epigenetic modifications are yet to be investigated in professional Mixed Martial Arts (MMA) fighters subject to highly stressful training involving repetitive head impacts. This study examined differences in DNA methylation and miRNA expression in elite MMA fighters compared to active controls. Global methylation differences between groups were assessed via a LINE-1 assay. At the same time, PCR arrays were used to estimate differential expression in samples of 21 fighters and 15 controls for 192 different miRNAs associated with inflammatory diseases. An Independent-Samples t-Test found no significant difference in LINE-1 methylation between groups. However, an Independent-Samples Mann-Whitney U Test revealed a significant upregulation in the expression of miR-155 in MMA fighter plasma. Since miR-155 has been recognized as an important regulator of neuroinflammation, this dysregulation suggests a possible epigenetic mechanism responsible for chronic inflammation associated with professional-level MMA training. Consistent with other published works, this study highlights the potential of miR-155 not only as a biomarker for monitoring long-term health risks linked to head trauma but also as a target to remediate the impact of chronic neuroinflammation.
## Introduction
Traumatic brain injury (TBI) often involves blunt force impacts to the head, causing rapid percussive or rotational damage to the brain [1]. This can result in bleeding, brain cell loss, and stretching of axons, often producing cognitive impairment, neuropsychiatric symptoms, dementia, and parkinsonism. Mild traumatic brain injury (mTBI) accounts for about $85\%$ of total cases worldwide, posing a major concern for healthcare providers tasked with diagnosing and treating patients suffering the various repercussions. mTBI is often undiagnosed, especially among contact sports athletes motivated to persist in competition and refuse to disclose their symptoms [1]. The impact of both TBI and mTBI on athletes has recently been highlighted by highly publicized cases of chronic traumatic encephalopathy (CTE) [2]. Many factors, including psychological stress, may contribute to the progressive tauopathy seen in CTE; however, it is still uncertain how frequency and severity of injury factor into the time course of neuropsychological decline. While recent diagnoses have been made in athletes in their 20s, CTE has most frequently been observed in the brains of older retired athletes since it can only be currently confirmed by observing the brains of deceased athletes. A recent study diagnosed CTE in 177 of the 202 ($87\%$) deceased American football players studied [3]. The severity of exposure to repetitive head impacts (RHI) necessary to place one at risk for CTE remains unclear. However, research supports that long-term inflammation mediated by microglial activation and tau pathology contributes significantly to disease progression and is likely associated with more prolonged exposure to repetitive head trauma [4]. Experimental research also supports the idea that this chronic neuroinflammation causes progressive neurodegeneration, which can potentially be treated long after the instance of TBI [5].
Noninvasive diagnosis of CTE remains problematic. Some current efforts have focused on dual and concomitant objectives: (i) explaining the underlying molecular mechanisms of diseases resulting from RHI and (ii) identifying molecular markers of disease progression [1,4,6,7]. For example, recent research has suggested persistent changes in DNA methylation with TBI. Exposure to RHI or an acute bout of exercise can rapidly change DNA methylation levels in associated tissues [8,9]. This epigenetic dysregulation could result in various diseases depending on which gene regions are differentially methylated, such as those associated with limbic system structures and Brain-Derived Neurotrophic Factor (BDNF) secretion [9].
Similarly, miRNA microarray and real-time qPCR (RT-qPCR) techniques have both been used to demonstrate the differential expression of microRNA (miRNAs) associated with the cerebral cortex (e.g., miR-21) following TBI in a rat model [10,11]. Various miRNAs circulating in plasma showed altered expression patterns soon after TBI in mice [12] and humans [13]. For instance, miR-155 is involved in the upregulation of microglia-mediated neuroimmune response, implicating its inhibition as a therapeutic method [14]. The overexpression of miR-155 has increased dendritic cell apoptosis and enhanced IL-12p70 production, which plays a role in Interferon-gamma (IFN-γ)-producing T cell development and natural killer cell activation [15]. *More* generally, the effectiveness of miRNAs as diagnostic markers for a wide range of diseases [16-19] or injuries [20] is a rapidly expanding area of research.
Based on these findings, the current study evaluated differences in DNA methylation and expression of miRNAs associated with chronic inflammation and various cancers between professional MMA fighters and a control group of non-contact sports athletes. In contrast to previous studies [1], the fighters tested were active, elite professionals that competed in the Ultimate Fighting Championship (UFC) or Bellator. They tend to experience RHI and are at risk for diseases associated with chronic brain inflammation [1]. While research into CTE has typically been observed in older retired athletes, little is known of epigenetic modification in younger elite professionals currently active in their careers. This may contribute to the risk of sustained neuroinflammatory disease experienced later in life.
Part of the data presented in this article was previously presented as a meeting abstract at the 2019 International Society of Sports Nutrition (ISSN) Annual Scientific Meeting in June 2019, at the 2019 Society for Neurosports Annual Scientific Meeting in November 2019, and at the 2022 International Behavioral Neuroscience Society (IBNS) Annual Scientific Meeting in June 2022.
## Materials and methods
Participants As part of a larger study investigating the effects of contact sport participation on inflammatory biomarkers and neurobehavioral performance, 21 male MMA fighters training for professional combat and 15 male age-matched professional athlete controls not experiencing RHI provided data for this analysis. Athletes were tested at the Nova Southeastern University (NSU) Health Profession's Annex in the Fight Science laboratory. All participants underwent an informed consent process following the Declaration of Helsinki and an approved IRB protocol submitted to the institutional review board of Nova Southeastern University (approval number: 2018-08-14).
Body composition Body composition was assessed with a dual-energy X-ray absorptiometry machine (DXA) (Model: Hologic Horizon W; Hologic Inc., Danbury CT, USA). Quality control calibration procedures were performed on a spine phantom. Subjects were instructed to come to the laboratory after at least a 3-hour fast and no prior exercise. Subjects wore typical athletic clothing and removed metal jewelry. The research participants were positioned supine on the DXA within the borders delineated by the scanning table. Each whole-body scan took approximately seven minutes.
DNA and RNA extractions Blood samples (3 mL) were collected from participants and centrifuged to separate plasma. Saliva samples (1 mL) were also collected via passive drool. DNA was extracted from saliva samples using the QIAamp DNA Investigator kit per manufacturer protocol. In contrast, RNA samples were extracted from the plasma samples using the miRNeasy Serum/Plasma Kit (both kits from QIAGEN, Valencia, CA, USA). All extractions were performed in a QIAcube instrument, and sample purity and concentration assessments were measured using a Nanodrop Lite Spectrophotometer (ThermoFisher Scientific, Waltham, Massachusetts, USA).
DNA methylation analysis Global methylation differences between groups were assessed via a LINE-1 assay (surrogate global DNA analysis), a product of Active Motif (Carlsbad, CA, USA), according to the assay protocol. 1 µL of each DNA sample was prepared for the LINE-1 methylation assay via a Msel digestion. Approximately 100 ng of digested sample DNA was used per well, and samples were run in triplicate.
miRNA expression analysis Each extracted RNA sample was converted to cDNA using the miScript II RT Kit (QIAGEN). 1 µL of each athlete's cDNA was combined into a pool with 10x of RNAse-free water. The Human Inflammatory Response & Autoimmunity and Human Serum & Plasma miScript miRNA PCR Arrays (QIAGEN) were used to estimate differential expression between MMA fighters and the control group for 192 miRNAs. Pooled samples (2 µL) were used as templates for each reaction, following the manufacturer's protocol. RT-qPCR was then used to quantify the expression of miR-155 in individuals of both groups by combining cDNA samples with SNORD61 and miR-155 primers, respectively, using miScript Primer Assays (QIAGEN). Dilutions (1:10) of all individual cDNA samples were used for these assays, and reactions were run in duplicates. The relative expression ratio for miR-155 in MMA fighters vs. matched controls was determined from quantitative PCR results to the control miRNA, SNORD61, and performed in the ROCHE LC96 Application Software [21].
Statistical analyses Age and body composition differences were determined via paired samples t-tests. The effect of group status on outcome measures was analyzed via Independent Samples t-tests. In the case of non-normal distribution, where the normality assumption was not met, an Independent-Samples Mann-Whitney U Test was carried out. All calculations were conducted using an SPSS statistical package (version 26, IBM, Armonk, NY, USA). All reported p-values are two-tailed with an a priori significance level of $p \leq 0.05.$
## Results
The sample sizes and respective body composition parameters are summarized in Table 1. The relatively small size of both groups is largely due to the challenges associated with identifying and recruiting individuals who met the stringent criteria for study enrollment (active and elite professional male athletes). Sample sizes, or apparent differences between both groups, did not impact the results described below. The statistical tests used for data interpretation were specifically selected to account for these conditions. As expected, both groups showed no significant difference in age, although there were variations in parameters associated with bone and lean mass composition (Table 1). High-quality DNA and RNA samples were obtained for all subjects. DNA methylation analysis suggested that global DNA methylation levels, measured using Long Interspersed Nucleotide Element 1 (LINE-1) repeats, comprise approximately $17\%$ of the total genome and act as a surrogate estimate for global DNA methylation levels, were similar between the two groups. As illustrated in Figure 1, an independent samples t-test found no significant difference between groups in LINE-1 methylation t[34] = 0.60, $$p \leq 0.55$$, $d = 0.20.$ In contrast, the miRNA PCR arrays indicated 28 miRNAs with differential expression between the pooled samples of MMA fighters and control athletes (Table 2). These miRNA dysregulations indicate that, as suggested by previous studies, unique molecular signatures may be detected in MMA fighters and are potentially related to the high frequency of RHI experienced by professional fighters relative to other athletes [1].
To further validate the preliminary miRNA expression dysregulation data, qPCR-based analyses were performed to measure differential expression in individual samples. These analyses used SNORD61 and miR-155 gene-specific primers. They assessed the expression of miR-155 relative to SNORD61, which was used as a reference with expression levels assumed to be constant in all conditions (control athletes and MMA fighters). As illustrated in Figure 2, miR-155 was found to be more expressed than SNORD61, and a Mann-Whitney U test found significantly higher expression of miR-155 in the group of MMA fighters (mean: 13.64, $$n = 11$$) compared to the group of active controls (mean: 8.10, $$n = 10$$) ($U = 26.00$, $$P \leq 0.043$$). This result confirms the high-throughput, pooled sample assays. The other miRNAs listed in Table 2 may also represent individual targets for further analyses of dysregulated gene expression patterns in MMA fighters.
**Figure 2:** *Comparative analysis of miR-155 expression levels in plasma samples collected from non-contact athletes and Mixed Martial Arts (MMA) fighters.A Mann-Whitney U test found miR-155 significantly overexpressed in the group of MMA fighters, indicating that the stress of professional MMA training may predispose fighters to display unique and potentially diagnostic molecular patterns. Error bars represent the standard error of the mean. Asterisk represents p < 0.05. MMA = Mixed Martial Arts.*
## Discussion
The current study sought to determine if professional MMA fighters experience dysregulated molecular patterns at the DNA or RNA levels. In the time following a typical sparring practice - which occurs twice a week for months leading up to a fight - these athletes likely experience molecular changes in measures related to muscle and bone synthesis, angiogenesis, and pro- and anti-inflammatory systems. While some of these changes may be part of beneficial adaptations to exercise, prolonged exposure to dysregulation of some of these measures may underlie tauopathy or other neuroinflammatory states observed in CTE, suggesting that altered molecular patterns may be mined for potential biomarkers associated with disease risk or progression.
*Data* generated throughout this study suggest that miRNA expression patterns have greater potential as a biomarker of neural integrity than DNA methylation. No significant difference in LINE-1 methylation between the two groups indicates that, despite the stress of MMA training, these fighters are not at any greater risk of developing dysregulation of global methylation (Figure 1). More precise measurements of DNA methylation at specific gene regions may reveal differences in MMA fighters relative to an athlete control group and support the growing body of literature linking epigenetic modifications with TBI [9,22] - including TBI experienced in contact sport competition [23]. While DNA methylation appeared stable among experimental groups, differential expression of select miRNAs was readily identified between MMA fighters and the matched athlete's control group (Table 2). These results not only serve to confirm a previously published study that highlighted changes in miRNA gene expression in amateur MMA fighters [1], but they are also consistent with previous efforts that assessed differential miRNA expression after TBI [13,24]. Several of the miRNAs found to be dysregulated in the present study, including miR-128, miR-15a, and miR-10b, were also differentially expressed in brain tissue samples from individuals diagnosed with CTE or Amyotrophic Lateral Sclerosis (ALS) [24].
Changes in miRNA expression following TBI have also been detected in plasma [13] and have implicated miR-155, especially in rodent models, where miR-155 levels increased after TBI (reviewed in [25]). In agreement with these observations, miR-155 up-regulation in serum has recently been associated with the number of hits to the heads (HTH) experienced by amateur MMA fighters [1]. Herein, a significant difference in the expression of miR-155 was measured (Figure 2) in professional fighters relative to controls, suggesting that these athletes may experience chronic overexpression of this miRNA and, therefore, may be at a greater risk for experiencing chronic inflammation or developing diseases such as CTE or ALS as a result of the stress and intensity associated with their training. The critical role of miR-155 as a regulator of neuroinflammation has been recently reviewed [25]. In addition to being dysregulated during various neuroinflammatory disorders, this miRNA was also demonstrated to promote microglia-mediated neuroinflammation by modulating the suppressor of cytokine signaling 1 (SOCS-1) protein, suggesting that the inhibition of miR-155 may be neuroprotective in the context of chronic inflammation [14]. Therefore, confirming the relationship between head trauma and miR-155 regulation could offer a potential biomarker of neuroinflammation and a therapeutic target for treating athletes suffering from chronic illness.
This study was limited in its specificity of data related to the risk of developing diseases these fighters may experience. Individualized data concerning the time of the last professional fight or the most recent concussion is worth considering in determining the timeframe of epigenetic responses to RHI experienced in professional combat. Moreover, the severity and frequency of RHI, factors likely relevant to the neurodegeneration observed in CTE, are challenging to assess. Fighters may underrepresent the number of undiagnosed concussions they have experienced, considering concussions synonymous with losing consciousness or other severe TBI symptoms. mTBI may be difficult for these fighters to self-assess; furthermore, they may be unwilling to disclose signs and symptoms of mTBI for fear of being held back from practice or competition. The predominant fighting style is also worthy of consideration: fighters with more striking experience may be at higher risk of suffering TBI than those who prefer grappling styles, which typically involve more wrestling maneuvers rather than strikes to the head. It remains unclear how body composition could affect the expression of miR-155. A 2021 study suggests that miR-155 mediates obesity-associated β cell dysregulation [26]. While the two groups in this study had significantly different lean masses, there was no difference in body fat (Table 1). Future research can confirm the relationship between miR-155 overexpression and RHI and the potential of this miRNA as a biomarker for individuals at risk for CTE.
Furthermore, work is underway to link the observed up-regulation of miR-155 with other serum-based biomarkers of neurodegeneration, inflammation, and RHI-caused trauma [27]. The expression of other miRNAs (e.g., those found in Table 2) in athletes at various points in their careers or after retirement is also being assessed. These analyses, focused on a more longitudinal approach, may provide the basis to calculate odds ratios based on the medical diagnosis of either the occurrence of concussions or the observation of disease symptoms. The results of this study serve as complimentary findings to this growing body of research.
## Conclusions
This study demonstrated overexpression of miR-155 in a group of professional MMA fighters; furthermore, the miRNA PCR arrays estimated group differences using pooled samples for 27 other miRNAs, which may be significantly altered with participation in professional MMA competition. While these fighters may experience a protective effect from the intense exercise and nutrition regiment they adhere to, accounts continually appear of older retired athletes suffering negative neurobehavioral consequences likely related to repetitive mTBI experienced throughout their careers. Epigenetic research continues to identify sensitive biomarkers involved in disease processes which may aid healthcare professionals in developing treatment plans to maintain the health and well-being of these athletes.
## References
1. LaRocca D, Barns S, Hicks SD. **Comparison of serum and saliva miRNAs for identification and characterization of mTBI in adult mixed martial arts fighters**. *PLoS One* (2019) **14** 0
2. McKee AC, Daneshvar DH, Alvarez VE, Stein TD. **The neuropathology of sport**. *Acta Neuropathol* (2014) **127** 29-51. PMID: 24366527
3. Mez J, Daneshvar DH, Kiernan PT. **Clinicopathological evaluation of chronic traumatic encephalopathy in players of American football**. *JAMA* (2017) **318** 360-370. PMID: 28742910
4. Cherry JD, Tripodis Y, Alvarez VE. **Microglial neuroinflammation contributes to tau accumulation in chronic traumatic encephalopathy**. *Acta Neuropathol Commun* (2016) **4** 112. PMID: 27793189
5. Faden AI, Wu J, Stoica BA, Loane DJ. **Progressive inflammation-mediated neurodegeneration after traumatic brain or spinal cord injury**. *Br J Pharmacol* (2016) **173** 681-691. PMID: 25939377
6. Cherry JD, Stein TD, Tripodis Y. **CCL11 is increased in the CNS in chronic traumatic encephalopathy but not in Alzheimer's disease**. *PLoS One* (2017) **12** 0
7. Oliver JM, Anzalone AJ, Stone JD. **Fluctuations in blood biomarkers of head trauma in NCAA football athletes over the course of a season**. *J Neurosurg* (2018) 1-8
8. Barrès R, Yan J, Egan B. **Acute exercise remodels promoter methylation in human skeletal muscle**. *Cell Metab* (2012) **15** 405-411. PMID: 22405075
9. Sagarkar S, Bhamburkar T, Shelkar G, Choudhary A, Kokare DM, Sakharkar AJ. **Minimal traumatic brain injury causes persistent changes in DNA methylation at BDNF gene promoters in rat amygdala: a possible role in anxiety-like behaviors**. *Neurobiol Dis* (2017) **106** 101-109. PMID: 28663119
10. Lei P, Li Y, Chen X, Yang S, Zhang J. **Microarray based analysis of microRNA expression in rat cerebral cortex after traumatic brain injury**. *Brain Res* (2009) **1284** 191-201. PMID: 19501075
11. Redell JB, Liu Y, Dash PK. **Traumatic brain injury alters expression of hippocampal microRNAs: potential regulators of multiple pathophysiological processes**. *J Neurosci Res* (2009) **87** 1435-1448. PMID: 19021292
12. Sharma A, Chandran R, Barry ES. **Identification of serum microRNA signatures for diagnosis of mild traumatic brain injury in a closed head injury model**. *PLoS One* (2014) **9** 0
13. Redell JB, Moore AN, Ward NH 3rd, Hergenroeder GW, Dash PK. **Human traumatic brain injury alters plasma microRNA levels**. *J Neurotrauma* (2010) **27** 2147-2156. PMID: 20883153
14. Cardoso AL, Guedes JR, Pereira de Almeida L, Pedroso de Lima MC. **miR-155 modulates microglia-mediated immune response by down-regulating SOCS-1 and promoting cytokine and nitric oxide production**. *Immunology* (2012) **135** 73-88. PMID: 22043967
15. Lu C, Huang X, Zhang X. **miR-221 and miR-155 regulate human dendritic cell development, apoptosis, and IL-12 production through targeting of p27kip1, KPC1, and SOCS-1**. *Blood* (2011) **117** 4293-4303. PMID: 21355095
16. Amaral FC, Torres N, Saggioro F. **MicroRNAs differentially expressed in ACTH-secreting pituitary tumors**. *J Clin Endocrinol Metab* (2009) **94** 320-323. PMID: 18840638
17. Hayes J, Peruzzi PP, Lawler S. **MicroRNAs in cancer: biomarkers, functions and therapy**. *Trends Mol Med* (2014) **20** 460-469. PMID: 25027972
18. Mayr M, Zampetaki A, Willeit P, Willeit J, Kiechl S. **MicroRNAs within the continuum of postgenomics biomarker discovery**. *Arterioscler Thromb Vasc Biol* (2013) **33** 206-214. PMID: 23325478
19. Rong H, Liu TB, Yang KJ. **MicroRNA-134 plasma levels before and after treatment for bipolar mania**. *J Psychiatr Res* (2011) **45** 92-95. PMID: 20546789
20. Laterza OF, Lim L, Garrett-Engele PW. **Plasma MicroRNAs as sensitive and specific biomarkers of tissue injury**. *Clin Chem* (2009) **55** 1977-1983. PMID: 19745058
21. Pfaffl MW. **A new mathematical model for relative quantification in real-time RT-PCR**. *Nucleic Acids Res* (2001) **29** 0
22. Treble-Barna A, Heinsberg LW, Puccio AM. **Acute brain-derived neurotrophic factor DNA methylation trajectories in cerebrospinal fluid and associations with outcomes following severe traumatic brain injury in adults**. *Neurorehabil Neural Repair* (2021) **35** 790-800. PMID: 34167372
23. Tarnowski M, Tomasiak P, Tkacz M, Zgutka K, Piotrowska K. **Epigenetic alterations in sports-related injuries**. *Genes (Basel)* (2022) **13** 1471. PMID: 36011382
24. Alvia M, Aytan N, Spencer KR. **MicroRNA alterations in chronic traumatic encephalopathy and amyotrophic lateral sclerosis**. *Front Neurosci* (2022) **16** 855096. PMID: 35663558
25. Zingale VD, Gugliandolo A, Mazzon E. **MiR- 155: an important regulator of neuroinflammation**. *Int J Mol Sci* (2021) **23** 90. PMID: 35008513
26. Gao H, Luo Z, Jin Z, Ji Y, Ying W. **Adipose tissue macrophages modulate obesity-Associated β cell adaptations through secreted miRNA-containing extracellular vesicles**. *Cells* (2021) **10** 2451. PMID: 34572101
27. Meier TB, Nelson LD, Huber DL, Bazarian JJ, Hayes RL, McCrea MA. **Prospective assessment of acute blood markers of brain injury in sport-related concussion**. *J Neurotrauma* (2017) **34** 3134-3142. PMID: 28699381
|
---
title: 'Evaluation of antibiotics resistance in Southern Iran in light of COVID‐19
pandemic: A retrospective observational study'
authors:
- Rahim Raoofi
- Negin Namavari
- Vahid Rahmanian
- Mohammad Hadi Dousthaghi
journal: Health Science Reports
year: 2023
pmcid: PMC10017310
doi: 10.1002/hsr2.1153
license: CC BY 4.0
---
# Evaluation of antibiotics resistance in Southern Iran in light of COVID‐19 pandemic: A retrospective observational study
## Abstract
### Background and Aims
Antimicrobial resistance (AMR) was taken as one of the high‐priority long‐lasting public health issues, although it might have been underrated in terms of COVID‐19 pandemic emergence. Regarding limited data on assessing the pandemic effect on AMR trend in Iran, this study aimed to describe the epidemiology of antibiotics resistance during the COVID pandemic in southern Iran.
### Methods
This descriptive study was conducted on 2675 patients' samples collected and processed in a referral COVID‐19 center hospital in southern Iran from March 21, 2019, to February 18, 2020 (prepandemic), and February 19, 2020, to March 21, 2021 (pandemic). Susceptibility test results in sensitivity and resistance levels were compared in prepandemic and pandemic periods.
### Results
Compared to prepandemic, the inpatient number has increased almost three times. On the other hand, there are around four times fewer outpatients now. More than $85\%$ of the specimens were found in urine samples. In all, $92.22\%$ of all bacteria samples were Gram‐negative isolates, with *Escherichia coli* accounting for $59.19\%$ of them. The change rate of Gram‐negative bacteria resistance to antimicrobials is an average of $7.74\%$ ($p \leq 0.001$). On the other hand, the average change rate of Gram‐positive bacteria resistant to antibiotics has decreased by $19.3\%$ ($$p \leq 008$$). As a forerunner among other Gram‐negative bacteria, the average change rate for *Pseudomonas aeruginosa* and *Klebsiella pneumonia* resistance to monitored antibiotics was $89\%$ and $66.3\%$, respectively ($p \leq 0.001$).
### Conclusion
During the Covid‐19 pandemic, the increase in AMR among Gram‐negative bacteria, particularly P. aeruginosa and K. pneumonia, was observed compared to the prepandemic. This further limits treatment options, and endangers global public health.
## INTRODUCTION
Antimicrobial resistance (AMR), as one of the most critical worldwide public health issues, should be intercepted as soon as possible. 1 AMR affects health care, and life quality eventuating in death and extra cost. 2 If there are no interventions, it was estimated that the annual death rate will reach 10 million in 2050 caused by AMR. 3 Considering the pathogen's resistance does not have any geographical boundary, AMR must not be taken as a bordered problem for just some countries or regions regarding either income or level of development. 4 In 2017, World Health Organization listed some high priority, and critical bacteria most of which belonged to Gram‐negative bacteria. These pathogens have multidrug‐resistant features and cause healthcare‐associated infections. 5 The COVID‐19 pandemic as a parallel issue to AMR is taken as a crucial health emergency. It is an acute problem; on the other hand, the AMR is the long‐lasting one. 6 Some comparative studies on AMR rates during COVID‐19 and before the pandemic has disclosed a significant change. 7, 8, 9, 10, 11 Taking action to slow down the spread of COVID‐19, such as social distancing, using physical barriers, and so forth, has led to a reduction in the spread of other infections, which resulted in less usage of antimicrobials. Hence, it was reported that patients with other infections prefer not to seek care in healthcare centers. 12, 13, 14 On the other hand, researchers empirically reused some medications, including some antibiotics, regarding their antiviral effects, to treat COVID‐19 patients, disregarding antimicrobial stewardship rules. 5, 12, 13, 14 For instance, azithromycin was prescribed to treat SARS‐CoV‐2. If their usage has not had any significant effect on treating COVID‐19, this matter has not had any consequence, but the AMR increased. 9 The bacterial co‐infection with COVID‐19 has increased the rate of antibiotic prescription in hospitalized patients, but there are no data on community antibiotic usage. 15, 16, 17
The mutual effect of AMR and COVID‐19 is unknown yet. 6 This study aimed to describe the epidemiology of antibiotics resistance in Jahrom District, Southern Iran during the COVID pandemic.
## Study design
This is a descriptive (retrospective observational) study in which data were collected from either inpatients or outpatients at a COVID‐19 referral hospital affiliated with Jahrom university of medical sciences. A total of 2675 patient samples were processed from March 21, 2019, to February 18, 2020 (prepandemic), and February 19, 2020, to March 21, 2021 (pandemic) among all referred patients to the hospital. In Iran, the pandemic officially started in mid‐February 2020. The data were divided into two categories: prepandemic and pandemic periods.
The clinical specimens of any positive culture of urine, blood, sputum, stool, wound, cerebrospinal fluid, aspiration, and pleural fluid/bronchoalveolar lavage (BAL) which had recorded an antibiogram were included. Any culture which had not have a recorded antibiogram was excluded.
## Sample identification and disc diffusion susceptibility testing method
Positive cultured samples in sterile saline were incubated to reach 0.5 McFarland (1.5 × 108 colony‐forming unit/mL) concentration. The colonies were incubated in Mueller‐Hinton agar for 24–48 h, depending on the sample type, which had been at 35–37°C. According to Clinical and Laboratory Standards Institute (CLSI) 2020 18 guidelines, the disc diffusion technique was used to assess the sensitivity of bacteria isolated from patient samples at two sensitive and resistant levels by measuring the size of the antibiotic disc's inhibitory growth zone. To identify the strains standard biochemical tests were done.
The antibiotics discs (PadtanTeb) containing CN: cephalexin 30 µg, CP: ciprofloxacin 5 µg, CRO: ceftriaxone 30 µg, CTX: cefotaxime 30 µg, FEP: cefepime 30 µg, GM: gentamicin 10 µg, SXT: cotrimoxazole $\frac{1.25}{23.75}$ µg, VA: vancomycin 30 µg.
## Data collection
The hospital medical records have been the basis of clinical features. The data were entered into an electronic pattern. The considered independent factors of the studied population have been age, sex, antibiotics, either sensitivity or resistance to antibiotics, and hospitalization status.
## Statistical analysis
IBM SPSS version 21 software was used for data analysis. The participants' demographic characteristics were expressed in frequency and percent. The association between antibiotic resistance and study variables was assessed using the chi‐square test and Student's t‐test at a significance level of 0.05. The change rate indicated by the difference level between prepandemic and the pandemic period of time divided by its frequency during the prepandemic period of time as what is shown by the following equation: [1] change rate=NPan−NpreNpre, where N pan is the number of frequency during the pandemic period of time and N pre the number of frequency during prepandemic period of time.
Supposing that the bacteria behavior about the evaluated antibiotics is independent, a weighted average was reported as overall monitored bacteria resistance to all evaluated antibiotics by the following equation: [2] overall resistance=(R1×n1)+(R2×n2)+⋯+(Rn×nn)n1+n2+⋯+nn, where R 1 stands for the antibiotic resistance of supposed isolated bacteria to first antibiotic and n 1 is the total sample number of the same isolated bacteria for which antibiogram test is reported for the first antibiotic. Moreover, R 2 stands for the antibiotic resistance of the same isolated bacteria to the second antibiotic, and n 2 is the total sample number of the same isolated bacteria for which the antibiogram test is reported for the second antibiotic. R n stands for the antibiotic resistance of the same isolated bacteria to the nth antibiotic and n n is the total sample number of the same isolated bacteria for which an antibiogram test is reported for the nth antibiotic. To establish an indicator to track changes in bacterial resistance patterns in light of the COVID‐19 presence, overall resistance has been calculated. A p value <0.05 was considered a significant level.
## Ethics approval
This research project was approved by the Ethics Committee of the Jahrom University of Medical Science, Fars, Iran (IR.JUMS.REC.1398.093).
## Clinical specimens and isolated pathogens and demographic features
The study population included 2675 patient samples, of whom 1778 and 897 were for prepandemic and pandemic, respectively. As compared to prepandemic, during the pandemic the inpatients' mean age had no significant change (prepandemic versus during the pandemic, 60.2 ± 22.90 versus 58.86 ± 17.59, $$p \leq 0.124$$). Unlike inpatients, the number of outpatients decreased. Although both genders had an equal inpatient portion each year, female patients possess more than $70\%$ of outpatients every year (Table 1).
**Table 1**
| Unnamed: 0 | Admission type | Prepandemica | Pandemicb | Change rate | p Value |
| --- | --- | --- | --- | --- | --- |
| Sample number | Inpatient | 155.0 | 470.0 | 203.23% | <0.001c |
| | Outpatient | 1623.0 | 427.0 | −73.69% | |
| Mean age | Inpatient | 60.2 | 58.86 | −2.23% | 0.124d |
| | Outpatient | 45.72 | 46.93 | 2.65% | 0.114d |
| Female | Inpatient | 82.0 | 241.0 | 193.90% | 0.164c |
| | Outpatient | 1187.0 | 335.0 | −71.78% | |
| Male | Inpatient | 73.0 | 229.0 | 213.70% | 0.124c |
| | Outpatient | 436.0 | 92.0 | −78.90% | |
The urine samples contained the most specimens, followed by sputum, blood, wound, aspiration, pleural fluid and BAL, cerebrospinal fluid, and stool (Table 2).
**Table 2**
| Specimens | Prepandemica | Prepandemica.1 | Pandemic b | Pandemic b.1 | Total | Total.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Urine | 1556 | 87.5% | 724 | 80.71% | 2280.0 | 85.23% |
| Sputum | 76 | 4.3% | 105 | 11.71% | 181.0 | 6.77% |
| Blood | 51 | 2.9% | 41 | 4.6% | 92.0 | 3.4% |
| Wound | 62 | 3.5% | 10 | 1.1% | 72.0 | 2.7% |
| Aspiration | 19 | 1.1% | 5 | 0.6% | 24.0 | 0.9% |
| Pleural fluid/BAL | 8 | 0.4% | 8 | 0.9% | 16.0 | 0.6% |
| CSF | 4 | 0.2% | 2 | 0.2% | 6.0 | 0.2% |
| Stool | 2 | 0.1% | 2 | 0.2% | 4.0 | 0.1% |
| Total | 1778 | 897 | 2675 | | | |
The most frequent Gram‐negative and Gram‐positive pathogens isolated in both years were *Escherichia coli* and *Staphylococcus aureus* specimens, respectively (Table 3). The order of the most prevalent bacteria was E. coli, Pseudomonas aeruginosa, Staphylococcus strains, Klebsiella pneumonia, Citrobacter, and *Acinetobacter baumannii* (Table 3).
**Table 3**
| Bacteria | Prepandemica | Prepandemica.1 | Pandemicb | Pandemicb.1 | Total | Total.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Escherichia coli | 547 | 30.76% | 306 | 34.11% | 853.0 | 31.89% |
| Pseudomonas aeruginosa | 168 | 9.45% | 114 | 12.71% | 282.0 | 10.54% |
| Staphylococcus strains | 57 | 3.2% | 50 | 5.6% | 107.0 | 4.00% |
| Klebsiella pneumoniae | 46 | 2.6% | 44 | 4.9% | 90.0 | 3.4% |
| Citrobacter freundii | 35 | 2% | 10 | 1.1% | 45.0 | 1.7% |
| Proteus | 25 | 1.4% | 13 | 1.4% | 38.0 | 1.4% |
| Acinetobacter baumannii | 13 | 0.7% | 1 | 0.1% | 14.0 | 0.5% |
| Enterobacter | 4 | 0.2% | 0 | 0% | 4.0 | 0.1% |
| Streptococcus | 3 | 0.2% | 2 | 0.2% | 5.0 | 0.2% |
| Salmonella | 0 | 0% | 2 | 0.2% | 2.0 | 0.1% |
| Shigella | 1 | 0.1% | 0 | 0% | 1.0 | 0.04% |
| Fungi | 51 | 2.9% | 47 | 5.2% | 98.0 | 3.7% |
| Yeast | 122 | 6.86% | 50 | 5.6% | 172.0 | 6.43% |
| Mix growthc | 637 | 35.83% | 258 | 28.76% | 895.0 | 33.46% |
| No growth | 69 | 3.9% | 0 | 0% | 69.0 | 2.6% |
| Total | 1778 | 897 | 2675 | | | |
Gram‐negative and ‐positive bacteria frequencies were 1329 and 112, respectively. E. coli, P. aeruginosa, and K. pneumonia were mainly detected from urine specimens. Moreover, *Staphylococcus strains* were predominant in blood and urine specimens. Furthermore, Citrobacter detection was more frequent in sputum specimens. A. baumannii was detected equally from sputum and blood specimens (Table 4).
**Table 4**
| Bacteria | Specimen | Specimen.1 | Specimen.2 | Specimen.3 | Specimen.4 | Specimen.5 | Specimen.6 | Specimen.7 | Specimen.8 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Bacteria | Urine | Sputum | Blood | Wound | Aspiration | Pleural fluid/BAL | CSF | Stool | Total |
| Escherichia coli | 34.96% (797) | 8.29% (15) | 31.5% (29) | 8.3% (6) | 20.83% (5) | 6.2% (1) | ‐ | ‐ | 31.89% (853) |
| Pseudomonas aeruginosa | 9.30% (212) | 22.10% (40) | 13% (12) | 13.9% (10) | (0) | 25% (4) | 66.7% (4) | ‐ | 10.54% (282) |
| Staphylococcus strains | 1.36% (31) | 3.87% (7) | 32.6% (30) | 31.9% (23) | 37.5% (9) | 31.2% (5) | 33.3% (2) | ‐ | 4.00% (107) |
| Klebsiella pneumoniae | 3.51% (80) | 3.31% (6) | 1.1% (1) | 1.4% (1) | (0) | 12.5% (2) | ‐ | ‐ | 3.36% (90) |
| Citrobacter freundii | 0.04% (1) | 8.29% (15) | 10.9% (10) | 18.1% (13) | 16.7% (4) | 6.2% (1) | ‐ | 25% (1) | 1.68% (45) |
| Proteus | 1.62% (37) | (0) | 1.1% (1) | ‐ | ‐ | ‐ | ‐ | ‐ | 1.42% (38) |
| Acinetobacter baumannii | ‐ | 3.87% (7) | 7.6% (7) | ‐ | ‐ | ‐ | ‐ | ‐ | 0.52% (14) |
| Enterobacter | ‐ | ‐ | 2.2% (2) | 2.8% (2) | ‐ | ‐ | ‐ | ‐ | 0.15% (4) |
| Streptococcus | 0.22% (5) | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | 0.19% (5) |
| Salmonella | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | 50% (2) | 0.07% (2) |
| Shigella | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | 25% (1) | 0.04% (1) |
| Fungi | 0.13% (3) | 39.23% (71) | ‐ | 23.6% (17) | 20.8% (5) | 12.5% (2) | ‐ | ‐ | 3.66% (98) |
| Yeast | 6.67% (152) | 9.94% (18) | ‐ | ‐ | 4.2% (1) | 6.2% (1) | ‐ | ‐ | 6.43% (172) |
| Mix growth | 39.17% (893) | 1.10% (2) | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | 33.46% (895) |
| No growth | 3.03% (69) | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ | 2.58% (69) |
| Total | 2280 | 181 | 92 | 72 | 24 | 16 | 6 | 4 | 2675 |
## Resistance profile for the isolate bacteria and the effect of COVID‐19
A significant overall resistance increase among Gram‐negative bacteria was observed, E. coli excluded. P. aeruginosa and K. pneumonia had a prominent overall resistance increase, Citrobacter for that matter, but in a milder trend. Although A. baumannii resistance has not had a significant increase, its resistance level was $100\%$, after the pandemic. Superficially, it might be interpreted that E. coli resistance has overall decreased. However, better scrutiny discloses that its resistance to gentamicin, cefepime, and cefotaxime has increased (Table 5).
## DISCUSSION
One of the future challenging public health issues as a subsequence of the COVID‐19 pandemic may have been an increase in AMR caused by indiscriminative antibiotic use. To the best of our knowledge, this is the first study evaluating the COVID‐19 pandemic's probable effect on AMR in Iran during the pandemic compared to prepandemic. Although there might be differences in the healthcare system set‐up of each country. It shows an overall increase in AMR by the pandemic presence, which has been shown in some other countries, including India, 8 Mexico, 18 Indonesia, 10 Serbia, 11 and so forth.
Considering the hospital was a COVID‐19 referral center, intensification in the inpatient admission, $203.69\%$, could be taken as a consequence of the COVID‐19 pandemic, but, by contrast, a drastic fall in the outpatients, $73.69\%$, shows the avoidance of referring to hospital during the pandemic, unless it were emergency. 7, 19 Mandatory lock‐down to create social distancing, personal fear of the contagious pandemic in the healthcare centers, lack of knowledge, concluded in preferring over‐the‐counter usage of antibiotics instead of referring to healthcare centers; moreover, remote medicine, the lack of guidelines and knowledge in facing the pandemic, disruption in research over AMR and antibiotics stewardship eventuated in overuse and over‐prescription of antibiotics. 15, 20 The current study showed an increase in resistance of Gram‐negative bacteria (Supporting Information: Figure 1), P. aeruginosa, K. pneumonia, A. baumannii, and Citrobacter, to most reported antibiotics by the pandemic presence, particularly in P. aeruginosa and K. pneumonia. A study in India on COVID‐19 patients reported an increase in resistance of P. aeruginosa to fourth generation of cephalosporins, including cefepime. 8 The same research also notes a rise in ciprofloxacin and Gentamicin resistance in K. pneumoniae. Moreover, research conducted in Mexico found that both K. pneumonia and P. aeruginosa were becoming more resistant to cefepime, ciprofloxacin, and gentamicin. 18 A recent study in Northeast Iran focusing on E. coli, P. aeruginosa, K. pneumonia, and A. baumannii strains showed a significant rise in resistance rate during 2020–2022. For instance, a $30\%$ increase in K. pneumonia to cefotaxime is reported. 21 Gram‐negative bacteria, specifically K. pneumoniae, are one of the main reasons for VAP in the intensive care units (ICU). 22 On the other hand, the COVID‐19 pandemic had overwhelmed ICU admission due to respiratory failure. Prophylactic empirical prescription of antibiotics to control the threat of bacterial co‐infection led to an increase in AMR. 23 Even though the risk of co‐infection was low, $70\%$ of admitted COVID‐19 patients in Bangladesh received prophylactic antibiotics, according to a study. Long hospitalization and use of medical devices in the ICU, lack of effective surveillance, and immunocompromised patients due to corticosteroid prescription all contributed to an increase in hospital‐acquired Infection. 24 Even more, the laboratory tests were in shortage 23 and nurse numbers were not in proportion to the patients, 25 which may lead to the empirical use of antibiotics concluded in AMR increase.
Remarkably, during the pandemic, A. baumannii resistance to all mentioned antibiotics has been $100\%$, albeit it was not a frequent microorganism. Likewise, but not at the same level, P. aeruginosa and Citrobacter have had more than $50\%$ resistance to the same antibiotics. It seems that these might be the most challenging post‐pandemic issues.
E. coli has different behavior in juxtaposition to other monitored Gram‐negative bacteria (Supporting Information: Figure 1), which might have emanated from the fact that the majority of E. coli specimens belong to urine for outpatients having a drastic fall the frequency during the pandemic. As observed by several studies in Indonesia 29 and Mexico, a rise in E. coli's sensitivity to cotrimoxazole, ciprofloxacin, and ceftriaxone may have resulted from a decrease in outpatients. 18 However, E. coli resistance to Gentamicin and cefepime, which are among the prescribed drugs during COVID‐19 in hospitals reported by some studies, 18 has increased.
The increasing Gram‐positive bacteria' resistance to cephalexin may were caused by irrational overuse of it. 8 The prevalence of Gram‐positive bacteria has decreased since COVID‐19's presence. 26 Although the number of Gram‐positive samples on which antibiogram tests were done in the current study has not been enough to distinguish a significant trend, the increase in Gram‐positive pathogens' resistance specifically methicillin‐resistance *Staphylococcus aureus* and vancomycin‐resistance enterococcus must be a subject of a detailed study to be determined.
An increase in resistance to the cephalosporin third or fourth generation, such as ceftriaxone, cefotaxime, and cefepime (Supporting Information: Figures 1 and 2), could have been in terms of the indiscriminate use of them reported by two studies in China and Peru, $68\%$ of COVID‐19 patients have had a history of azithromycin, and ceftriaxone administration before admission. Furthermore, $33\%$ of them have had a self‐medication history. 27, 28 The present study has some limitations, including a short period, and limited data; some antimicrobials were not evaluated in all classes, such as carbapenem, penicillin, methicillin, and azithromycin, which are some of the most important antibiotics to consider. Only the data of urine colony count were reported. The clinical status of patients, such as SARS‐COV‐2 infection was not reported. The information about prescribed antimicrobials during the COVID‐19 pandemic was not available. A strong point of the present study would be that all the data were collected from a COVID‐19 referral center. Furthermore, available data allowing us to compare prepandemic, and the pandemic period from the same center make the study more reliable about COVID‐19's effect on AMR. Focusing on some high‐priority, and critical bacteria turns this study into an outstanding one.
## CONCLUSION
Although the accurate effect of COVID‐19 on AMR is not distinguishable yet, an increase in AMR was observed, particularly in Gram‐negative bacteria among which P. aeruginosa and K. pneumonia had a tremendous increase. It may have stemmed from excessive and inappropriate utilization of antibiotics. The lack of fully developed stewardship may have had an intensive effect on AMR increase. More strength surveillance may reduce irrational prescriptions of antibiotics. In addition, well‐developed guidelines to manage COVID‐19 patients, especially the mild ones, and appropriate diagnostic kits are recommended for detecting AMR early on to reduce empirical therapy. Moreover, the biomarkers to differentiate viral and bacterial infections avoid inappropriate antibiotic use. Further studies are required to determine bacterial co‐infection risk factors of COVID‐19 patients. The appliance of a hospital infection control team under a strict protocol can reduce the infection transmission, subsequently AMR. Public awareness to reduce self‐medication, and simultaneously, restriction of antibiotic accessibility over the counter may legally have an impressive effect on their consumption.
## AUTHOR CONTRIBUTIONS
Rahim Raoofi: Conceptualization; investigation; methodology; project administration; supervision; validation; writing—review & editing. Negin Namavari: Conceptualization; data curation; formal analysis; investigation; methodology; validation; visualization; writing—original draft. Vahid Rahmanian: Formal analysis; investigation; methodology; software; visualization; writing—original draft; writing—review & editing. Mohammad Hadi Dousthaghi: Conceptualization; data curation; formal analysis; investigation; software; validation; writing—original draft.
## CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.
## TRANSPARENCY STATEMENT
The lead author Negin Namavari affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.
## DATA AVAILABILITY STATEMENT
The authors acknowledge that data supporting the findings of this study are available in the article [and/or] its supplementary material.
## References
1. Brinkac L, Voorhies A, Gomez A, Nelson KE. **The threat of antimicrobial resistance on the human microbiome**. *Microb Ecol* (2017) **74** 1001-1008. PMID: 28492988
2. Harvard T, Aminov R, Naylor NR. **A brief history of the antibiotic era: lessons learned and challenges for the future**. *Antimicrob Resist Infect Control* (2018) **7** 1-17. PMID: 29312658
3. Medina M‐j, Legido‐Quigley H, Hsu LY, Masys AJ, Izurieta R, Reina Ortiz M. **Antimicrobial Resistance in One Health**. *Global Health Security: Recognizing Vulnerabilities, Creating Opportunities* (2020) 209-229. DOI: 10.1007/978-3-030-23491-1_10
4. Hay SI, Rao PC, Dolecek C. **Measuring and mapping the global burden of antimicrobial resistance**. *BMC Med* (2018) **16** 78. PMID: 29860943
5. Jubeh B, Breijyeh Z, Karaman R. **Resistance of Gram‐positive bacteria to current antibacterial agents and overcoming approaches**. *Molecules* (2020) **25**. DOI: 10.3390/molecules25122888
6. Nieuwlaat R, Mbuagbaw L, Mert D. **Coronavirus disease 2019 and antimicrobial resistance: parallel and interacting health emergencies**. *Cad Saude Publ* (2020) **12** 1-30
7. Caruso P, Maiorino MI, Macera M. **Antibiotic resistance in diabetic foot infection: how it changed with COVID‐19 pandemic in a tertiary care center**. *Diabetes Res Clin Pract* (2021) **175**. PMID: 33845049
8. Saini V, Jain C, Singh NP. **Paradigm shift in antimicrobial resistance pattern of bacterial isolates during the covid‐19 pandemic**. *Antibiotics* (2021) **10** 954. PMID: 34439004
9. Seabra G, Ventura Mendes RF, dos Santos Amorim LFV. **Azithromycin use in COVID‐19 patients: implications on the antimicrobial resistance**. *Curr Top Med Chem* (2021) **21** 677-683. PMID: 34028347
10. Lai CC, Chen SY, Ko WC. **Antibiotics susceptibility of**. *Antibiotics* (2021) **10** 1-8
11. Despotovic A, Milosevic B, Cirkovic A. **The impact of covid‐19 on the profile of hospital‐acquired infections in adult intensive care units**. *Antibiotics* (2021) **10** 1146. PMID: 34680727
12. Rawson TM, Moore LSP, Castro‐Sanchez E. **COVID‐19 and the potential long‐term impact on antimicrobial resistance**. *J Antimicrob Chemother* (2020) **75** 1681-1684. PMID: 32433765
13. Lazzerini M, Barbi E, Apicella A, Marchetti F, Cardinale F, Trobia G. **Delayed access or provision of care in Italy resulting from fear of COVID‐19**. *Lancet Child Adolesc Health* (2020) **4** e10-e11. PMID: 32278365
14. Haberman R, Axelrad J, Chen A. **Covid‐19 in immune‐mediated inflammatory diseases—case series from New York**. *N Engl J Med* (2020) **383** 85-88. PMID: 32348641
15. Nieuwlaat R, Mbuagbaw L, Mertz D. **Coronavirus disease 2019 and antimicrobial resistance: parallel and interacting health emergencies**. *Clin Infect Dis* (2021) **72** 1657-1659. PMID: 32544232
16. Vaughn VM, Gandhi TN, Petty LA. **Empiric antibacterial therapy and community‐onset bacterial coinfection in patients hospitalized with coronavirus disease 2019 (COVID‐19): a multi‐hospital cohort study**. *Clin Infect Dis* (2021) **72** e533-e541. PMID: 32820807
17. Stevens RW, Jensen K, O'Horo JC, Shah A. **Antimicrobial prescribing practices at a tertiary‐care center in patients diagnosed with COVID‐19 across the continuum of care**. *Infect Control Hosp Epidemiol* (2021) **42** 89-92. PMID: 32703323
18. LE L‐J, D F‐R, R F‐C. **Increment antimicrobial resistance during the COVID‐19 pandemic: results from the invifar network**. *Microb Drug Resist* (2022) **28** 338-345. PMID: 34870473
19. Oseran AS, Nash D, Kim C. **The untold toll — the pandemic's effects on patients without Covid‐19**. *Am J Manag Care* (2020) **28** 294-295
20. Huttner BD, Catho G, Pano‐Pardo JR, Pulcini C, Schouten J. **COVID‐19: don't neglect antimicrobial stewardship principles!**. *Clin Microbiol Infect* (2020) **26** 808-10. PMID: 32360446
21. Khoshbakht R, Kabiri M, Neshani A. **Assessment of antibiotic resistance changes during the Covid‐19 pandemic in northeast of Iran during 2020–2022: an epidemiological study**. *Antimicrob Resist Infect Control* (2022) **11** 121. PMID: 36182905
22. Póvoa HCC, Chianca GC, Iorio NLPP. **COVID‐19: an alert to ventilator‐associated bacterial pneumonia**. *Infect Dis Ther* (2020) **9** 417-20. PMID: 32474891
23. Rothe K, Feihl S, Schneider J. **Rates of bacterial co‐infections and antimicrobial use in COVID‐19 patients: a retrospective cohort study in light of antibiotic stewardship**. *Eur J Clin Microbiol Infect Dis* (2021) **40** 859-869. PMID: 33140176
24. Ferreira RL, Da Silva BCM, Rezende GS. **High prevalence of multidrug‐resistant**. *Front Microbiol* (2019) **9**. DOI: 10.3389/fmicb.2018.03198
25. Ghanizadeh A, Najafizade M, Rashki S, Marzhoseyni Z, Motallebi M. **Genetic diversity, antimicrobial resistance pattern, and biofilm formation in**. *BioMed Res Int* (2021) **2021**. DOI: 10.1155/2021/2347872
26. Hirabayashi A, Kajihara T, Yahara K, Shibayama K, Sugai M. **Impact of the COVID‐19 pandemic on the surveillance of antimicrobial resistance**. *J Hosp Infect* (2021) **117** 147-156. PMID: 34562548
27. Barros‐Sevillano JS, Sandoval CP, Alcarraz‐Mundial LS, Barboza JJ. **Automedicación en tiempos de COVID‐19. Una perspectiva desde Perú**. *Gac Med Mex* (2021) **157**
28. Li H, Liu Z, Ge J. **Scientific research progress of COVID‐19/SARS‐CoV‐2 in the first five months**. *J Cell Mol Med* (2020) **24** 6558-6570. PMID: 32320516
|
---
title: Maternal and perinatal outcomes in mixed antenatal care modality implementing
telemedicine in the southwestern region of Colombia during the COVID-19 pandemic
authors:
- María Fernanda Escobar
- Juan Carlos Gallego
- María Paula Echavarria
- Paula Fernandez
- Leandro Posada
- Shirley Salazar
- Isabella Gutierrez
- Juliana Alarcon
journal: BMC Health Services Research
year: 2023
pmcid: PMC10017345
doi: 10.1186/s12913-023-09255-4
license: CC BY 4.0
---
# Maternal and perinatal outcomes in mixed antenatal care modality implementing telemedicine in the southwestern region of Colombia during the COVID-19 pandemic
## Abstract
### Introduction
Contingency measures due to the COVID-19 pandemic limited access to routine prenatal care for pregnant women, increasing the risk of pregnancy complications due to poor prenatal follow-up, especially in those patients at high obstetric risk. This prompted the implementation and adaptation of telemedicine.
### Objective
We aim to evaluate the maternal and perinatal outcomes of patients who received prenatal care in-person and by telemedicine.
### Methods
We conducted a retrospective observational cohort study of pregnant women who received exclusive in-person and alternate (telemedicine and in-person) care from March to December 20,202, determining each group's maternal and neonatal outcomes.
### Results
A total of 1078 patients were included, 156 in the mixed group and 922 in the in-person group. The patients in the mixed group had a higher number of prenatal controls (8 (6–9) vs 6 (4–8) $p \leq 0.001$), with an earlier gestational age at onset (7.1 (6–8.5) vs 9.3 (6.6–20.3), $p \leq 0.001$), however, they required a longer hospital stay (26 (16,$67\%$) vs 86 (9,$33\%$), $$p \leq 0.002$$) compared to those attended in-person; there were no significant differences in the development of obstetric emergencies, maternal death or neonatal complications.
### Discussion
Incorporating telemedicine mixed with in-person care could be considered as an alternative for antenatal follow-up of pregnant women in low- and middle-income countries with barriers to timely and quality health care access.
## Introduction
Maternal and perinatal health are fundamental pillars of a country's public health, being one of the Sustainable Development Goals, so efforts focused on their improvement must always be present [1]. In 2017, the global maternal mortality ratio (MMR) was 211 per 100,000 live births (LB), with almost $99\%$ of deaths registered in low- and middle-income countries (LMIC) [2]; for this same year, Colombia recorded an MMR of 50.7 per 100,000 LB [3]. In 2021, two years after the outbreak of the novel coronavirus pandemic, the MMR in Colombia was 78.3 deaths per 100,000 LB, and nine territorial entities registered a higher than 100 per 100,000 LB [4]. Since the COVID-19 pandemic, a global increase in maternal mortality, near-miss mortality, perinatal mortality, and neonatal morbidity is observed, being more significant in LMICs [5, 6].
Therefore, in confinement settings with a high social cost and economic detriment, it is urgent to prioritize equitable access to high-quality maternal care [5, 7]. Prenatal care has been associated with a decrease in maternal and perinatal mortality and complications related to pregnancy such as hypertensive disorders, intrauterine growth restriction, and preterm delivery [8–11]. The measures taken to address the health emergency by COVID-19 such as the adaptation of the infrastructure and the redistribution of health personnel to support and optimize emergency rooms, intensive care units, and hospitalization services, as well as the relocation in telework of those health workers with high-risk comorbidities, led to a lack of coverage and a reduction of quality prenatal care increasing pregnancy complications [11, 12]. To confront the problem, different organizations encouraged the use of telehealth [11, 13, 14], the use of these new tools in the setting of pregnant women has focused on the management of complications such as gestational diabetes, hypertensive disorders, and obesity [15–19]. Evidence on synchronous telemedicine in prenatal care is limited; however, it has shown similar results compared with exclusively in-person prenatal care [20–23].
Fundación Valle del *Lili is* a quaternary level university hospital located in Cali, which receives highly complex patients referred from the southwestern areas of Colombia. At the beginning of the mandatory confinement in Colombia, the hospital implemented an outpatient telehealth system through videoconference tools [24], which allowed the follow-up of patients from 65 medical specialties, including obstetric patients. The main objective of this research was to compare maternal and perinatal outcomes between pregnant patients who received their prenatal care in an alternate check-up (telemedicine and In-person) and those who received it exclusively in-person.
## Design
We conducted a retrospective observational cohort study of obstetrics patients attended for prenatal care from March 1 to December 31, 2020, at the Fundación Valle del Lili, divided into two groups: Mixed or Telemedicine cohort: those patients who received prenatal check-ups alternating between telemedicine and in-person, with compliance with at least one consultation via telemedicine. In-person cohort: patients who received exclusively in-person prenatal check-ups.
Patients were free to choose if they wanted in-person or telemedicine care according to preference. Patients who needed vital signs or a physical examination to define the treatment plan were directed to receive an in-person check-up. However, no specific medical condition was considered a contraindication for telemedicine follow-up. Records with incomplete prenatal care data and patients who performed check-ups at other institutions were excluded.
The institutional biomedical research ethics committee approved the protocol of this study, informed consent was not required for the study as it was classified as risk-free according to national resolution (No. 008430 of 1993, article 11, numeral A) of the Ministry of Health and Social Protection of Colombia.
## Data collection and variables
We used the ICD-10 codes to identify the pregnant women who received prenatal follow-up at the institution in the period from March 1 to December 31, 2020, the statistics and information management department in conjunction with the telemedicine program identified the patients who received care by telemedicine and those who received exclusively in-person care. We then collected data from the institutional electronic medical record, which were registered in the BdClinic database software. Table 1 shows the variables that were considered to evaluate maternal and neonatal outcomes. Table 1Maternal and neonatal variables considered to determine outcomesMaternal VariablesNeonatal Variables Delivery route or termination of pregnancy Early hospital dischargea Admission to the high complexity obstetrics unit (HCOU)b Access to the intensive care unit Development of obstetric emergency Maternal death Newborn birth weight Need for hospitalization after childbirth Admission to the neonatal intensive care unit *Neonatal deatha* Discharge within the first 48 h after deliveryb High dependency unit for the management of pregnancies classified as high obstetric risk
## Implementation of telemedicine in antenatal care
In response to the health emergency caused by the COVID-19 pandemic, the Fundación Valle del Lili implemented and adapted telemedicine strategies by developing the " Siempre" (Always) program as an alternative for outpatient care, avoiding the spread and exposure to the virus.
The "Siempre" program uses the Microsoft Teams platform, which establishes real-time video calls between the obstetrician-gynecologist and the pregnant women. The clinical practice guidelines of the Colombian Ministry of Health and Social Protection recommend at least ten prenatal check-ups for nulliparous women and seven for multiparous women [25]; however, this guide does not contemplate telemedicine interventions, so the recommendations of the American Society of Gynecology and Obstetrics—Maternal–Fetal Medicine delivered in the guide "MFM guidance for COVID-19" (Table 2) were considered [26]. Women could have more or fewer evaluations than those established, depending on the risk classification according to the validated scale of Herrera and Hurtado [27] and on clinical needs; those patients with high obstetric risk check-ups were programmed monthly until 37 weeks of gestational age, then check-ups were every 15 days until the delivery. Table 2Telemedicine care program integrated into prenatal care in Fundación Valle del LiliGestational ageObstetrics and gynecology follow-upUltrasounds and LabsVia In-PersonVia Telemedicine< 11 weeksX11 – 13 weeksXX18 – 24 weeksXX28 weeksX32 weeksXX36 weeksXX (As required)38 weeks to BirthXPostpartumX Throughout the medical attention, the obstetrician-gynecologist updates the patient's current condition, asks about the presence of symptoms or warning signs, the need for emergency consultations or hospitalizations since the last check-up, and analyzes the last examinations performed. The information is recorded in the institutional electronic medical record system, through which the request for procedures, paraclinical or imaging tests, referral for evaluation by other necessary specialties, and formulation of medications are made. The modality and date of the next appointment are decided with the patient at the end of the video call, and the documents generated are sent in PDF format to each patient's e-mail address.
## Statistical analysis
The research conducted by Yvonne Butler et. al [22], estimated an incidence of cesarean delivery of $15\%$ in pregnant women who were attended in-person and $13\%$ in those attended via telemedicine; with these data, a sample size of 9,448 (4,724 for each group) was calculated for a confidence level of $95\%$ and a power of $80\%$. However, the institutional department of statistics and information management indicated that during the defined period a total of 1808 prenatal care were performed, being 190 via telemedicine, which offers a power of $81\%$ with a $95\%$ confidence level to identify $15\%$ of cesarean deliveries in in-person consultations and $23\%$ via telemedicine, with an Unexposed/Exposed ratio of $25\%$, requiring 940 patients attended exclusively in person.
A total of 1,618 in-person visits were carried out in the determined period, so a simple random sampling was performed by enumerating each patient, and then the 922 patients were randomly obtained using the Random number generator Comprehensive Version on the calculator.net website.
The Shapiro–Wilk test was used to evaluate the normality distribution of the numerical variables, taking a p-value < 0.05 as a significance value; thus, medians and their respective interquartile ranges were used to describe these variables. The qualitative variables were summarized through percentages and presented in frequency tables. The Mann–Whitney-Wilcoxon statistical tests were used to evaluate the difference in the sociodemographic or clinical numerical variables between cohorts, while the differences in the qualitative variables were evaluated using the Chi2 test.
The Chi2 test will be used to determine the differences in each maternal and perinatal outcome incidence. In case of being a dichotomous outcome and the expected value is less than 5 in any category, Fisher's test will be used. Statistically significant differences will be considered if the p-value < 0.05.
## Results
A total of 1808 patients received antenatal care check-ups (190 via telemedicine, and 1618 in-person) from March 1 to December 31, 2020. After using simple random sampling for the in-person group and applying inclusion and exclusion criteria we obtained 156 ($14.5\%$) patients attended by a Mixed modality, being the exposed cohort; and 922 ($85.5\%$) patients in-person modality, being the control cohort (Fig. 1).Fig. 1Distribution of patients with antenatal check-ups by telemedicine versus exclusively in-person modality of attention The median age was 31 (IQR 28–35), and most of the participants lived in urban areas (95,$27\%$); in the mixed group 63,$46\%$ had a bachelor’s degree being higher than the in-face group (48,$26\%$, $$p \leq 0.041$$) the other sociodemographic characteristics were no statistically significant differences between both cohorts. Women in the mixed group started their routine antenatal care earlier with a median gestational age at onset of 7.1 weeks (IQR 6—8.5) versus 9.3 weeks (IQR 6.6—20.3) in the in-person group ($p \leq 0.001$); they also presented a greater number of antenatal appointment received than the in-person group (8 [IQR 6–9] vs. 6 [IQR 4–8], $p \leq 0.001$) (Table 3).Table 3Sociodemographic and clinical characteristics of the pregnant women in the mixed group vs. In-person groupVariablesTotal ($$n = 1078$$)Mixed antenatal control ($$n = 156$$)In-person antenatal control ($$n = 922$$)P valueAge, years, median (IQR)31 (28—35)31 (28—35)31 (28—35)0.7608Area of residence, n (%) Urban area1027 [95,27]151 [96,79]876 [95,01]0.555 Rural area28 [2,60]3 [1,92]25 [2,71] No data23 [2,13]2 [1,23]21 [2,28]Occupation, n (%) Unemployed51 [4,73]9 [5,77]42 [4,56]0.263 Employed659 [61,13]107 [68,59]552 [59,87] Independent100 [9,28]10 [6,41]90 [9,76] Housewife119 [11,04]14 [8,97]105 [11,39] No data149 [13,82]16 [10,26]133 [14,43]Scholarship, n (%) Basic5 [0,46]0 [0]5 [0,54]0.041 High school256 [23,75]28 [17,95]228 [24,73] Bachelor degree544 [50,46]99 [63,46]445 [48,26] Master's or doctorate68 [6,31]9 [5,77]59 [6,40] No data205 [19,02]20 [12,82]185 [20,07] Number of antenatal contacts, median (IQR)7 (4 – 8)8 (6—9)6 (4—8)< 0,001* Gestational age at the time of admission to routine antenatal care, weeks (IQR)8.5 (6.4 – 17.3)7.1 (6 – 8.5)9.3 (6.6 – 20.3)< 0,001* Gestational age at pregnancy termination a (weeks), median (IQR)38.2 (37.3 – 39)38.2 (37.2 – 39)38.2 (34.7—39)0,498 High-risk pregnancy, n (%)702 [65,12]108 [69,23]594 [64,43]0.244 Emergency admission, n (%)529 [49,07]69 [44,23]460 [49,89]0.191 Gravidity, median (IQR)2 (1—2)2 (1—2)2 (1—2)0.189 Twin pregnancy, n (%)17 [1,58]2 [1,28]16 [1,63]0.749Abbreviations: IQR Interquartile range, AC Antenatal control* p-value less than 0.001 We used the Herrera & Hurtado scale to classify the biopsychosocial risk of the pregnant woman, finding an elevated percentage of high-risk pregnancies in the total sample (65,$12\%$), without differences between cohorts (Mixed 108 (69,$23\%$) vs. In-person 594 (64,$43\%$), $$p \leq 0.244$$). Likewise, admission to the emergency room for obstetric causes during the gestational period was similar for both cohorts (Mixed 69 (44,$23\%$) vs. In-person 460 (49,$89\%$), $$p \leq 0.191$$) (Table 3).
Regarding maternal outcomes, 72,$26\%$ of the total patients required cesarean section as a way of delivery, without difference between both groups (Mixed 117 ($75\%$) vs. In-person 662 (71,$8\%$), $$p \leq 0.424.$$ After the attention of the obstetric event, both groups had a higher percentage of women with an early hospital discharge, but with a higher frequency in the prolongation of the hospitalization time in the mixed group (mixed 26 (16,$67\%$) vs. in-person 86 (9,$33\%$) $$p \leq 0.002$$). Additionally, we evaluated the need for admission to the High complexity obstetric unit for the surveillance and care of childbirth, evidencing a higher frequency of admission in the in-person group (mixed 39 ($25\%$) vs. in-person 332 (36,$01\%$), $$p \leq 0.005$$). None of the other maternal outcomes evaluated reached a statistically significant difference between cohorts (Table 4).Table 4Maternal outcomes of women treated via telemedicine vs. In-personVariableTotal ($$n = 1078$$)Mixed antenatal control ($$n = 156$$)In-person antenatal control ($$n = 922$$)P valueFinal state of pregnancy a, n (%) Vaginal delivery265 [24,58]32 [20,51]233 [25,27]0.424 Cesarean section779 [72,26]117 [75]662 [71,8] Abortion34 [3,15]7 [4,49]27 [2,92]Hospital discharge, n (%) Not hospitalized29 [2,69]8 [5,13]21 [2,28]0.002 Late hospital discharge c112 [10,39]26 [16,67]86 [9,33] Early hospital discharged b935 [86,73]122 [78,21]813 [88,18] Admission to HCOU, n (%)371 [34,42]39 [25]332 [36,01]0.005 Admission to ICU, n (%)38 [3,53]6 [3,85]32 [3,47]0.838 Obstetric emergency, n (%)405 [37,57]63 [40,38]342 [37,09]0.845 Maternal mortality, n (%)000N/AAbbreviations: HCOU High complexity obstetric unit, ICU Intensive care unit. N/A: *Not applya* This variable included: Abortion, voluntary interruption of pregnancy, vaginal delivery, and cesarean sectionb Late hospital discharge: before 48 h after deliveryc Early hospital discharge: within the first 48 h after delivery Finally, no statistically significant differences were identified between groups in the evaluated neonatal outcomes. There was a greater number of newborns with an adequate birth weight (2500—3999 g: mixed 127 (81,$41\%$) vs. in-person 792 (85,$9\%$), $$p \leq 0.171$$), most of the newborns did not require hospitalization (Mixed 104 (66,$67\%$) vs. in-person 621 (67,$35\%$), $$p \leq 0.435$$), a lower percentage required admission to the ICU / Neonatal (mixed 17 (10,$90\%$) vs. in-person 122 (13,$63\%$), $$p \leq 0.453$$) and only 29 deaths occurred (mixed 7 (4,$49\%$) vs. in-person 22 (2,$38\%$), $$p \leq 0.088$$) (Table 5).Table 5Neonatal outcomes of women attended mixed vs. In-person modalitiesVariableTotal ($$n = 1078$$)Mixed antenatal control ($$n = 156$$)In-person antenatal control ($$n = 922$$)P valueNewborn weight, n (%) < 500 gr6 [0,56]2 [1,28]4 [0,43]0.171 500—999 gr4 [0,37]2 [1,28]2 [0,22] 1000—1499 gr3 [0,28]0 [0]3 [0,33] 1500—2499 gr86 [7,98]15 [9,62]71 [7,70] 2500—3999 gr919 [85,25]127 [81,41]792 [85,90] ≥ 4000 gr40 [3,71]6 [3,85]34 [3,69]Hospital discharge, n (%) Not hospitalized725 [67,25]104 [66,67]621 [67,35]0.435 Late hospital discharge a160 [14,84]19 [12,18]141 [15,29] Early hospital discharged b134 [12,43]23 [14,74]111 [12,04] No data59 [5,47]10 [6,41]49 [5,31] Admission to Neonatal ICU, n (%)139 [12,89]17 [10,90]122 [13,63]0.453Perinatal mortality, n (%) No perinatal mortality1024 [94,99]146 [93,56]878 [95,23]0.088 Before childbirth23 [2,13]6 [3,85]17 [1,84] Intrapartum2 [0,19]1 [0,64]1 [0,11] Postpartum4 [0,37]04 [0,43]Abbreviations: ICU Intensive care unita Late hospital discharge: before 48 h after deliveryb Early hospital discharge: within the first 48 h after delivery
## Discussion
The maternal and perinatal outcomes of women who received attention in mixed modality prenatal care (in-person plus telemedicine) were similar to those recorded in women with exclusively in-person care. These results indicate no adverse impact with the incorporation of mixed modality telemedicine in antenatal follow-up compared with exclusively in-person visits. Inclusive, telemedicine allowed earlier admission to the program and a more significant number of evaluations related to greater maternal satisfaction and decreased adverse outcomes, such as perinatal mortality [28].
For pregnant women, the COVID-19 pandemic and government restrictions imposed to prevent the spread of the disease led to delayed entry to prenatal care, less care, and increased related complications [12, 29]. The disruption of services focused on maternal and newborn health, mainly in low and middle-income countries, leads to a significant increase in the number of maternal and perinatal deaths, and even a reduction in coverage and an increase in waiting times of $10\%$ can contribute to 253,000 perinatal deaths and 12,200 additional maternal deaths [30]. Faced with situations such as the current contingency, seeking strategies to avoid this lack of continuity in services is essential to prevent indirect complications. Because of that, we highlight how an outpatient teleconsultation program allowed an earlier admission to the antenatal program and a higher number of evaluations, managing to face the current health problem. This is how different organizations promoted the integration of telemedicine tools into prenatal control programs, leading to restructure the traditional in-person services programs [30, 31].
Although the pandemic promoted telemedicine, this is not a new practice in prenatal care. Butler Tobah et al. conducted a randomized clinical trial to evaluate the effectiveness of an antenatal program for low-risk women, which reduced the number of in-person care and were replaced by virtual visits and monitoring devices without differences between groups and with had greater satisfaction and less stress related to pregnancy in the mixed group [22]. With the arrival of the pandemic, multiple health centers around the world implemented new care protocols that promoted social distancing using telehealth tools. However, the evidence so far has focused on the description of these programs carried out in developed countries, with little information regarding the clinical outcomes of patients and newborns [32]. Prospective studies are therefore needed to allow a better comparison of prenatal care modalities among different risk groups of pregnant women in low- and middle-income countries.
Even in high-income countries, one of the concerns of incorporating telemedicine in prenatal care programs is the applicability of this modality in high-risk pregnant women. In our case, we have highlighted how most of the women treated fell into this category. Since no specific clinical condition limited the use of any of the care modalities, no difference was found in this condition between both cohorts. Telemedicine tools have proven useful to reduce in-person care when these are not viable in pregnancies with low and high risk with specific conditions such as gestational diabetes, cardiovascular diseases, and fetal and genetic disorders, leading to avoiding unnecessary exposure of women to hospital environments [21, 33]. Published studies found no significant differences in the incidence of intrauterine growth restriction, preeclampsia, gestational diabetes, or neonatal morbidity and mortality [34]. Our results are aligned, having found no differences in the causes of obstetric emergency at the time of birth. The high cesarean rate reported in both groups ($61.1\%$ in the telemedicine group and $65.5\%$ in the in-person group) is related, above the national average reported in 2019 ($44.5\%$) [35], possibly due to the high clinical complexity of the women managed in our institution. Likewise, the length of hospital stay for the care of the obstetric event (which was longer in the telemedicine group) may be associated with the way of termination of pregnancy since cesarean delivery has been identified as an independent factor to prolong the stay, with an average follow-up between 2.5—9.3 days [36].
With these results, one of the most significant discussions is the possibility of extending this care modality to the Colombian territory and other LMICs. The coverage of antenatal programs in Colombia continues to be low, with $62.9\%$ of all births with less than seven evaluations, and $4.8\%$ without any attention, according to national reports for 2019 [37]. Additionally, according to the country's Ministry of Information Technologies and Communications, in Colombia, there is a wide gap in fixed internet access between regions, with the access of 25.3 connections for every 100 residents in Bogotá, up to 5 for every 100 inhabitants in the Guaviare or Guainía, departments with the lowest human development index [38]. Telemedicine coverage is more complex in these areas with less internet access, and paradoxically, the highest MMRs are concentrated. Duryea et al. [ 23] implemented a prenatal care program where synchronous telemedicine was incorporated using only audio and compared perinatal outcomes with those who received conventional management. Consistent with what was found in our research, the telemedicine group had an earlier admission to prenatal care (11 weeks) and a more significant number of evaluations (9.8 vs. 9.4). They found no differences in neonatal mortality, admission to the NICU, hypertensive disorders, postpartum hemorrhage, or type of delivery. They demonstrated how it is possible to expand antenatal services' coverage through different telehealth modalities allowing early admission and increased follow-up during gestation, without increasing adverse clinical outcomes, which could be an alternative to explore in regions with low internet speed.
## Limitations
The retrospective nature of the research leads to a higher risk of biases, which we tried to control by including all pregnant women evaluated with at least one assessment via telemedicine in the exposed cohort and using random sampling to select the control group. Additionally, after applying exclusion criteria, we could not reach the sample size initially proposed, which could limit the identification of statistically significant differences between groups. Therefore, it would be important to carry out prospective studies comparing the modalities of care in low and middle-income countries, making it possible to evaluate the results in relation to temporality.
## Conclusions
The results of this research are encouraging by showing telemedicine as a tool that increases access to healthcare, leads an earlier entry into prenatal control, and allows continuing with the delivery of an effective and quality prenatal control, even if it is only a single check-up; showing itself as an alternative when in-person evaluation is not the first option. Although this program was quickly implemented during the COVID-19 contingency and understanding the limitations of telehealth to perform a physical examination that is essential in pregnant patients, the benefits found may prompt the use of telemedicine to expand coverage of the prenatal control programs in LMIC. The following steps will be to build and validate safe, quality alternating care models based on this evidence, which will accelerate the adoption of technology and reduce the use of technologies for the care of pregnant women.
## References
1. Kruk ME, Gage AD, Arsenault C. **High-quality health systems in the Sustainable Development Goals era: time for a revolution**. *Lancet Glob Health* (2018.0) **6** e1196-e1252. DOI: 10.1016/S2214-109X(18)30386-3
2. 2.WHO, Unicef, UNFPA, And WBG, D. UNP, World Health Organization. Trends in maternal mortalityto 20172000UNICEF, UNFPA, World Bank Group and the United Nations Population DivisionWHO2019. *to 2017* (2000.0) 2019
3. 3.Instituto Nacional de Salud. INFORME DE EVENTO MORTALIDAD MATERNA, COLOMBIA, AÑO 2017. Bogotá; 2018 Apr.
4. 4.Instituto Nacional de Salud de Colombia. Boletín Epidemiológico Semanal Semana epidemiológica 52. 26 de diciembre del 2021 al 1 de enero de 2022. https://www.ins.gov.co/buscador-eventos/BoletinEpidemiologico/2021_Boletin_epidemiologico_semana_52.pdf
5. Chmielewska B, Barratt I, Townsend R, Kalafat E, van der Meulen J. **Effects of the COVID-19 pandemic on maternal and perinatal outcomes: a systematic review and meta-analysis**. *Lancet Glob Health* (2021.0) **9** e759-e772. DOI: 10.1016/S2214-109X(21)00079-6
6. Villar J, Ariff S, Gunier RB, Thiruvengadam R. **Maternal and neonatal morbidity and mortality among pregnant women with and without COVID-19 infection: the intercovid multinational cohort study**. *JAMA Pediatr* (2021.0) **175** 817-826. DOI: 10.1001/jamapediatrics.2021.1050
7. Bonaccorsi G, Pierri F, Cinelli M, Flori A, Galeazzi A. **Economic and social consequences of human mobility restrictions under COVID-19**. *Proc Natl Acad Sci U S A* (2020.0) **117** 15530-15535. DOI: 10.1073/pnas.2007658117
8. Bishai DM, Cohen R, Alfonso YN, Adam T, Kuruvilla S, Schweitzer J. **Factors contributing to maternal and child mortality reductions in 146 low- and middle-income countries between 1990 and 2010**. *PLoS One* (2016.0) **11** e0144908. DOI: 10.1371/journal.pone.0144908
9. Shiferaw K, Mengiste B, Gobena T, Dheresa M. **The effect of antenatal care on perinatal outcomes in Ethiopia: A systematic review and meta-analysis**. *PLoS One* (2021.0) **16** 1-19. DOI: 10.1371/journal.pone.0245003
10. **Countdown to 2030: tracking progress towards universal coverage for reproductive, maternal, newborn, and child health**. *Lancet* (2018.0) **391** 1538-1548. DOI: 10.1016/S0140-6736(18)30104-1
11. Bhutta ZA, Black RE. **Global maternal, newborn, and child health–so near and yet so far**. *N Engl J Med* (2013.0) **369** 2226-2235. DOI: 10.1056/NEJMra1111853
12. Roberton T, Carter ED, Chou VB, Stegmuller AR, Jackson BD, Tam Y. **Early estimates of the indirect effects of the COVID-19 pandemic on maternal and child mortality in low-income and middle-income countries: a modelling study**. *Lancet Glob Heal* (2020.0) **8** e901-8. DOI: 10.1016/S2214-109X(20)30229-1
13. Turrentine M, Ramirez M, Monga M, Gandhi M, Swaim L, Tyer-Viola L. **Rapid deployment of a drive-through prenatal care model in response to the coronavirus disease 2019 (COVID-19) pandemic**. *Obstet Gynecol* (2020.0) **136** 29-32. DOI: 10.1097/AOG.0000000000003923
14. Kasaven LS, Saso S, Barcroft J, Yazbek J, Joash K, Stalder C. **Implications for the future of obstetrics and gynaecology following the COVID-19 pandemic: a commentary**. *BJOG An Int J Obstet Gynaecol* (2020.0) **127** 1318-1323. DOI: 10.1111/1471-0528.16431
15. Van Den Heuvel JFM, Groenhof TK, Veerbeek JHW, Van Solinge WW, Lely AT, Franx A. **eHealth as the next-generation perinatal care: An overview of the literature**. *J Med Internet Res* (2018.0) **20** e202. DOI: 10.2196/jmir.9262
16. Xie W, Dai P, Qin Y, Wu M, Yang B, Yu X. **Effectiveness of telemedicine for pregnant women with gestational diabetes mellitus: An updated meta-analysis of 32 randomized controlled trials with trial sequential analysis**. *BMC Pregnancy Childbirth* (2020.0) **20** 1-14. DOI: 10.1186/s12884-020-02892-1
17. Kalafat E, Benlioglu C, Thilaganathan B, Khalil A. **Home blood pressure monitoring in the antenatal and postpartum period: A systematic review meta-analysis**. *Pregnancy Hypertens* (2020.0) **19** 44-51. DOI: 10.1016/j.preghy.2019.12.001
18. Khalil A, Perry H, Lanssens D, Gyselaers W. **Telemonitoring for hypertensive disease in pregnancy**. *Expert Rev Med Devices* (2019.0) **16** 653-61. DOI: 10.1080/17434440.2019.1640116
19. Ferrara A, Hedderson MM, Brown SD, Ehrlich SF, Tsai AL, Feng J. **A telehealth lifestyle intervention to reduce excess gestational weight gain in pregnant women with overweight or obesity (GLOW): a randomized, parallel-group, controlled trial**. *Lancet Diabetes Endocrinol* (2020.0) **8** 490-500. DOI: 10.1016/S2213-8587(20)30107-8
20. Greiner AL. **Telemedicine applications in obstetrics and gynecology**. *Clin Obstet Gynecol* (2017.0) **60** 853-866. DOI: 10.1097/GRF.0000000000000328
21. Palmer K, Davies-Tuck M, Tanner M, Rindt A, Papacostas K, Giles M. **Widespread implementation of a low-cost telehealth service in the delivery of antenatal care during the COVID-19 pandemic: an interrupted time-series analysis**. *Lancet* (2021.0) **398** 41-52. DOI: 10.1016/S0140-6736(21)00668-1
22. Butler Tobah YS, LeBlanc A, Branda ME, Inselman JW, Morris MA, Ridgeway JL. **Randomized comparison of a reduced-visit prenatal care model enhanced with remote monitoring**. *Am J Obstet Gynecol* (2019.0) **221** 638.e1-638.e8. DOI: 10.1016/j.ajog.2019.06.034
23. Duryea EL, Adhikari EH, Ambia A, Spong C, McIntire D, Nelson DB. **Comparison between in-person and audio-only virtual prenatal visits and perinatal outcomes**. *JAMA Netw Open* (2021.0) **4** 1-9. DOI: 10.1001/jamanetworkopen.2021.5854
24. Escobar MF, Henao JF, Prieto D, Echavarria MP, Gallego JC. **Teleconsultation for outpatient care of patients during the Covid-19 pandemic at a University Hospital in Colombia**. *Int J Med Inform* (2021.0) **155** 104589. DOI: 10.1016/j.ijmedinf.2021.104589
25. 25.Centro Nacional de Investigación en Evidencia y Tecnologías en Salud CINETS. Guías de Práctica Clínica para la Prevención, Detección Temprana y Tratamiento de las Complicaciones del Embarazo, Parto o Puerperio para uso de Profesionales de Salud. 2013. 84 p.
26. Boelig RC, Saccone G, Bellussi F, Berghella V. **MFM guidance for COVID-19**. *Am J Obstet Gynecol MFM* (2020.0) **2** 100106. DOI: 10.1016/j.ajogmf.2020.100106
27. Herrera JA, Gao E, Shahabuddin AKM, Lixia D, Wei Y, Faisal M. **Evaluación periódica del riesgo biopsicosocial prenatal en la predicción de las complicaciones maternas y perinatales en Asia 2002–2003**. *Colomb Med* (2006.0) **37** 6-14
28. 28.World Health Organization. WHO Recommendations on Antenatal Care for a Positive Pregnancy Experience: Summary. Geneva, Switzerland: WHO 2018. Highlights and Key Messages from the World Health Organization’s 2016 Global Recommendations for Routine Antenatal Care. 2018. Available from: https://apps.who.int/iris/bitstream/handle/10665/259947/WHO-RHR-18.02-eng.pdf. [Cited 27 Jul 2021].
29. Goyal M, Singh P, Singh K, Shekhar S, Agrawal N, Misra S. **The effect of the COVID-19 pandemic on maternal health due to delay in seeking health care: Experience from a tertiary center**. *Int J Gynecol Obstet* (2021.0) **152** 231-235. DOI: 10.1002/ijgo.13457
30. 30.Coronavirus ( COVID-19 ) Infection in Pregnancy. R Coll Obstet Gynaecol. 2021;(February):1–98. Available from: https://www.rcog.org.uk/globalassets/documents/guidelines/2021-02-19-coronavirus-covid-19-infection-in-pregnancy-v13.pdf
31. **Maintaining essential health services: operational guidance for the COVID-19 context**. *World Heal Organ* (2020.0) **1** 1-55
32. Montagnoli C, Zanconato G, Ruggeri S, Cinelli G, Eugenio A. **Restructuring maternal services during the covid-19 pandemic: Early results of a scoping review for non-infected women**. *Midwifery* (2021.0) **94** 102916. DOI: 10.1016/j.midw.2020.102916
33. Aziz A, Zork N, Aubey JJ, Baptiste CD, D’alton ME, Emeruwa UN. **Telehealth for high-risk pregnancies in the setting of the COVID-19 pandemic**. *Am J Perinatol* (2020.0) **37** 800-808. DOI: 10.1055/s-0040-1712121
34. Leighton C, Conroy M, Bilderback A, Kalocay W, Henderson JK, Simhan HN. **Implementation and impact of a maternal-fetal medicine telemedicine program**. *Am J Perinatol* (2019.0) **36** 751-758. DOI: 10.1055/s-0038-1675158
35. 35.DANE. DIRECCIÓN DE CENSOS Y DEMOGRAFÍA ESTADÍSTICAS VITALES - EEVV [Internet]. Vol. 2019, CIFRAS DEFINITIVAS AÑO 2019. 2020. Available from: https://www.dane.gov.co/files/investigaciones/poblacion/cifras-definitivas-2019.pdf
36. Campbell OMR, Cegolon L, Macleod D, Benova L. **Length of stay after childbirth in 92 countries and associated factors in 30 low- and middle-income countries: compilation of reported data and a cross-sectional analysis from nationally representative surveys**. *PLoS Med* (2016.0) **13** 1-24. DOI: 10.1371/journal.pmed.1001972
37. 37.Departamento Administrativo Nacional de Estadística. Boletín Técnico Estadísticas Vitales (EEVV). Bogotá D.C; 2022 Feb.
38. 38.Ministerio de Tecnologías de La Información y Comunicaciones. ¿Cómo está el país en conexiones de internet?. 2020. Available from: https://www.mintic.gov.co/portal/inicio/Sala-de-prensa/MinTIC-en-los-medios/151654:Como-esta-el-pais-en-conexiones-de-internet. [Cited 27 Jul 2021].
|
---
title: 'Acetyl-CoA carboxylase inhibitor increases LDL-apoB production rate in NASH
with cirrhosis: prevention by fenofibrate'
authors:
- Mohamad Dandan
- Julia Han
- Sabrina Mann
- Rachael Kim
- Kelvin Li
- Hussein Mohammed
- Jen-Chieh Chuang
- Kaiyi Zhu
- Andrew N. Billin
- Ryan S. Huss
- Chuhan Chung
- Robert P. Myers
- Marc Hellerstein
journal: Journal of Lipid Research
year: 2023
pmcid: PMC10017426
doi: 10.1016/j.jlr.2023.100339
license: CC BY 4.0
---
# Acetyl-CoA carboxylase inhibitor increases LDL-apoB production rate in NASH with cirrhosis: prevention by fenofibrate
## Body
Over 90 million Americans have nonalcoholic fatty liver disease (NAFLD), a condition characterized by excessive liver fat and chronic inflammation [1, 2]. The cause is unclear but it is associated with obesity, diabetes, and metabolic syndrome [3]. In a subset of individuals with NAFLD, progression to nonalcoholic steatohepatitis (NASH), cirrhosis, and hepatocellular carcinoma may occur [3, 4]. Dyslipidemias are common in NAFLD patients and are associated with increased risk of cardiovascular disease and progression to NASH [5, 6, 7]. Hypertriglyceridemia is particularly common in NAFLD and can be influenced by pharmacological treatment [5, 8].
An attractive therapeutic target for NAFLD is inhibition of acetyl-CoA carboxylase (ACC), which catalyzes the rate limiting step of hepatic de novo lipogenesis (DNL) and regulates fatty acid oxidation [8, 9]. Interestingly, observations in NASH patients in phase 2 clinical trials, ACC knockout mouse models, and preclinical models exhibited not only reductions in liver fat content but also hypertriglyceridemia [8, 10, 11, 12, 13, 14, 15, 16]. The latter was unexpected, as reduced hepatic malonyl-CoA production by ACC inhibition was anticipated both to reduce synthesis and to increase oxidation of fatty acids in the liver [17]. In addition, ACC inhibitors (ACCi) in NASH patients has been reported to increase apoB-containing lipoproteins as well as VLDL particle number, triglyceride (TG) content, and secretion [8, 10, 11, 12, 13, 14, 15, 16].
A key question is whether this effect of ACCi acts on the liver through increased production of apoB100-containing particles, or on tissue clearance of plasma lipids, or apoB-containing lipoproteins, as these may confer differences in atherogenicity and suggest different treatment approaches. Data from animal models have suggested that ACCi can cause both changes in hepatic lipid metabolism [8] and peripheral lipoprotein lipase activity [11] but definitive data in humans are not available. During the process of metabolic conversion of VLDL to LDL, apoB100, the main structural protein of VLDL and LDL particles, remains intact, whereas receptor-mediated uptake removes the entire particle, including apoB [18, 19]. Accordingly, LDL-apoB production and clearance kinetics may be useful as a window into the upstream behavior and dynamics of apoB-containing particles and may suggest the tissue site of action of ACCi that alters plasma lipid and lipoprotein levels. In addition, LDL-apoB production rates are of interest in their own right in context of potential atherogenicity [5, 6, 7].
The half-life of VLDL-apoB is rapid (hours) while LDL-apoB exhibits a half-life of 2–5 days [20, 21, 22, 23, 24, 25]. Stable isotopic metabolic tracers such as heavy water (2H2O) can be used to measure synthesis and removal rates of blood proteins, including apolipoproteins [24, 26] such as apoB100 in VLDL and LDL [25]. In humans, 2H-label in body water rapidly equilibrates throughout all tissues and 2H-label rapidly enters free nonessential amino acids during intermediary metabolic processes, but not into peptide-bound amino acids [26], thereby allowing newly synthesized proteins to be labeled and measured.
Here, as part of studies to measure the effects of ACCi treatment on hepatic DNL [12], NASH patients were given heavy water before and after experimental treatment with the ACCi, firsocostat. We measured the kinetics of LDL-apoB in plasma to explore the underlying metabolic mechanisms associated with reported hypertriglyceridemia and changes of apoB particle number in ACCi-treated patients [8, 10, 11, 12, 14, 16]. The primary questions were whether ACCi treatment alters the total production rate and/or the replacement rate constant (clearance) of apoB-containing particles and whether stage of liver disease influences the apoB kinetic response to ACCi treatment. In addition, we evaluated the preventive effects of concurrent therapy with the lipid-lowering agent fenofibrate and firsocostat on plasma TG concentrations and apoB kinetics.
## Abstract
Treatment with acetyl-CoA carboxylase inhibitors (ACCi) in nonalcoholic steatohepatitis (NASH) may increase plasma triglycerides (TGs), with variable changes in apoB concentrations. ACC is rate limiting in de novo lipogenesis and regulates fatty acid oxidation, making it an attractive therapeutic target in NASH. Our objectives were to determine the effects of the ACCi, firsocostat, on production rates of plasma LDL-apoB in NASH and the effects of combined therapy with fenofibrate. Metabolic labeling with heavy water and tandem mass spectrometric analysis of LDL-apoB enrichments was performed in 16 NASH patients treated with firsocostat for 12 weeks and in 29 NASH subjects treated with firsocostat and fenofibrate for 12 weeks. In NASH on firsocostat, plasma TG increased significantly by $17\%$ from baseline to week 12 ($$P \leq 0.0056$$). Significant increases were also observed in LDL-apoB fractional replacement rate (baseline to week 12: 31 ± 20.2 to 46 ± $22.6\%$/day, $$P \leq 0.03$$) and absolute synthesis rate (ASR) (30.4–45.2 mg/dl/day, $$P \leq 0.016$$) but not plasma apoB concentrations. The effect of firsocostat on LDL-apoB ASR was restricted to patients with cirrhosis (21.0 ± 9.6 at baseline and 44.2 ± 17 mg/dl/day at week 12, $$P \leq 0.002$$, $$n = 8$$); noncirrhotic patients did not change (39.8 ± 20.8 and 46.3 ± 14.8 mg/dl/day, respectively, $$P \leq 0.51$$, $$n = 8$$). Combination treatment with fenofibrate and firsocostat prevented increases in plasma TG, LDL-apoB fractional replacement rate, and ASR. In summary, in NASH with cirrhosis, ACCi treatment increases LDL-apoB100 production rate and this effect can be prevented by concurrent fenofibrate therapy.
## Graphical abstract
Graphical abstract: ACCi treatment alone (green arrow) leads to increased apoB particle production in the liver as shown by increased production of LDL-apoB. Fenofibrate combined with ACCi treatment (red inhibitory symbol) prevented the increased LDL-apoB particle production.
## Reagents
Hyclone molecular grade water was obtained from GE health care. Sodium chloride, formic acid, acetonitrile, and methanol were obtained from Thermo Fisher Scientific. Tris base buffer, ethylenediaminetetraacetic acid, acetic acid, ammonium bicarbonate, tris(2-carboxyethyl) phosphine, iodoacetamide, and proteomics grade trypsin were obtained from Sigma-Aldrich.
## Patient treatment, characteristics, and clinical measurements
Adults 18–75 years of age with suspected NASH were studied in a phase 2a clinical trial of the ACCi, firsocostat, and fenofibrate (ClinicalTrials.gov Identifier: NCT02781584). All NASH subjects ($$n = 20$$) were administered 20 mg of firsocostat orally once daily for 12 weeks [12, 14]. Of the 20 NASH subjects, 10 had F2-F3 fibrosis and 10 had cirrhosis (F4). The noncirrhotic NASH subjects treated with firsocostat subjects were enrolled with noninvasive tests using the following parameters: Screening FibroTest® < 0.75, unless a historical liver biopsy within 12 months of screening does not reveal cirrhosis, MRI-estimated proton density fat fraction with ≥ $10\%$ steatosis, magnetic resonance elastography (MRE) with liver stiffness ≥2.88 kPa, or historical liver biopsy within 12 months of screening consistent with NASH (defined as the presence of steatosis, inflammation, and ballooning) and with stage 2–3 fibrosis according to the NASH Clinical Research Network classification (or equivalent). For cirrhotic NASH subjects treated with firsocostat, patients must have a clinical diagnosis of NAFLD and have at least one of the following criteria (a–d): a) Screening MRE with liver stiffness ≥4.67 kPa, b) A historical FibroScan® ≥ 14 kPa within 6 months of Screening, c) Screening FibroTest® ≥ 0.75, and d) A historical liver biopsy consistent with stage 4 fibrosis according to the NASH Clinical Research Network classification (or equivalent). Additional details of patient clinical characteristics have been also described elsewhere [12, 14]. For the cohort of NASH subjects treated with fenofibrate and firsocostat combination therapy, all subjects had hypertriglyceridemia (TG > 150 and < 500 mg/dl) and advanced fibrosis (F3-F4) due to NASH, as determined by historic liver biopsy or liver stiffness by MRE ≥ 3.64 kPa or transient elastography (FibroScan; Echosens, Paris, France) ≥ 9.9 kPa [12]. A historical liver biopsy was conducted within 6 months of screening consistent with NASH and bridging fibrosis (F3) or within 12 months of screening consistent with NASH and compensated cirrhosis (F4) in the opinion of the investigator. All patients were either pretreated with a low (48 mg) or high dose (145 mg) of fenofibrate once daily for two weeks, then a combination of firsocostat 20 mg daily plus fenofibrate at 48 mg/day ($$n = 14$$) or 145 mg/day ($$n = 15$$) for 24 weeks (supplemental Fig. S3A).
## Heavy water labeling protocol and measurements
Heavy water labeling was performed as part of labeling studies to investigate hepatic DNL and fibrogenesis [12]. Plasma samples were taken at day 3 (baseline) and again during week 11 of treatment for LDL-apoB kinetics. 2H2O was administered for seven days in each of the labeling periods, with subjects drinking 50 ml of $70\%$ 2H2O three times daily. During each labeling period, average body 2H2O enrichments rose to ∼0.01 fractional enrichment ($1\%$) (see supplemental Fig. S1), as previously described [12]. Blood samples were drawn after 12 hours of overnight fasting. Heavy water enrichments in each subject were analyzed by distillation followed by acetone exchange and measured via gas chromatography mass spectrometry [26].
## Sample preparation
Lipoproteins were isolated via preparative ultracentrifugation [27]. NativePAGE™ Novex® Bis-Tris Gels using XCell™ SureLock™ Mini-Cell from Life Technologies was employed to further purify LDL-apoB100 from other apoB100-containing lipoproteins. The LDL-apoB100 band was excised, subjected to an in gel tryptic digest, and desalted using a C-18 SPEC tip prior to submission for mass spectrometry kinetic analysis (Thermo Fisher Scientific, In-gel Tryptic Digestion Kit).
## Serum TG and apoB measurements
Serum metabolic markers including TGs, total cholesterol, LDL-C, HDL-C, and total apoB were measured through a central laboratory (Covance, Indianapolis, IN).
## Mass spectrometry and mass isotopomer distribution analysis for calculation of LDL-apoB kinetics
LDL-apoB100 kinetics were analyzed in plasma samples obtained from subjects after 3 days of 2H2O labeling at baseline and at week 12 of ACCi treatment or fenofibrate + ACCi. LC-MS/MS was performed, as previously described, to obtain fractional replacement rates of apoB100 [26]. Briefly, tryptic peptides from apoB100 were analyzed in data-dependent MS/MS mode for peptide identification and in MS mode for peptide isotopomer analysis on an Agilent 6550 ion funnel quadrupole time-of-flight mass analyzer coupled with HPLC-Chip/MS interface. Acquired MS/MS spectra were extracted and searched against the UniProtKB/Swiss-Prot human protein database (20,265 proteins, UniProt.org, release 2013_05) using Spectrum Mill Proteomics Workbench Rev B.06.00.203 software (Agilent Technologies, https://proteomics.broadinstitute.org/millhome.htm). Peptide filtering criteria included: baseline abundance of 30,000 counts, ± $5\%$ of the predicted isotopomer, and false discovery rate of $1\%$. Peptide sequences provided information of elemental composition. A filtered list of apoB100 peptides was collapsed into a nonredundant peptide formula database containing peptide elemental composition, mass, and retention time. This database was used to extract mass isotopomer abundances (M0-M3) of multiple apoB100 peptides from MS-only acquisition files with the Find-by-Formula algorithm in MassHunter Qualitative Analysis software (Rev B.07.00, Agilent Technologies, https://www.agilent.com/en-us/support/software-informatics/masshunter-workstation-software/masshunter-workstation/masshunter-qualitative-analysis-b-07-00-service-pack-%28sp$2\%$29). Isotopic enrichment was calculated as a metric of 2H-label incorporation from the isotopomer abundances of the peptides quantified by LC-MS, expressed as change in fractional abundance (relative intensity) of the monoisotopic peak compared to natural abundance. This is termed as excess M0 or EM0 [26]. Mass Isotopomer Distribution Analysis was used to establish the isotopomer distribution pattern and enrichments in each peptide of newly synthesized apoB100. This calculation is as described previously [26] and incorporates the measured AUC of body 2H2O enrichments prior to each sample and the number of biosynthetically labile C-H bonds in each specific peptide. These parameters allow calculation of fractional synthesis (f), representing the proportion of LDL-apoB100 molecules present that were newly synthesized from the ratio of the peptide enrichment measured in the sample to the peptide enrichment for a newly synthesized peptide determined by Mass Isotopomer Distribution Analysis [24, 26, 28, 29]. Fractional replacement rates (FRR, %/day) were then calculated as described previously [26, 28, 29].FRR=−ln(1−f)/t Half-lives (days) were calculated ast$\frac{1}{2}$=ln[2]/FRR.
Absolute synthesis rates (ASR, mg/dl/day) were calculated by multiplying the FRR by plasma apoB100 concentration (mg/dl). We used the measured total plasma apoB100 concentration in this calculation because it is a more reliable metric of apoB100 pool size than LDL-apoB content, by avoiding potential variability of recovery through LDL isolation, and because over $90\%$ of plasma apoB is in LDL [18, 19, 20, 21, 30]. The data analysis was handled with Microsoft excel version 16.28 and Graphpad Prism version 9.2.0 (https://www.graphpad.com/scientific-software/prism/).
## Search parameters and acceptance criteria (MS/MS and/or peptide fingerprint data)
The software used for peak list generation was Agilent MassHunter Qualitative Analysis release version B.07.00. Spectrum Mill released version B.06.00.203 was the search engine for proteomic analysis based on MS/MS identifications. The sequence database searched for human protein identifications was Uniprot Release 2013_05 [31]. Twenty thousand two hundred and sixty-five was the number of entries searched in the data base. Trypsin proteolysis was used. Two missed cleavages were permitted. Carbamidomethylation (C) was for fixed modifications. Acetylated lysine (K), oxidized methionine (M), N-terminal pyroglutamic acid (N-termQ), deamidated asparagine (N), and hydroxylated prolines (P) were for variable modifications. Twenty ppm and thirty ppm were the mass tolerance for precursor ions and fragment ions, respectively. The threshold score was $30\%$ based on the minimum match peak intensity for accepting individual spectra. One percent global false discovery rate was determined by algorithms of the Spectrum Mill software (https://proteomics.broadinstitute.org/millhome.htm) and validated at the peptide and protein levels.
## Experimental design and statistical rationale
Data are presented as means ± SEM or SD as indicated in each figure. Statistical significance was calculated by a mixed model ANOVA with Tukey's multiple comparisons test. To establish differences in clinical measurements between healthy, noncirrhotic, and cirrhotic NASH patients in Table 1, statistical significance was evaluated by Kruskal-Wallis rank sum test or Fisher’s exact test. To address differences between noncirrhotic and cirrhotic NASH patients, statistical significance was computed by Wilcoxon rank sum test, Pearson’s Chi-squared test, or Wilcoxon rank sum exact test. Changes in synthesis rates (FRR or ASR) between groups were compared using a paired t test in the same subjects. Unpaired student t test with or without a Welch’s correction was used for specific comparisons as explicitly stated in the figures. Linear regression and Spearman nonparametric correlation analyses were implemented using GraphPad Prism version 9.2.0 for Mac (GraphPad Software, La Jolla, CA).Table 1Clinical and metabolic characteristics of healthy, noncirrhotic and cirrhotic NASH patientsCharacteristicHealthy,$$n = 10$$aNoncirrhotic,$$n = 28$$aCirrhotic,$$n = 22$$aP (3-Group comparison)bP (Noncirrhotic vs. cirrhotic)cAge, years32 [25, 37]59 [47, 64]60 [52, 64]<0.0010.8Male6 ($60\%$)12 ($43\%$)8 ($36\%$)0.50.6Nonhispanic ethnicity0 ($0\%$)11 ($39\%$)12 ($55\%$)0.0080.3Diabetes0 ($0\%$)18 ($64\%$)17 ($77\%$)<0.0010.3BMI, kg/m∧225.6 (23.4, 27.5)34.3 (31.5, 36.9)34.2 (29.9, 36.3)<0.0010.7ALT, U/L14 [11, 20]46 [33, 82]40 [32, 55]<0.0010.3AST, U/L15 [13, 16]42 [28, 71]46 [27, 56]<0.001>0.9GGT, U/L18 [12, 20]36 [27, 70]81 [36, 147]<0.0010.021ALP, U/L60 [48, 63]71 [58, 86]78 [59, 113]0.0270.4Albumin, g/dL4.60 (4.30, 4.68)4.60 (4.40, 4.90)4.50 (4.43, 4.68)0.40.2Platelets, x10∧3/uL224 [207, 250]261 [201, 290]175 [150, 228]0.0080.003Bilirubin, mg/dL0.55 (0.40, 0.60)0.49 (0.32, 0.70)0.56 (0.45, 0.75)0.50.3Bile Acid, umol/LNA6 [5, 7]11 [6, 16]0.0020.002MRI-PDFF, %NA15 [12, 20]10 [5, 13]<0.001<0.001MRE, kPaNA3.21 (2.82, 3.59)5.77 (5.00, 7.00)<0.001<0.001FIB-4NA1.15 (0.99, 1.77)1.92 (1.43, 2.59)0.0170.017FibroSure/FibrotestNA0.23 (0.16, 0.54)0.50 (0.39, 0.66)0.0030.003ELFNA9.56 (8.96, 9.91)10.48 (9.73, 11.41)<0.001<0.001APRINA0.50 (0.30, 0.69)0.69 (0.41, 1.00)0.100.10Hyaluronic acid, ng/mLNA55 [27, 98]112 [70, 232]0.0010.001PIII-NP, ng/mLNA9 [7, 12]13 [10, 17]0.0170.017TIMP-1, ng/mLNA260 [222, 306]313 [257, 389]0.0310.031Glucose, mg/dL86 [83, 90]116 [104, 138]115 [100, 160]<0.001>0.9HOMA-IRNA5 [4, 9]8 [5, 15]0.0840.084HbA1c, %NA6.35 (5.80, 7.03)6.60 (5.75, 7.65)0.40.4Insulin, uIU/mL7 [4, 11]19 [14, 31]30 [20, 38]<0.0010.038Proinsulin, pmol/L4 [3, 6]14 [6, 27]17 [9, 39]<0.0010.15Triglycerides, mg/dL98 [78, 107]159 [133, 242]162 [130, 218]<0.0010.7HDL cholesterol, mg/dLNA42 [38, 50]38 [32, 45]0.130.13NonHDL cholesterol, mg/dLNA138 [126, 156]124 [113, 171]0.40.4VLDL triglycerides, mg/dLNA95 [86, 157]100 [64, 134]0.50.5ApoA1, mg/dLNA143 [127, 162]126 [117, 148]0.0540.054ApoB, mg/dL91 [86, 95]100 [86, 113]86 [77, 119]0.50.4Adiponectin, ng/mLNA3,284 [2,290, 4,307]2,795 [2,120, 3,983]0.50.5Leptin, pg/mLNA25,770 [14,305, 39,357]28,937 [15,939, 38,103]0.90.9Free fatty acid, mEq/L0.25 (0.20, 0.30)0.50 (0.38, 0.60)0.50 (0.33, 0.68)<0.0010.6Beta-hydroxybutyrate, mg/dL0.70 (0.70, 0.78)0.90 (0.90, 0.90)1.10 (0.90, 1.40)<0.0010.027Data are expressed median value (interquartile range) or as a percentage, n (%). To determine whether these groups differ between each other, statistical significance was evaluated by Kruskal-Wallis rank sum test, or Fisher’s exact test. To address differences between non-cirrhotic and cirrhotic NASH patients, statistical significance was calculated by Wilcoxon rank sum test, or Pearson’s Chi-squared test, or Wilcoxon rank sum exact test. ALT, alanine amino transferase; ALP, alkaline phosphatase; APRI, AST to platelet ratio index; AST, aspartate amino transferase; ELF, enhanced liver fibrosis test; FIB-4, fibrosis-4; GGT, gamma-glutamyl transpeptidase; HbA1c, hemoglobin A1c; HOMA-IR, homeostatic model assessment for insulin resistance; MRE, magnetic resonance elastography; NASH, nonalcoholic steatohepatitis; PIII-NP, Type III procollagen peptide; PDFF, proton density fat fraction; TIMP-1, tissue inhibitor of metalloproteinase-1.aMedian (IQR); n (%).bKruskal-Wallis rank sum test; Fisher's exact test.cWilcoxon rank sum test; Pearson's Chi-squared test; Wilcoxon rank sum exact test.
## Study oversight
This study was approved by the institutional review board or independent ethics committees at all participating sites and was conducted in compliance with the Declaration of Helsinki, Good Clinical Practice guidelines, and local regulatory requirements.
## Clinical and biochemical characteristics of healthy, noncirrhotic, and cirrhotic NASH patients
To establish patient population demographics of defined NASH subjects with clinical correlates of fibrosis and cirrhosis, common clinical and metabolic characteristics were evaluated and compared in healthy subjects and noncirrhotic or cirrhotic NASH patients (Table 1). NASH patients displayed common hallmarks of metabolic syndrome such as elevated plasma TGs, free fatty acids, ketone bodies, hyperglycemia, hyperinsulinemia, insulin resistance, and rates of diabetes as compared to healthy controls ($P \leq 0.001$). Markers of liver damage such as alanine amino transferase, aspartate amino transferase, gamma-glutamyl transpeptidase ($P \leq 0.001$), and alkaline phosphatase (0.027) were all elevated in NASH as compared to healthy controls. MRI-proton density fat fraction showed that hepatic liver fat was lower in cirrhotic NASH subjects than in noncirrhotic NASH subjects ($P \leq 0.001$). Noninvasive markers of liver cirrhosis including MRE ($P \leq 0.001$), Fib-4 ($$P \leq 0.17$$), FibroSure/Fibrotest ($$P \leq 0.003$$), Enhanced Liver Fibrosis test ($P \leq 0.001$), hyaluronic acid ($$P \leq 0.001$$), PIII-NP ($$P \leq 0.17$$), and tissue inhibitor of metalloproteinase-1 ($$P \leq 0.031$$) were all elevated in NASH patients with cirrhosis compared to noncirrhotic NASH subjects.
## ACCi treatment increases fasting plasma TGs in patients with NASH
At baseline, plasma TG concentrations were significantly higher in noncirrhotic and cirrhotic NASH patients than in healthy subjects ($P \leq 0.001$), whereas apoB content did not differ ($$P \leq 0.5$$) among groups (Table 1). For the 20 patients with NASH, mean (± SD) plasma TG increased $17\%$, from 180 ± 79 mg/dl at baseline to 211 ± 83 mg/dl at week 12 of firsocostat treatment ($$P \leq 0.0056$$, Fig. 1A). In subgroup analysis, changes in TG were not statistically significant among the 10 noncirrhotic NASH patients (197.4 ± 84.4 at baseline vs. 229.4 ± 78.9 mg/dl at week 12, $$P \leq 0.1276$$; Fig. 1B), while significant increases were observed among the 10 cirrhotic patients (163.3 ± 73.3 at baseline vs. 192.5 ± 86.2 mg/dl at week 12, $$P \leq 0.0014$$; Fig. 1C).Fig. 1Plasma TG concentrations in NASH patients given ACCi. A: Plasma triglyceride (TG, mg/dl ± SD) concentrations in both noncirrhotic and cirrhotic NASH patients were 180 ± 79 at baseline and 211 ± 83 at week 12 of ACCi treatment. At week 12 plasma TG displayed a significant increase of $17\%$ ($$P \leq 0.0056$$) as compared to baseline. B: Plasma triglyceride (TG, mg/dl ± SD) concentrations in non-cirrhotic NASH patients were 197.4 ± 84.4 at baseline and 229.4 ± 78.9 at week 12 of ACCi treatment ($$P \leq 0.1276$$, $$n = 10$$). C: Plasma triglyceride (TG, mg/dl ± SD) concentrations in cirrhotic NASH patients were 163.3 ± 73.3 at baseline and 192.5 ± 86.2 at week 12 of ACCi treatment ($$P \leq 0.0014$$, $$n = 10$$). Data are expressed as mean ± SD. Statistical significance was calculated by paired t test, ∗P ≤ 0.05. ACCi, acetyl-CoA carboxylase inhibitor; NASH, nonalcoholic steatohepatitis.
## LDL-apoB synthesis rates at baseline in NASH and healthy control subjects
We measured apoB FRR in VLDL and LDL particles by 2H2O labeling combined with LC-MS/MS analysis at baseline in the different groups. The earliest time point available for analysis was at day 3 of heavy water administration. ApoB FRRs were monitored in each lipoprotein fraction after preparative ultracentrifugation. On day 3, fractional synthesis measured in VLDL-apoB in NASH subjects had reached or exceeded $100\%$ values, which precluded inference of VLDL-apoB kinetics. LDL-apoB had an average fractional synthesis of $68\%$ ± $19\%$, representing an average FRR of $37\%$/day or a half-life (t½) of just under 2 days. LDL-apoB FRRs were not different between healthy subjects, noncirrhotic, and cirrhotic NASH patients (supplemental Fig. S2A), which excludes a difference in LDL-apoB clearance efficiency (half-life), but LDL-apoB ASRs, calculated from plasma apoB100 concentrations multiplied by the FRR of LDL-apoB in each subject [26, 30], were significantly lower in cirrhotic versus noncirrhotic NASH subjects ($$P \leq 0.03$$, supplemental Fig. S2B).
## ACCi treatment did not affect plasma apoB concentrations in NASH patients
Among patients with NASH ($$n = 20$$), mean (± SD) plasma apoB concentrations did not differ between baseline and week 12 of ACCi therapy (106 ± 8 vs. 106 ± 9 mg/dl, $$P \leq 0.9$$; Fig. 2A). Similar findings were observed in the subgroup analyses of noncirrhotic (115.5 ± 38.4 vs. 117.5 ± 34.7 mg/dl, $$P \leq 0.6461$$; Fig. 2B) and cirrhotic patients (96.3 ± 35.3 and 94.9 ± 41.1 mg/dl, $$P \leq 0.6550$$; Fig. 2C).Fig. 2Plasma-apoB concentrations, LDL-apoB fractional replacement rates (FRR) and apoB absolute synthesis rates (ASR) in noncirrhotic and cirrhotic NASH patients at baseline and after 12 weeks of ACCi. A: Plasma apoB concentrations in combined cirrhotic and noncirrhotic subjects were 106 ± 8 and 106 ± 9 mg/dl (mean ± SD, $$P \leq 0.9$$, $$n = 20$$) at baseline and week 12, respectively. B: *Noncirrhotic plasma* apoB concentrations were 115.5 ± 38.4 and 117.5 ± 34.7 mg/dl (mean ± SD, $$P \leq 0.6461$$, $$n = 10$$) at baseline and week 12, respectively. C: *Cirrhotic plasma* apoB concentrations were 96.3 ± 35.3 and 94.9 ± 41.1 mg/dl (mean ± SD, $$P \leq 0.6550$$, $$n = 10$$) at baseline and week 12, respectively. D: LDL-apoB FRR values ± SD in combined cirrhotic and noncirrhotic subjects were 31 ± 20.2 and 46 ± $22.6\%$/day ($$P \leq 0.03$$, $$n = 16$$) at baseline and week 12, respectively. E: Noncirrhotic LDL-apoB FRR values ± SD were 38.5 ± 22.6 and 40.5 ± $14.6\%$/day at baseline and week 12, respectively, ($$P \leq 0.8197$$, $$n = 8$$). F: Cirrhotic LDL-apoB FRR values ± SD were 23.5 ± 15.4 and 51.38 ± $28.6\%$/day ($$P \leq 0.006$$, $$n = 8$$) at baseline and week 12, respectively. G: Plasma-apoB ASR values in combined cirrhotic and noncirrhotic subjects were 30.4 ± 18.4 and 45.2 ± 15.4 mg/dl/day ($$P \leq 0.016$$, $$n = 16$$) at baseline and week 12, respectively. H: Noncirrhotic plasma-apoB ASR values ± SD were 39.8 ± 20.8 and 46.3 ± 14.8 mg/dl/day at baseline and week 12 ($$P \leq 0.5060$$, $$n = 8$$), respectively. I: Cirrhotic plasma-apoB ASR values were 21.0 ± 9.6 and 44.2 ± 17 mg/dl/day at baseline and week 12 ($$P \leq 0.0021$$, $$n = 8$$), respectively. Plasma-apoB ASR was calculated as LDL-apoB100 FRR (fraction/day) x plasma-apoB concentration (mg/dl). Data are expressed as mean values ± SD. Statistical significance was calculated by a paired t test, ∗P ≤ 0.05. ACCi, acetyl-CoA carboxylase inhibitor; ASR, absolute synthesis rate; NASH, nonalcoholic steatohepatitis.
## ACCi treatment increases ASRs of plasma apoB in NASH patients with cirrhosis
ApoB FRRs and ASRs were measured from plasma apoB100 concentrations and the FRR of LDL-apoB in each subject. For the overall NASH population, mean (± SD) LDL-apoB FRR increased significantly from baseline to week 12 of ACCi therapy (31 ± 20.2 vs. 46 ± $22.6\%$/day, $$P \leq 0.03$$, $$n = 16$$, Fig. 2D), representing mean half-lives of 2.2 and 1.5 days, respectively. Subgroup analysis revealed that significant effects were restricted to NASH patients with cirrhosis. Specifically, mean (± SD) LDL-apoB FRR at baseline and week 12 of ACCi therapy were 38.5 ± $22.6\%$/day and 40.5 ± $14.6\%$/day among noncirrhotic subjects ($$P \leq 0.8197$$, $$n = 8$$, Fig. 2E), representing mean half-lives of 1.8 and 1.7 days as compared with 23.5 ± 15.4 and 51.38 ± $28.6\%$/day, representing mean half-lives of 2.9 and 1.4 days, respectively, in cirrhotic NASH subjects ($$P \leq 0.006$$, $$n = 8$$, Fig. 2F). Similar observations were made with respect to plasma-apoB ASR. The combined group of cirrhotic and noncirrhotic NASH patients still exhibited a significant $47\%$ increase in plasma-apoB ASR from baseline to week 12 of ACCi therapy (30.4 ± 18.4 vs. 45.2 ± 15.4 mg/dl/day, $$P \leq 0.016$$, $$n = 16$$, Fig. 2G). Mean (± SD) plasma-apoB ASRs at baseline and week 12 of ACCi therapy were 39.8 ± 20.8 and 46.3 ± 14.8 mg/dl/day among noncirrhotic subjects ($$P \leq 0.51$$, $$n = 8$$, Fig. 2H), respectively, as compared with 21.0 ± 9.6 and 44.2 ± 17 mg/dl/day, respectively, among cirrhotic subjects ($$P \leq 0.002$$, $$n = 8$$, Fig. 2I).
## Effects of concurrent fenofibrate plus firsocostat therapy on LDL-apoB kinetics in NASH
We evaluated the effects of the PPAR-α agonist, fenofibrate, in combination with firsocostat on apoB kinetics in NASH subjects (supplemental Fig. S3A). Patients were pretreated with fenofibrate 48 mg/day or 145 mg/day for 2 weeks prior to adding firsocostat and LDL-apoB kinetics were sampled by heavy water labeling during the first three days of fenofibrate monotherapy. Mean (± SD) LDL-apoB FRR were 31 ± 20, 38 ± 32, and 38 ± $27\%$/day for the untreated, 3 days of fenofibrate 48 mg/day, and 3 days of fenofibrate 145 mg/day groups, respectively (all nonsignificant, $$P \leq 0.74$$–0.99 between each group by ANOVA, supplemental Fig. S3B). The absence of significant differences between groups suggests that three days of fenofibrate treatment did not influence acute LDL-apoB kinetics. There were no significant differences in LDL-apoB FRRs and ASRs between baseline values (after 3 days of fenofibrate treatment) and values after 12 weeks of combination therapy with fenofibrate plus the ACCi (supplemental Fig. S4A, B). ACCi in combination with both doses of fenofibrate treatment, versus ACCi alone in the mixed noncirrhotic and cirrhotic NASH patients, significantly lowered mean LDL-apoB FRR (± SEM) (33 ± 4 from 46 ± $6\%$/day, $$P \leq 0.032$$, Fig. 3A) and LDL-apoB ASR (± SEM) (34 ± 4 from 45 ± 4 mg/dl/day, $$P \leq 0.026$$, Fig. 3B). The change in LDL-apoB FRR ± SEM from baseline to ACCi treatment alone, and from baseline to ACCi plus fenofibrate in the two fenofibrate groups combined (48 mg/day and 145 mg/day) were 15 ± 6 and −2 ± $5\%$/day, respectively ($$P \leq 0.028$$, Fig. 3C). Additionally, the change in LDL-apoB ASR (mean ± SEM) from baseline to ACCi treatment alone and from baseline to ACCi plus fenofibrate in the two fenofibrate combined groups (48 mg/day and 145 mg/day) were 15 ± 5 and 3 ± 4 mg/dl/day, respectively ($$P \leq 0.04$$, Fig. 3D). Subgroup analysis of the change in LDL-apoB ASR revealed nonsignificant effects between groups (supplemental Fig. S5B), except for the change in LDL-apoB FRRs from baseline to ACCi alone in the cirrhotic group compared to baseline versus ACCi + low dose fibrate in subjects with no cirrhosis ($$P \leq 0.05$$, supplemental Fig. S5A).Fig. 3The effect of either low or high dose of fenofibrate in combination with ACCi on LDL-apoB FRR and ASR in NASH patients. A: LDL-apoB FRR (mean values ± SEM) for ACCi alone in both noncirrhotic and cirrhotic subjects versus, ACCi + two combined fenofibrate (48 mg/day and 145 mg/day) doses were 46 ± 6 and 33 ± $4\%$/day, respectively. ACCi+fenofibrate lowered LDL-apoB FRR than the ACCi treated group alone, $$P \leq 0.032.$$ B: LDL-apoB ASR (mean values ± SEM) for ACCi alone in both noncirrhotic and cirrhotic subjects versus, ACCi + two combined fenofibrate (48 mg/day and 145 mg/day) doses were 45 ± 4 and 34 ± 4 mg/dl/day, respectively. ACCi+fenofibrate lowered LDL-apoB ASR to near baseline levels than the ACCi treated group alone, $$P \leq 0.026.$$ C: Change in LDL-apoB FRR ± SEM from baseline to ACCi treatment in NASH subjects with both noncirrhosis and cirrhosis, and from baseline to ACCi + two combined fibrate doses (48 mg+145 mg) were 15 ± 6 and −2 ± $5\%$/day ($$P \leq 0.028$$), respectively. D: Change in LDL-apoB ASR ± SEM from baseline to ACCi treatment in NASH subjects with both noncirrhosis and cirrhosis, and from baseline to ACCi + two combined fibrate doses (48 mg+145 mg) were 15 ± 5 and 3 ± 4 mg/dl/day ($$P \leq 0.04$$), respectively. Data are expressed as mean ± SEM. Statistical significance was evaluated by one-tailed unpaired t test, ∗P ≤ 0.05. ACCi, acetyl-CoA carboxylase inhibitor; ASR, absolute synthesis rate; FRR, fractional replacement rate; NASH, nonalcoholic steatohepatitis.
## Correlation between changes in plasma TG and plasma apoB content or kinetics
Changes in plasma-apoB100 and TG levels were compared to changes in plasma apoB100 kinetics from baseline to 12 weeks of ACCi treatment. While changes in TG and plasma-apoB100 content between baseline and week 12 were significantly correlated ($r = 0.47$, $$P \leq 0.018$$, Table 2), no significant correlations were observed between changes in TG concentrations and plasma-apoB kinetics, with a trend in the negative direction. Interestingly, the change in LDL-apoB FRR and plasma apoB concentration from baseline to week 12 of ACCi treatment displayed a borderline significant association ($$P \leq 0.053$$) and negative correlation (r = −0.42).Table 2Correlations between the change in plasma triglycerides, plasma-apoB, LDL-apoB FRR, and LDL-apoB ASR at week 12 as compared to baseline after ACCi treatmentCorrelationsSpearman rPTriglyceride concentration versus plasma-apoB concentration0.47a0.018Triglyceride concentration versus LDL-apoB FRR−0.180.25Triglyceride concentration versus LDL-apoB ASR−0.140.31Plasma-apoB concentration versus LDL-apoB FRR−0.420.053Data are expressed as a spearman r correlation value with corresponding P value. Change in mean plasma apoB concentrations, triglyceride concentrations, and LDL-apoB FRR or LDL-apoB ASR kinetic values from baseline to week 12 were calculated by the following formula, week 12 values - baseline value.aP ≤ 0.05 based on spearman correlation analysis. ACCi, acetyl-CoA carboxylase inhibitor; ASR, absolute synthesis rate; FRR, fractional replacement rate.
## DISCUSSION
Treatment with ACC inhibitors leads to hypertriglyceridemia in a subset of patients with NASH [8, 10, 11, 12, 13, 14, 15, 16]. Our goals were to determine whether treatment with a liver specific ACCi, firsocostat [32], which has been reported to increase plasma TGs and variably to increase apoB concentrations, are associated with altered LDL-apoB particle production rate or half-life (clearance), whether the stage of liver disease alters the LDL-apoB kinetic response to ACCi therapy and whether concurrent treatment with a fibrate can prevent changes in LDL-apoB kinetics.
Endogenously derived TG are trafficked in the blood primarily in VLDL. During the process of metabolic conversion of VLDL to LDL, the main structural protein of these particles, apoB100, remains intact, whereas receptor-mediated uptake removes the intact particle, which includes nonexchangeable apoB100 that persists throughout the lifetime of a LDL particle [19]. The export of VLDL from the liver into circulation is reliant on hepatic apoB synthesis [19]. The majority (∼$90\%$) of circulating apoB100 resides in LDL. The half-life of VLDL is hours, whereas LDL, for which VLDL is the intravascular precursor, has a half-life of days [20, 21, 22, 23, 24, 25]. In this study, blood was not sampled during the first day of heavy water labeling to measure VLDL-apoB kinetics but the sample collected at 3 days of labeling enabled analysis of LDL-apoB kinetics.
We did not measure VLDL-TG or VLDL-apoB production rates, but the finding of LDL-apoB overproduction is important in its own right for at least two reasons. First, LDL-apoB production rate is of interest in terms of metabolic site of drug action and potential atherogenic risk [5, 6, 7, 33]. Second, in other clinical settings such as type 2 diabetes, hypertriglyceridemia is driven primarily by hepatic overproduction of apoB-containing molecules [34, 35, 36]. It is therefore reasonable to infer overproduction of apoB-containing particles from overproduction of LDL-apoB particles by the liver [24, 30], at least as a hypothesis-generating observation. The half-life of ∼2 days measured here for apoB in the LDL fraction is consistent with prior reports [20, 21, 22, 23, 24, 25].
In NASH patients in this study, plasma concentrations of TG but not apoB increased after 12 weeks of treatment with the ACCi firsocostat (Figs. 1A and 2A). Our primary finding here is the significant increase in ASR of LDL-apoB at week 12 of firsocostat treatment, restricted to the subgroup of NASH patients with cirrhosis (Fig. 2F, I). The finding that the replacement rate constant (FRR) was more rapid, not slower, argues against an LDL-apoB clearance defect induced by ACCi.
Some technical points about lipoprotein kinetics are worth noting. ASR is the product of FRR and pool size [18, 19, 20, 21, 30] and represents the biochemical production rate, expressed in units of mass per time. An increase in pool size with no slowing of half-life means that the change in pool size is due to higher production rates, not slower removal rates. Indeed, this is the main information gained from a metabolic labeling study of this type and the result here was unambiguous. Hypertriglyceridemia might have involved no change in apoB100 turnover, which would have suggested altered plasma lipid turnover without a change in particle metabolism (e.g., a lipoprotein lipase or ApoCIII effect) [11]. The correlation between relative changes in plasma-apoB and TG concentrations at week 12 of firsocostat treatment suggests a relationship between increased particle production and plasma TG concentrations (Table 2). Since our study did not directly assess VLDL conversion into LDL, however, we cannot directly confirm this hypothesis from our study.
We did not explore potential molecular signals, but these observations should be considered in context of previous reports describing higher VLDL secretion in a genetic model of ACC ablation (ACC double knock-out mice) and hypertriglyceridemia in humans treated with different ACCi compounds [8, 10, 11, 12, 13, 14, 15, 16]. Reduced PUFAs in the liver have been reported after ACCi treatment [8]. PUFAs act as key regulators of SREBP-1C activity by inhibiting its proteolytic processing [37]. PUFA reduction, specifically omega 3- and 6-containing PUFAs such as arachidonic acid and docosahexaenoic acid were suggested to activate SREBP-1c and reduce PPARα activity [8] and PUFA supplementation in the double knockout ACC mice normalized TG levels [8].
Kim et al. also showed upregulation of downstream genes to SREBP-1C, such as glycerol phosphate acyl transferase-1, associated with increased VLDL secretion in liver specific ACC knockout mice [8]. In fasted overnight rodents treated with an ACCi after Poloxamer 405 administration to inhibit lipolytic clearance of TG rich lipoproteins, Goedeke, and colleagues reported an increase in VLDL secretion rates [11]. Insulin resistance in NASH patients [1, 2, 3, 4] is associated with elevated NEFAs flux, as shown by elevated levels of NEFAs here (Table 1), providing an alternate pool of fatty acids for TG synthesis [8]. Our data in humans are consistent with an effect of ACCi on hepatic TG and apoB particle production as the site of action.
Moreover, Loomba et. al. conducted nuclear magnetic resonance lipoprotein analysis in a similar cohort of NASH patients treated with firsocostat in a 12-weeks phase 2a study [14]. While increased particle number and TG concentration of VLDL were observed over 1 week of ACCi treatment, the number of small LDL particles, total cholesterol, HDL-C, and LDL-C concentrations and particle number, along with glycemic parameters, did not change during the study in comparison to placebo [14]. In a multivariate analysis adjusting for demographics and lipids at baseline, grade 3 or 4 hypertriglyceridemia (>500 mg/dl) during ACCi treatment was associated with patients whose baseline plasma TG levels were over 250 mg/dl. Importantly, despite ongoing treatment with firsocostat, treatment with fibrates or fish oil led to resolution of grade 3 or 4 hypertriglyceridemia [14]. The utility of fibrates to mitigate ACCi-induced hypertriglyceridemia has been evaluated in several clinical studies in addition to studies in rodents [11, 13, 15]. In a proof-of-concept study of NASH patients with hypertriglyceridemia (>150 mg/dl) and advanced (F3-F4) fibrosis, Lawitz et al. showed that a 2-weeks course of preemptive therapy with fenofibrate (48 or 145 mg) prevented any increase in TG after 24 weeks of fenofibrate and firsocostat combination therapy [13]. They also confirmed that fenofibrate (145 mg) prevents TG elevations in the setting of combination therapy with firsocostat and the farnesoid X receptor agonist, cilofexor, in hypertriglyceridemic patients with NASH [13]. These data suggest that concurrent ACCi and fenofibrate treatment may prevent the increase in plasma TG with ACCi treatment alone.
To explore these clinical observations, we evaluated whether combination therapy with fenofibrate and firsocostat also altered LDL-apoB production rates. When the data for both doses of fenofibrate with ACCi treatment and the stages of NASH were combined, apoB FRR and ASR were significantly lower for combined treatment than for ACCi treatment alone in the cross-sectional analysis (Fig. 3A, B), although the change in LDL-apoB kinetics in the low or high fibrate doses + ACCi treatments were not significantly different from ACCi treatment alone in either subgroup classified by cirrhosis state alone (supplemental Fig. S5A, B). Our longitudinal analyses were potentially confounded by the initial 3 days of fibrate therapy at baseline but longitudinal comparisons supported the cross-sectional analyses. There were no apparent effects of the initial 3 days of fibrate treatment on apoB FRR (supplemental Fig. S3B) and the addition of ACCi treatment did not increase apoB FRR (supplemental Fig. S4A, B). Moreover, we observed a significant reduction in the change in LDL-apoB FRR and ASR with the ACCi + two combined fibrate doses (Fig. 3C, D). Previous lipoprotein kinetic studies have been reported in men with metabolic syndrome treated with fibrates alone. In one study, treatment with fenofibrate 200 mg/day for 5 weeks led to increased fractional catabolic rate and decreased pool size of apoB-containing particles [38]. Caslake et. al. reported decreased VLDL particle size after fenofibrate treatment, as well as increased LDL-apoB degradation by the receptor route but not for receptor independent routes [39]. Lastly, fenofibrate treatment lowered VLDL-apoB concentrations and secretion rates in NAFLD patients, consistent with our data showing suppressed LDL-apoB ASR with concurrent fenofibrate and ACCi treatment (Fig. 3B, [40]).
We observed that NASH patients with cirrhosis have lower apoB ASR than noncirrhotic and healthy counterparts (supplemental Fig. S2B). Both noncirrhotic NASH and cirrhotic NASH patients were insulin resistant as indicated by homeostatic model assessment for insulin resistance but the cirrhotic NASH displayed higher levels of circulating insulin (Table 1). Differences in insulin signaling could potentially explain alterations in basal apoB secretion rates [34, 35, 36, 41]. The cirrhotic NASH patents also showed lower hepatic fat than noncirrhotic NASH patients (Table 1 and supplemental Fig. S2B), which might lower apoB secretion [5, 6, 7, 41]. Further studies are warranted to understand these differences in LDL-apoB metabolism.
Importantly, we observed different effects of the ACCi on lipid and lipoprotein metabolism between noncirrhotic and cirrhotic NASH patients [12, 14]. After ACCi treatment, cirrhotic NASH patients displayed elevated plasma TG levels and LDL-apoB FRR and ASR as compared to noncirrhotic subjects (Figs. 1C and 2F, I). The explanation for these differential effects remains uncertain. These data could indicate some restoration of hepatocellular function with ACCi treatment in cirrhosis [42]. Data to support this hypothesis include improvements in liver function, including decreased liver fat content, liver stiffness, neuroinflammatory activity, and fibrosis [11, 12, 13, 14, 15, 42]. In rodents treated with ACCi’s, decreased expression of markers of macrophage activation and fibrosis such as C-C motif chemokine 2 and collagen alpha-1(I) chain were observed [42]. Lower staining of α-smooth muscle actin, and cluster of differentiation 3, markers of hepatic stellate cell activation, as well as fibrogenesis, and T cell activation have been reported [16, 42].
In conclusion, we show here that treatment with the ACCi firsocostat significantly increases the synthesis rate of apoB-containing LDL particles in NASH subjects with cirrhosis, without a significant increase in plasma apoB concentrations (Graphical abstract). These results suggest that the site of action of previously reported effects of ACCi treatment on plasma TG concentrations is the liver. Fenofibrate combination therapy prevented the increased LDL-apoB particle production induced by firsocostat therapy.
## Data availability
All data can be viewed in the manuscript. Any or additional data is available upon reviewer’s request from the corresponding author.
## Supplemental data
This article contains supplemental data.
Data filter 1 Data filter 2 Data filter 3 Data filter 4 Data filter 5 Data filter 6 Data filter 7 Data filter 8 Data filter 9 Data filter 10 Data filter 11 Summary table 1 Summary table 2 Summary table 3 Summary table 4 Summary table 5 Summary table 6 Summary table 7 Summary table 8 Summary table 9 Summary table 10 Summary table 11 Supplemental materials
## Conflict of interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests.
## Author contributions
J.- C. C., K. Z., A. N. B., R. S. H., C. C., R. P. M., and M. H. conceptualization; J.- C. C., K. Z., A. N. B., R. S. H., C. C., R. P. M., and M. H. methodology; J.- C. C., K. Z., A. N. B., R. S. H., C. C., R. P. M., and M. H. software; M. D. and K. Z. data curation; M. D. and K. Z. writing and original draft preparation; M. D. and K. Z. visualization; M. D. and K. Z. investigation; J.-C. C. and M. H. supervision; J.-C. C. and M. H. validation; J.-C. C., R. P. M., and M. H. writing-review and editing.
## Funding and additional information
M. H. received grant support from $\frac{10.13039}{100005564}$Gilead sciences.
## References
1. McCullough A.J.. **Update on nonalcoholic fatty liver disease**. *J. Clin. Gastroenterol.* (2002) **34** 255-262. PMID: 11873108
2. Lindenmeyer C.C., McCullough A.J.. **The natural history of Nonalcoholic Fatty Liver Disease-an evolving view**. *Clin. Liver Dis.* (2018) **22** 11-21. PMID: 29128051
3. Marengo A., Jouness R.I., Bugianesi E.. **Progression and natural history of Nonalcoholic Fatty Liver Disease in adults**. *Clin. Liver Dis.* (2016) **20** 313-324. PMID: 27063271
4. Farrell G.C., Larter C.Z.. **Nonalcoholic fatty liver disease: from steatosis to cirrhosis**. *Hepatology* (2006) **43** S99-S112. PMID: 16447287
5. Chatrath H., Vuppalanchi R., Chalasani N.. **Dyslipidemia in patients with nonalcoholic fatty liver disease**. *Semin. Liver Dis.* (2012) **32** 22-29. PMID: 22418885
6. Corey K.E., Misdraji J., Gelrud L., Zheng H., Chung R.T., Krauss R.M.. **Nonalcoholic steatohepatitis is associated with an atherogenic lipoprotein subfraction profile**. *Lipids Health Dis.* (2014) **13** 100. PMID: 24952382
7. Jiang Z.G., Robson S.C., Yao Z.. **Lipoprotein metabolism in nonalcoholic fatty liver disease**. *J. Biomed. Res.* (2013) **27** 1-13. PMID: 23554788
8. Kim C.W., Addy C., Kusunoki J., Anderson N.N., Deja S., Fu X.. **Acetyl CoA carboxylase inhibition reduces hepatic steatosis but elevates plasma triglycerides in mice and humans: a bedside to bench investigation**. *Cell Metab.* (2017) **26** 394-406.e6. PMID: 28768177
9. Tong L., Harwood H.J.. **Acetyl-coenzyme A carboxylases: versatile targets for drug discovery**. *J. Cell Biochem.* (2006) **99** 1476-1488. PMID: 16983687
10. Alkhouri N., Lawitz E., Noureddin M., DeFronzo R., Shulman G.I.. **GS-0976 (Firsocostat): an investigational liver-directed acetyl-CoA carboxylase (ACC) inhibitor for the treatment of non-alcoholic steatohepatitis (NASH)**. *Expert Opin. Investig. Drugs* (2020) **29** 135-141
11. Goedeke L., Bates J., Vatner D.F., Perry R.J., Wang T., Ramirez R.. **Acetyl-CoA carboxylase inhibition reverses NAFLD and hepatic insulin resistance but promotes hypertriglyceridemia in rodents**. *Hepatology* (2018) **68** 2197-2211. PMID: 29790582
12. Lawitz E.J., Coste A., Poordad F., Alkhouri N., Loo N., McColgan B.J.. **Acetyl-CoA carboxylase inhibitor GS-0976 for 12 weeks reduces hepatic de novo lipogenesis and steatosis in patients with nonalcoholic steatohepatitis**. *Clin. Gastroenterol. Hepatol.* (2018) **16** 1983-1991.e3. PMID: 29705265
13. Lawitz E.J., Bhandari B.R., Ruane P.J., Kohli A., Harting E., Ding D.. **Fenofibrate mitigates hypertriglyceridemia in nonalcoholic steatohepatitis patients treated with Cilofexor/Firsocostat**. *Clin. Gastroenterol. Hepatol.* (2022). DOI: 10.1016/j.cgh.2021.12.044
14. Loomba R., Kayali Z., Noureddin M., Ruane P., Lawitz E.J., Bennett M.. **GS-0976 reduces hepatic steatosis and fibrosis markers in patients with Nonalcoholic Fatty Liver Disease**. *Gastroenterology* (2018) **155** 1463-1473.e6. PMID: 30059671
15. Loomba R., Noureddin M., Kowdley K.V., Kohli A., Sheikh A., Neff G.. **Combination therapies including cilofexor and firsocostat for bridging fibrosis and cirrhosis attributable to NASH**. *Hepatology (Baltimore, Md.)* (2021) **73** 625-643
16. Calle R.A., Amin N.B., Carvajal-Gonzalez S., Ross T.T., Bergman A., Aggarwal S.. **ACC inhibitor alone or co-administered with a DGAT2 inhibitor in patients with non-alcoholic fatty liver disease: two parallel, placebo-controlled, randomized phase 2a trials**. *Nat. Med.* (2021) **27** 1836-1848. PMID: 34635855
17. McGarry J.D., Mannaerts G.P., Foster D.W.. **A possible role for malonyl-CoA in the regulation of hepatic fatty acid oxidation and ketogenesis**. *J. Clin. Invest.* (1977) **60** 265-270. PMID: 874089
18. Vance D.E., Vance J.E.. *Biochemistry of Lipids, Lipoproteins, and Membranes* (2008)
19. Elovson J., Chatterton J.E., Bell G.T., Schumaker V.N., Reuben M.A., Puppione D.L.. **Plasma very low-density lipoproteins contain a single molecule of apolipoprotein B**. *J. Lipid Res.* (1988) **29** 1461-1473. PMID: 3241122
20. Kesäniemi Y.A., Beltz W.F., Grundy S.M.. **Comparisons of metabolism of apolipoprotein B in normal subjects, obese patients, and patients with coronary heart disease**. *J. Clin. Invest.* (1985) **76** 586-595. PMID: 3861622
21. Cohn J.S., Wagner D.A., Cohn S.D., Millar J.S., Schaefer E.J.. **Measurement of very low density and low-density lipoprotein apolipoprotein (Apo) B-100 and high-density lipoprotein Apo A-I production in human subjects using deuterated leucine. Effect of fasting and feeding**. *J. Clin. Invest.* (1990) **85** 804-811. PMID: 2107210
22. Bilz S., Wagner S., Schmitz M., Bedynek A., Keller U., Demant T.. **Effects of atorvastatin versus fenofibrate on apoB-100 and apoA-I kinetics in mixed hyperlipidemia**. *J. Lipid Res.* (2004) **45** 174-185. PMID: 14523053
23. Matthan N.R., Jalbert S.M., Lamon-Fava S., Dolnikowski G.G., Welty F.K., Barrett H.R.. **TRL, IDL, and LDL apolipoprotein B-100 and HDL apolipoprotein A-I kinetics as a function of age and menopausal status**. *Arterioscler. Thromb. Vasc. Biol.* (2005) **25** 1691-1696. PMID: 15933247
24. Busch R., Kim Y.K., Neese R.A., Schade-Serin V., Collins M., Awada M.. **Measurement of protein turnover rates by heavy water labeling of nonessential amino acids**. *Biochim. Biophys. Acta* (2006) **1760** 730-744. PMID: 16567052
25. Beysen C., Angel T.E., Hellerstein M.K., Turner S.M., Krentz A., Weyer C., Hompesch M.. *Translational Research Methods in Diabetes, Obesity, and Nonalcoholic Fatty Liver Disease* (2019)
26. Holmes W.E., Angel T.E., Li K.W., Hellerstein M.K.. **Dynamic proteomics: in vivo proteome-wide measurement of protein kinetics using metabolic labeling**. *Met. Enzymol.* (2015) **561** 219-276
27. Lindgren F.T., Nichols A.V., Freeman N.K., Wills R.D., Wing L., Gullberg J.E.. **Analysis of low-density lipoproteins by preparative ultracentrifugation and refractometry**. *J. Lipid Res.* (1964) **5** 68-74. PMID: 14173332
28. Hellerstein M.K., Neese R.A.. **Mass isotopomer distribution analysis at eight years: theoretical, analytic, and experimental considerations**. *Am. J. Physiol.* (1999) **276** E1146-E1170. PMID: 10362629
29. Price J.C., Holmes W.E., Li K.W., Floreani N.A., Neese R.A., Turner S.M.. **Measurement of human plasma proteome dynamics with (2)H(2)O and liquid chromatography tandem mass spectrometry**. *Anal. Biochem.* (2012) **420** 73-83. PMID: 21964502
30. Welty F.K., Lichtenstein A.H., Barrett P.H., Dolnikowski G.G., Schaefer E.J.. **Human apolipoprotein (Apo) B-48 and ApoB-100 kinetics with stable isotopes**. *Arterioscler. Thromb. Vasc. Biol.* (1999) **19** 2966-2974. PMID: 10591677
31. **UniProt: a worldwide hub of protein knowledge**. *Nucl. Acids Res.* (2019) **47** D506-D515. PMID: 30395287
32. Harriman G., Greenwood J., Bhat S., Huang X., Wang R., Paul D.. **Acetyl-CoA carboxylase inhibition by ND-630 reduces hepatic steatosis, improves insulin sensitivity, and modulates dyslipidemia in rats**. *Proc. Natl. Acad. Sci. U. S. A.* (2016) **113** E1796-E1805. PMID: 26976583
33. Imai N., Cohen D.E.. **Trimming the fat: Acetyl-CoA carboxylase inhibition for the management of NAFLD**. *Hepatology* (2018) **68** 2062-2065. PMID: 30076622
34. Haas M.E., Attie A.D., Biddinger S.B.. **The regulation of ApoB metabolism by insulin**. *Trends Endocrinol. Metab.* (2013) **24** 391-397. PMID: 23721961
35. Goldberg I.J.. **Clinical review 124: diabetic dyslipidemia: causes and consequences**. *J. Clin. Endocrinol. Metab.* (2001) **86** 965-971. PMID: 11238470
36. Sparks J.D., Sparks C.E., Adeli K.. **Selective hepatic insulin resistance, VLDL overproduction, and hypertriglyceridemia**. *Arterioscler. Thromb. Vasc. Biol.* (2012) **32** 2104-2112. PMID: 22796579
37. Shimano H., Sato R.. **SREBP-regulated lipid metabolism: convergent physiology — divergent pathophysiology**. *Nat. Rev. Endocrinol.* (2017) **13** 710-730. PMID: 28849786
38. Watts G.F., Barrett P.H., Ji J., Serone A.P., Chan D.C., Croft K.D.. **Differential regulation of lipoprotein kinetics by atorvastatin and fenofibrate in subjects with the metabolic syndrome**. *Diabetes* (2003) **52** 803-811. PMID: 12606523
39. Caslake M.J., Packard C.J., Gaw A., Murray E., Griffin B.A., Vallance B.D.. **Fenofibrate and LDL metabolic heterogeneity in hypercholesterolemia**. *Arterioscler. Thromb.* (1993) **13** 702-711. PMID: 8485122
40. Fabbrini E., Mohammed B.S., Korenblat K.M., Magkos F., McCrea J., Patterson B.W.. **Effect of fenofibrate and niacin on intrahepatic triglyceride content, very low-density lipoprotein kinetics, and insulin action in obese subjects with nonalcoholic fatty liver disease**. *J. Clin. Endocrinol. Metab.* (2010) **95** 2727-2735. PMID: 20371660
41. Charlton M., Sreekumar R., Rasmussen D., Lindor K., Nair K.S.. **Apolipoprotein synthesis in nonalcoholic steatohepatitis**. *Hepatol. (Baltimore, Md* (2002) **35** 898-904
42. Ross T.T., Crowley C., Kelly K.L., Rinaldi A., Beebe D.A., Lech M.P.. **Acetyl-CoA carboxylase inhibition improves multiple dimensions of NASH pathogenesis in model systems**. *Cell Mol. Gastroenterol. Hepatol.* (2020) **10** 829-851. PMID: 32526482
|
---
title: The patient perspective on remote monitoring of implantable cardiac devices
authors:
- Henrike A. K. Hillmann
- Claudius Hansen
- Oliver Przibille
- David Duncker
journal: Frontiers in Cardiovascular Medicine
year: 2023
pmcid: PMC10017432
doi: 10.3389/fcvm.2023.1123848
license: CC BY 4.0
---
# The patient perspective on remote monitoring of implantable cardiac devices
## Abstract
### Aims
Remote monitoring for patients with cardiac implantable electronic devices (CIEDs) is well established in clinical routine and recommended by current guidelines. Nevertheless, data regarding patients’ perceptions are limited. Therefore, this study aims to analyze the patient perspectives on the remote monitoring of cardiac devices in Germany.
### Methods and results
Patients with CIEDs and remote monitoring of all current manufacturers from three German centers were asked to participate. The questionnaire consisted of 37 questions regarding the patients’ individual use and perspectives on remote monitoring. Survey participation was anonymous and on a voluntary basis. A total of 617 patients ($71.6\%$ men) participated. Most patients reported feeling well informed ($69.3\%$) and reported having unchanged or improved coping ($98.8\%$) since the start of remote monitoring. At least $39.7\%$ of patients experienced technical problems regarding the transmitter, whereas most patients ($60.3\%$) reported that they never noted technical issues. Older patients had significantly less interest than younger patients in using their own smartphones for data transfer ($p \leq 0.001$).
### Conclusion
Patients with remote follow-up of CIED reported that they felt well informed about the remote monitoring approach. Remote monitoring can support coping with their disease. With remote monitoring, patients experienced a prolongation of intervals of in-person follow-up visits, and especially younger patients would appreciate smartphone-based data transfer of their CIEDs.
## Introduction
Remote monitoring in patients with implantable cardiac devices is associated with reduced hospitalization rates (1–3), increased survival rates [2, 3], and a reduction of necessary healthcare resources [1, 4]. It is non-inferior to in-person follow-up visits regarding patients with single or dual chamber pacemaker (PM) devices and activated automatic threshold algorithms [5], as well as patients with an implantable cardioverter-defibrillator (ICD) (4, 6–8). Thus, remote monitoring is recommended for patients with a cardiac implantable electronic device (CIED) having problems attending in-person follow-up visits or struggling with chronic device-related problems to expand the time between two in-person follow-up visits [9, 10]. Patients with an ICD managed via remote monitoring experience fewer inappropriate ICD shocks [8, 11, 12], which is an important aspect as ICD shocks are known to increase patient mortality [13]. Therefore, current guidelines recommend remote monitoring for ICDs to minimize the occurrence of inappropriate shocks [14]. Moreover, remote monitoring can reduce the time between an event and reaction [11, 15, 16], with a concomitant decrease in adverse outcomes [3, 17] due to clinical problems or technical issues. Furthermore, remote monitoring is cost-effective [18, 19], reduces workload [20], and is time-effective [21]. Most patients have an unchanged or improved quality of life after device implantation [22] and are satisfied while using remote monitoring [23]. Nevertheless, detailed information regarding patients’ perceptions and individual concerns toward remote monitoring as well as possible future perspectives is limited. This study aims to analyze the patients’ perception of remote monitoring and possible future perspectives on remote monitoring of CIEDs.
## Methods
Patients with remote monitoring of a CIED of all current manufacturers in three German centers (Hannover Heart Rhythm Center, Hannover; Heart & Vascular Center, Göttingen; CCB, Frankfurt) were invited to fill in a patient questionnaire sent by mail. Survey participation was anonymous and on a voluntary basis. Questionnaires that had been returned between July and November 2021 were included in the analysis. The present study was conducted in compliance with the Declaration of Helsinki.
## Questionnaire
The questionnaire consisted of 37 questions regarding patients’ baseline characteristics, individual use, and opinion on remote monitoring, as well as the patients’ opinion regarding future remote monitoring perspectives (refer to Appendix Table S1). Remote monitoring includes three concepts [9]: [1] remote follow-up with scheduled interrogations of the CIED, [2] remote monitoring with unscheduled data transfers due to upcoming events or alerts, and [3] patient-initiated follow-ups with unscheduled data transfers due to symptomatic arrhythmias or other issues. Hereafter, the term “remote monitoring” is going to include all those three aspects. Due to the type of question, there were three response options: single-choice, multiple-choice, or free text. The questionnaire could be filled in paper-based or web-based, according to patients’ preferences. No questions were set as mandatory. Patients with a cardiac resynchronization therapy defibrillator (CRT-D) were included in the category of patients with ICD, whereas patients with a cardiac resynchronization therapy pacemaker (CRT-P) were included in the group of patients with PM.
## Statistical analysis
Categorial data are presented as numbers and percentages. Percentages given were calculated due to the total amount of given answers per question (refer to Appendix Table S1). Continuous data are presented as median (P25;P75). Wilcoxon test, Mann–Whitney U-test, or Kruskal–Wallis test was used for between-group comparisons, as appropriate. Statistical data analysis was performed using SPSS (version 27, IBM, Armonk, United States). P-values of < 0.05 were considered statistically significant.
## Results
A total of 617 patients (437 men, $71.6\%$) from three electrophysiological centers in Germany participated between July and November 2021. Two hundred and twenty-five questionnaires were received paper-based, and 392 questionnaires were received web-based.
## Baseline characteristics
Baseline characteristics are presented in Table 1.
**Table 1**
| Parameter | n = 617 (%) |
| --- | --- |
| Male, n (%) | 437 (71.6) |
| Age range, n (%) | |
| 10–29 years | 7 (1.1) |
| 30–49 years | 50 (8.1) |
| 50–69 years | 262 (42.8) |
| 70–89 years | 279 (45.6) |
| 90–99 years | 14 (2.3) |
| Device, n (%) | |
| ICD incl. CRT-D | 424 (70.3) |
| PM incl. CRT-P | 72 (11.9) |
| Implantable loop recorder | 107 (17.7) |
| Residence, n (%) | |
| City | 337 (55.8) |
| Rural area | 252 (41.7) |
| Other | 15 (2.5) |
## Remote monitoring and transmitter handling
Most patients said that they felt well informed about this tool (fully agree: $27.2\%$, $$n = 101$$; agree: $42.1\%$, $$n = 156$$; neutral: $26.4\%$, $$n = 98$$; do not agree: $4.6\%$, $$n = 17$$; do not agree absolutely: $1.1\%$, $$n = 4$$). The majority of patients answered not to have any concerns in regard to their telecardiological monitoring ($$n = 436$$, $92.0\%$), while 38 patients indicated having concerns regarding this aspect ($8.0\%$). Patients’ view on remote monitoring is summarized in Figure 1. To receive information about remote monitoring and the transmitter (multiple answers possible), all patients would prefer a personal conversation. One hundred and sixty-two patients ($43.9\%$) would rather have a brochure, whereas 136 patients ($36.9\%$) would prefer an instruction manual. Mobile applications, instruction videos, websites, or a set of frequently asked questions were chosen by 87 ($23.6\%$), 85 ($23.0\%$), 79 ($21.4\%$), and 75 ($20.3\%$) patients, respectively. Since using remote monitoring, most patients answered to experience an improved or unchanged coping with their disease (Figure 2). Asked for individual reasons, patients who answered to have improved coping since using remote monitoring stated to feel better monitored, reassured, and/or secure. Patients with a worse coping, e.g., stated to experience a higher focus on the disease than before and that everything feels too complicated. Patients reported that remote monitoring resulted in a prolongation of intervals of in-person follow-up visits for device interrogation as well as in-person follow-up visits at the attending cardiologist (Table 2). Technical problems regarding transmitter handling were reported by 196 patients ($39.7\%$), whereas 298 patients ($60.3\%$) never noted any technical problems. In case of questions regarding the transmitter, most patients refer to their cardiologist ($$n = 286$$; $62.3\%$) or another medical specialist ($$n = 75$$; $16.3\%$). Other chosen contacts were the manufacturer ($$n = 42$$; $9.2\%$), relatives ($$n = 15$$; $3.3\%$), emergency service ($$n = 5$$; $1.1\%$), an association ($$n = 1$$; $0.2\%$), or nobody ($$n = 35$$; $7.6\%$). Four hundred and twenty-eight ($84.9\%$) of the participants indicate to handle their transmitter on their own without any assistance from a third person, whereas 76 participants ($15.1\%$) answered to get assistance. If assistance was given (multiple answers were possible), most patients chose “a member of the family” as a helping person ($$n = 66$$; $86.8\%$). Most participants never take the transmitter with them if they go on a trip for more than 24 h ($$n = 285$$, $65.1\%$). Other chosen answers were less often (“Yes, always,” $$n = 71$$, $16.2\%$; “Yes, in cases of >48 h,” $$n = 19$$, $4.3\%$; “Yes, in case of >1 week,” $$n = 45$$, $10.3\%$; “Other,” $$n = 18$$, $4.1\%$). The majority answered not to consider the transmitter a limitation ($$n = 445$$, $95.3\%$), whereas 22 patients answered to see the transmitter as a limitation ($4.7\%$). Individual reasons given were cumbersome equipment, permanent illumination in the bedroom, having to take their transmitter with them when leaving home, or the necessity of an internet connection.
**Figure 1:** *Patients’ perception of remote monitoring. (A) Importance of given advantages regarding remote monitoring. As patients had the possibility to choose multiple answers, they had to sort the given answers according to their importance. The importance is represented in colors. (B) Patients’ ideas on the advantages and disadvantages of remote monitoring. ICD = implantable cardioverter-defibrillator; CIED = cardiac implantable device.* **Figure 2:** *Patients’ answers on the change regarding coping since the use of remote monitoring.* TABLE_PLACEHOLDER:Table 2 *In this* survey, previous data regarding the reduction of in-person follow-up visits due to remote monitoring [5, 20] could be confirmed due to a significant reduction ($p \leq 0.001$) of in-person follow-up visits not only regarding interrogations of the CIEDs but also regarding the number of in-person visits of the attending cardiologist. In accordance, patients in this survey chose less time spent on consultations as an important aspect regarding remote monitoring. Nevertheless, reassurance and continuous monitoring with the possibility of a fast objectification of symptoms and the possibility of being quickly informed in case of problems were named as the most important advantages due to remote monitoring. As patient involvement in coping becomes more important [24] and patient-reported outcome instruments to quantify the quality of life are evolving [25], this is an important point to consider. Accordingly, most patients answered to experience unchanged ($54.8\%$) or improved ($44.0\%$) coping since the start of remote monitoring.
Due to the COVID-19 pandemic with concomitant restrictions regarding the possibility of in-person visits and face-to-face follow-ups, remote monitoring of CIEDs became more important, and a significant increase in the number of patients using remote monitoring for their CIEDs has been observed [26]. The REMOTE-CIED randomized trial reported no differences in “patient-reported health status and ICD acceptance” between patients with remote monitoring follow-ups and in-clinic follow-ups [7]. Recent studies have evaluated the quality of life in patients with ICDs and results have not shown a significant decrease in quality of life, in general [22, 27, 28]. Nevertheless, it has been shown that inappropriate shocks in patients with ICDs reduce the quality of life [22]. Remote monitoring, therefore, has the potential to additionally increase the quality of life and reassurance in patients with a CIED and especially ICDs.
Most patients are satisfied using remote monitoring and prefer remote monitoring over in-person follow-ups (29–31). Concerns, disadvantages, and reasons for a worse coping regarding remote monitoring recorded during this survey such as data concerns, the fear of unauthorized access regarding the CIED through a third party, as well as the insecurity about the transmitter functionality may help physicians and manufacturers to further improve and evolve remote monitoring. On the one hand, focusing on patient education regarding remote monitoring and the handling of the transmitter may reduce individual anxiety and the number of patients unsatisfied with remote monitoring and, therefore, improve the quality of life. On the other hand, the idea to manage follow-ups via remote monitoring should be discussed with eligible patients individually due to shared decision-making.
Concerning patient education, patients answered to prefer a brochure ($43.9\%$), or instruction manual ($36.9\%$) as the medium to get information about remote monitoring and the transmitter but also accepted other tools such as mobile apps or websites with 23.6 and $21.4\%$, respectively. Brochures and instruction manuals should already be available for every patient. The development of other tools such as apps or patient websites could further improve the patients’ educational level [32, 33].
The majority of patients in this survey did not consider the transmitter as a limitation ($95.3\%$) and $60.3\%$ of the patients answered that they never experienced any technical issues. Nevertheless, $39.7\%$ of the patients stated to note technical issues. This should be taken seriously, as a working transmitter is crucial for continuous data transfer.
Most patients in this survey answered to refer to their attending cardiologist or other medical specialists in case of questions. As the ones most referred to, cardiologists and cardiology nurses should be carefully trained regarding troubleshooting and use of the different transmitters. Physicians responsible for interrogations should be aware of patients who do not send data and should subliminally reach out to those patients. Doing this could not only improve the patient’s quality of life but also further improve the continuity and quality of remote monitoring while minimizing adverse events due to insufficient monitoring.
## Future perspectives
A total of 383 patients answered to have a smartphone ($73.0\%$), whereas 142 do not have a smartphone ($27.0\%$). Most patients answered to use their smartphone for phone calls ($$n = 373$$, $96.9\%$), taking pictures ($$n = 305$$, $79.2\%$), or mobile internet use ($$n = 315$$, $81.8\%$; multiple answers possible). Other answers were less frequent (to play games: $$n = 73$$, $19.0\%$; other: $$n = 68$$, $17.7\%$). Two hundred and thirty-one patients reported using their smartphone several times a day ($60.2\%$), whereas 86 patients reported permanent use ($22.4\%$), 29 patients used several times a week ($7.6\%$), 24 patients used one time a day ($6.3\%$), and 14 patients (almost) never used their smartphone ($3.7\%$). Two hundred and ninety-nine ($81.5\%$) patients answered to download applications. Patients’ answers on the possible use of a smartphone for data transfer regarding remote monitoring are summarized in Figure 3.
**Figure 3:** *Patients’ opinions regarding the use of a smartphone for remote monitoring. Patients’ interest in using their own smartphone with an application instead of the transmitter for data transmission. Patients had to choose between 1 and 10 with 1 meaning “no interest” to 10 meaning “high interest”–median (P25; P75) 7.0 (2.0; 10.0) (A). Patients’ responses to the question if they would, asked today, prefer a smartphone with an application or a transmitter next to their bed for data transmission (B). Advantages (C) and disadvantages (D) of using a smartphone for remote monitoring.*
Younger patients had a significantly higher preference for using smartphone-based remote monitoring instead of a transmitter ($p \leq 0.001$). There was no difference when analyzing gender ($$p \leq 0.192$$) or residence ($$p \leq 0.355$$). Details on the interest in using the own smartphone instead of the transmitter are shown in Figure 4.
**Figure 4:** *Interest in using the own smartphone instead of the transmitter for data transmission, divided into groups regarding different parameters (A–D). Patients had to choose between 1 and 10 with 1 meaning “no interest” to 10 meaning “high interest.” p-values of < 0.05 were considered statistically significant. The small circles represent the median. ICD = implantable cardioverter defibrillator; PM = pacemaker; CRT = cardiac resynchronization therapy; ILR = implantable loop recorder.*
Most patients stated to have concerns regarding data safety ($$n = 156$$, $40.8\%$). Other concerns were battery consumption of the implanted device ($$n = 109$$, $28.5\%$) and the smartphone ($$n = 57$$, $14.9\%$), memory capacity ($$n = 79$$, $20.7\%$), data consumption ($$n = 72$$, $18.9\%$), and other concerns ($$n = 37$$, $9.7\%$). Other concerns were fear of mobile phone obsession, the necessity to always have their phone with them and always be online, fear of being hacked (data, mail, and implanted device), and the fear of complex handling for older patients. One hundred and forty-five patients ($38.0\%$) answered to have no concerns.
Asked for information that should be shown in a smartphone-based application (multiple answers possible), most patients chose “battery status” of the implantable device ($$n = 289$$, $84.3\%$), “validation of connectivity” ($$n = 255$$, $74.3\%$), or “information regarding technical issues” ($$n = 241$$, $70.3\%$; Figure 5). Twenty-two patients ($6.4\%$) answered “no information,” whereas 47 patients ($13.7\%$) answered to wish for “other” information.
**Figure 5:** *Patient’s responses on information that should be shown in an application regarding remote monitoring of CIEDs. Multiple answers were possible. CIED = cardiac implantable electronic device.*
The majority of patients ($73.0\%$) answered to use a smartphone, while only a minority of patients were younger than 60 years ($$n = 159$$). Thus, age is not necessarily a restriction regarding the use of smartphone-based techniques [34]. Accordingly, most patients participated in this survey via an online platform and not via the paper-based form. A total of $48.3\%$ of patients answered to prefer a smartphone-based application to a transmitter next to their bed for data transfer. Previous studies have shown that smartphone-based remote monitoring has the potential to improve the success rates of scheduled data transfers [35] and to improve patients’ compliance and connectivity [36]. Therefore, application-based data transfer may improve the management of remote monitoring in several ways: Feedback regarding successful or unsuccessful data transfers as well as a schedule for measurements regarding remote follow-ups could be integrated. Patients could be directly informed in case of events and possible next steps could be recommended within such an application. Moreover, patient education could be further enhanced with information regarding remote monitoring or specific situations. According to the patients who participated in this survey, information on battery status, validation of connectivity, and information regarding technical issues should be taken into consideration when developing such a tool. Nevertheless, patients’ concerns and disadvantages regarding such an idea should not be ignored. Especially, data safety is an important concern that should be noted [37]. Not all patients will have access to data transfer and remote monitoring via a smartphone. As shown in this survey, this may especially be relevant for older patients. Since current guidelines recommend remote monitoring for patients with difficulties attending in-person visits, this might be an issue that should be targeted.
As remote monitoring is known to reduce hospitalizations due to heart failure, application-based alerts can help patients and physicians to adapt and improve heart failure management [3, 4, 6, 7, 38, 39]. CIEDs may be helpful to contribute to optimized heart failure management. Nevertheless, studies have shown that with one parameter alone there may be no significant improvement in outcomes [38, 39]. Algorithms including multiple parameters may be more appropriate but have to be evaluated (40–42).
## Discussion
This is the first survey regarding the patient perspective on remote monitoring in Germany, not only focusing on patient satisfaction but also addressing advantages, concerns, and future perspectives perceived by the individual patient.
The key findings of this analysis are as follows:
## Limitations
This survey has several limitations regarding the respondents as well as the designed questionnaire. As the survey was anonymous and on a voluntary basis, the study population includes an expectable selection bias and baseline, as well as follow-up data on the patient group are scarce. Having only included patients already receiving remote monitoring, the study design did not provide a control group and does not allow comparisons of patients with remote monitoring vs. those without remote monitoring care, and thus, recall bias may be present. Nevertheless, the overall sample size provides important insights into a large cohort of patients followed-up by remote monitoring.
## Conclusion
In this national patient survey, remote monitoring led to a prolongation of intervals of in-person follow-up visits, while most patients reported unchanged or improved coping since the start of remote monitoring and answered to be open to new technologies regarding data transfer. Nevertheless, some patients reported concerns about remote monitoring as well as worse coping since the start of remote monitoring. The medical specialist was named as the person mostly referred to in case of problems. Thus, both the patients’ and the physicians’ education are important to improve the continuity and quality of remote monitoring while minimizing adverse events due to insufficient monitoring and therefore to improve the quality of life in patients with CIEDs and remote monitoring.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. All participants volunteered to participate in this anonymous data collection.
## Author contributions
HH acquisitioned the data, performed the analysis, interpreted the data, and drafted the manuscript. CH and OP revised the manuscript and provided substantial intellectual content. DD analyzed the data and drafted and revised the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
Abbott Medical supported the conduction of the survey by executing the postage of the questionnaires to patients and designing the online questionnaire, and also provided open access funding, but did not have any influence on patient selection, data analysis, nor writing or editing the manuscript. OP received lecture honorary from Abbott Medical, Biotronik, Medtronic, Zoll. DD received modest lecture honorary, travel grants, and/or a fellowship grant from Abbott, Astra Zeneca, Biotronik, Boehringer Ingelheim, Boston Scientific, Bristol Myers Squibb, CVRx, Medtronic, Microport, Pfizer, and Zoll.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcvm.2023.1123848/full#supplementary-material
## References
1. Piccini JP, Mittal S, Snell J, Prillinger JB, Dalal N, Varma N. **Impact of remote monitoring on clinical events and associated health care utilization: a nationwide assessment**. *Heart Rhythm* (2016) **13** 2279-86. DOI: 10.1016/j.hrthm.2016.08.024
2. Portugal G, Cunha P, Valente B, Feliciano J, Lousinha A, Alves S. **Influence of remote monitoring on long-term cardiovascular outcomes after cardioverter-defibrillator implantation**. *Int J Cardiol* (2016) **222** 764-8. DOI: 10.1016/j.ijcard.2016.07.157
3. Simone AD, Leoni L, Luzi M, Amellone C, Stabile G, Rocca VL. **Remote monitoring improves outcome after ICD implantation: the clinical efficacy in the management of heart failure (EFFECT) study**. *Europace* (2015) **17** 1267-75. DOI: 10.1093/europace/euu318
4. Boriani G, Costa AD, Quesada A, Ricci RP, Favale S, Boscolo G. **Effects of remote monitoring on clinical outcomes and use of healthcare resources in heart failure patients with biventricular defibrillators: results of the MORE-CARE multicentre randomized controlled trial**. *Eur J Heart Fail* (2017) **19** 416-25. DOI: 10.1002/ejhf.626
5. Mabo P, Victor F, Bazin P, Ahres S, Babuty D, Costa AD. **A randomized trial of long-term remote monitoring of pacemaker recipients (the COMPAS trial)**. *Eur Heart J* (2012) **33** 1105-11. DOI: 10.1093/eurheartj/ehr419
6. Chiu CSL, Timmermans I, Versteeg H, Zitron E, Mabo P, Pedersen SS. **Effect of remote monitoring on clinical outcomes in European heart failure patients with an implantable cardioverter-defibrillator: secondary results of the REMOTE-CIED randomized trial**. *Europace* (2021) **24** 256-67. DOI: 10.1093/europace/euab221
7. Versteeg H, Timmermans I, Widdershoven J, Kimman GJ, Prevot S, Rauwolf T. **Effect of remote monitoring on patient-reported outcomes in European heart failure patients with an implantable cardioverter-defibrillator: primary results of the REMOTE-CIED randomized trial**. *Europace* (2019) **21** 1360-8. DOI: 10.1093/europace/euz140
8. Guédon-Moreau L, Lacroix D, Sadoul N, Clémenty J, Kouakam C, Hermida JS. **A randomized study of remote follow-up of implantable cardioverter defibrillators: safety and efficacy report of the ECOST trial**. *Eur Heart J* (2013) **34** 605-14. DOI: 10.1093/eurheartj/ehs425
9. Glikson M, Nielsen JC, Kronborg MB, Michowitz Y, Auricchio A, Barbash IM. **2021 ESC guidelines on cardiac pacing and cardiac resynchronization therapy**. *Eur Heart J* (2021) **42** 3427-520. DOI: 10.1093/eurheartj/ehab364
10. Slotwiner D, Varma N, Akar JG, Annas G, Beardsall M, Fogel RI. **HRS expert consensus statement on remote interrogation and monitoring for cardiovascular implantable electronic devices**. *Heart Rhythm* (2015) **12** e69-e100. DOI: 10.1016/j.hrthm.2015.05.008
11. Parthiban N, Esterman A, Mahajan R, Twomey DJ, Pathak RK, Lau DH. **Remote monitoring of implantable cardioverter-defibrillators a systematic review and meta-analysis of clinical outcomes**. *J Am Coll Cardiol* (2015) **65** 2591-600. DOI: 10.1016/j.jacc.2015.04.029
12. Duncker D, Michalski R, Müller-Leisse J, Zormpas C, König T, Veltmann C. **Devicebasiertes Telemonitoring**. *Herzschrittmacherther Elektrophysiol* (2017) **28** 268-78. DOI: 10.1007/s00399-017-0521-3
13. Poole JE, Johnson GW, Hellkamp AS, Anderson J, Callans DJ, Raitt MH. **Prognostic importance of defibrillator shocks in patients with heart failure**. *N Engl J Med* (2008) **359** 1009-17. DOI: 10.1056/nejmoa071098
14. Zeppenfeld K, Tfelt-Hansen J, de RM, Winkel BG, Behr ER, Blom NA. **ESC guidelines for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death**. *Eur Heart J* (2022) **43** 3997-4126. DOI: 10.1093/eurheartj/ehac262
15. Ricci RP, Morichelli L, Santini M. **Remote control of implanted devices through home MonitoringTM technology improves detection and clinical management of atrial fibrillation**. *Europace* (2009) **11** 54-61. DOI: 10.1093/europace/eun303
16. Crossley GH, Boyle A, Vitense H, Chang Y, Mead RH, Investigators C. **The CONNECT (clinical evaluation of remote notification to reduce time to clinical decision) trial the value of wireless remote monitoring with automatic clinician alerts**. *J Am Coll Cardiol* (2011) **57** 1181-9. DOI: 10.1016/j.jacc.2010.12.012
17. Akar JG, Bao H, Jones PW, Wang Y, Varosy PD, Masoudi FA. **Use of remote monitoring is associated with lower risk of adverse outcomes among patients with implanted cardiac defibrillators**. *Circ Arrhythm Electrophysiol* (2015) **8** 1173-80. DOI: 10.1161/circep.114.003030
18. Ricci RP, Vicentini A, D’Onofrio A, Sagone A, Rovaris G, Padeletti L. **Economic analysis of remote monitoring of cardiac implantable electronic devices: results of the health economics evaluation registry for remote follow-up (TARIFF) study**. *Heart Rhythm* (2017) **14** 50-7. DOI: 10.1016/j.hrthm.2016.09.008
19. Perl S, Stiegler P, Rotman B, Prenner G, Lercher P, Anelli-Monti M. **Socio-economic effects and cost saving potential of remote patient monitoring (SAVE-HM trial)**. *Int J Cardiol* (2013) **169** 402-7. DOI: 10.1016/j.ijcard.2013.10.019
20. García-Fernández FJ, Asensi JO, Romero R, Lozano IF, Larrazabal JM, Ferrer JM. **Safety and efficiency of a common and simplified protocol for pacemaker and defibrillator surveillance based on remote monitoring only: a long-term randomized trial (RM-ALONE)**. *Eur Heart J* (2019) **40** 1837-46. DOI: 10.1093/eurheartj/ehz067
21. Ricci RP, Morichelli L, D’Onofrio A, Calò L, Vaccari D, Zanotto G. **Effectiveness of remote monitoring of CIEDs in detection and treatment of clinical and device-related cardiovascular events in daily practice: the HomeGuide registry**. *Europace* (2013) **15** 970-7. DOI: 10.1093/europace/eus440
22. Januszkiewicz Ł, Barra S, Providencia R, Conte G, de AC, Chun JKR. **Long-term quality of life and acceptance of implantable cardioverter-defibrillator therapy: results of the European heart rhythm association survey**. *Europace* (2022) **24** 860-7. DOI: 10.1093/europace/euac011
23. Timmermans I, Meine M, Szendey I, Aring J, Roldán JR, Erven L. **Remote monitoring of implantable cardioverter defibrillators: patient experiences and preferences for follow-up**. *Pacing Clin Electrophysiol* (2019) **42** 120-9. DOI: 10.1111/pace.13574
24. Su D, Michaud TL, Estabrooks P, Schwab RJ, Eiland LA, Hansen G. **Diabetes management through remote patient monitoring: the importance of patient activation and engagement with the technology**. *Telemed J E Health* (2019) **25** 952-9. DOI: 10.1089/tmj.2018.0205
25. Anker SD, Agewall S, Borggrefe M, Calvert M, Caro JJ, Cowie MR. **The importance of patient-reported outcomes: a call for their comprehensive integration in cardiovascular clinical trials**. *Eur Heart J* (2014) **35** 2001-9. DOI: 10.1093/eurheartj/ehu205
26. Simovic S, Providencia R, Barra S, Kircanski B, Guerra JM, Conte G. **The use of remote monitoring of cardiac implantable devices during the COVID-19 pandemic: an EHRA physician survey**. *Europace* (2021) **24** 473-80. DOI: 10.1093/europace/euab215
27. Pedersen SS, Sears SF, Burg MM, van den BKC. **Does ICD indication affect quality of life and levels of distress?**. *Pacing Clin Electrophysiol* (2009) **32** 153-6. DOI: 10.1111/j.1540-8159.2008.02196.x
28. Gopinathannair R, Lerew DR, Cross NJ, Sears SF, Brown S, Olshansky B. **Longitudinal changes in quality of life following ICD implant and the impact of age, gender, and ICD shocks: observations from the INTRINSIC RV trial**. *J Interv Card Electrophysiol* (2017) **48** 291-8. DOI: 10.1007/s10840-017-0233-y
29. Petersen HH, Larsen MCJ, Nielsen OW, Kensing F, Svendsen JH. **Patient satisfaction and suggestions for improvement of remote ICD monitoring**. *J Interv Card Electrophysiol* (2012) **34** 317-24. DOI: 10.1007/s10840-012-9675-4
30. Watanabe E, Kasai A, Fujii E, Yamashiro K, Brugada P. **Reliability of implantable cardioverter defibrillator home monitoring in forecasting the need for regular office visits, and patient perspective**. *Circ J* (2013) **77** 2704-11. DOI: 10.1253/circj.cj-13-0387
31. Marzegalli M, Lunati M, Landolina M, Perego GB, Ricci RP, Guenzati G. **Remote monitoring of CRT-ICD: the multicenter Italian CareLink evaluation-ease of use, acceptance, and organizational implications**. *Pacing Clin Electrophysiol* (2008) **31** 1259-64. DOI: 10.1111/j.1540-8159.2008.01175.x
32. Duncker D, Svennberg E, Deharo JC, Costa FM, Kommata V. **The ‘afibmatters.org’ educational website for patients with atrial fibrillation from the European heart rhythm association**. *Europace* (2021) **23** 1693-7. DOI: 10.1093/europace/euab098
33. Kommata V, Deharo JC, Drossart I, Foldager D, Svennberg E, Vernooy K. **The ‘myrhythmdevice.org’ educational website for patients with implanted cardiac devices from the European heart rhythm association**. *Europace* (2022) **24** 1713-5. DOI: 10.1093/europace/euac137
34. Svennberg E, Tjong F, Goette A, Akoum N, Biaise LD, Bordachar P. **How to use digital devices to detect and manage arrhythmias: an EHRA practical guide**. *Europace* (2022) **24** 979-1005. DOI: 10.1093/europace/euac038
35. Tarakji KG, Zaidi AM, Zweibel SL, Varma N, Sears SF, Allred J. **Performance of first pacemaker to use smart device app for remote monitoring**. *Heart Rhythm O2* (2021) **2** 463-71. DOI: 10.1016/j.hroo.2021.07.008
36. Manyam H, Burri H, Casado-Arroyo R, Varma N, Lennerz C, Klug D. **Smartphone-based cardiac implantable electronic device remote monitoring: improved compliance and connectivity**. *Eur Heart J Digit Health* (2022) **4** 43-52. DOI: 10.1093/ehjdh/ztac071
37. Nielsen JC, Lin YJ, Figueiredo MJ d O, Shamloo AS, Alfie A, Boveda S. **European heart rhythm association (EHRA)/Heart Rhythm Society (HRS)/Asia Pacific Heart Rhythm Society (APHRS)/Latin American Heart Rhythm Society (LAHRS) expert consensus on risk assessment in cardiac arrhythmias: use the right tool for the right outcome, in the right population**. *Europace* (2020) **22** 1147-8. DOI: 10.1093/europace/euaa065
38. Böhm M, Drexler H, Oswald H, Rybak K, Bosch R, Butter C. **Fluid status telemedicine alerts for heart failure: a randomized controlled trial**. *Eur Heart J* (2016) **37** 3154-63. DOI: 10.1093/eurheartj/ehw099
39. van VDJ, Braunschweig F, Conraads V, Ford I, Cowie MR, Jondeau G. **Intrathoracic impedance monitoring, audible patient alerts, and outcome in patients with heart failure**. *Circulation* (2011) **124** 1719-26. DOI: 10.1161/circulationaha.111.043042
40. Guerra F, D’Onofrio A, Ruvo ED, Manzo M, Santini L, Giubilato G. **Decongestive treatment adjustments in heart failure patients remotely monitored with a multiparametric implantable defibrillators algorithm**. *Clin Cardiol* (2022) **45** 670-8. DOI: 10.1002/clc.23832
41. López-Azor JC, de la TN, Carmena MDGC, Pérez PC, Munera C, MarcoClement I. **Clinical utility of HeartLogic, a multiparametric Telemonitoring system, in heart failure**. *Cardiac Fail Rev* (2022) **8** e13. DOI: 10.15420/cfr.2021.35
42. Calò L, Bianchi V, Ferraioli D, Santini L, Russo AD, Carriere C. **Multiparametric implantable cardioverter-defibrillator algorithm for heart failure risk stratification and management: an analysis in clinical practice**. *Circ Heart Fail* (2020) **14** e008134. DOI: 10.1161/circheartfailure.120.008134
|
---
title: Influence of state of health and personality factors of resilience and coping
in healthy subjects and those with diabetes
authors:
- Cristina Rivera-Picón
- María Hinojal Benavente-Cuesta
- María Paz Quevedo-Aguado
- Raúl Juárez-Vela
- Jesús Martinez-Tofe
- Juan Luis Sánchez-González
- Pedro Manuel Rodríguez-Muñoz
journal: Frontiers in Public Health
year: 2023
pmcid: PMC10017435
doi: 10.3389/fpubh.2023.1074613
license: CC BY 4.0
---
# Influence of state of health and personality factors of resilience and coping in healthy subjects and those with diabetes
## Abstract
### Introduction
Currently, the most common chronic metabolic disease in our society is Diabetes Mellitus. The diagnosis of Diabetes Mellitus supposes an impact for the patient, since it requires a modification in the lifestyle, which demands a great capacity for adaptation and modification of habits. The aim of the study was to determine whether personality factors and health status influence resilience and coping strategies in a sample of healthy and diabetic subjects.
### Methodology
The sample included a total of 401 subjects (201 patients with Diabetes and 200 without pathology). The instruments applied for data collection were: *Sociodemographic data* questionnaire, the Resilience Scale, the Coping Strategies Questionnaire and The “Big Five” factor taxonomy. The data collection period was approximately 2 years (between February 2018 and January 2020).
### Results
Certain personality factors, such as Emotional Stability, Integrity, Conscientiousness and Extraversion, were positively related to Resilience. Additionally, Emotional Stability, Integrity, and Extraversion were positively associated with Rational Coping. On the other hand, emotional stability, agreeableness and extraversion were negatively related to emotional coping. In relation to health status, the absence of pathology is related to the use of rational strategies more than to the diagnosis of diabetes. Therefore, the participants in this study present different psychological patterns depending on personality and health status.
### Conclusions
The present study shows that the subjects of the sample present different psychological patterns depending on Personality and health status.
## 1. Introduction
According to the latest studies, the most frequent chronic metabolic disease in our society is diabetes mellitus [1]. In 2002, the WHO announced a worldwide prevalence of diabetes of $3\%$, which corresponds to 170 million people in the world diagnosed with this pathology. It was even estimated that this figure would double by 2025 [2]. Today, these forecasts have already been exceeded. The latest figures provided by the International Diabetes Federation (IDF), corresponding to 2019, showed that $9.3\%$ of adults have diabetes, which corresponds to a total of 463 million people. They also indicated that 1.1 million children and adolescents under the age of 20 live with type 1 diabetes. In addition, the IDF estimates that in 2030, 578 million adults will be living with the disease. In 2045, it is estimated that the figure will rise to 700 million [3].
These data are of great importance because the diagnosis, prognosis and treatment of the disease have a great emotional impact on the patient. This is associated with the need to assume a pathology that will accompany the subject throughout their life and the subject's obligation to modify their life habits in order to obtain a better quality of life, thus reducing complications deriving from it [4].
There is even research that details the psychological repercussions that accompany diabetes [5, 6]. Thus, it has been stated that said pathology can be associated with depression and anxiety. These two diseases can arise regardless of the type of diabetes, especially in the presence of clinical complications. Therefore, it is essential that health workers use programmes in their clinical practice that address the emotional demands detected. Even the American Diabetes Association has incorporated new medical care recommendations, with the intention of including the assessment of the psychological and social situation of subjects diagnosed with diabetes [7, 8].
Based on the aforementioned theoretical aspects, some research has been oriented toward the identification of those psychological mediators that could contribute to achieving a better quality of life in subjects with chronic disease. Thus, the impact of resilience, personality and coping strategies on the health of subjects with chronic pathologies has been studied.
Therefore, research studying resilience and coping strategies has increased (9–11). Resilience is defined from the health field as the ability of individuals to maintain health and quality of life in a dynamic and challenging environment. Therefore, it is considered relevant variable for in the area of health due to its capacity to buffer stress (12–16). In the study by Pasantes et al. interventions were carried out with diabetic patients aimed at promoting their level of resilience. These incidents had a positive result on hemoglobin A1C levels [17]. The type of coping strategies that people use to adapt to their illness can anticipate the impact caused by said illness. Therefore, certain coping styles can mediate and buffer the effects of stress. It is stated that active coping strategies are positively related to health [18].
In addition, personality may also play a fundamental role in the way subjects deal with the disease, directly influencing their wellbeing. Thus, personality modulates the way in which people face and adapt to a chronic disease, favoring the development of resilience and the use of coping strategies (19–21). The psychological aspect of diabetes is considered an important part of the treatment and management of this condition in the modern world. Thus, the assessment of personality traits can play a substantial role in the proper treatment of diabetics. The study by Esmaeilinasab et al. was determined that extraversion in diabetic patients is associated with better disease control [21]. In addition, this may be relevant if we take into account that some studies indicate that patients with diabetes have different personality traits than subjects without pathology [22]. In the context of this study, personality is approached through the Big Five model in Spanish, which hierarchically orders five personality factors: emotional stability, agreeableness, integrity, conscientiousness and extraversion [23].
Based on the scientific evidence outlined above, the objective of this study was to determine whether health status and personality factors influence resilience and coping strategies in a sample of healthy and diabetic subjects. For this reason, we consider it essential to know which diseases predict a worse adaptation, as well as those personality characteristics that favor the development of resilience, in order to focus on effective and individualized health programme.
The novelty of this study is justified in the use of a clinical sample. This allowed us to assess the influence of the Personality, but also the health status of the subjects to explain the Resilience and Coping. In most studies, only the relationships between these psychological variables have been investigated, through samples with healthy population, such as university students [24, 25].
## 2.1. Aim and design of the study
The aim of this study was to determine whether health status and personality factors influence resilience and coping strategies in a sample of healthy and diabetic subjects. The study had a non-experimental cross-sectional design with a correlational objective.
## 2.2. Participants
These samples were selected at the University Assistance Complex of Salamanca. Four hundred and thirty six subjects participated in the study, of which 35 were excluded for not completing the informed consent or not completing the questionnaires. The total sample consisted of 401 subjects (Figure 1).
**Figure 1:** *Sample selection flowchart.*
The study participants were 200 healthy subjects and 201 patients with diabetes ($$n = 401$$). The majority of it is made up of men ($$n = 285$$) and are mainly aged between 44 and 50 years ($$n = 118$$). Most of the subjects are married/in a couple ($$n = 247$$) and only 85 subjects have higher education (Table 1).
**Table 1**
| Unnamed: 0 | Diabetes | Diabetes.1 | Healthy | Healthy.1 | Total | Total.1 | Ji | TE | p |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | N | % | N | % | N | % | | | |
| N° participants | 201 | 50.1% | 200 | 49.9% | 401 | 100% | | | |
| Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex |
| Woman | 58 | 28.9% | 58 | 29.0% | 116 | 28.9% | 0.001 | 0.002 | 0.975 |
| Man | 143 | 71.1% | 142 | 71.0% | 285 | 71.6% | | | |
| Age | Age | Age | Age | Age | Age | Age | Age | Age | Age |
| 43 years or younger | 46 | 22.9% | 59 | 29.5% | 105 | 26.2% | 2.588 | 0.080 | 0.460 |
| 44 to 50 years | 63 | 31.3% | 55 | 27.5% | 118 | 29.4% | | | |
| From 51 to 55 years old | 44 | 21,9% | 38 | 19.0% | 82 | 20.4% | | | |
| 56 years or older | 48 | 23,9% | 48 | 24.0% | 96 | 23.9% | | | |
| Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status |
| Married/couple | 123 | 61.2% | 124 | 62.0% | 247 | 61.6% | 4.574 | 0.107 | 0.102 |
| Single/widowed/other | 64 | 31.8% | 51 | 25.5% | 115 | 28.7% | | | |
| Separated/divorced | 14 | 07.0% | 25 | 12.5% | 39 | 9.7% | | | |
| Level of studies | Level of studies | Level of studies | Level of studies | Level of studies | Level of studies | Level of studies | Level of studies | Level of studies | Level of studies |
| Secondary or lower | 160 | 79.6% | 156 | 78.0% | 316 | 78.8% | 0.154 | 0.020 | 0.695 |
| Superior | 41 | 20.4% | 44 | 22.0% | 85 | 21.2% | | | |
Pearson's χ2 test was used, using Cramer's V to determine the effect size. Thus, it was detected that the subsamples of our study, no significant differences were detected in the sociodemographic variables ($p \leq 0.05$) (Table 1).
In both subsamples, to participate in the project, the following inclusion criteria must be met: The subjects can be of legal age and participate voluntarily in the study. In the case of subjects with diabetes, an additional inclusion criterion was having a confirmed diagnosis of said disease, regardless of its stage. An additional inclusion criterion in healthy patients was that they were not diagnosed with any disease. The exclusion criteria were: suffering from a disease that would prevent the patient from completing the study, not agree to participate in the study and have been diagnosed with an affective pathology that could bias the results.
## 2.3. Data collection
The samples were selected following a quota sample with equivalent age ranges, sex and educational level, with the aim of achieving homogeneous sub-samples. The sample was collected at the University Assistance Complex of Salamanca. The selection of the subsample made up of diabetic patients was carried out in the Diabetes Unit of the Clinical Hospital of Salamanca and the Internal Medicine hospitalization wards of the same hospital.
After obtaining the sample of subjects with diabetes, the sample of healthy subjects was selected. The selection of this subsample was carried out in different Salamanca health centers (namely, “Periurbana Sur” and “Capuchinos” Health Centers). These patients voluntarily participated in the study after attending their scheduled appointment in the nursing consultation.
The data collection period was ~2 years (between February 2018 and January 2020), through the instruments detailed below.
## 2.3.1. Sociodemographic data questionnaire
Sociodemographic data were collected through an instrument made up of a series of questions of a socio-demographic nature and information on the presence of diabetes.
## 2.3.2. Personality: “Big Five” factor taxonomy
To assess personality factors, the “Taxonomic Proposal of the Big Five in Spanish” was produced by Iraegui and Quevedo-Aguado [26]. This research consisted of a psycholinguistic approach to the study of personality following the “Big Five hypothesis.” Principal component factor analysis was applied to the 150 mini-markers finally identified in this research as personality descriptors. The Kaiser rule was employed to select the number of factors to retain and varimax normalization was used as the rotation method.
The five factor solution required ten iterations for convergence and explained $19.36\%$ of the total variance with a Cronbach's α of 0.88. For our study we have used a reduced scale of 50 personality descriptors, ten for each factor (five positive and five negative). These descriptors were chosen based on their correlations with the corresponding factor. In this investigation, the use of the reduced version was chosen due to its brevity and its adequate psychometric properties. The global reliability of the instrument is α = 0.884, finding each of the five factors in indices that oscillate between α = 0.079 and α = 0.89 [26].
In this scale, the subjects have to evaluate these 50 descriptors depending on whether they are suitable or not for defining their personality traits. The response range was from 0, not suitable, to 4, very suitable. A total score was obtained for each of the factors.
## 2.3.3. Coping strategies questionnarie
The Coping Strategies Questionnaire scale was designed by the authors Sandín and Chorot in 2002. This questionnaire contains a scale made up of 42 items, which score from 0 (never) to 4 (almost always). Through this scale, two general dimensions of coping can be measured: emotional coping and rational coping. Also, based on this general classification, it allows assessing seven more specific coping dimensions. Thus, emotional coping includes negative self-focused coping and overt emotional expression. Rational coping included problem-solving coping, positive reappraisal, and seeking social support. Each coping factor/dimension includes seven items, with the total variance explained by the seven factors being $55.3\%$ [27].
## 2.3.4. Resilience scale
Wagnild and Young created The Resilience Scale in 1993, adapted by Novella in 2002 into Spanish [28]. This scale has 25 items. Each item ranges from 1 (strongly disagree) to 7 (strongly agree).
The scale assesses five resilience factors: personal satisfaction, equanimity, feeling good alone, self-confidence, and perseverance. Global internal consistency was measured using Cronbach's α coefficient (α = 0.88).
## 2.4. Ethical considerations
This study received a positive report from the Clinical Research Ethics Committee of the University Hospital of Salamanca PIO$\frac{02}{01}$/2018.
## 2.5. Data analysis
Statistical analysis was performed using International Business Machines' (IBM) Statistical Package for the Social Sciences (SPSS) version 25 (IBM Corp., Armonk, NY, USA). To determine whether personality factors and health status influence resilience and coping strategies in a sample of healthy subjects and those with diabetes, linear regression analysis was performed.
This technique requires the fulfillment of five assumptions: independence, non-collinearity, linearity, homoscedasticity and normality. After verifying these assumptions, the linear regression analysis was applied. For this, the variables were grouped into two blocks. One of them contained the dummy variables, and the other the rest of the variables. For the first block the method was introduced, while for the second the stepwise regression was obtained. To study the fit of the model, the coefficient of determination (R2) and the adjusted coefficient of determination (RA2) were used. In all statistical test, testing was significant when $p \leq 0.05.$
## 3.1. Linear regression
For each dependent variable, a linear regression analysis was performed. These de-pendent variables (DVs) were resilience, rational coping and emotional coping. In contrast, the independent variables (IVs) were the factors of personality and the state of health of the subjects. In categorical VI with two or more levels, it was necessary to create dummy variables.
## 3.1.1. Resilience
In relation to resiliene, the regression revealed that the best model is the one in which four variables were included (R2 = 0.848, RA2 = 0.719, F (df1, df2) = 20.047 [1, 393], p ≤ 0.001). These predictor variables included in the final model explained $71.6\%$ of the variance in DVs.
Table 2 shows that emotional stability, integrity, conscientiousness and extraversion were positively related to resilience. Personality factor 1 (emotional stability) was the one with the greatest weight ($B = 0.883$, β = 0.457, $t = 11.998$, $p \leq 0.001$). In relation to health status, the dummy of healthy subjects was not significant ($B = 0.990$, β = 0.875, $t = 1.137$, $$p \leq 0.256$$).
**Table 2**
| Unnamed: 0 | B | Std error | β | t | p |
| --- | --- | --- | --- | --- | --- |
| Constant | 32.589 | 3.938 | | 8.279 | < 0.001 |
| F1 emotional stability factor | 0.883 | 0.074 | 0.457 | 11.998 | < 0.001 |
| F3 integrity factor | 0.663 | 0.115 | 0.221 | 5.756 | < 0.001 |
| F4 conscientiousness factor | 0.449 | 0.099 | 0.175 | 5.531 | < 0.001 |
| F5 extraversion factor | 0.384 | 0.078 | 0.166 | 4.936 | < 0.001 |
| Healthy subjects | 0.99 | 0.875 | 0.031 | 1.137 | 0.256 |
Finally, the existence of atypical and predominant cases was assessed. Table 3 shows the value of the most extreme cases in different measures. Two atypical cases are detected in RV but neither of them is predominant.
**Table 3**
| Unnamed: 0 | Maximum | Minimum | Cases out of range |
| --- | --- | --- | --- |
| Typified residues | 4.166 | −2.971 | 2 |
| Waste diversion | 4.288 | −3.103 | 2 |
| Leverage | 0.075 | - | 0 |
| Cook's distance | 0.11 | - | 0 |
## 3.1.2. Rational coping
The regression revealed that the best model is the one in which four variables were included (R2 = 0.776, RA2 = 0.598, F (df1, df2) = 22.527 [1, 393], $p \leq 0.001$). These predictor variables included in the final model explained $59.8\%$ of the RV variance.
Table 4 shows that emotional stability, integrity and extraversion are positively related to rational coping. In relation to the state of health, the dummy of healthy subjects was significant, being the variable that had the greatest weight ($B = 4.244$, β = 0.153, $t = 4.746$, p ≤ 0.001).
**Table 4**
| Unnamed: 0 | B | Std error | β | t | p |
| --- | --- | --- | --- | --- | --- |
| Constant | 23.016 | 2.702 | | 8.521 | < 0.001 |
| F1 emotional stability factor | 0.554 | 0.073 | 0.333 | 7.582 | < 0.001 |
| F3 integrity factor | 0.859 | 0.101 | 0.331 | 8.507 | < 0.001 |
| F5 extraversion factor | 0.491 | 0.08 | 0.247 | 6.161 | < 0.001 |
| Healthy subjects | 4.244 | 0.894 | 0.153 | 4.746 | < 0.001 |
Finally, the existence of atypical and predominant cases was assessed. Table 5 shows the value of the most extreme cases in different measures. Two atypical cases are detected, but neither is predominant.
**Table 5**
| Unnamed: 0 | Maximum | Minimum | Cases out of range |
| --- | --- | --- | --- |
| Typified residues | 4.116 | −2.971 | 2 |
| Waste diversion | 4.288 | −3.103 | 4 |
| Leverage | 0.075 | - | 0 |
| Cook's distance | 0.11 | - | 0 |
## 3.1.3. Emotional coping
The regression revealed that the best model is the one in which three variables were included (R2 = 0.457, RA2 = 0.243, F (df1, df2) = 5.539 [1, 394], $$p \leq 0.019$$). These predictor variables included in the final model explained $24.3\%$ of the variance in DVs.
Table 6 shows that emotional stability, agreeableness and extraversion were negatively related to emotional coping. In relation to health status, healthy subjects have negative coefficients, but this variable was not significant (B = −0.375, β = 0.033, t = −0.733, $$p \leq 0.464$$).
**Table 6**
| Unnamed: 0 | B | Std error | β | t | p |
| --- | --- | --- | --- | --- | --- |
| Constant | 25.231 | 1.309 | | 19.278 | < 0.001 |
| F1 Emotional Stability factor | −0.257 | 0.042 | −0.378 | −6.125 | < 0.001 |
| F2 Agreeableness factor | −0.181 | 0.048 | −0.204 | −3.783 | < 0.001 |
| F5 Extraversion factor | −0.106 | 0.046 | −0.130 | −2.323 | 0.021 |
| Healthy subjects | −0.375 | 0.511 | −0.033 | −0.733 | 0.464 |
Finally, the existence of atypical and predominant cases was assessed. Table 7 shows the value of the most extreme cases in different measures. Two atypical cases are detected, but neither of them is predominant.
**Table 7**
| Unnamed: 0 | Maximum | Minimum | Cases out of range |
| --- | --- | --- | --- |
| Typified residues | 4.166 | −2.973 | 2 |
| Waste diversion | 4.288 | −3.103 | 3 |
| Leverage | 0.07 | - | 0 |
| Cook's distance | 0.11 | - | 0 |
## 4. Discussion
Different studies have confirmed that the diagnosis of a chronic pathology and individual differences in personality traits can influence the development and maintenance of resilience and coping strategies (20, 26–29). Then, these results are compared with those of our project.
The results obtained in our research show that emotional stability, integrity, responsibility and extraversion were positively related to resilience. We can highlight the fact that emotional stability turned out to be a significant predictor in all models. In addition, it was the variable with the highest weight for the resilience variable.
In a similar vein, numerous studies reflect the negative association between resilience and neuroticism (19, 25, 30–35). In order to compare this argument with the results of our re-search, it should be noted that the subjects with a low score in neuroticism are located on the opposite side of the emotional stability factor. Therefore, the results of the cited authors coincide with those of our study: the emotional stability factor was the one that was most related to higher levels of resilience.
Other research also reflects the positive relationship between extraversion, integrity and conscientiousness with levels of resilience (33, 35–39). This evidence also coincides with the results presented in our study, which reflect that the three factors contributed significantly to the prediction of resilience.
Also, the results of our research, in relation to the study of coping strategies, revealed that emotional stability, integrity and extraversion were positively related to rational coping. Emotional stability, agreeableness and extraversion were negatively associated with emotional coping. Other authors have also addressed the relationship between personality factors and coping strategies. The research found on this study topic indicates that the way in which an individual faces problem is influenced by their personality traits [40, 41]. Thus, Mirnics et al. [ 42] found in their research that emotional stability was the trait that most significantly predicted coping strategies. Thus, emotional stability was associated positively with rational strategies and negatively with emotional strategies. In addition, extraversion and conscientiousness were found to be positively related to the use of rational strategies. Other authors, such as Afshar et al. [ 43], found similar results in their research, pointing out that subjects with a higher level of extraversion, integrity and emotional stability frequently use more rational strategies. This evidence is in line with the results obtained in our study. We also found results similar to those obtained in our research in the work of Leszko et al. [ 44], which indicated that agreeableness is negatively associated with emotional coping.
Therefore, the results obtained in our research are consistent with published studies on personality factors that predict resilience and coping. In relation to health status, our research shows that the absence of a pathology predicted a greater use of rational coping strategies.
However, there is little research focused on predicting the levels of resilience and coping strategies used based on the health status of the subjects. However, it has been described that the diagnosis of a chronic disease is a stress factor, hindering the development of resilience and threatening the coping capacity of the individual [14]. Our results predicted that healthy and diabetic subjects would not present differences in the resilience variable. Thus, subjects with diabetes have learned how to face, overcome and transform themselves in the face of adversity.
It should be noted that there is very little research with which we can compare these results. Thus, few studies compare the level of resilience of subjects with diabetes and healthy subjects. However, we found similar results to those presented in our project in the research carried out by Novaes [45], which was conducted with a sample of subjects with diabetes mellitus and healthy subjects. In this study, it was found that there were no significant differences in the level of resilience between the groups. This finding coincides with that obtained in our study [45].
However, other studies have shown that healthy subjects have a higher level of resilience than patients with chronic pathologies [14]. Thus, we consider it necessary to develop more research that evaluates and com-pares resilience in specific chronic diseases. It should be noted that each chronic disease has very different characteristics in terms of its development and therapeutic plan. For this reason, it is necessary to carry out more studies that compare the level of resilience according to the state of health. We consider it essential to investigate and learn about the pathologies associated with lower levels of resilience, since different studies coincide in believing that resilient people are more capable of coping with disease processes, both their own and those of others, and emerge stronger from the situation [13, 46].
Finally, the results of this study show that the state of health was also related to the type of coping strategies. The subsample of healthy subjects presented a greater use of rational cutting strategies. These strategies are associated with positive coping, coping with stress and trauma differently between individuals [47]. Furthermore, rational coping, characterized by the mobilization of the patient to deal with the disease, is associated with greater adaptation to the disease and a higher quality of life (46–49). However, we find opposite conclusions in other studies, which state that diabetic subjects more frequently use rational coping strategies [50, 51].
## 4.1. Limitations
We point out as the main limitation that personality and health status only explained $24.3\%$ of emotional coping. Therefore, variables that help improve our predictions are missing. However, there are investigations that state that resilience and gender can also predict the type of emotional coping used [52]. Future research should take into account these variables not included in the models, which may be relevant for predicting the variables that are not well-explained.
## 5. Conclusion
Subjects present different psychological patterns depending on personality and health status. This conclusion may be useful in clinical practice for developing strategies, individually, focused on individuals with certain personality characteristics that predict a greater risk of maladjustment to their disease. Also, in an individualized way, strategies could be developed focused on individuals with certain personality characteristics that predict a greater risk of maladjustment to their disease. Therefore, the conclusions of this study show the importance of developing individualized health programs to address diabetes. However, it would be important to expand the study with other chronic diseases.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by El Comité Ético de Investigación Con Medicamentos del Área de Salud de Salamanca. Código CEIC: PlO$\frac{02}{01}$/2018. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
Conceptualization: CR-P, MHB-C, and PMR-M; methodology: CR-P and JM-T; software: JM-T and JLS-G; validation: MHB-C, RJ-V, and CR-P; formal analysis: CR-P, RJ-V, and JLS-G; investigation: RJ-V and PMR-M; resources: PMR-M; data curation: CR-P, JLS-G, and JM-T; writing-original draft preparation: CR-P and PMR-M; writing-review and editing: CR-P and MPQ-A; visualization: CR-P and JM-T; supervision: JM-T and PMR-M; project administration: CR-P; funding acquisition: RJ-V, JM-T, and JLS-G. All authors have read and agreed to the published version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Nadal J, Cases M, Puente D. **Epidemiología y control clínico de la diabetes mellitus tipo 2 y sus comorbilidades en España (estudio e Control)**. *Medicina Cl* (2021.0) **147** 1-7. DOI: 10.1016/S0025-7753(17)30618-8
2. Ramirez S, Villa-Ruano N, García D. **Epidemiología genética sobre las teorías causales y la patogénesis de la diabetes mellitus tipo 2**. *Gaceta Médica de México* (2017.0) **153** 864-874. DOI: 10.24875/GMM.17003064
3. Saeedi P, Petersohn I, Salpea P, Malanda B, Karuranga S, Unwin N. **Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: results from the International Diabetes Federation Diabetes Atlas, 9 th edition**. *Diab Res Clin Pract* (2019.0) **157** 107843. DOI: 10.1016/j.diabres.2019.107843
4. Beléndez M, Lorente I, Maderuelo M. **Emotional distress and quality of life in people with diabetes and their families**. *Gaceta Sanitaria* (2015.0) **29** 300-3. DOI: 10.1016/j.gaceta.2015.02.005
5. Samaniego RA, García I, Sánchez FM, Del Río ML, Esparza ÓA. **Coping and its relationship with quality of life in patients with type 2 diabetes mellitu**. *European Journal of Health Research* (2018.0) **4** 19-29. DOI: 10.30552/ejhr.v4i1.87
6. 6.Wild S Roglic G Green A Global Global prevalence of Diabetes estimates for the year 2000 and projections for 2030. Diabetes Care. (2004) 27:1047–53. 10.2337/diacare.27.5.104715111519. *Diabetes Care* (2004.0) **27** 1047-53. DOI: 10.2337/diacare.27.5.1047
7. Azzallini S, Vera BP, Vidal V, Benvenuto A, Ludmila F. **Depression and anxiety in patients with diabetes type 2 and its relationship with the coping strategies used for the adherence to treatments**. *Anuario de Investigaciones.* (2015.0) **22** 287-91
8. Torales J, Jara G, Ruiz C, Villalba J. *Aspectos Psicopatológicos en el Paciente con Diabetes*
9. Ferrer L, Kirchner T. **How do adolescents with Adjustment Disorder cope with stressful situations? Relationship with suicidal risk**. *Revista de psiquiatria y salud mental* (2020.0) **13** 63-72. DOI: 10.1016/j.rpsm.2018.11.002
10. Orozco-Gómez Á, Castiblanco-Orozco L. **Factores Psicosociales e Intervención Psicológica en Enfermedades Crónicas No transmisibles**. *Revista Colombiana de Psicol* (2015.0) **24** 203-17. DOI: 10.15446/rcp.v24n1.42949
11. Quiceno JM, Vinaccia S. **Resilience: a perspective from chronic disease in the adult population**. *Pensamiento Psicol.* (2011.0) **9** 69-82
12. Garrido-Hernansaiz H, Murphy P. *AIDS and Behavior* (2017.0) **21** 3260-70. DOI: 10.1007/s10461-017-1870-y
13. González I, Eduarda DS, Paiva L, Rossi LA, Dantas S, Alcalá D. **Anxiety, depression, resilience and self-esteem in individuals with cardiovascular diseases**. *Revista Latino-Americana de Enfermagem* (2016.0) **24** 2-10. DOI: 10.1590/1518-8345.1405.2836
14. Gheshlagh RG, Ebadi A, Dalvandi A, Rezaei M, Tabrizi KN. **A systematic study of resilience in patients with chronic physical diseases**. *Nurs Midwifery Stud* (2017.0) **6** 36401. DOI: 10.5812/nmsjournal.36401
15. Plascencia JC, Castellanos C. **Resilience assessment in mexicans diagnosed with hiv: a comparative study**. *Salud y Sociedad* (2019.0) **10** 52-64. DOI: 10.22199/S07187475.2019.0001.00004
16. Fernández-Álvarez N, Fontanil Y, Alcedo Á. **Resilience and associated factors in women survivors of Intimate Partner Violence: a systematic review**. *Anal. Psicol* (2022.0) **38** 1631. DOI: 10.6018/analesps.461631
17. Pesantes MA, Lazo-Porras M, Abu Dabrh AM, Ávila-Ramírez JR, Caycho M, Villamonte GY. **Resilience in vulnerable populations with type 2 diabetes mellitus and hypertension: a systematic review and meta-analysis.**. *The Canadian journal of cardiolog* (2015.0) **31** 3. DOI: 10.1016/j.cjca.2015.06.003
18. Pedraza GL, Vega CZ. **Stress, coping, emotions and therapeutic adherence in diabetic patients**. *Eureka.* (2018.0) **15** 173-85
19. Gong Y, Shi J, Ding H, Zhang M, Kang C, Wang K, Yu Y, Wei J, Wang S. **Personality traits and depressive symptoms: the moderating and mediating effects of resilience in Chinese adolescents**. *Journal of Affective Disorders* (2020.0) **265** 611-7. DOI: 10.1016/j.jad.2019.11.102
20. Morell-Mengual V, Ruiz-Palomino E, Giménez-García CJ, Castro-Calvo I. **The influence of personality in the perception of health care of Spanish youth**. *Int J Develop Edu Psycho* (2016.0) **2** 173-80. DOI: 10.17060/ijodaep.2016.n1.v2.199
21. Oshio A, Taku K, Hirano M, Saeed G. **Resilience and big five personality traits: a meta-analysis**. *Persona Individ Diff* (2018.0) **127** 54-60. DOI: 10.1016/j.paid.2018.01.048
22. Esmaeilinasab M, Ebrahimi M, Mokarrar MH, Rahmati L, Mahjouri MY, Arzaghi SM. **Type II diabetes and personality; a study to explore other psychosomatic aspects of diabetes**. *J Diab Metabol Disord* (2016.0) **15** 3. DOI: 10.1186/s40200-016-0281-3
23. Van Dooren FE, Denollet J, Verhey FR, Stehouwer CD, Sep SJ, Henry RM. **Psychological and personality factors in type 2 diabetes mellitus, presenting the rationale and exploratory results from The Maastricht Study, a population-based cohort study**. *BMC Psychiatry* (2016.0) **16** 22. DOI: 10.1186/s12888-016-0722-z
24. Richards, M. **Optimism and Resilience in Adolescents**. (2022.0) **7** 259. DOI: 10.32351/rca.v7.259
25. Ramírez-Fernández E, Ortega-Martínez AR, Calero-García MJ. **Optimism as a mediator between resilience and affective states in older adults**. *Estudios de Psicolog* (2018.0) **39** 1-19. DOI: 10.1080/02109395.2018.1486360
26. Iraegui A, Quevedo-Aguado MP. **Psycholinguistic approach to the study of personality in Spanish: A taxonomic proposal**. *Iberpsicolog* (2002.0) 7
27. Sandín B, Chorot B.. **The Coping Strategies Questionnaire: Development and preliminary validation**. *Revista de Piscopatoogía Piscologia Clínico* (2002.0) **8** 39-54. DOI: 10.5944/rppc.vol.8.num.1.2003.3941
28. Novella A. *Incremento De La Resiliencia Luego De La Aplicación De Un Programa De Psicoterapia Breve En Madres Adolescentes* (2002.0)
29. Kalisch R, Müller M, Tüscher O. **A conceptual framework for the neurobiological study of resilience**. *Behav Brain Sci* (2015.0) **38** 1-79. DOI: 10.1017/S0140525X1400082X
30. Linnemann P, Berger K. **Associations between outcome resilience and sociodemographic factors, childhood trauma, personality dimensions and self-rated health in middle-aged adults**. *Int J Behav Med* (2022.0) **29** 796-806. DOI: 10.1007/s12529-022-10061-1
31. Martínez-Martí M.L, Ruch W.. **Character strengths predict resilience over and above positive affect, self-efficacy, optimism, social support, self-esteem, and life satisfaction.**. *J Posit Psychol* (2016.0) **12** 1-10. DOI: 10.1080/17439760.2016.1163403
32. Vizoso C, Arias-Gundín O. **Relationship between resilience, optimism, and engagement in future educators**. *Int J Edu Res Innovat.* (2019.0) **11** 33-46
33. Genise G, Genise N, Gómez M, Humeniuk A, Muiños FJ. **Relationship Between Psychological Resilience and Personality Factors in Adolescent Population**. *Revista Latinoamericana de Ciencia Psicológica* (2018.0) **10** 1-17. DOI: 10.5872/psiencia/10.3.21
34. Holden CL. **Characteristics of Veterinary Students: Perfectionism, Personality Factors, and Resilience**. *Journal of Veterinary Medical Education ADVANCE* (2020.0) **3** e0918111r. DOI: 10.3138/jvme.0918-111r
35. Shi M, Liu L, Wang ZY, Wang L. **The mediating role of resilience in the relationship between big five personality and anxiety among Chinese medical students: a cross-sectional study**. *PLoS ONE* (2015.0) **10** 1-12. DOI: 10.1371/journal.pone.0119916
36. González N, Valdez JL. **Resilience and personality in adults**. *Revista Electrónica de Psicolog* (2011.0) **14** 295-316
37. Jalilianhasanpour R, Williams B, Gilman I, Burke MJ, Glass S, Fricchione G. **Resilience linked to personality dimensions, alexithymia and affective symptoms in motor functional neurological disorders**. *J Psycho Res* (2018.0) **107** 55-61. DOI: 10.1016/j.jpsychores.2018.02.005
38. Rudow D, Lacoviello B, Charney D. **Resilience and personality traits among living liver and kidney donors**. *Progress Transplant* (2014.0) **24** 82-90. DOI: 10.7182/pit2014448
39. Soriano J, Monsalve V. **Profiles of personality and resilience in chronic pain: utility of the CD-RISC-10 to discriminate between resilient and vulnerable types**. *Revista de La Sociedad Española Del Dolor* (2019.0) **26** 72-80. DOI: 10.20986/resed.2018.3670/2018
40. Hengartner MP, Linden D. **Van Der, Bohleber L, Von Wyl A. Big five personality traits and the general factor of personality as moderators of stress and coping reactions following an emergency alarm on a Swiss university campus**. *Stress and Health* (2016.0) **33** 34-44. DOI: 10.1002/smi.2671
41. Pereyra R. **Coping and stress within the framework of the five personality factors model**. *Review study PSocial.* (2017.0) **3** 39-45
42. Mirnics Z, Bagdy G, Surány Z, Gonda X. **The relationship between the big five personality dimensions and acute psychopathology: Mediating and moderating effects of coping strategies**. *Psychiatr Danub.* (2013.0) **25** 379-88. PMID: 24247050
43. Afshar H, Roohafza H, Keshteli AH, Mazaheri M. **The association of personality traits and coping styles according to stress level**. *Journal of Research in Medical Sciences.* (2015.0) **20** 353-8. PMID: 26109990
44. Leszko M, Iwansky R, Jerzebinska A. **The relationship between personality traits and coping styles among first-time and recurrent prisoners in Poland**. *Front Psychol* (2020.0) **10** 1-8. DOI: 10.3389/fpsyg.2019.02969
45. Novaes LE. **Stress, resiliência e apoio social em indivíduos com hipertensão e diabetes mellitus**. *Revista de Psicolog* (2019.0) **28** 1-13. DOI: 10.5354/0719-0581.2019.53954
46. Amar J, Martínez M, Utria L. **New approach to health considering the resilience**. *Salud Uninorte.* (2013.0) **29** 124-33
47. Theleritis C, Psarros C, Mantonakis L, Roukas D, Papaioannou A, Paparrigopoulos T. **Coping and its relation to PTSD in Greek firefighters**. *J Nerv Mental Dis* (2020.0) **208** 252-9. DOI: 10.1097/NMD.0000000000001103
48. Furukori-Yasui N, Murakami H, Otaka H, Nakayama H, Murabayashi M, Muzushiri S. **Coping behaviors and depressive status in individuals with type 2 diabetes mellitus**. *Annals General Psychiatry* (2019.0) **18** 1-8. DOI: 10.1186/s12991-019-0235-5
49. Sanjuán P, Ávila M. **Coping and motivation as predictors of subjective and psychological well-being**. *Revista de Psicopatolog* (2016.0) **21** 1-10. DOI: 10.5944/rppc.vol.21.num.1.2016.15401
50. Féki I, Turki M, Zitoun I, Sellami R, Baati I, Masmoudi J. **Dépression et stratégies de coping chez les sujets âgés atteints de diabète de type 2**. *L'Encephale* (2019.0) **45** 320-6. DOI: 10.1016/j.encep.2019.01.005
51. Jourdan YY. **Coping and quality of life in type 1 and 2 diabetic subjects from Argentina**. *Revista ALAD.* (2016.0) **6** 29-40
52. Muñoz-Alonzo HM, González-Aguilar D, Ponce ME, Samayoa V, Paniagua WO. **Coping and resilience in the context of oncological health care in Guatemala**. *Ciencias Sociales y Humanidades* (2018.0) **5** 9-18
|
---
title: Development and preliminary validation of Cancer-related Psychological Flexibility
Questionnaire
authors:
- Mei-jun Ou
- Xiang-hua Xu
- Hong Chen
- Fu-rong Chen
- Shuai Shen
journal: Frontiers in Psychology
year: 2023
pmcid: PMC10017436
doi: 10.3389/fpsyg.2023.1052726
license: CC BY 4.0
---
# Development and preliminary validation of Cancer-related Psychological Flexibility Questionnaire
## Abstract
The Cancer-related Psychological Flexibility Questionnaire (CPFQ) was developed and validated for assessing cancer patients’ psychological flexibility, including attitudes and behavior toward cancer. In a systematic process, the CPFQ identified four factors through principal component analysis and confirmatory factor analysis: Cancer Acceptance, Cancer Avoidance, Activity Engagement, and Valued Action. The results of this study reveal that the CPFQ has a clear factor structure and good psychometric properties. The specific nature of cancer and the need for a specific measure of cancer patient psychological flexibility make this questionnaire valuable for research on psychological flexibility in cancer patients.
## Introduction
Cancer is a leading cause of death and a worldwide public health issue. According to the GLOBOCAN 2020 estimates of cancer incidence and mortality worldwide in 185 countries, ~19.3 million new cancer cases and 10.0 million deaths occurred in 2020. Moreover, an estimated 28.4 million cases are expected in 2040 (Sung et al., 2021). Cancer is a chronic and life-threatening illness, and most cancer patients must undergo comprehensive anticancer treatment, including surgery, radiotherapy, and chemotherapy. Existing evidence showed that cancer patients often endure treatment-related toxicities and permanent functional impairment, which lead to multiple symptoms (Neufeld et al., 2017; Cuthbert et al., 2020; Gravier et al., 2020; Lage et al., 2020; Raphael et al., 2020). Cuthbert et al. reported that $60\%$ and $80\%$ of patients with a cancer diagnosis suffered from anxiety and low well-being, respectively (Cuthbert et al., 2020). Almost half of cancer survivors ($43.6\%$) experienced a fear of relapse, and $32.1\%$ had a severe/pathological fear of relapse in Singapore (Mahendran et al., 2021). A negative emotional state could significantly reduce the quality of life of cancer patients (Liu et al., 2021; Phoosuwan and Lundberg, 2022).
Many cancer patients adopt negative coping styles such as avoidance and pessimism when confronted with treatment-related toxicities, impaired functions, and distorted body image (Zhang et al., 2020). They might struggle to eliminate or fight cancer-related discomfort symptoms, including pain, fatigue, nausea, vomiting, dyspnea, and edema. Some live in this self-conceptualization and think they are useless and cumbersome. They experience a diminished sense of self-worth and show avoidance and withdrawal tendencies, such as avoiding discussing disease-related issues and reducing daily activities and social interactions. Coping styles of cancer survivors plays a predictive role in psychological symptoms, psychological well-being, and quality of life. The patients with an avoidance coping style showed higher cancer distress, anxiety, depression, and lower quality of life (Cheng et al., 2019). The better we understand the mechanism that underlies cancer patients’ avoidance coping style, the better we can reduce their psychological symptoms and improve their quality of life.
Psychological flexibility (PF), a new concept in clinical psychology, is defined as the ability to stay in contact with the present moment and pursue behavioral goals based on personal values and situational contexts, despite the presence of distress (Kashdan et al., 2020; Cherry et al., 2021). Psychological flexibility is a core construct of the Hexaflex model of acceptance and commitment therapy (ACT), one of the third wave of cognitive behavioral therapy (Hulbert-Williams et al., 2015). Acceptance and commitment therapy is based on Hexaflex model which is composed of six core components: acceptance, cognitive defusion, self as context, being present, values, and committed action. Acceptance and commitment therapy aims to improve the coping style and diminish the impact of adverse stressor events by deconstructing the individual experience in the context of personal values, enabling acceptance of both positive and negative responses (Hulbert-Williams et al., 2015), producing adaptive behavior change by enhancing PF (Hayes et al., 2019; Hofmann and Hayes, 2019). Many studies found that PF is associated with adaptive personality traits, including higher conscientiousness, openness to experiences, and lower negative emotionality, which may be the primary therapeutic mechanism of ACT (Hayes et al., 2006; Bryan et al., 2015). Previous studies indicated that a higher PF predicted lower anxiety, depression, and aversive emotional states in patients with breast cancer (Berrocal Montiel et al., 2016). A higher PF also resulted in a higher meaning in life in patients with thyroid cancer (Lv et al., 2021). In patients with prostate cancer, PF significantly predicted psychological distress and quality of life and moderated the relationship between the fear of recurrence and psychological distress (Sevier-Guy et al., 2021). Lucas et al. reported that PF was important for mental health and had a direct, positive effect on life satisfaction among community residents (Lucas and Moore, 2020). In conclusion, PF is a common protective factor across different contexts and populations. Therefore, quantitative assessment of PF for cancer patients can predict their psychological status and quality of life and evaluate the effect of ACT.
Psychological flexibility currently has a wide range of measurement tools. The most popular general measure of PF was AAQ-II, a version of the Acceptance and Action Questionnaire (AAQ; Cherry et al., 2021). AAQ-II has been widely used to measure PF in different contexts and populations, such as cancer patients (Lv et al., 2021) and community residents (Pyszkowska and Ronnlund, 2021). However, AAQ-II measures experiential avoidance (EA), which is an unwillingness to face unwanted experiences and acting to avoid them, and fails to capture core elements of PF (Kashdan et al., 2020). Experiential avoidance measured by AAQ-II is only a component of PF and cannot fully represent PF. In addition, there are many specific assessment tools for PF adapted from AAQ, such as the Acceptance and Action Diabetes Questionnaire (Gregg et al., 2007), the Chronic Pain Acceptance Questionnaire (CPAQ; Fish et al., 2010), and the Psychological Flexibility in Epilepsy Questionnaire (PFEQ; Burket et al., 2021), which are used to measure the PF of patients with diabetes mellitus, chronic pain, and epilepsy, respectively. However, there is no specialized assessment tool for PF in cancer patients.
Cancer is a chronic disease with long-term complex treatment and high physical and psychological burdens. Cancer is a life-threatening disease with high incidence, destruction of integrity, and high recurrence risk. Cancer patients are at risk for several comorbid psychological problems, such as anxiety, depression, and fear (Cuthbert et al., 2020; Lage et al., 2020; Raphael et al., 2020). Moreover, patients with cancer tend to confuse negative emotions and thoughts with objective facts, and immerse themselves in negative automatic thoughts, which aggravate negative emotions and form a vicious circle. Therefore, the PF of cancer patients may differ from that of patients with other non-cancer diseases owing to the characteristics of the tumor. The measure of PF in cancer patients is helpful in understanding their psychological process and coping style so that psychological interventions can be implemented to enhance PF, decrease psychosocial distress, and pursue a more meaningful and healthy life. In addition, it might reduce the specificity and sensitivity if general psychological flexibility assessment tools were used to measure PF of cancer patients. Hence, developing a self-reporting tool that specifically addresses PF in relation to cancer for both research and clinical purposes is necessary.
The present study aimed to [1] develop a tool to measure PF in cancer patients and identify its latent structure [Study 1], and [2] confirm the structure, and explore the validity of the questionnaire by using the Meaning in Life Questionnaire, the Templer’s Death Anxiety Scale, and the Acceptance and Action Questionnaire II [Study 2].
## Participants
Participants were recruited using a convenience sampling method from a tertiary cancer hospital in Hunan Province, China. Patients were included if they: (a) were aged over 18 years old, (b) had a diagnosis of cancer and awareness of it; (c) had normal cognitive function and were able to read and write; (d) could complete the survey; and (e) were willing to participate and provide informed consent. A total of 250 questionnaires were distributed. Finally, 231 patients completed the survey, with a valid response rate of $92.4\%$. Of all the patients, 115 were men, aged 19–89 years, with an average age of 56.4 ± 11.3 years. Most of them were married ($90\%$) ($$n = 208$$), $4.3\%$ ($$n = 10$$) were single, and $5.6\%$ ($$n = 13$$) were divorced or widowed. As for educational background, $27.7\%$ ($$n = 64$$) completed primary school or below, $42.8\%$ ($$n = 99$$) junior high school, $14.3\%$ ($$n = 33$$) senior high school, and $15.2\%$ ($$n = 35$$) college and above. Regarding cancer stages, $4.3\%$ ($$n = 10$$) of the patients had stage I, $24.2\%$ ($$n = 56$$) stage II, $39.0\%$ ($$n = 99$$) stage III, $15.6\%$ ($$n = 36$$) stage IV, and $16.9\%$ ($$n = 39$$) were unreported.
Inpatients were invited to participate in this survey from a tertiary cancer hospital in Hunan Province, China. The selection criteria of the participants were consistent with Study 1. A total of 285 questionnaires were sent to patients, and 252 patients completed the whole questionnaire, with a valid response rate of $88.4\%$. Of all the patients, 130 were men and were aged between 22 and 90 with an average age of 56.0 ± 11.3 years. Eighty six percent of patients ($$n = 217$$) were married, $5.2\%$ ($$n = 13$$) were single, and $8.7\%$ ($$n = 22$$) were divorced or widowed. As for educational background, $27.8\%$ ($$n = 70$$) completed primary school or below, $40.1\%$ ($$n = 101$$) junior high school, $17.5\%$ ($$n = 44$$) senior high school, and $14.7\%$ ($$n = 37$$) college and above. The cancer stages of the sample were as follows: $5.9\%$ ($$n = 15$$) patients had stage I, $25\%$ ($$n = 63$$) stage II, $42.5\%$ ($$n = 107$$) stage III, $13.9\%$ ($$n = 35$$) stage IV, and $12.7\%$ ($$n = 32$$) were not reported.
To validate the model, the subjects-to-parameters ratio could not be lower than 5:1, and the total number of subjects needed to be over 200. The new sample ($$n = 252$$) in Study 2 reached a 5.7:1 subjects-to-parameters ratio, which was appropriate for testing a model with 44 parameters, consisting of 19 factor loadings, 19 error variances, and six factor correlations.
## Item generation of the pilot Cancer-related Psychological Flexibility Questionnaire
There were four steps to generate the items of the pilot Cancer-related Psychological Flexibility Questionnaire. The four steps are item generation, scoring methodology, expert consultation, and pilot test, described as follows:
## Item generation
*The* generation of items was based mainly on the following principles. [ 1] It came from the analysis of the Hexaflex theoretical framework and a large number of literature reviews. According to the model, PF included two processes, which consist of six core components: mindfulness and acceptance processes (acceptance, cognitive defusion, and self as context) and commitment and behavior change processes (being present, values, and committed action). [ 2] It followed items of other measurement tools of psychological flexibility, such as the CPAQ (Fish et al., 2010) and the Multidimensional Psychological Flexibility Inventory (MPFI; Rolffs et al., 2018). [ 3] Semi-structured interviews with open-ended questions were conducted with a representative sample of cancer patients to understand their feelings and responses after a cancer diagnosis. The interview outline revised by experts was as follows: ① Please describe your experience or feeling about cancer; ② What influences or changes have tumors brought to your life, including daily life, work, social interaction, family relations, etc.?; ③ What do you do in the face of cancer?; ④What are your main concerns?; and ⑤ What are your plans for the future?. An experienced interviewer conducted one-to-one interviews in an independent and quiet room. The entire interview process was recorded. The interviewer transcribed and analyzed the interview results on the day of the interview and stopped the interview after sufficient information was gathered. Finally, 18 cancer patients were interviewed. Four themes were extracted: negative emotions (distress, shame, frustration, anxiety, and self-blame), avoidance coping (avoid disease, social isolation, workplace alienation, and meaningless life), positive coping (accept reality, cooperate with treatment, and cherish life), and future plan (adjust lifestyle, assume roles, and go with the flow).
If an item reflected one of the six core components mentioned above, it was included in the potential items pool. Following these guidelines, 32 potential items were created to reflect the construct of PF. After study group discussion, some similar items were deleted or merged, and 14 items were retained.
## Scoring methodology
A 5-point Likert-type scale that ranged from “never true” to “always true” was used. Most items were reverse scored, with “never true” score as 5 points and “always true” scored as 1 point. A few items were positively scored. The total score was the sum of all items, with higher points representing better cancer-related PF.
## Expert consultation
After creating the potential items pool, expert consultation was conducted by sending an email to assess the accuracy and importance of the items and proposing modification suggestions. We selected 15 psychology experts from the ACT field; however, 12 experts ended up being involved in the consultation. There were five men and seven women, with ages ranging from 30 to 55 years (with an average of 43.1 ± 8.6 years). They had been involved in psychology for at least 5 years, and most are currently active in the ACT field. Regarding academic qualifications, one was an undergraduate, and the rest had a master’s or doctorate degree. Each expert evaluated the items independently.
In the first round of consultation, the experts recommended we add some items about self as context and being present, and split some items with double meanings, so that the number of items increased to 23 after this consultation. We then conducted the second round of expert consultation. After this consultation, we adjusted the items appropriately, modified the ambiguous items, adjusted the order of the items, and selected the most representative items. For example, “Even if I feel ill, I can still live a normal life” changed to “Even if I feel ill, I still try to live like before I got sick,” and “I experience a lot of pain when I think about or feel certain things because of my tumor” changed to “I feel pain for suffering from a tumor.” Meanwhile, according to experts’ suggestions, we put together items that expressed the same concept. Finally, the initial questionnaire with 23 items was generated after two rounds of consultations.
## Pilot test
The pilot test took a sample of 15 inpatients from a tertiary cancer hospital, which was used to clarify ambiguous items, and delete items that were hard to understand or with multiple meanings. No incomprehension or ambiguity were discovered. As a result, a pilot questionnaire with 23 items was left.
## Procedure
Two master’s students from the research team distributed the survey face-to-face between November and December 2021. Before the survey, participants were informed about the purpose of this study, the requirements for participation, potential risks/benefits, and their right to terminate participation at any time. The researchers started the survey once informed consent was obtained. The survey was conducted anonymously, and participants participated in the survey free of charge.
## Data analysis
Data analysis was performed using the IBM SPSS software version 26.0 (IBM, Armonk, NY, United States). First, item-total correlations were used to test whether all items were consistent with the questionnaire. Inconsistent items were deleted based on the results. Second, the cases were divided into a high score group (the first $27\%$) and a low score group (the last $27\%$) according to the total score of the CPFQ, and then the scores of all items in the two groups were compared. Items with no significant differences indicating a lack of identification were deleted. Third, Kaiser-Meyer-Olkin and Bartlett’s test of sphericity was used to test whether the data were appropriate for factor analysis. Fourth, principal component analysis (PCA) was used to explore the latent structure of the CPFQ. The criteria for dimensions and item selection were as follows (Wu, 2010): [1] eigenvalues >1; [2] factors containing three or more items; [3] items load strongly (>0.40) onto factors; and [4] items do not cross-load onto two or more factors.
Data analysis was performed using Amos version 23.0, SPSS version 26.0, and Mplus version 8.3. The construct validity was identified by the confirmatory factor analysis (CFA), and the criteria for indexes that were used to assess the goodness of fit of the model as follows: 1 < χ2/df < 3, comparative fit index (CFI) > 0.90, goodness-of-fit index (GFI) > 0.90, and root-mean-square error of approximation (RMSEA) < 0.08 (Wu, 2010). The equivalence of the model across gender was examined by invariance testing, and the criterion for indices that were used to evaluate the adequacy of the fit of the model as follows: ΔCFI was <0.01, ΔRMSEA was <0.015 (Cheung and Rensvold, 2002).
Descriptive statistical analysis was used to examine the mean, standard variation, skewness, and kurtosis of the four factors. In order to assess the concurrent validity of the questionnaire, Pearson’s r between CPFQ, MIL, and T-DAF was calculated. In order to assess the convergent validity of the questionnaire, Pearson’s r between CPFQ and AAQ-II was calculated. Internal consistency of the CPFQ was examined using Cronbach’s alpha coefficient. Finally, split-half reliability R was evaluated by the correlation coefficient r between the odd and even items ($R = 2$r/1+r).
## Results and discussion
Based on the item analysis, the following four items were removed because the correlation coefficient (r) between the item and the total score was 0.233, 0.189, 0.254, and 0.280, respectively: Item 7: “The tumor made me realize what is important in life”; item 8: “We still live a happy life although we are in distress”; item 10: “I try not to think about the changes that cancer treatment may bring”; and item 16: “Even if I am ill, I still try to attend family, friends, or classmate gatherings.” *Item analysis* ranked the total scores of 231 patients from low to high and assigned the first $27\%$ as the low score group and the last $27\%$ as the high score group. The t-test of two independent samples was used to detect the differences between the 23 items in the high and low score groups. The results showed no significant difference on item 7 and item 8, which further indicated that the identification of these items was low and should be deleted.
In the Kaiser-Meyer-Olkin test, an r-value of 0.831 indicated that the data was suitable for factor analysis. A Bartlett test of sphericity (χ2 = 2879.375, df = 171, $p \leq 0.001$) indicated that the analysis model was appropriate. Therefore, it was acceptable to adopt factor analysis to test the construct reliability of this scale.
Applying PCA and varimax orthogonal rotation, we set parameters and extracted four factors with eigenvalues >1 and a cumulative variance interpretation rate of $68.939\%$. Four factors all contained at least four items, and the loading of each item was more than 0.59 (see Figure 1). According to the content of the items, the four factors were named as cancer acceptance (six items, $M = 20.08$, SD = 5.64, skewness = −0.262, kurtosis = −0.376), cancer avoidance (four items, $M = 10.31$, SD = 2.97, skewness = −0.048, kurtosis = −0.239), activity engagement (five items, $M = 18.05$, SD = 3.70, skewness = −0.059, kurtosis = −0.501), and valued action (four items, $M = 16.01$, SD = 2.63, skewness = −0.402, kurtosis =0.824). The items of the original Chinese form are shown in Appendix.
**Figure 1:** *Confirmatory factor analysis in Study 2, and rotated(promax) factor loading in Study 1 with principal component analysis.*
Following a series of data analyses, Study 1 resulted in a 19-item scale with four factors. This proposed model evaluated the PF of cancer patients, and its structure was inconsistent with the existing measuring tools for PF. For example, the CPAQ-8 contains two factors, that is pain willingness and activity engagement (Fish et al., 2010). The Personalized Psychological Flexibility Index includes three dimensions (avoidance, acceptance, and harnessing; Kashdan et al., 2020). This four-factor structure of the CPFQ was based on one sample; hence study 2 was conducted to validate the factor structure in another dataset.
## Study 2: Validation
To validate the four-factor structure and test the validity and reliability of the 19-item CPFQ, Study 2 collected another dataset. A CFA was conducted to test the four-factor model of PF. Furthermore, previous studies indicated a positive correlation between PF and life meaning but a negative relationship between anxiety and experiential avoidance (Lv et al., 2021). Therefore, meaning in life (assessed by the Meaning in Life Questionnaire) and death anxiety (assessed by the Templer’s Death Anxiety Scale) were used to evaluate the concurrent validity of the CPFQ, and experiential avoidance (assessed by the AAQ II) was used to evaluate the convergent validity of the CPFQ.
## Procedure and measures
The survey was distributed face-to-face by three master’s students from the research team between January and March 2022. Before the survey, participants were informed about the purpose of this study, the requirements of participation, potential risks/benefits, and their right to terminate participation at any time. The researchers started the survey once informed consent was obtained. In order to evaluate concurrent validity and convergent validity, patients were required to complete the Meaning in Life Questionnaire, the Templer’s Death Anxiety Scale, and the AAQ II.
## Meaning in life questionnaire
Meaning in life (MIL) was measured by the Meaning in Life Questionnaire (MLQ; Steger et al., 2006). The questionnaire contained the following two subscales: [1] The presence of meaning (MLQ-P), which assessed the extent to which meaning is experienced in a respondent’s life using statements such as “I understand my life’s meaning,” and [2] search for meaning (MLQ-S), which assessed a respondent’s desire to find and understand MIL using statements such as “I am searching for meaning in my life.” The original questionnaire had 10 items (five items for each of the two subscales) scored using a 7-point Likert scale ranging from one point (absolutely untrue) to seven points (absolutely true). Higher scores on the MLQ suggested that respondents were more likely to perceive and find MIL. Chinese scholars had previously translated and modified the questionnaire. The Chinese version, with five items for the MLQ-P and four items for the MLQ-S, was reported to have satisfactory reliability and validity (Liu and Gan, 2010). Finally, the MLQ was used to analyze the concurrent validity of the CPFQ.
## Templer’s death anxiety scale
The Templer’s Death Anxiety Scale (T-DAS; Templer, 1970) assessed death anxiety and was used to analyze the concurrent validity of the CPFQ. The scale consisted of 15 items with dichotomous responses (true/false). Nine items were forward scored, and six were reverse scored, and higher scores indicated greater death anxiety. This scale was reported to have test–retest reliability of 0.83 and reasonable internal consistency of 0.76. This scale had been translated into multiple languages and used in several countries. The Chinese version of T-DAS contained four factors. These were labeled [1] Stress and pain, [2] Emotion, [3] Cognitive, and [4] Awareness of Time Passing. The translated measure demonstrated good reliability and validity with an estimated internal consistency of Cronbach’s α = 0.71 (Yang et al., 2012).
## Acceptance and action questionnaire II
The Acceptance and Action Questionnaire II (AAQ-II; Bond et al., 2011) was a general measure of experiential avoidance and was used to analyze convergent validity with CPFQ. AAQ-II was developed by Bond et al. in 2011, a unidimensional scale with seven items based on the seven points Likert scale, ranging from one (never true) to seven (always true). The total score was summed over the seven items, with higher scores representing greater experiential avoidance and lower PF. AAQ-II had good test–retest reliability and good internal consistency (α = 0.88). The Chinese version of AAQ-II was modified by Cao et al. in 2013 (Cao et al., 2013), which had established a good content validity index, and acceptable internal consistency with Cronbach’s α = 0.88.
## Construct validity analysis
To obtain the construct validity of the four-factor structure developed from Study 1, CFA with the maximum likelihood method was conducted by Amos 23.0. The results showed a good fit to the data of Study 2, χ2 = 297.572, χ2/df = 2.343, $p \leq 0.001$, CFI = 0.948, GFI = 0.900, RMSEA = 0.073, $90\%$ CI = 0.062–0.084. Configural or factorial invariance analysis was conducted by Mplus 8.3 across gender group to determine whether the factor structure of the CPFQ was the same for both men and women. The results revealed the model fit the data reasonably well, with ΔCFI <0.01, and ΔRMSEA <0.015 (Table 1). Therefore, formal and measurement invariance across gender was evidenced for the CPFQ.
**Table 1**
| Invariance level | χ 2 | df | χ2/df | CFI | RMSEA | ΔCFI | ΔRMSEA |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Configural | 442.154 | 284 | 1.557 | 0.876 | 0.066 | | |
| Metric | 466.107 | 299 | 1.559 | 0.869 | 0.067 | −0.007 | 0.001 |
| Scalar | 491.209 | 314 | 1.564 | 0.861 | 0.067 | −0.008 | −0.001 |
The standardized coefficients of each path are shown in Figure 1. Descriptive analysis revealed that the distributions were relatively normal for the overall CPFQ ($M = 63.75$, SD = 10.65, skewness = 0.112, kurtosis = −0.016) and its four factors: cancer acceptance ($M = 19.80$, SD = 5.73, skewness = −0.320, kurtosis = −0.187), cancer avoidance ($M = 10.15$, SD = 3.31, skewness = 0.290, kurtosis = 0.110), activity engagement ($M = 17.70$, SD = 3.67, skewness = 0.029, kurtosis = −0.430), and valued action ($M = 16.10$, SD = 2.63, skewness = −0.352, kurtosis = 0.656).
## Concurrent and convergent validity analysis
Pearson’s correlational analysis was conducted to explore the association between the CPFQ and other measures. The overall CPFQ was significantly positively associated with the presence of meaning ($r = 0.519$), search for meaning ($r = 0.257$), and MIL ($r = 0.456$), and negatively associated with death anxiety (rs = −0.449 to −0.591), and experiential avoidance (r = −0.704), which provided evidence that the overall CPFQ was measuring the essence of psychological flexibility and could estimate mental and behavioral health of cancer patients. Table 2 shows that most correlation coefficients were significant, ranging from −0.159 to.747, except for Cancer Avoidance. Cancer Avoidance was weakly correlated with AAQ-II (r = −0.175, $p \leq 0.01$) and not correlated with other measures.
**Table 2**
| Unnamed: 0 | MLQ-P | MLQ-S | MLQ | DAS Stress and pain | T-DAS Emotion | T-DAS Cognitive | T-DAS Awareness of time passing | T-DAS | AAQ-II |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Mean | 24.62 | 19.10 | 43.71 | 2.83 | 2.08 | 1.67 | 1.03 | 7.60 | 25.56 |
| SD | 4.681 | 4.227 | 7.707 | 1.673 | 1.718 | 1.041 | 0.820 | 4.332 | 9.641 |
| r with | | | | | | | | | |
| Cancer Acceptance | 0.397** | 0.084 | 0.287** | −0.538** | −0.513** | −0.574** | −0.465** | −0.637** | −0.747** |
| Cancer Avoidance | 0.116 | 0.012 | 0.077 | −0.067 | −0.056 | −0.039 | −0.010 | −0.059 | −0.175** |
| Activity Engagement | 0.496** | 0.323** | 0.478** | −0.405** | −0.291** | −0.395** | −0.409** | −0.444** | −0.561** |
| Valued Action | 0.399** | 0.392** | 0.457** | −0.333** | −0.254** | −0.159* | −0.219** | −0.309** | −0.219** |
| Overall CPFQ | 0.519** | 0.257** | 0.456** | −0.532** | −0.457** | −0.496** | −0.449** | −0.591** | −0.704** |
## Reliability analysis
The Cronbach’s α coefficient of the whole CPFQ was 0.885, and the Cronbach’s α coefficient for Cancer Acceptance, Cancer Avoidance, Activity Engagement, and Valued Action was 0.927, 0.874, 0.823, and 0.849, respectively, which indicated that the items were internally consistent. The odd and even items were summed and the correlation coefficient was statistically significant with $r = 0.898$, and split-half reliability was 0.946.
## Effects of gender, age, and cancer stage on CPFQ
Table 3 shows the effects of gender, age, and cancer stage on the overall CPFQ and its four dimensions. Independent t-tests showed a significant difference in Valued Action (t = −2.590, $p \leq 0.05$) and no significant difference in the overall CPFQ and the other three dimensions between men and women. In terms of cancer stage, ANOVA displayed statistical differences in the overall CPFQ and its three dimensions (F > 5.789, $p \leq 0.005$), except for Cancer Avoidance. Table 3 shows that age was related to Activity Engagement, Valued Action, and the overall CPFQ (F > 2.784, $p \leq 0.018$).
**Table 3**
| Variables | n | Cancer Acceptance | Cancer Avoidance | Activity Engagement | Valued Action | Overall CPFQ |
| --- | --- | --- | --- | --- | --- | --- |
| Gender | Gender | Gender | Gender | Gender | Gender | Gender |
| Male | 130 | 19.35 (5.36) | 10.26 (3.34) | 17.63 (3.80) | 15.69 (2.57) | 62.94 (10.59) |
| Female | 122 | 20.27 (6.08) | 10.03 (3.30) | 17.77 (3.55) | 16.54 (2.63) | 64.61 (10.68) |
| t | | −1.266 | 0.547 | −0.301 | −2.590 | −1.250 |
| Value of p | | 0.207 | 0.585 | 0.764 | 0.010 | 0.212 |
| Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) |
| 20–30 | 3 | 16.33 (7.37) | 7.67 (4.04) | 19.00 (5.29) | 18.00 (3.46) | 61.00 (12.29) |
| 31–40 | 29 | 19.24 (6.58) | 10.93 (3.47) | 19.00 (3.55) | 16.69 (2.58) | 65.86 (12.13) |
| 41–50 | 42 | 20.33 (6.36) | 10.05 (3.87) | 18.69 (4.13) | 17.12 (2.87) | 66.19 (11.49) |
| 51–60 | 94 | 20.66 (4.62) | 10.47 (3.58) | 17.27 (3.38) | 15.78 (2.52) | 64.17 (9.32) |
| 61–70 | 65 | 19.52 (6.14) | 9.74 (2.51) | 17.92 (3.42) | 15.82 (2.49) | 63.00 (10.41) |
| >70 | 19 | 16.68 (5.45) | 9.42 (2.43) | 14.68 (3.16) | 15.26 (2.45) | 56.05 (10.61) |
| F | | 1.963 | 1.231 | 4.590 | 2.784 | 2.894 |
| Value of p | | 0.085 | 0.295 | 0.001 | 0.018 | 0.015 |
| Cancer stage | Cancer stage | Cancer stage | Cancer stage | Cancer stage | Cancer stage | Cancer stage |
| I | 15 | 24.07 (4.27) | 10.27 (3.67) | 22.13 (3.11) | 19.40 (1.40) | 75.87 (7.92) |
| II | 63 | 20.30 (4.48) | 10.48 (3.64) | 18.51 (2.84) | 15.83 (2.57) | 65.11 (9.39) |
| III | 107 | 19.65 (5.56) | 9.98 (2.94) | 17.39 (3.67) | 16.01 (2.39) | 63.04 (9.80) |
| IV | 35 | 17.20 (7.23) | 9.80 (3.28) | 16.11 (3.80) | 15.57 (2.51) | 58.69 (12.78) |
| F | | 5.789 | 0.438 | 12.126 | 10.257 | 10.664 |
| Value of p | | 0.001 | 0.726 | <0.001 | <0.001 | <0.001 |
Overall, the 19-item CPFQ with four dimensions showed a good model fit in a second Chinese sample. The results demonstrated good reliability, construct validity, concurrent validity, and convergent validity of the CPFQ. The overall CPFQ and its dimensions were positively correlated with MIL, and negatively correlated with death anxiety and experiential avoidance. However, Cancer Avoidance showed non-significant correlations with MIL and death anxiety. This outcome may be due to cancer patients’ characteristics, whose attitudes toward cancer change over time. Future studies are needed to continue to validate the four-factor structure of the CPFQ in a bigger sample.
This study suggested women showed higher Valued Action. In China, most women undertake more roles and responsibilities than men, as they take care of many people. Even if they do not accept cancer, avoid cancer issues, and avoid socializing, they still do something based on their values, such as taking the doctor’s advice to live longer.
It is worth noting that age was correlated with the CPFQ, especially for activity engagement and valued action. The younger the patient was, the more willing they were to participate in activities and do something worthwhile. However, in terms of the overall CPFQ, patients aged 31–50 show higher PF. The possible reason is that middle-aged patients are mentally more mature than younger patients and more responsible than older people. Therefore, considering the small sample size of some subgroups, future studies are necessary to explore this interesting phenomenon further.
Furthermore, the current study also revealed that patients with advanced cancer had lower PF than those with early cancer. It may be because early cancer is easier to treat, has a better prognosis, and is easier to recovery. Patients acquire posttraumatic growth after this life event and tend to cherish work, life, families, and friends more. It is interesting that irrespective of the cancer stage, they all had an attitude of avoiding cancer. One possible explanation for this phenomenon is that according to Chinese traditional culture, Chinese people are extremely sensitive about their illness and consider cancer a disgrace.
In this study, we recruited patients with different types of cancer, but we did not analyze the impact of cancer types on PF. Because there were many types of cancer and the sample size was not relatively small, cancer was difficult to classify by types. If the classification is very specific, the sample size of each category will be very small, and if classified by tumor region, the severity of diseases in the same category will vary greatly, e.g., head and neck cancer includes oral and thyroid cancer; however, the severity of these diseases is completely different. Although the extent of the impact of different types of cancer on PF is unknown, it clearly has an impact. Because different types of cancer have different symptoms and prognoses, symptoms and prognoses will affect PF.
## General discussion
The current study describes the development and preliminary validation of the CPFQ, an instrument to measure PF in cancer patients. Initial scale development resulted in a 23-item instrument, which was reduced to 19 items based on item-scale correlations and PCA. After item analysis and PCA, a four-factor structure of the CPFQ indicated four dimensions of psychological flexibility of cancer patients: Cancer Acceptance, Cancer Avoidance, Activity Engagement, and Valued Action. The PCA revealed a four-dimension questionnaire consistent with the concept of PF (Cherry et al., 2021). Confirmatory factor analysis indicated a good model fit on the four-factor structure; in other words, the construct validity was satisfactory. Concurrent validity was expressed as correlations between CPFQ, MIL, and death anxiety (T-DAS) were moderate. Convergent validity, as these constructs were supposed to share some common features, expressed as the correlation between CPFQ and AAQ-II was acceptable. The internal consistency and split-half reliability were beyond the specified standard. The results showed that the CPFQ had a clear factor structure and good psychometric properties in Chinese samples. Therefore, the questionnaire is valuable and beneficial for research on the PF of cancer patients.
The CPFQ reflects both attitudes and behaviors toward cancer. The four dimensions of the CPFQ represented PF in terms of cancer acceptance, cancer avoidance, social contact, and behavior orientation. The ability to live a valuable life despite a cancer diagnosis is a type of PF related to cancer. Different from other life events, cancer is a life-threatening disease, and individuals’ responses should be different from other stress events. Therefore, the PF of cancer patients may have its own essence and characteristics. Examining the four specific dimensions of the CPFQ, the former two dimensions mainly assessed the patients’ attitudes toward cancer, and the latter two dimensions mainly measured their behavioral tendencies after a cancer diagnosis. These contents reflected not only the nature of PF, such as acceptance, cognitive fusion, being present, values, and action, but also the characteristics of cancer patients.
Differing from other measures of psychological flexibility, the CPFQ measures [1] individuals’ psychological response to cancer and their attitudes toward the psychological response; [2] one’s emotional and behavioral tendencies when thinking of cancer treatment; and [3] individuals’ social interaction and behavior change after cancer. Compared with other questionnaires for measuring PF, the CPFQ has similarities and differences. For example, the CPAQ includes two dimensions, namely pain willingness (feeling little need to avoid or control painful experiences) and activity engagement (the degree to which one engages in life’s activities regardless of pain; McCracken et al., 2004; Fish et al., 2010). The similarities are that both questionnaires measure patients’ attitudes and behaviors toward diseases (cancer vs. chronic pain). The difference is that the content of CPFQ is more comprehensive, including not only psychological responses and behavioral orientation to diseases, but also avoidance reactions and valued actions to diseases.
The CPFQ contributes significantly by providing a valuable tool that measures components of psychological flexibility and verifies psychological flexibility from cancer patients’ perspectives. As described in the “Materials and Methods” section, the items of the CPFQ were generated from both the literature review and the theoretical definition of psychological flexibility. We also refer to some items from other measurement instruments of psychological flexibility, such as the CPAQ (Fish et al., 2010) and MPFI (Rolffs et al., 2018). Moreover, we interviewed cancer patients about their feelings, attitudes, and behavior change after a cancer diagnosis. Hence, we support that the CPFQ is a questionnaire with a solid theoretical foundation and comprehensive content.
In summary, a new measurement instrument of PF was developed and validated in two samples. To our knowledge, this is the first cancer-specific psychological flexibility measurement that includes attitude and behavior components, which could provide a more accurate assessment of PF among cancer patients and help health care providers develop personalized and targeted interventions in the PF field. The CPFQ was a reliable and valid tool to evaluate the PF of cancer patients with a four-factor structure: Cancer Acceptance, Cancer Avoidance, Activity Engagement, and Valued Action. Moreover, this questionnaire has a good readability and a reasonable length with 19 items. We believe that the CPFQ can be used as a valuable measurement in the psychological flexibility field.
## Limitations and future directions
The present study forms a preliminary version of the CPFQ. However, there are still some limitations. First, there may be sampling bias. The samples of Study 1 and Study 2 were from the same hospital, and it would be important to verify the reliability and validity of the current questionnaire among different groups. Future studies could apply this questionnaire to other groups, such as cancer patients from a general hospital, to validate our results. Second, one dimension of the questionnaire, namely Cancer Avoidance, showed unsatisfactory validity values, which needs further exploration in future research. Finally, our study did not evaluate test–retest reliability, because most inpatients had been discharged at the time of the retest, and there may exist a deviation between the online questionnaire survey and the face-to-face survey. When applying the questionnaire in the future, a small sample (such as 30) of cancer patients could be selected for the re-test reliability test.
## Conclusion
The CPFQ includes four dimensions: Cancer Acceptance, Cancer Avoidance, Activity Engagement, and Values Actions, and it was proven to be a reliable and valid measure of psychological flexibility in cancer patients.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
M-jO, X-hX, and SS contributed to the original idea and concepts. HC and F-rC completed data collection and analysis. M-jO wrote the first draft. X-hX and SS revised the manuscript. All authors approved the final version of manuscript for submission.
## Funding
This work was supported by grants from Hunan Provincial Natural Science Foundation of China (grant No. 2021JJ40327 and 2018JJ6110).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Berrocal Montiel C., Rivas Moya T., Venditti F., Bernini O.. **On the contribution of psychological flexibility to predict adjustment to breast cancer**. *Psicothema* (2016) **28** 266-271. DOI: 10.7334/psicothema2015.271
2. Bond F. W., Hayes S. C., Baer R. A., Carpenter K. M., Guenole N., Orcutt H. K.. **Preliminary psychometric properties of the acceptance and action questionnaire–II: a revised measure of psychological inflexibility and experiential avoidance**. *Behav. Ther.* (2011) **42** 676-688. DOI: 10.1016/j.beth.2011.03.007
3. Bryan C. J., Ray-Sannerud B., Heron E. A.. **Psychological flexibility as a dimension of resilience for posttraumatic stress, depression, and risk for suicidal ideation among air force personnel**. *J. Contextual Behav. Sci.* (2015) **4** 263-268. DOI: 10.1016/j.jcbs.2015.10.002
4. Burket L., Parling T., Jansson-Frojmark M., Reuterskiold L., Ahlqvist J., Shanavazh S.. **Development and preliminary evaluation of the psychometric properties of the psychological flexibility in epilepsy questionnaire (PFEQ)**. *Epilepsy Behav.* (2021) **115** 107685. DOI: 10.1016/j.yebeh.2020.107685
5. Cao J., Ji Y., Zhu Z. H.. **Reliability and validity of the Chinese version of the acceptance and action questionnaire-second edition (AAQ-II) in college students**. *Chin. Ment. Health J.* (2013) **27** 873-877. DOI: 10.3969/j.issn.1000-6729.2013.11.014
6. Cheng C. T., Ho S. M. Y., Liu W. K., Hou Y. C., Lim L. C., Gao S. Y.. **Cancer-coping profile predicts long-term psychological functions and quality of life in cancer survivors**. *Support Care Cancer* (2019) **27** 933-941. DOI: 10.1007/s00520-018-4382-z
7. Cherry K. M., Hoeven E. V., Patterson T. S., Lumley M. N.. **Defining and measuring "psychological flexibility": a narrative scoping review of diverse flexibility and rigidity constructs and perspectives**. *Clin. Psychol. Rev.* (2021) **84** 101973. DOI: 10.1016/j.cpr.2021.101973
8. Cheung G. W., Rensvold R. B.. **Evaluating goodness-of-fit indexes for testing measurement invariance**. *Struct. Equ. Modeling* (2002) **9** 233-255. DOI: 10.1207/S15328007SEM0902_5
9. Cuthbert C. A., Boyne D. J., Yuan X., Hemmelgarn B. R., Cheung W. Y.. **Patient-reported symptom burden and supportive care needs at cancer diagnosis: a retrospective cohort study**. *Support Care Cancer* (2020) **28** 5889-5899. DOI: 10.1007/s00520-020-05415-y
10. Fish R. A., McGuire B., Hogan M., Morrison T. G., Stewart I.. **Validation of the chronic pain acceptance questionnaire (CPAQ) in an internet sample and development and preliminary validation of the CPAQ-8**. *Pain* (2010) **149** 435-443. DOI: 10.1016/j.pain.2009.12.016
11. Gravier A. L., Shamieh O., Paiva C. E., Perez-Cruz P. E., Muckaden M. A., Park M.. **Meaning in life in patients with advanced cancer: a multinational study**. *Support Care Cancer* (2020) **28** 3927-3934. DOI: 10.1007/s00520-019-05239-5
12. Gregg J. A., Callaghan G. M., Hayes S. C., Glenn-Lawson J. L.. **Improving diabetes self-management through acceptance, mindfulness, and values: a randomized controlled trial**. *J. Consult. Clin. Psychol.* (2007) **75** 336-343. DOI: 10.1037/0022-006X.75.2.336
13. Hayes S. C., Hofmann S. G., Stanton C. E., Carpenter J. K., Sanford B. T., Curtiss J. E.. **The role of the individual in the coming era of process-based therapy**. *Behav. Res. Ther.* (2019) **117** 40-53. DOI: 10.1016/j.brat.2018.10.005
14. Hayes S. C., Luoma J. B., Bond F. W., Masuda A., Lillis J.. **Acceptance and commitment therapy: model, processes and outcomes**. *Behav. Res. Ther.* (2006) **44** 1-25. DOI: 10.1016/j.brat.2005.06.006
15. Hofmann S. G., Hayes S. C.. **The future of intervention science: process-based therapy**. *Clin. Psychol. Sci.* (2019) **7** 37-50. DOI: 10.1177/2167702618772296
16. Hulbert-Williams N. J., Storey L., Wilson K. G.. **Psychological interventions for patients with cancer: psychological flexibility and the potential utility of acceptance and commitment therapy**. *Eur. J. Cancer Care* (2015) **24** 15-27. DOI: 10.1111/ecc.12223
17. Kashdan T. B., Disabato D. J., Goodman F. R., Doorley J. D., McKnight P. E.. **Understanding psychological flexibility: a multimethod exploration of pursuing valued goals despite the presence of distress**. *Psychol. Assess.* (2020) **32** 829-850. DOI: 10.1037/pas0000834
18. Lage D. E., El-Jawahri A., Fuh C. X., Newcomb R. A., Jackson V. A., Ryan D. P.. **Functional impairment, symptom burden, and clinical outcomes among hospitalized patients with advanced cancer**. *J. Natl. Compr. Cancer Netw.* (2020) **18** 747-754. DOI: 10.6004/jnccn.2019.7385
19. Liu S. S., Gan Y. Q.. **Reliability and validity of the Chinese version of the meaning in life questionnaire**. *Chin. Ment. Health J.* (2010) **24** 5. DOI: 10.3969/j.issn.1000-6729.2010.06.021
20. Liu Y. J., Schandl A., Markar S., Johar A., Lagergren P.. **Psychological distress and health-related quality of life up to 2 years after oesophageal cancer surgery: nationwide population-based study**. *BJS Open* (2021) **5** zraa038. DOI: 10.1093/bjsopen/zraa038
21. Lucas J. J., Moore K. A.. **Psychological flexibility: positive implications for mental health and life satisfaction**. *Health Promot. Int.* (2020) **35** 312-320. DOI: 10.1093/heapro/daz036
22. Lv J., Zhu L., Wu X., Yue H., Cui X.. **Study on the correlation between postoperative mental flexibility, negative emotions, and quality of life in patients with thyroid cancer**. *Gland Surg.* (2021) **10** 2471-2476. DOI: 10.21037/gs-21-424
23. Mahendran R., Liu J., Kuparasundram S., Simard S., Chan Y. H., Kua E. H.. **Fear of cancer recurrence among cancer survivors in Singapore**. *Singap. Med. J.* (2021) **62** 305-310. DOI: 10.11622/smedj.2020007
24. McCracken L. M., Vowles K. E., Eccleston C.. **Acceptance of chronic pain: component analysis and a revised assessment method**. *Pain* (2004) **107** 159-166. DOI: 10.1016/j.pain.2003.10.012
25. Neufeld N. J., Elnahal S. M., Alvarez R. H.. **Cancer pain: a review of epidemiology, clinical quality and value impact**. *Future Oncol.* (2017) **13** 833-841. DOI: 10.2217/fon-2016-0423
26. Phoosuwan N., Lundberg P. C.. **Psychological distress and health-related quality of life among women with breast cancer: a descriptive cross-sectional study**. *Support Care Cancer* (2022) **30** 3177-3186. DOI: 10.1007/s00520-021-06763-z
27. Pyszkowska A., Ronnlund M.. **Psychological flexibility and self-compassion as predictors of well-being: mediating role of a balanced time perspective**. *Front. Psychol.* (2021) **12** 671746. DOI: 10.3389/fpsyg.2021.671746
28. Raphael D., Frey R., Gott M.. **Distress in post-treatment hematological cancer survivors: prevalence and predictors**. *J. Psychosoc. Oncol.* (2020) **38** 328-342. DOI: 10.1080/07347332.2019.1679320
29. Rolffs J. L., Rogge R. D., Wilson K. G.. **Disentangling components of flexibility via the Hexaflex model: development and validation of the multidimensional psychological flexibility inventory (MPFI)**. *Assessment* (2018) **25** 458-482. DOI: 10.1177/1073191116645905
30. Sevier-Guy L. J., Ferreira N., Somerville C., Gillanders D.. **Psychological flexibility and fear of recurrence in prostate cancer**. *Eur. J. Cancer Care (Engl)* (2021) **30** e13483. DOI: 10.1111/ecc.13483
31. Steger M. F., Frazier P., Oishi S., Kaler M.. **The meaning in life questionnaire: assessing the presence of and search for meaning in life**. *J. Couns. Psychol.* (2006) **53** 80-93. DOI: 10.1037/0022-0167.53.1.80
32. Sung H., Ferlay J., Siegel R. L., Laversanne M., Soerjomataram I., Jemal A.. **Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA Cancer J. Clin.* (2021) **71** 209-249. DOI: 10.3322/caac.21660
33. Templer D. I.. **The construction and validation of a death anxiety scale**. *J. Gen. Psychol.* (1970) **82** 165-177. DOI: 10.1080/00221309.1970.9920634
34. Wu M. L.. *Practice of questionnaire statistical analysis: operation and application of SPSS* (2010)
35. Yang H., Han L. S., Guo H. M.. **Cultural adjustment and application of the death anxiety scale**. *Chinese J. Pract. Nurs.* (2012) **28** 5. DOI: 10.3760/cma.j.issn.1672-7088.2012.31.028
36. Zhang L., Lu Y., Qin Y., Xue J., Chen Y.. **Post-traumatic growth and related factors among 1221 Chinese cancer survivors**. *Psychooncology* (2020) **29** 413-422. DOI: 10.1002/pon.527
|
---
title: Suppression of SMOC2 alleviates myocardial fibrosis via the ILK/p38 pathway
authors:
- Huang Rui
- Fang Zhao
- Lei Yuhua
- Jiang Hong
journal: Frontiers in Cardiovascular Medicine
year: 2023
pmcid: PMC10017443
doi: 10.3389/fcvm.2022.951704
license: CC BY 4.0
---
# Suppression of SMOC2 alleviates myocardial fibrosis via the ILK/p38 pathway
## Abstract
### Background
Fibrosis of the myocardium is one of the main pathological changes of adverse cardiac remodeling, which is associated with unsatisfactory outcomes in patients with heart disease. Further investigations into the precise molecular mechanisms of cardiac fibrosis are urgently required to seek alternative therapeutic strategies for individuals suffering from heart failure. SMOC2 has been shown to be essential to exert key pathophysiological roles in various physiological processes in vivo, possibly contributing to the pathogenesis of fibrosis. A study investigating the relationship between SMOC2 and myocardial fibrosis has yet to be conducted.
### Methods
Mice received a continuous ISO injection subcutaneously to induce cardiac fibrosis, and down-regulation of SMOC2 was achieved by adeno-associated virus-9 (AAV9)-mediated shRNA knockdown. Neonatal fibroblasts were separated and cultured in vitro with TGFβ to trigger fibrosis and infected with either sh-SMOC2 or sh-RNA as a control. The role and mechanisms of SMOC2 in myocardial fibrosis were further examined and analyzed.
### Results
SMOC2 knockdown partially reversed cardiac functional impairment and cardiac fibrosis in vivo after 21 consecutive days of ISO injection. We further demonstrated that targeting SMOC2 expression effectively slowed down the trans-differentiation and collagen deposition of cardiac fibroblasts stimulated by TGFβ. Mechanistically, targeting SMOC2 expression inhibited the induction of ILK and p38 in vivo and in vitro, and ILK overexpression increased p38 phosphorylation activity and compromised the protective effects of sh-SMOC2-mediated cardiac fibrosis.
### Conclusion
Therapeutic SMOC2 silencing alleviated cardiac fibrosis through inhibition of the ILK/p38 signaling, providing a preventative and control strategy for cardiac remodeling management in clinical practice.
## Introduction
Adverse cardiac remodeling, defined as changes in left ventricular (LV) anatomical structure and function, is a common complication developed in the diseased heart as a result of an array of intra- and extra cardiac pathophysiological conditions and is correlated with a poor outcome and shortened survival in hospitalized individuals presenting with heart conditions (1–3). Implications from substantial evidence suggest that abnormal hypertrophy of existing cardiomyocytes and perpetual activation of cardiac fibroblasts are fundamental pathomechanisms of heart failure [4]. Previous studies have demonstrated the involvement of AKT and MAPK pathways in myocardial remodeling and myocardial fibrosis, which has been widely recognized [5, 6]. However, fibrosis of the myocardium is a complex pathological process with a multifactorial etiology, and further investigation into the specific in-depth molecular mechanisms is still imperative for improving prognoses [4, 7].
SMOC2, SPARC-related modular calcium-binding protein 2, which encodes a secreted modular protein containing a pair of FS domains and an EC domain and predominantly resides in the kidneys, lungs, myocardium, skeletal muscles, and ovaries, forms part of the BM-40/SPARC/osteonectin protein family as SMOC1 [8, 9]. As a result of binding to cell surface receptors, cytokines, and proteases, SMOC2 regulates the cell-matrix interaction, facilitating ischemic myocardial microvascular regeneration and tumor cell growth (8, 10–12). SMOC2 is implicated in some preclinical studies as contributing to the regulation of fibrotic disorders (13–16). Study results previously suggested that SMOC2 could stimulate renal fibroblast-to-myofibroblast conversion or transition, facilitate extracellular matrix synthesis, and ultimately result in kidney fibrosis [13]. As evidence for this finding, Xin et al. [ 14] disclosed that silencing SMOC2 could affect renal inflammation, fibrosis, and function in chronic kidney disease (CKD) mice. A subsequent study discovered that SMOC2 knockout was available to mitigate bleomycin-induced pulmonary fibrosis [15]. In addition, a network analysis of expression profile data has identified that SMOC2 may be implicated in the pathogenesis of heart failure [17], which may allude to a possible connection between SMOC2 and cardiac fibrosis despite the lack of studies demonstrating this link.
Studies have shown that SMOC2 exerts its physiological effects in vivo in part by transducing integrin-linked kinase (ILK)-mediated intracellular signaling cascades [18, 19]. SMOC2 is also involved in activating ILK directly, enabling it to propel the cell cycle forward [20]. Overexpression of ILK increases the collagen type I expression in cardiac fibroblasts by activating nuclear factor-κB (NF-κB) while silencing ILK resulted in the reverse effect [21]. Additionally, p38-MAPK, a generally accepted pathway involving fibrosis, can also be activated by ILK (22–24). Researchers found that ILK directly activated P38 MAPK to influence osteoblading effects, which could be reversed if ILK was targeted to be silenced [24].
As yet, it is unclear if ILK contributes to cardiac fibrosis by stimulating p38. Consequently, we designed and conceived this study to investigate the effect of SMOC2 on cardiac fibrosis and to understand the detailed mechanisms involved.
## Reagents
Transforming growth factor-β (TGF-β, 763104) was purchased from Biolegend (San Diego, CA, USA). Isoprenaline (ISO, I5627) was obtained from Sigma-Aldrich and prepared in DD water. The SMOC2 (SC-376104, 1:500 for western blot and 1:100 for satining) and ILK (SC-20019, 1:500) antibodies were obtained from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA). Antibodies against collagen type I (Col I, 1:1000), collagen type III (Col III, 1:1000), phospho-JNK (Tyr185,1:1000), JNK (1:2000), phospho-p38 MAPK (Thr180/Tyr182,1:1000), t-p38 (1:1000), phospho-AKT (1:5000), t-AKT (1:5000) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH,1:4000) were obtained from Proteintech (Wuhan, China). Anti-α smooth muscle actin (a-SMA, 1:100 for staining) antibody was purchased from Abcam (Cambridge, UK).
## Animals and treatments
All animal experimental procedures were carried out under the supervision of the Animal Care and Use Committee of Renmin Hospital of Wuhan University and in compliance with established guidelines published by the National Institutes of Health of United States. Eight-week-old male C57BL/6 mice (body weight 23.5 ± 2 g) were provided by the Experimental Animal Center of the Three Gorges University (Hubei, China) for this study, and were acclimated to the experimental scenarios for 5–7 days before they were used. Under standard environmental conditions, all animals were fed a pellet diet and had access to water without restrictions. Mice were subcutaneously injected with ISO (10 mg/kg for 3 days and 5 mg/kg for 11 days) to induce cardiac fibrosis (25–27). For the control group, an equal volume of normal saline solution was administered. Four weeks before ISO injection, mice received a single intravenous injection of adeno-associated virus 9 (AAV9) carrying small hairpin RNA against SMOC2 (sh-SMOC2) and its corresponding negative control (shRNA) generated by Obio technology (Shanghai, China) to verify its role in vivo. Animals were randomly assigned to four groups: Control + sh-RNA, Control + sh-SMOC2, ISO + sh-RNA, and ISO + sh-SMOC2. Animals were sacrificed after an excessive inhalation of CO2, having their hearts and tibias harvested and measured to calculate the heart weight (HW)/body weight (BW) and HW/tibia length (TL).
## Echocardiography
Following modeling, echocardiographic evaluations were conducted under light anesthesia with a Vinno 6th ultrasound device with Doppler imaging (VINNO6, Vinno Corporation, China). At the level of the left ventricular short-axis papillary muscle, M-mode tracking was used to measure the following parameters: left ventricular end-systolic diameter (LVEDs), left ventricular end-diastolic diameter (LVEDd), left ventricular posterior wall thickness (LVPW), interventricular septal thickness (IST), ejection fraction (EF), and left fractional ventricular shortening (FS).
## Cell culture and treatment
A mixture of trypsin and collagenase II was used to extract CFs from the hearts of 1–3 days old mice. CFs were cultured in a medium with $10\%$ FBS (fetal bovine serum) at 37°C with $5\%$ CO2 after removing non-adherent cells after 1 h. Non-adherent cells were removed after 1 h, and the adherent CFs were cultured in DMEM/F12 medium containing $10\%$ FBS (fetal bovine serum) at 37°C with $5\%$ CO2. The following experiments were conducted with 2–3 passages of CFs. Following 8 h of serum-free DMEM/F12 culture, cells were stimulated with TGFβ. In order to knock down the SMOC2, CFs were transfected with short hairpin (sh)RNA adenovirus against SMOC2 performed by Shanghai Jikai Gene Chemical Technology Co., Ltd. As a negative control, adenovirus expressing short hairpin (sh)GFP was used. To confirm the role of ILK, CFs were transfected with adenovirus carrying ILK (Ad-ILK) or green fluorescent protein (Ad-GFP).
## Determination of collagen contents
Commercial ELISA kits (collagen I and collagen III ELISA kits from Uscnlife, Wuhan, China) were used to measure collagen I and III content in the cell culture supernatant after cardiac fibroblasts were incubated with or without TGFβ.
## Histological analysis and immunofluorescence
Heart tissue samples were fixed in $4\%$ paraformaldehyde and embedded in paraffin before being cut into 5-micron sections. Heart tissue morphology and cardiac fibrosis were assessed with hematoxylin and eosin (HE) and Masson staining, respectively. Quantitative analysis was performed using Image J software.
CFs were stained with immunofluorescence in a well-cultured condition. Briefly, cell permeabilization was enhanced after fixation with $4\%$ paraformaldehyde for 20 min, followed by incubation with 50–100 μl of disruption buffer for 10 min at room temperature. Cell coverslips were incubated with anti-α-SMA and anti-SMOC2 overnight at 4°C. The next day, cells were incubated with goat anti-mouse (SMOC2) pre-adsorbed or goat anti-rabbit (α-SMA) pre-adsorbed secondary antibodies for 1 h. Nuclei were subsequently stained with 4′,6-diamino-2-phenylindole (DAPI) and images were taken with a fluorescence microscope.
## Western blot and RT-PCR
The heart tissues or cells were prepared to extract total proteins with RIPA lysis buffer (Servicebio, Wuhan, China). The extracts were resolved via 8 and $12\%$ sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to PVDF membranes (Millipore Corp. Bedford, MA, USA) using standard protocols. And then, the membranes were incubated with skim milk at room temperature for 1 h. Membranes were incubated with specific primary antibodies at 4°C overnight, followed by bio-tinylated secondary antibodies (goat anti-mouse HRP and goat anti-rabbit HRP) for 2 h at room temperature. The membranes were washed with TBST three times between each step mentioned above. Finally, the blots were detected by a hypersensitive ECL reagent (Biosharp, Beijing, China).
Total RNA was extracted using TRIzol reagent (Servicebio, Wuhan, China), and a cDNA synthesis kit was carried out to synthesize cDNA following the manufacturer’s protocol. Real-time PCR was carried out using EnTurboTM SYBR Green PCR SuperMix Kit (ELK Biotechnology). The mRNA levels were normalized to GAPDH. Molecular biological tests were conducted by individuals who were unaware of the treatment conditions of the animals and cells.
## Statistics analysis
SPSS 26.0 software was used for analysis. All results are presented as mean ± standard error of the mean, and t-test was used for comparison between two groups, and one-way ANOVA was used for three or more groups, followed by post L-S-D test. Differences were considered statistically significant with $p \leq 0.05.$
## SMOC2 was highly expressed in ISO-induced mouse model
We first investigated whether SMOC2 expression was altered in ISO-induced mouse model. Mice were treated daily with ISO or an equal volume of saline as control for 21 days as described. It was found that SMOC2 protein levels increased markedly in the ISO group, as was the accumulation of collagen types I and III (Figures 1A–D). Notably, RT-PCR further revealed increased expression level of SMOC2 in the ISO group (Figure 1E). The above observations strongly implicate that SMOC2 may play a critical role in regulating ISO-induced myocardial remodeling.
**FIGURE 1:** *SMOC2 expression was increased in ISO-induced mice hearts. (A–D) Western blot image and quantitative results of the SMOC2, collagen I, and collagen III in the control group and the ISO group (n = 6). (E) RT-PCR analysis of SMOC2 mRNA levels in the two groups (n = 6). *p < 0.05.*
## SMOC2 knockdown partially reversed the myocardial function loss and myocardial fibrosis in response to ISO stimulation in vivo
We next sought to explore the specific effects of SMOC2 in ISO-induced cardiac fibrosis. Considering the up-regulation of SMOC2 expression in the ISO-induced mouse model, we constructed a SMOC2-knockdown AAV9 system to infect mice as previously described.
As shown in Figures 2A–C, hearts of animals administered sh-SMOC2 showed a significant decrease in SMOC2 mRNA and protein expression after 4 weeks compared to hearts of mice treated with sh-RNA. Mice infected with sh-RNA subjected to ISO injection for 3 weeks developed cardiac hypofunction, manifested as a decrease in EF, FS, and elevation in LVEDd, LVEDs, HW/TL, and HW/BW (Figures 2D–J). Notably, cardiac function was partially reversed in mice injected with sh-SMOC2 (Figures 2D–J). Neither ISO administration nor infection with sh-SMOC2 affected ventricular wall thickness, however (Figures 2K, L).
**FIGURE 2:** *SMOC2 knockdown partially reversed the myocardial function loss. (A,B) SMOC2 expression after sh-RNA or sh-SMOC2 injection in mice heart (n = 6). (C) mRNA levels of SMOC2 after sh-RNA or sh-SMOC2 injection in mice heart (n = 6). (D) Representative images of echocardiographic measurement of cardiac function in mice. (E–L) Echocardiographic measurements of ejection fraction (EF), fractional shortening (FS), left ventricular internal diameter at end-systole (LVIDs), left ventricular internal diameter at end-diastole (LVIDd), HW/TL, HW/BW, interventricular septal thickness at diastole (IVS), and left ventricular posterior wall thickness at diastole (n = 8). *p < 0.05 vs. the control + sh-RNA group. #p < 0.05 vs. the ISO + sh-RNA group. n.s., non-significant.*
In addition, morphological examination revealed that mice experienced more severe myofibrillar disarrangement and fibrosis after 3 weeks of ISO injection (Figures 3A, B). Moreover, these histological changes weakened after SMOC2 deficiency (Figures 3A, B). To substantiate this further, we detected biomarkers of myocardial fibrosis. As expected, collagen I and collagen III protein and mRNA levels were down-regulated post-sh-SMOC2 injection (Figures 3C–G). These findings implied an underlying physiopathological impact of SMOC2 in modulating cardiac fibrosis, and SMOC2 knockdown in mice can partially reverse the myocardial function reduction and myocardial fibrosis in response to ISO stimulation in vivo.
**FIGURE 3:** *SMOC2 knockdown partially reversed the myocardial fibrosis in response to ISO Stimulation in vivo. (A,B) Representative images of HE and Masson staining and quantified analysis (n = 6). (C–E) Representative images of Western blots and the quantitative data (n = 6). (F,G) The relative mRNA levels of Collagen I, Collagen III (n = 3). *p < 0.05 vs. the control + sh-RNA group. #p < 0.05 vs. the ISO + sh-RNA group. n.s., non-significant.*
## Targeting SMOC2 expression attenuates myocardial fibrosis via the ILK and p38 pathway
Next, we went on to achieve further insight into the potential mechanisms accounting for our previous findings. Pre-existing research has identified that SMOC2 could activate downstream ILK [20, 28], which was reported to be involved in the fibrotic pathological process [21]. Consequently, we focused on the expression of ILK. It was revealed that, compared with the control group, the ISO + shRNA injection group had a significantly higher ILK level, and, importantly, the effect of ILK production can be relieved by SMOC2 knockdown (Figures 4A, B). Given that the AKT and AMPK pathways have been implicated in the pathophysiology of fibrosis, we examined whether these pathways are involved in the process of SMOC2-medicated myocardial fibrosis. It was found that p-AKT, p-p38-MAPK, and p-JNK-MAPK were significantly elevated in the ISO + sh-RNA group, while the p-p38 was decreased in the ISO + sh-SMOC2 group (Figures 4A, D). Strikingly, knocking down SMOC2 did not affect the expression of p-JNK and p-AKT (Figures 4C, E and Supplementary Figures 1A–F). Collectively, these findings demonstrate that the ILK and p38 pathway is likely to participate in the ISO-induced cardiac fibrotic process in vivo.
**FIGURE 4:** *Targeting SMOC2 expression attenuates myocardial fibrosis via the ILK/p38 pathway. (A–E) Representative Western blots and quantitative results of ILK, t-AKT, p-AKT, t-JNK, p-JNK, t-p38, and p-p38 (n = 6). *p < 0.05. n.s., non-significant.*
## SMOC2 is up-regulated in TGFβ-induced cardiac fibroblasts
To gain further insight into the regulation of SMOC2 in cardiac fibrosis, we conducted in vitro experiments subsequently. Neonatal myocardial fibroblasts were separated by a differential adherent as described to explore the role of SMOC2 in TGFβ-induced fibroblast activation. The results showed TGFβ induced fibroblasts activation in a dose-dependent and time-dependent manner. After being treated with TGF β (10 ng/ml) for 24 h, the protein concentrations of SMOC2 reached the highest (Figures 5A–F). Consequently, the cells were treated with TGFβ at 10 ng/ml for 24 h before being harvested in subsequent analysis.
**FIGURE 5:** *SMOC2 is up-regulated in TGFβ-induced cardiac fibroblasts. (A,B) Representative Western blots and quantitative results of SMOC2 in cardiac fibroblasts induced by different concentrations of TGFβ (n = 6). (C) mRNA levels of SMOC2 in cardiac fibroblasts induced by different concentrations of TGFβ (n = 3). (D,E) Representative Western blots and quantitative results of SMOC2 after induction of TGFβ at different times (n = 6). (F) mRNA levels of SMOC2 in cardiac fibroblasts after induction of TGFβ at different times (n = 3). *p < 0.05. n.s., non-significant.*
## Knocking SMOC2 improves the proliferation, migration, differentiation and exacerbates fibrosis of fibroblasts in vitro
In order to discuss the potential effects and mechanisms underlying the pro-fibrotic effect of SMOC2, we infected CFs with sh-SMOC2 or sh-RNA as control. As shown in Figures 6A–C, after being treated with TGFβ for 24 h, SMOC2 protein and mRNA levels increased. However, these changes could be relieved post sh-SMOC2 infection. Furthermore, these findings were confirmed by an immunofluorescence stain for SMOC2 (Figures 6D, F).
**FIGURE 6:** *Targeting SMOC2 expression improves the proliferation, migration, and differentiation and exacerbates fibrosis of fibroblasts in vitro. (A,B) SMOC2 expression after sh-RNA or sh-SMOC2 infection (n = 6). (C) mRNA levels of SMOC2 after sh-RNA or sh-SMOC2 injection (n = 6). (D–G) Immunofluorescence staining of SMOC2 and α-SMA for the indicated groups (n = 6). (H,I) Representative images of the wound scratch assay at 0 and 24 h (n = 5). (J–N) Collagen I and collagen III expressions after sh-RNA or sh-SMOC2 infection in cardiac fibroblasts (n = 6). *p < 0.05 vs. the control + sh-RNA group. #p < 0.05 vs. the TGF β + sh-RNA group. n.s., non-significant.*
The up-regulation of α-SMA expression, as a hallmark marker for differentiated myofibroblasts, was examined using immunofluorescence staining. As anticipated, the level of α-SMA was elevated in TGFβ + shRNA group, while SMOC2 down-regulation reduced the level of α-SMA (Figures 6E, G). Further analysis revealed that knocking SMOC2 could decrease the proliferation of TGFβ-induced CFs, as indicated by wound-healing scratch experiments (Figures 6H, I). Further analysis implicated that the protein and mRNA levels of collagen I and collagen III increased in the TGFβ + shRNA group, while knockdown of SMOC2 expression significantly inhibited type I and type III collagen production (Figures 6J–N).
Following that, we measured the collagen production of fibroblasts in the media. In agreement with the results above, the protein levels of fibrotic biomarkers such as type I and Type III collagen in the medium were elevated in the TGFβ group, whereas SMOC2 knockdown could relieve these changes (Supplementary Figures 2A, B).
## Targeting SMOC2 expression ameliorates fibrosis via the ILK/p38 pathway in vitro
In order to explore the mechanisms of anti-fibrotic effects of sh-SMOC2, we also detected ILK, AKT, MAPK pathways. Consistent with the findings in vivo, the ILK level was elevated after being treated with TGFβ, which could be down-regulated by sh-SMOC2 administration (Figures 7A, B). In-depth studies also found that p-AKT, P-JNK, and p-P38 were significantly up-regulated, while the sh-SMOC2 infection significantly reduced p-p38 without affecting the phosphorylation levels of p-AKT and p-JNK (Figures 7A, C–E and Supplementary Figures 3A–F).
**FIGURE 7:** *Targeting SMOC2 expression ameliorates fibrosis via the ILK/p38 pathway in vitro. (A–E) Representative Western blots and quantitative results of ILK, t-AKT, p-AKT, t-JNK, p-JNK, t-p38, and p-p38 (n = 6). *p < 0.05 vs. the control + sh-RNA group. #p < 0.05 vs. the ISO + sh-RNA group. n.s., non-significant. (F) Representative images of the wound scratch assay post Ad-ILK infection. (G) Representative images of immunohistochemical staining for α-SMA (n = 4). (H–M) Representative Western blots and quantitative results post Ad-ILK infection (n = 3). *p < 0.05. n.s., non-significant. #p < 0.05 vs. the sh-RNA + Ad-GFP group. ##p < 0.05 vs. the sh-SMOC2 + Ad-GFP group.*
Next, we also verified whether ILK participated in the activation of p-p38 in CFs in vitro, and we infected CFs with Ad-ILK to overexpress ILK. In line with the above findings, the anti-fibrotic effect of sh-SMOC2 was knocked over after ILK overexpression, and the scratch experiment also suggested that ILK further promoted the migration of CFS and activated p-p38 (Figure 7F). To further confirm this notion, the expression of α-SMA by IHC staining was examined. The results also identified that Ad-ILK worsened the transformation of CFs (Figure 7G). Results of in vitro studies matched those of in vivo studies. The level of SMOC2 did not change due to overexpression of ILK (Figures 7H, I), but ILK, p-p38, collagen I and collagen III were significantly up-regulated (Figures 7H, J, K–M and Supplementary Figures 4A–D).
## Discussion
To the best of our knowledge, this is the first report on the relationship between SMOC2 and myocardial fibrosis. Our study implied that the therapeutic silencing of SMOC2 can improve ISO-induced myocardial fibrosis and heart function reduction in vivo, and ameliorate proliferation and collagen deposition of CFs induced by TGFβ in vitro, and the down-regulation of ILK/p38 may be the underlying mechanism.
In addition to being associated with unfavorable outcomes, progressive cardiac fibrosis is one of the key hallmarks of heart failure [3, 29]. Several etiopathogenic mechanisms contribute to myocardial fibrosis, a complex and multifactorial condition [5, 6]. At present, the treatment of myocardial fibrosis is still an intractable clinical problem that demands to be further addressed [30, 31]. Currently, only a few pharmacotherapy options, such as treatments targeting the increased activity of the renin-angiotensin-aldosterone system (RAAS) and sympathetic nervous system (SNS), are available to manage cardiac remodeling and fibrosis [31, 32]. Despite exhibiting excellent efficacy, in some situations, the clinical application of these drugs has been restricted due to side effects or intolerance among patients [31]. A better understanding of the molecular mechanisms of cardiac fibrosis is thus essential for developing novel therapeutic approaches.
SMOC2 is an encoded modular secretion protein that is known to influence cytokine activity, destabilize cell-substrate attachment, and modulate cell differentiation and cell cycle [8, 13, 16, 33, 34]. A recently published study indicates that SMOC2 is an independent prognostic marker in patients with colorectal cancers [35]. SMOC2 was shown to be a promising biomarker of kidney fibrosis in patients with CKD and was able to be used by researchers as a prognostic indicator [36]. Previous studies have also demonstrated that SMOC2 is highly expressed upon kidney injury or in CKD models and stimulates the production of extracellular matrix, which can be ameliorated by infecting sh-RNA targeting SMOC2 [14, 16, 36]. In addition, SMOC2 also participates in pulmonary fibrosis and hepatic steatosis caused by a high-fat diet targeting TGFβ1 pathways [13, 15]. However, no research has examined the potential relationship between SMOC2 and cardiac fibrosis, nor have the relevant mechanisms been explored in more depth. The present study confirms for the first time that SMOC2 may contribute to the pathogenic process of cardiac fibrosis, which could be rescued at least partially by the deletion of SMOC2 expression in vivo and vitro.
ILK, a serine-threonine-protein kinase that exerts essential effects on cell transduction and molecular scaffolding, is known as an in vivo target of SMOC2 [18, 19, 37]. Previously accumulation of evidence identified that ILK was associated with some cardiomyopathy phenotypes, as well as involved in cardiac remodeling [37]. It was revealed that ILK adaptively increased in the pressure overloaded cardiac hypertrophy model and could further participate in the activation of CFs (21, 37–39). Overexpressed ILK promotes the cardiac fibrotic process by up-regulation of nuclear factor-κB (NF-κB) in cardiac fibroblasts that activate fibrotic-related genes such as CTGF and collagen I. On the other hand, ILK knockdown by siRNA could lead to attenuation of this fibrotic action [21]. Therefore, it is estimated that ILK may contribute to pathological cardiac remodeling. In line with previous studies, we found that ILK was overactivated adaptively in sustained profibrogenic factors-induced mice fibrosis model to respond to the up-regulation of SMOC2.
However, other findings provided discrepant observations. ILK may be one of the key determinants that contributed to the protective effect of myocardial re-modeling after MI during Cardiac Shock Wave Therapy [40]. What’s more, ILK overexpression also minimized cardiac remodeling and improved reperfusion after ischemia [41]. And these were in agreement with the corresponding findings of another study [42]. Researchers have recently published a study showing that targeting the regulatory ILK and relevant pathways ameliorates cardiac function in dilated myocardial animals induced by DOX, strongly implicating increased expression of ILK may ameliorate the prognosis in patients with HF [43]. A mouse model of the ILK knockout was also shown to alter myocardial electrical properties, potentially resulting in fatal arrhythmias [37]. Taken together, these studies suggest an intricate relationship between ILK and cardiac fibrosis and remodeling, and future research is still desperately warranted to shed light on this issue.
Evidence is mounting that the MAPK pathway, such as p38 and JNK$\frac{1}{2}$, functions in regulating cardiac apoptosis, hypertrophy, and fibrosis (44–47). And p38 was thought to be a dominant regulator in myocardial fibrosis [48]. Additionally, it is interesting to note that several studies have identified the AKT pathway as being associated with fibrosis of the heart and collagen synthesis [5]. The results of our present study are consistent with the previous observations, showing that AKT, p38, and JNK$\frac{1}{2}$ MAPKs were slightly activated following the pro-fibrotic stimuli in vivo and vitro. However, unexpectedly, SMOC2 knockdown impaired the expression of p38 without affecting AKT or JNK$\frac{1}{2}$ pathways, revealing p38 as a contributor to the development of SMOC2-mediated cardiac fibrosis. It was well-established that p38 phosphorylation in CFs could drive myofibroblasts to synthesize and formate collagen, which could be reversed when p38 was inhibited [6, 45]. P38 has been implicated in a variety of pathogenetic processes: such as inflammation [49], osteogenic differentiation [50], apoptosis [51], and cancer differentiation [52]. Specifically, p38 could be activated by several regulators thus exerting its physiological effects as a result. Zhang et al. [ 6] discovered that p38 was mediated by ribosomal protein S5 (RPS5) in press overload-induced cardiac fibrosis mice, whereas matrine administration alleviated this process. Additionally, Meijles et al. [ 53] revealed that p38 was regulated by apoptosis signal-regulating kinase 1 (AKS1) during the process of fibrosis and subsequent deterioration of hypertensive heart disease.
Interestingly, in a study, scholars found that ILK/p38 MAPK pathway can regulate the osteogenic effect on the surface of the a-C coated titanium [24]. They found that ILK was significantly up-regulated on the surface of the C-Ti, and siRNA targeting ILK reduced p38 phosphorylation and osteogenic differentiation in the C-Ti surface, which shed light on the research into the relationship between ILK and p38 in the present study. As anticipated, overexpression of ILK in vivo or vitro would result in increased phosphorylation of p38 MAPK accompanied by elevated markers of cardiac fibrosis.
## Conclusion and limitations
Our study identified that SMOC2 is involved in collagen deposition, cardiac fibroblasts activation, and, ultimately, cardiac fibrosis for the first time in response to pro-fibrotic stimuli such as ISO and TGFβ and ILK/p38 signaling pathway may be responsible for this important process (Figure 8). Furthermore, therapeutic SMOC2 silencing has protective effects against cardiac fibrosis by inhibiting ILK/p38 signaling, validating the critical role of SMOC2 in mediating cardiac fibrosis, and providing promising prevention and control strategies for cardiac remodeling management in clinical practice.
**FIGURE 8:** *The proposed working model of targeting SMOC2 expression on pathological cardiac fibrosis.*
There are several limitations to this study, despite its promising results. Firstly, we only applied the cardiac fibrosis model induced by ISO, and did not evaluate other cardiac fibrosis models. Secondly, a human validation of the findings was not performed. Moreover, an important limitation to this study is the use of neonatal cardiac fibroblasts. Finally, this study did not address the role of SMOC2 in cardiomyocytes and its role in cardiac hypertrophy remains to be defined. Therefore, the results of the study should be interpreted with caution.
## Data availability statement
The original contributions presented in this study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The animal study was reviewed and approved by the Animal Care and Use Committee of Renmin Hospital of Wuhan University.
## Author contributions
HR and FZ performed the experiments and analyzed the data. HR, LY, and JH designed the experiments, supervised and conceptualized the study, and wrote and edited the manuscript. HR and FZ wrote and edited the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcvm.2022.951704/full#supplementary-material
## References
1. Kwong RY, Bax JJ. **Unraveling the complex processes of adverse cardiac remodeling.**. (2019) **12**
2. Curley D, Lavin Plaza B, Shah AM, Botnar RM. **Molecular imaging of cardiac remodelling after myocardial infarction.**. (2018) **113**
3. McLaughlin S, McNeill B, Podrebarac J, Hosoyama K, Sedlakova V, Cron G. **Injectable human recombinant collagen matrices limit adverse remodeling and improve cardiac function after myocardial infarction.**. (2019) **10**. DOI: 10.1038/s41467-019-12748-8
4. Frangogiannis NG. **Cardiac fibrosis.**. (2021) **117** 1450-88. PMID: 33135058
5. Ma ZG, Yuan YP, Zhang X, Xu SC, Wang SS, Tang QZ. **Piperine attenuates pathological cardiac fibrosis via PPAR-γ/AKT pathways.**. (2017) **18** 179-87. DOI: 10.1016/j.ebiom.2017.03.021
6. Zhang X, Hu C, Zhang N, Wei WY, Li LL, Wu HM. **Matrine attenuates pathological cardiac fibrosis via RPS5/p38 in mice.**. (2021) **42** 573-84. DOI: 10.1038/s41401-020-0473-8
7. Filippatos G, Angermann CE, Cleland JGF, Lam CSP, Dahlström U, Dickstein K. **Global differences in characteristics, precipitants, and initial management of patients presenting with acute heart failure.**. (2020) **5** 401-10. PMID: 31913404
8. Maier S, Paulsson M, Hartmann U. **The widely expressed extracellular matrix protein SMOC-2 promotes keratinocyte attachment and migration.**. (2008) **314** 2477-87. DOI: 10.1016/j.yexcr.2008.05.020
9. Vannahme C, Gösling S, Paulsson M, Maurer P, Hartmann U. **Characterization of SMOC-2, a modular extracellular calcium-binding protein.**. (2003) **373** 805-14. PMID: 12741954
10. Rocnik EF, Liu P, Sato K, Walsh K, Vaziri C. **The novel SPARC family member SMOC-2 potentiates angiogenic growth factor activity.**. (2006) **281** 22855-64. DOI: 10.1074/jbc.M513463200
11. Wong GS, Rustgi AK. **Matricellular proteins: priming the tumour microenvironment for cancer development and metastasis.**. (2013) **108** 755-61. DOI: 10.1038/bjc.2012.592
12. Schellings MW, Pinto YM, Heymans S. **Matricellular proteins in the heart: possible role during stress and remodeling.**. (2004) **64** 24-31. DOI: 10.1016/j.cardiores.2004.06.006
13. Yuting Y, Lifeng F, Qiwei H. **Secreted modular calcium-binding protein 2 promotes high fat diet (HFD)-induced hepatic steatosis through enhancing lipid deposition, fibrosis and inflammation via targeting TGF-β1.**. (2019) **509** 48-55. DOI: 10.1016/j.bbrc.2018.12.006
14. Xin C, Lei J, Wang Q, Yin Y, Yang X, Moran Guerrero JA. **Therapeutic silencing of SMOC2 prevents kidney function loss in mouse model of chronic kidney disease.**. (2021) **24**. DOI: 10.1016/j.isci.2021.103193
15. Luo L, Wang CC, Song XP, Wang HM, Zhou H, Sun Y. **Suppression of SMOC2 reduces bleomycin (BLM)-induced pulmonary fibrosis by inhibition of TGF-β1/SMADs pathway.**. (2018) **105** 841-7. DOI: 10.1016/j.biopha.2018.03.058
16. Gerarduzzi C, Kumar RK, Trivedi P, Ajay AK, Iyer A, Boswell S. **Silencing SMOC2 ameliorates kidney fibrosis by inhibiting fibroblast to myofibroblast transformation.**. (2017) **2**. DOI: 10.1172/jci.insight.90299
17. Li D, Lin H, Li L. **Multiple feature selection strategies identified novel cardiac gene expression signature for heart failure.**. (2020) **11**. DOI: 10.3389/fphys.2020.604241
18. Shi Q, Bao S, Song L, Wu Q, Bigner DD, Hjelmeland AB. **Targeting SPARC expression decreases glioma cellular survival and invasion associated with reduced activities of FAK and ILK kinases.**. (2007) **26** 4084-94. DOI: 10.1038/sj.onc.1210181
19. Barker TH, Baneyx G, Cardó-Vila M, Workman GA, Weaver M, Menon PM. **SPARC regulates extracellular matrix organization through its modulation of integrin-linked kinase activity.**. (2005) **280** 36483-93. DOI: 10.1074/jbc.M504663200
20. Liu P, Lu J, Cardoso WV, Vaziri C. **The SPARC-related factor SMOC-2 promotes growth factor-induced cyclin D1 expression and DNA synthesis via integrin-linked kinase.**. (2008) **19** 248-61. DOI: 10.1091/mbc.e07-05-0510
21. Thakur S, Li L, Gupta S. **NF-κB-mediated integrin-linked kinase regulation in angiotensin II-induced pro-fibrotic process in cardiac fibroblasts.**. (2014) **107** 68-75. DOI: 10.1016/j.lfs.2014.04.030
22. Ying YT, Ren WJ, Tan X, Yang J, Liu R, Du AF. **Annexin A2-mediated internalization of staphylococcus aureus into bovine mammary epithelial cells requires its interaction with clumping factor B.**. (2021) **9**. DOI: 10.3390/microorganisms9102090
23. Wang W, Liu Q, Zhang Y, Zhao L. **Involvement of ILK/ERK1/2 and ILK/p38 pathways in mediating the enhanced osteoblast differentiation by micro/nanotopography.**. (2014) **10** 3705-15. DOI: 10.1016/j.actbio.2014.04.019
24. Yue G, Song W, Xu S, Sun Y, Wang Z. **Role of ILK/p38 pathway in mediating the enhanced osteogenic differentiation of bone marrow mesenchymal stem cells on amorphous carbon coating.**. (2019) **7** 975-84. DOI: 10.1039/c8bm01151f
25. Jiang XH, Wu QQ, Xiao Y, Yuan Y, Yang Z, Bian ZY. **Evodiamine prevents isoproterenol-induced cardiac fibrosis by regulating endothelial-to-mesenchymal transition.**. (2017) **83** 761-9. DOI: 10.1055/s-0042-124044
26. Li N, Zhou H, Ma ZG, Zhu JX, Liu C, Song P. **Geniposide alleviates isoproterenol-induced cardiac fibrosis partially via SIRT1 activation in vivo and in vitro.**. (2018) **9**. DOI: 10.3389/fphar.2018.00854
27. Li C, Ying S, Wu X, Zhu T, Zhou Q, Zhang Y. **CTRP1 aggravates cardiac fibrosis by regulating the NOX2/P38 pathway in macrophages**. (2022) **24** 732-40. DOI: 10.22074/cellj.2022.557327.1043
28. Liu P, Pazin DE, Merson RR, Albrecht KH, Vaziri C. **The developmentally-regulated Smoc2 gene is repressed by Aryl-hydrocarbon receptor (Ahr) signaling.**. (2009) **433** 72-80. DOI: 10.1016/j.gene.2008.12.010
29. Aminzadeh MA, Tseliou E, Sun B, Cheng K, Malliaras K, Makkar RR. **Therapeutic efficacy of cardiosphere-derived cells in a transgenic mouse model of non-ischaemic dilated cardiomyopathy.**. (2015) **36** 751-62. DOI: 10.1093/eurheartj/ehu196
30. González A, Schelbert EB, Díez J, Butler J. **Myocardial interstitial fibrosis in heart failure: biological and translational perspectives.**. (2018) **71** 1696-706. PMID: 29650126
31. Park S, Nguyen NB, Pezhouman A, Ardehali R. **Cardiac fibrosis: potential therapeutic targets.**. (2019) **209** 121-37. PMID: 30930180
32. AlQudah M, Hale TM, Czubryt MP. **Targeting the renin-angiotensin-aldosterone system in fibrosis.**. (2020) **9** 92-108
33. Long F, Shi H, Li P, Guo S, Ma Y, Wei S. **A SMOC2 variant inhibits BMP signaling by competitively binding to BMPR1B and causes growth plate defects.**. (2021) **142**. DOI: 10.1016/j.bone.2020.115686
34. Peeters T, Monteagudo S, Tylzanowski P, Luyten FP, Lories R, Cailotto F. **SMOC2 inhibits calcification of osteoprogenitor and endothelial cells.**. (2018) **13**. DOI: 10.1371/journal.pone.0198104
35. Jang BG, Kim HS, Bae JM, Kim WH, Kim HU, Kang GH. **SMOC2, an intestinal stem cell marker, is an independent prognostic marker associated with better survival in colorectal cancers.**. (2020) **10**. DOI: 10.1038/s41598-020-71643-1
36. Schmidt IM, Colona MR, Kestenbaum BR, Alexopoulos LG, Palsson R, Srivastava A. **Cadherin-11, Sparc-related modular calcium binding protein-2, and Pigment epithelium-derived factor are promising non-invasive biomarkers of kidney fibrosis.**. (2021) **100** 672-83. DOI: 10.1016/j.kint.2021.04.037
37. Quang KL, Maguy A, Qi XY, Naud P, Xiong F, Tadevosyan A. **Loss of cardiomyocyte integrin-linked kinase produces an arrhythmogenic cardiomyopathy in mice.**. (2015) **8** 921-32. DOI: 10.1161/CIRCEP.115.001668
38. Sofia RR, Serra AJ, Silva JA, Antonio EL, Manchini MT, Oliveira FA. **Gender-based differences in cardiac remodeling and ILK expression after myocardial infarction.**. (2014) **103** 124-30. DOI: 10.5935/abc.20140113
39. Xie J, Lu W, Gu R, Dai Q, Zong B, Ling L. **The impairment of ILK related angiogenesis involved in cardiac maladaptation after infarction.**. (2011) **6**. DOI: 10.1371/journal.pone.0024115
40. Yang W, He Y, Gan L, Zhang F, Hua B, Yang P. **Cardiac shock wave therapy promotes arteriogenesis of coronary micrangium, and ILK is involved in the biomechanical effects by proteomic analysis.**. (2018) **8**. DOI: 10.1038/s41598-018-19393-z
41. Mu D, Zhang XL, Xie J, Yuan HH, Wang K, Huang W. **Intracoronary transplantation of mesenchymal stem cells with overexpressed integrin-linked kinase improves cardiac function in porcine myocardial infarction.**. (2016) **6**. DOI: 10.1038/srep19155
42. Mao Q, Lin C, Gao J, Liang X, Gao W, Shen L. **Mesenchymal stem cells overexpressing integrin-linked kinase attenuate left ventricular remodeling and improve cardiac function after myocardial infarction.**. (2014) **397** 203-14. PMID: 25134935
43. Gu R, Bai J, Ling L, Ding L, Zhang N, Ye J. **Increased expression of integrin-linked kinase improves cardiac function and decreases mortality in dilated cardiomyopathy model of rats.**. (2012) **7**. DOI: 10.1371/journal.pone.0031279
44. Wang S, Ding L, Ji H, Xu Z, Liu Q, Zheng Y. **The Role of p38 MAPK in the development of diabetic cardiomyopathy.**. (2016) **17**
45. Aschar-Sobbi R, Izaddoustdar F, Korogyi AS, Wang Q, Farman GP, Yang F. **Increased atrial arrhythmia susceptibility induced by intense endurance exercise in mice requires TNFα.**. (2015) **6**. DOI: 10.1038/ncomms7018
46. Silva TA, Ferreira LFC, Pereira MCS, Calvet CM. **Differential role of TGF-β in extracellular matrix regulation during trypanosoma cruzi-host cell interaction.**. (2019) **20**. DOI: 10.3390/ijms20194836
47. Sun F, Duan W, Zhang Y, Zhang L, Qile M, Liu Z. **Simvastatin alleviates cardiac fibrosis induced by infarction via up-regulation of TGF-β receptor III expression.**. (2015) **172** 3779-92. DOI: 10.1111/bph.13166
48. Yang HX, Sun JH, Yao TT, Li Y, Xu GR, Zhang C. **Bellidifolin ameliorates isoprenaline-induced myocardial fibrosis by regulating TGF-β1/Smads and p38 signaling and preventing NR4A1 cytoplasmic localization.**. (2021) **12**. DOI: 10.3389/fphar.2021.644886
49. Lasola JJM, Cottingham AL, Scotland BL, Truong N, Hong CC, Shapiro P. **Immunomodulatory nanoparticles mitigate macrophage inflammation via inhibition of PAMP interactions and lactate-mediated functional reprogramming of NF-κB and p38 MAPK.**. (2021) **13**. DOI: 10.3390/pharmaceutics13111841
50. Meng L, Yuan L, Ni J, Fang M, Guo S, Cai H. **Mir24-2-5p suppresses the osteogenic differentiation with Gnai3 inhibition presenting a direct target via inactivating JNK-p38 MAPK signaling axis.**. (2021) **17** 4238-53. DOI: 10.7150/ijbs.60536
51. Chen Y, Chen Y, Tang C, Zhao Q, Xu T, Kang Q. **RPS4Y1 promotes high glucose-induced endothelial cell apoptosis and inflammation by activation of the p38 MAPK signaling.**. (2021) **14** 4523-34. DOI: 10.2147/DMSO.S329209
52. Bao S, Ji Z, Shi M, Liu X. **EPB41L5 promotes EMT through the ERK/p38 MAPK signaling pathway in esophageal squamous cell carcinoma.**. (2021) **228**. DOI: 10.1016/j.prp.2021.153682
53. Meijles DN, Cull JJ, Markou T, Cooper STE, Haines ZHR, Fuller SJ. **Redox regulation of cardiac ASK1 (apoptosis signal-regulating kinase 1) controls p38-MAPK (mitogen-activated protein kinase) and orchestrates cardiac remodeling to hypertension.**. (2020) **76** 1208-18. DOI: 10.1161/HYPERTENSIONAHA.119.14556
|
---
title: Differences in peripheral and central metabolites and gut microbiome of laying
hens with different feather-pecking phenotypes
authors:
- Chao Wang
- Yaling Li
- Haoliang Wang
- Miao Li
- Jinsheng Rong
- Xindi Liao
- Yinbao Wu
- Yan Wang
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC10017472
doi: 10.3389/fmicb.2023.1132866
license: CC BY 4.0
---
# Differences in peripheral and central metabolites and gut microbiome of laying hens with different feather-pecking phenotypes
## Abstract
### Background
Feather pecking (FP) is a maladaptive behavior in laying hens that is associated with numerous physiological traits, including those involving the central neurotransmitter system and the immune system, which have been identified in many species as being regulated by the gut microbiota via the “microbiota-gut-brain” (MGB) axis. Yet, it is unknown whether and how gut microbiota influences FP by regulating multiple central neurotransmission systems and immune system.
### Methods
This study was measured the prevalence of severe FP (SFP) in the commercial layer farm. The chicken flock with the highest frequency of SFP were selected for FP phenotype identification. Nontargeted metabolomics was performed to investigated the differences in the peripheral and central metabolites and 16S rDNA sequencing was performed to investigated the differences in gut microbiome of laying hens with different FP phenotypes. Correlation analysis was performed to determine the potential mechanism by which the disturbed gut microbiota may modulate host physiology and behavior.
### Results
The results showed that pullets (12 weeks of age) showed significantly higher SFP frequencies than chicks (6 weeks of age) and adults (22 weeks of age; $p \leq 0.05$). Compared to neutrals (N), peckers (P) exhibited the stress-induced immunosuppression with the increased plasma levels of corticosterone and norepinephrine, and the decreased plasma levels of IgA, IL-1, IL-6 and tumor necrosis factor α ($p \leq 0.05$). In the cecum, the relative abundances of Bacteroides and Gemmiger were higher in the P group, while Roseburia, Ruminococcus2, Anaerostipes, Lachnospiracea_incertae_sedis and Methanobrevibacter were more enriched in the N group. Moreover, increased plasma levels of L-tryptophan, beta-tyrosine and L-histidine were found in the P group ($p \leq 0.05$). Notably, in the P group, hippocampal levels of L-tryptophan, xanthurenic acid, L-histidine and histamine were improved and showed a positive association with L-glutamic acid levels. Plasma levels of L-tryptophan, beta-tyrosine and L-histidine were both positively correlated with Bacteroides abundance but negatively correlated with Methanobrevibacter abundance.
### Conclusion
Overall, these findings suggest that the development of FP may be affected by the gut microbiota, which regulates the central glutamatergic nerve system by altering the metabolism of tryptophan, histidine and tyrosine.
## Introduction
Feather pecking (FP) is a maladaptive behavior with an identified prevalence of $80\%$ in all laying hens housing systems (Gunnarsson, 1999). FP was divided into gentle feather peck (GFP) and severe feather pecking (SFP). SFP, a detrimental type of FP, can cause feather loss and skin damage, and in some cases, this can escalate to severe injuries and cannibalism, while GFP is suggested to be similar to social exploration without damage (Kops et al., 2013). What’s more, SFP can spread rapidly through learning among chickens, severely damaging animal welfare and causing economic losses (Rodenburg et al., 2013). Therefore, FP, especially SFP in laying hens is one of the most important unsolved behavioral issues in modern agriculture.
FP is multifactorial and has been linked to numerous behavioral characteristics, such as fearfulness, stress sensitivity and depression, but also to the central and peripheral physiological characteristics (Rodenburg et al., 2013). Deficiency or redundancy in the central serotonergic system can predispose birds to develop FP, while birds with high FP tendency generally have low rates of central serotonin (5-HT) and dopamine (DA) turnover at a young age but high turnover in the brain at an adult age (de Haas and van der Eijk, 2018). The brain transcriptomes of laying hens divergently selected for FP reveal the potential role of GABAergic and glutamatergic neurotransmitter systems in the development of FP (Falker-Gieske et al., 2020). Birds selected for high FP (HFP) and low FP (LFP) differ in innate and adapted immune characteristics (van der Eijk et al., 2019a,b). These studies above suggest that the occurrence of FP may be related to alterations in multiple central neurotransmission systems and immune system, however, there is a lack of clear evidence for the cause of these alterations.
The gastrointestinal tract is a complex ecosystem containing a large number of resident microorganisms that have been found to play an important role in the maintenance of host behavior via the “microbiota-gut-brain” (MGB) axis (Zheng et al., 2019). Recent evidence suggests that alterations in gut microbiota composition, via, for example, anti-, pre-or probiotic treatment, affect anxiety, stress sensitivity and fearfulness (Desbonnet et al., 2010; Ait-Belgnaoui et al., 2014). Moreover, the regulatory effects of the gut microbiota on serotonergic, dopaminergic, GABAergic and glutamatergic neurotransmitter systems in the central nervous system (CNS) and immune system have also been identified by many studies (Miller et al., 2011; Zheng et al., 2019; Hasebe et al., 2021). Notably, a growing number of studies are focusing on the potential role of the regulation of the gut microbiota in FP development. Laying hens that are divergently selected for FP (HFP and LFP) show significant differences in gut microbiota composition (Birkl et al., 2018; van der Eijk et al., 2019c). Early-life transplantation of microbiota from HFP birds influences the behavioral and physiological characteristics that are related to FP (van der Eijk et al., 2020). Collectively, these findings highlight the novel possibility that disturbances of gut microbiota or the MGB axis may contribute to the onset of FP in laying hens. Yet, it is unknown whether and how gut microbiota influences FP by multiple central neurotransmission systems and immune system. Metabolism is an important pathway of two-way communication between gut microorganisms and the brain (Zheng et al., 2019), however, there is a lack of systematic studies on the metabolism in laying hens with different FP phenotypes (including pecker, victim and neutral).
Outcomes from numerous research methods, including improving the house climate (light intensity, temperature, humidity and sound) and foraging condition (rearing intensity, feed shape and nonstarch polysaccharide concentration), indicate that the appropriate rearing environment plays a key role in alleviating the development of FP in laying hens (Lambton et al., 2010; Gilani et al., 2013). The physically, nutritionally, sensorially and socially restricted environment in which the majority of commercial laying hens hatch and live can be a powerful social and environmental chronic stressor that could induce a high occurrence rate of SFP (Maes et al., 2009). Therefore, the aim of this research was to investigate the prevalence of SFP in the commercial layer farm, as well as the metabolic characteristics and the gut microbial characteristics of laying hens with different FP phenotypes, and to reveal the potential correlation between gut microbiota and FP.
To address this issue, the frequency of SFP in the three stages of laying hens, which include chick, pellet and adult was investigated in a commercial layer farm. The chicken flock with the highest frequency of SFP were selected for FP phenotype identification and laboratory analysis. 16S ribosomal RNA (16S rRNA) gene sequencing analysis was performed to reveal the difference in the gut microbial communities of laying hens with different FP phenotypes. We conducted nontarget metabolomic analysis of the plasma and hippocampus to investigate metabolic changes in the peripheral and central systems. Finally, correlation analysis was performed to determine the potential mechanism by which the disturbed gut microbiota may modulate host physiology and behavior.
## Animals and housing conditions
This research was conducted in a commercial layer farm (Guangdong Lvyang Agricultural Co., LTD, Guangdong Province, China). First, preliminary FP observation was conducted for three consecutive days to select the chicken flock with the highest frequency of SFP among three different stages of chicken flocks (chick, pellet and adult stages of chicken) of beak-trimming Hyline gray laying hens. Specifically, 6 cages were randomly selected from laying hens at 6 weeks of age (40 chicks per cage), 12 weeks of age (25 pullets per cage) and 22 weeks of age (10 adults per cage), and pullet flocks at 12 weeks of age showed more SFP (Figure 1). Therefore, a total of 200 12-weeks-old beak-trimmed birds (8 cages, 25 birds per cage) were then selected, individually identified using numbered silicone backpacks (6 × 6 × 0.5 cm; Birkl et al., 2017) and transferred to the top cage without disturbing the flock. Birds were kept in battery layer cages (120 l*60 W*45 H cm for chicks, 100 l*50*45 H cm for pullets and 120 l*60 W*45 H cm for adults) and reared under conventional management conditions on a commercial farm. One camera (HIKVISION DS-2CD2T55(D)-I3, Hangzhou, China) was installed 1 meter above each cage to enable a full view of the cage. The cages were arranged at intervals, so it was certain that there was no visual contact between the different cages. At 6 weeks of age, the light was on for 9 h, from 8:00 until 17:00. At 12 weeks of age, the light was provided from 6:00 to 22:00, and this stayed the same throughout the laying period. Birds had ad libitum access to well water and commercial layer feed. The animals used in this study were treated in accordance with the approval of the Scientific Ethics Committee of South China Agricultural University under approved permit number SYXK2014-0136.
**Figure 1:** *The SFP frequencies of birds at different stage of laying hens. The statistically significant differences between different groups are indicated as asterisks (*p < 0.05, **p < 0.01, and ***p < 0.001).*
## Behavioral observations
For the preliminary FP observation, each cage was video recorded for 10 min (1 × in the morning between 8 and 11 am, 1 × in the after between 3 and 5 pm) on 3 consecutive days. FP was divided into GFP (subdivided into exploratory FP and stereotyped FP) and SFP as adapted from (Birkl et al., 2017). SFP was defined as follows: “A bird grips and pulls or tears vigorously at a feather of another bird with her beak, causing the feather to lift up, break or be pulled out. The recipient reacts to the peck by vocalizing, moving away or turning toward the pecking bird.” The number of SFP events was observed using the video and recoded at the cage level. After individual identification and transfer, birds were provided 2 weeks before the experiment started. Backpacks were fastened around the wings via two elastic straps secured to the backpacks with metal eyelets (Mindus et al., 2021). At 14 weeks of age, FP behavior was observed at an individual level. SFP was observed from video recordings, and each observation lasted 20 min and was performed once in the morning (8:30–11:30) and once in the afternoon (14:00–17:00). The number of SFP events, either given or received, was summed over 4 consecutive days, thus including one morning and one afternoon observation with a total observation period of 40 min, and was used to identify FP phenotypes (Daigle et al., 2015). When a bird gave more than one and received zero or one severe FP, it was defined as a pecker (P). When a bird gave and received zero or one severe feather peck, it was defined as neutral (N). We did not include victims or feather pecker-victims in this study. All-occurrence sampling was used to record initiators and recipients of SFP interactions. An occurrence was defined as a sequence of uninterrupted behavior lasting more than 4 s aimed at the same bird (Birkl et al., 2017). All behavioral observations were performed by a trained, blinded observer.
## Sample collection
After FP observation, one pecker and one natural were chosen from each cage, for a total of 16 hens. Blood samples were collected from the wing vein using EDTA-coated vacutainer tubes. Blood samples were stored on ice (maximum of 4 h) until centrifugation (4°C, 2,500 rpm, 15 min) for plasma separation. Plasma was aliquoted into 1.5 ml microtubes and stored at-80°C until further determination of plasma stress and immune indices. After blood sampling, the birds were euthanized by cervical dislocation to obtain the contents of the duodenum, ileum and caeca. Gut contents were stored in 2 ml cryovials at-80°C until further analysis. The hippocampus, considered the memory and learning area in both mammals and birds, was quickly sampled on ice and stored in liquid nitrogen (Colombo and Broadbent, 2000).
## Measurement of plasma stress and immune indices
The plasma levels of interleukin-1 (IL-1), interleukin-6 (IL-6), immunoglobulin A (IgA), immunoglobulin G (IgG), tumor necrosis factor α (TNF-α), and corticosterone (CORT) (Item No#: YJ059829, YJ042757, YJ002792, YJ042771, YJ002790, YJ059881, Shanghai Enzyme Linked Biotechnology Co., Ltd., Shanghai, China) were measured using a commercially available ELISA kit following the manufacturer’s protocol. The optical density of each sample was read at 450 nm using Nessler’s reagent spectrophotometry (Shanghai Ao Yi Technology Co., Ltd., Shanghai).
## DNA extraction and 16S rRNA gene sequencing
The total bacterial DNA of each sample of gut content was extracted using the QIAamp PowerFecal Pro DNA Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. DNA quality was examined by electrophoresis on a $1\%$ agarose gel, and the final DNA concentration of each sample was determined using an ultrafine, ultraviolet spectrophotometer (Shanghai Ao Yi Technology Co. Ltd., Shanghai). DNA samples were stored at-80°C until further analysis.
The 16S rRNA gene amplicons were used to determine diversity and compare the structures of bacterial communities among samples to reveal taxonomic composition. Next-generation sequencing library preparations and Illumina NoveSeq sequencing were conducted at Novogene Co., Ltd., Tianjin, China. The V3-V4 hypervariable regions of the 16S rRNA genes were amplified with forward primers containing the sequence 5’-CCTAYGGGRBGCASCAG-3′ and reverse primers containing the sequence 5’-GGACTACNNGGGTATCTAAT-3′. Furthermore, indexed adapters were added to both ends of the 16S rDNA amplicons to generate indexed libraries ready for downstream NGS sequencing on the Illumina NovaSeq platform. The quality of each DNA library was validated with an Agilent 2,100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, United States), and the concentration was measured by a Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Carlsbad, CA, United States). DNA libraries were multiplexed and loaded on an Illumina NovaSeq6000 instrument according to the manufacturer’s instructions (Illumina, San Diego, CA, United States). Sequencing was performed using the 2 × 300 bp paired-end configuration. Image analysis and base calling were performed by NovaSeq Control Software embedded in the NovaSeq instrument.
## Nontargeted metabolomics
Thawed plasma samples at 4°C were vortexed for 1 min after thawing and mixed evenly as described in previous research (Zelena et al., 2009) with some modifications. For metabolite extraction, cold methanol (stored at-20°C) was added at a ratio of 400 μl per 100 μl of plasma. The samples were then vortexed for 1 min and centrifuged for 10 min at 4°C and 13,000 rpm. The samples were kept on ice between the steps. The supernatant was transferred to a new 2 ml centrifuge tube and dissolved in 150 μl of 2-chloro-l-phenylalanine (4 ppm) solution prepared with $80\%$ methanol water (stored at 4°C). The supernatant was filtered with a 0.22 μm membrane and inserted into HPLC vials for analysis. To extract hippocampal metabolites, samples weighing 60 mg were ground at 50 Hz for 1 min in 1 ml of cold tissue extract ($75\%$ 9:1 methanol: chloroform, $25\%$ H2O) with 3 steel balls (Want et al., 2013). After room temperature ultrasound for 30 min and ice bath for 30 min, the samples were centrifuged at 12000 rpm and 4°C for the supernatant. The supernatant was then redissolved with 200 μl of $50\%$ acetonitrile solution prepared with 2-amino-3-(2-chlorophenyl)-propionic acid (4 ppm; stored at 4°C) and filtered through a 0.22 μm membrane for Liquid chromatography-mass spectrometry (LC–MS) detection.
All samples were analyzed by LC–MS. Liquid chromatography (LC) analysis was performed on a Vanquish UHPLC System (Thermo Fisher Scientific, United States). Chromatography was carried out with an ACQUITY UPLC ® HSS T3 (150 × 2.1 mm, 1.8 μm; Waters, Milford, MA, United States). The column temperature was maintained at 40°C. The flow rate and injection volume were set at 0.25 ml/min and 2 μl, respectively. For LC-ESI (+)-MS analysis, the mobile phases consisted of (C) $0.1\%$ formic acid in acetonitrile (v/v) and (D) $0.1\%$ formic acid in water (v/v). Separation was conducted under the following gradient: 0 ~ 1 min, $2\%$ C; 1 ~ 9 min, $2\%$ ~ $50\%$ C; 9 ~ 12 min, $50\%$ ~ $98\%$ C; 12 ~ 13.5 min, $98\%$ C; 13.5 ~ 14 min, $98\%$ ~ $2\%$ C; 14 ~ 20 min, $2\%$ C. For LC-ESI (−)-MS analysis, the analytes were carried out with (A) acetonitrile and (B) ammonium formate (5 mM). Separation was conducted under the following gradient: 0 ~ 1 min, $2\%$ A; 1 ~ 9 min, $2\%$ ~ $50\%$ A; 9 ~ 12 min, $50\%$ ~ $98\%$ A; 12 ~ 13.5 min, $98\%$ A; 13.5 ~ 14 min, $98\%$ ~ $2\%$ A; 14 ~ 17 min, $2\%$ A. Mass spectrometric (MS) detection of metabolites was performed on a Q Exactive (Thermo Fisher Scientific, United States) with an ESI ion source. Simultaneous MS1 and MS/MS (Full MS-ddMS2 mode, data-dependent MS/MS) acquisition was used. The parameters were as follows: sheath gas pressure, 30 arb; aux gas flow, 10 arb; spray voltage, 3.50 kV and-2.50 kV for ESI(+) and ESI(−), respectively; capillary temperature, 325°C; MS1 range, m/z 81–1,000; MS1 resolving power, 70,000 FWHM; number of data-dependent scans per cycle, 10; MS/MS resolving power, 17,500 FWHM; normalized collision energy, $30\%$; dynamic exclusion time, automatic.
## Behavior observation data analysis
The SFP frequencies were determined per individual cage per min and averaged at the cage level. IBM SPSS Statistic 26 software (IBM Corp., Armonk, NY) was used to compare the SFP frequencies by one-way analysis of variance (ANOVA). Multiple comparisons were conducted using the Duncan method. The data are presented as the means ± standard errors (SE). Significant differences were reported as those with $p \leq 0.05.$
## Plasma stress and immune data analysis
The plasma stress and immune data were analyzed with independent sample T tests using IBM SPSS Statistic 26 software (IBM Corp., Armonk, NY). The data are presented as the means ± SE. Significant differences were reported as those with $p \leq 0.05.$
## 16S rRNA gene sequencing data analysis
The 16S rRNA gene sequencing data were analyzed using the QIIME data analysis package (V1.9.11) in R software (version 4.0.3) and R studio (version 1.3.1093). The forward and reverse reads were joined to form joined sequences. After removing barcode and primer sequences, the reads of each sample were spliced using FLASH V 1.2.7.2 Quality filtering of the joined sequences was performed. Sequences that did not fulfill the following standards were discarded: sequence length < 200 bp, no ambiguous bases, and mean quality score ≥ 15. The remaining sequences were compared with the reference database (RDP Gold database) using the UCHIME algorithm3 to detect chimeras. Sequences with chimeric sequences were removed from further analysis. Filtered sequences were grouped into operational taxonomic units (OTUs) using the clustering program VSEARCH (V2.15.14) against the Silva 119 database at $97\%$ sequence identity. The Ribosomal Database Program (RDP) classifier was used to assign taxonomic categories to all OTUs at a confidence threshold of 0.8. The α and β diversity analyzes were conducted using USEARCH (V10.0.2405).
Alpha diversity (α-diversity) indices, including the Shannon index for diversity6 and the Chao1 index for richness,7 were calculated by QIIME (1.9.1) from rarefied samples. Difference analyzes of the alpha diversity index, parametric tests and nonparametric tests were conducted. Because there were only two groups, the T test was used for the difference analysis of the alpha diversity index. Beta diversities were calculated using unweighted UniFrac distances. The unweighted pair group method with arithmetic mean was used to generate dendrograms from the beta diversity distance matrix. The principal coordinate analysis (PCoA) and the significance analysis of microbial structure were performed using the OmicStudio tools at https://www.omicstudio.cn/tool. The PCoA analysis was performed based on the unweighted unifrac distance and the p value was calculated using the analysis of similarity (ANOSIM). The key bacterial taxa responsible for discrimination between the two groups were identified using linear discriminant analysis (LDA) effective size (LEfSe).8 The threshold of the logarithmic LDA score was 3.5.
## Nontargeted metabolomic data analysis
The raw data were first converted to mzXML format by MSConvert in the ProteoWizard software package (v3.0.8789) and processed using XCMS for feature detection, retention time correction and alignment. The metabolites were identified by accuracy mass (< 30 ppm) and MS/MS data, which were matched with HMDB,9 massbank,10 LipidMaps,11 mzclound12 and KEGG.13 The robust LOESS signal correction (QC-RLSC) was applied for data normalization to correct for any systematic bias. After normalization, only ion peaks with relative standard deviations (RSDs) less than $30\%$ in QC were kept to ensure proper metabolite identification.
Ropls software was used for all multivariate data analyzes and modeling. Data were mean-centered using scaling. Models were generated on principal component analysis (PCA), orthogonal partial least-square discriminant analysis (PLS-DA) and partial least-square discriminant analysis (OPLS-DA). The metabolic profiles could be visualized as a score plot, where each point represents a sample. The corresponding loading plot and S-plot were generated to provide information on the metabolites that influence the clustering of the samples. All the models evaluated were tested for overfitting with permutation tests. The descriptive performance of the models was determined by R2X (cumulative; perfect model: R2X (cum) = 1) and R2Y (cumulative; perfect model: R2Y (cum) = 1) values, while their prediction performance was measured by Q2 (cumulative; perfect model: Q2 (cum) = 1) and a permutation test. The permuted model should not be able to predict classes: R2 and Q2 values at the Y-axis intercept must be lower than those of Q2 and the R2 of the nonpermuted model. OPLS-DA allowed the determination of discriminating metabolites using the variable importance on projection (VIP). The p value, variable importance projection (VIP) produced by OPLS-DA and fold change (FC) were applied to discover the contributable variables for classification. Finally, results with a p value < 0.05 and VIP values > 1 were considered to be statistically significant metabolites. Differential metabolites were subjected to pathway analysis by MetaboAnalyst,14 which combines results from powerful pathway enrichment analysis with pathway topology analysis. The metabolites identified in the metabolomics analysis were then subjected to KEGG pathway analysis for biological interpretation of higher-level systemic functions. The metabolites and corresponding pathways were visualized using the KEGG Mapper tool.
## Twelve-week-old birds showed more serious SFP
The SFP frequencies of birds at different stages are shown in Figure 1. Pullets showed significantly higher SFP frequencies than chicks and adults ($p \leq 0.05$). Compared to adults, pullets were kept at a higher rearing density. A previous study identified that the high rearing density of laying hens results in FP at the pullet stage and not the adult stage in hens (30 weeks of age) on commercial farms (Bestman et al., 2009). In this study, in addition to evaluating rearing density, continuous inspection and isolation of victims during commercial farming were investigated and may have been a potential factor contributing to the decrease in SFP at the adult stage of laying hens. The pullet stage is the stage of fastest weight gain, with higher stress vulnerability and more serious SFP than in chicks (Rodenburg et al., 2004). Therefore, pullets were selected for further investigation in this study.
## Stress-induced immunosuppression in peckers
After 4 consecutive days of behavioral observation, one pecker and one natural were chosen from each cage, resulting in a total of 16 hens, for further investigation. The plasma levels of stress and immune indices were measured by ELISA (Figure 2). The plasma levels of IgA, IL-1, IL-6 and TNF-α in the P group were significantly lower than those in the N group ($p \leq 0.05$), while the IgG level showed no significant difference. Plasma CORT and Norepinephrine (NE) (Supplementary Figure S6) levels, which are measured in bird stress indices, were increased in the P group ($p \leq 0.05$). The increased stress hormone levels measured in the P group suggest that the FP birds were in a state of stress, and stress hormone signaling has been identified as the final common pathway involved in regulating the pathophysiological status and behavior of animals (Dallman et al., 1994). Various unpredicted extreme or mild chronic stresses, such as noise, mixing chicken breeds, strengthening or weakening light intensity and chicken transfer, have been found to contribute to the development of FP in chickens (Maes et al., 2009). The outcomes from numerous experiments indicate that unpredicted stress is associated with many detrimental behaviors, including depression-like and anxiety-like behavior, via the immune system in both humans and animals (Ait-Belgnaoui et al., 2014). In this research, the suppressed immune system, which was associated with decreased levels of IgA and proinflammatory cytokines (IL-1, IL-6 and TNF-α), may have been the result of central anti-inflammatory cytokines (IL-10) expression caused by long-term stress (Mormède et al., 2003). Moreover, elevated stress hormone levels can lead to disrupted gut barrier function and altered commensal bacteria (Maes et al., 2013).
**Figure 2:** *The plasma level of stress (CORT) and immune indices (IgA, IgG, IL-1, IL-6 and TNF-α). IgA = Immunoglobulin A, IgG = Immunoglobulin G, IL-1 = interleukin-1, IL-6 = interleukin-6, TNF-α = tumor necrosis factor-α, CORT = corticosterone. The values in the violin plot are the means ± SE (n = 8). The statistically significant differences between the P and the N group are indicated as different letter (a, b; p < 0.05). P = feather pecker, N = neutra.*
## Changes in the intestinal microbiota community in peckers
In total, we obtained 967,536 high-quality reads across all cecal samples, and these reads were clustered into 509 OTUs at $97\%$ sequence similarity. In the duodenum and ileum, 809,736 and 785,292 high-quality sequences and 788 and 817 OTUs were obtained, respectively. Most rarefaction curves tended to approach the saturation plateau, suggesting that the sequencing depth was sufficient to cover the whole bacterial diversity (Supplementary Figures S1A–C). Alpha diversity (α-diversity) showed that both the richness index (Chao 1) and diversity index (Shannon) of the duodenum, ileum and cecum did not differ between the P group and the N group (Figures 3A,B). To determine whether the microbial composition of birds with FP was substantially different from that of the N group, we carried out β-diversity analysis. Based on the unweighted UniFrac distance, PCoA revealed a significant difference in the cecal microbiota community ($p \leq 0.05$) but no significant difference in the duodenum and ileum ($p \leq 0.05$; Figure 3C; Supplementary Figures S2A,B). Therefore, we focused on the characteristics of the cecal microbial community in the rest of this study. At the phylum level, the top five phyla identified were Firmicutes, Bacteroidetes, Proteobacteria, Euryarchaeota and Fusobacteria in the cecum of laying hens (Figure 4A). Among these phyla, Firmicutes and Bacteroidetes were dominant. At the genus level, the majority of 16S rRNA amplicons belonged to Bacteroides, Faecalibacterium, Methanobrevibacter, Ruminococcus2, Alistipes and Lactobacillus (Figure 4B). To further identify the cecal microbiota responsible for discriminating peckers from neutrals, we carried out LEfSe. This analysis identified 7 differential microorganisms responsible for the discrimination between the two groups at the genus level (Figures 4C,D). The relative abundances of Bacteroides and Gemmiger were higher in the P group, while Roseburia, Ruminococcus2, Anaerostipes, Lachnospiracea_incertae_sedis and Methanobrevibacter were more enriched in the N group ($p \leq 0.05$). These data imply that the structure of the cecal microbiota was disordered in peckers.
**Figure 3:** *The diversity of gut microbiota of laying hen with different FP phenotype. (A) Chao 1 index of cecum (n = 8), duodenum (n = 6) and ileum (n = 6); (B) Shannon index of cecum, duodenum and ileum; (C) The principal co-ordinates (PCoA) analysis of cecum based on the unweighted unifrac distance. P = feather pecker, N = neutral.* **Figure 4:** *The relative abundance of cecal bacterial composition at the phylum and genus level and LefSe analysis of cecal microbiota. (A) The Relative abundance of cecal bacterial composition at phylum level; (B) The Relative abundance of cecal bacterial composition at genus level; (C) and (D) LEfSe analysis of cecal microbiota. The threshold of the logarithmic LDA score was 3.5. P = feather pecker, N = neutral.*
The gut microbiota, known as the second brain, has been found to exert its effect on the brain and behavior by regulating catabolism or the nervous system based on the “gut-brain axis” (Zheng et al., 2019). To date, an increasing number of investigations have focused on whether gut microbiota affect feather pecking in laying hens. Early-life microbiota transplantation showed a long-term influence on depression-like and anxiety-like behavior related to FP in laying hens (van der Eijk et al., 2020). The gut microbiota analysis of LFP and HFP birds revealed significantly different diversity and composition, including increased abundance of Lactobacillus and decreased abundance of Clostridiales in HFP birds (van der Eijk et al., 2019c). Moreover, ingestion of *Lactobacillus rhamnosus* retards chronic stress-induced FP in chickens, suggesting the important role of gut microorganisms in relieving FP in birds (Mindus et al., 2021). The features of the gut microbiota found in this research showed a different pattern from those observed in previous studies, which could have been the result of multiple factors, including genetics, nutrition and stress (Maes et al., 2013; Tremblay et al., 2021). Bacteroides, Ruminococcus2 and Methanobrevibacter, the main enriched microorganisms, were found to have a higher or lower relative abundance in peckers, which has a close association with depression-like and anxiety-like behavior (Chen et al., 2021; Zhang et al., 2022). Knowledge regarding the underlying associations between feather pecking and these differential microbes are vague and require further metabolomics analysis.
## Changes in the metabolic profile in peripheral and central organisms in peckers
We further performed nontargeted metabolomics to determine whether the metabolic state reflected in the plasma and hippocampus were paralleled by an altered gut microbiota. To further explore the metabolic distinctions between the P and N groups, the multivariate statistical analysis was performed (Supplementary Figures S3–S5). Furthermore, a total of 94 metabolites were changed significantly in the P groups. The P group had 89 metabolites with higher levels and 5 metabolites with lower levels than in the N group ($p \leq 0.05$; Supplementary Figure S6; Supplementary Table S1). Tryptophan, as the precursor of the central major inhibitory neurotransmitter 5-HT, can pass through the blood brain barrier (BBB) and affect brain function and behavior (Hasebe et al., 2021). Tryptophan-5-HT deficiency has been identified to be involved in the development of many maladaptive behaviors in birds, such as aggression and FP (de Haas and van der Eijk, 2018). In the P group, however, the plasma concentration of L-tryptophan was significantly higher than that in the N group by 2.19-log2−fold ($p \leq 0.05$). Moreover, indole and quinolinic acid, downstream metabolites of tryptophan, also had log2-fold increases of 1.78 and 2.36, respectively ($p \leq 0.05$). NE, which acts as a neurotransmitter and as a hormone, is usually activated after exposure to stress and has been found to have increased levels in the blood of HFP birds in response to manual restraint (Korte et al., 1997). In our research, the levels of NE and its precursor tyrosine were higher in the P group than in the N group ($p \leq 0.05$). Outcomes from various experiments suggest that cognition and memory function are tightly related to fluctuations in the histidine level in the peripheral system (Holeček, 2020). In the P group, the L-histidine concentration was more enriched in plasma than in the N group ($p \leq 0.05$). Notably, peripheral histidine and aromatic amino acids (AAA, tyrosine, phenylalanine and tryptophan), which had increased levels in P plasma, can be transported into the brain via large neutral amino acids (LNAAs) transporter and affect the CNS neurotransmitters histamine, 5-HT, glutamic acid and GABA (Oldendorf et al., 1988). To further investigate the characteristics of the metabolic profile in peckers, we conducted KEGG pathway enrichment analysis of differential metabolites via MetabAnaylst. Specifically, 9 metabolic pathways, including ‘glycine, serine and threonine metabolism’, ‘alanine, aspartate and glutamate metabolism’, ‘beta-alanine metabolism’, ‘histidine metabolism’ and ‘tyrosine metabolism’, were identified (Figure 5A; Supplementary Table S2). The peripheric change including glycine, serine and threonine metabolism’, ‘alanine, aspartate and glutamate metabolism’, ‘beta-alanine metabolism’, ‘histidine metabolism’ and ‘tyrosine metabolism’ have previously reported to be associated with depression (Goto et al., 2017; Liu et al., 2020; Johnston et al., 2021; Solís-Ortiz et al., 2021). Next, the association between KEGG metabolic pathways and differential metabolites was exhibited by a network plot (Figure 5B). These enriched metabolic pathways exhibit a complicated interaction involving metabolites, including L-tryptophan, L-histidine and NE.
**Figure 5:** *Changes in the metabolic profile in plasma obtained from FP birds. (A) Bubble chart using the top 25 KEGG pathways enriched by the differential metabolites in plasma; (B) Network plot using the top 10 KEGG pathways enriched from the differential metabolites in plasma.*
To determine the central effects of the changes in plasma metabolism and the cecal microbial community, hippocampal metabolomics analysis was performed. For subsequent differential metabolite screening and pathway enrichment analysis, multivariate statistical analysis was performed (Supplementary Figures S7–S9). According to the criteria: p value < 0.05 and VIP values > 1, 51 differential metabolites were detected in the metabolic profiles of the two groups. In the P group, the levels of 47 metabolites were higher and the levels of 4 metabolites were lower than in the N group (Supplementary Figure S10; Supplementary Table S3). It is worth noting that the L-glutamic acid concentration showed a log2-fold increase of 1.51 in the hippocampus in group P ($p \leq 0.05$). Glutamic acid, which acts as the main excitatory neurotransmitter in the CNS, plays an extensive and key role in the maintenance of brain functions, including emotion and cognition, therefore affecting numerous behaviors, such as aggression, depression and anxiety (Gerhard et al., 2016; Zheng et al., 2019). In the hippocampus of peckers, the concentration of L-tryptophan was also higher than that in the N group ($p \leq 0.05$). The levels of 5-HT, which is an important product of tryptophan and central neurotransmitters and plays an important role in regulating the onset of FP in laying hens (de Haas and van der Eijk, 2018), did not differ between the two groups. Instead, the levels of xanthurenic acid, one crucial metabolite in the kynurenine (KYN) pathway of tryptophan, significantly increased in the P group ($p \leq 0.05$). The KYN pathway is an alternate tryptophan breakdown pathway that, under physiological conditions, metabolizes tryptophan (> $95\%$) into KYN and an array of downstream neuroactive metabolites, including xanthurenic acid (Danielski et al., 2018). Xanthurenic acid, an endogenous kynurenine, is a known vesicular glutamic acid transport (VGLUT) inhibitor and has also been proposed as a mGlu$\frac{2}{3}$ receptor agonist (Neale et al., 2013). Previous studies have found stereoselective blood–brain barrier transport of histidine by in vivo or in vitro experiments (Nowak et al., 1997; Yamakami et al., 1998). Histamine, which is synthesized from the amino acid histidine through oxidative decarboxylation by histidine decarboxylase in the brain, exerts complicated effects in depression-like and anxiety-like behavior (Raber, 2005). In the present study, the hippocampal levels of L-histidine and histamine were increased in the P group ($p \leq 0.05$). The decreased NE concentration shown in group P ($p \leq 0.05$) may have been the result of the competition between peripheral tyrosine (the precursor of NE), other AAAs (phenylalanine and tryptophan) and histidine to LNAA carriers to be transported into the brain (Oldendorf et al., 1988). The KEGG pathway enrichment analysis of differential metabolites showed 9 significantly enriched metabolic pathways, including ‘Histidine metabolism’, ‘Phenylalanine, tyrosine and tryptophan biosynthesis’, ‘Arginine and proline metabolism’, ‘Tryptophan metabolism’ and ‘Aminoacyl-tRNA biosynthesis’ (Figure 6A; Supplementary Table S4). The central alteration, including ‘Histidine metabolism’, ‘Phenylalanine, tyrosine and tryptophan biosynthesis’, and ‘Tryptophan metabolism’ have reported to be associated with depression-like behaviors (Yoshikawa et al., 2014; Wang et al., 2022). The network diagram showed the association between differential metabolites and enrichment pathways (Figure 6B). L-glutamic acid shows a complex relationship with multiple metabolic pathways, suggesting the potentially important role of L-glutamic acid in FP development.
**Figure 6:** *The turbulence of metabolic profile in hippocampus of FP bird. (A) Bubble chart from the top 25 KEGG pathways enriching with the differential metabolites in plasma; (B) The network plot from the top 10 KEGG pathways enriching with the differential metabolites in hippocampus.*
## Correlation analysis
To investigate potential associations among the gut microbiome, plasma physiological index, plasma metabolites and hippocampal metabolites, correlation analysis was conducted. Figure 7A shows the results of Spearman’s correlation analysis performed using the levels of hippocampal differential metabolites and Mantel’s test performed using the levels of hippocampal differential metabolites and plasma metabolic pathways. L-glutamic acid levels showed a significant positive correlation with L-histidine, histamine, L-tryptophan and xanthurenic acid levels in the hippocampus (Spearman r > 0.6, $p \leq 0.01$). A previous study reported an association between glutamic acid and histidine levels (Ritz et al., 2002). The main pathway of histidine catabolism begins with deamination catalyzed by histidase to urocanate and leads through 4-imidazolone-5-propionate and formiminoglutamate to glutamic acid, while the alternative pathways of histidine catabolism include transamination to imidazolepyruvate and decarboxylation to histamine. Increased L-histidine and histamine levels were also found in a rat model of glutamic acid excitotoxicity and other neurodegenerative disorders (Fang et al., 2014). The present study revealed an increase in xanthurenic acid concentration but not in 5-HT levels. Xanthurenic acid is a known VGLUT inhibitor and has also been proposed as a mGlu$\frac{2}{3}$ receptor agonist (Neale et al., 2013). These results indicated that the development of FP is potentially correlated with changes in the glutamatergic system in the CNS. Furthermore, the levels of multiple differential metabolites, including L-histidine, histamine, L-tryptophan and xanthurenic acid, in the hippocampus were significantly correlated with ‘glycine, serine and threonine metabolism’, ‘tyrosine metabolism’ and ‘histidine metabolism’ in plasma. Collectively, these findings indicated that the disturbed glutamatergic system was potentially associated with the differential metabolites involved in ‘glycine, serine and threonine metabolism’ or ‘histidine metabolism’ in plasma. Therefore, a total of 16 differential metabolites, such as L-histidine, L-tryptophan and beta-tyrosine, which participate in ‘glycine, serine and threonine metabolism’, ‘tyrosine metabolism’ and ‘histidine metabolism’, were chosen to perform Spearman correlation analysis with cecal flora genera and plasma physiological indices (Figure 7B). Bacteroides and Methanobrevibacter abundance showed the closest association with the plasma metabolite profile. L-tryptophan, L-histidine and beta-tyrosine levels were both positively correlated with Bacteroides abundance but negatively correlated with Methanobrevibacter abundance. Bacteroides, the most prominent genus that was enriched in the P group, was associated with abnormal behaviors of the host. The gut microbiome from major depressive disorder patients was found to be enriched with the genus Bacteroides, and these microbes were found to be associated with increased anxiety and depression-like behavior and impaired hippocampal neurogenesis in rats subjected to fecal microbiota transplantation (Zhang et al., 2022). Outcomes from numerous experimental methods, including dietary tryptophan restriction and histidine supplementation, have revealed that Bacteroides play an important role in histidine metabolism and tryptophan metabolism (Zapata et al., 2018; Kang et al., 2020). More OTUs were enriched in the genus Methanobrevibacter, which had decreased abundance in the P group, and its function is closely related to anxiety and depression-like behavior (Chen et al., 2021). Previous research found a negative association between the abundance of Methanobrevibacter and dietary tryptophan levels (Rao et al., 2021). Elevated levels of stress hormones, including CORT and N, can also lead to disrupted gut barrier function and altered commensal bacteria (Maes et al., 2013). In this research, the increase in the levels of NE and CORT, which are stress hormones, was positively correlated with Bacteroides abundance. However, the potential association between NE, CORT and *Bacteroides is* unclear. To our knowledge, NE levels were found only to change with a significant negative correlation with Bacteroides abundance, which was reported in research on depression (Chen et al., 2021).
**Figure 7:** *Correlation analysis between gut microorganisms, metabolites and plasma physiological indices. (A) Pairwise comparisons of hippocampus metabolites, with a color gradient denoting the Spearman’s correlation coefficients. Plasma KEGG pathways enriched with differential metabolites was correlated with each hippocampus metabolites by partial Mantel tests. Curve width represents the significant correlation coefficients (p < 0.05) of the partial Mantel tests. (B) Spearman’s correlation analysis between differential microorganism, plasma differential metabolites and plasma physiological indices.*
## Conclusion
Taken together, our results demonstrated the different patterns of the gut microbiota, metabolism and immune system and revealed the potential association between FP, the gut microbiota and the glutamatergic neurotransmitter system. In this research, peckers were found to suffer from long-term stress with a suppressed immune system. Disturbances in the cecal microbiota, including increased Bacteroides abundance and decreased Methanobrevibacter abundance, were found in peckers. The abundances of the two microorganisms showed significant correlations with the plasma levels of L-tryptophan, beta-tyrosine and L-histidine, which may further affect the hippocampal levels of metabolites involved in the glutamatergic neurotransmitter system, including L-glutamic acid, L-tryptophan, xanthurenic acid, and L-histidine. In conclusion, the findings of this study have provided a new insight into developing novel biotherapeutic strategies for alleviating FP in laying hens.
## Data availability statement
The data presented in the study are deposited in NCBI Sequence Read Archive (SRA) repository (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA933384), accession number PRJNA933384.
## Ethics statement
The animal study was reviewed and approved by the Scientific Ethics Committee of South China Agricultural University.
## Author contributions
CW contributed to the data analysis, investigation, and drafting the manuscript. YL, HW, and ML were responsible for behavioral observation, breeding and sampling. JR provided technical support. XL and YBW were responsible for supervision. YW contributed to the conceptualization, project administration and critical revision of the manuscript. All the authors have read and approved the final manuscript.
## Funding
This work was supported by the National Natural Science Foundation of China [31972610], the Construction Project of Modern Agricultural Science and Technology Innovation Alliance in Guangdong Province (2022KJ128 and 2023KJ128) and the earmarked fund for Modern Agro-industry Technology Research System (CARS-40).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2023.1132866/full#supplementary-material
## References
1. Ait-Belgnaoui A., Colom A., Braniste V., Ramalho L., Marrot A., Cartier C.. **Probiotic gut effect prevents the chronic psychological stress-induced brain activity abnormality in mice**. *Neurogastroenterol. Motil.* (2014) **26** 510-520. DOI: 10.1111/nmo.12295
2. Bestman M., Koene P., Wagenaar J. P.. **Influence of farm factors on the occurrence of feather pecking in organic reared hens and their predictability for feather pecking in the laying period**. *Appl. Anim. Behav. Sci.* (2009) **121** 120-125. DOI: 10.1016/j.applanim.2009.09.007
3. Birkl P., Bharwani A., Kjaer J. B., Kunze W., McBride P., Forsythe P.. **Differences in cecal microbiome of selected high and low feather-pecking laying hens**. *Poult. Sci.* (2018) **97** 3009-3014. DOI: 10.3382/ps/pey167
4. Birkl P., Franke L., Rodenburg T. B., Ellen E., Harlander-Matauschek A.. **A role for plasma aromatic amino acids in injurious pecking behavior in laying hens**. *Physiol. Behav.* (2017) **175** 88-96. DOI: 10.1016/j.physbeh.2017.03.041
5. Chen Y., Xiao N., Chen Y., Chen X., Zhong C., Cheng Y.. **Semen sojae praeparatum alters depression-like behaviors in chronic unpredictable mild stress rats via intestinal microbiota**. *Food Res. Int.* (2021) **150** 110808. DOI: 10.1016/j.foodres.2021.110808
6. Colombo M., Broadbent N.. **Is the avian hippocampus a functional homologue of the mammalian hippocampus?**. *Neurosci. Biobehav. Rev.* (2000) **24** 465-484. DOI: 10.1016/S0149-7634(00)00016-6
7. Daigle C. L., Rodenburg T. B., Bolhuis J. E., Swanson J. C., Siegford J. M.. **Individual consistency of feather pecking behavior in laying hens: once a feather pecker always a feather pecker?**. *Front. Vet. Sci.* (2015) **2** 6. DOI: 10.3389/fvets.2015.00006
8. Dallman M. F., Akana S. F., Levin N., Walker C. D., Bradbury M. J., Suemaru S.. **Corticosteroids and the control of function in the hypothalamo-pituitary-adrenal (HPA) axis**. *Ann. N. Y. Acad. Sci.* (1994) **746** 22-31. DOI: 10.1111/j.1749-6632.1994.tb39206.x
9. Danielski L. G., Giustina A. D., Goldim M. P., Florentino D., Mathias K., Garbossa L.. **Vitamin B6 reduces neurochemical and long-term cognitive alterations after polymicrobial sepsis: involvement of the kynurenine pathway modulation**. *Mol. Neurobiol.* (2018) **55** 5255-5268. DOI: 10.1007/s12035-017-0706-0
10. de Haas E. N., van der Eijk J. A.. **Where in the serotonergic system does it go wrong? Unravelling the route by which the serotonergic system affects feather pecking in chickens**. *Neurosci. Biobehav. Rev.* (2018) **95** 170-188. DOI: 10.1016/j.neubiorev.2018.07.007
11. Desbonnet L., Garrett L., Clarke G., Kiely B., Cryan J. F., Dinan T.. **Effects of the probiotic**. *Neuroscience* (2010) **170** 1179-1188. DOI: 10.1016/j.neuroscience.2010.08.005
12. Falker-Gieske C., Mott A., Preuß S., Franzenburg S., Bessei W., Bennewitz J.. **Analysis of the brain transcriptome in lines of laying hens divergently selected for feather pecking**. *BMC Genomics* (2020) **21** 595-514. DOI: 10.1186/s12864-020-07002-1
13. Fang Q., Hu W. W., Wang X. F., Yang Y., Lou G. D., Jin M. M.. **Histamine up-regulates astrocytic glutamate transporter 1 and protects neurons against ischemic injury**. *Neuropharmacology* (2014) **77** 156-166. DOI: 10.1016/j.neuropharm.2013.06.012
14. Gerhard D. M., Wohleb E. S., Duman R. S.. **Emerging treatment mechanisms for depression: focus on glutamate and synaptic plasticity**. *Drug Discov. Today* (2016) **21** 454-464. DOI: 10.1016/j.drudis.2016.01.016
15. Gilani A. M., Knowles T. G., Nicol C. J.. **The effect of rearing environment on feather pecking in young and adult laying hens**. *Appl. Anim. Behav. Sci.* (2013) **148** 54-63. DOI: 10.1016/j.applanim.2013.07.014
16. Goto T., Tomonaga S., Toyoda A.. **Effects of diet quality and psychosocial stress on the metabolic profiles of mice**. *J. Proteome Res.* (2017) **16** 1857-1867. DOI: 10.1021/acs.jproteome.6b00859
17. Gunnarsson S.. **Effect of rearing factors on the prevalence of floor eggs, cloacal cannibalism and feather pecking in commercial flocks of loose housed laying hens**. *Br. Poult. Sci.* (1999) **40** 12-18. DOI: 10.1080/00071669987773
18. Hasebe K., Kendig M. D., Morris M. J.. **Mechanisms underlying the cognitive and behavioural effects of maternal obesity**. *Nutrients* (2021) **13** 240. DOI: 10.3390/nu13010240
19. Holeček M.. **Influence of histidine administration on ammonia and amino acid metabolism: a review**. *Physiol. Res.* (2020) **69** 555-564. DOI: 10.33549/physiolres.934449
20. Johnston C. S., Jasbi P., Jin Y., Bauer S., Williams S., Fessler S. N.. **Daily vinegar ingestion improves depression scores and alters the metabolome in healthy adults: a randomized controlled trial**. *Nutrients* (2021) **13** 4020. DOI: 10.3390/nu13114020
21. Kang M., Yin J., Ma J., Wu X., Huang K., Li T.. **Effects of dietary histidine on growth performance, serum amino acids, and intestinal morphology and microbiota communities in low protein diet-fed piglets**. *Mediat. Inflamm.* (2020) **2020** 1240152-7. DOI: 10.1155/2020/1240152
22. Kops M. S., de Haas E. N., Rodenburg T. B., Ellen E. D., Korte-Bouws G. A., Olivier B.. **Effects of feather pecking phenotype (severe feather peckers, victims and non-peckers) on serotonergic and dopaminergic activity in four brain areas of laying hens (**. *Physiol. Behav.* (2013) **120** 77-82. DOI: 10.1016/j.physbeh.2013.07.007
23. Korte S. M., Beuving G., Ruesink W. I. M., Blokhuis H. J.. **Plasma catecholamine and corticosterone levels during manual restraint in chicks from a high and low feather pecking line of laying hens**. *Physiol. Behav.* (1997) **62** 437-441. DOI: 10.1016/S0031-9384(97)00149-2
24. Lambton S. L., Knowles T. G., Yorke C., Nicol C. J.. **The risk factors affecting the development of gentle and severe feather pecking in loose housed laying hens**. *Appl. Anim. Behav. Sci.* (2010) **123** 32-42. DOI: 10.1016/j.applanim.2009.12.010
25. Liu T., Zhou N., Xu R., Cao Y., Zhang Y., Liu Z.. **A metabolomic study on the anti-depressive effects of two active components from**. *Artif. Cells Nanomed. Biotechnol.* (2020) **48** 718-727. DOI: 10.1080/21691401.2020.1774597
26. Maes M., Kubera M., Leunis J. C., Berk M., Geffard M., Bosmans E.. **In depression, bacterial translocation may drive inflammatory responses, oxidative and nitrosative stress (O&NS), and autoimmune responses directed against O&NS-damaged neoepitopes**. *Acta Psychiatr. Scand.* (2013) **127** 344-354. DOI: 10.1111/j.1600-0447.2012.01908.x
27. Maes M., Yirmyia R., Noraberg J., Brene S., Hibbeln J., Perini G.. **The inflammatory & neurodegenerative (I&ND) hypothesis of depression: leads for future research and new drug developments in depression**. *Metab. Brain Dis.* (2009) **24** 27-53. DOI: 10.1007/s11011-008-9118-1
28. Miller B. J., Buckley P., Seabolt W., Mellor A., Kirkpatrick B.. **Meta-analysis of cytokine alterations in schizophrenia: clinical status and antipsychotic effects**. *Biol. Psychiatry* (2011) **70** 663-671. DOI: 10.1016/j.biopsych.2011.04.013
29. Mindus C., van Staaveren N., Fuchs D., Gostner J. M., Kjaer J. B., Kunze W.. *Sci. Rep.* (2021) **11** 19538-19515. DOI: 10.1038/s41598-021-98459-x
30. Mormède C., Castanon N., Médina C., Moze E., Lestage J., Neveu P. J.. **Chronic mild stress in mice decreases peripheral cytokine and increases central cytokine expression independently of IL-10 regulation of the cytokine network**. *Neuroimmunomodulation* (2003) **10** 359-366. DOI: 10.1159/000071477
31. Neale S. A., Copeland C. S., Uebele V. N., Thomson F. J., Salt T. E.. **Modulation of hippocampal synaptic transmission by the kynurenine pathway member xanthurenic acid and other VGLUT inhibitors**. *Neuropsychopharmacology* (2013) **38** 1060-1067. DOI: 10.1038/npp.2013.4
32. Nowak J. Z., Zawilska J. B., Woldan-Tambor A., Sçk B., Voisin P., Lintunen M.. **Histamine in the chick pineal gland: origin, metabolism, and effects on the pineal function**. *J. Pineal Res.* (1997) **22** 26-32. DOI: 10.1111/j.1600-079X.1997.tb00299.x
33. Oldendorf W. H., Crane P. D., Braun L. D., Gosschalk E. A., Diamond J. M.. **pH dependence of histidine affinity for blood-brain barrier carrier transport systems for neutral and cationic amino acids**. *J. Neurochem.* (1988) **50** 857-861. DOI: 10.1111/j.1471-4159.1988.tb02991.x
34. Raber J.. **Histamine receptors as potential therapeutic targets to treat anxiety and depression**. *Drug Dev. Res.* (2005) **65** 126-132. DOI: 10.1002/ddr.20015
35. Rao Z., Li J., Shi B., Zeng Y., Liu Y., Sun Z.. **Dietary tryptophan levels impact growth performance and intestinal microbial ecology in weaned piglets via tryptophan metabolites and intestinal antimicrobial peptides**. *Animals* (2021) **11** 1. DOI: 10.3390/ani11030817
36. Ritz M. F., Schmidt P., Mendelowitsch A.. **17β-estradiol effect on the extracellular concentration of amino acids in the glutamate excitotoxicity model in the rat**. *Neurochem. Res.* (2002) **27** 1677-1683. DOI: 10.1023/A:1021695213099
37. Rodenburg T. B., Buitenhuis A. J., Ask B., Uitdehaag K. A., Koene P., Van Der Poel J. J.. **Genetic and phenotypic correlations between feather pecking and open-field response in laying hens at two different ages**. *Behav. Genet.* (2004) **34** 407-415. DOI: 10.1023/B:BEGE.0000023646.46940.2d
38. Rodenburg T. B., Van Krimpen M. M., De Jong I. C., De Haas E. N., Kops M. S., Riedstra B. J.. **The prevention and control of feather pecking in laying hens: identifying the underlying principles**. *Worlds Poult. Sci. J.* (2013) **69** 361-374. DOI: 10.1017/S0043933913000354
39. Solís-Ortiz S., Arriaga-Avila V., Trejo-Bahena A., Guevara-Guzmán R.. **Deficiency in the essential amino acids l-isoleucine, l-leucine and l-histidine and clinical measures as predictors of moderate depression in elderly women: a discriminant analysis study**. *Nutrients* (2021) **13** 3875. DOI: 10.3390/nu13113875
40. Tremblay A., Lingrand L., Maillard M., Feuz B., Tompkins T. A.. **The effects of psychobiotics on the microbiota-gut-brain axis in early-life stress and neuropsychiatric disorders**. *Prog. Neuropsychopharmacol. Biol. Psychiatry* (2021) **105** 110142. DOI: 10.1016/j.pnpbp.2020.110142
41. van der Eijk J. A., de Vries H., Kjaer J. B., Naguib M., Kemp B., Smidt H.. **Differences in gut microbiota composition of laying hen lines divergently selected on feather pecking**. *Poult. Sci.* (2019c) **98** 7009-7021. DOI: 10.3382/ps/pez336
42. van der Eijk J. A., Lammers A., Kjaer J. B., Rodenburg T. B.. **Stress response, peripheral serotonin and natural antibodies in feather pecking genotypes and phenotypes and their relation with coping style**. *Physiol. Behav.* (2019a) **199** 1-10. DOI: 10.1016/j.physbeh.2018.10.021
43. van der Eijk J. A., Rodenburg T. B., de Vries H., Kjaer J. B., Smidt H., Naguib M.. **Early-life microbiota transplantation affects behavioural responses, serotonin and immune characteristics in chicken lines divergently selected on feather pecking**. *Sci. Rep.* (2020) **10** 2750-2713. DOI: 10.1038/s41598-020-59125-w
44. van der Eijk J. A., Verwoolde M. B., de Vries Reilingh G., Jansen C. A., Rodenburg T. B., Lammers A.. **Chicken lines divergently selected on feather pecking differ in immune characteristics**. *Physiol. Behav.* (2019b) **212** 112680. DOI: 10.1016/j.physbeh.2019.112680
45. Wang D., Wu J., Zhu P., Xie H., Lu L., Bai W.. **Tryptophan-rich diet ameliorates chronic unpredictable mild stress induced depression-and anxiety-like behavior in mice: the potential involvement of gut-brain axis**. *Food Res. Int.* (2022) **157** 111289. DOI: 10.1016/j.foodres.2022.111289
46. Want E. J., Masson P., Michopoulos F., Wilson I. D., Theodoridis G., Plumb R. S.. **Global metabolic profiling of animal and human tissues via UPLC-MS**. *Nat. Protoc.* (2013) **8** 17-32. DOI: 10.1038/nprot.2012.135
47. Yamakami J., Sakurai E., Sakurada T., Maeda K., Hikichi N.. **Stereoselective blood-brain barrier transport of histidine in rats**. *Brain Res.* (1998) **812** 105-112. DOI: 10.1016/S0006-8993(98)00958-5
48. Yoshikawa T., Nakamura T., Shibakusa T., Sugita M., Naganuma F., Iida T.. **Insufficient intake of L-histidine reduces brain histamine and causes anxiety-like behaviors in male mice**. *J. Nutr.* (2014) **144** 1637-1641. DOI: 10.3945/jn.114.196105
49. Zapata R. C., Singh A., Ajdari N. M., Chelikani P. K.. **Dietary tryptophan restriction dose-dependently modulates energy balance, gut hormones, and microbiota in obesity-prone rats**. *Obesity* (2018) **26** 730-739. DOI: 10.1002/oby.22136
50. Zelena E., Dunn W. B., Broadhurst D., Francis-McIntyre S., Carroll K. M., Begley P.. **Development of a robust and repeatable UPLC-MS method for the long-term metabolomic study of human serum**. *Anal. Chem.* (2009) **81** 1357-1364. DOI: 10.1021/ac8019366
51. Zhang Y., Fan Q., Hou Y., Zhang X., Yin Z., Cai X.. **Bacteroides species differentially modulate depression-like behavior via gut-brain metabolic signaling**. *Brain Behav. Immun.* (2022) **102** 11-22. DOI: 10.1016/j.bbi.2022.02.007
52. Zheng P., Zeng B., Liu M., Chen J., Pan J., Han Y.. **The gut microbiome from patients with schizophrenia modulates the glutamate-glutamine-GABA cycle and schizophrenia-relevant behaviors in mice**. *Sci. Adv.* (2019) **5** eaau8317. DOI: 10.1126/sciadv.aau8317
|
---
title: A bibliometric analysis of systematic reviews and meta-analyses in ophthalmology
authors:
- Yihang Fu
- Yuxiang Mao
- Shuangyan Jiang
- Sheng Luo
- Xiaoyun Chen
- Wei Xiao
journal: Frontiers in Medicine
year: 2023
pmcid: PMC10017479
doi: 10.3389/fmed.2023.1135592
license: CC BY 4.0
---
# A bibliometric analysis of systematic reviews and meta-analyses in ophthalmology
## Abstract
### Background
Bibliometric analysis is a quantitative method which applies mathematical and statistical tools to evaluate the inter-relationships and impacts of publications, authors, institutions and countries in a specific research area. Systematic reviews and meta-analyses (SRMAs) are summaries of the best available evidence to address a specific research question via comprehensively literature search, in-depth analysis and synthesis of results. To date, there have been several studies summarizing the publication trends of SRMAs in research specialties, however, none has conducted specifically in ophthalmology. The purpose of this study is to establish the scientometric landscape of SRMAs published in the field of ophthalmology over time.
### Methods
We retrieved relevant ophthalmological SRMAs and the corresponding bibliometric parameters during 2000 to 2020 from Web of Science Core Collection. Bibliometric analysis was performed using bibliometrix package. Pre-registration and guideline compliance of each article was independently assessed by two investigators.
### Results
A total of 2,660 SRMAs were included, and the average annual growth rate was $21.26\%$. China and the United States were the most productive countries, while Singapore was the country with the highest average citations per document. Wong TY was not only the most productive, but also the most frequently cited author. The most productive affiliation was National University of Singapore ($$n = 236$$). Systematic reviews and meta-analyses output in most subspecialties had steadily increased with retina/vitreous ($$n = 986$$), glaucoma ($$n = 411$$) and cornea/external diseases ($$n = 303$$) constantly as the most dominant fields. Rates of pre-registration and guideline compliance had dramatically increased over time, with 20.0 and $63.5\%$ of article being pre-registered and reported guideline in 2020, respectively. However, SRMAs published on ophthalmology journals tended to be less frequently pre-registered and guideline complied than those on non-ophthalmology journals (both $p \leq 0.001$).
### Conclusion
The annual output of SRMAs has been rapidly increasing over the past two decades. China and the United States were the most productive countries, whereas Singapore has the most prolific and influential scholar and institution. Raising awareness and implementation of SRMAs pre-registration and guideline compliance is still necessary to ensure quality, especially for ophthalmology journals.
## 1. Introduction
Through comprehensive literature search, critical assessment, and synthesis of all related and trustworthy studies on a specific subject, systematic reviews are more rigorous than traditional narrative reviews, and are regarded as the best summaries of existing evidence (1–3). More importantly, they are capable of updating the knowledge in certain field as well as proposing future research directions [3]. Featured by a replicable and methodical presentation and methodology, systematic reviews could be either quantitative (meta-analysis) or qualitative [1, 4]. Meta-analysis is a quantitative statistical approach to aggregate the results of original studies. It can remarkably increase statistical strength, and precisely estimate the effect size, thus overcoming the limitation of sample size of individual studies (5–7). Additionally, it is able to explore the sources of heterogeneity as well as determine subgroups connected to the factor of interest [8]. High-quality systematic reviews and meta-analyses (SRMAs) have potentials to underpin evidence-based clinical guidelines, and to inform decision-making. Therefore, well-conducted SRMAs are placed at the very top of the evidence pyramid in most current hierarchies [9]. In the field of ophthalmology, reliable SRMAs have been increasingly identified and served as the backbone of the practice guidelines, such as the Preferred Practice Pattern (PPP) issued by the American Academy of Ophthalmology [10, 11]. A prior survey in 2012 reported that SRMAs in ophthalmology had been actively performed, particularly in the domains of retina and glaucoma [12]. Nonetheless, great advances have been taking place thereafter in evidence-based medicine, including the improvement of methodology for meta-analysis, adoption of principles of evidence-based medicine in precision medicine, and the implementation of prospective registration. For example, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statements [13] were first released in 2009 and were updated lately [14]. In 2011, the United Kingdom Centre for Reviews and Dissemination (CRD) at the University of York in England launched an international platform for prospective register of systematic reviews, PROSPERO [15], which aims to minimize reporting bias, reduce waste from unintended duplication, and increase transparency of SRMAs. Overall, these efforts have greatly helped standardize and improve the quality of SRMAs.
Bibliometric analysis is a quantitative method which applies mathematical and statistical tools to evaluate the inter-relationships and impacts of publications, authors, institutions and countries in a specific research area [16]. Through extracting and analyzing the metrics of each publication including author, institution, country, and keywords, bibliometric analysis is able to determine the development trends or future research directions. Compared with conventional narrative reviews by experts, which often subjectively focus on the progress in a specific research field, bibliometric analysis is advantageous in objectively, comprehensively, and quantitatively summarizing the whole topic based on the information best available [16]. More important, with various visualization approaches, the results are displayed in more intuitive and comprehensible ways, which enables scholars to gain a one-stop overview, identify knowledge gaps, uncover emerging trends, and explore the intellectual structure of a specific domain. Given these advantages, bibliometric analysis has gained immense popularity in biomedical research in recent years. For instance, there have been several bibliometric analyses in specialties of medicine and dentistry comprehensively depicting the research trends and hotspots (17–20). In the field of ophthalmology, a number of bibliometric analyses have been published regarding to certain ocular diseases [21, 22], and treatment or diagnostic modalities [23, 24]. However, none has done on all available SRMAs in ophthalmology.
We herein performed this bibliometric analysis on SRMAs in ophthalmology published within the last two decades to explore the trends and patterns of publication, tracking impact and collaboration at author-, institution- and country-levels. These data would help ophthalmologists and scholars grasp the current state of development characteristics of the domain and guide future ophthalmological evidence-based research.
## 2.1. Data sources and search strategies
We designed a two-step approach to retrieve all relevant publications from 2000 to 2020. The first step aimed to extract all possible keywords in varying ophthalmological subspecialties. To this end, we searched the Web of Science-Core Collection (WoSCC) database which is maintained by Clarivate Analytics1 using the following parameters: TS = (“meta analysis” OR “meta analyses” OR “systematic review” OR “systematic reviews”), time span = “from 2000 to 2020,” language = “English,” Web of Science category = “Ophthalmology,” type = “article, review or early access.” TS here represents topic, meaning the search of the mentioned words in the title, abstract, and keyword lists. In this step, a total of 1,498 publications were obtained. We downloaded all original records and analyzed frequency of author’s keywords using the R package - bibliometrix. Keywords that had an occurrence ≥3 times were arbitrarily defined as the core ophthalmology-related keywords, yielding 128 terms (Supplementary material). Based on this collection, we conducted the second-round search in the WoSCC database. The retrieval strategy was set as: TS = (each of the 128 core keywords) AND TS = (“meta analysis” OR “meta analyses” OR “systematic review” OR “systematic reviews”). Other parameters, including publication year, language and literature type, were set identical to those in the first-round search. Both literature retrieval and raw data collection were performed on a single day (October 1, 2021).
## 2.2. Data cleaning
Raw metadata comprising 5,063 records in the 2nd round of literature search were downloaded from WoSCC. The dataset contained complete information of each publication for bibliometric analysis, including literature title, abstract, author list, journal name, keywords, publication year, countries/regions, affiliations, reference list, and citations. To assess eligibility of literature, the raw dataset was transformed and exported to Microsoft Excel 2017 using bibliometrix package, and then independently evaluated by two investigators (W.X. and Y.F.). Reports meeting any of the following criteria were excluded: [1] non-English publication; [2] publication year beyond 2000–2020; [3] retracted article. Full texts were then gone through to classify the subspecialty, and to evaluate whether they had pre-registered and/or adhered to reporting guidelines. Pre-registration was defined if the study prospectively registered its protocol on any of the major public platforms [25], including PROSPERO, the Registry of Systematic Reviews/Meta-Analyses in Research Registry, INPLASY, and Open Science Framework (OSF) Registries and protocols.io. Guideline compliance was assessed through checking whether the study used widely accepted statements, guidelines or checklists, like PRISMA and its extensions (PRISMA-IDP, PRISMA-NMA, etc.), and Meta-analyses Of Observational Studies in Epidemiology (MOOSE). Ophthalmological subspecialty was classified as 12 sections according to the subcategories listed by EyeWiki,2 an online resource launched by the American Academy of Ophthalmology Academy. Any disagreement between two raters was resolved through discussion. The cleaned dataset containing 2,660 publications were finally used for the subsequent bibliometric analysis.
## 2.3. Bibliometric analysis and visualization
Bibliometric analysis was conducted using the R-package bibliometrix and its shiny web-interface biblioshiny [26]. Their integrative and powerful functions allow scholars easily conduct various scientometric analysis, from data importing and conversion, filtering, to various analytics and plots for different levels of metrics. The cleaned dataset was imported to biblioshiny to generate descriptive statistics, including the productivity and citation by author, affiliation, and country. Authors collaboration was illustrated with a network plot using the function biblioNetwork of bibliometrix. Only co-authorship ≥20 was retained in the network. Community structure within the network was uncovered using the Louvain clustering algorithm.
## 2.4. Statistical analysis
Descriptive statistics were directly extracted from the R-package bibliometrix. The proportions of pre-registered study and guideline complied study in ophthalmology and non-ophthalmology journals were compared with Pearson’s Chi-squared test. All statistical analyses and data visualization were performed using the R programming language (version 4.0.5) on RStudio’s open-source software (RStudio, Boston, MA). A p value <0.05 was considered as statistically significant.
## 2.5. Patient and public involvement
No patient was involved in this study.
## 3.1. Characteristics of identified literature
The process of literature identification was illustrated in Figure 1. A total of 5,063 records were initially obtained from the WoSCC database. After excluding ineligible items, 4,866 records were manually assessed for titles and abstracts. We ruled out non-ophthalmology literatures ($$n = 2$$,048) and non-SRMAs ($$n = 158$$), yielding 2,660 SRMAs for final analysis.
**Figure 1:** *Flow diagram demonstrates the screening process of systematic reviews and meta-analyses in ophthalmology.*
## 3.2. Productivity and impact by author and institution
The amount of literature had consistently increased over time, and the average annual growth rate was $21.26\%$. A total of 9,522 authors contributed to all 2,660 publications, resulting in average of 0.28 documents per author. The top 10 productive authors were: Wong TY ($$n = 95$$), Mitchell P ($$n = 51$$), Cheng CY ($$n = 50$$), Jonas JB ($$n = 41$$), Chen LJ ($$n = 39$$), Hewitt AW ($$n = 39$$), Wang JJ ($$n = 38$$), Klaver CCW ($$n = 37$$), Hammond CJ ($$n = 36$$), Li Y ($$n = 36$$) (Figure 2A). There was apparent overlap between top 10 cited and productive authors (Figure 2B). The most frequently cited author was Wong TY ($$n = 694$$), followed by Mitchell P ($$n = 467$$), Cheng CY ($$n = 329$$), Jonas JB ($$n = 279$$), Aung T ($$n = 258$$), Hewitt AW ($$n = 254$$), Hammond CJ ($$n = 234$$), Saw SM ($$n = 232$$), van Duijn CM ($$n = 226$$), Vingerling JR ($$n = 225$$). In terms of the top 10 relevant affiliations, National University of Singapore produced the most publications ($$n = 236$$); other highly productive affiliations included University of Melbourne ($$n = 158$$), and University of Sydney ($$n = 114$$, Figure 2C). The collaboration patterns at author-level were analyzed with bibliometrix package. The overall collaboration index was 3.63. Authors with extensive collaboration fell into an interconnected network (Figure 2D). Notably, all top 10 productive authors appeared in this network (Figure 2D, highlighted in red).
**Figure 2:** *Author production and collaboration. (A) Top 10 productive authors. (B) Top 10 productive authors. (C) Top 10 relevant affiliations. (D) Collaboration network at author-level. The thickness of the line is proportional to the strength of co-authorship. Top 10 productive authors are highlighted in red.*
## 3.3. Country production
There were 88 countries/regions that contributed to publish ophthalmological SRMAs from 2000 to 2020. A heat map visualized the geographical distribution of publication by country (Figure 3A). Overall, the United States and China were two countries with the most publications, other productive countries including those in the East Asia & Pacific, Europe & Central Asia, North America. However, countries in Latin America & Caribbean, Middle East & North Africa, South Asia and Sub-Saharan Africa were extremely underrepresented. As for corresponding authors’ country, we found that China ranked first with nearly 1,000 publications, followed by the United States, the United Kingdom, Australia and Canada (Figure 3B). These top 5 countries published approximately $70\%$ of all articles. The percentage of multiple country publication (MCP) was used to reflect the inter-country collaboration. The vast majority were single country publication (SCP) in most top productive countries, including China, the United States and the United Kingdom. Of note, great percentages of MCP were observed in Singapore and Switzerland. Regarding the citation of SRMAs, we found that the United Kingdom, the United States and China were the top three most-cited countries with over 10,000 citations per country, followed by Australia and the Netherlands (both >5,000 citations) (Figure 3C). As for average citations per document, top five countries were Singapore (116.38 times), Switzerland (86.04 times), the United Kingdom (58.21 times), Australia (55.04 times), and Austria (54.62 times).
**Figure 3:** *Country production and citation. (A) A heat map is presented displaying the number of publications from different countries according to the occurrence of authors. (B) Corresponding author’s country. (C) Top 10 cited countries. SCP, single country publication; MCP, multiple country publication.*
## 3.4. Ophthalmological subspecialties
The counts of SRMA in various ophthalmological subspecialties were listed in Figure 4A. Retina/vitreous ($$n = 986$$), glaucoma ($$n = 411$$), cornea/external diseases ($$n = 303$$), cataract/anterior segment ($$n = 189$$), and pediatric ophthalmology/strabismus ($$n = 183$$) were top five represented subspecialties. Approximately half of SRMAs were published in ophthalmology journals in most subspecialties, with an exception of Oncology/Pathology and Neuro−ophthalmology/Orbit, where SRMAs were more frequently published in non-ophthalmology journals. Year wise publication in various subspecialties was shown in Figure 4B. The number of publications in most subspecialties show a rising trend in general, especially after 2011 (Figure 4B). We then analyzed the proportion of different types of study content (Figure 4C) in major subspecialties. Treatment related SRMA accounted for about a half in each subspecialty, followed by those on epidemiology, genetics, and diagnosis. A Sankey plot revealed the relationship between subspecialties and various bibliometric properties (Figure 4D).
**Figure 4:** *Publication in subspecialties. (A) Numbers of publications in different ophthalmological subspecialties. (B) The trends of publications in five major ophthalmological subspecialties from 2000 to 2020. (C) The proportion of different types of study content in five major ophthalmological subspecialties. (D) A Sankey diagram reveals the relationship between subspecialties and different bibliometric indicators.*
## 3.5. Pre-registration and guideline-compliance of systematic reviews and meta-analyses in ophthalmology
The percentage of pre-registered studies increased dramatically from 0 in 2012 to $20.0\%$ in 2020 (Figure 5A). The rising trend significantly accelerated after 2016. Similarly, a gradual increase showed in the percent of guideline-complied studies from 2010 onwards, which reached up to $63.5\%$ in 2020 (Figure 5B). There showed significantly lower percentages of pre-registered and guideline-complied studies published on ophthalmology journals compared with those on non-ophthalmology journals (Table 1, $5.6\%$ vs. $10.2\%$ for pre-registration, $36.6\%$ vs. $45.1\%$ for guideline-compliance, both $p \leq 0.001$).
**Figure 5:** *The percentage of pre-registered studies (A) and guideline complied studies (B) indexed by web of science core collection over time.* TABLE_PLACEHOLDER:Table 1
## 4. Discussion
The present study aimed to construct the scientometric landscape of SRMA publications in ophthalmology over the period 2000 to 2020. Our data revealed the annual output gradually increased in ophthalmology as a whole, and also in most ophthalmological subspecialties, particularly in the domains of retinal/vitreous, glaucoma and cornea/external disease. China and the United States were the most productive countries, whereas Singapore was the country having the most prolific and influential scholar and institution. International collaboration was intense among high-impact authors. Finally, we found that the rates of pre-registration and reporting guideline compliance in ophthalmology SRMAs have been steadily increasing since 2012, yet leaving room for improvement.
The number of published SRMAs in ophthalmology has substantially increased over the past two decades. The average annual growth rate of publication was $21.26\%$, approximately 5-fold greater than the growth of overall scientific publication ($4.10\%$) [27]. We observed a striking 47-fold longitudinal increase of literature indexed in WoSCC from 2000 to 2020. The reasons for such increase are to be determined, but may in part attribute to the proliferation of SRMAs conducted by more researchers worldwide, especially those in China, as commonly seen in other medical specialties [28].
Geographically, around $70\%$ of SRMAs in ophthalmology were published by scholars in top five productive countries (i.e., China, the United States, the United States, Australia, and Canada). Of note, China ranked first in output of articles, making China the most productive country on SRMA research in ophthalmology. In parallel, five out of top 10 productive affiliations were from China. Compared to rank 3rd of China in 2010 [12], one can readily discern that China has experienced a rapid growth of production within the latest decade (2010–2020), exceeding any other countries. Similar findings were also observed in several other medical specialties than ophthalmology [29, 30]. However, increased research output in China did not lead to simultaneous increase in international collaboration and the academic influence, as explicitly indicated by low percentage of MCP ($15\%$) and low average citation per publication. In terms of the impact of country-level, a noteworthy country is Singapore. It has the most productive institution, numerous high-yield and high-impact scholars, and a close network of collaborations among these scholars, altogether making outstanding contributions to the application of evidence-based medicine in ophthalmology.
Not surprisingly, the field of retina/vitreous section remains the most intensely researched area in evidence-based medicine. Indeed, two out of top five leading causes of global blindness in people aged ≥50 years were retinal diseases (age-related macular degeneration, and diabetic retinopathy) [31]. In the past two decades, clinical study on fundus diseases, especially on DR and ARMD, has been the focus of global ophthalmological research, with extensive investment of manpower, financial resources and funds. It is foreseeable that this trend will continue in the future.
The mass proliferation of SRMA publication has raised concerns about the quality and rigor of reports. A study estimated that only $3\%$ of SRMAs are methodologically sound, non-redundant, thus provide useful clinical information [32]. To minimize bias and increase transparency, systematic reviews are best to be prospectively registered on one hand, and are strictly adhered to set reporting guidelines on the other. Prospective registration on public platforms (e.g., PROSPERO) is a means to publish details about a research project before its commencement thus allowing evidence users to assess whether all steps of the research have been performed and reported as planned. Complete reporting, adhering to guidelines, for example PRISMA, allows readers to assess the appropriateness of the methods, and therefore the trustworthiness of the findings. In ophthalmology, our data showed that the percentages of pre-registered and guideline compliant SRMAs have significantly increased since 2012. This might be largely attributed to the fact that some scientific journals mandatorily require the authors to include a completed checklist in their submission to aid the editorial process and reader. As the rates of pre-registered and guideline compliant SRMAs were still much lower in ophthalmology journals than those in general medical journals, it is strongly recommended that ophthalmology journal editors to make the PRISMA checklist and pre-registration mandatory for all submissions of SRMAs to ensure their reliability and rigor. Further work to understand the barriers to pre-registration and guideline compliance uptake in ophthalmology is also required to address the gap identified by this study.
There has been an increasing number of bibliometric analyses on certain ocular disease or treatment modality of ophthalmology in recent years (33–39). These documents systematically revealed the productivity as well as collaborations of institutions, journals, and countries, making monitor the development of a specific field possible. Furthermore, they helped researchers or clinicians to master the research trends precisely and quickly, thereby aiding in conduct further studies. However, only a handful of bibliometric analyses focused on field of ophthalmology as a whole, and investigate the distribution and gap across subspecialties. In 2012, Chen et al. [ 12] analyzed the SRMAs published in ophthalmology from 1988 to 2010, and found the most heavily reported topics being retina and glaucoma. More than half of SRMAs were published in ophthalmology journals (about $60\%$). These trends have remained the same during the past decade. As this study shown, the most representative topics as well as the proportion of SRMAs published in ophthalmological journals almost unchanged. Another more recent bibliometric analysis investigated all available ophthalmological literature from 2017 to 2021, and found that epidemiology, prevention, screening, and treatment of ocular diseases were served as the hotspots. Moreover, artificial intelligence, drug development, and fundus diseases were acted as new research trends [40], indicating that non-ophthalmology knowledge were increasingly involved in ophthalmology research in the recent 5 years. In parallel, SRMAs on these themes are expected to increase in future.
Several limitations of this study should be addressed. First, one should note that some studies might have actually been compliant with guidelines regardless of no mention of using any specific statements in their full texts. Conversely, studies that stated use of particular guidelines may not necessarily satisfy all specific domains of its statement. As in-depth scoring and evaluating the quality of each SRMA using specific tools (e.g., A Measurement Tool to Assess systematic Reviews [AMSTAR] [41]) is beyond the scope of this study, future research is warranted to perform such analysis focusing on a particular subject. Second, we only searched WoSCC database which may not encompass the entirety of SRMA literatures in ophthalmology.
## 5. Conclusion
The annual output of SRMAs has been rapidly increasing over the past two decades. China and the United States were the most productive countries, whereas Singapore had the most prolific and influential scholar and institution. Raising awareness and implementation of SRMAs pre-registration and guideline compliance is still necessary to ensure quality, especially for ophthalmology journals.
## Data availability statement
We would like to remain the previous data availability statement, “*Original data* were obtained from the Web of Science-Core Collection (WoSCC) database. Dataset of the present study was available from the Dryad repository (http://datadryad.org/) with the link: https://doi.org/10.5061/dryad.fxpnvx0vw.
## Author contributions
WX and XC: conception, design, and administrative support. WX, YF, and SJ: data analysis and interpretation. YF, YM, SJ, SL, XC, and WX: manuscript writing, collection, and assembly of data. All authors read and approved the final manuscript.
## Funding
This study was supported by the National Natural Science Foundation of China [81600751]; the Natural Science Foundation of Guangdong Province, China (2017A030313613, 2016A030310230); and the Pearl River Nova Program of Guangzhou [201806010167].
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmed.2023.1135592/full#supplementary-material
## References
1. Cronin P, Kelly AM, Altaee D, Foerster B, Petrou M, Dwamena BA. **How to perform a systematic review and meta-analysis of diagnostic imaging studies**. *Acad Radiol* (2018) **25** 573-93. DOI: 10.1016/j.acra.2017.12.007
2. Pati D, Lorusso LN. **How to write a systematic review of the literature**. *HERD* (2018) **11** 15-30. DOI: 10.1177/1937586717747384
3. Gupta S, Rajiah P, Middlebrooks EH, Baruah D, Carter BW, Burton KR. **Systematic review of the literature: best practices**. *Acad Radiol* (2018) **25** 1481-90. DOI: 10.1016/j.acra.2018.04.025
4. Siddaway AP, Wood AM, Hedges LV. **How to do a systematic review: a best practice guide for conducting and reporting narrative reviews, meta-analyses, and meta-syntheses**. *Annu Rev Psychol* (2019) **70** 747-70. DOI: 10.1146/annurev-psych-010418-102803
5. Lee YH. **An overview of meta-analysis for clinicians**. *Korean J Intern Med* (2018) **33** 277-83. DOI: 10.3904/kjim.2016.195
6. Walker E, Hernandez AV, Kattan MW. **Meta-analysis: its strengths and limitations**. *Cleve Clin J Med* (2008) **75** 431-9. DOI: 10.3949/ccjm.75.6.431
7. Egger M, Smith GD, Phillips AN. **Meta-analysis: principles and procedures**. *BMJ* (1997) **315** 1533-7. DOI: 10.1136/bmj.315.7121.1533
8. Gotzsche PC. **Why we need a broad perspective on meta-analysis. It may be crucially important for patients**. *BMJ* (2000) **321** 585-6. DOI: 10.1136/bmj.321.7261.585
9. Murad MH, Asi N, Alsawas M, Alahdab F. **New evidence pyramid**. *Evid Based Med* (2016) **21** 125-7. DOI: 10.1136/ebmed-2016-110401
10. Golozar A, Chen Y, Lindsley K, Rouse B, Musch DC, Lum F. **Identification and description of reliable evidence for 2016 American Academy of ophthalmology preferred practice pattern guidelines for cataract in the adult eye**. *JAMA Ophthalmol* (2018) **136** 514-23. DOI: 10.1001/jamaophthalmol.2018.0786
11. Saldanha IJ, Lindsley KB, Lum F, Dickersin K, Li T. **Reliability of the evidence addressing treatment of corneal diseases: a summary of systematic reviews**. *JAMA Ophthalmol* (2019) **137** 775-85. DOI: 10.1001/jamaophthalmol.2019.1063
12. Chen H, Jhanji V. **Survey of systematic reviews and meta-analyses published in ophthalmology**. *Br J Ophthalmol* (2012) **96** 896-9. DOI: 10.1136/bjophthalmol-2012-301589
13. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JP. **The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration**. *BMJ* (2009) **339** b2700. DOI: 10.1136/bmj.b2700
14. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews**. *BMJ* (2021) **372** n71. DOI: 10.1136/bmj.n71
15. Booth A, Clarke M, Dooley G, Ghersi D, Moher D, Petticrew M. **The nuts and bolts of PROSPERO: an international prospective register of systematic reviews**. *Syst Rev* (2012) **1** 2. DOI: 10.1186/2046-4053-1-2
16. Donthu N, Kumar S, Mukherjee D, Pandey N, Lim WM. **How to conduct a bibliometric analysis: an overview and guidelines**. *J Bus Res* (2021) **133** 285-96. DOI: 10.1016/j.jbusres.2021.04.070
17. Ahmad P, Slots J. **A bibliometric analysis of periodontology**. *Periodontol* (2000) **85** 237-40. DOI: 10.1111/prd.12376
18. Arshad AI, Ahmad P, Karobari MI, Asif JA, Alam MK, Mahmood Z. **Antibiotics: a bibliometric analysis of top 100 classics**. *Antibiotics (Basel)* (2020) **9** 219. DOI: 10.3390/antibiotics9050219
19. Ahmad P, Della Bella E, Stoddart MJ. **Applications of bone morphogenetic proteins in dentistry: a bibliometric analysis**. *Biomed Res Int* (2020) **2020** 1-12. DOI: 10.1155/2020/5971268
20. Shuaib W, Acevedo JN, Khan MS, Santiago LJ, Gaeta TJ. **The top 100 cited articles published in emergency medicine journals**. *Am J Emerg Med* (2015) **33** 1066-71. DOI: 10.1016/j.ajem.2015.04.047
21. Nichols JJ, Jones LW, Morgan PB, Efron N. **Bibliometric analysis of the meibomian gland literature**. *Ocul Surf* (2021) **20** 212-4. DOI: 10.1016/j.jtos.2021.03.004
22. Koh B, Banu R, Nusinovici S, Sabanayagam C. **100 most-cited articles on diabetic retinopathy**. *Br J Ophthalmol* (2021) **105** 1329-36. DOI: 10.1136/bjophthalmol-2020-316609
23. Wang S, Yang K, Wang Y, Xu L, Gu Y, Fan Q. **Trends in research on corneal cross linking from 2001 to 2020: a bibliometric analysis**. *Clin Exp Optom* (2022) 1-7. DOI: 10.1080/08164622.2022.2038013
24. Boudry C, Al Hajj H, Arnould L, Mouriaux F. **Analysis of international publication trends in artificial intelligence in ophthalmology**. *Graefes Arch Clin Exp Ophthalmol* (2022) **260** 1779-88. DOI: 10.1007/s00417-021-05511-7
25. Pieper D, Rombey T. **Where to prospectively register a systematic review**. *Syst Rev* (2022) **11** 8. DOI: 10.1186/s13643-021-01877-1
26. Aria M, Cuccurullo C. **Bibliometrix: an R-tool for comprehensive science mapping analysis**. *J Inf Secur* (2017) **11** 959-75. DOI: 10.1016/j.joi.2017.08.007
27. Bornmann L, Haunschild R, Mutz R. **Growth rates of modern science: a latent piecewise growth curve approach to model publication numbers from established and new literature databases**. *Hum Soc Sci Commun* (2021) **8** 903. DOI: 10.1057/s41599-021-00903-w
28. Yang Z, Wu Q, Wu K, Fan D. **Scientific publications on systematic review and meta-analysis from Chinese authors: a 10-year survey of the English literature**. *Front Med* (2012) **6** 94-9. DOI: 10.1007/s11684-012-0181-y
29. Jiang JL, Zhang JX, Li RR, Zhao ZQ, Ye XL. **Research trends of systematic review/meta-analysis on acupuncture therapy: a bibliometric analysis**. *J Pain Res* (2021) **14** 561-73. DOI: 10.2147/JPR.S290516
30. Shi J, Gao Y, Ming L, Yang K, Sun Y, Chen J. **A bibliometric analysis of global research output on network meta-analysis**. *BMC Med Inform Decis Mak* (2021) **21** 144. DOI: 10.1186/s12911-021-01470-5
31. **Causes of blindness and vision impairment in 2020 and trends over 30 years, and prevalence of avoidable blindness in relation to VISION 2020: the right to sight: an analysis for the global burden of disease study**. *Lancet Glob Health* (2021) **9** e144-60. DOI: 10.1016/S2214-109X(20)30489-7
32. Ioannidis J. **The mass production of redundant, misleading, and conflicted systematic reviews and meta-analyses**. *Milbank Q* (2016) **94** 485-514. DOI: 10.1111/1468-0009.12210
33. Dong LY, Yin M, Kang XL. **Bibliometric network analysis of glaucoma**. *Genet Mol Res* (2014) **13** 3577-85. DOI: 10.4238/2014.May.9.1
34. Zhao Y, Huang L, Xiang M, Li Q, Miao W, Lou Z. **Trends in conjunctivochalasis research from 1986 to 2017: a bibliometric analysis**. *Medicine (Baltimore)* (2018) **97** e12643. DOI: 10.1097/MD.0000000000012643
35. Efron N, Morgan PB, Jones LW, Nichols JJ. **Bibliometric analysis of the refractive error field**. *Clin Exp Optom* (2021) **104** 641-3. DOI: 10.1080/08164622.2021.1880868
36. Xu J, Zhao F, Fang J, Shi M, Pan J, Sun W. **Mapping research trends of chronic ocular graft-versus-host disease from 2009 to 2020: a bibliometric analysis**. *Int Ophthalmol* (2022) **42** 3963-76. DOI: 10.1007/s10792-022-02380-9
37. Efron N, Morgan PB, Jones LW, Nichols JJ. **Bibliometric analysis of the keratoconus literature**. *Clin Exp Optom* (2022) **105** 372-7. DOI: 10.1080/08164622.2021.1973866
38. Nichols JJ, Jones L, Morgan PB, Efron N. **Bibliometric analysis of the orthokeratology literature**. *Cont Lens Anterior Eye* (2021) **44** 101390. DOI: 10.1016/j.clae.2020.11.010
39. Efron N, Jones LW, Morgan PB, Nichols JJ. **Bibliometric analysis of the literature relating to scleral contact lenses**. *Cont Lens Anterior Eye* (2021) **44** 101447. DOI: 10.1016/j.clae.2021.101447
40. Tan Y, Zhu W, Zou Y, Zhang B, Yu Y, Li W. **Hotspots and trends in ophthalmology in recent 5 years: bibliometric analysis in 2017-2021**. *Front Med (Lausanne)* (2022) **9** 988133. DOI: 10.3389/fmed.2022.988133
41. Shea BJ, Reeves BC, Wells G, Thuku M, Hamel C, Moran J. **AMSTAR 2: a critical appraisal tool for systematic reviews that include randomised or non-randomised studies of healthcare interventions, or both**. *BMJ* (2017) **358** j4008. DOI: 10.1136/bmj.j4008
|
---
title: Association of healthy lifestyle with incident cardiovascular diseases among
hypertensive and normotensive Chinese adults
authors:
- Jian Su
- Houyue Geng
- Lulu Chen
- Xikang Fan
- Jinyi Zhou
- Ming Wu
- Yan Lu
- Yujie Hua
- Jianrong Jin
- Yu Guo
- Jun Lv
- Pei Pei
- Zhengming Chen
- Ran Tao
journal: Frontiers in Cardiovascular Medicine
year: 2023
pmcid: PMC10017485
doi: 10.3389/fcvm.2023.1046943
license: CC BY 4.0
---
# Association of healthy lifestyle with incident cardiovascular diseases among hypertensive and normotensive Chinese adults
## Abstract
### Background
Whether lifestyle improvement benefits in reducing cardiovascular diseases (CVD) events extend to hypertensive patients and whether these benefits differ between hypertensive and normotensive individuals is unclear. This study aimed to investigate the associations of an overall healthy lifestyle with the subsequent development of CVD among participants with hypertension and normotension.
### Methods
Using data from the Suzhou subcohort of the China Kadoorie Biobank study of 51,929 participants, this study defined five healthy lifestyle factors as nonsmoking or quitting for reasons other than illness; nonexcessive alcohol intake; relatively higher physical activity level; a relatively healthy diet; and having a standard waist circumference and body mass index. We estimated the associations of these lifestyle factors with CVD, ischemic heart disease (IHD) and ischemic stroke (IS).
### Results
During a follow-up of 10.1 years, this study documented 6,151 CVD incidence events, 1,304 IHD incidence events, and 2,243 IS incidence events. Compared to those with 0–1 healthy lifestyle factors, HRs for those with 4–5 healthy factors were 0.71 ($95\%$ CI: 0.62, 0.81) for CVD, 0.56 ($95\%$ CI: 0.42, 0.75) for IHD, and 0.63 ($95\%$ CI: 0.51, 0.79) for IS among hypertensive participants. However, we did not observe this association among normotensive participants. Stratified analyses showed that the association between a healthy lifestyle and IHD risk was stronger among younger participants, and the association with IS risk was stronger among hypertensive individuals with lower household incomes.
### Conclusion
Adherence to a healthy lifestyle pattern is associated with a lower risk of cardiovascular diseases among hypertensive patients, but this benefit is not as pronounced among normotensive patients.
## Introduction
Cardiovascular diseases (CVD) continue to be the leading cause of death and disability globally [1]. Moreover, CVD contribute tremendously to the disease burden in China; more than $40\%$ of deaths are attribute to CVD. Ischemic heart disease (IHD) and ischemic stroke (IS) constitute the largest proportions in CVD deaths [2]. Meanwhile, as one of the most important independent risk factors for CVD, hypertension had a prevalence rate of $27.5\%$ among Chinese adults in 2018 [3]. A third of CVD deaths among hypertensive patients are caused by high blood pressure [4].
It has been demonstrated that avoiding smoking [5], nonexcessive alcohol consumption [6], engaging in adequate physical activity [7, 8], following a healthy diet (9–11), and maintaining a healthy body shape [12] can prevent many cases of CVD in general populations. A healthy lifestyle including these factors was associated with an approximately $43.2\%$ reduction in IHD incidence and a $39.1\%$ reduction in IS incidence among Chinese adults according to previous study [13]. However, there is still insufficient research evidence to confirm whether the control of CVD by lifestyle improvement could be extrapolated to hypertensive patients. In addition, it is also unclear whether these healthy lifestyle habits differ between hypertensive and normotensive individuals.
Therefore, this large prospective cohort study aimed to investigate the associations of healthy lifestyle with incidence of CVD, IHD, and IS among Chinese adults with and without hypertension.
## Study population
We used participants’ data from the China Kadoorie Biobank (CKB) study in Wuzhong District of Suzhou city, Jiangsu Province. Detailed descriptions of the CKB cohort have been previously published (14–16). We collected informed consent from all participants who completed questionnaires administered by interviewers and had physical measurements taken. The CKB study was approved by the Ethical Review Committee of the Chinese Center for Disease Control and Prevention (Beijing, China), and the Oxford Tropical Research Ethics Committee (University of Oxford, UK).
Overall, a total of 53,269 participants aged 30–79 years were eligible for inclusion. After excluding individuals who had self-reported previous medical histories of cancer ($$n = 331$$), stroke ($$n = 466$$), heart disease ($$n = 574$$) and outliers of duration of hypertension ($$n = 8$$), current analysis included 51,921 participants.
## Assessment of lifestyle factors
In the baseline questionnaire, a variety of lifestyle factors were assessed. Questions about cigarette smoking included smoking status (never, former, or current smoker), amount of daily cigarette smoking for current smokers, and the reason for quitting and years since quitting for former smokers. Alcohol consumption included drinking status (never, former, occasionally, monthly, weekly, or daily); drinkers who drank once or more per week were asked how much alcohol they consumed on a typical drinking day over the past year. The information about physical activity included the common type (occupational-, commuting-, domestic-, and leisure time-related domains) and duration of activities in the past year. Based on the metabolic equivalent tasks (METs) for each activity, we calculated the daily level of physical activity by multiplying the MET value for each activity by hours spent on each activity and summing the MET-hours for all activities [17]. Dietary data was collected by a qualitative food frequency questionnaire including 12 conventional food groups in China to assess the habitual dietary intake during the past year. The relative validity and reproducibility of qualitative and quantitative food frequency questionnaires (FFQs) have been validated in previous studies [18]. Weight, height, and waist circumference (WC) were measured by trained investigator using calibrated instruments. We calculated body mass index (BMI) as weight (kg)/(height (m)2).
## Assessment of covariates and hypertension
Baseline questionnaire also collected sociodemographic information (age, sex, marital status, highest education level, household income, and occupation), personal and family medical history, time of sedentary behavior, consumption of preserved vegetable and use of antihypertensive drugs. Participants reporting at least one first-degree relative with stroke or heart attack were defined as having a family history of these diseases. Participants were asked how many hours per week they spent watching TV or reading to calculate the time of sedentary behavior per week. Trained staff members used a UA-779 digital monitor to measure blood pressure at least twice, using the mean of the 2 measurements for analyses. Self-reported diabetes or screen-detected diabetes were considered as diabetes mellitus [19]. Screen-detected diabetes was defined as measured nonfasting blood glucose ≥11.1 mmol/L or fasting blood glucose ≥7.0 mmol/L [20]. Participants with self-reported diagnosis of hypertension by a registered physician, measured systolic blood pressure ≥ 140 mmHg, measured diastolic blood pressure ≥ 90 mmHg, or self-reported use of antihypertensive medication at baseline were classified as having hypertension [4]. Duration of hypertension was calculated as age at baseline minus age at diagnosis of hypertension. If hypertension was ascertained by blood pressure at baseline, the duration of hypertension would be considered as 0 year.
## Definition of healthy lifestyle
Smoking status, alcohol intake, physical activity, diet, and body shape have been proven to be closely related to the risks of CVD. We included these five lifestyle factors to define a healthy lifestyle (21–24). The healthy group regarding smoking status was defined as nonsmokers or individuals who stopped smoking not resulting from illness [25] because there may be a misleadingly elevated risk while including those who quitted smoking due to illness. For alcohol consumption, the healthy group was defined as never drinkers, weekly drinkers, and moderate daily drinkers (i.e., drinking <25 g of pure alcohol for men and < 15 g for women per day) [26]. The healthy group for physical activity was defined as those whose physical activity level was above median after taking age- (<50 years, 50–59 years, and ≥ 60 years) and sex-specific into account. For diet, according to the Chinese Dietary Guidelines and previous findings [10, 11, 27], six food items were taken into consideration, including vegetables, fruits, eggs, red meat, grains and fish. We created a diet score according to the following criteria: eating vegetables daily, eating fruits daily, eating eggs ≥4 days every week, eating red meat 1–6 days every week, eating grains weekly, eating fish weekly. A score of 1 for those who met the criteria for each food group, a score of 0 otherwise. The diet score ranged from 0 to 6. The healthy group included participants who scored 4 to 6. For body shape, we took body weight and fat into consideration to reflect energy balance. The healthy group was defined as having a moderate BMI (18.5 ≤ BMI ≤ 27.9 kg/m2) and WC (WC < 90 cm for men and < 85 cm for women). The healthy lifestyle score ranged from 0 to 5. To avoid extreme groups with limited cases, we subsequently categorized the lifestyle scores into four groups (0–1, 2, 3, and 4–5).
## Ascertainment of outcomes
Information on CVD incidence cases since baseline recruitment was ascertained from local disease and death registries, the health insurance system, and active follow-up [14]. The health insurance system has almost universal coverage and includes detailed descriptions of diagnosis. Street committees conduct annual surveys to supplement the morbidity information of uninsured participants. Trained investigators blinded to baseline information coded all cases with the 10th revision of the International Classification of Diseases (ICD-10). Major cardiovascular events (for stroke, IHD) were reviewed and integrated centrally by cardiovascular specialists from China and the UK.
The primary outcomes were incidences of total cardiovascular diseases (CVD), ischemic heart disease (IHD) and ischemic stroke (IS). Total CVD included all circulatory diseases coded as “I” in ICD-10 (e.g., stroke, any type of heart disease, peripheral vascular disease) and were coded as I00 to I99. IHD and IS were coded as I20 to I25 and I63, which were subdivisions of total CVD.
## Statistical analysis
Participants contributed person-years from enrollment into the study until the diagnosis of CVD, loss to follow-up, or December 31, 2017, whichever came first. A Cox proportional hazards model was used to estimate the hazard ratios (HRs) and $95\%$ confidence intervals (CIs) for the associations of individual and combined lifestyle factors with risks of incidence of total CVD, IHD and IS among participants with or without hypertension. The Cox model was stratified by age at baseline in 5-year intervals.
All lifestyles were included when analyzing individual lifestyle factors. Model 1 was adjusted for sex. Model 2 was additionally adjusted for education level, occupation, marital status, family history of heart attack or stroke, time of sedentary behavior, and usage of antihypertensive drugs. Similarly, we made the same adjustments in the analysis of combined lifestyle factors. We treated the number of healthy lifestyle factors as a continuous variable to analyze the linear trend. Analyses were further stratified by age, sex, education level, household income, occupation, time of sedentary behavior, and usage of antihypertensive drugs for hypertension. The likelihood ratio test including the cross-product term was used to estimate multiplicative interactions. To demonstrate the robustness of our findings, we conducted several sensitivity analyses. First, participants who had diabetes at baseline were excluded. Second, participants whose outcome occurred in the first 2 years of follow-up were excluded. Third, participants whose BMI < 18.5 kg/m2 were excluded. Forth, for alcohol consumption, the healthy group was redefined as moderate drinking.
R software (version 4.1.0) was used to perform the statistical analyses. A two-sided $p \leq 0.05$ was considered statistically significant.
## Baseline characteristics
Table 1; Supplementary Table S1 show the characteristics of the study participants with or without hypertension (mean age 51.87 years, $58.14\%$ women). Of the 20,194 hypertensive participants, 27.18, 37.84 and $25.31\%$ had 2, 3, and ≥ 4 healthy lifestyle factors, while for the 31,727 normotensive participants, 20.91, 38.77 and $34.62\%$ had 2, 3, and ≥ 4 healthy lifestyle factors. Women were more likely to adhere to a healthy lifestyle. Married participants tended to adhere to fewer healthy lifestyle behaviors.
**Table 1**
| Baseline characteristics | Number of healthy lifestyle factors | Number of healthy lifestyle factors.1 | Number of healthy lifestyle factors.2 | Number of healthy lifestyle factors.3 | p valueb |
| --- | --- | --- | --- | --- | --- |
| Baseline characteristics | 0–1 | 2 | 3 | 4–5 | p valueb |
| Hypertension | | | | | |
| No. of participants | 1,953 (9.67) | 5,488 (27.18) | 7,642 (37.84) | 5,111 (25.31) | |
| Age, years | 53.40 (9.85) | 56.47 (9.78) | 56.89 (9.82) | 56.36 (9.75) | <0.01 |
| Women | 23 (1.18) | 2,074 (37.79) | 4,853 (63.50) | 4,293 (84.00) | <0.01 |
| Married | 1,849 (94.67) | 5,027 (91.60) | 6,819 (89.23) | 4,548 (88.98) | <0.01 |
| High school and above | 258 (13.21) | 466 (8.49) | 584 (7.64) | 452 (8.84) | <0.01 |
| Household income ≥20,000 RMB/year | 1,507 (77.16) | 3,718 (67.75) | 4,981 (65.18) | 3,346 (65.47) | <0.01 |
| Family history of heart attack or stroke | 491 (25.14) | 1,410 (25.69) | 2,067 (27.05) | 1,294 (25.32) | 0.29 |
| Low-risk lifestyle factorsa | | | | | |
| Nonsmoking | 40 (2.05) | 2,374 (43.26) | 5,653 (73.97) | 4,910 (96.07) | <0.01 |
| Nonexcessive alcohol intake | 831 (42.55) | 4,492 (81.85) | 7,410 (96.96) | 5,102 (99.82) | <0.01 |
| Being physically active | 232 (11.88) | 1,443 (26.29) | 3,584 (46.90) | 4,379 (85.68) | <0.01 |
| Healthy dietary habits | 38 (1.95) | 341 (6.21) | 1,138 (14.89) | 1,906 (37.29) | <0.01 |
| Healthy body weight and fat | 561 (28.73) | 2,326 (42.38) | 5,141 (67.27) | 4,715 (92.25) | <0.01 |
| Normotension | | | | | |
| No. of participants | 1,807 (5.70) | 6,634 (20.91) | 12,302 (38.77) | 10,984 (34.62) | |
| Age, years | 49.44 (9.33) | 50.08 (9.55) | 49.25 (9.56) | 48.13 (9.47) | <0.01 |
| Women | 20 (1.11) | 1,701 (25.64) | 7,534 (61.24) | 9,687 (88.19) | <0.01 |
| Married | 1,737 (96.13) | 6,307 (95.07) | 11,587 (94.19) | 10,344 (94.17) | <0.01 |
| High school and above | 243 (13.45) | 788 (11.88) | 1,139 (9.26) | 1,117 (10.17) | <0.01 |
| Household income ≥20,000 RMB/year | 1,459 (80.74) | 5,092 (76.76) | 9,487 (77.12) | 8,697 (79.18) | <0.01 |
| Family history of heart attack or stroke | 336 (18.59) | 1,154 (17.40) | 2,180 (17.72) | 1,939 (17.65) | 0.67 |
| Low-risk lifestyle factorsa | | | | | |
| Nonsmoking | 28 (1.55) | 1,865 (28.11) | 8,286 (67.35) | 10,519 (95.77) | <0.01 |
| Nonexcessive alcohol intake | 733 (40.56) | 5,331 (80.36) | 12,028 (97.77) | 10,961 (99.79) | <0.01 |
| Being physically active | 147 (8.14) | 1,727 (26.03) | 5,278 (42.90) | 9,103 (82.88) | <0.01 |
| Healthy dietary habits | 31 (1.72) | 340 (5.13) | 1,435 (11.66) | 4,290 (39.06) | <0.01 |
| Healthy body weight and fat | 711 (39.35) | 4,005 (60.37) | 9,879 (80.30) | 10,555 (96.09) | <0.01 |
## Associations of individual healthy lifestyle factors with the incidence of cardiovascular diseases
During a median follow-up period of 10.1 years, 6,151 incidence of total CVD cases, 1,304 IHD cases, and 2,243 IS cases were documented. When categorizing the five lifestyle factors into healthy and unhealthy groups (reference group), nonsmoking, being physically active, healthy body weight and fat were each independently associated with a 16, 8, and $10\%$ lower risk of the incidence of total CVD, 18, 20, and $18\%$ lower risk of incident IHD, and 26, 3, and $13\%$ lower risk of incident IS among participants with hypertension, respectively. Those associations were only observed between healthy dietary habits and incident CVD among normotensive participants (Table 2).
**Table 2**
| Unnamed: 0 | Hypertension (n = 20,202) | Hypertension (n = 20,202).1 | Hypertension (n = 20,202).2 | Normotension (n = 31,727) | Normotension (n = 31,727).1 | Normotension (n = 31,727).2 |
| --- | --- | --- | --- | --- | --- | --- |
| | Cases/PYs | Model 1 | Model 2 | Cases/PYs | Model 1 | Model 2 |
| Total CVD | | | | | | |
| Nonsmoking | 2,572/133,556 | 0.83 (0.75, 0.93) | 0.84 (0.75, 0.93) | 1,354/225,411 | 0.87 (0.74, 1.03) | 0.87 (0.73, 1.02) |
| Nonexcessive alcohol intake | 3,579/181,386 | 1.10 (0.99, 1.23) | 1.06 (0.95, 1.18) | 1,958/313,726 | 1.11 (0.95, 1.31) | 1.10 (0.94, 1.30) |
| Physically active | 1,728/99,070 | 0.88 (0.82, 0.93) | 0.92 (0.85, 0.99) | 1,049/175,700 | 0.91 (0.84, 0.99) | 0.97 (0.88, 1.07) |
| Healthy dietary habits | 627/34,930 | 0.89 (0.82, 0.97) | 0.89 (0.81, 0.97) | 336/65,796 | 0.84 (0.74, 0.94) | 0.84 (0.74, 0.95) |
| Healthy body weight and fat | 2,512/129,899 | 0.85 (0.80, 0.91) | 0.90 (0.84, 0.96) | 1,639/272,008 | 0.90 (0.81, 1.00) | 0.90 (0.82, 1.00) |
| Ischemic Heart Disease | | | | | | |
| Nonsmoking | 566/140,577 | 0.84 (0.67, 1.05) | 0.82 (0.66, 1.03) | 262/228,943 | 0.83 (0.58, 1.18) | 0.85 (0.59, 1.22) |
| Nonexcessive alcohol intake | 803/190,924 | 1.33 (1.03, 1.70) | 1.25 (0.97, 1.61) | 398/318,630 | 1.85 (1.21, 2.82) | 1.85 (1.21, 2.82) |
| Being physically active | 343/104,042 | 0.75 (0.66, 0.86) | 0.80 (0.68, 0.95) | 201/178,438 | 0.88 (0.72, 1.07) | 0.92 (0.73, 1.15) |
| Healthy dietary habits | 155/36,540 | 1.02 (0.85, 1.21) | 0.93 (0.77, 1.13) | 75/ 66,615 | 0.97 (0.75, 1.24) | 1.03 (0.79, 1.35) |
| Healthy body weight and fat | 534/136,510 | 0.78 (0.68, 0.89) | 0.82 (0.71, 0.94) | 315/276,162 | 0.84 (0.67, 1.05) | 0.84 (0.67, 1.05) |
| Ischemic Stroke | | | | | | |
| Nonsmoking | 975/139,064 | 0.73 (0.62, 0.88) | 0.74 (0.62, 0.88) | 442/228,501 | 1.03 (0.78, 1.36) | 1.05 (0.80, 1.39) |
| Nonexcessive alcohol intake | 1,372/188,928 | 0.97 (0.81, 1.14) | 0.92 (0.78, 1.10) | 626/318,008 | 1.06 (0.81, 1.41) | 1.07 (0.81, 1.41) |
| Being physically active | 683/102,747 | 0.90 (0.81, 1.00) | 0.97 (0.86, 1.09) | 337/178,105 | 0.92 (0.79, 1.07) | 0.96 (0.80, 1.14) |
| Healthy dietary habits | 232/36,278 | 0.86 (0.74, 0.98) | 0.88 (0.76, 1.02) | 103/66,564 | 0.80 (0.65, 0.99) | 0.84 (0.67, 1.04) |
| Healthy body weight and fat | 966/134,935 | 0.82 (0.74, 0.91) | 0.87 (0.79, 0.97) | 533/275,553 | 0.97 (0.81, 1.16) | 0.97 (0.81, 1.16) |
## Association of a healthy lifestyle pattern with the incidence of cardiovascular diseases
When considering healthy lifestyle factors jointly, compared to those with ≤1 healthy lifestyle scores, the adjusted HRs ($95\%$ CIs) of those with 4–5 scores were 0.71 ($95\%$ CI: 0.62, 0.81) for the incidence of total CVD, 0.56 ($95\%$ CI: 0.42, 0.75) for the incidence of IHD, and 0.63 ($95\%$ CI: 0.51, 0.79) for the incidence of IS among hypertensive patients (all p for trend <0.01) (Table 3). When evaluated ordinally, participants having a 1-score increment were related to a greater magnitude of total CVD, IHD and IS risk lowering among hypertensive patients than among normotensive patients. However, there were no significant multiplicative interactions between blood pressure and lifestyle factors on CVD incidence (p for interaction = 0.18 for total CVD, 0.16 for IHD, 0.06 for IS).
**Table 3**
| Category | Lifestyle score categoryb | Lifestyle score categoryb.1 | Lifestyle score categoryb.2 | Lifestyle score categoryb.3 | p for trend | HR (95% CI) per score point | p for interactionc |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Category | 0–1 | 2 | 3 | 4–5 | p for trend | HR (95% CI) per score point | p for interactionc |
| Hypertension | | | | | | | |
| Total CVD | | | | | | | 0.18 |
| Cases/PYs | 385/19,310 | 1,164/54,960 | 1,559/77,410 | 896/53,492 | | | |
| Model 1 | 1.00 | 0.85 (0.76, 0.96) | 0.79 (0.70, 0.89) | 0.66 (0.58, 0.76) | <0.01 | 0.89 (0.86, 0.92) | |
| Model 2a | 1.00 | 0.85 (0.76, 0.96) | 0.81 (0.72, 0.91) | 0.71 (0.62, 0.81) | <0.01 | 0.91 (0.88, 0.95) | |
| Ischemic Heart Disease | | | | | | | 0.16 |
| Cases/PYs | 94/20,265 | 257/57,890 | 346/81,647 | 184/56,022 | | | |
| Model 1 | 1.00 | 0.74 (0.58, 0.94) | 0.68 (0.53, 0.87) | 0.53 (0.40, 0.70) | <0.01 | 0.86 (0.80, 0.93) | |
| Model 2 | 1.00 | 0.74 (0.58, 0.95) | 0.70 (0.54, 0.90) | 0.56 (0.42, 0.75) | <0.01 | 0.88 (0.81, 0.95) | |
| Ischemic Stroke | | | | | | | 0.06 |
| Cases/PYs | 160/20,106 | 447/57,227 | 619/80,705 | 328/55,426 | | | |
| Model 1 | 1.00 | 0.78 (0.64, 0.93) | 0.73 (0.61, 0.88) | 0.57 (0.47, 0.71) | <0.01 | 0.86 (0.81, 0.91) | |
| Model 2 | 1.00 | 0.78 (0.65, 0.95) | 0.76 (0.63, 0.92) | 0.63 (0.51, 0.79) | <0.01 | 0.88 (0.83, 0.94) | |
| Normotension | | | | | | | |
| Total CVD | | | | | | | |
| Cases/PYs | 131/19,002 | 528/70,474 | 850/132,814 | 638/11,9,917 | | | |
| Model 1 | 1.00 | 1.00 (0.82, 1.21) | 0.89 (0.74, 1.08) | 0.80 (0.65, 0.98) | <0.01 | 0.91 (0.87, 0.96) | |
| Model 2 | 1.00 | 1.00 (0.82, 1.21) | 0.91 (0.75, 1.11) | 0.84 (0.68, 1.04) | 0.01 | 0.93 (0.88, 0.98) | |
| Ischemic Heart Disease | | | | | | | |
| Cases/PYs | 24/19,321 | 99/71,709 | 182/134,933 | 118/121,643 | | | |
| Model 1 | 1.00 | 1.02 (0.65, 1.60) | 1.08 (0.69, 1.69) | 0.86 (0.54, 1.39) | 0.31 | 0.94 (0.84, 1.06) | |
| Model 2 | 1.00 | 1.05 (0.67, 1.65) | 1.16 (0.74, 1.81) | 0.97 (0.59, 1.57) | 0.73 | 0.98 (0.87, 1.10) | |
| Ischemic Stroke | | | | | | | |
| Cases/PYs | 32/19,331 | 180/71,428 | 271/134,663 | 206/121,464 | | | |
| Model 1 | 1.00 | 1.39 (0.95, 2.02) | 1.18 (0.81, 1.73) | 1.10 (0.74, 1.64) | 0.14 | 0.94 (0.86, 1.02) | |
| Model 2 | 1.00 | 1.39 (0.95, 2.03) | 1.22 (0.83, 1.79) | 1.17 (0.78, 1.76) | 0.39 | 0.96 (0.87, 1.05) | |
## Stratified analyses
When stratified by age, sex, education level, household income, occupation, duration of sedentary behavior, and antihypertensive drug use, the analyses yielded consistent results (Figure 1). For participants with hypertension, adults younger than 65 years had a stronger inverse association between healthy lifestyle scores and IHD risk (p for interaction <0.01), and adults with annual household incomes less than 20,000 RMB/year had a stronger inverse association between healthy lifestyle scores and IS risk as well (p for interaction = 0.05).
**Figure 1:** *Stratified analysis of the association of incident major cardiovascular diseases (CVD) with each 1-unit increment in healthy lifestyle factors in the hypertensive population. This multivariable model was adjusted for sex, education level, marital status, household income, family history of heart attack or stroke, consumption of preserved vegetable, occupation, sedentary behavior, usage of antihypertensive drugs, and duration of hypertension at baseline. The p for interaction was calculated using multiplicative interaction terms and the likelihood ratio test.*
## Sensitivity analyses
Several sensitivity analyses were performed by excluding participants who had diabetes at baseline (Supplementary Table S2), excluding participants whose disease outcomes occurred in the first 2 years of follow-up (Supplementary Table S3), excluding participants whose BMI < 18.5 kg/m2 (Supplementary Table S4), and only considering moderate drinking as healthy (Supplementary Table S5). The risk estimates did not have materially changes among sensitivity analyses.
## Principal findings
This large, prospective cohort study of Chinese adults examined the associations of healthy lifestyle scores (i.e., nonsmoking, nonexcessive alcohol intake, being physically active, having a relatively healthy dietary habit, healthy body weight and fat) with the incidence of total CVD, IHD, and IS. Compared with 0 or 1 ideal lifestyle factors, hypertensive patients having a score of 4 or 5 showed a 29, 44, and $37\%$ reduction in the risk of total CVD, IHD, and IS, which was higher than that of normotensive individuals.
## Comparison with other studies
Our findings in hypertensive patients are consistent with previous studies in the general population (13, 23, 28–31). In Nurses’ Health Study of 15 to 20 years follow-up data, the relative risk (RR) for the healthy lifestyle factors including nonsmoking, daily moderate alcohol consumption, moderate-to-vigorous physical activity, a healthy diet, and BMI under 25 kg/m2 was 0.25 ($95\%$ CI: 0.14, 0.44) for total CVD incidence, 0.17 ($95\%$ CI: 0.07, 0.41) for coronary heart disease (CHD) incidence [28], and 0.19 ($95\%$ CI: 0.09, 0.40) for IS incidence [29]. In Swedish cohorts, a healthy pattern combination of a healthy diet, being physically active, nonsmoking, moderate daily drinking was associated with a population attributable risk of $79\%$ ($95\%$ CI: 34, $93\%$) in myocardial infarction (MI) events among men [23] and a $92\%$ ($95\%$ CI: 72, $98\%$) in MI events among women [30]. Lv et al. combined five healthy lifestyles (normal BMI and waist-to-hip ratio, participation in physical exercise, a diet rich in vegetables and fruits and limited in red meat, nonsmoking, and moderate alcohol consumption) to quantify their impacts on IHD and IS incidence in a Chinese population [13]. The HR for having 5 to 6 healthy lifestyle factors was 0.50 ($95\%$ CI: 0.41, 0.60) for IHD incidence and 0.50 ($95\%$ CI: 0.40, 0.64) for IS incidence during the 7.2-year follow-up. However, this protective effect was reduced in the normotensive population in this study. Generally, although many prospective studies have demonstrated the significance of lifestyle interventions for the prevention of CVD, they might have missed patients who already had hypertension. Meanwhile, because adherence to a healthy lifestyle can also reduce the risk of hypertension [32, 33], people without hypertension are recommended to follow a healthy lifestyle as well. Furthermore, due to the potential mediating effect of lipid profile on lifestyle and CVD [34], lipid was not included as a confounder in models, which was consistent with other studies (13, 28–30).
In stratified analyses, the association between healthy lifestyle and IHD risk was stronger among younger participants, and the association with IS risk was stronger among adults with lower household incomes, which was consistent with previous studies [35]. These results indicated that people could obtain larger benefits if they adopted healthy lifestyles at an early age or have a lower socioeconomic status. The possible reason may be that individuals of different ages and socioeconomic statuses perceive and choose healthy lifestyles differently, such as people who may choose not to smoke or drink because of financial constraints.
Previous studies have found that compared to people with moderate alcohol consumption, nondrinkers have an increased risk of CVD [13, 36]. Nevertheless, compared to nondrinking, moderate drinking can increase the risk of cancer and injury [37, 38]. Therefore, regarding overall human health, our study considered nondrinking as a healthy lifestyle. By classifying both nondrinking and moderate drinking participants into low-risk groups, we found that nonexcessive drinking had no independent protective effect on CVD. Meanwhile, the association between healthy lifestyle and CVD slightly changed after only considering moderate drinking as healthy. The reason may be that genetic variants involved in alcohol metabolism (such as rs671 variant, which is common in east Asian populations and can slow the decomposition of acetaldehyde to causes cardiovascular damage) were different in the two groups [39].
## Public health impact
For the primary prevention of CVD, this study’s healthy lifestyle pattern provides a positive framework. Our findings contribute valuable information to the prevention of CVD by five modifiable lifestyle factors in hypertensive populations. However, in this study, only less than one-third participants adopted 4 or 5 healthy lifestyles. From a public health perspective, individuals, especially hypertensive patients, could refer to our findings to better understand the significance of CVD prevention and develop healthy lifestyle habits in reference to our findings.
## Strengths and limitations
The strengths of this study included a prospective study design, a relatively large sample size of population, controlling for potential confounding factors, and the use of measured anthropometric information to provide more accurate estimates of blood pressure, BMI and WC [14]. Meanwhile, the present study has several limitations. First, lifestyle behaviors were self-reported, which may lead to some misclassification. However, there is no evidence that this type of exposure misclassification is differentially associated with CVD. Second, we created a healthy lifestyle score by using baseline lifestyle information, there is no measurement on the persistence of lifestyles during the follow-up. However, the re-survey conducted during the follow-up showed that there was good agreement between the baseline and re-survey for lifestyle variables [14]. Third, confounding such as genetic susceptibility, detailed medication use, or salt and sugar-sweetened beverage intake could not be entirely ruled out. Unmeasured or unknown factors could still cause residual confounding. Forth, some individuals who self-reported taking blood pressure medications at baseline may have met their blood pressure goals at follow-up, which may weaken the difference of protective effects of healthy lifestyles between hypertensive and normotensive population. In addition, information on adherence and persistence to antihypertensive drugs in hypertensive participants could not be confirmed during the follow-up. However, this study calculated the correlation coefficient between healthy lifestyle scores and use of blood pressure medications, and the correlation coefficient is 0.054, indicating that there was little association between taking blood pressure medications and healthy lifestyle scores at baseline (Supplementary Material S5). Finally, this study was observational, and further RCTs are needed to confirm the causal nature of the associations.
## Conclusion
This prospective cohort study of Chinese adults provided evidence that adopting a healthy lifestyle pattern, including abstinence from or cessation of smoking, nondaily drinking or daily moderate drinking, adequate physical activity, adherence to a healthy diet, and having a standard BMI and WC, is related to a significantly lower risk of the incidence of total CVD, IHD, and IS among hypertensive participants, but this association is not as pronounced among normotensive individuals.
## Data availability statement
The datasets presented in this article are not readily available because the data that support the findings of this study are available from the Department of the China Kadoorie Biobank, but restrictions apply to the availability of these data, which were used under license for the current study and are not publicly available. Data are, however, available from the authors upon reasonable request and with the permission of the Department of the China Kadoorie Biobank. Requests to access the datasets should be directed to https://www.ckbiobank.org/site.
## Ethics statement
The studies involving human participants were reviewed and approved by Ethical Review Committee of the Chinese Center for Disease Control and Prevention (Beijing, China) and the Oxford Tropical Research Ethics Committee, University of Oxford (UK). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
JS and HG designed the study. HG performed the data analyses and drafted the manuscript. JS revised the data analyses. JS, LC, XF, JZ, MW, and RT critically revised the manuscript for important intellectual content. YL, YH, JJ, YG, JL, PP, and ZC edited and proofread the manuscript. All authors read and approved the final manuscript.
## Funding
This work was supported by grants from the National Natural Science Foundation of China (82192900, 81390540, and 91846303), grants from the National Key Research and Development Program of China (2016YFC0900500), grants from the Kadoorie Charitable Foundation in Hong Kong and grants from the Wellcome Trust (088158/Z/09/Z, 104085/Z/14/Z) in the UK.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcvm.2023.1046943/full#supplementary-material
## References
1. Diseases GBD, Injuries C. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990-2019: a systematic analysis for the global burden of disease study 2019**. *Lancet* (2020.0) **396** 1204-22. DOI: 10.1016/S0140-6736(20)30925-9
2. Liu S, Li Y, Zeng X, Wang H, Yin P, Wang L. **Burden of cardiovascular Diseases in China, 1990-2016: findings from the 2016 global burden of disease study**. *JAMA Cardiol* (2019.0) **4** 342-52. DOI: 10.1001/jamacardio.2019.0295
3. Zhang M, Wu J, Zhang X, Hu CH, Zhao ZP, Li C. **Prevalence and control of hypertension in adults in China, 2018**. *Zhonghua Liu Xing Bing Xue Za Zhi* (2021.0) **42** 1780-9. DOI: 10.3760/cma.j.cn112338-20210508-00379
4. Lewington S, Lacey B, Clarke R, Guo Y, Kong XL, Yang L. **The burden of hypertension and associated risk for cardiovascular mortality in China**. *JAMA Intern Med* (2016.0) **176** 524-32. DOI: 10.1001/jamainternmed.2016.0190
5. **National Center for Chronic Disease Prevention and Health Promotion (US) Office on Smoking and Health**. *The Health Consequences of Smoking—50 Years of Progress: A Report of the Surgeon General* (2014.0)
6. Ronksley PE, Brien SE, Turner BJ, Mukamal KJ, Ghali WA. **Association of alcohol consumption with selected cardiovascular disease outcomes: a systematic review and meta-analysis**. *BMJ* (2011.0) **342** d671. DOI: 10.1136/bmj.d671
7. Zhou TY, Su J, Tao R, Qin Y, Zhou JY, Lu Y. **The association between daily total physical activity and risk of cardiovascular disease among hypertensive patients: a 10-year prospective cohort study in China**. *BMC Public Health* (2021.0) **21** 517. DOI: 10.1186/s12889-021-10551-z
8. Bennett DA, Du H, Clarke R, Guo Y, Yang L, Bian Z. **Association of physical activity with risk of major cardiovascular Diseases in Chinese men and women**. *JAMA Cardiol* (2017.0) **2** 1349-58. DOI: 10.1001/jamacardio.2017.4069
9. Martinez-Gonzalez MA, Bes-Rastrollo M. **Dietary patterns, Mediterranean diet, and cardiovascular disease**. *Curr Opin Lipidol* (2014.0) **25** 20-6. DOI: 10.1097/MOL.0000000000000044
10. Qin C, Lv J, Guo Y, Bian Z, Si J, Yang L. **Associations of egg consumption with cardiovascular disease in a cohort study of 0.5 million Chinese adults**. *Heart* (2018.0) **104** 1756-63. DOI: 10.1136/heartjnl-2017-312651
11. Du H, Li L, Bennett D, Guo Y, Key TJ, Bian Z. **Fresh fruit consumption and major cardiovascular disease in China**. *N Engl J Med* (2016.0) **374** 1332-43. DOI: 10.1056/NEJMoa1501451
12. Zheng W, McLerran DF, Rolland B, Zhang XL, Inoue M, Matsuo K. **Association between body-mass index and risk of death in more than 1 million Asians**. *New Engl J Med* (2011.0) **364** 719-29. DOI: 10.1056/NEJMoa1010679
13. Lv J, Yu C, Guo Y, Bian Z, Yang L, Chen Y. **Adherence to healthy lifestyle and cardiovascular diseases in the Chinese population**. *J Am Coll Cardiol* (2017.0) **69** 1116-25. DOI: 10.1016/j.jacc.2016.11.076
14. Chen Z, Chen J, Collins R, Guo Y, Peto R, Wu F. **China Kadoorie biobank of 0.5 million people: survey methods, baseline characteristics and long-term follow-up**. *Int J Epidemiol* (2011.0) **40** 1652-66. DOI: 10.1093/ije/dyr120
15. Chen Z, Lee L, Chen J, Collins R, Wu F, Guo Y. **Cohort profile: the Kadoorie study of chronic disease in China (KSCDC)**. *Int J Epidemiol* (2005.0) **34** 1243-9. DOI: 10.1093/ije/dyi174
16. Li LM, Lv J, Guo Y, Collins R, Chen JS, Peto R. **The China Kadoorie biobank: related methodology and baseline characteristics of the participants**. *Zhonghua Liu Xing Bing Xue Za Zhi* (2012.0) **33** 249-55. PMID: 22613372
17. 17.WHO. What is moderate-intensity and vigorous-intensity physical activity? (2010). Available at: https://apps.who.int/iris/bitstream/handle/10665/44399/9789241599979_eng.pdf;jsessionid=5F2D865F5054413F31D74CC5295FEFB3?sequence=1 (Accessed, 2022).
18. Qin C, Guo Y, Pei P, Du H, Yang L, Chen Y. **The relative validity and reproducibility of food frequency questionnaires in the China Kadoorie biobank study**. *Nutrients* (2022.0) **14** 794. DOI: 10.3390/nu14040794
19. Wang M, Gong WW, Hu RY, Pan J, Lv J, Guo Y. **Associations between stressful life events and diabetes: findings from the China Kadoorie biobank study of 500, 000 adults**. *J Diabetes Investig* (2019.0) **10** 1215-22. DOI: 10.1111/jdi.13028
20. American DA. **Diagnosis and classification of diabetes mellitus**. *Diabetes Care* (2013.0) **36** S67-74. DOI: 10.2337/dc13-S067
21. Chomistek AK, Chiuve SE, Eliassen AH, Mukamal KJ, Willett WC, Rimm EB. **Healthy lifestyle in the primordial prevention of cardiovascular disease among young women**. *J Am Coll Cardiol* (2015.0) **65** 43-51. DOI: 10.1016/j.jacc.2014.10.024
22. Guasch-Ferre M, Li Y, Bhupathiraju SN, Huang T, Drouin-Chartier JP, Manson JE. **Healthy lifestyle score including sleep duration and cardiovascular disease risk**. *Am J Prev Med* (2022.0) **63** 33-42. DOI: 10.1016/j.amepre.2022.01.027
23. Akesson A, Larsson SC, Discacciati A, Wolk A. **Low-risk diet and lifestyle habits in the primary prevention of myocardial infarction in men: a population-based prospective cohort study**. *J Am Coll Cardiol* (2014.0) **64** 1299-306. DOI: 10.1016/j.jacc.2014.06.1190
24. Zhu N, Yu C, Guo Y, Bian Z, Han Y, Yang L. **Adherence to a healthy lifestyle and all-cause and cause-specific mortality in Chinese adults: a 10-year prospective study of 0.5 million people**. *Int J Behav Nutr Phys Act* (2019.0) **16** 98. DOI: 10.1186/s12966-019-0860-z
25. Chen Z, Peto R, Zhou M, Iona A, Smith M, Yang L. **Contrasting male and female trends in tobacco-attributed mortality in China: evidence from successive nationwide prospective cohort studies**. *Lancet* (2015.0) **386** 1447-56. DOI: 10.1016/S0140-6736(15)00340-2
26. Yang YX, Wang XL, Leong PM, Zhang HM, Yang XG, Kong LZ. **New Chinese dietary guidelines: healthy eating patterns and food-based dietary recommendations**. *Asia Pac J Clin Nutr* (2018.0) **27** 908-13. DOI: 10.6133/apjcn.072018.03
27. Wang SS, Lay S, Yu HN, Shen SR. **Dietary guidelines for Chinese residents (2016): comments and comparisons**. *J Zhejiang Univ Sci B* (2016.0) **17** 649-56. DOI: 10.1631/jzus.B1600341
28. Stampfer MJ, Hu FB, Manson JE, Rimm EB, Willett WC. **Primary prevention of coronary heart disease in women through diet and lifestyle**. *New Engl J Med.* (2000.0) **343** 16-22. DOI: 10.1056/Nejm200007063430103
29. Chiuve SE, Rexrode KM, Spiegelman D, Logroscino G, Manson JE, Rimm EB. **Primary prevention of stroke by healthy lifestyle**. *Circulation* (2008.0) **118** 947-54. DOI: 10.1161/CIRCULATIONAHA.108.781062
30. Akesson A, Weismayer C, Newby PK, Wolk A. **Combined effect of low-risk dietary and lifestyle behaviors in primary prevention of myocardial infarction in women**. *Arch Intern Med* (2007.0) **167** 2122-7. DOI: 10.1001/archinte.167.19.2122
31. Knoops KT, de Groot LC, Kromhout D, Perrin AE, Moreiras-Varela O, Menotti A. **Mediterranean diet, lifestyle factors, and 10-year mortality in elderly European men and women: the HALE project**. *JAMA* (2004.0) **292** 1433-9. DOI: 10.1001/jama.292.12.1433
32. Shah S, Mac Donald CJ, El Fatouhi D, Mahamat-Saleh Y, Mancini FR, Fagherazzi G. **The associations of the Palaeolithic diet alone and in combination with lifestyle factors with type 2 diabetes and hypertension risks in women in the E3N prospective cohort**. *Eur J Nutr* (2021.0) **60** 3935-45. DOI: 10.1007/s00394-021-02565-5
33. Valenzuela PL, Carrera-Bastos P, Galvez BG, Ruiz-Hurtado G, Ordovas JM, Ruilope LM. **Lifestyle interventions for the prevention and treatment of hypertension**. *Nat Rev Cardiol* (2021.0) **18** 251-75. DOI: 10.1038/s41569-020-00437-9
34. Geng T, Zhu K, Lu Q, Wan Z, Chen X, Liu L. **Healthy lifestyle behaviors, mediating biomarkers, and risk of microvascular complications among individuals with type 2 diabetes: a cohort study**. *PLoS Med* (2023.0) **20** e1004135. DOI: 10.1371/journal.pmed.1004135
35. Zhang YB, Pan XF, Chen J, Cao A, Xia L, Zhang Y. **Combined lifestyle factors, all-cause mortality and cardiovascular disease: a systematic review and meta-analysis of prospective cohort studies**. *J Epidemiol Community Health* (2021.0) **75** jech-2020-214050-9. DOI: 10.1136/jech-2020-214050
36. Bell S, Daskalopoulou M, Rapsomaniki E, George J, Britton A, Bobak M. **Association between clinically recorded alcohol consumption and initial presentation of 12 cardiovascular diseases: population based cohort study using linked health records**. *BMJ* (2017.0) **356** j909. DOI: 10.1136/bmj.j909
37. Smyth A, Teo KK, Rangarajan S, O'Donnell M, Zhang X, Rana P. **Alcohol consumption and cardiovascular disease, cancer, injury, admission to hospital, and mortality: a prospective cohort study**. *Lancet* (2015.0) **386** 1945-54. DOI: 10.1016/S0140-6736(15)00235-4
38. Bagnardi V, Rota M, Botteri E, Tramacere I, Islami F, Fedirko V. **Light alcohol drinking and cancer: a meta-analysis**. *Ann Oncol* (2013.0) **24** 301-8. DOI: 10.1093/annonc/mds337
39. Cho Y, Shin SY, Won S, Relton CL, Davey Smith G, Shin MJ. **Alcohol intake and cardiovascular risk factors: a Mendelian randomisation study**. *Sci Rep* (2015.0) **5** 18422. DOI: 10.1038/srep18422
|
---
title: Discovery of drug–omics associations in type 2 diabetes with generative deep-learning
models
authors:
- Rosa Lundbye Allesøe
- Agnete Troen Lundgaard
- Ricardo Hernández Medina
- Alejandro Aguayo-Orozco
- Joachim Johansen
- Jakob Nybo Nissen
- Caroline Brorsson
- Gianluca Mazzoni
- Lili Niu
- Jorge Hernansanz Biel
- Cristina Leal Rodríguez
- Valentas Brasas
- Henry Webel
- Michael Eriksen Benros
- Anders Gorm Pedersen
- Piotr Jaroslaw Chmura
- Ulrik Plesner Jacobsen
- Andrea Mari
- Robert Koivula
- Anubha Mahajan
- Ana Vinuela
- Juan Fernandez Tajes
- Sapna Sharma
- Mark Haid
- Mun-Gwan Hong
- Petra B. Musholt
- Federico De Masi
- Josef Vogt
- Helle Krogh Pedersen
- Valborg Gudmundsdottir
- Angus Jones
- Gwen Kennedy
- Jimmy Bell
- E. Louise Thomas
- Gary Frost
- Henrik Thomsen
- Elizaveta Hansen
- Tue Haldor Hansen
- Henrik Vestergaard
- Mirthe Muilwijk
- Marieke T. Blom
- Leen M. ‘t Hart
- Francois Pattou
- Violeta Raverdy
- Soren Brage
- Tarja Kokkola
- Alison Heggie
- Donna McEvoy
- Miranda Mourby
- Jane Kaye
- Andrew Hattersley
- Timothy McDonald
- Martin Ridderstråle
- Mark Walker
- Ian Forgie
- Giuseppe N. Giordano
- Imre Pavo
- Hartmut Ruetten
- Oluf Pedersen
- Torben Hansen
- Emmanouil Dermitzakis
- Paul W. Franks
- Jochen M. Schwenk
- Jerzy Adamski
- Mark I. McCarthy
- Ewan Pearson
- Karina Banasik
- Simon Rasmussen
- Søren Brunak
- Philippe Froguel
- Philippe Froguel
- Cecilia Engel Thomas
- Ragna Haussler
- Joline Beulens
- Femke Rutters
- Giel Nijpels
- Sabine van Oort
- Lenka Groeneveld
- Petra Elders
- Toni Giorgino
- Marianne Rodriquez
- Rachel Nice
- Mandy Perry
- Susanna Bianzano
- Ulrike Graefe-Mody
- Anita Hennige
- Rolf Grempler
- Patrick Baum
- Hans-Henrik Stærfeldt
- Nisha Shah
- Harriet Teare
- Beate Ehrhardt
- Joachim Tillner
- Christiane Dings
- Thorsten Lehr
- Nina Scherer
- Iryna Sihinevich
- Louise Cabrelli
- Heather Loftus
- Roberto Bizzotto
- Andrea Tura
- Koen Dekkers
- Nienke van Leeuwen
- Leif Groop
- Roderick Slieker
- Anna Ramisch
- Christopher Jennison
- Ian McVittie
- Francesca Frau
- Birgit Steckel-Hamann
- Kofi Adragni
- Melissa Thomas
- Naeimeh Atabaki Pasdar
- Hugo Fitipaldi
- Azra Kurbasic
- Pascal Mutie
- Hugo Pomares-Millan
- Amelie Bonnefond
- Mickael Canouil
- Robert Caiazzo
- Helene Verkindt
- Reinhard Holl
- Teemu Kuulasmaa
- Harshal Deshmukh
- Henna Cederberg
- Markku Laakso
- Jagadish Vangipurapu
- Matilda Dale
- Barbara Thorand
- Claudia Nicolay
- Andreas Fritsche
- Anita Hill
- Michelle Hudson
- Claire Thorne
- Kristine Allin
- Manimozhiyan Arumugam
- Anna Jonsson
- Line Engelbrechtsen
- Annemette Forman
- Avirup Dutta
- Nadja Sondertoft
- Yong Fan
- Stephen Gough
- Neil Robertson
- Nicky McRobert
- Agata Wesolowska-Andersen
- Andrew Brown
- David Davtian
- Adem Dawed
- Louise Donnelly
- Colin Palmer
- Margaret White
- Jorge Ferrer
- Brandon Whitcher
- Anna Artati
- Cornelia Prehn
- Jonathan Adam
- Harald Grallert
- Ramneek Gupta
- Peter Wad Sackett
- Birgitte Nilsson
- Konstantinos Tsirigos
- Rebeca Eriksen
- Bernd Jablonka
- Mathias Uhlen
- Johann Gassenhuber
- Tania Baltauss
- Nathalie de Preville
- Maria Klintenberg
- Moustafa Abdalla
journal: Nature Biotechnology
year: 2023
pmcid: PMC10017515
doi: 10.1038/s41587-022-01520-x
license: CC BY 4.0
---
# Discovery of drug–omics associations in type 2 diabetes with generative deep-learning models
## Abstract
The application of multiple omics technologies in biomedical cohorts has the potential to reveal patient-level disease characteristics and individualized response to treatment. However, the scale and heterogeneous nature of multi-modal data makes integration and inference a non-trivial task. We developed a deep-learning-based framework, multi-omics variational autoencoders (MOVE), to integrate such data and applied it to a cohort of 789 people with newly diagnosed type 2 diabetes with deep multi-omics phenotyping from the DIRECT consortium. Using in silico perturbations, we identified drug–omics associations across the multi-modal datasets for the 20 most prevalent drugs given to people with type 2 diabetes with substantially higher sensitivity than univariate statistical tests. From these, we among others, identified novel associations between metformin and the gut microbiota as well as opposite molecular responses for the two statins, simvastatin and atorvastatin. We used the associations to quantify drug–drug similarities, assess the degree of polypharmacy and conclude that drug effects are distributed across the multi-omics modalities.
Clinical multi-omics data are integrated and analyzed using a generative deep-learning model.
## Main
Drug-response patterns in individuals with complex disease, such as type 2 diabetes (T2D), are intricate. Multiple organs and confounders are typically involved including comorbidities and polypharmacy1,2. Conversely, treatment with one or more drugs and the associated polypharmacy effects can have considerable impact on the molecular profile of the individual; however, such changes are still largely unknown3. The increasing availability of deep phenotyping and multi-omics screening has proven to be beneficial in the characterization of T2D and other diseases4–7, and offer the opportunity to gain mechanistic insights on the action of drugs on disease processes.
Cohort studies can be highly useful for investigating associations between drugs and molecular phenotypes, and can be used to tailor the design of randomized control studies to assess direct causal relationships8. Common approaches to analysis of cohort data apply univariate statistical methods, linear and logistic regression, dimensionality reduction and clustering analyses. However, when expanding to multi-omics data such analyses are not straightforward and traditional methods of data interpretation are insufficient to exploit the full scope of multi-modality data.
Here we investigate vertical data integration, where multiple omics datasets have been generated for the same samples. Challenges that must be overcome include integration of data across multiple continuous and discrete data modalities, efficient handling of missing data or even large missing parts of specific data types, differences in dimensionality, modality-specific noise and how to extract associations across data modalities9–11. There are several strategies for vertical integration of multi-modal datasets, such as element-wise addition of one dataset at a time, learning individual representations for each dataset before fusion, or multi-dimensional fusion where representations are learned from the input data altogether9,12–14. Examples are multi-omics factor analysis (MOFA), iCluster, and data integration analysis for biomarker discovery using latent components (DIABLO) implemented in mixOmics, which can integrate multiple modalities11,14–16. However, these methods primarily focus on discovering factors or latent variables that can be used for visualization, clustering, or prediction of disease.
We have previously developed a deep-learning framework on the basis of variational autoencoders (VAE)17,18 for integration and binning of large amounts of unstructured metagenomics data19. Specifically, a VAE is based on deep neural networks and learns to transform high-dimensional data into a lower-dimensional space, termed a latent representation. During this process the two networks of the VAE learn the structure of input data and associations between the input variables. In our previous study, we found that the VAE could learn to integrate two datasets without any prior knowledge or statistical model19. Similarly, others have shown the capabilities of VAEs as integrative models for extracting the underlying signal in data for improving clustering and prediction12,20–23, as well as for handling large proportions of missing data24. We, therefore, speculated that such a model could be used to integrate even deeper cohort-level multi-omics datasets. While previous studies have primarily focused on stratifying patients using the underlying latent representation22,25,26 we were also interested in whether we could acquire insights into the complex relationships that the network learns through data integration.
For this purpose, we exploited that the decoder of the VAE is a generative model. Thus, the final trained decoder will be able to generate new examples of data from the learned latent distribution. On the basis of this principle, a variety of generative models have been used to generate new examples of data, such as single-cell RNA data and artificial human chromosomes27,28. Additionally, when combined with Bayesian decision theory they have been used for analysis of single-cell RNA data on the basis of variational inference29–31. Generative models also allow investigation of the effect that a virtual perturbation of the input data will have on the generated examples. For instance, Yeo et al. trained a generative model on single-cell RNA time-series data and then perturbed the input data to identify the effect of the perturbation on the output of the generative model32. Similarly, a recent study used the generative model of a VAE trained on protein evolutionary data to predict the effect that genetic variants have on the fitness of human proteins33. For our multi-modal data, we hypothesized that the generative ability of the VAE would allow us to identify associations between, for example, patient exposures and omics features.
We therefore developed a framework that is based on VAEs that we applied to a cohort of 789 people with newly diagnosed T2D with extensive multi-omics characterization. These modalities included genomics, transcriptomics, proteomics, metabolomics, and microbiomes as well as data on medication, diet questionnaires, and clinical measurements. Our method was able to integrate multi-omics data with clinical and categorical data and was resistant to systematic biases in the data as well as large amounts of missing data. Using an ensemble of generative VAE models, feature perturbation, univariate statistical methods, and Bayesian decision theory we identify cross omics associations. We compared the drug multi-omics profiles and showed that different drugs are associated with unique clinical and molecular profiles. Our method, multi-omics variational autoencoders (MOVE) is freely available, easily scalable, can integrate any number of categorical and continuous datasets, and able to identify features to multi-omics associations.
## Designing a VAE for multi-omics data integration
We used a dataset of 789 newly diagnosed T2D individuals with extensive multi-omics characterization (Supplementary Table 1). In total the data included 8,807 variables per individual with median missingness within an omics dataset of less than $5\%$ except for metagenomics data where two thirds of the individuals [532] did not have any data (Supplementary Data 1 and Supplementary Fig. 1). Therefore, these individuals had up to $24.7\%$ missingness across the multi-omics data. For the clinical data missingness was higher with a per individual median of $14\%$ and $7\%$ for continuous and categorical clinical data, respectively. We designed the MOVE framework to be flexible in relation to the number of input data types and to be able to handle both continuous and categorical features (Fig. 1a). To identify the optimal hyperparameters that would capture the structure of the data without losing the ability to generalize on unseen individuals, we initially divided the dataset into training and test sets. We then measured the ability of the models to reconstruct the input as well as the stability when refitting the model to the data several times (Supplementary Figs. 2–4). The median reconstruction accuracies were between 0.95–1 and the final models were highly stable when retrained five times with average change of cosine similarities in the latent space of 0.037. Thus, the VAE models were able to reconstruct the data with high accuracy across the individuals (Supplementary Fig. 5).Fig. 1Integrating multi-omics data with a VAE.a, Principle of integration and analysis approach using MOVE. Individual-level non-omics and multi-omics data were used as input to a VAE. The optimal network hyperparameters were estimated from the summed test set error across all individuals in the test (test likelihood), training reconstruction accuracy, and model stability. Significant drug–omics associations were identified by perturbing drug status from no [0] to yes [1] for all individuals that were not already administered the drug. b, UMAP representation of the latent representation from the 789 people with newly diagnosed T2D. Individuals were colored according to their z-scaled Matsuda index from low (blue), average (yellow), and high (red). c, Overlap in significant drug–omics associations between standard t-test (two-sided, Benjamini–Hochberg FDR < 0.01) on the input data, MOVE t-test (multi-stage Bonferroni-corrected, P adjust < 0.05) and MOVE Bayes approaches (FDR Bayes < 0.05). The different methods of multiple testing correction corresponded to FDR of 0.05 on the ground-truth dataset. The overlap between MOVE t-test and MOVE Bayes was used for further analysis ($$n = 573$$). d, The number of significant associations found between drugs and features in the multi-omics datasets using MOVE t-test and MOVE Bayes (purple), t-test (green) or ANOVA (orange). See c for information on the tests. e, Fraction of features in the multi-omics datasets that was found by MOVE to be significantly associated with at least one drug ($$n = 20$$). The lower and upper hinges correspond to the first and third quartiles. The upper and lower whiskers extend from the hinge to the highest and lowest values, respectively, but no further than 1.5× interquartile range from the hinge. Data beyond the ends of whiskers are outliers and are plotted individually. Source data
## The latent space contains important clinical signatures
To illustrate how well the model captured the structure of the clinical data, we analyzed the neural network weights connected to the input variables of the encoder. Here we found the majority of the clinical and dietary variables to be among the top 50 most important (Supplementary Fig. 6). This was also the case when we investigated how the continuous features impacted the positioning of the individuals in the latent space using a Shapley additive explanation (SHAP) analysis34, whereas for discrete features we found T2D-associated genetic variants as well as clinically related features to be important (Supplementary Fig. 7). Then, we investigated how individuals would be differentiated by characteristics such as insulin sensitivity quantified by the Matsuda index (Fig. 1b). Here we found a trend of the Matsuda Index correlating with the two uniform manifold approximation and projection (UMAP) dimensions using Pearson’s correlation coefficient (PCC) of 0.34 and −0.35 for dimensions one and two, respectively. Using k-nearest-neighbor (kNN) regression on the latent representation we found that R2 for Matsuda Index ($k = 5$) was 0.70 compared to 0.37–0.38 when using residualized data or dimensionality reduction using principal component analysis (PCA) and that this trend was consistent for larger k (Supplementary Figs. 8 and 9). This indicated that the MOVE latent representation captured a clinical signal that was not as easily identified from the residualized data or by using PCA for dimensionality reduction. Furthermore, we did not find any strong local effects of missingness (R2 = 0.05 at $k = 5$) and only small effects of age (R2 < 0.01, $k = 100$). Similarly, we used a kNN classifier to investigate the effect of the confounders sex and recruitment center on the global structure of the latent representation. These achieved accuracies of 0.58 and 0.25 for sex and center, respectively, which should be compared to by-chance accuracies of 0.50 and 0.17, respectively (Supplementary Figs. 10 and 11). If we used non-residualized data, that is, when not correcting for confounding effects including age, sex, and center, we observed larger effects (Supplementary Figs. 10 and 11). This demonstrates the ability of the VAE to integrate heterogeneous data but also that substantial confounding factors can influence the latent representation.
## Extracting drug to clinical and multi-omics associations
We then investigated if the model had learned associations between the clinical, drug and multi-omics data. To do this, we developed an approach that is based on perturbating input features one at a time (Fig. 1a). For instance, to identify associations between a particular drug and all other features, we simulated that we gave the drug to each of the individuals that did not receive the drug. In addition to excluding individuals that were already receiving the drug we also excluded individuals taking a drug of the same therapeutic drug-class in the anatomical therapeutic chemical classification (ATC) system (Supplementary Table 2). We then assessed if the change in each of the feature reconstructions was significantly different compared to when passing the original data through the model (Fig. 1a). Because VAE models are stochastic, we used results across an ensemble of models and developed two different approaches to identify significant associations. One approach was based on applying t-tests with Bonferroni correction across four different models, where each model was refitted 10 times (MOVE t-test), while we also, inspired by earlier variational work29–31, used Bayesian decision theory and a single model refitted 30 times (MOVE Bayes). To identify different parameters of the approaches that would allow for comparison across and to standard methods (t-test, analysis of variance (ANOVA)), we applied them to two datasets consisting of randomized clinical, drug and multi-omics data. Our findings showed that MOVE t-test and MOVE Bayes had good performance to identify drug–omics associations compared with t-test and ANOVA at a ground-truth false discovery rate (FDR) of 0.05 (Supplementary Fig. 12 and Supplementary Table 3 and Methods).
## MOVE identifies drug and multi-omics associations
We then applied the MOVE framework to identify drug associations in the DIRECT multi-modal data. The two methods, MOVE t-test and MOVE Bayes, identified 3,143 and 763 significant associations to the multi-omics and clinical features, respectively (Supplementary Tables 4–6 and Supplementary Data 2–4). We analyzed the intersection of the two approaches and found that 573 of the 763 ($75\%$) of the significant associations were found by both methods (Fig. 1c). Making a conservative choice, we used the associations identified by both methods for further analyses. When compared to traditional tests such as the Student’s t-test and ANOVA we found this to add $211\%$ more significant associations, from 184 to 573 (Fig. 1d). In addition, the significant associations identified by MOVE were distributed across the drugs (two-sided t-test, $$P \leq 0.016$$) and not only for the drugs administered to most individuals such as Simvastatin, Atorvastatin, and Metformin. For instance, MOVE identified a median of 20 associations per drug compared to 1 for t-test and 0 for ANOVA, highlighting that our method was more sensitive for extracting associations for drugs given to a smaller number of individuals (Supplementary Tables 5 and 6). Among the multi-omics datasets, we found that the largest number of significant drug associations was to the metabolomics, clinical, and transcriptomics data with an average of six associations per drug (Fig. 1e and Supplementary Fig. 13). When normalizing for all possible associations, the highest fraction of associations was to the clinical data ($8\%$) followed by targeted and untargeted metabolomics with an average of $5.1\%$ and $2.8\%$ of the features associated to a drug, respectively. Finally, we investigated if our results could be driven by disease subtypes within the T2D cohort. To do this, we used four archetype clusters from Wesolowska–Andersen and Brorsson et al.7 that were based on clustering from 32 clinical features. Here we found that a median of $6.5\%$ of the significant drug–omics associations were specific to one of the subgroups indicating that the associations were not primarily driven by the archetypes (Supplementary Table 7).
## Changes in T2D biomarkers were associated with metformin
We then investigated drug and multi-omics interactions (Fig. 2a and Supplementary Figs. 14–18), and initially focused on expected clinical drug interactions. For instance, for metformin, we identified 88 significant clinical and multi-omics interactions across all the datasets. When investigating associations across the individuals we found low intra-patient variability indicating that the changes were stable (Fig. 2b and Supplementary Fig. 19). We found that metformin was significantly associated with 12 clinical markers of T2D such as insulin clearance, active GLP-1, glucose levels from mixed-meal glucose tolerance test, glucose sensitivity, and blood pressure (Fig. 2a and Supplementary Data 2–4). The directions of some of the associations were opposite to the expected metformin effects, that is, metformin was associated with decreased glucose sensitivity at baseline (average Z-score change −0.029, confidence intervals [−0.030, −0.029]). This could be due to confounding by indication in terms of the study design where newly diagnosed T2D individuals that have been prescribed metformin are expected to have more severe clinical T2D values compared to individuals not needing medical treatments35,36. Therefore, since all individuals have T2D the confounding effect of their diabetic status could not be disentangled from the effect of metformin. When investigating the multi-omics associations of metformin we found two of the seven associated proteins (ERAP2 and CD40L) could be linked to the immune system (Fig. 3a and Supplementary Data 4). Similarly, for the transcriptomics data we found CXCL8 and CD177 to be altered by metformin where the former has been shown to be altered in healthy individuals and cancer patients37–39. In the targeted metabolomics data we identified a significant enrichment of metabolites associated with aminoacyl-tRNA biosynthesis (hypergeometric test, $$P \leq 2.2$$ × 10−4, FDR corrected). This pathway has previously been associated with metformin in functional pathway analysis of microbial change in mice40. Finally, for the untargeted metabolomics data, metformin had the highest number of associations of any drug (22 associations) indicating that new metabolic effectors of metformin treatment could potentially be identified (Supplementary Fig. 17 and Supplementary Table 4).Fig. 2Significant associations between drugs, clinical, and multi-omics features.a, Significant associations between drugs and clinical features. Effects are given as effect size (z-scaled units) from negative (blue) to positive (red). Significant associations identified by both MOVE t-test and MOVE Bayes are indicated using a star. Features (y-axis) and drugs (x-axis) are clustered using hierarchical clustering on the basis of Euclidean distances. b, As in a but showing per individual-level associations of metformin to multi-omics features demonstrating that associations are highly stable across individuals. Features (y-axis) and newly diagnosed T2D individuals (x-axis).Source dataFig. 3Drug associations with metagenomics species and drug–drug similarities.a, Display of effect sizes (z-scaled units) for (outer to inner) metformin, simvastatin, atorvastatin, omeprazole, lansoprazole, paracetamol, and codeine. Only significant associations to any of the drugs are shown and effect size is visualized as brown (negative), gray (none), and green (positive). Selected omics features are indicated. The Gene Ontologies element represents significantly over-represented Gene Ontology terms using transcriptomics (hypergeometric test, FDR < 0.05) (green). The innermost ring indicates SHAP importance for the individual features in the encoding from input data to the latent representation. b, Effect size (z-scaled units) (x-axis) of the human gut metagenomics species that were significantly associated with metformin (orange) or omeprazole (teal). c, Drug–drug similarities by comparing drug-response profiles across the multi-omics datasets. Cosine similarity indicated from no similarity (blue) to identical profiles (red). d, Average effect (z-score) of drugs for the omics datasets. All 20 drugs are shown, however, only metformin (red), omeprazole (purple), atorvastatin (green), and simvastatin (blue) are indicated. All other drugs are colored gray without a text label. e, Distribution of multi-omics ranks for the different drugs. The ranks are determined as a number between 1–20 (drugs) on the basis of the average effect size from d. The boxes are colored according to number of individuals taking a particular drug from 0 (white) to 323 (purple). There was no correlation between rank scores and number of individuals taking a drug (PCC = 0.14). The lower and upper hinges correspond to the first and third quartiles. The upper and lower whiskers extend from the hinge to the highest and lowest values, respectively, but no further than 1.5× interquartile range from the hinge. Data beyond the ends of whiskers are outliers and are plotted individually. Source data
## Association of metformin and omeprazole with gut microbiota
Recent studies have shown how drug intake can influence the human gut microbiome composition41,42. Here we found metformin and omeprazole to be the only drugs to have significant associations to the metagenomics data with an increase of eleven metagenomics species as well as a decrease of six other species (Fig. 3b). Remarkably, the findings of increased *Escherichia coli* and decreased levels of *Intestinibacter bartlettii* and Peptostreptococcaceae sp. have been reported in healthy individuals taking metformin in an intervention study43 (Supplementary Data 4). As the study first reporting the findings was performed in healthy individuals, the changes are most likely not explained by other factors than metformin treatment. For omeprazole, a protein pump inhibitor (PPI), we identified three *Streptococcus species* to be significantly increased (Streptococcus sp., Streptococcus parasanguinis, and Streptococcus vestibularis) (Supplementary Data 4). Previous work by others has specifically shown PPIs to influence the abundance of *Streptococcus parasanguinis* and vestibularis in the human gut44. Interestingly, both omeprazole and lansoprazole target the K-transporter ATPase alpha channel 1 and increases pH in the stomach. The two drugs, however, have different speed to effect rates where omeprazole elicits its effect with a slower rate compared to lansoprazole45. This, in combination with more individuals being administered omeprazole [125] compared to lansoprazole [57], could explain why we identified significant alterations of gut microbiota for omeprazole and not lansoprazole.
## Statins were associated with decreased low-density lipoprotein and cholesterol
Next, we investigated associations between the two statins, simvastatin, and atorvastatin, which are widely used to treat high blood cholesterol by lowering low-density lipoprotein (LDL)46. In agreement with their potential to treat dyslipidemia, we found both LDL and overall cholesterol levels to be significantly associated and decreased with average LDL z-score change of −0.039 (CI [−0.040, −0.038]) and −0.015 (CI [−0.016, −0.014]) for simvastatin and atorvastatin, respectively (Supplementary Data 4). This effect could be a consequence of many of the participants having been administered statins before their T2D diagnosis (simvastatin median duration 1.9 years and atorvastatin median duration 1.7 years; Supplementary Table 8), thereby increasing the chance of observing the effect of the drug with reduced confounding by indication. Interestingly, we noticed that besides the downregulation of LDL and general cholesterol levels some of the remaining clinical associations were not similar. Simvastatin was associated with an increase in the health marker high-density lipoprotein (HDL) cholesterol whereas atorvastatin had a decrease. This agrees with known effects of the two statins on HDL, where simvastatin and atorvastatin, respectively, increase and decrease HDL levels with increasing doses47.
## Different molecular profiles of simvastatin and atorvastatin
When investigating the multi-omics associations, the two statins had diverse effects across the omics data (Fig. 3a and Supplementary Figs. 14–18 and 20). In agreement with the analysis of the clinical data, we found simvastatin to be significantly associated with downregulation of cholesterol homeostasis (Hypergeometric test, $$P \leq 0.005$$, FDR) and lipid transportation pathways (Hypergeometric test, $$P \leq 0.002$$, FDR) from the enrichment analysis of the associated transcripts (Fig. 3a and Supplementary Data 4 and 5). Specifically, we identified changes in LDLR, SREBF2, ABCA1, and ABCG1 expression, previously associated with simvastatin usage and accumulation of fatty acid and triglyceride in the liver through different pathways48–52 (Supplementary Data 4). In the proteomics data of atorvastatin, we identified known associations to FADS1 (ref. 53), as well as EIF2AK3, which has been reported associated with cholesterol homeostasis54,55. Additionally, two insulin growth factor binding proteins (IGFBP1 and IGFBP4) were associated with atorvastatin and IGFBP4 for simvastatin as well (Supplementary Data 4). These have previously been reported specifically for people with T2D and atorvastatin use54,56. Finally, in the targeted metabolomics data, we identified simvastatin to be associated with an increase in glycine levels, which in low systemic concentration has been associated with obesity and T2D57 (Supplementary Data 4). Furthermore, we observed a decrease of several phosphatidylcholines (11 of 17 decreased metabolites), and an increase of sphingomyelin and ceramide (2 of 11 increased metabolites), a ratio which has previously been shown to be altered with high doses of simvastatin compared to other statins58 (Supplementary Data 2–4). For atorvastatin, we observed a non-significant decrease of glycine levels and that the overall ratio of sphingomyelin and ceramide decreased (4 of 13 decreased metabolites).
## Drug polypharmacy and similarity across multi-omics data
We then investigated similarities between drugs and their multi-omics associations. Overall, we observed four clusters containing three to six drugs each and found that some of the drugs within a cluster could potentially be associated with polypharmacy (Fig. 3c). Therefore, we investigated the impact of a drug–drug combination on the associations and found a correlation between overall drug association similarity and the individuals taking the two drugs (PCC 0.75, P value of 2.2 × 10−35). This finding indicates possible polypharmacy effects introduced by taking the two drugs together resulting in a higher drug–drug similarity across all clinical and multi-omics changes. However, some of the similarities might to some extent be driven by overlapping patient groups and non-drug-related similarities such as the underlying reason for taking the drug. An example could be the drug similarity cluster of Ramipril, Acetylsalicylic Acid, Bisoprolol, Amlodipine and Atorvastatin, which can be linked to cardiovascular diseases. Furthermore, the drugs that had the most similar drug and multi-omics associations were codeine and paracetamol with a cosine similarity of 0.78. Most (38 of 46) of the individuals in the cohort taking codeine were also taking paracetamol while a large fraction of individuals (52 of 90) was only taking paracetamol. We therefore cannot rule out that the correlated multi-omics profiles of the two drugs could be driven by the partial overlap leading to similar latent representation and model reconstructions. Finally, we investigated known drug–drug interactions and association with drug multi-omics profiles; however, found no statistically significant correlations (Supplementary Note and Supplementary Fig. 21).
## The effects of drugs are widespread across the omics data
Currently, there are widespread efforts in investigating drugs and gut microbiome interactions suggesting that the microbiome is a potential target and mediator of drug effect42,59,60. As we investigated several multi-omics datasets besides the gut microbiome (metagenomics), we can compare the effect size of the drugs across the omics datasets. Interestingly, we found that the gut microbiome was the dataset with the second fewest number of statistically significant hits across the drugs with 17 significant associations (Supplementary Table 4 and Supplementary Fig. 13). Only diet and wearable data had fewer associations [11]; transcriptomics, proteomics, targeted, and untargeted metabolomics had between 44–134 significant associations. We then asked if the effect size of the drugs were different across datasets and determined the cumulative effect size of the drugs in the respective multi-omics datasets. Here we found that the average effect sizes in transcriptomics and metagenomics data were the lowest for all drugs, and that those in the metagenomics dataset were significantly lower compared to all other omics datasets but transcriptomics (ANOVA, Tukey HSD test, adjusted $P \leq 0.05$) (Fig. 3d and Supplementary Table 9). When we subset to significant drug–omics associations, of which the gut microbiome only had two drugs with significant associations (metformin and omeprazole), we found that the effect of these two drugs were similar or lower compared to the effect sizes of the other multi-omics datasets (Supplementary Fig. 22). Finally, we investigated if this could be caused by increased uncertainty when learning and reconstructing a given modality but only found small correlations with PCCs of −0.15 to 0.16 between modality uncertainty and inferred effect sizes in a modality (Supplementary Table 10). Overall, this observation implies that the multi-omics response to drug stimuli are not only targeting the gut microbiome and that multiple omics datasets should be included when attempting to understand drug effects.
## Ranking the impact of drugs in multi-omics data
Finally, we investigated the effect sizes of the individual drugs across the multi-omics datasets. We found that metformin and omeprazole, in general, had the most pronounced effects on the multi-omics data (cumulative rank scores) and that the two statins ranked 14 and 20 out of the 20 drugs (Fig. 3e) where simvastatin had the lowest overall rank of cumulative effect sizes. This analysis was not confounded by the number of individuals taking a particular drug as there was no correlation (PCC = 0.14) between the number of individuals and drug effect. This was opposed to when investigating only significant associations where statins ranked 2 and 4 with high effect sizes (Supplementary Figs. 22 and 23). This observation may indicate that statins had fewer strong effects, whereas, for instance, both metformin and omeprazole with the highest average rank had larger systemic effects.
## Discussion
Here we show that it is possible to use unsupervised deep learning to integrate and extract associations from a deeply phenotyped cohort of people with T2D. While existing methods for vertical integration of multi-omics data focus on encoding the data to factors or latent representations that can be used for clustering and classification, we took this further by using the generative capacity of VAE models. In comparison to traditional univariate statistical tests, MOVE can identify significant drug–omics associations for a wider selection of drugs. We believe that these improvements come from the ability of the generative models to infer multi-omics changes for individuals not receiving a drug thus increasing power.
Previous work to stratify the newly diagnosed T2D individuals from this cohort used 32 clinical features to identify four archetypes representing different T2D subtypes7. In addition, they used metformin status of the individuals to investigate if the subgroups were confounded by metformin treatment and found no significant impact on the clusters and their multi-omics correlations. In contrast to their work, we added medication data on 19 additional drugs and used all data as input to our unsupervised deep-learning model allowing the model to learn from all inputs simultaneously. Thus, we were able to identify associations between the drugs and multi-omics data, including for metformin indicating the importance of vertical integration.
The cross-sectional design and clinical data-guided medical decisions make it difficult to assess the directionality of drug associations and further complicates causal inference. Hence, it is not possible to draw causal conclusions on drug effects; however, the results can be considered as input to design informed studies as well as randomized clinical control studies. In the future, expansion with longitudinal multi-omics data and modeling time could add more information on the causality of the drugs by investigating the long-term effects and associations32.
Similarly, our approach opens up for individualized analysis of patients in an N-of-1 approach61. It is well-known in health care that often selecting a drug or treatment in a situation at the same time excludes performing the control experiment of using another drug. Using MOVE, we can in principle ask what would happen if we gave the patient a drug and compare to the result of choosing another drug. Our cohort size is limited, but for larger cohorts of tens to hundreds of thousands of patients this could potentially be powerful to identify molecular associations and treatment outcomes for individual patients.
Finally, we emphasize that our approach is, of course, not limited to drug associations; in principle, all the omics data could be assessed for associations across the datasets. We therefore believe that our generative method opens new possibilities in big multi-omics data analysis for discoveries of potential new biomarkers, carrying out gedankenexperiments, and investigating potential direct effects of drugs in high dimensionality molecular data that leads to testable hypotheses.
## The cohort
The cohort and available data included in the study are described in detail in Koivula et al.62,63 and Wesolowska–Andersen and Brorsson et al. ( ref. 7). In brief, we used the newly diagnosed sub-cohort of the IMI-DIRECT study consisting of 789 participants. Fifty-eight percent of participants was male and participants had the following characteristics at baseline: age 62 (8.1) years; body mass index 30.5 (5.0) kg m−2; fasting glucose 7.2 (1.4) mmol l−1; 2 h glucose 8.6 (2.8) mmol l−1. Participants were diagnosed within 2 years before recruitment and had glycated hemoglobin (HbA1c) < 60.0 mmol mol−1 (<$7.6\%$) within the previous 3 months. All samples represent distinct individuals. Furthermore, while Wesolowska–Andersen and Brorsson et al.7 used data from baseline and follow up at 18 and 36 months we only used baseline data for modeling. In addition to the baseline data from Wesolowska–Andersen and Brorsson, we carried out extensive curation and harmonization of the medication records included in the electronic case forms by the research nurses in the different recruitment centers and thus used standardized ATC annotated medication data for the individuals (see further detail below). Approval for the study protocol was obtained from each of the regional research ethics review boards separately (Lund, Sweden: 20130312105459927; Copenhagen, Denmark: H-1-2012-166 and H-1-2012-100; Amsterdam, Netherlands: NL40099.029.12; Newcastle, Dundee, and Exeter, UK: 12/NE/0132) and all participants provided written informed consent at enrollment. The research conformed to the ethical principles for medical research involving human participants outlined in the declaration of Helsinki. Further details about the data generation can be found in Wesolowska–Andersen and Brorsson et al.7.
## Pre-processing of data
From the clinical, environmental, and questionnaire data only variables with variation across the dataset that were present in at least $10\%$ of the individuals were included. The genomic data was included as the genotypes of risk alleles identified in Mahajan et al.64. In total 393 risk alleles were identified in our cohort out of the 403 associations mentioned in the paper. The genotypes were included as homozygous for risk allele, heterozygote, not having the allele, or missing if the locus was not identified for the individual. Diet data was included as 47 features on self-reported total intake of macronutrients and vitamins across a 24-h period. The wearables measured with an accelerometer included 25 measurements that summarize the movement and heart rate during the day. Transcriptomics data (RNA sequencing) from fasting whole blood samples were processed with RailRNA (v0.2.4b)65 to obtain scaled counts for all samples and only the most variable genes were included. The variable genes were selected by calculating the standard deviation across all individuals for each gene and selecting genes with an above-average standard deviation. Both targeted and untargeted metabolomics data in fasting plasma were included for all measurements passing quality control. In the proteomics data, all measurements within the measurable range based on the OLINK antibody panel were included and residualized for plate layout. The metagenomics data was only available for approximately one-third [256] of the individuals and were included as normalized read counts of identified Metagenomic Species66. Categorical data, including questionnaire responses, drug data, and genomics, was one-hot encoded. The continuous data were residualized by the collection center as the data was collected from six different European countries and, thus, handled by different nurses and lab technicians, as well as differences in the time-of-day samples were taken, which could have a large effect on the measurements. Additionally, the data were residualized for age and sex as these could be biological non-disease-related confounders in the data. Lastly, each continuous dataset was z-scale normalized per feature to ensure that each feature was distributed around zero.
## Classification of drugs using the ATC system
The ATC system is the WHO classification system for therapeutic drugs. The system has a hierarchical structure, where the topmost level, ‘level 1—Anatomical main group’, specifies the target organ or tissue, and the lowermost level, ‘level 5—chemical substance’, specifies the active chemical compound. The three levels in between specify the therapeutic, pharmacological, and chemical levels, respectively. We, therefore, mapped all drugs to the lowest possible level to prevent information loss. A total of 4,155 entries could be mapped to level 5. For 55 entries, only a higher-level mapping was possible owing to lack of specificity and 43 entries could not be mapped to the ATC system, either because of the compound not existing in the database, for example nutraceutical compounds, or when we were unable to identify which drug was registered for the participant. The ATC system does not only specify compound names, but also administration route and daily dosages for over half of level 5 entries. However, owing to uncertainty of the reliability of the registered dosages, only drug names and administration routes were used for mapping. In instances where the administration route was not available, the drug was mapped by drug name only.
## Drug data collection and clean-up
The study participants were asked to register their current drug usage at screening and baseline. Drug names were registered as free text together with administration route, dosage and frequency, and indication. Metformin was recorded separately from other anti-diabetic and non-anti-diabetic drugs. The collected data was variable in quality, using both generic and brand names, which were in many cases specific to the country of the participant. The data was cleaned in four steps: [1] removal of special characters, company names, formulations, and other non-relevant information; [2] automatic mapping to the PubChem database; [3] manual mapping to generic drug names; and [4] mapping to the ATC system. Indications of placebo use, for example participation in clinical drug trials, were noted as such. Only active compounds were included and consequently, possible brand variation was ignored, including for dietary supplements. Drug combinations were mapped, when possible, to the ATC code specifying said combination. However, when the specificity of the proposed ATC code was less specific than the registered drugs, the drug combinations were mapped to individual ATC codes, that is, ‘Perindopril’ (C09AA04) and ‘Indapamide’ (C03BA11) was used instead of ‘Perindopril and diuretics’ (C09BA04). Entries were mapped to ATC codes with the administration route when possible and otherwise mapped without the administration route. Dosage information was not used in the mapping process. In the manual mapping process, $99.4\%$ of terms were assigned and a total of 359 drugs and drug combinations were identified. A total of 339 drugs ($94.4\%$) was mapped to 441 ATC codes.
## Design of the VAE
The VAE framework was constructed to account for a variable number of fully connected hidden layers in both the encoder and decoder and a latent layer that samples from a Gaussian distribution N[0, 1] of two vectors of size NL representing the means, µ, and standard deviations, σ. Each hidden layer included both batch normalization and dropout67 and with leaky rectified linear units (LeakyReLU)68 as activation function. Each dataset was concatenated to one input layer of both categorical and continuous variables. To allow for dataset-specific weights the error calculation was done separately for each dataset. Here we applied cross-entropy loss for categorical data and mean squared error for continuous data as implemented in PyTorch69. The loss was normalized by dataset input size and batch size. Deviance from the Gaussian distribution was penalized by adding the Kullback–Leibler divergence (KLD) to the loss. The final loss was defined as\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L = \mathbf{W}_{\mathrm{cat}} \times \mathbf{E}_{\mathrm{cat}} + \mathbf{W}_{\mathrm{con}} \times \mathbf{E}_{\mathrm{con}} + \mathbf{W}_{\mathrm{KLD}} \times \mathrm{KLD}$$\end{document}L=Wcat×Ecat+Wcon×Econ+WKLD×KLD Here, Ecat and Econ are vectors of normalized reconstruction error for each of the continuous and categorical datasets. Wcat and Wcon are vectors as well of the same length as the errors to introduce dataset-specific weights. We applied an equal weight of 1 for all datasets except for continuous clinical data where we used a weight of 2. WKLD is a weight put on the KLD defined as WKLD = β × NL−1 for which we used a β of 0.0001 for the final model. The KLD was defined as\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{KLD} = {\sum} { - \frac{1}{2}(1 + \ln \left(\sigma \right) - \mu ^2 - \sigma)}$$\end{document}KLD=∑−12(1+lnσ−μ2−σ) To efficiently handle missing data for the continuous features we encoded them as mean values across a particular feature during training and excluded the missing data points during back-propagation. With the data being z-score normalized the mean value is represented as zero. For the categorical features, we included them as a zero vector and the ignore index feature in the cross-entropy implementation in PyTorch was used to not include errors for missing data in the back-propagation. The VAE model was trained with the Adam optimizer70, with a mini-batch size of 10 and increasing batch size with a factor of 1.25 during training after every 50 epochs. The number of training epochs was set to 200 on the basis of early stopping on the test set as described below. Additionally, we trained the model using warm-up by first including the full KLD after 10 epochs slowly increasing the weight at epochs 4, 6, and 8. The latent representation of each patient was obtained by passing them through the trained VAE and extracting the µ layer. The VAE was implemented using PyTorch69 (v.1.7.0) and run using a GPU running CUDA (v.10.2.89).
## Hyperparameter optimization for multi-omics integration
We initially divided the dataset into training ($90\%$) and test ($10\%$) sets to identify the optimal hyperparameter settings to efficiently capture the data structure without losing the ability to generalize on the test data (Supplementary Figs. 2 and 3). We tested different combinations of sizes of hidden layers, the number of hidden layers, size of latent space, dropout, and weight on the KLD. We then evaluated the model on the basis of both test log-likelihood and reconstruction accuracy. For the number of hidden neurons, the variations used were 200, 500, 800, 1,000, and 1,200, with the number of layers ranging between 1 and 5. The tested latent sizes were between 20 and 400 as well as dropout of $10\%$, $20\%$, and $30\%$ and KLD weights of 0.001, 0.0001, and 0.0001. We defined an accurate reconstruction for categorical variables as the class with the highest probability corresponding to the class given by the input. For continuous variables, the accuracy was assessed by comparing the reconstructed array with the input array using cosine similarity for each individual instead of using exact matching. For both categorical and continuous data only non-missing values were used when calculating the accuracy in the reconstruction. We chose the number of training epochs on the basis of when the optimal test likelihood was achieved during testing rounded up to the nearest 100 epochs to ensure sufficient training to learn the complexity of the data. Here we found that more complex models, with higher numbers of hidden neurons and layers, resulted in worse performance on the test set (Supplementary Fig. 2) and that models with more than one hidden layer were unable to provide a decent reconstruction on the training data without overfitting. The only exception was the size of the latent representation, which gave a worse performance with smaller sizes (<50) and equally good performance for larger sizes (from 100 to 400) (Supplementary Fig. 3). For the five best performing models, stability was measured to choose the final model. The stability of the model was evaluated by repeating training with the same hyperparameters and calculating the difference in cosine similarity of the latent space to all other individuals. If the model produced the same result the average change in cosine similarity should be zero. The model with the average change closest to zero was then considered the most stable. The final hyperparameters were set to be one hidden layer of 2,000 neurons, a latent size of at least 100, and a $10\%$ dropout for regularization.
## Evaluating feature importance
Feature importance was extracted from the weights of the network for the models with only one hidden layer and because the input data was z-score normalized calculated as\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_i = \mathop {\sum }\limits_{$j = 1$}^{n_{\mathrm{hidden}}} \left| {w_{ij}} \right|$$\end{document}Ii=∑$j = 1$nhiddenwijwhere *Ii is* the ith feature input and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left| {w_{ij}} \right|$$\end{document}wij is the absolute value of the weight from ith input to the jth hidden neuron. To assess the actual impact on the latent representation an adaptation of the SHAP19 analysis was applied. The difference in model performance was assessed as the absolute differences of the latent representation when changing each input to missing for all individuals and passing it through the trained model.
## Extracting significant drug associations
Drug associations were extracted by perturbation of the input data after training the final model on all individuals. Thus, for each drug we changed the drug status for all individuals with ‘not receiving’ to ‘receiving’. Importantly, we only included individuals that did not receive the specific drug or another drug within the same therapeutic subgroup (ATC level 2). Then, for each drug change, we compared the change in reconstructions to when we passed the original (un-perturbed) data through the network. In other words we determined the differences that the network infers from the change in drug status that during training was learned from all individuals receiving the drug. We used two strategies for this, one was based on an ensemble of Student’s t-tests using benchmarked thresholds, and another was based on Bayesian decision theory. Both approaches were benchmarked against randomized datasets where all the input data matrices were shuffled on rows and columns. We simulated effects in the shuffled data by randomly sampling a combination of a drug, a multi-omics dataset, and a feature within that omics dataset. For each combination, we then sampled an effect from the standard normal distribution N[0,1] and added this value to the omics feature whenever the selected drug was taken by an individual. We, therefore, did not expect that all effects would be significant in the statistical tests because we sample from N[0,1] and some effects will be close to 0. We added a total of 100 effects to the shuffled data and repeated the entire procedure to generate two shuffled datasets each with their unique added effects. Additionally, we investigated if the number of significant associations, effect size estimates and model uncertainty in the reconstruction were not biased by individual dataset uncertainties. This was done by calculating PCCs between the average estimated effect size across all 20 drugs and the difference between model input and the reconstructions for each of the omics features.
## Significant associations using MOVE t-test
To evaluate if the change in the reconstruction was significant, we first determined the expected average change when passing the original and perturbed data through the model ten times. On the basis of these averages, we used a Student’s t-test for related samples as implemented in Python SciPy (v.1.3.1)71 between the baseline and drug-perturbed data for all non-missing continuous data. All P values were subsequently Bonferroni-corrected independently for each drug, and we applied a significance threshold of adjusted $P \leq 0.05.$ We repeated the entire analysis with retraining of the model 10 times for each of four latent sizes (150, 200, 250, and 300). Associations were only included for analysis if they were significant for at least three of the four latent sizes and in at least five out of ten of the repeats. Therefore, reported P values were the averaged P value across the 10 replicate and 4 model tests, that is a total of 40 two-sided Bonferroni-corrected t-tests. The change in reconstruction, what we report as effect size, was calculated as the average difference across the 10 replicates and 4 model tests and were reported with $95\%$ confidence intervals.
## Significant associations using Bayes decision theory
For the method that was based on Bayesian decision theory we used an approach inspired by single-cell variational inference29 and Lopez et al.31. We trained VAE models with a latent size of 150 neurons and benchmarked the approach using different latent sizes and ensembling 1, 5, 10, 20, 30, 35, 40, or 50 models, which we termed refits. For the refits we averaged the reconstructions and used these to obtain the posteriors for the non-perturbed data and each of the drug perturbations. Thus, for VAE ensemble refit i, individual n, feature f, and drug d we define the variational reconstructions as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat x_{infd}$$\end{document}x^infd. By averaging across VAE refits, we obtain estimates of the average posteriors \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat x_{nfd}$$\end{document}x^nfd. Then, for each drug d we compare between two models: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M_d^f$$\end{document}Mdf where feature f is significantly associated with the drug, and the alternative model \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M_0^f$$\end{document}M0f where feature f is not significantly associated with drug d. Hence, we evaluate how often \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left| {\hat x_{nfd} - \hat x_{nf0}} \right| > 0$$\end{document}x^nfd−x^nf0>0 and calculate Bayes factors (K) as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$K = {{{\mathrm{log}}}}_e\left| {\frac{{\mathrm{P}\left({M_d^f|\hat x_{fd},\,\hat x_{f0}} \right)}}{{\mathrm{P}(M_0^f|\hat x_{fd},\,\hat x_{f0})}}} \right|$$\end{document}K=logePMdf∣x^fd,x^f0P(M0f∣x^fd,x^f0) We ranked the associated features according to K (ref. 72). We set a FDR of α by accepting associations (n) between features and a drug until the cumulative evidence of P(M0) across accepted features for the drug was above the threshold. Since \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{P}(M_0^f)=(1-\mathrm{P}(M_d^f))$$\end{document}P(M0f)=(1−P(Mdf)) we accepted drug-feature associations while the cumulative evidence E is lower than α\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E = \mathop {\sum }\limits_f \frac{{(1 - \mathrm{P}(M_d^f))}}{n} < \alpha$$\end{document}E=∑f(1−P(Mdf))n<α
## Benchmarking of t-test, MOVE t-test and MOVE Bayes
To be able to compare the number of significant associations between methods we used the two randomized datasets to estimate FDR from the ground truth, that is the added drug–omics effects (Supplementary Table 3). Here we found that a t-test with Benjamini–Hochberg FDR of 0.01 had ground-truth FDR of 0.00 and 0.06 on the two randomized datasets, corresponding to 52 and 67 true positives as well as 0 and 4 false positives, respectively. For MOVE t-test, we benchmarked the number of refits of the 4 models and found 10 refits to have a ground-truth FDR of 0.02 and 0.06, with 48 and 61 true positives as well as 1 and 3 false positives, respectively. For MOVE Bayes we benchmarked the number of refits for a model with 150 latent neurons and found FDR from the cumulative evidence to be well aligned with FDR of the ground truth. Using Bayes FDR of 0.05 we found 30 refits to have ground-truth FDR of 0.02 and 0.05, respectively. Across the two shuffled datasets 42 and 59 true positives were found by all three methods (Supplementary Fig. 12).
## Calculation of drug associations using other methods
We compared our findings to associations identified with standard statistical approaches using Student’s t-test for unrelated samples and an ANOVA between two groups of individuals ‘not receiving’ and ‘receiving’ each drug. Here we used Benjamini–Hochberg correction for FDR73 with an adjusted $P \leq 0.01.$ Additionally, we tested if a least absolute shrinkage and selection operator (LASSO) model was able to identify features with significant impact on predicting the ‘not receiving’ or ‘receiving’ groups for each drug. However, the LASSO model was unable to converge possibly owing to the high input feature dimensionality. All statistical tests were done with Python SciPy (v.1.3.1)71.
## Drug effect size and similarities across omics data
Drug effect sizes were determined as the difference between the baseline and drug-perturbed variational reconstructions, that is, as the average difference across the VAE ensemble refits reported with $95\%$ confidence intervals. Drug similarities were calculated as the cosine similarity as implemented in Python SciPy (v.1.3.1)71 between the average effect sizes on all features identified as significantly associated for at least one of the drugs both across and within each dataset. The difference was only calculated for non-missing data and individuals not already on the drug or a drug in the same ACT group. The rank of drug effect sizes was determined for each omics dataset ranking the effect sizes from 1 to 20. A rank of 20 indicates that the drug had the highest average effect size in this omics dataset compared to the other drugs. Correlations between multi-omics profiles and number of individuals taking the drug pair were calculated from the fraction of individuals that overlapped between the two drugs.
## Molecular-focused analysis of the multi-omics data
To get a better understanding of the molecular profiles identified in the associations for the transcriptomics and proteomics data we tested for enriched Gene Ontology terms as well as molecular pathways. For the transcriptomics data, we assessed the molecular patterns of biological processes and pathways from Reactome74 (v.3.7) using the significantly associated genes for each drug against a background list of all genes included in the data integration. We used WebGestaltR75 (v.0.4.4) for the analysis with default settings (hypergeometric test) and evaluated all results with an FDR < 0.05. The targeted metabolomics data was analyzed for potential metabolite enrichments using MetaboAnalyst76 (v.5) over-representation analysis using a hypergeometric test and FDR of 0.05. We investigated both enrichments in known pathways in the KEGG database as well as enrichment of chemical structures sub-, main- and super-class levels. For all analyses, we used the included panel of targeted metabolites as the reference data.
## Association differences within diabetes archetypes
As mentioned, previous work by Wesolowska–Andersen and Brorsson et al. performed archetype analysis of the multi-omics data with only metformin medication data7. Here they based the archetypes on clinical markers and identified four distinct and one ‘mixed’ T2D archetypes with clinical and omics profiles. To investigate if these distinct archetypes differed in their drug associations we used a t-test on the average effect size change for the individuals of each archetype against the remaining individuals. The analysis was only done for the significant drug associations for each drug. All analysis was only done for individuals not taking the drug or a drug within the same ATC therapeutical class similarly to the main analysis.
## Drug–drug interactions
We used an in-house drug–drug interaction compendium generated from publicly available sources (Supplementary Table 11) to assess whether drug combinations had been reported previously to be interacting or not77. The compendium contains interactions from 26 different datasets of pharmacovigilance, clinically oriented information, schemas for NLP corpora, and drug–Cytochrome P450 relationships sources. For 12 of the drug–drug pairs in our dataset we could identify drug–drug interactions with reported severity (major, moderate, minor, possible, undetermined, and none) indicating clinical significance.
## Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
## Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41587-022-01520-x.
## Supplementary information
Supplementary InformationSupplementary Notes, Supplementary Figs. 1–23, and Supplementary Tables 1–11. Reporting Summary Supplementary Data 1Features that were included from the different omics datasets. Supplementary Data 2MOVE t-test, adjusted P values and effect sizes. Supplementary Data 3MOVE Bayes, adjusted P values. Supplementary Data 4Significant associations in both MOVE t-test and MOVE Bayes. Supplementary Data 5Over-representation analysis using Gene Ontologies.
The online version contains supplementary material available at 10.1038/s41587-022-01520-x.
## Source data
Source Data Fig. 1Source data for Fig. 1c–e. Source Data Fig. 2Source data for the Fig. 2a. Source Data Fig. 3Source data for Fig. 3a–e.
## Peer review information
Nature Biotechnology thanks Yasuhiro Kojima and Elin Nyman for their contribution to the peer review of this work.
## References
1. Fares H, DiNicolantonio JJ, O’Keefe JH, Lavie CJ. **Amlodipine in hypertension: a first-line agent with efficacy for improving blood pressure and patient outcomes**. *Open Heart* (2016.0) **3** e000473. DOI: 10.1136/openhrt-2016-000473
2. Hu JX, Thomas CE, Brunak S. **Network biology concepts in complex disease comorbidities**. *Nat. Rev. Genet.* (2016.0) **17** 615-629. DOI: 10.1038/nrg.2016.87
3. Austin RP. **Polypharmacy as a risk factor in the treatment of type 2 diabetes**. *Diabetes Spectr.* (2006.0) **19** 13-16. DOI: 10.2337/diaspect.19.1.13
4. Zhou W. **Longitudinal multi-omics of host–microbe dynamics in prediabetes**. *Nature* (2019.0) **569** 663-671. DOI: 10.1038/s41586-019-1236-x
5. Hasin Y, Seldin M, Lusis A. **Multi-omics approaches to disease**. *Genome Biol.* (2017.0) **18** 83. DOI: 10.1186/s13059-017-1215-1
6. Gudmundsdottir V. **Whole blood co-expression modules associate with metabolic traits and type 2 diabetes: an IMI-DIRECT study**. *Genome Med.* (2020.0) **12** 109. DOI: 10.1186/s13073-020-00806-6
7. Wesolowska-Andersen A. **Four groups of type 2 diabetes contribute to the etiological and clinical heterogeneity in newly diagnosed individuals: an IMI DIRECT study**. *Cell Reports Medicine* (2022.0) **3** 100477. DOI: 10.1016/j.xcrm.2021.100477
8. Song JW, Chung KC. **Observational studies: cohort and case-control studies**. *Plast. Reconstr. Surg.* (2010.0) **126** 2234-2242. DOI: 10.1097/PRS.0b013e3181f44abc
9. Picard M, Scott-Boyer M-P, Bodein A, Périn O, Droit A. **Integration strategies of multi-omics data for machine learning analysis**. *Comput. Struct. Biotechnol. J.* (2021.0) **19** 3735-3746. DOI: 10.1016/j.csbj.2021.06.030
10. Nicora G, Vitali F, Dagliati A, Geifman N, Bellazzi R. **Integrated multi-omics analyses in oncology: a review of machine learning methods and tools**. *Front. Oncol.* (2020.0) **10** 1030. DOI: 10.3389/fonc.2020.01030
11. Rohart F, Gautier B, Singh A, Lê Cao K-A. **mixOmics: an R package for ’omics feature selection and multiple data integration**. *PLoS Comput. Biol.* (2017.0) **13** e1005752. DOI: 10.1371/journal.pcbi.1005752
12. Chung NC. **Unsupervised classification of multi-omics data during cardiac remodeling using deep learning**. *Methods* (2019.0) **166** 66-73. DOI: 10.1016/j.ymeth.2019.03.004
13. Kriebel AR, Welch JD. **UINMF performs mosaic integration of single-cell multi-omic datasets using nonnegative matrix factorization**. *Nat. Commun.* (2022.0) **13** 780. DOI: 10.1038/s41467-022-28431-4
14. Argelaguet R. **Multi-omics factor analysis—a framework for unsupervised integration of multi-omics data sets**. *Mol. Syst. Biol.* (2018.0) **14** e8124. DOI: 10.15252/msb.20178124
15. Shen R, Olshen AB, Ladanyi M. **Integrative clustering of multiple genomic data types using a joint latent variable model with application to breast and lung cancer subtype analysis**. *Bioinformatics* (2009.0) **25** 2906-2912. DOI: 10.1093/bioinformatics/btp543
16. Singh A. **DIABLO: an integrative approach for identifying key molecular drivers from multi-omics assays**. *Bioinformatics* (2019.0) **35** 3055-3062. DOI: 10.1093/bioinformatics/bty1054
17. 17.Kingma, D. P. & Welling, M. Auto-Encoding Variational Bayes. Preprint at arXiv10.48550/arXiv.1312.6114 (2013).
18. 18.Rezende, D. J., Mohamed, S. & Wierstra, D. Stochastic backpropagation and approximate inference in deep generative models. Preprint at arXiv10.48550/arXiv.1401.4082 (2014).
19. Nissen JN. **Improved metagenome binning and assembly using deep variational autoencoders**. *Nat. Biotechnol.* (2021.0) **39** 555-560. DOI: 10.1038/s41587-020-00777-4
20. Ding J, Condon A, Shah SP. **Interpretable dimensionality reduction of single cell transcriptome data with deep generative models**. *Nat. Commun.* (2018.0) **9** 2002. DOI: 10.1038/s41467-018-04368-5
21. Chaudhary K, Poirion OB, Lu L, Garmire LX. **Deep learning-based multi-omics integration robustly predicts survival in liver cancer**. *Clin. Cancer Res.* (2018.0) **24** 1248-1259. DOI: 10.1158/1078-0432.CCR-17-0853
22. Zhang L. **Deep learning-based multi-omics data integration reveals two prognostic subtypes in high-risk neuroblastoma**. *Front. Genet.* (2018.0) **9** 477. DOI: 10.3389/fgene.2018.00477
23. Cao Z-J, Gao G. **Multi-omics single-cell data integration and regulatory inference with graph-linked embedding**. *Nat. Biotechnol.* (2022.0) **40** 1458-1466. DOI: 10.1038/s41587-022-01284-4
24. 24.Mattei, P.-A. & Frellsen, J. MIWAE: deep generative modelling and imputation of incomplete data. In Proceedings of the 36th International Conference on Machine Learning 4413–4423 (PMLR, 2019).
25. Way GP, Greene CS. **Extracting a biologically relevant latent space from cancer transcriptomes with variational autoencoders**. *Pac. Symp. Biocomput.* (2018.0) **23** 80-91. PMID: 29218871
26. Allesøe RL. **Deep learning-based integration of genetics with registry data for stratification of schizophrenia and depression**. *Sci. Adv.* (2022.0) **8** eabi7293. DOI: 10.1126/sciadv.abi7293
27. 27.Ghahramani, A., Watt, F. M. & Luscombe, N. M. Generative adversarial networks simulate gene expression and predict perturbations in single cells. Preprint at bioRxiv10.1101/262501 (2018).
28. Yelmen B. **Creating artificial human genomes using generative neural networks**. *PLoS Genet.* (2021.0) **17** e1009303. DOI: 10.1371/journal.pgen.1009303
29. Lopez R, Regier J, Cole MB, Jordan MI, Yosef N. **Deep generative modeling for single-cell transcriptomics**. *Nat. Methods* (2018.0) **15** 1053-1058. DOI: 10.1038/s41592-018-0229-2
30. Gayoso A. **A Python library for probabilistic analysis of single-cell omics data**. *Nat. Biotechnol.* (2022.0) **40** 163-166. DOI: 10.1038/s41587-021-01206-w
31. 31.Lopez, R., Boyeau, P., Yosef, N., Jordan, M. I. & Regier, J. Decision-making with auto-encoding variational Bayes. In Proceedings of the 34th International Conference on Neural Information Processing Systems 5081–5092 (Curran Associates Inc., 2020).
32. Yeo GHT, Saksena SD, Gifford DK. **Generative modeling of single-cell time series with PRESCIENT enables prediction of cell trajectories with interventions**. *Nat. Commun.* (2021.0) **12** 3222. DOI: 10.1038/s41467-021-23518-w
33. Frazer J. **Disease variant prediction with deep generative models of evolutionary data**. *Nature* (2021.0) **599** 91-95. DOI: 10.1038/s41586-021-04043-8
34. 34.Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. in Advances in Neural Information Processing Systems 30 (eds Guyon, I. et al.) 4765–4774 (Curran Associates, 2017).
35. Hirst JA, Farmer AJ, Ali R, Roberts NW, Stevens RJ. **Quantifying the effect of metformin treatment and dose on glycemic control**. *Diabetes Care* (2012.0) **35** 446-454. DOI: 10.2337/dc11-1465
36. Knowler WC. **Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin**. *N. Engl. J. Med.* (2002.0) **346** 393-403. DOI: 10.1056/NEJMoa012512
37. Ustinova M. **Metformin strongly affects transcriptome of peripheral blood cells in healthy individuals**. *PLoS One* (2019.0) **14** e0224835. DOI: 10.1371/journal.pone.0224835
38. Xiao Z, Wu W, Poltoratsky V. **Metformin suppressed CXCL8 expression and cell migration in HEK293/TLR4 cell line**. *Mediators Inflamm.* (2017.0) **2017** 6589423. DOI: 10.1155/2017/6589423
39. Bruno S. **Metformin inhibits cell cycle progression of B-cell chronic lymphocytic leukemia cells**. *Oncotarget* (2015.0) **6** 22624-22640. DOI: 10.18632/oncotarget.4168
40. Ma W. **Metformin alters gut microbiota of healthy mice: implication for its potential role in gut microbiota homeostasis**. *Front. Microbiol.* (2018.0) **9** 1336. DOI: 10.3389/fmicb.2018.01336
41. Forslund K. **Disentangling type 2 diabetes and metformin treatment signatures in the human gut microbiota**. *Nature* (2015.0) **528** 262-266. DOI: 10.1038/nature15766
42. Vieira-Silva S. **Statin therapy is associated with lower prevalence of gut microbiota dysbiosis**. *Nature* (2020.0) **581** 310-315. DOI: 10.1038/s41586-020-2269-x
43. Bryrup T. **Metformin-induced changes of the gut microbiota in healthy young men: results of a non-blinded, one-armed intervention study**. *Diabetologia* (2019.0) **62** 1024-1035. DOI: 10.1007/s00125-019-4848-7
44. Vich Vila A. **Impact of commonly used drugs on the composition and metabolic function of the gut microbiota**. *Nat. Commun.* (2020.0) **11** 362. DOI: 10.1038/s41467-019-14177-z
45. Shin JM, Munson K, Vagin O, Sachs G. **The gastric HK-ATPase: structure, function, and inhibition**. *Pflugers Arch.* (2009.0) **457** 609-622. DOI: 10.1007/s00424-008-0495-4
46. **Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials**. *Lancet* (2010.0) **376** 1670-1681. DOI: 10.1016/S0140-6736(10)61350-5
47. Barter PJ, Brandrup-Wognsen G, Palmer MK, Nicholls SJ. **Effect of statins on HDL-C: a complex process unrelated to changes in LDL-C: analysis of the VOYAGER database**. *J. Lipid Res.* (2010.0) **51** 1546-1553. DOI: 10.1194/jlr.P002816
48. Aguayo-Orozco A. **sAOP: linking chemical stressors to adverse outcomes pathway networks**. *Bioinformatics* (2019.0) **35** 5391-5392. DOI: 10.1093/bioinformatics/btz570
49. Margerie D. **Hepatic transcriptomic signatures of statin treatment are associated with impaired glucose homeostasis in severely obese patients**. *BMC Med. Genomics* (2019.0) **12** 80. DOI: 10.1186/s12920-019-0536-1
50. Gilbert R, Al-Janabi A, Tomkins-Netzer O, Lightman S. **Statins as anti-inflammatory agents: a potential therapeutic role in sight-threatening non-infectious uveitis**. *Porto Biomed J* (2017.0) **2** 33-39. DOI: 10.1016/j.pbj.2017.01.006
51. Aguayo-Orozco A, Bois FY, Brunak S, Taboureau O. **Analysis of time-series gene expression data to explore mechanisms of chemical-induced hepatic steatosis toxicity**. *Front. Genet.* (2018.0) **9** 396. DOI: 10.3389/fgene.2018.00396
52. Kennedy MA. **ABCG1 has a critical role in mediating cholesterol efflux to HDL and preventing cellular lipid accumulation**. *Cell Metab.* (2005.0) **1** 121-131. DOI: 10.1016/j.cmet.2005.01.002
53. Ishihara N. **Atorvastatin increases**. *Mol. Med. Rep.* (2017.0) **16** 4756-4762. DOI: 10.3892/mmr.2017.7141
54. Ferretti G, Bacchetti T, Banach M, Simental-Mendía LE, Sahebkar A. **Impact of statin therapy on plasma MMP-3, MMP-9, and TIMP-1 concentrations: a systematic review and meta-analysis of randomized placebo-controlled trials**. *Angiology* (2017.0) **68** 850-862. DOI: 10.1177/0003319716688301
55. Orekhov AN. **Role of phagocytosis in the pro-inflammatory response in LDL-induced foam cell formation; a transcriptome analysis**. *Int. J. Mol. Sci.* (2020.0) **21** 817. DOI: 10.3390/ijms21030817
56. Osório J. **Statins and T2DM—an IGF link?**. *Nat. Rev. Endocrinol.* (2013.0) **9** 187-187. DOI: 10.1038/nrendo.2013.33
57. Alves A, Bassot A, Bulteau A-L, Pirola L, Morio B. **Glycine metabolism and its alterations in obesity and metabolic diseases**. *Nutrients* (2019.0) **11** 1356. DOI: 10.3390/nu11061356
58. Snowden SG. **High-dose simvastatin exhibits enhanced lipid-lowering effects relative to simvastatin/ezetimibe combination therapy**. *Circ. Cardiovasc. Genet.* (2014.0) **7** 955-964. DOI: 10.1161/CIRCGENETICS.114.000606
59. Forslund SK. **Combinatorial, additive and dose-dependent drug-microbiome associations**. *Nature* (2021.0) **600** 500-505. DOI: 10.1038/s41586-021-04177-9
60. Zimmermann M, Zimmermann-Kogadeeva M, Wegmann R, Goodman AL. **Mapping human microbiome drug metabolism by gut bacteria and their genes**. *Nature* (2019.0) **570** 462-467. DOI: 10.1038/s41586-019-1291-3
61. Lillie EO. **The**. *Per. Med.* (2011.0) **8** 161-173. DOI: 10.2217/pme.11.7
62. Koivula RW. **Discovery of biomarkers for glycaemic deterioration before and after the onset of type 2 diabetes: rationale and design of the epidemiological studies within the IMI DIRECT Consortium**. *Diabetologia* (2014.0) **57** 1132-1142. DOI: 10.1007/s00125-014-3216-x
63. Koivula RW. **Discovery of biomarkers for glycaemic deterioration before and after the onset of type 2 diabetes: descriptive characteristics of the epidemiological studies within the IMI DIRECT Consortium**. *Diabetologia* (2019.0) **62** 1601-1615. DOI: 10.1007/s00125-019-4906-1
64. Mahajan A. **Fine-mapping type 2 diabetes loci to single-variant resolution using high-density imputation and islet-specific epigenome maps**. *Nat. Genet.* (2018.0) **50** 1505-1513. DOI: 10.1038/s41588-018-0241-6
65. Nellore A. **Rail-RNA: scalable analysis of RNA-seq splicing and coverage**. *Bioinformatics* (2017.0) **33** 4033-4040. DOI: 10.1093/bioinformatics/btw575
66. Nielsen HB. **Identification and assembly of genomes and genetic elements in complex metagenomic samples without using reference genomes**. *Nat. Biotechnol.* (2014.0) **32** 822-828. DOI: 10.1038/nbt.2939
67. Srivastava N, Hinton G, Krizhevsky A, Sutskever I, Salakhutdinov R. **Dropout: a simple way to prevent neural networks from overfitting**. *J. Mach. Learn. Res.* (2014.0) **15** 1929-1958
68. 68.Maas, A. L., Hannun, A. Y. & Ng, A. Y. Rectifier nonlinearities improve neural network acoustic models. In Proc. 30th International Conference on Machine Learning (eds Dasgupta, S. & McAllester, D.) (JMLR, 2013).
69. 69.Paszke, A. et al. PyTorch: an imperative style, high-performance deep learning library. In Advances in Neural Information Processing Systems Vol. 32 (eds Wallach, H. et al.) (Curran Associates, 2019).
70. 70.Kingma, D. P. & Ba, J. Adam: a method for stochastic optimization. Preprint at arXiv10.48550/arXiv.1412.6980 (2014).
71. Virtanen P. **SciPy 1.0: fundamental algorithms for scientific computing in Python**. *Nat. Methods* (2020.0) **17** 261-272. DOI: 10.1038/s41592-019-0686-2
72. Kass RE, Raftery AE. **Bayes factors**. *J. Am. Stat. Assoc.* (1995.0) **90** 773-795. DOI: 10.1080/01621459.1995.10476572
73. Benjamini Y, Hochberg Y. **Controlling the false discovery rate: a practical and powerful approach to multiple testing**. *J. R. Stat. Soc. Ser. B Stat. Methodol.* (1995.0) **57** 289-300
74. Jassal B. **The reactome pathway knowledgebase**. *Nucleic Acids Res.* (2020.0) **48** D498-D503. PMID: 31691815
75. Liao Y, Wang J, Jaehnig EJ, Shi Z, Zhang B. **WebGestalt 2019: gene set analysis toolkit with revamped UIs and APIs**. *Nucleic Acids Res.* (2019.0) **47** W199-W205. DOI: 10.1093/nar/gkz401
76. Chong J, Xia J. **MetaboAnalystR: an R package for flexible and reproducible analysis of metabolomics data**. *Bioinformatics* (2018.0) **34** 4313-4314. DOI: 10.1093/bioinformatics/bty528
77. Leal Rodríguez C. **Drug interactions in hospital prescriptions in Denmark: prevalence and associations with adverse outcomes**. *Pharmacoepidemiol. Drug Saf.* (2022.0) **31** 632-642. DOI: 10.1002/pds.5415
|
---
title: Macrophage polarization markers in subcutaneous, pericardial, and epicardial
adipose tissue are altered in patients with coronary heart disease
authors:
- Bianca Papotti
- Trine Baur Opstad
- Sissel Åkra
- Theis Tønnessen
- Bjørn Braathen
- Charlotte Holst Hansen
- Harald Arnesen
- Svein Solheim
- Ingebjørg Seljeflot
- Nicoletta Ronda
journal: Frontiers in Cardiovascular Medicine
year: 2023
pmcid: PMC10017535
doi: 10.3389/fcvm.2023.1055069
license: CC BY 4.0
---
# Macrophage polarization markers in subcutaneous, pericardial, and epicardial adipose tissue are altered in patients with coronary heart disease
## Body
**GRAPHICAL ABSTRACT:** *Pictures were created by combining images from Smart Servier Medical Art (https://smart-servier.com, accessed on 20.09.22). Servier Medical Art by Servier is licensed under a Creative Commons Attribution 3.0 Unported License (https://creativecommons.org/licenses/by/3.0/).*
## Abstract
### Background
Epicardial and pericardial adipose tissue (EAT and PAT) surround and protect the heart, with EAT directly sharing the microcirculation with the myocardium, possibly presenting a distinct macrophage phenotype that might affect the inflammatory environment in coronary heart disease (CHD). This study aims to investigate the expression of genes in different AT compartments driving the polarization of AT macrophages toward an anti-inflammatory (L-Galectin 9; CD206) or pro-inflammatory (NOS2) phenotype.
### Methods
EAT, PAT, and subcutaneous (SAT) biopsies were collected from 52 CHD patients undergoing coronary artery bypass grafting, and from 22 CTRLs undergoing aortic valve replacement. L-Galectin9 (L-Gal9), CD206, and NOS2 AT gene expression and circulating levels were analyzed through RT-PCR and ELISA, respectively.
### Results
L-Gal9, CD206, and NOS2 gene expression was similar in all AT compartments in CHD and CTRLs, as were also L-Gal9 and CD206 circulating levels, while NOS2 serum levels were higher in CHD ($$p \leq 0.012$$ vs. CTRLs). In CTRLs, NOS2 expression was lower in EAT vs. SAT ($$p \leq 0.007$$), while in CHD patients CD206 expression was lower in both SAT and EAT as compared to PAT ($$p \leq 0.003$$, $$p \leq 0.006$$, respectively), suggestive of a possible macrophage reprogramming toward a pro-inflammatory phenotype in EAT. In CHD patients, NOS2 expression in SAT correlated to that in PAT and EAT ($$p \leq 0.007$$, both), CD206 expression correlated positively to L-Gal9 ($p \leq 0.001$) only in EAT, and CD206 expression associated with that of macrophage identifying markers in all AT compartments ($p \leq 0.001$, all). In CHD patients, subjects with LDL-C above 1.8 mmol/L showed significantly higher NOS2 expression in PAT and EAT as compared to subjects with LDL-C levels below ($p \leq 0.05$), possibly reflecting increased cardiac AT pro-inflammatory activation. In SAT and PAT, CD206 expression associated with BMI in both CHD and CTRLs ($p \leq 0.05$, all), and with L-Gal9 in EAT, however only in CTRLs ($$p \leq 0.002$$).
### Conclusion
CHD seems to be accompanied by an altered cardiac, and especially epicardial AT macrophage polarization. This may represent an important pathophysiological mechanism and a promising field of therapy targeting the excessive AT inflammation, in need of further investigation.
## 1. Introduction
The main cause of coronary heart disease (CHD) is atherosclerosis [1], where chronic inflammatory processes and obesity play pivotal roles. The heart is surrounded by adipose tissue (AT) that is both epicardial (EAT) and pericardial (PAT), with EAT embryonically originating from the splanchnopleuric mesoderm and PAT from thoracic mesenchyme [2]. Anatomically, EAT is localized between the myocardium and the visceral pericardium and covers about $80\%$ of the heart surface [3]. It shares the microcirculation with the myocardial tissue and secretes various molecules through paracrine and “vasocrine” [4] mechanisms that may protect coronary arteries [3, 5, 6]. PAT is separated from the heart by pericardium and, differently from EAT, receives blood supply from non-coronary arteries. Its role as a source of cardiac biochemical mediators is still a matter of debate [7]. EAT exerts various anti-atherogenic and anti-inflammatory activities and regulates cardiac thermogenesis [8], while PAT apparently has less anti-inflammatory activity [9]. Disturbances in AT composition and extension associate to adipocytes’ hypertrophy, insulin resistance and pro-inflammatory processes [10]. According to some studies, the EAT volume is increased in coronary artery disease (CAD) patients [11, 12], although this was not confirmed by others [13], and may be a predictor of cardio-metabolic risk [14]. Similarly, also increased PAT volume has been shown to associate with CV risk, lipid disorders, hypertension and obesity [7, 9]. AT secretes a wide variety of adipokines, cytokines and chemokines [15, 16], thus being at the crossroad between metabolism and immunity. In addition to adipocytes, various resident immune cell populations can be found in AT, including adipose tissue macrophages (ATMs) [17], characterized by an high degree of phenotypic plasticity [17]. Based on their inflammatory profile, ATMs may be classified, like other macrophages, as M1 pro-, and M2 anti-inflammatory elements [18]. It is well accepted that ATMs in the tissues of lean subjects mainly exhibit the so-called alternatively activated M2 profile, releasing anti-inflammatory cytokines, known to impact positively on insulin sensitivity, angiogenesis and tissue repair [19]. On the contrary, in obese subjects, ATMs are mainly polarized toward a classically activated M1 profile, known to produce pro-inflammatory cytokines, capable of inhibiting the normal insulin signaling in adipocytes [20]. However, it is well accepted that macrophage polarization in vivo is a highly dynamic process that goes beyond the simplified M1/M2 classification [18]. At this regard, Hirata and co-workers reported that in patients with CAD, EAT has an altered M1/M2 polarization [21], suggesting that the ATM phenotype may have an impact on cardiovascular disease (CVD), but to date the studies assessing potential differences between macrophage polarization in the different cardiac ATs are still limited.
The present study aims at investigating the polarization status of ATMs in EAT, PAT, and subcutaneous AT (SAT) from patients with CHD and from control patients with aortic valvular disease, to explore possible alterations specific to CHD and for each AT type. To this purpose, we measured gene expression of the following markers in the different cardiac AT compartments: (I) L-Galectin 9 (L-Gal 9), mainly produced in AT by macrophages and T cells [22], promoting anti-inflammatory T-regulatory cell activity and macrophage polarization toward an anti-inflammatory phenotype [23, 24]; (II) CD206, a cell marker widely used to identify M2 macrophages [21, 25]; and (III) nitric oxide synthase 2 (NOS2), a biomarker tightly connected with M1 macrophages, being transcribed following pro-inflammatory stimuli and involved in the effector molecule nitric oxide (NO) biosynthesis [26]. Gene expression of the markers was related to their circulating protein levels and to patient clinical characteristics, including lipid profile and anthropometrics. In parallel, the expression of cell markers specific for macrophages, T lymphocytes and endothelial cells was also evaluated. Our hypothesis was that the different cardiac AT compartments might present a distinct pattern of macrophage polarization markers, with possible impact on CHD development.
## 2.1. Patients
Fifty-two patients with CHD undergoing coronary artery bypass grafting and 22 control subjects (CTRLs) undergoing aortic valve replacement without evidence of CHD were recruited from December 2016 to May 2018 at Oslo University Hospital, Ullevål, Oslo (Norway). Before surgery, all patients gave written informed consent and the experimental protocol was approved by the Regional Ethics Committee of North Norway (#$\frac{2016}{411}$), following the Declaration of Helsinki. The inclusion criterium for the pathological condition under study was the indication for coronary artery bypass surgery due to CHD; with no specific restrictions to inclusion, except for lack of consent to participate to the study and the use of medications known to overtly interfere with inflammation (e.g., steroids). The study is registered at clinicaltrials.gov with the code NCT02760914. Briefly, during the open-chest surgery and before starting the extracorporeal circulation, representative biopsies (approximately 0.5–1.5 cm) from SAT (pre-sternally at the middle of the sternum), PAT (ventrally to the pericardium next to the aorta) and EAT (area between the right coronary artery and the pulmonary artery) were isolated and immediately deep-frozen at −80°C until RNA extraction. Before anesthesia, arterial blood samples were collected. More details have previously been published [27].
## 2.2. Laboratory analyses
Total RNA was extracted from SAT, PAT, and EAT by processing samples through the RNeasy Lipid Tissue Mini Kit (Qiagen, GmbH, Germany), following the manufacturer’s instructions. RNA concentration and purity was determined using a NanoDropTM 1,000 Spectroscophotometer (SaveenWerner, Sweden), observing a mean concentration of 29.3 ng/mL and a purity of 1.7 (absorbance ratio at 260 and 280 nm). cDNA was then retro-transcribed starting from 5 ng/mL of RNA for each sample and using the qScript™ cDNA superMix commercial kit (Quanta Biosciences, United States). Gene expression analyses were measured with TaqMan® assays (Applied Biosystems, CA, United States), as follows: L-Galectin 9 (Hs00247135_m1), CD206 (Hs00267207_m1) and NOS2 (Hs01075529_m1). Real-Time qPCR was performed using the TaqMan® Universal PCR Master Mix (cat. n. 4,324,018) on a ViiATM7 instrument (Applied Biosystems, CA, United States). The ΔΔCt method was applied to determine the mRNA levels in each reaction, using the β2-Microglobulin (Hs99999907_m1) as the normalizer internal gene, and expressed as relative quantification (RQ) to a reference sample [28]. Gene expression of CD163, CD68, CD3, and CD31, representing macrophages, T cells and endothelial cells, respectively, was analyzed through RT-PCR to determine the different cell types present in each AT sample, as previously described [27].
Whole blood from each patient was centrifuged at 2,500 g for 10` and the isolated serum was conserved at −80°C until use. Commercially available enzyme-linked immunosorbent assays (ELISA) were used to determine the circulating levels of L-Gal 9 (R&D Systems, NE, US), CD206 (RayBotech, GA, United States) and NOS2 (LifeSpan BioSciences, WA, United States), following the manufacturer’s instructions. All samples were analyzed as duplicates. The intra-assay coefficient of variation (CV) in our laboratory were 6.7, 3.5, and $5.8\%$, respectively. Routine patient’s analyses were performed by conventional laboratory methods.
## 2.3. Statistical analyses
The characteristics of patients are reported as numbers and percentages or as median values and 25th and 75th percentiles. The vast majority of the variables were skewed distributed, therefore non-parametric analyses were performed, including the Mann–Whitney U-test to compare the two groups, the Friedmans’ test coupled to the Wilcoxon signed-rank test to compare gene expression in the individual AT compartments. The Spearmann’s rho was used for correlations and trend lines applied in figures when being statistically significant. Furthermore, the Chi-square test was used for differences in categorical variables between groups. A value of $p \leq 0.05$ was considered statistically significant, and Bonferroni correction for multiple comparisons was applied as specified. Statistical analyses were performed using SPSS version 28 (SPSS Inc., IL, United States).
## 3.1. Patients characteristics
Clinical and laboratory parameters of the 74 recruited subjects, 52 CHD and 22 CTRLs are reported in Table 1. The proportion of men was higher in the CHD group, as well as glomerular filtration rate (GFR), HbA1c levels and the use of aspirin, beta-blockers, statins and other lipid-lowering drugs. The latter are possibly responsible for the lower total-and LDL-cholesterol levels found in CHD patients compared to CTRLs.
**Table 1**
| Unnamed: 0 | CTRLs (n = 22) | CHD (n = 52) | p value |
| --- | --- | --- | --- |
| Age (years) | 69 (63, 71.5) | 66.5 (62,71.8) | |
| Male (%) | 11 (50%) | 40 (76.92%) | 0.04 |
| Smoker (previous/current) | 10 (45.45%) | 31 (59.6%) | |
| Weight (Kg) | 82.5 (77.5, 107.0) | 85 (70.3, 95.5) | |
| Height (m) | 1.75 (1.7, 1.8) | 1.77 (1.7, 1.8) | |
| Waist (cm) | 90 (88, 101) | 92 (86, 98) | |
| BMI (kg/m2) | 28.4 (24.6, 31.6) | 27.3 (23.8, 30.1) | |
| SBP (mmhg) | 140 (115, 163) | 140 (125, 160) | |
| DBP (mmhg) | 79 (70, 86) | 80 (70, 87) | |
| Cardiovascular status | | | |
| Previous AMI (%) | 2 (9.1%) | 20 (38.5%) | 0.025 |
| Angina (%) | 0 (0%) | 24 (46.2%) | <0.001 |
| PCI (%) | 0 (0%) | 20 (38.5%) | 0.002 |
| Hypertension (%) | 9 (40.9%) | 28 (53.8%) | |
| Diabetes type I and II (%) | 3 (13.6%) | 14 (26.9%) | |
| Heart failure (%) | 1 (4.5%) | 3 (5.8%) | |
| Medications | | | |
| Aspirin (%) | 9 (40.9%) | 45 (86.5%) | <0.001 |
| Other antiplatelet (%) | 0 (0%) | 14 (26.9%) | |
| ACEi/ATII (%) | 11 (50%) | 24 (46.15%) | |
| Beta-blockers (%) | 6 (27.3%) | 32 (61.5%) | 0.015 |
| Statins (%) | 11 (50%) | 37 (71.2%) | |
| Lipid-lowering agents (%) | 1 (4.54%) | 10 (19.2%) | |
| NSAIDs (%) | 0 (0%) | 2 (3.9%) | |
| Insulin (%) | 0 (0%) | 6 (11.5%) | |
| Anti-diabetic drugs (%) | 3 (13.6%) | 11 (21.2%) | |
| Diuretics (%) | 5 (22.7%) | 7 (13.46%) | |
| Laboratory values | | | |
| hsCRP (mg/L) | 1.00 (1.0, 2.0) | 0.91 (0.49, 1.77) | |
| Troponin T (ng/L) | 11.5 (9, 25) | 13 (9, 22) | |
| TC (mmol/L) | 3.9 (2.8, 4.6) | 3.1 (2.7, 3.4) | 0.026 |
| HDL-C (mmol/L) | 1.10 (0.9, 1.3) | 0.97 (0.8, 1.1) | |
| LDL-C (mmol/L) | 2.18 (1.8, 2.9) | 1.83 (1.4, 2.2) | 0.02 |
| Triglycerides (mmol/L) | 1.02 (0.9, 1.7) | 1.22 (1.0, 1.8) | |
| Glucose (mmol/L) | 5.6 (4.9, 6.3) | 5.6 (4.9, 6.6) | |
| HbA1c (mmol/mol) | 36 (33, 38) | 39 (36, 51) | 0.011 |
| GFR (%) | 80 (68, 91) | 90 (75, 95) | 0.03 |
| Carbamide (mmol/L) | 6.2 (5.1, 7.3) | 5.5 (4.7, 6.7) | |
| Creatinine (μmol/L) | 80 (64.8, 91) | 76 (67, 85) | |
| Uric acid (μmol/L) | 317 (258, 396) | 313 (271, 362) | |
## 3.2. Gene expression and circulating proteins in CHD and CTRLs
L-Gal 9, CD206, and NOS2 gene expression in the three AT compartments was overall similar in CHD patients and CTRLs (Table 2). Circulating L-Gal 9 and CD206 levels were similar in the two populations, while NOS2 circulating levels were significantly higher in CHD patients as compared to CTRLs ($$p \leq 0.012$$).
**Table 2**
| Unnamed: 0 | Unnamed: 1 | CTRLs | CHD | p value |
| --- | --- | --- | --- | --- |
| AT L-Gal9 | | | | |
| | SAT | 0.63 (0.4, 1.18) | 0.86 (0.47, 1.25) | 0.52 |
| | PAT | 0.70 (0.47, 1.45) | 1.08 (0.55, 1.47) | 0.23 |
| | EAT | 0.70 (0.56, 0.97) | 0.92 (0.48, 1.26) | 0.46 |
| AT CD206 | | | | |
| | SAT | 0.62 (0.32, 1.00) | 0.57 (0.36, 0.78) | 0.80 |
| | PAT | 0.73 (0.31, 1.11) | 0.72 (0.37, 1.11) | 0.61 |
| | EAT | 0.54 (0.43, 0.95) | 0.57 (0.34, 0.81) | 0.58 |
| AT NOS2 | | | | |
| | SAT | 1.13 (0.78, 2.20) | 0.9 (0.38, 1.74) | 0.29 |
| | PAT | 0.59 (0.43, 1.11) | 0.67 (0.34, 1.26) | 0.93 |
| | EAT | 0.49 (0.37, 0.74) | 0.52 (0.31, 0.90) | 0.70 |
| Circulating levels | Circulating levels | Circulating levels | Circulating levels | Circulating levels |
| L-Gal 9 | ng/mL | 6.21 (5.28, 7.72) | 5.88 (5.01, 7.56) | 0.98 |
| CD206 | ng/mL | 312.04 (234.68, 474.86) | 374.11 (228.71, 895.11) | 0.24 |
| NOS2 | pg/mL | 380.59 (254.11, 497.51) | 573.05 (383.21, 935.67) | 0.01 |
## 3.3. Expression of the selected genes in the AT compartments
In CTRLs, L-Gal 9, and CD206 were similarly expressed in the three AT compartments (Figures 1A,B), whereas NOS2 showed the lowest expression in EAT (Figure 1C), in which values were significantly lower than those relative to SAT ($$p \leq 0.007$$). In CHD patients, no significant differences were found for L-Gal 9 and NOS2 expression (Figures 1D,F), although EAT showed the lowest NOS2 values. CD206 expression was significantly lower in SAT and EAT as compared to PAT ($$p \leq 0.003$$ and $$p \leq 0.006$$, respectively; Figure 1E).
**Figure 1:** *Gene expression of L-Gal 9, CD206, and NOS2 in SAT, PAT, and EAT compartments. Data are reported as RQ values and box plots indicate the median values, 25th and 75th percentiles, while error bars report 10th and 90th percentiles. Gene expression in AT compartments was evaluated through the Friedman’s test coupled to the Wilcoxon signed-rank test and a value of p < 0.05 was considered statistically significant. L-Gal9 expression was similar in all ATs in CTRLs (A) and CHDs (D); CD206 was similarly expressed in CTRLs (B), but significantly lower in SAT and EAT as compared to PAT in CHDs (E); NOS2 expression was significantly lower in EAT vs. SAT in CTRLs (C), but similarly expressed among ATs in CHDs (F)**p < 0.01.*
## 3.4. Correlations in gene expression of L-gal 9, CD206, and NOS2 between the different AT compartments
In the CTRL group, L-Gal 9 expression in SAT positively correlated to that in EAT ($r = 0.765$, $p \leq 0.001$), while CD206 expression in SAT correlated to its expression in PAT ($r = 0.579$, $$p \leq 0.006$$) (Figure 2A). In CHD patients, L-Gal 9, CD206, and NOS2 expression in SAT positively correlated with their expression in EAT ($r = 0.411$, $$p \leq 0.004$$; $r = 0.393$, $$p \leq 0.004$$ and $r = 0.542$, $$p \leq 0.007$$, respectively) (Figure 2B; Supplementary Figures 1A,C,E). In addition, CD206 and NOS2 expression in SAT positively correlated with their expression in PAT ($r = 0.431$, $$p \leq 0.002$$ and $r = 0.556$, $$p \leq 0.007$$, respectively) (Figure 2B; Supplementary Figures 1B,D).
**Figure 2:** *Correlations between the expression of L-Gal9, CD206, and NOS2 in SAT, PAT, and EAT compartments in CTRLs (A) and CHD patients (B). Correlations are reported as heatmap following Spearman ρ analyses, and those still significant after Bonferroni correction are marked in bold.*
No significant correlations were found between circulating L-Gal 9, CD206, and NOS2 levels and their corresponding gene expression in SAT, PAT, and EAT in neither CTRL nor CHD patients (Supplementary Table 1).
## 3.5. Correlations between expression of L-gal 9, CD206, and NOS2 within each AT compartment
When analyzing for the inter-relationships between gene expression of the actual markers within each AT, significant correlations (Bonferroni’s corrected; 18 comparisons, $p \leq 0.002$) were found between L-Gal 9 and CD206 within SAT ($r = 0.647$, $$p \leq 0.002$$) in CTRLs (Figure 2A; Supplementary Figure 2A) and within EAT ($r = 0.561$, $p \leq 0.001$) in CHD (Figure 2B; Supplementary Figures 2B).
## 3.6. Correlations between L-gal 9, CD206, and NOS2 expression and cell markers in the AT compartments in CHD patients
CD163 and CD68, CD3, and CD31, the most representative markers of macrophages, T cells and endothelial cells, respectively, were analyzed in SAT, PAT, and EAT of CHD patients to detect the specific cellular subtypes in each AT compartment. The previously analyzed markers were detectable in the three AT compartments, with an overall similar distribution [27]. When matching these data with L-Gal 9, CD206, and NOS2, we found that L-Gal 9 expression in PAT and EAT positively correlated to CD3 expression ($r = 0.37$, $$p \leq 0.008$$ and $r = 0.373$, $$p \leq 0.007$$, respectively; although no longer statistically significant after Bonferroni’s correction; Figure 3). After testing for multiple corrections, L-Gal 9 expression correlated still significantly to CD68 in EAT ($r = 0.502$, $p \leq 0.0001$; Figure 3), while CD206 expression correlated positively to CD163 and CD68 in all AT compartments ($p \leq 0.0001$, all; Figure 3) and with CD31 in PAT ($r = 0.493$, $p \leq 0.0001$, Figure 3).
**Figure 3:** *Correlations between L-Gal 9, CD206, and NOS2 expression and the cell markers CD68, CD163, CD3, and CD31 in SAT, PAT, and EAT in CHD patients. Correlations are reported as heatmap following Spearman ρ analyses, and those still significant after Bonferroni correction (36 comparisons; p < 0.0014) are marked in bold.*
## 3.7. L-gal 9, CD206, and NOS2 AT expression according to lipid profile and anthropometric characteristics in CHD patients
Higher NOS2 expression was found in both PAT and EAT in CHD subjects with LDL-C levels above compared to those with levels below the median value of 1.8 mmol/L (Figure 4). Also, CD206 expression in PAT was higher in patients with LDL-C above median (Figure 4).
**Figure 4:** *Gene expression of L-Gal 9, CD206, and NOS2 in CHD patients with LDL-C levels below and above the median value (1.8 mmol/L). Data are reported as RQ values and box plots indicate the median values, 25th and 75th percentiles, while error bars report 10th and 90th percentiles. Mann–Whitney U-test was applied and a value of p < 0.05 was considered statistically significant. Above median vs. below median: PAT: higher expression of CD206 and NOS2. In EAT: higher expression of NOS2. *p < 0.05.*
Next, we investigated the association between L-Gal 9, CD206, and NOS2 gene expression in the different compartments and the most relevant anthropometric parameters, also dividing gene expression data in quartiles to show potential cut-off values. We observed in CTRLs only a significant positive correlations between EAT L-Gal 9 expression and BMI (Figure 5A), weight and waist (Supplementary Table 2). In both CTRLs and CHD patients, significant correlations were found between CD206 gene expression in SAT and PAT and subjects’ BMI (Figures 5B–E), although the correlation in PAT appeared less strong and not statistically significant after Bonferroni’s correction, as also did the correlation between CD206 gene expression in SAT and PAT in CTRLs.
**Figure 5:** *Gene expression of L-Gal 9 in CTRLs (A) and CD206 in CTRLs (B,C) and CHD patients (D,E) either stratified based on subject’s BMI quartiles (left graph) or as correlation with continuous BMI values (right graph). Data are reported as RQ values and box plots indicate the median values, 25th and 75th percentiles, while error bars report 10th and 90th percentiles. Spearman ρ analyses were performed and only significant correlations after Bonferroni’s correction (27 comparisons; p < 0.0018) are reported.*
## 3.8. Correlations between circulating levels of L-gal 9, CD206, NOS2, and serum lipids and CRP
No significant correlations were found between circulating levels of L-Gal 9, CD206 and NOS2 and serum lipids and CRP in the CTRL group. In CHD patients, L-Gal 9 circulating levels correlated inversely with HDL-C (r = −0.458, $p \leq 0.001$; Figure 6A) and positively with LDL-C ($r = 0.316$, $$p \leq 0.025$$; Figure 6B), TG ($r = 0.525$, $p \leq 0.001$; Figure 6C) and hsCRP ($r = 0.369$, $$p \leq 0.007$$; Figure 6D). However, after Bonferroni’s correction, the correlation to LDL-C lost its statistical significance (Supplementary Table 3). CD206 and NOS2 circulating levels were not correlated to any serum lipid and to hsCRP, and none of the circulating markers correlated significantly to anthropometric measures (Supplementary Table 3).
**Figure 6:** *Correlations between L-Gal 9 circulating levels and HDL-C (A), LDL-C (B), TG (C), and hsCRP (D) in CHD patients. Spearman ρ analyses were performed and Bonferroni’s correction was applied (6 comparisons; p < 0.008).*
## 4. Discussion
In this clinical case–control study we aimed to explore expression of genes linked to the polarization status of macrophages in SAT, PAT, and EAT in CHD patients undergoing coronary artery bypass grafting and in controls undergoing aortic valve replacement.
Our main findings were that expression of the selected genes, L-Gal 9, CD206 and NOS2, that to some degree reflects the macrophage polarization process, (I) did not differ between CHD and control patients; (II) showed a few differences between the various AT compartments: NOS2 was least expressed in control subjects EAT, whereas CD206 was most expressed in PAT of CHD patients; (III) associated with the presence of macrophages in all compartments, except for NOS2; (IV) in specific cases associated with LDL-C levels and modestly with BMI; (V) did not associate with circulating levels of the markers.
The lack of differences in AT gene expression between CHD and CTRLs might be explained considering that the control cohort was composed of patients undergoing aortic valve replacement, thus presenting a cardiopathy. In fact, very recently it was reported that such patients possess pro-inflammatory activation of ascending aortic AT and EAT [29]. Another explanation might be that serum total cholesterol and LDL-C were higher in CTRLs compared to CHD patients, and some CTRLs had hypertension and diabetes mellitus. Moreover, a large part of the recruited CHD patients were under treatment with aspirin and statins, both known to exert an anti-inflammatory activity [30, 31], thus possibly further attenuating potential differences.
When comparing gene expression levels of the investigated molecules in the different AT compartments, we did not find differences in L-Gal 9 expression between SAT, PAT and EAT in either CTRLs or CHD patients. Several in vitro studies have demonstrated that L-Gal 9 is involved in the modulation of a variety of biological processes, including cell aggregation and adhesion, regulation of T cell pools, as well as the modulation of macrophages polarization [23, 24]. Only a few studies have evaluated L-Gal 9 expression in AT, and demonstrated that it is predominantly expressed in the stromal vascular fraction (SVF), composed of various cell types. It is possible that, overall, L-Gal 9 expression in the various AT compartments is similarly and tightly controlled. We observed lower CD206 expression in SAT and EAT compared to PAT in both groups, statistically significant in CHD patients only, possibly due to the lower number of CTRLs. This is in line with the different origin and vascularization of PAT [2, 7]. Lower CD206 expression in EAT and SAT might be important because in concomitance with atherosclerosis progression the number of M2 macrophages in the plaques decreases [32] in various tissues including AT, with consequent formation of crown-like structures, a typical hallmark of chronic fat tissue inflammation that possibly hints CHD [33]. We found NOS2 to be less expressed in EAT with respect to the other compartments, although statistical significance was reached only compared to SAT in the control group. Very recently, it was reported that adipocytes in EAT generated exosomes containing NOS2 [34]. Thus, it is possible that the NOS2 we measured was adipocyte-derived, rather than of macrophage origin. Accordingly, its expression did not correlate with any macrophage, T cell or endothelial cell marker. The lower NOS2 expression in EAT in controls points to a specific, stricter regulation of this molecule in EAT, which might be lost in CHD (Graphical abstract, point a). Moreover, the EAT and SAT expression of NOS2 was positively associated only in the CHD group, in which also L-Gal 9 and CD206 were positively inter-correlated in EAT, suggesting activation of a compensatory anti-inflammatory mechanism specifically occurring in EAT in presence of CHD (Graphical abstract, point b).
In CHD patients, specifically in EAT, L-Gal 9 expression positively and strongly correlated to CD68 expression, a cellular marker identifying macrophages [35], thus the L-Gal 9 findings described above seems to be mainly related to its macrophage expression. As a consequence, the positive association observed specifically in EAT between L-Gal 9 and CD206 expression in CHD patients might confirm our hypothesis of a L-Gal 9 signaling toward an anti-inflammatory M2 profile in CHD patients, to compensate a local cardiac AT low-grade inflammation [36] (Graphical abstract, point b). Furthermore, specifically in PAT and EAT, a positive, albeit milder association was observed between L-Gal 9 and the expression of T cell receptor CD3. This is particularly of interest as several studies have demonstrated that L-Gal 9 acts also by binding and promoting the activity of death receptor 3, involved in the expansion of regulatory T cells number and activity [37], known to exert a fundamental role in AT homeostasis [38]. CD206 gene expression was strongly associated with that of CD163 and CD68 in all AT analyzed. This could be expected as both CD206 and CD163 are commonly expressed by AT resident macrophages, mostly characterized by an M2-like signature [39, 40], while CD68 is commonly used as a macrophage-specific cell marker [35], regardless of the cell phenotype. Interestingly, specifically in PAT, we observed a positive association between CD31 expression, representing endothelial cells, and CD206. Notably, it has previously been demonstrated in humans that the vascular density and amount of endothelial cells in visceral AT is higher as compared to SAT, with a higher angiogenic and inflammatory profile [41]. Furthermore, a pre-clinical model of infarcted rats showed that the administration of stem cells isolated from PAT promoted myogenic differentiation, with consequent efficient cardiac repair [42]. Hence, the positive association that we observed between CD206 and CD31 expression selectively in PAT in CHD patients may be due to a possible reparatory mechanism.
As for circulating concentration of the investigated molecules, no difference in L-Gal 9 levels between CHD and controls was found. Increased levels of L-Gal 9 have been described in a wide range of pathologic conditions, like autoimmune and infectious diseases and in diabetes mellitus (43–45). Conversely, reduced levels were reported in patients with acute coronary syndrome as compared to patients without CAD [43, 44]. We found L-Gal 9 circulating levels to be positively associated with LDL-C, TG, and hsCRP concentrations, and inversely correlated to HDL-C, only in CHD patients. As previously hypothesized, L-Gal 9 may increase as a compensatory response to the inflammatory environment in CHD subjects. We found no differences in CD206 circulating levels between CHD and control patients. To our knowledge, no study has to date been reported on CD206 serum concentrations in subjects with CVDs. In addition, no correlations were found between serum CD206 concentration and serologic and clinical parameters, neither in CHD nor in CTRLs subjects. Circulating levels of NOS2 were significantly higher in CHD patients vs. CTRLs. NOS2, also known as inducible Nitric Oxide Synthase (iNOS), is a key enzyme synthesizing nitric oxide (NO) under specific inflammatory conditions, including atherosclerosis [46, 47]. Indeed, while small amounts of NO produced by eNOS are known to exhibit atheroprotective effects [48], enhanced NO production upon NOS2 activation leads to cytotoxicity and oxidative stress [49], thus potentially contributing to CVD development. Recent studies revealed that NOS2 can be found in circulating macrovesicles in its inactive form, while activated when macrovesicles are internalized into target cells [50]. As such, septic intensive care unit patients displayed a strong increase in NOS2 circulating levels as compared to non-septic, followed by decreased levels upon effective therapy [50]. Within this picture, the observed slight increase of NOS2 concentration in CHD patients might be compatible with low-grade inflammation (Graphical abstract, point a). NOS2 serum levels did not correlate to any serologic or clinical parameter. We could not show any significant correlation between circulating levels of L-Gal 9, CD206 and NOS2 and their corresponding gene expression in the three AT compartments, suggesting that tissue-specific events may not be reflected in circulatory levels, as previously reported [12].
We observed a modest association between L-Gal 9, CD206 and NOS2 with BMI. SAT expression of CD206 associated significantly with BMI and weight in CHD patients, but not in CTRLs, probably due to the lower number in this group. On the other hand, the positive association found specifically in CTRLs between EAT L-Gal 9 expressions and BMI, weight and waist circumference might be a physiologic compensatory anti-inflammatory mechanisms in response to increased amount of AT. This mechanisms might be lost in CHD patients, possibly contributing to the establishment of a pro-inflammatory environment in this specific AT compartment. The relationship between AT macrophage polarization and anthropometric measures has not been fully investigated yet. It has been reported that pro-inflammatory SAT macrophages were increased with BMI, in parallel with a decrease in their anti-inflammatory counterpart; however this was not found in subjects with BMI <30 kg/m2 and in VAT [51], thus highlighting that this expression pattern specifically occurs in obese SAT.
To our knowledge, very few studies have investigated the influence of circulating cholesterol levels on AT inflammation. The administration of a diet rich in saturated fatty acids in mice led to a significant increase in LDL-C and AT expression of CD206 and CD11c, the latter used to identify M1 macrophages, together with the activation of nuclear factor-kappa B (NF-kB), thus suggesting a general increased AT inflammation [52]. However, to date, no human studies have been reported. Interestingly, we observed that specifically in CHD patients with LDL-C levels above the median (1.8 mmol/L), PAT and EAT NOS2 expressions were significantly increased as compared to subjects with LDL-C levels below median. These observations suggest that CHD subjects with high LDL-C may be characterized by an increased inflammatory environment occurring selectively in cardiac ATs (Graphical abstract, point c).
There are several limitations in our study, mainly related to the impossibility to exclude any degree of subclinical atherosclerosis in the control group, as discussed. Moreover, we cannot fully rule out a potential AT dysfunction in our recruited controls. Finding an appropriate control group for this kind of investigation is challenging. Cardiac surgery indications are a mandatory prerequisite for cardiac AT biopsy collection, which necessarily are based on some sort of cardiac disease, potentially impacting on, or associated with AT dysfunction. Also, a high proportion of CHD patients were treated with aspirin and statins [30, 31], which possibly impacts macrophage polarization. Furthermore, we do not have any imaging information, such as CT scans, to measure the AT volume in the various cardiac and subcutaneous locations, and we have made a selection of polarization markers. Finally, being an observational study, only the associations and not the causality could be investigated. On the other side, the strength of this experimental design is the concomitant availability of the two compartments of cardiac AT in addition to the pre-sternal subcutaneous AT, thus allowing us to separately study the macrophage polarization at a molecular level in the different AT compartments.
In conclusion, through the analysis of macrophage polarization markers in pericardial, epicardial and subcutaneous AT, this study suggests that CHD patients might be characterized by an increased low-grade inflammation specifically occurring in EAT. A compensatory anti-inflammatory mechanism involving the L-Gal 9-CD206 axis might be a possible consequence, as the tight regulation of pro-inflammatory NOS2 signaling observed in the controls may partly be lost in CHD patients. Hence, cardiac, and especially EAT macrophage polarization might be considered a promising field of investigation to target more precisely the inflammatory status in cardiovascular disease.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by Regional Ethics Committee of North Norway (#$\frac{2016}{411}$). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
BP carried out the experiments, acquired the data, and wrote the first draft of the manuscript and performed the statistical analyses. TT, BB, SÅ, and CH recruited the subjects, performed the assessments of patients, and critically reviewed the manuscript for intellectual content. IS, TO, HA, SS, and NR conceived and designed the study and handled the founding. IS, TO, and NR handled supervision and critically reviewed the manuscript for intellectual content. All the authors read and approved the final manuscript.
## Funding
We thank the Italian Society of Pharmacology (Italy) for the scholarship to BP and the Stein Erik Hagens Foundation for Clinical Heart Research, Oslo (Norway) for financial support. The funder had no role in the design of the study, in the collection, analyses, or interpretation of data, in the writing of the manuscript or in the decision to publish the results.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcvm.2023.1055069/full#supplementary-material
## References
1. Shao C, Wang J, Tian J, Tang Y. **Coronary artery disease: from mechanism to clinical practice**. *Adv Exp Med Biol* (2020) **1177** 1-36. DOI: 10.1007/978-981-15-2517-9_1
2. Perez-Miguelsanz J, Jiménez-Ortega V, Cano-Barquilla P, Garaulet M, Esquifino AI, Varela-Moreiras G. **Early appearance of Epicardial adipose tissue through human development**. *Nutrients* (2021) **13** 2906. DOI: 10.3390/nu13092906
3. Ansaldo AM, Montecucco F, Sahebkar A, Dallegri F, Carbone F. **Epicardial adipose tissue and cardiovascular diseases**. *Int J Cardiol* (2019) **278** 254-12. DOI: 10.1016/j.ijcard.2018.09.089
4. Yudkin JS, Eringa E, Stehouwer CDA. **“Vasocrine” signalling from perivascular fat: a mechanism linking insulin resistance to vascular disease**. *Lancet (London, England)* (2005) **365** 1817-20. DOI: 10.1016/S0140-6736(05)66585-3
5. Wu Y, Zhang A, Hamilton DJ, Deng T. **Epicardial fat in the maintenance of cardiovascular health**. *Methodist Debakey Cardiovasc J* (2017) **13** 20-4. DOI: 10.14797/mdcj-13-1-20
6. Iacobellis G. *Epicardial and pericardial fat: close, but very different obesity* (2009)
7. Si Y, Cui Z, Liu J, Ding Z, Han C, Wang R. **Pericardial adipose tissue is an independent risk factor of coronary artery disease and is associated with risk factors of coronary artery disease**. *J Int Med Res* (2020) **48** 737. DOI: 10.1177/0300060520926737
8. Iacobellis G, Bianco AC. **Epicardial adipose tissue: emerging physiological, pathophysiological and clinical features**. *Trends Endocrinol Metab* (2011) **22** 450-7. DOI: 10.1016/j.tem.2011.07.003
9. Bragina AE, Tarzimanova AI, Osadchiy KK, Rodionova YN, Bayutina DA, Bragina GI. **Relationship of pericardial fat tissue with cardiovascular risk factors in patients without cardiovascular diseases**. *Metab Syndr Relat Disord* (2021) **19** 524-12. DOI: 10.1089/met.2021.0045
10. Ahmed B, Sultana R, Greene MW. **Adipose tissue and insulin resistance in obese**. *Biomed Pharmacother* (2021) **137** 111315. DOI: 10.1016/j.biopha.2021.111315
11. Shan D, Dou G, Yang J, Wang X, Wang J, Zhang W. **Epicardial adipose tissue volume is associated with high risk plaque profiles in suspect CAD patients**. *Oxidative Med Cell Longev* (2021) **2021** 1-10. DOI: 10.1155/2021/6663948
12. Djaberi R, Schuijf JD, van Werkhoven JM, Nucifora G, Jukema JW, Bax JJ. **Relation of epicardial adipose tissue to coronary atherosclerosis**. *Am J Cardiol* (2008) **102** 1602-7. DOI: 10.1016/j.amjcard.2008.08.010
13. van Meijeren AR, Ties D, de Koning M-SLY, van Dijk R, van Blokland IV, Lizana Veloz P. **Association of epicardial adipose tissue with different stages of coronary artery disease: a cross-sectional UK biobank cardiovascular magnetic resonance imaging substudy**. *Int J Cardiol Hear Vasc* (2022) **40** 101006. DOI: 10.1016/j.ijcha.2022.101006
14. Villasante Fricke AC, Iacobellis G. **Epicardial adipose tissue: clinical biomarker of cardio-metabolic risk**. *Int J Mol Sci* (2019) **20** 5989. DOI: 10.3390/ijms20235989
15. Bradley D, Shantaram D, Smith A, Hsueh WA. **Adipose tissue T regulatory cells: implications for health and disease**. *Adv Exp Med Biol* (2021) **1278** 125-9. DOI: 10.1007/978-981-15-6407-9_8
16. Kunz HE, Hart CR, Gries KJ, Parvizi M, Laurenti M, Dalla Man C. **Adipose tissue macrophage populations and inflammation are associated with systemic inflammation and insulin resistance in obesity**. *Am J Physiol Endocrinol Metab* (2021) **321** E105-21. DOI: 10.1152/ajpendo.00070.2021
17. Kane H, Lynch L. **Innate immune control of adipose tissue homeostasis**. *Trends Immunol* (2019) **40** 857-2. DOI: 10.1016/j.it.2019.07.006
18. Viola A, Munari F, Sánchez-Rodríguez R, Scolaro T, Castegna A. **The metabolic signature of macrophage responses**. *Front Immunol* (2019) **10** 1462. DOI: 10.3389/fimmu.2019.01462
19. Ruggiero AD, Key C-CC, Kavanagh K. **Adipose tissue macrophage polarization in healthy and unhealthy obesity**. *Front Nutr* (2021) **8** 625331. DOI: 10.3389/fnut.2021.625331
20. Eshghjoo S, Kim DM, Jayaraman A, Sun Y, Alaniz RC. **Macrophage polarization in atherosclerosis**. *Genes (Basel)* (2022) **13** 756. DOI: 10.3390/genes13050756
21. Hirata Y, Tabata M, Kurobe H, Motoki T, Akaike M, Nishio C. **Coronary atherosclerosis is associated with macrophage polarization in epicardial adipose tissue**. *J Am Coll Cardiol* (2011) **58** 248-5. DOI: 10.1016/j.jacc.2011.01.048
22. Karlsson M, Zhang C, Méar L, Zhong W, Digre A, Katona B. **A single-cell type transcriptomics map of human tissues**. *Sci Adv* (2021) **7**. DOI: 10.1126/sciadv.abh2169
23. Lv R, Bao Q, Li Y. **Regulation of M1-type and M2-type macrophage polarization in RAW264.7 cells by Galectin-9**. *Mol Med Rep* (2017) **16** 9111-9. DOI: 10.3892/mmr.2017.7719
24. Yu J, Zhu R, Yu K, Wang Y, Ding Y, Zhong Y. **Galectin-9: a suppressor of atherosclerosis?**. *Front Immunol* (2020) **11** 604265. DOI: 10.3389/fimmu.2020.604265
25. Igarashi Y, Nawaz A, Kado T, Bilal M, Kuwano T, Yamamoto S. **Partial depletion of CD206-positive M2-like macrophages induces proliferation of beige progenitors and enhances browning after cold stimulation**. *Sci Rep [Internet]* (2018) **8** 14567. DOI: 10.1038/s41598-018-32803-6
26. Kieler M, Hofmann M, Schabbauer G. **More than just protein building blocks: how amino acids and related metabolic pathways fuel macrophage polarization**. *FEBS J* (2021) **288** 3694-14. DOI: 10.1111/febs.15715
27. Åkra S, Seljeflot I, Braathen B, Bratseth V, Hansen CH, Arnesen H. **The NLRP3 inflammasome activation in subcutaneous, epicardial and pericardial adipose tissue in patients with coronary heart disease undergoing coronary by-pass surgery**. *Atheroscler Plus [Internet]* (2022) **48** 47-54. DOI: 10.1016/j.athplu.2022.03.005
28. Livak KJ, Schmittgen TD. **Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) method**. *Methods* (2001) **25** 402-8. DOI: 10.1006/meth.2001.1262
29. Shi K, Anmin R, Cai J, Qi Y, Han W, Li M. **Ascending aortic perivascular adipose tissue inflammation associates with aortic valve disease**. *J Cardiol* (2022) **80** 240-8. DOI: 10.1016/j.jjcc.2022.04.004
30. Weidmann L, Obeid S, Mach F, Shahin M, Yousif N, Denegri A. **Pre-existing treatment with aspirin or statins influences clinical presentation, infarct size and inflammation in patients with de novo acute coronary syndromes**. *Int J Cardiol* (2019) **275** 171-8. DOI: 10.1016/j.ijcard.2018.10.050
31. Taqueti VR, Ridker PM. **Lipid-lowering and anti-inflammatory benefits of statin therapy: more than meets the plaque**. *Cardiovasc Imag* (2017) **10** e006676. DOI: 10.1161/CIRCIMAGING.117.006676
32. Bisgaard LS, Mogensen CK, Rosendahl A, Cucak H, Nielsen LB, Rasmussen SE. **Bone marrow-derived and peritoneal macrophages have different inflammatory response to oxLDL and M1/M2 marker expression – implications for atherosclerosis research**. *Sci Rep [Internet].* (2016) **6** 35234. DOI: 10.1038/srep35234
33. Fan W, Si Y, Xing E, Feng Z, Ding Z, Liu Y. **Human epicardial adipose tissue inflammation correlates with coronary artery disease**. *Cytokine [Internet]* (2023) **162** 156119. DOI: 10.1016/j.cyto.2022.156119
34. Lin J-R, Ding Q, Xu L-L-Q, Huang J, Zhang Z-B, Chen X-H. **Brown adipocyte ADRB3 mediates Cardioprotection via suppressing Exosomal iNOS**. *Circ Res* (2022) **131** 133-7. DOI: 10.1161/CIRCRESAHA.121.320470
35. Jia Q, Morgan-Bathke ME, Jensen MD. **Adipose tissue macrophage burden, systemic inflammation, and insulin resistance**. *Am J Physiol Endocrinol Metab* (2020) **319** E254-64. DOI: 10.1152/ajpendo.00109.2020
36. Pierzynová A, Šrámek J, Cinkajzlová A, Kratochvílová H, Lindner J, Haluzík M. **The number and phenotype of myocardial and adipose tissue CD68+ cells is associated with cardiovascular and metabolic disease in heart surgery patients**. *Nutr Metab Cardiovasc Dis* (2019) **29** 946-5. DOI: 10.1016/j.numecd.2019.05.063
37. Madireddi S, Eun S-Y, Mehta AK, Birta A, Zajonc DM, Niki T. **Regulatory T cell-mediated suppression of inflammation induced by DR3 signaling is dependent on Galectin-9**. *J Immunol* (2017) **199** 2721-8. DOI: 10.4049/jimmunol.1700575
38. Feuerer M, Herrero L, Cipolletta D, Naaz A, Wong J, Nayer A. **Lean, but not obese, fat is enriched for a unique population of regulatory T cells that affect metabolic parameters**. *Nat Med* (2009) **15** 930-9. DOI: 10.1038/nm.2002
39. Rőszer T. **Understanding the mysterious M2 macrophage through activation markers and effector mechanisms**. *Mediat Inflamm* (2015) **2015** 816460-16. DOI: 10.1155/2015/816460
40. Nawaz A, Tobe K. **M2-like macrophages serve as a niche for adipocyte progenitors in adipose tissue**. *J Diabetes Investig* (2019) **10** 1394-00. DOI: 10.1111/jdi.13114
41. Villaret A, Galitzky J, Decaunes P, Estève D, Marques M-A, Sengenès C. **Adipose tissue endothelial cells from obese human subjects: differences among depots in angiogenic, metabolic, and inflammatory gene expression and cellular senescence**. *Diabetes* (2010) **59** 2755-63. DOI: 10.2337/db10-0398
42. Wang X, Liu X, Zhang H, Nie L, Chen M, Ding Z. **Reconstitute the damaged heart via the dual reparative roles of pericardial adipose-derived flk-1+ stem cells**. *Int J Cardiol* (2016) **202** 256-4. DOI: 10.1016/j.ijcard.2015.09.002
43. Xie J-H, Zhu R-R, Zhao L, Zhong Y-C, Zeng Q-T. **Down-regulation and clinical implication of Galectin-9 levels in patients with acute coronary syndrome and chronic kidney disease**. *Curr Med Sci* (2020) **40** 662-12. DOI: 10.1007/s11596-020-2238-5
44. Zhu R, Liu C, Tang H, Zeng Q, Wang X, Zhu Z. **Serum Galectin-9 levels are associated with coronary artery disease in Chinese individuals**. *Mediat Inflamm* (2015) **2015** 457167-13. DOI: 10.1155/2015/457167
45. Kurose Y, Wada J, Kanzaki M, Teshigawara S, Nakatsuka A, Murakami K. **Serum galectin-9 levels are elevated in the patients with type 2 diabetes and chronic kidney disease**. *BMC Nephrol* (2013) **14** 23. DOI: 10.1186/1471-2369-14-23
46. Wilmes V, Scheiper S, Roehr W, Niess C, Kippenberger S, Steinhorst K. **Increased inducible nitric oxide synthase (iNOS) expression in human myocardial infarction**. *Int J Legal Med* (2020) **134** 575-1. DOI: 10.1007/s00414-019-02051-y
47. Habib SS, Al-Regaiey KA, Al-Khlaiwi T, Habib SM, Bashir S, Al-Hussain F. **Serum inducible and endothelial nitric oxide synthase in coronary artery disease patients with type 2 diabetes mellitus**. *Eur Rev Med Pharmacol Sci* (2022) **26** 3695-02. DOI: 10.26355/eurrev_202205_28865
48. Lee J, Bae EH, Ma SK, Kim SW. **Altered nitric oxide system in cardiovascular and renal diseases**. *Chonnam Med J* (2016) **52** 81-90. DOI: 10.4068/cmj.2016.52.2.81
49. Lind M, Hayes A, Caprnda M, Petrovic D, Rodrigo L, Kruzliak P. **Inducible nitric oxide synthase: good or bad?**. *Biomed Pharmacother* (2017) **93** 370-5. DOI: 10.1016/j.biopha.2017.06.036
50. Webber RJ, Sweet RM, Webber DS. **Inducible nitric oxide synthase in circulating microvesicles: discovery, evolution, and evidence as a novel biomarker and the probable causative agent for sepsis**. *J Appl Lab Med* (2019) **3** 698-1. DOI: 10.1373/jalm.2018.026377
51. Lesna IK, Cejkova S, Kralova A, Fronek J, Petras M, Sekerkova A. **Human adipose tissue accumulation is associated with pro-inflammatory changes in subcutaneous rather than visceral adipose tissue**. *Nutr Diabetes* (2017) **7** e264. DOI: 10.1038/nutd.2017.15
52. Enos RT, Davis JM, Velázquez KT, McClellan JL, Day SD, Carnevale KA. **Influence of dietary saturated fat content on adiposity, macrophage behavior, inflammation, and metabolism: composition matters**. *J Lipid Res* (2013) **54** 152-3. DOI: 10.1194/jlr.M030700
|
---
title: Identification of co-expressed central genes and transcription factors in atherosclerosis-related
intracranial aneurysm
authors:
- Quan Zhang
- Hengfang Liu
- Min Zhang
- Fang Liu
- Tiantian Liu
journal: Frontiers in Neurology
year: 2023
pmcid: PMC10017537
doi: 10.3389/fneur.2023.1055456
license: CC BY 4.0
---
# Identification of co-expressed central genes and transcription factors in atherosclerosis-related intracranial aneurysm
## Abstract
### Background
Numerous clinical studies have shown that atherosclerosis is one of the risk factors for intracranial aneurysms. Calcifications in the intracranial aneurysm walls are frequently correlated with atherosclerosis. However, the pathogenesis of atherosclerosis-related intracranial aneurysms remains unclear. This study aims to investigate this mechanism.
### Methods
The Gene Expression Omnibus (GEO) database was used to download the gene expression profiles for atherosclerosis (GSE100927) and intracranial aneurysms (GSE75436). Following the identification of the common differentially expressed genes (DEGs) of atherosclerosis and intracranial aneurysm, the network creation of protein interactions, functional annotation, the identification of hub genes, and co-expression analysis were conducted. Thereafter, we predicted the transcription factors (TF) of hub genes and verified their expressions.
### Results
A total of 270 common (62 downregulated and 208 upregulated) DEGs were identified for subsequent analysis. Functional analyses highlighted the significant role of phagocytosis, cytotoxicity, and T-cell receptor signaling pathways in this disease progression. Eight hub genes were identified and verified, namely, CCR5, FCGR3A, IL10RA, ITGAX, LCP2, PTPRC, TLR2, and TYROBP. Two TFs were also predicted and verified, which were IKZF1 and SPI1.
### Conclusion
Intracranial aneurysms are correlated with atherosclerosis. We identified several hub genes for atherosclerosis-related intracranial aneurysms and explored the underlying pathogenesis. These discoveries may provide new insights for future experiments and clinical practice.
## 1. Introduction
Intracranial aneurysm (IA) is a prevalent disease that affects ~$3\%$ of the population [1]. The rupture of an IA leads to subarachnoid hemorrhage, with a high risk of morbidity and death. Although the pathogenesis of IA remains unclear, it may be closely related to atherosclerosis (AS). An increasing number of studies have found that IAs are frequently complicated by AS, resulting in a worse prognosis. Killer-Oberpfalzer et al. found atherosclerotic lesions in all their deaths from cystic IA [2]. Evidence supports the hypothesis that atherosclerosis, inflammation, and degenerative changes in aneurysm walls play considerable roles in the development of IA, and the presence of atherosclerotic plaques in the aneurysm wall may contribute to the degeneration and rupture of IA [3, 4]. Aneurysm wall enhancement increases the instability of IA [5]; it is related to an increased level of atherogenic proteins and a decreased level of anti-atherosclerotic proteins, and atherosclerosis can be detected when these enhanced arterial walls are observed in vitro [6, 7]. In addition, inflammation-related atherosclerotic changes and neovascularization of the aneurysm wall have been found in larger unruptured IAs [4], and increased lipid infiltration was observed in the ruptured cerebral aneurysm wall [8]. In summary, atherosclerosis-related intracranial aneurysms (AS-related IA) are more unstable and require further care in clinical practice.
Although atherosclerosis is considered one of the risk factors for IA, the pathogenic mechanism of the complication of IA and AS remains unknown but may be connected to inflammation, smooth muscle cell (SMC) proliferation, and macrophage phagocytosis [9]. Currently, histological studies have demonstrated that the SMC phenotype, lipoprotein buildup, and production of foam cells in intracranial aneurysm walls are similar to the alterations in atherosclerotic artery walls (10–12), indicating that there may be some common mechanisms leading to the occurrence of these two diseases and triggering the onset of AS-related IA. The increased rupture risk of AS-related IA may be mainly caused by atherosclerosis-induced phenotypic modulation of SMC in the aneurysm media layer [13]. The adventitia of intracranial aneurysms comprises collagen fibers, encasing the media layer primarily composed of the SMC and extracellular matrix (ECM), while the intima is the invasive site of atherosclerotic plaques. When atherosclerosis occurs on the wall of intracranial aneurysms, the SMC of the media layer transforms to the matrix remodeling phenotype, resulting in ECM dysfunctional remodeling and the destruction of elastic fibers [14]. These pathological alterations reduce the stability of the media layer of the aneurysm wall, increasing the risk of AS-related IA rupture (Figure 1).
**Figure 1:** *Structural alterations of common rupture sites in AS-related IA.*
With the gradual revelation of the close association between IA and AS, there is still a lack of effective treatments, and new strategies are urgently required to prevent corresponding adverse prognoses. This study aimed to identify the transcriptome signature of AS-related IA. We retrieved differentially expressed genes (DEGs) of IA and AS from the Gene Expression Omnibus (GEO) database and used integrative bioinformatics tools to uncover functional pathways, potential hub genes, and transcription factors. Our findings are expected to shed new insights into the pathogenic mechanisms and treatments of IAs complicated with atherosclerosis.
## 2.1. Data source
GEO (http://www.ncbi.nlm.nih.gov/geo/) is a vast online database containing various high-throughput sequencing data types. We downloaded sequencing datasets of IA (GSE75436) and atherosclerotic vascular specimens (GSE100927) from the GEO database. GSE75436 comprised 15 IA wall tissues and 15 matched control superficial temporal artery walls. GSE100927 comprised 69 atherosclerotic samples and 35 control arteries without atherosclerosis.
## 2.2. Differential expression analysis
First, the acquired data were normalized, background adjusted, and log2 transformed; the probes without gene annotation were removed, and the values of duplicate probes were averaged. We performed differential gene expression analysis on GSE75436 and GSE100927 using the “limma” R package (https:/www.bioconductor.org/packages/3.5/bioc/html/limma.html) [15]. DEGs were identified as genes with an adjusted p-value of <0.05 and a |logFC| of >1. Subsequently, we used Venny2.1 (http://bioinfogp.cnb.csic.es/tools/venny/index.html) to generate a Venn diagram of the intersection of the DEGs in these two datasets. *Removing* genes with opposite expression trends, we obtained the common DEGs (co-DEGs) in these two diseases.
## 2.3. Enrichment analyses of DEGs
Gene Ontology (GO) is a database that describes the related biological processes, molecular functions, and cellular components for gene collections. Kyoto Encyclopedia of Genes and Genomes (KEGG) is supported by a database containing functional annotation and gene pathway enrichment across multiple species. The co-DEGs were submitted to enrichment analyses using the clusterProfiler package [16], with an adjusted p-value of <0.05 serving as the screening criteria. The results were displayed using the Ggplot2 package (https://ggplot2.tidyverse.org).
## 2.4. Protein–protein interaction network construction and module analysis
The Search Tool for the Retrieval of Interacting Genes (STRING; http://string-db.org) is a database that can search for potential relationships between proteins [17] using the STRING database to construct a PPI network of co-DEGs with a combined score of >0.40 for meaningful interactions. To visualize the PPI network, we imported the results into Cytoscape (http://www.cytoscape.org) [18]; subsequently, the MCODE plugin was used to find the potential meaningful gene modules in co-DEGs. Finally, we conducted an enrichment analysis on the most valuable gene modules.
## 2.5. Selection and analysis of candidate hub genes
Using the cytoHubba plugin in Cytoscape, we analyzed the entire PPI network. DEGs were analyzed in cytoHubba using 12 algorithms (MCC, DMNC, MNC, Degree, EPC, BottleNeck, EcCentricity, Closeness, Radiality, Betweenness, Stress, and ClusteringCoefficient), and each algorithm recorded the top 20 genes. Upset charts were constructed, calculating the frequency of gene registrations. Subsequently, we selected the genes reported six times or more as potential hub genes. GeneMANIA (http://www.genemania.org/) [19] is a dependable instrument for determining gene correlations. Candidate hub genes were imported into GeneMANIA for analysis.
## 2.6. Verification and analysis of hub genes
Two external datasets, GSE43292 (AS) and GSE122897 (IA), were used to verify the expression of candidate hub genes. GSE43292 included 32 carotid atherosclerotic plaques paired with 32 distant macroscopic control tissue samples; GSE122897 included 44 IA samples and 16 control intracranial cortical arterials. Data between groups were compared using the mean t-test, with a p-value of <0.05 as the standard for significant differences. The candidate genes that passed the verification were considered hub genes, and enrichment analysis was performed on them.
## 2.7. Prediction and verification of transcription factors
ChEA3 (https://maayanlab.cloud/chea3/) is an internet transcription factor (TF) enrichment analysis instrument that can predict the regulatory relationship between TFs and their corresponding target genes [20]. The ENCODE TF target library contains ChIP-seq experiments from humans and mice. We imported hub genes into the ChEA3 database, set library = “ENCODE”, and predicted the top 10 TFs corresponding to them. Thereafter, we verified the expression levels of these TFs in GSE75436 and GSE100927.
## 3.1. Differential expression analysis
Differential expression analysis revealed that GSE75436 included 2,389 DEGs comprising 1,374 upregulated DEGs and 1,015 downregulated DEGs; GSE100927 contained 442 DEGs consisting of 323 upregulated DEGs and 119 downregulated DEGs (Figure 2A). DEGs with the same expression trend in IA and AS were considered co-DEGs. Taking the intersection of DEGs and removing genes that had opposite expression trends in the two diseases, we obtained 208 upregulated co-DEGs and 62 downregulated co-DEGs (Figure 2B).
**Figure 2:** *Results of differential expression analysis and intersection Venn diagram. (A) The volcano map of GSE75436 and GSE100927. Bright red color indicates upregulated genes; blue color indicates downregulated ones. (B) From the DEGs of the two datasets, 270 co-DEGs were selected.*
## 3.2. Enrichment analyses of DEGs
We performed enrichment analysis on 270 co-DEGs to explore their potential biological functions and pathways. According to GO analysis, these genes were primarily enriched in leukocyte-mediated immunity, leukocyte cell-cell adhesion, immune receptor activity, and endocytic vesicle, which were related to leukocyte immunity and cell phagocytosis (Figure 3B). Meanwhile, co-DEGs were substantially related to the phagosome, lipid, and atherosclerosis, and the B-cell receptor signaling pathway, as determined by KEGG analysis (Figure 3C). These results provide additional evidence that atherosclerosis induces the development of IA and reflect that immune response, phagocytosis, lipid accumulation, and other factors are implicated in the onset and progression of AS-related IA.
**Figure 3:** *PPI network and enrichment analysis of common DEGs. (A) PPI network of common DEGs. Red signifies upregulated genes, while blue denotes downregulated genes. (B, C) The outcomes of GO and KEGG pathway enrichment analyses. It was deemed significant if the adjusted P-value of <0.05.*
## 3.3. Construction and module analysis of the PPI network
A PPI network of co-DEGs with 227 nodes and 2,379 interaction pairings was established (Figure 3A). Using the MCODE plugin (set K-core = 2, degree cutoff = 2, max depth = 100, and node score cutoff = 0.2), we analyzed this network and identified the most significant gene module (score = 28.857, 36 genes, and 505 interaction pairs) (Figure 4A). Interestingly, all genes in this module are upregulated co-DEGs, and heatmaps display their expression levels in GSE75436 (Figure 4B) and GSE100927 (Figure 4C). Moreover, we performed an enrichment analysis on this gene module. GO analysis revealed that these genes were predominantly engaged in immunologic inflammation and cell response regulation (Figure 4D), and KEGG analysis revealed their involvement in a toll-like receptor signaling pathway, a chemokine signaling route, lipid and atherosclerosis, and other pathways (Figure 4E).
**Figure 4:** *MCODE-identified genes module and corresponding enrichment analysis results. (A) One essential module of gene clustering. (B, C) Distribution of gene expression levels in the module. (D, E) The genes module underwent GO and KEGG analysis. It was deemed significant if the adjusted P-value < 0.05.*
## 3.4. Selection and analysis of candidate hub genes
We analyzed the PPI network of co-DEGs using 12 algorithms in the cytoHubba plugin of Cytoscape software, and each algorithm obtained the top 20 candidate hub genes. According to the upset plot (Figure 5A), genes simultaneously selected by six or more algorithms were considered candidate hub genes, namely, PTPRC, TNF, ITGAM, TYROBP, IL1B, CSF1R, FCGR3A, IRF8, LCP2, TLR2, CYBB, CCR5, ITGAX, IL10RA, and C1QA, which were all upregulated genes. Based on the GeneMANIA database, we created gene co-expression networks and demonstrated their associated functions. *These* genes exhibited a complicated PPI network with $88.81\%$ co-expression, $7.24\%$ co-localization, $2.45\%$ prediction, and $1.51\%$ shared protein domains (Figure 5B).
**Figure 5:** *Candidate hub gene co-expression network and upset plot. (A) The intersection of the top 20 genes within 12 methods of the cytoHubba module. (B) GeneMANIA was used to analyze candidate hub genes and associated co-expression genes.*
## 3.5. Verification and analysis of hub genes
We verified the expression of 15 candidate hub genes with GSE43292 (AS) and GSE122897 (IA). All candidate hub genes were significantly upregulated in GSE43292 (Figure 6). While eight genes were remarkably upregulated in GSE122897, and the expression of candidate hub genes showed an overall upward trend in IA, except for IL1B and TNF (Figure 7). *Eight* genes were confirmed as hub genes of AS-related IA, namely, CCR5, FCGR3A, IL10RA, ITGAX, LCP2, PTPRC, TLR2, and TYROBP. Supported by the GeneCards database (https://www.genecards.org/), Table 1 displays their real names and associated functions. Interestingly, each hub gene corresponds to the MCODE algorithm's most significant gene module. Following the enrichment analysis of these genes, GO analysis demonstrated that they are mainly involved in the immune response, macrophage phagocytic function, and cytotoxicity (Figure 8A). Subsequently, their involvement in natural killer cell-mediated cytotoxicity, FcR-mediated phagocytosis, and the T cell receptor (TCR) signaling pathway were revealed by KEGG analysis (Figure 8B).
**Figure 6:** *The hub genes expression level in GSE43292. The mean t-test is used for the comparison of the two datasets. Statistical significance was determined when the P-value < 0.05. AS, atherosclerotic plaque; CON, control artery samples. ***p < 0.001.* **Figure 7:** *The hub genes expression level in GSE122897. The mean t-test is used for the comparison of the two datasets. Statistical significance was determined when the P-value < 0.05. IA, intracranial aneurysm samples; CON, control intracranial cortical arterials. *p < 0.05; **p < 0.01. ns, p > 0.05.* TABLE_PLACEHOLDER:Table 1 **Figure 8:** *Chord diagram for enrichment analysis of hub genes. (A, B) GO and KEGG analysis of hub genes. The circles on the right and left, respectively, indicate pathways and the corresponding hub genes.*
## 3.6. Exploration and authentication of TFs
We predicted the top 10 TFs that may regulate hub gene expression using the ChEA3 database (Figure 9A). Thereafter, we verified the expression of these TFs in datasets. In total, two TFs, IKZF1 and SPI1, were significantly upregulated in GSE75436 (Figure 9B) and GSE100927 (Figure 9C). Subsequently, a network diagram of the TFs and hub genes was constructed. A total of five hub genes (CCR5, IL10RA, LCP2, TYROBP, and PTPRC) were coregulated by these two TFs (Figure 9D).
**Figure 9:** *Exploration and verification of TFs and their regulatory networks. (A) Top 10 related TFs predicted using the ChEA3 database. (B, C) TFs expression level in GSE75436 and GSE100927. The mean t-test is used for the comparison of the two datasets. Statistical significance was determined when the P-value < 0.05. IA, intracranial aneurysm wall tissue; AS, atherosclerotic samples; CON, control arteries. *p < 0.05; **p < 0.01; ***p < 0.001. (D) The regulatory network of TFs. The hub genes were highlighted in red, whereas TFs were highlighted in yellow.*
## 4. Discussion
This study determined new targets for preventing and treating AS-related IA, revealing the potential biological mechanism in the pathogenesis. A total of 270 co-DEGs were identified, and then eight hub genes were selected and verified, namely, CCR5, FCGR3A, IL10RA, ITGAX, LCP2, PTPRC, TLR2, and TYROBP. After the enrichment analysis of hub genes, we found that these genes were principally enriched in cytotoxicity, phagocytosis, and TCR signaling pathways. SPI1 and IKZF1, two transcription factors, were discovered to be significant for the development of this disease, and they jointly regulate five hub genes, namely, CCR5, IL10RA, LCP2, TYROBP, and PTPRC.
The enrichment analysis of hub genes showed that NK cell-mediated cytotoxicity, FcγR-mediated phagocytosis, and the TCR signaling pathway might play considerable roles in the pathogenesis of AS-related IA. The atherosclerotic tissue is rich in NK cells that express many biomarkers, such as IFN-γ. NK cell activation may play a significant role in the exacerbation of AS [21], suggesting that the increased cytotoxicity mediated by NK cells may contribute to the development of AS. Moreover, the migration ability of NK cells in the peripheral blood of patients with IA is also enhanced [22]. NK cells may aggregate and activate in the aneurysm wall during the pathogenesis, which mediates cytotoxicity to promote disease progression. In addition, phagocytosis appears to be activated during pathogenesis. FcγR is a receptor for the Fc portion of IgG, which activates the mitogen-activated protein kinase signaling pathway by mediating low-density lipoprotein immune complexes (LDL-ICs), thereby activating macrophages [23, 24]. Subsequently, macrophage infiltration and its polarization toward the M1 phenotype increase the risk of IA pathogenesis and rupture [25]. Inflammatory macrophages in the arterial wall can uptake LDL-ICs through FcγRI and transform them into foam cells [26], forming atherosclerotic plaques. FcγR may activate inflammatory macrophages and promote their phagocytosis to induce the aggregation of foam cells, leading to the development of atherosclerotic plaque in the aneurysm wall, inducing the deterioration of IA. The activation of the TCR signaling pathway will lead to the cascade reaction of the PKCθ-IKK-NFκB pathway [27], stimulating the NFκB-mediated inflammatory response and participating in the pathology process of AS-related IA.
As a receptor of inflammatory CC-chemokines, CCR5 can increase intracellular calcium ion levels to transduce signals. CCR5 transports blood monocytes to atherosclerotic plaques to promote disease progression [28]. Cipriani et al. treated AS model mice with a CCR5 antagonist, which resulted in a $70\%$ reduction in plaque volume and a $50\%$ attenuation of monocyte/macrophage infiltration [29]. T cells in the wall of IA express high levels of the chemokine receptor CCR5 [30]. We speculate that CCR5 may promote disease progression by participating in the chemotactic process of inflammatory macrophages in the aneurysm wall. FCGR3A, also known as CD16, is involved in mediating cytotoxicity. Previous studies have found more CD16+ intermediate monocytes in patients with IA [31]. Decreased CD16 monocyte subsets are also associated with a decrease in subclinical AS [32]. Combined with our study, FCGR3A may induce disease development by promoting the cytotoxic effect of monocytes and macrophages. The protein encoded by IL10RA is a receptor for interleukin 10 (IL10); It can mediate immunosuppressive signals and reduce inflammatory responses. Patients with IA exhibit a decrease in IL-10, suggesting that the low IL-10 level in vivo may be associated with the development of IA [33]. Moreover, the IL10RA was highly expressed in AS [34], indicating that the IL10RA-mediated inhibition of inflammation is similarly active in AS. Therefore, IL10RA may play a protective role in AS-related IAs by mediating the anti-inflammatory effect of IL-10, and activating the expression of IL10RA can effectively prevent the occurrence of AS-related IA. ITGAX, also known as CD11c, can mediate cell–cell interactions in inflammation, monocyte adhesion, and chemotaxis. The decrease in CD11c + cells can reduce the progression of abdominal aortic aneurysms (AAA) [35]. According to our results, ITGAX may have a similar function in IA. Simultaneously, the high expression of ITGAX is a prominent feature of unstable carotid atherosclerotic plaques. CD11c + macrophages gather in vulnerable plaques, resulting in the deterioration of AS [36]. The effect of ITGAX on AS-related IA may depend on affecting the adhesion and chemotaxis of monocytes.
PTPRC, also known as CD45, encodes a leukocyte antigen that regulates the immune response of T and B cells. PTPRC is involved in the progression of AS as a regulatory T cell-related gene [37]. Hosaka et al. found the infiltration of CD45+ cells in IA walls [38], reflecting the involvement of PTPRC in the pathogenesis of IA. PTPRC may promote the inflammatory response of AS-related IA by regulating immune lymphocytes. TLR2 activates the inflammatory response and cytokine secretion by activating the TLR2-Myd88-NF-κB pathway and can also activate immune cells to promote apoptosis. Multiple studies have demonstrated that the TLR2-Myd88-NF-κB pathway is activated in IA and AS [39, 40]. TLR2 may promote the pathogenesis of AS-related IA mainly by activating the inflammatory response mediated by the TLR2-Myd88-NF-κB pathway and apoptosis. TYROBP, also known as DAP12, encodes a protein that can activate TFs, such as NF-κB, and promote cellular inflammatory responses. Previous studies have found that the expression level of TYROBP is significantly upregulated in the atherosclerotic tissue and AAA, and TYROBP promotes the pathogenesis of AAA through the activation of the NK cell-mediated cytotoxicity pathway [41, 42]. Combined with our results, TYROBP may also have such a role in the progression of AS-related IA, which may lead to disease by activating NF-κB and affecting NK cell-mediated cytotoxicity. The association between LCP2 and AS-related IA is still unclear, and the underlying mechanism requires additional investigation.
Subsequently, we predicted and verified the TFs of hub genes. Among the top 10 TFs predicted from the ChEA3 database, IKZF1 and SPI1 passed the verification, and they jointly regulated five hub genes, namely, CCR5, IL10RA, LCP2, TYROBP, and PTPRC. IKZF1 is considered a transcriptional regulator of hematopoietic differentiation and participates in the development of lymphocytes, B cells, and T cells [43]. IKZF1 can promote the production of the inflammatory cytokine INF-γ by regulating the balance of Th1/Th2 [44]. Increased IFN-γ levels can be observed in patients with IA, especially when the aneurysm ruptures [33]. Moreover, the severity of AS is associated with genetic polymorphisms in the arterial IFN-γ gene [45]. IFN-γ affects immune cells, endothelial cells, and SMCs [46, 47], leading to the progression of AS and may play a similar role in the pathogenesis of AS-related IA. IKZF1 is involved in regulating the expression of CCR5, IL10RA, LCP2, and TYROBP and can also promote the production of INF-γ to affect the pathogenesis process. IKZF1 may become a new therapeutic target. Another TF, SPI1, regulates LCP2, PTPRC, and TYROBP. SPI1 was significantly upregulated in aortic atherosclerotic plaques in Tibetan minipigs [48]. In addition, SPI1 was identified as a significant regulator in peripheral blood samples of patients with IA [49], which also corresponds to the results obtained in our study. Due to the lack of relevant studies, the relationship between SPI1 and AS-related IA still needs to be explored.
Our findings contribute to elucidating the mechanism of the link between IA and AS. However, there are still some flaws in our study. First, the findings of our retrospective study need to be further confirmed with external data. Second, hub genes need to be further verified experimentally in in vitro models. Complementing these shortcomings is the focus of our future study.
## 5. Conclusion
We identified and verified eight hub genes and two TFs for AS-related IA, providing new study directions and therapeutic targets for this disease. *Eight* genes, namely, CCR5, FCGR3A, IL10RA, ITGAX, LCP2, PTPRC, TLR2, and TYROBP, are considered hub genes, and their pathway enrichment results focus on phagocytosis, NK cell-mediated cytotoxicity, and the TCR signaling pathway. Moreover, IKZF1 and SPI1 were identified as the TFs of hub genes and jointly involved in regulating the expression of five genes, namely, CCR5, IL10RA, LCP2, TYROBP, and PTPRC.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the patients/participants or patients/participants' legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.
## Author contributions
QZ completed the data analysis and the writing of the manuscript. HL critically revised the manuscript. MZ checked for omissions in the study and provided comments. FL and TL participated in the formulation of the draft study design. The final manuscript has been read and approved by all authors.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Thompson BG, Brown RD Jr, Amin-Hanjani S, Broderick JP, Cockroft KM, Connolly ES Jr. **Guidelines for the management of patients with unruptured intracranial aneurysms: a guideline for healthcare professionals from the American Heart Association/American Stroke Association**. *Stroke.* (2015) **46** 2368-400. DOI: 10.1161/STR.0000000000000070
2. Killer-Oberpfalzer M, Aichholzer M, Weis S, Richling B, Jones R, Virmani R. **Histological analysis of clipped human intracranial aneurysms and parent arteries with short-term follow-up**. *Cardiovasc Pathol.* (2012) **21** 299-306. DOI: 10.1016/j.carpath.2011.09.010
3. Hashimoto Y, Matsushige T, Shimonaga K, Hosogai M, Kaneko M, Ono C. **Vessel wall imaging predicts the presence of atherosclerotic lesions in unruptured intracranial aneurysms**. *World Neurosurg.* (2019) **132** e775-82. DOI: 10.1016/j.wneu.2019.08.019
4. Matsushige T, Shimonaga K, Ishii D, Sakamoto S, Hosogai M, Hashimoto Y. **Vessel wall imaging of evolving unruptured intracranial aneurysms**. *Stroke.* (2019) **50** 1891-4. DOI: 10.1161/STROKEAHA.119.025245
5. Vergouwen MDI, Backes D, van der Schaaf IC, Hendrikse J, Kleinloog R, Algra A. **Gadolinium enhancement of the aneurysm wall in unruptured intracranial aneurysms is associated with an increased risk of aneurysm instability: a follow-up study**. *Am J Neuroradiol.* (2019) **40** 1112-6. DOI: 10.3174/ajnr.A6105
6. Ishii D, Matsushige T, Sakamoto S, Shimonaga K, Akiyama Y, Okazaki T. **Decreased antiatherogenic protein levels are associated with aneurysm structure alterations in MR vessel wall imaging**. *J Stroke Cerebrovasc Dis.* (2019) **28** 2221-7. DOI: 10.1016/j.jstrokecerebrovasdis.2019.05.002
7. Ishii D, Choi A, Piscopo A, Mehdi Z, Raghuram A, Zanaty M. **Increased concentrations of atherogenic proteins in aneurysm sac are associated with wall enhancement of unruptured intracranial aneurysm**. *Transl Stroke Res.* (2022) **13** 577-82. DOI: 10.1007/s12975-021-00975-5
8. Ou C, Qian Y, Zhang X, Liu J, Liu W, Su H. **Elevated lipid infiltration is associated with cerebral aneurysm rupture**. *Front Neurol.* (2020) **11** 154. DOI: 10.3389/fneur.2020.00154
9. Oka M, Shimo S, Ohno N, Imai H, Abekura Y, Koseki H. **Dedifferentiation of smooth muscle cells in intracranial aneurysms and its potential contribution to the pathogenesis**. *Sci Rep.* (2020) **10** 8330. DOI: 10.1038/s41598-020-65361-x
10. Coen M, Burkhardt K, Bijlenga P, Gabbiani G, Schaller K, Kövari E. **Smooth muscle cells of human intracranial aneurysms assume phenotypic features similar to those of the atherosclerotic plaque**. *Cardiovasc Pathol.* (2013) **22** 339-44. DOI: 10.1016/j.carpath.2013.01.083
11. Ollikainen E, Tulamo R, Lehti S, Lee-Rueckert M, Hernesniemi J, Niemelä M. **Smooth muscle cell foam cell formation, apolipoproteins, and ABCA1 in intracranial aneurysms: implications for lipid accumulation as a promoter of aneurysm wall rupture**. *J Neuropathol Exp Neurol.* (2016) **75** 689-99. DOI: 10.1093/jnen/nlw041
12. Frösen J, Tulamo R, Paetau A, Laaksamo E, Korja M, Laakso A. **Saccular intracranial aneurysm: pathology and mechanisms**. *Acta Neuropathol.* (2012) **123** 773-86. DOI: 10.1007/s00401-011-0939-3
13. Starke RM, Chalouhi N, Ding D, Raper DM, Mckisic MS, Owens GK. **Vascular smooth muscle cells in cerebral aneurysm pathogenesis**. *Transl Stroke Res.* (2014) **5** 338-46. DOI: 10.1007/s12975-013-0290-1
14. Texakalidis P, Sweid A, Mouchtouris N, Peterson EC, Sioka C, Rangel-Castilla L. **Aneurysm formation, growth, and rupture: the biology and physics of cerebral aneurysms**. *World Neurosurg.* (2019) **130** 277-84. DOI: 10.1016/j.wneu.2019.07.093
15. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W. **Limma powers differential expression analyses for RNA-sequencing and microarray studies**. *Nucleic Acids Res.* (2015) **43** e47. DOI: 10.1093/nar/gkv007
16. Yu G, Wang LG, Han Y, He QY. **clusterProfiler: an R package for comparing biological themes among gene clusters**. *OMICS.* (2012) **16** 284-7. DOI: 10.1089/omi.2011.0118
17. Szklarczyk D, Gable AL, Lyon D, Junge A, Wyder S, Huerta-Cepas J. **STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets**. *Nucleic Acids Res.* (2019) **47** D607-13. DOI: 10.1093/nar/gky1131
18. Smoot ME, Ono K, Ruscheinski J, Wang PL, Ideker T. **Cytoscape 28: new features for data integration and network visualization**. *Bioinformatics.* (2011) **27** 431-2. DOI: 10.1093/bioinformatics/btq675
19. Warde-Farley D, Donaldson SL, Comes O, Zuberi K, Badrawi R, Chao P. **The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene function**. *Nucleic Acids Res* (2010) **38** W214-20. DOI: 10.1093/nar/gkq537
20. Keenan AB, Torre D, Lachmann A, Leong AK, Wojciechowicz ML, Utti V. **ChEA3: transcription factor enrichment analysis by orthogonal omics integration**. *Nucleic Acids Res.* (2019) **47** W212-24. DOI: 10.1093/nar/gkz446
21. Bonaccorsi I, Spinelli D, Cantoni C, Barillà C, Pipitò N, De Pasquale C. **Symptomatic carotid atherosclerotic plaques are associated with increased infiltration of natural killer (NK) cells and higher serum levels of NK activating receptor ligands**. *Front Immunol.* (2019) **10** 1503. DOI: 10.3389/fimmu.2019.01503
22. Ge P, Liu C, Chan L, Pang Y, Li H, Zhang Q. **High-dimensional immune profiling by mass cytometry revealed the circulating immune cell landscape in patients with intracranial aneurysm**. *Front Immunol.* (2022) **13** 922000. DOI: 10.3389/fimmu.2022.922000
23. Virella G, Muñoz JF, Galbraith GM, Gissinger C, Chassereau C, Lopes-Virella MF. **Activation of human monocyte-derived macrophages by immune complexes containing low-density lipoprotein**. *Clin Immunol Immunopathol.* (1995) **75** 179-89. DOI: 10.1006/clin.1995.1069
24. Huang Y, Jaffa A, Koskinen S, Takei A, Lopes-Virella MF. **Oxidized LDL-containing immune complexes induce Fc gamma receptor I-mediated mitogen-activated protein kinase activation in THP-1 macrophages**. *Arterioscler Thromb Vasc Biol.* (1999) **19** 1600-7. DOI: 10.1161/01.ATV.19.7.1600
25. Muhammad S, Chaudhry SR, Dobreva G, Lawton MT, Niemelä M, Hänggi D. **Vascular macrophages as therapeutic targets to treat intracranial aneurysms**. *Front Immunol.* (2021) **12** 630381. DOI: 10.3389/fimmu.2021.630381
26. Lopes-Virella MF, Binzafar N, Rackley S, Takei A, La Via M, Virella G. **The uptake of LDL-IC by human macrophages: predominant involvement of the Fc gamma RI receptor**. *Atherosclerosis.* (1997) **135** 161-70. DOI: 10.1016/S0021-9150(97)00157-3
27. Shah K, Al-Haidari A, Sun J, Kazi JU. **T cell receptor (TCR) signaling in health and disease**. *Signal Transduct Target Ther.* (2021) **6** 412. DOI: 10.1038/s41392-021-00823-w
28. Tacke F, Alvarez D, Kaplan TJ, Jakubzick C, Spanbroek R, Llodra J. **Monocyte subsets differentially employ CCR2, CCR5, and CX3CR1 to accumulate within atherosclerotic plaques**. *J Clin Invest.* (2007) **117** 185-94. DOI: 10.1172/JCI28549
29. Cipriani S, Francisci D, Mencarelli A, Renga B, Schiaroli E, D'Amore C. **Efficacy of the CCR5 antagonist maraviroc in reducing early, ritonavir-induced atherogenesis and advanced plaque progression in mice**. *Circulation.* (2013) **127** 2114-24. DOI: 10.1161/CIRCULATIONAHA.113.001278
30. Joy MT, Ben Assayag E, Shabashov-Stone D, Liraz-Zaltsman S, Mazzitelli J, Arenas M. **CCR5 Is a therapeutic target for recovery after stroke and traumatic brain injury**. *Cell* (2019) **176** 1143-57. DOI: 10.1016/j.cell.2019.01.044
31. Wang J, Cao Y. **Characteristics of circulating monocytes at baseline and after activation in patients with intracranial aneurysm**. *Hum Immunol.* (2020) **81** 41-7. DOI: 10.1016/j.humimm.2019.11.003
32. Poitou C, Dalmas E, Renovato M, Benhamo V, Hajduch F, Abdennour M. **CD14dimCD16+ and CD14+CD16+ monocytes in obesity and during weight loss: relationships with fat mass and subclinical atherosclerosis**. *Arterioscler Thromb Vasc Biol.* (2011) **31** 2322-30. DOI: 10.1161/ATVBAHA.111.230979
33. Zhang HF, Zhao MG, Liang GB Yu CY, He W, Li ZQ. **Dysregulation of CD4(+) T cell subsets in intracranial aneurysm**. *DNA Cell Biol.* (2016) **35** 96-103. DOI: 10.1089/dna.2015.3105
34. Cagnin S, Biscuola M, Patuzzo C, Trabetti E, Pasquali A, Laveder P. **Reconstruction and functional analysis of altered molecular pathways in human atherosclerotic arteries**. *BMC Genomics.* (2009) **10** 13. DOI: 10.1186/1471-2164-10-13
35. Okuno K, Cicalese S, Eguchi S. **Depletion of CD11c+ cell attenuates progression of abdominal aortic aneurysm**. *Clin Sci.* (2020) **134** 33-7. DOI: 10.1042/CS20191083
36. Edsfeldt A, Swart M, Singh P, Dib L, Sun J, Cole JE. **Interferon regulatory factor-5-dependent CD11c+ macrophages contribute to the formation of rupture-prone atherosclerotic plaques**. *Eur Heart J.* (2022) **43** 1864-77. DOI: 10.1093/eurheartj/ehab920
37. Xia M, Wu Q, Chen P, Qian C. **Regulatory T cell-related gene biomarkers in the deterioration of atherosclerosis**. *Front Cardiovasc Med.* (2021) **8** 661709. DOI: 10.3389/fcvm.2021.661709
38. Hosaka K, Downes DP, Nowicki KW, Hoh BL. **Modified murine intracranial aneurysm model: aneurysm formation and rupture by elastase and hypertension**. *J Neurointerv Surg.* (2014) **6** 474-9. DOI: 10.1136/neurintsurg-2013-010788
39. Wang Y, Chen L, Tian Z, Shen X, Wang X, Wu H. **CRISPR-Cas9 mediated gene knockout in human coronary artery endothelial cells reveals a pro-inflammatory role of TLR2**. *Cell Biol Int.* (2018) **42** 187-93. DOI: 10.1002/cbin.10885
40. Zhang X, Wan Y, Feng J, Li M, Jiang Z. **Involvement of TLR2/4-MyD88-NF-κB signaling pathway in the pathogenesis of intracranial aneurysm**. *Mol Med Rep.* (2021) **23** 230. DOI: 10.3892/mmr.2021.11869
41. Wang HM, Gao JH, Lu JL. **Pravastatin improves atherosclerosis in mice with hyperlipidemia by inhibiting TREM-1/DAP12**. *Eur Rev Med Pharmacol Sci.* (2018) **22** 4995-5003. DOI: 10.26355/eurrev_201808_15640
42. Hinterseher I, Schworer CM, Lillvis JH, Stahl E, Erdman R, Gatalica Z. **Immunohistochemical analysis of the natural killer cell cytotoxicity pathway in human abdominal aortic aneurysms**. *Int J Mol Sci.* (2015) **16** 11196-212. DOI: 10.3390/ijms160511196
43. Georgopoulos K, Winandy S, Avitahl N. **The role of the Ikaros gene in lymphocyte development and homeostasis**. *Annu Rev Immunol.* (1997) **15** 155-76. DOI: 10.1146/annurev.immunol.15.1.155
44. Hu SJ, Wen LL, Hu X, Yin XY, Cui Y, Yang S. **IKZF1: a critical role in the pathogenesis of systemic lupus erythematosus?**. *Mod Rheumatol.* (2013) **23** 205-9. DOI: 10.3109/s10165-012-0706-x
45. Esperança JC, Miranda WR, Netto JB, Lima FS, Baumworcel L, Chimelli L. **Polymorphisms in IL-10 and INF-γ genes are associated with early atherosclerosis in coronary but not in carotid arteries: a study of 122 autopsy cases of young adults**. *BBA Clin.* (2015) **3** 214-20. DOI: 10.1016/j.bbacli.2015.02.005
46. McLaren JE, Ramji DP. **Interferon gamma: a master regulator of atherosclerosis**. *Cytokine Growth Factor Rev.* (2009) **20** 125-35. DOI: 10.1016/j.cytogfr.2008.11.003
47. Tabas I, Lichtman AH. **Monocyte-macrophages and T cells in atherosclerosis**. *Immunity.* (2017) **47** 621-34. DOI: 10.1016/j.immuni.2017.09.008
48. Pan Y, Yu C, Huang J, Rong Y, Chen J, Chen M. **Bioinformatics analysis of vascular RNA-seq data revealed hub genes and pathways in a novel Tibetan minipig atherosclerosis model induced by a high fat/cholesterol diet**. *Lipids Health Dis.* (2020) **19** 54. DOI: 10.1186/s12944-020-01222-w
49. Zhao H, Li ST, Zhu J, Hua XM. **Wan L. Analysis of peripheral blood cells' transcriptome in patients with subarachnoid hemorrhage from ruptured aneurysm reveals potential biomarkers**. *World Neurosurg.* (2019) **129** e16-22. DOI: 10.1016/j.wneu.2019.04.125
|
---
title: Muscle progenitor cells are required for skeletal muscle regeneration and prevention
of adipogenesis after limb ischemia
authors:
- Hasan Abbas
- Lindsey A. Olivere
- Michael E. Padgett
- Cameron A. Schmidt
- Brian F. Gilmore
- Timothy J. McCord
- Kevin W. Southerland
- Joseph M. McClung
- Christopher D. Kontos
journal: Frontiers in Cardiovascular Medicine
year: 2023
pmcid: PMC10017542
doi: 10.3389/fcvm.2023.1118738
license: CC BY 4.0
---
# Muscle progenitor cells are required for skeletal muscle regeneration and prevention of adipogenesis after limb ischemia
## Abstract
Skeletal muscle injury in peripheral artery disease (PAD) has been attributed to vascular insufficiency, however evidence has demonstrated that muscle cell responses play a role in determining outcomes in limb ischemia. Here, we demonstrate that genetic ablation of Pax7+ muscle progenitor cells (MPCs) in a model of hindlimb ischemia (HLI) inhibited muscle regeneration following ischemic injury, despite a lack of morphological or physiological changes in resting muscle. Compared to control mice (Pax7WT), the ischemic limb of Pax7-deficient mice (Pax7Δ) was unable to generate significant force 7 or 28 days after HLI. A significant increase in adipose was observed in the ischemic limb 28 days after HLI in Pax7Δ mice, which replaced functional muscle. Adipogenesis in Pax7Δ mice corresponded with a significant increase in PDGFRα+ fibro/adipogenic progenitors (FAPs). Inhibition of FAPs with batimastat decreased muscle adipose but increased fibrosis. In vitro, Pax7Δ MPCs failed to form myotubes but displayed increased adipogenesis. Skeletal muscle from patients with critical limb threatening ischemia displayed increased adipose in more ischemic regions of muscle, which corresponded with fewer satellite cells. Collectively, these data demonstrate that Pax7+ MPCs are required for muscle regeneration after ischemia and suggest that muscle regeneration may be an important therapeutic target in PAD.
## 1. Introduction
Peripheral artery disease (PAD) is caused by atherosclerosis of the peripheral arteries, most commonly the legs. PAD affects over 200 million individuals globally, and it is a major contributor to disease burden in both developing and developed countries [1, 2]. Current treatment options are limited to surgical and percutaneous revascularization approaches [3], both of which have minimal impact on long-term morbidity and mortality [4, 5]. The clinical course of PAD ranges from the milder manifestation of intermittent claudication (IC), resulting in pain with ambulation that resolves with rest, to the more severe critical limb threatening ischemia (CLTI), characterized by pain at rest, either with or without tissue necrosis [3]. Although CLTI affects only 10–$15\%$ of patients with PAD, it results in a substantial burden on the health care system, as these patients often progress to limb amputation and have significantly greater morbidity and mortality [6, 7]. Although therapeutic approaches to PAD primarily target revascularization and tissue perfusion, it has been observed that patients with similar degrees of atherosclerotic vascular occlusion often present with markedly different severity of disease [8, 9], suggesting that blood flow alone may not determine clinical outcomes.
Recent evidence from our group and others supports the idea that skeletal muscle responses to tissue ischemia, and not solely the vascular supply, play an important role in determining the muscle response to limb ischemia (9–12). In mice subjected to hind limb ischemia (HLI), a model of PAD, the genetic background strongly influences outcomes. For example, C57BL/6 mice display not only robust angiogenesis but also a muscle regenerative response that typically leads to full recovery from HLI. In stark contrast, HLI in BALB/c mice typically results in muscle degeneration and auto-amputation, and even muscle that does survive fails to recover function, i.e., force generation [9]. Although this genetic difference was previously attributed to differences in collateral vessel density (13–16), muscle progenitor cells (MPCs) isolated from these strains of mice display markedly different responses to experimental ischemia in vitro, independent of blood supply. This finding is consistent with the differential responses observed in vivo and it suggests muscle cell-specific determinants of the response to ischemia. Although the mechanisms by which skeletal muscle responds to ischemia remain poorly understood, a genetic variant in at least one gene, Bag3, has been linked to this differential ischemic response in mice [12]. However, it is not known whether these effects are at the level of mature muscle cells or MPCs.
MPCs, commonly known as satellite cells, lie between the basal lamina and plasma membrane of skeletal muscle cells and are critical regulators of postnatal myofiber regeneration [17, 18]. MPCs are defined by expression of the Pax3 homolog Pax7, and they serve as a unipotent stem cell population for myogenesis following injury [19]. However, these cells have a limited capacity for self-renewal, and repeated replication cycles may result in depletion of the satellite cell pool [20]. The development of genetically modified mouse models to ablate MPCs has allowed investigation of the role of Pax7+ MPCs in various disease states. In particular, mice inducibly expressing Diphtheria toxin A (DTA) only in Pax7+ cells have been used to demonstrate a requirement for these MPCs in muscle regeneration in a variety of conditions. Most of these studies have been performed using cytotoxic injury models, such as cardiotoxin, freeze injury, or BaCl2 injury [21]. However, it is known that different modes of injury have unique characteristics. For example, glycerol injury results in a more adipogenic phenotype compared to other modes of injury [22]. Although we and others have characterized the skeletal muscle response to ischemia [9, 10, 12, 23, 24], the role of MPCs in this process remains unexplored.
In addition to MPCs, the discovery of a novel subpopulation of fibro/adipogenic progenitors (FAPs) in mature skeletal muscle [25] has led to considerable focus on the role of these cells in pathological skeletal muscle conditions. Histologically and functionally, these cells can be identified in skeletal muscle by their expression of platelet-derived growth factor receptor alpha (PDGFRα) and signaling by this receptor in pathophysiology (26–28). FAPs isolated from skeletal muscle were able to cause white fat infiltration in diseased but not in healthy muscle because myofibers have a significant inhibitory effect on the differentiation of FAPs [29]. This observation suggests an environmental contribution to FAP cell fate. FAP expansion has also been shown to regulate the MPC pool during muscle regeneration in addition to playing a critical role in skeletal muscle homeostasis [27].
Here, we used genetically modified mouse models to explore the role of Pax7+ MPCs in the skeletal muscle response to hind limb ischemia. We demonstrate a near complete absence of skeletal muscle regeneration after HLI in mice following ablation of Pax7+ satellite cells. Furthermore, ischemic, Pax7-deficient muscle displayed a dramatic increase in adipogenesis that was driven at least in part by FAPs. Consistent with these findings in mice, decreased MPC numbers and increased adipogenesis were observed in more ischemic regions of skeletal muscle of CLTI patients. These findings demonstrate the requirement for Pax7+ MPCs in ischemic skeletal muscle regeneration, and they provide important new insights into the pathogenesis of PAD.
## 2.1. Mouse lines and tamoxifen treatment
For satellite cell genetic ablation experiments, Pax7-CreERT2 mice (Jackson Labs Stock: 017763, B6.Cg-Pax7tm1(cre/ERT2)Gaka/J) were crossed to ROSA26DTA mice (Jackson Labs Stock: 009669, C.129P2(B6)-Gt(ROSA)26Sortm1(DTA)Lky/J). Both lines of mice had been backcrossed to C57BL/6 J mice for at least 8 generations at the time of these studies. Mice were given sterile-filtered tamoxifen (Sigma T5648) or corn oil at 75 mg/kg body weight via an intraperitoneal route for 5 days. Following this initial treatment, mice were given tamoxifen (Envigo Teklad Tamoxifen Diet TD.130855) or a control-matched diet (Envigo Teklad global $16\%$ protein diet 2016S) to continue treatment at a lower dose of 50 mg/kg body weight. All mice were used at 8–12 weeks of age unless stated otherwise.
## 2.2. Hindlimb ischemia surgery and perfusion imaging
Hindlimb Ischemia surgery was performed as described previously [10, 30]. Briefly, mice were anaesthetized on a heated pad with inhaled isoflurane (1–$3\%$) in oxygen (1.5 L/min). Prior to surgery, the mice were scanned with a Laser Doppler Perfusion Imager (LDPI, Moor Instruments United States) to quantify baseline perfusion in the hindlimbs. Using sterile surgical instruments (sterilized by autoclaving), a 1-cm incision was made just below the inguinal ligament. Subcutaneous fat was removed and the femoral artery was separated from the neurovascular bundle, taking care not to perforate the femoral vein. A 7–0 silk non-absorbable suture (Sharpoint) was used to ligate the femoral artery above the bifurcation of the lateral circumflex femoral artery, and a ligature was also made below the superficial caudal epigastric artery but above the bifurcation of the popliteal artery. The wound was then closed using an absorbable Vicryl 5–0 suture (Ethicon). A post-operative LDPI scan was then performed to verify complete occlusion of the artery. Mice were provided appropriate pain relief and monitored after surgery to ensure animal welfare.
## 2.3. Tissue collection and muscle processing
Mice were deeply anesthetized with inhaled isoflurane as described above, and the tibialis anterior (TA) and extensor digitorum longus (EDL) muscles were isolated and frozen on liquid nitrogen in Optimal Cutting Temperature (OCT) medium, while the gastrocnemius muscles were flash frozen in liquid nitrogen and stored at −80°C for later tissue analysis. Mice were euthanized by exsanguination or bilateral thoracotomy while still under anesthesia. Tissue sections (8 μm or 30 μm) were cut on a Leica 3150S cryostat at −21° to −25°C and stored on Superfrost Slides.
## 2.4. Immunofluorescence microscopy and image analysis
Tissue sections were fixed in $4\%$ paraformaldehyde (PFA) followed by permeabilization in $0.2\%$ Triton X-100 in Phosphate-Buffered Saline (PBS). After washes in PBS, slides were blocked with $5\%$ normal goat serum (NGS) in PBS for 1 h. Dilutions of antibodies used for immunostaining are listed in the Table 1. For Pax7 immunofluorescence staining, a goat anti-mouse IgG blocking antibody (Jackson Immunoresearch 115–007-003) was used in blocking buffer, and an antigen retrieval step was performed by heating slides in a Cuisinart CPC-6001000-watt pressure cooker at high setting in 10 mM sodium citrate, $0.05\%$ Tween 20, pH 6.0, and allowed to return to room temperature over 20 min prior to incubation with the primary antibody. Slides were incubated overnight at 4°C with antibodies of interest at the dilutions listed in Table 1. The following day, sections were washes 3 times in PBS, followed by incubation at room temperature for 1 h with appropriate Alexa Fluor-conjugated secondary antibody. Slides were then washed 3 times with PBS or PBST followed by a 5-min incubation with a nuclear stain (DAPI or Hoechst) as indicated. After a final PBS rinse, slides were mounted in either Vectashield Antifade Mounting Medium (H1000) or Prolong Gold Antifade Mountant (Invitrogen P36962) and allowed to cure overnight. Brightfield microscopy and epifluorescence microscopy were both performed on a Zeiss Upright AxioImager, while all confocal microscopy was performed on the Zeiss 780 or Zeiss 880 Inverted Confocal Microscope. All image analysis was performed in Zeiss Zen software, IMARIS, or ImageJ with identical thresholds and blinding performed for all signal quantification.
**Table 1**
| Antibody | Manufacturer | Catalog Number | Dilution | Species |
| --- | --- | --- | --- | --- |
| Pax7 | DSHB | Pax7 | 1:50 | Mouse IgG1 |
| Dystrophin | Thermo | RB-9024 | 1:100 | Rabbit |
| Type IIa fibers | DSHB | SC-71 | 1:100 | Mouse IgG1 |
| Type I fibers | DSHB | BA-F8 | 1:100 | Mouse IgG2b |
| Type IIb fibers | DSHB | BF-F3 | 1:100 | Mouse IgM |
| Embryonic myosin heavy chain | DSHB | F1.652 | 1:50 | Mouse IgG1 |
| CD31 | BioRad | MCA2388 | 1:200 | Rat |
| Dystrophin | Abcam | ab3149 (discontinued) | 1:100 | Mouse IgG1 |
| Perilipin | Cell Signaling Technologies | 9349S | 1:200 | Rabbit |
| PDGFRα | Cell Signaling Technologies | 3174S | 1:800 | Rabbit |
| Myogenin | DSHB | F5D | 1:50 | Mouse IgG1 |
| Myosin heavy chain | DSHB | A4.1025 | 1:200 | Mouse IgG2a |
## 2.5. Hematoxylin and eosin staining
Sections were brought to room temperature, then fixed for 10 min in $10\%$ Neutral Buffered Formalin (NBF). The slides were washed with distilled water followed by staining of nuclei with Meyer’s hematoxylin for 4 min. Slides were rinsed under running tap water for 10 min and then differentiated with $0.3\%$ acid-alcohol. An additional rinse with tap water and Scott’s tap water substitute was used to further enhance coloration of the nuclei. Samples were then briefly incubated in $90\%$ ethanol (EtOH) and then stained with alcoholic eosin for 30 s. Slides were then dehydrated in $100\%$ EtOH followed by 2 rinses in xylene or xylene substitute for 2 min each before mounting with Cytoseal resin-based mounting medium. The slides were then allowed to cure overnight prior to imaging.
## 2.6. Muscle contractile measurements
Contractile muscle force was measured as described previously [31]. Briefly, single EDL muscles were isolated and ligated with a 5–0 silk suture at each tendon and maintained in a physiological saline solution (pH 7.6) containing 119 mM NaCl, 5 mM KCl, 1 mM MgSO4, 5 mM NaHCO3, 1.25 mM CaCl2, 1 mM KH2PO4, 10 mM HEPES, and 10 mM dextrose at 30°C under aeration with $95\%$ O2/$5\%$ CO2 throughout the experiment. Muscles were mounted in a bath within the force transducer (Aurora 300B-LR) operated in isometric mode. A 5-min equilibration was performed, during which single twitches were elicited every 30 s with 0.5 msec electrical pulses. Isometric tension was evaluated by 250 msec trains of pulses delivered at 10, 20, 40, 60, 80, 100 and 120 Hz. After the experimental protocol, muscle length was determined with a digital caliper and muscle mass was measured after removing liquid. The cross-sectional area for each muscle was measured, and muscle density was determined as the muscle mass (g) divided by the product of its length (Lo, mm) and cross-sectional area (mm2), expressed in g/mm3. Muscle output was then expressed as isometric tension (N/cm2) determined by dividing the developed tension (N) by the muscle cross-sectional area. In the case of atrophied muscle, absolute tension was used as the measure of force because the cross-sectional muscle area is no longer a reliable measure due to change in muscle density.
## 2.7. Oil red O staining
Tissue sections were fixed in $10\%$ NBF for 4 min and briefly washed under running tap water for 1 to 10 min. After rinsing with $60\%$ isopropanol, samples were stained with freshly prepared and filtered oil red O working solution (Oil Red O powder [Sigma] in $60\%$ isopropanol) for 15 min, then rinsed again with $60\%$ isopropanol. The samples were then lightly stained with Meyer’s hematoxylin and rinsed with distilled water. Slides were mounted in aqueous glycerine jelly and imaged within 2 h.
## 2.8. BODIPY staining
Frozen sections were fixed with $4\%$ PFA in PBS for 10 min followed by 2 washes with PBST for 5 min each. The slides were then incubated for 60 min with 1 μg/ml BODIPY $\frac{493}{503}$ (Invitrogen D3922) in PBST. Following incubation, the slides were washed twice with PBST for 5 min each, then twice with PBS for 5 min each. Slides were finally mounted with Prolong Diamond Antifade Mountant with DAPI (Invitrogen P36962) and imaged immediately.
## 2.9. MicroCT/DiceCT
EDLs were isolated from hindlimbs of mice and fixed immediately in $10\%$ NBF solution overnight. Muscles were then stained using the diceCT protocol, as described previously [32]. Briefly, muscles were incubated in Lugol’s Iodine for 2 nights. The muscles were then scanned in a fixed container at low power in a Nikon XTH 225 ST microCT scanner at a 10-μm or 14-μm resolution. Images were then reconstructed using Nikon automated reconstruction software and analyzed using Avizo to delineate soft tissue densities in false colors.
## 2.10. Batimastat treatment
Pax7-CreERT2; ROSA26DTA mice were all given tamoxifen (75 mg/kg body weight) via i.p. injection for 5 days prior to surgery. 1 day prior to surgery, half the mice were given batimastat (30 mg/kg body weight) as a 3 mg/ml suspension in sterile-filtered PBS with $0.01\%$ Tween-80 via i.p. injection, and the other half were injected with vehicle only. Batimastat injections were given daily until muscle was isolated. Following surgery, the mice were switched to a tamoxifen diet and the muscle was harvested 7 days post-operatively.
## 2.11. Fast Green/Sirius Red staining
Samples were placed in $0.04\%$ Fast Green (Sigma) for 15 min then washed with distilled water. Sections were then incubated in $0.1\%$ Fast Green and $0.04\%$ Sirius Red (Sigma) in saturated picric acid for 30 min. Samples were dehydrated through serial 70, 90, and $100\%$ Ethanol washes and cleared in xylene for 2 min before mounting using Cytoseal mounting medium. Positive and negative controls were run simultaneously to validate the specificity of this assay for collagen.
## 2.12. Myoblast isolation
Hindlimb muscles from mice were dissected, rinsed briefly in sterile PBS, and placed in a 10-cm dish containing DMEM +$1\%$ penicillin/streptomycin (pen/strep). Thereafter, all steps were performed in a biosafety cabinet under sterile conditions. Muscles were cleaned of excess connective tissue and tendons and transferred to a new 10-cm dish containing 5 mL DMEM +$1\%$ pen/strep. The muscle was minced with razor blades for >10 min then transferred to a 50-ml centrifuge tube using a wide-bore pipet. Samples were centrifuged in a tabletop centrifuge for 2 min at 800× g. The medium was aspirated, cells were resuspended in 18 mL DMEM + pen/strep, and 2 ml pronase ($1\%$ solution) was added and the mixture was digested for 1 h at 37°C on a Nutator. The cells were then centrifuged for 3 min at 800 × g and the medium was aspirated. Muscle was then suspended in 10 mL DMEM +$10\%$ FBS + pen/strep and triturated 20 times to loosen cells. The supernatant was filtered through a Steriflip 100-μm vacuum filter and washed with 5 ml DMEM with $10\%$ FBS + pen/strep. The cells were then centrifuged 5 min at 1,000× g and resuspended in 10 mL of growth medium (Ham’s F10 with $20\%$ FBS + pen/strep) and plated on collagen-coated plates.
## 2.13. In vitro myogenic and adipogenic differentiation
Differentiation of isolated myoblasts was stimulated by plating the cells on entactin-collagen-laminin-coated plates in differentiation medium (DMEM supplemented with $2\%$ horse serum, $1\%$ pen/strep, $0.2\%$ amphotericin B, and $0.01\%$ human insulin/transferrin/selenium). Effects of hypoxia were determined by placing cells in a hypoxia chamber (Billups-Rothenberg) at $0\%$ O2 ($95\%$ N2, $5\%$ CO2). Control cells were maintained in normoxia ($21\%$ O2, $5\%$ CO2). Medium was changed daily to ensure cell viability. Adipogenic differentiation was induced by incubating cells for 48 h in medium containing $10\%$ FBS, 0.5 mM isobuylmethylxanthine, 125 nM indomethacin, 1 μM dexamethosone, 850 nM insulin, 1 nM T3 with or without 1 μM rosiglitazone. After 48 h, cells were switched to medium containing $10\%$ FBS, 850 nM insulin, 1 nM T3, and 1 μM rosiglitazone. Cells were placed in either normoxic or hypoxic conditions as described above, and medium was changed every other day to ensure cell viability.
## 2.14. Human skeletal muscle acquisition
Critical limb threatening ischemia (CLTI) patients undergoing above-or below-knee amputations were consented according to an Institutional Review Board (IRB)-approved protocol to donate skeletal muscle tissue from the amputated limb. Muscle samples were collected under sterile conditions in the operating room from both the proximal and distal ends of the gastrocnemius muscle and oriented cross-sectionally in OCT and frozen on liquid nitrogen. The samples were then sectioned in a cryostat and stored at −80°C for subsequent analysis.
## 2.15. Statistical analysis
For each of the analyses, a script was used to blind the reviewer to either the images or the animal treatments to ensure no bias in the analysis. For a comparison of 2 groups, a two-way Student’s t-test was performed in GraphPad Prism, and statistical significance was established at $p \leq 0.05.$ For multiple group comparisons, an ANOVA was first performed to determine whether an effect was present, followed by a t-test for multiple groups with a correction for multiple group testing in GraphPad Prism. Significance was once again established by a corrected p value <0.05.
## 3. Results
To study the role of Pax7+ MPCs in a mouse model of PAD, we crossed Pax7-CreERT2 mice to ROSA26DTA mice. To ablate satellite cells, we injected tamoxifen (Pax7Δ) or corn oil as a control (Pax7WT) for 5 days, followed by femoral artery ligation to induce HLI. Perfusion imaging demonstrated an identical injury and similar perfusion of the ischemic hind limb up to 28 days after HLI surgery in both groups (Supplementary Figure S1C). To maintain MPC ablation, mice were fed a diet supplemented with either corn oil or tamoxifen. To validate the model, muscle sections from the non-ischemic tibialis anterior (TA) muscle were stained for the satellite cell marker Pax7, and in the tamoxifen-treated mice there was a complete absence of satellite cells (Supplementary Figures S1A,B), demonstrating successful ablation of all satellite cells within skeletal muscle.
## 3.1. Satellite cell ablation does not alter resting muscle morphology or physiology
To examine whether satellite cell ablation resulted in changes in resting muscle, we isolated the TA muscle from the contralateral, non-ischemic limbs of Pax7Δ and Pax7WT mice after HLI and compared the skeletal muscle histologically by H&E staining. Muscle morphology and architecture appeared similar in Pax7Δ and Pax7WT mice (Figure 1A), although total cross-sectional area of the TA muscle was significantly reduced in Pax7Δ mice (4.48 ± 0.24 mm2 vs. 5.49 ± 0.07 mm2, $$p \leq 0.0043$$), possibly due to greater muscle hypertrophy in Pax7WT mice after disuse of the ischemic limb. Importantly, however, ex vivo force generation of extensor digitorum longus (EDL) muscle did not differ between Pax7Δ and Pax7WT mice (Figure 1B), demonstrating that absence of satellite cells does not alter resting muscle physiology. Lastly, we examined whether deletion of the endogenous skeletal muscle progenitor cell pool affects muscle fiber type distribution. Staining and quantification of both slow-twitch type 1 fibers, which are highly oxidative, and more glycolytic type IIa, IIb, and IId/x fibers (which are more abundant in TA muscle) demonstrated no differences between the groups, demonstrating that loss of Pax7+ MPCs does not cause a shift in myofiber metabolism at rest (Figures 1C,D).
**Figure 1:** *Pax7+ MPC ablation does not alter resting muscle morphology 1 week after ischemia. (A) H&E stains of skeletal muscle from mice with ablated satellite cells are not distinguishable from those with intact satellite cells (n = 3–4 per group). (B) EDL muscles from mice (n = 5 per group) were isolated, and their ability to generate force was measured on a force transducer. Ablation of satellite cells did not impair the ability of resting skeletal muscle to generate force. Data shown are means +/− SEM. (C) Representative immunostains of type I, IIa, and IIb myofibers in non-ischemic limbs of Pax7WT and Pax7Δ mice. (D) Quantification of relative percentages of each myofiber type in non-ischemic muscle of Pax7WT and Pax7Δ mice. Pax7+ MPC ablation did not alter non-ischemic resting muscle fiber type distribution 1 week after ischemia (n = 3–4 per group). Type IId/x myofibers were quantified by lack of staining for the other three markers. All data shown are means +/− SEM; p = ns for all comparisons. Scale bar = 100 μm.*
## 3.2. Satellite cell ablation in ischemic muscle results in complete absence of regeneration 1 week after ischemia
To determine the effect of MPC ablation after ischemia, Pax7Δ and Pax7WT mice were subjected to unilateral HLI and examined 7 days later. Following ablation of satellite cells, markers of skeletal muscle regeneration (embryonic myosin heavy chain expression and centralized myonuclei) were absent in the ischemic limb of Pax7Δ mice (Figures 2A–C). To exclude the possibility that genetic ablation of MPCs with DTA had a non-specific effect on the vasculature, muscle sections were stained for the endothelial cell marker PECAM (CD31). Not only was the endothelium intact, but the total endothelial area relative to the muscle area was in fact increased in Pax7Δ mice (Figure 2D), suggesting a possible vascular compensation for the muscle loss. Satellite cell activation and proliferation normally occur after muscle injury in general and are observed after limb ischemia as well. 1 week after HLI, Pax7WT mice displayed a significant increase in the number of Pax7+ cells in ischemic TA muscle in contrast to the contralateral, non-ischemic limb, confirming normal satellite cell activation in this model (Figures 2E,F). As expected, this activation was absent in Pax7Δ mice lacking satellite cells, consistent with their inability to regenerate muscle following ischemic injury (Figures 2E,F).
**Figure 2:** *Ablation of Pax7+ MPCs in mice results in a complete lack of a muscle regenerative response 1 week after HLI. (A) Muscle regeneration was examined by staining for embryonic myosin heavy chain (eMHC, red) and endothelial cells (CD31, green), top, and for centralized myonuclei by H&E, bottom. (B–D) Quantification of centralized myonuclei (B) and eMHC (C) demonstrates a complete lack of regenerative response to ischemia. Quantification of CD31 area (D) demonstrates an increase in endothelial area relative to muscle area in Pax7Δ mice (n = 3–4 per group). (E,F) 1 week after HLI surgery, there was a significant increase in the number of Pax7+ cells per high power field in the ischemic TA of Pax7WT mice but not in muscle of Pax7Δ mice. Compared to resting muscle, there was a 10-15-fold increase in the number of Pax7+ cells in injured Pax7WT muscle, consistent with activation of satellite cells following injury (n = 3 per group). Scale bar = 100 μm. All data shown are means +/− SEM. **p < 0.01; ***p < 0.001, by 2-sided t-test.*
## 3.3. Chronic satellite cell ablation in ischemic muscle results in complete absence of regeneration 1 month after ischemia
To investigate the effects of satellite cell ablation on long-term muscle recovery from ischemia, Pax7Δ and Pax7WT mice were subjected to HLI and followed for 14 and 30 days after surgery. Consistent with responses observed in parental C57BL/6 mice, ischemic Pax7WT mice displayed improved muscle architecture at day 14 post-HLI. Expression of eMHC had resolved by this time point, although there were still centralized myonuclei, and an inflammatory infiltrate was still present in the interstitial spaces between muscle fibers (Figure 3A). These features were further improved by day 30, with near compete resolution of inflammation (Figure 3B). In contrast, Pax7Δ mice displayed a persistent absence of muscle regeneration with an accompanying increase in cellularity characteristic of ongoing inflammation (Figures 3A,B). Strikingly, muscle of late stage ischemic Pax7Δ mice displayed a dramatic increase in adipose observed both histologically and by microCT (Figure 3A,B; Supplementary Figure S2), which was also evidenced grossly by the inability of whole muscle tissue to sink in aqueous solution (Supplementary Figure S2A). Whereas distinct, individual myofibers were visualized by microCT in control muscle (Supplementary Figure S2B), EDL muscle from Pax7Δ mice was markedly atrophied and displayed significant soft tissue adipogenic changes (Supplementary Figure S2B). These findings suggested that the chronic absence of satellite cells after ischemic injury resulted not only in a loss of muscle regeneration but also a shift in the cellular makeup of injured muscle. Persistent satellite cell ablation in Pax7Δ mice 30 days after HLI was verified by Pax7 immunostaining (Figures 3C,D). In the non-ischemic limb of Pax7WT mice, satellite cell numbers were similar to the day 7 timepoint, whereas satellite cell number diminished significantly in the ischemic limb by day 30 (~4/hpf compared to ~35/hpf on day 7 post-HLI) and was only slightly higher than in the non-ischemic limb at this stage (Figures 3C,D).
**Figure 3:** *Sustained deletion of satellite cells results in long-term prevention of muscle regeneration after HLI. (A,B) H&E staining and quantification of regenerating fibers, demonstrated by myofibers with centralized myonuclei, of the ischemic TA muscle at 14 days (A) and 30 days (B) after ischemia demonstrated a complete lack of regeneration (n = 3–4 per group). (C,D) 30 days after HLI Pax7+ cells per high power field were significantly increased (2-fold) in the ischemic relative to the non-ischemic TA muscle of Pax7WT mice, although their numbers were diminished compared to 7 days post-HLI. Pax7+ cells were persistently absent in Pax7Δ mice (n = 4 per group). Scale bar = 100 μm. All data shown are means +/− SEM. ***p < 0.001, ****p < 0.0001 by 2-sided t-test.*
## 3.4. Long-term satellite cell ablation in ischemic muscle results in impaired force generation
Because long-term satellite cell ablation resulted in markedly abnormal muscle tissue morphology, we tested ex vivo muscle force generation to determine the functional effects of this injury. *Force* generation in Pax7Δ and Pax7WT mice correlated with histological findings, as there was a significant impairment in both maximal force generation and the time-tension force integral in EDL muscle of Pax7Δ mice compared to that of Pax7WT mice 30 days after ischemia (Figures 4A,B). In stark contrast, force generation in the non-ischemic EDL mirrored that observed on day 7 post-HLI (Figures 4C,D), confirming that resting skeletal muscle is unaffected by satellite cell ablation even after 30 days.
**Figure 4:** *Pax7+ MPC ablation impairs force generation 30 days after HLI. (A) The maximum force generated by the ischemic EDL muscle was significantly lower (p < 0.0001 by 2-way ANOVA) in Pax7Δ mice. (B) Maximum force was unchanged in the non-ischemic limb of Pax7Δ mice. (C,D) The time-tension integral, a measure of work done in a single contraction, of muscle 30 days after HLI mirrored the maximum force data in both ischemic (C) and non-ischemic TA muscle (D) (n = 4–5 per group).*
## 3.5. Ablation of Pax7+ MPCs in mice results in marked fat infiltration of skeletal muscle following ischemia
A key feature of muscle injury is that different modes of injury can result in varying regenerative responses. For example, unlike cardiotoxin-mediated injury, glycerol injection induces a more adipogenic change to the muscle [22]. In contrast, the mdx mouse, a genetic model of muscular dystrophy, fails to accurately recapitulate many of the adipogenic changes observed in patients with muscular dystrophy. The lipid deposition seen in Pax7Δ mice 30 days after HLI is reminiscent not only of that of patients with muscular dystrophy but also of patients with CLTI [33]. To investigate the adipogenic changes that occur in skeletal muscle following ischemic injury, we used two different complementary lipid stains, oil red O and BODIPY $\frac{493}{503}$, to examine fatty changes 7 days after HLI. Oil red O staining showed a small amount of fat deposition in the control Pax7WT TA muscle, which was significantly increased in Pax7Δ muscle (Figure 5A), and these findings were mirrored by the BODIPY staining (data not shown). The increased fat deposition in Pax7Δ muscle after long-term injury resulted in the need to cut thicker tissue (~30 μm) sections, which also resulted in what appeared to be increased non-specific oil red O (Figure 5A) and BODIPY staining (data not shown). To overcome this issue, we immunostained for perilipin, which is selectively localized to the periphery of lipid droplets and thus specifically marks adipose accumulation. Perilipin staining also revealed a significant increase in adipogenesis in Pax7Δ TA muscle compared to that of Pax7WT mice at 7 and 14 days post-HLI (Figures 5A,B), and this difference persisted out to day 30 post-HLI (Figure 5C). These findings demonstrate that the lack of Pax7+ MPCs results in aberrant lipid accumulation, which may contribute to the pathogenesis of PAD.
**Figure 5:** *Ablation of Pax7+ MPCs in mice results in marked fat infiltration within skeletal muscle following ischemia. (A) Oil red O (top) and Perilipin (bottom) staining of the ischemic TA muscle demonstrated significantly increased lipid staining in Pax7Δ mice compared to Pax7WT 7 days after HLI surgery (n = 3–4 per group). (B,C) Perilipin staining and quantification of adipose in the ischemic TA muscle 14 days (B) and 30 days (C) after HLI surgery demonstrated increased lipid staining in Pax7Δ mice compared to Pax7WT (n = 3–4 per group). Scale bars = 1 mm. All data shown are means +/-SEM. * represents p < 0.05, **p < 0.01; ***p < 0.001 by 2-sided t-test.*
## 3.6. Fibro/adipogenic progenitors are significantly increased in Pax7Δ mice after ischemia
To begin to elucidate the origins of the adipogenic changes observed after ischemia in Pax7Δ mice, we explored the potential contribution of fibro/adipogenic progenitor cells (FAPs) to the phenotype. FAPs have been shown to induce adipogenic changes in skeletal muscle in limb girdle muscular dystrophy type II [33] and in other pathological conditions [22]. Additionally, FAPs have been shown to drive adipogenic changes in a variety of metabolic and cardiovascular disorders [26, 27, 34]. Staining ischemic muscle from Pax7Δ and Pax7WT mice for the FAP marker PDGFRα (35–37) demonstrated a significant increase in FAPs in Pax7Δ mice that progressively increased over time after ischemia (Figures 6A–C), consistent with the observed temporal increase in adipogenesis (Figure 5). In contrast, PDGFRα+ area was unchanged in Pax7WT muscle at all timepoints after ischemia. These findings suggest that increased ischemic skeletal muscle adipogenesis following MPC ablation is driven by FAPs.
**Figure 6:** *Ablation of Pax7+ MPCs in mice results in a significant increase in FAPs in skeletal muscle following ischemia. (A-C) PDGFRα staining of the ischemic TA muscle demonstrated significantly increased FAP staining in Pax7Δ mice compared to Pax7WT 7 days (A), 14 days (B), and 30 days (C) after HLI surgery (n = 3–5 per group). Scale bars = 1 mm. All data shown are means +/-SEM. *p < 0.05; **p < 0.01; ***p < 0.001 by 2-sided t-test.*
## 3.7. Batimastat, an FAP inhibitor, limits adipogenesis and promotes fibrosis after ischemia in the absence of satellite cells
Batimastat is a non-specific MMP inhibitor that has been shown to prevent adipogenesis both in isolated FAP cells in vitro and in skeletal muscle in vivo [22, 33]. We reasoned that if FAPs contribute to adipogenesis after HLI in the absence of satellite cells, then treating ischemic mice with batimastat should limit the amount of lipid deposition. Indeed, treatment of Pax7Δ mice with batimastat during recovery from HLI resulted in a significant decrease in oil red O+ and perilipin+ area compared to that observed in vehicle-treated Pax7Δ mice (Figure 7A). Notably, this change was accompanied by a corresponding increase in fibrosis (Figure 7B). Despite this clear difference in phenotype, batimastat did not alter the number of FAPs, as indicated by the lack of a difference in PDGFRα staining (Supplementary Figure S3), consistent with previous reports [33]. Collectively, these findings suggest that in the absence of satellite cells, ischemia drives FAPs to promote adipogenesis, which may play an important role in the pathophysiology of PAD.
**Figure 7:** *Inhibiting FAPs with batimastat reduces adipogenesis and increases fibrosis after HLI in the absence of satellite cells. (A) Batimastat treatment significantly decreased total fat in Pax7Δ ischemic TA muscle as determined by oil red O and perilipin staining 7 days after HLI surgery. (B) Fast Green/Sirius Red staining demonstrated a corresponding significant increase in collagen content in Pax7Δ ischemic TA muscle, consistent with a switch from adipogenesis to fibrosis after inhibition of FAPs (n = 4–6 per group). Scale bars = 100 μm. All data shown are normalized means +/− SEM. **p < 0.01; ***p < 0.001 by 2-sided t-test.*
## 3.8. Isolation and differentiation of myoblasts following satellite cell ablation in vivo results in defective myogenesis and increased adipogenesis in vitro
It is well known that myoblasts isolated from whole muscle tissue retain their ability to differentiate and fuse into mature skeletal myotubes in vitro. Pax7+ satellite cells comprise a small percentage (<$10\%$) of the mononuclear cells isolated from muscle tissue that have the potential to differentiate into muscle (i.e., MPCs). Once MPCs are isolated and plated in vitro, satellite cells rapidly lose expression of Pax7 and differentiate into MyoD-expressing committed myoblasts [38]. Prior studies have demonstrated that deletion of Pax7+ satellite cells in vitro, after plating, does not impair myoblast differentiation [39]. To our knowledge, however, no studies have examined the effect of in vivo ablation of Pax7+ cells on subsequent myoblast differentiation in vitro and whether this might influence isolated myoblasts to differentiate toward an adipogenic lineage. To test this, mice were treated with either tamoxifen or corn oil for 5 days to ablate Pax7+ cells in vivo, then muscle was harvested and mononuclear cells/myoblasts were isolated and plated in vitro. When cultured in muscle differentiation medium, only cells from Pax7WT mice were able to form mature myotubes, as evidenced by expression of the myogenic regulatory factor myogenin and myosin heavy chain (MHC) (Figure 8A). To determine whether MPCs isolated from Pax7WT or Pax7Δ mice have an increased propensity to differentiate into adipocytes, cells were plated in adipogenic medium. Because increased adipogenesis in Pax7Δ mice was observed in vivo in the setting of ischemia, cells were incubated for 12 days under hypoxic conditions to simulate ischemia. Compared to cells from Pax7WT mice, cells isolated from Pax7Δ mice had an increased propensity to form adipocytes, as demonstrated by oil red O staining (Figure 8B). These findings suggest that in the absence of Pax7+ cells, Pax7− cells with the potential to fuse and differentiate into muscle are driven toward an adipocyte lineage, although it is unclear whether these cells are FAPs or if they are derived from some other progenitor cell population.
**Figure 8:** *Myoblasts isolated from Pax7-depleted muscle fail to differentiate under hypoxic conditions and display increased adipogenesis. (A) Pax7WT myoblasts in differentiation medium expressed myosin heavy chain (MHC) under hypoxia whereas Pax7Δ cells failed to fuse and did not express the early differentiation marker myogenin or MHC. (B) When grown in adipogenic medium, Pax7Δ myoblasts had a higher propensity to form oil red O+ lipid droplets. Similar results were observed in three independent experiments. Scale bars = 100 μm.*
## 3.9. Critical limb ischemia patients have increased adipogenesis and fewer satellite cells in regions of greater ischemia
To examine whether the adipogenic changes observed in our preclinical model are also seen clinically, we obtained skeletal muscle tissue from CLTI patients undergoing limb amputation. In this setting, tissue that is farthest from the amputation site (distal) is typically the most ischemic, whereas proximal tissue, closer to the amputation site is less ischemic and often relatively healthy. Paired proximal and distal gastrocnemius muscle samples were obtained from 10 CLTI patients undergoing amputation, and adipose area was determined by perilipin staining. Distal, more ischemic muscle displayed significantly greater adipose area (Figure 9A). Because the increase in adipogenic area in our preclinical model was caused by the ablation of satellite cells prior to ischemia, we investigated whether the increased adipogenesis in the regions of greater ischemia corresponded with a loss or reduction in the number of Pax7+ cells. Immunostaining for Pax7 was performed on paired proximal and distal skeletal muscle sections from each subject. Although Pax7+ cells were still present in all subjects’ distal muscle, we observed significantly fewer Pax7+ cells in distal vs. proximal tissue (Figure 9B). These findings support the possibility that chronic limb ischemia results in loss of satellite cell number and/or satellite cell dysfunction, which leads to increased skeletal muscle adipogenesis and may contribute to the pathogenesis of PAD in general and CLTI in particular.
**Figure 9:** *CLTI patients have increased adipogenesis in more ischemic muscle regions that correspond with decreased Pax7+ cell numbers. (A) Perilipin staining in the gastrocnemius muscle of CLTI patients (n = 10) revealed significantly greater fat deposition in more distal ischemic regions. Scale bar = 1 mm (B) More ischemic distal regions of the same patients in panel (A) had significantly fewer Pax7+ cells. Scale bar = 100 μm. All data shown are paired values from the same patient. **p < 0.01 by a 2-sided ratio paired t-test.*
## 4. Discussion
Although surgical and endovascular approaches to revascularization represent the primary strategy to treat PAD, outcomes remain poor, particularly in CLTI, which results in high rates of subsequent amputation [40, 41]. Moreover, while experimental pro-angiogenic approaches to improve limb perfusion have shown great promise in preclinical models of hindlimb ischemia, they have proven suboptimal in clinical experience [42, 43]. We hypothesized that these poor outcomes might be explained, at least in part, by non-vascular etiologies of CLTI. Our prior results supported this hypothesis by demonstrating that skeletal muscle cell responses to ischemia are independent of blood supply and are strongly influenced by genetic background. However, the role of skeletal muscle regeneration in the response to ischemia and, in particular, the role of muscle progenitor cells in this process, remained unknown. Here, we have demonstrated an absolute requirement for Pax7+ skeletal muscle satellite cells in muscle regeneration following ischemic injury. Furthermore, by continuously feeding mice a tamoxifen-containing diet over 30 days post-HLI, we ensured that there was no repopulation of the satellite cell pool [39], and we demonstrated that the regenerative response to ischemia was entirely muscle-dependent. Although one prior study raised the possibility that, following a critical juvenile period, satellite cells were dispensable for regeneration in the postnatal phase, our results are consistent with studies that demonstrate an absolute requirement for satellite cells during regeneration [44, 45], in our case following ischemia-induced muscle injury. Our data demonstrate that complete recovery from ischemia follows a similar time course as skeletal muscle injuries that are cytotoxic and cryogenic in nature [21, 23].
Staining for endothelial cells in mice lacking satellite cells verified that vascular cells were not targeted non-specifically by DTA after tamoxifen treatment and, therefore, that the observed injury was not likely due to loss of vascular supply. Somewhat surprisingly, we found that capillary density was in fact increased in Pax7Δ mice. Although the mechanisms responsible for this effect are not clear, it is possible that capillary proliferation occurred as a compensatory response to the increased tissue destruction [46]. One caveat in interpreting this result is that decreased muscle area due to atrophy could have falsely increased apparent vascular density. Future studies will be necessary to fully elucidate the nature of the endothelial response during this process, including examination of endothelial cell proliferation, angiogenesis, and collateralization, which are known to occur in the setting of hindlimb ischemia [47].
Using several complementary approaches (oil red O, BODIPY, perilipin), we demonstrated the novel and important finding that in the absence of Pax7+ satellite cells, ischemia induces marked lipid deposition within skeletal muscle. This observation distinguishes the injury in this model from that seen in murine models of muscular dystrophy and cardiotoxin injury, which lack similar adipogenesis. Although the mdx mouse model lacks the extreme fat deposition that is observed in DMD patients [29], a “humanized” mdx model with shortened telomeres and mitochondrial defects did show greater adipogenic changes [48]. These lipid deposits are presumed to be pathogenic, because many skeletal muscle diseases are characterized by increased adipogenesis [49]. Notably, the adipose deposition observed after complete loss of satellite cells in Pax7Δ mice recapitulated findings seen in muscle tissue samples of CLTI patients, who displayed increased adipogenesis in more ischemic, distal regions of the amputated limb. The mice used in this study were on a C57BL/6 background, a strain in which the skeletal muscle is known to be relatively resistant to ischemic injury [10]. Strikingly, the absence of satellite cells completely abrogated the protective effect conferred by C57BL/6 genetic factors, suggesting that satellite cell loss or dysfunction contributes to the CLTI phenotype. Consistent with this observation, we found that more ischemic distal regions of CLTI muscle had significantly fewer Pax7+ satellite cells. It is important to note that the mouse phenotype was induced by the complete ablation of satellite cells after tamoxifen treatment, although it is unclear whether partial loss of Pax7+ cells would result in a similar phenotype. Although satellite cells were still present in more ischemic regions of CLTI tissue, it is possible that they were dysfunctional and unable to contribute to regeneration. Satellite cell dysfunction may not manifest as a decrease in absolute number, but there may instead be epigenetic, post-transcriptional, and/or post-translational alterations that affect satellite cells’ ability to effectively promote regeneration in CLTI patients. Alternatively, the reduction in Pax7+ cell number with ischemia in CLTI may result from a loss due to satellite cell exhaustion reminiscent of phenotypes seen in DMD patients. Future experiments will be necessary to elucidate the exact role that satellite cells play in the pathogenesis of PAD. Gene expression profiling of satellite cells in PAD patients with claudication or CLTI may identify a specific genetic signature that defines the pathophysiology of satellite cells in these conditions. The observed correlation between preclinical and clinical adipose deposition in the setting of limb ischemia supports the biological and clinical relevance of these findings.
FAPs have been shown to play a role in obesity-associated skeletal muscle dysfunction as well as in denervated skeletal muscle [34, 50]. We hypothesized that FAPs were responsible for the increased adipogenesis after ischemia in Pax7Δ mice. To explore this possibility, we treated ischemic Pax7Δ mice with batimastat, a small molecule inhibitor of fibroblast activation protein, a dual specificity serine protease. Batimastat has been shown to inhibit adipogenesis resulting from FAP cell differentiation into adipocytes in both isolated FAPs in culture and skeletal muscle in vivo in a model of limb girdle muscular dystrophy [33]. Indeed, we observed a decrease in the degree of adiposity after batimastat treatment, and this was accompanied by a corresponding increase in fibrosis, supporting the possibility that FAP differentiation into adipocytes was responsible for the observed ischemic lipid deposition. Future studies, such as lineage tracing using an FAP marker like PDGFRα [25], will be necessary to conclusively determine whether FAPs or other progenitor cell types contribute to this fat infiltration. Definitively establishing that FAPs are responsible for the increased skeletal muscle adiposity in the setting of ischemia would likely require a genetic approach, such as ablation of PDGFRα+ FAPs. However, ablation of both Pax7+ cells and PDGFRα+ cells would likely have complex effects that may be difficult to interpret.
Several important questions arise regarding the mechanisms responsible for both the adipogenesis and the switch to a fibrotic phenotype after batimastat treatment. First, what are the paracrine signaling pathways between satellite cells and other muscle progenitor cells, including FAPs, that drive normal myogenesis? Pax7+ cells account for a small percentage of total cells in muscle tissue, yet in typical muscle cell isolates, a number of mononuclear cell types have the capacity to fuse and differentiate into myotubes in vitro, suggesting that the presence of satellite cells confers on other MPCs (e.g., myoblasts, pericytes, FAPs) the ability to differentiate into functional muscle. This likely involves paracrine signaling mechanisms that remain to be fully elucidated, although PDGF-BB and DLL4 have been implicated in driving pericytes toward a myogenic lineage [51]. Second, what are the mechanisms that drive the increased adipogenesis in the absence of satellite cells? *Does a* suppressive signal from satellite cells to FAPs normally prevent adipogenesis, or does the absence of satellite cells activate another pathway to drive adipogenesis? Third, and equally important, does ischemia contribute to these processes, since adipogenesis does not occur in the non-ischemic limb, or are these pathways driven by aberrant regeneration? Future studies will be necessary to elucidate these mechanisms, and it is hoped that such information would lead to the eventual development of therapies for diseases of aberrant muscle stem cell number and/or function, such as CLTI and DMD. Batimastat provides a potential starting point for development of drugs to inhibit adipogenic changes in skeletal muscle. Although an increase in fibrosis in CLTI in place of adipose tissue may not translate into optimal clinical outcomes, it provides an initial strategy to redirect aberrant MPC differentiation and possibly prevent pathological adipogenesis.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by Duke University Institutional Review Board. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Duke University Institutional Animal Care and Use Committee.
## Author contributions
HA and CK designed the research study. HA conducted all in vivo and in vitro experiments and performed data analysis. LO, BG, and KS isolated human skeletal muscle and assisted in human muscle staining and experiments. MP performed animal husbandry, genotyping and HLI surgeries. TM performed histological data analysis and assisted with editing the manuscript. CS and JM conducted muscle force generation experiments. HA wrote the manuscript, and CK co-wrote and edited the manuscript. All authors contributed to the article and approved the submitted version.
## Funding
This study was supported in part by NIH grants HL124444, HL118661, and HL156009 to CK, HL125695 to JM, and by a grant from the Duke University School of Medicine to CK for microCT studies through the Shared Materials Instrumentation Facility. BG was supported by grant F32 HL136125 from the NIH. KS was supported in part by a KL2 award through the Duke Clinical and Translational Science Award TR002553 from the NIH. LO was the recipient of a Eugene A. Stead Student Research Scholarship and a Poindexter Scholars in Basic Sciences Award from the Duke University School of Medicine.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcvm.2023.1118738/full#supplementary-material
## References
1. Hirsch AT, Haskal ZJ, Hertzer NR, Bakal CW, Creager MA, Halperin JL. **ACC/AHA 2005 Practice Guidelines for the management of patients with peripheral arterial disease (lower extremity, renal, mesenteric, and abdominal aortic): a collaborative report from the American Association for Vascular Surgery/Society for Vascular Surgery, Society for Cardiovascular Angiography and Interventions, Society for Vascular Medicine and Biology, Society of Interventional Radiology, and the ACC/AHA Task Force on Practice Guidelines (Writing Committee to Develop Guidelines for the Management of Patients With Peripheral Arterial Disease): endorsed by the American Association of Cardiovascular and Pulmonary Rehabilitation; National Heart, Lung, and Blood Institute; Society for Vascular Nursing; TransAtlantic Inter-Society Consensus; and Vascular Disease Foundation**. *Circulation* (2006) **113** e463-654. DOI: 10.1161/CIRCULATIONAHA.106.174526
2. Fowkes FGR, Rudan D, Rudan I, Aboyans V, Denenberg JO, McDermott MM. **Comparison of global estimates of prevalence and risk factors for peripheral artery disease in 2000 and 2010: a systematic review and analysis**. *Lancet* (2013) **382** 1329-40. DOI: 10.1016/S0140-6736(13)61249-0
3. Rooke TW, Hirsch AT, Misra S, Sidawy AN, Beckman JA, Findeiss L. **Management of patients with peripheral artery disease (compilation of 2005 and 2011 ACCF/AHA guideline recommendations): a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines**. *J Am Coll Cardiol* (2013) **61** 1555-70. DOI: 10.1016/j.jacc.2013.01.004
4. Taylor SM, Cull DL, Kalbaugh CA, Senter HF, Langan EM, Carsten CG. **Comparison of interventional outcomes according to preoperative indication: a single center analysis of 2,240 limb revascularizations**. *J Am Coll Surg* (2009) **208** 770-8. DOI: 10.1016/j.jamcollsurg.2009.01.025
5. Norgren L, Hiatt WR, Dormandy JA, Nehler MR, Harris KA, Fowkes FG. **Inter-society consensus for the management of peripheral arterial disease (TASC II)**. *J Vasc Surg* (2007) **45** S5-S67. DOI: 10.1016/j.jvs.2006.12.037
6. Dormandy J, Heeck L, Vig S. **The fate of patients with critical leg ischemia**. *Semin Vasc Surg* (1999) **12** 142-7. PMID: 10777241
7. Dormandy JA, Rutherford RB. **Management of peripheral arterial disease (PAD). TASC Working Group. TransAtlantic Inter-Society Concensus (TASC)**. *J Vasc Surg* (2000) **31** S1-S296. PMID: 10666287
8. Mätzke S, Lepäntalo M. **Claudication does not always precede critical leg ischemia**. *Vasc Med* (2001) **6** 77-80. DOI: 10.1177/1358836X0100600202
9. McClung JM, McCord TJ, Southerland K, Schmidt CA, Padgett ME, Ryan TE. **Subacute limb ischemia induces skeletal muscle injury in genetically susceptible mice independent of vascular density**. *J Vasc Surg* (2015) **64** 1101-1111.e2. DOI: 10.1016/j.jvs.2015.06.139
10. McClung JM, McCord TJ, Keum S, Johnson S, Annex BH, Marchuk DA. **Skeletal muscle-specific genetic determinants contribute to the differential strain-dependent effects of hindlimb ischemia in mice**. *Am J Pathol* (2012) **180** 2156-69. DOI: 10.1016/j.ajpath.2012.01.032
11. McClung JM, Reinardy JL, Mueller SB, McCord TJ, Kontos CD, Brown DA. **Muscle cell derived angiopoietin-1 contributes to both myogenesis and angiogenesis in the ischemic environment**. *Front Physiol* (2015) **6** 161. DOI: 10.3389/fphys.2015.00161
12. McClung JM, McCord TJ, Ryan TE, Schmidt CA, Green TD, Southerland KW. **BAG3 (Bcl-2-associated athanogene-3) coding variant in mice determines susceptibility to ischemic limb muscle myopathy by directing autophagy**. *Circulation* (2017) **136** 281-96. DOI: 10.1161/circulationaha.116.024873
13. Chalothorn D, Clayton JA, Zhang H, Pomp D, Faber JE. **Collateral density, remodeling, and VEGF-A expression differ widely between mouse strains**. *Physiol Genomics* (2007) **30** 179-91. DOI: 10.1152/physiolgenomics.00047.2007
14. Clayton JA, Chalothorn D, Faber JE. **Vascular endothelial growth factor-A specifies formation of native collaterals and regulates collateral growth in ischemia**. *Circ Res* (2008) **103** 1027-36. DOI: 10.1161/CIRCRESAHA.108.181115
15. Chalothorn D, Faber JE. **Strain-dependent variation in collateral circulatory function in mouse hindlimb**. *Physiol Genomics* (2010) **42** 469-79. DOI: 10.1152/physiolgenomics.00070.2010
16. Wang S, Zhang H, Wiltshire T, Sealock R, Faber JE. **Genetic dissection of the Canq1 locus governing variation in extent of the collateral circulation**. *PLoS One* (2012) **7** e31910. DOI: 10.1371/journal.pone.0031910
17. Seale P, Sabourin LA, Girgis-Gabardo A, Mansouri A, Gruss P, Rudnicki MA. **Pax7 is required for the specification of myogenic satellite cells**. *Cells* (2000) **102** 777-86. DOI: 10.1016/S0092-8674(00)00066-0
18. Wang YX, Rudnicki MA. **Satellite cells, the engines of muscle repair**. *Nat Rev Mol Cell Biol* (2012) **13** 127-33. DOI: 10.1038/nrm3265
19. Feige P, Brun CE, Ritso M, Rudnicki MA. **Orienting muscle stem cells for regeneration in homeostasis, aging, and disease**. *Cell Stem Cell* (2018) **23** 653-64. DOI: 10.1016/j.stem.2018.10.006
20. Motohashi N, Asakura A. **Muscle satellite cell heterogeneity and self-renewal**. *Front Cell Dev Biol* (2014) **2** 1. DOI: 10.3389/fcell.2014.00001
21. Hardy D, Besnard A, Latil M, Jouvion G, Briand D, Thépenier C. **Comparative study of injury models for studying muscle regeneration in mice**. *PLoS One* (2016) **11** e0147198. DOI: 10.1371/journal.pone.0147198
22. Kopinke D, Roberson EC, Reiter JF. **Ciliary hedgehog signaling restricts injury-induced Adipogenesis**. *Cells* (2017) **170** 340-51.e12. DOI: 10.1016/j.cell.2017.06.035
23. Mohiuddin M, Lee NH, Moon JY, Han WM, Anderson SE, Choi JJ. **Critical limb ischemia induces remodeling of skeletal muscle motor unit, myonuclear-, and mitochondrial-domains**. *Sci Rep* (2019) **9** 9551. DOI: 10.1038/s41598-019-45923-4
24. Ryan TE, Yamaguchi DJ, Schmidt CA, Zeczycki TN, Shaikh SR, Brophy P. **Extensive skeletal muscle cell mitochondriopathy distinguishes critical limb ischemia patients from claudicants**. *JCI Insight* (2019) **3** e123235. DOI: 10.1172/jci.insight.123235
25. Joe AWB, Yi L, Natarajan A, Le Grand F, So L, Wang J. **Muscle injury activates resident fibro/adipogenic progenitors that facilitate myogenesis**. *Nat Cell Biol* (2010) **12** 153-63. DOI: 10.1038/ncb2015
26. Arrighi N, Moratal C, Clément N, Giorgetti-Peraldi S, Peraldi P, Loubat A. **Characterization of adipocytes derived from fibro/adipogenic progenitors resident in human skeletal muscle**. *Cell Death Dis* (2015) **6** e1733. DOI: 10.1038/cddis.2015.79
27. Wosczyna MN, Konishi CT, Perez Carbajal EE, Wang TT, Walsh RA, Gan Q. **Mesenchymal stromal cells are required for regeneration and homeostatic maintenance of skeletal muscle**. *Cell Rep* (2019) **27** 2029-35.e5. DOI: 10.1016/j.celrep.2019.04.074
28. Mueller AA, van Velthoven CT, Fukumoto KD, Cheung TH, Rando TA. **Intronic polyadenylation of PDGFRα in resident stem cells attenuates muscle fibrosis**. *Nature* (2016) **540** 276-9. DOI: 10.1038/nature20160
29. Uezumi A, Fukada S, Yamamoto N, Takeda S, Tsuchida K. **Mesenchymal progenitors distinct from satellite cells contribute to ectopic fat cell formation in skeletal muscle**. *Nat Cell Biol* (2010) **12** 143-52. DOI: 10.1038/ncb2014
30. Padgett ME, McCord TJ, McClung JM, Kontos CD. **Methods for acute and subacute murine hindlimb ischemia**. *J Vis Exp* (2016) **112** e54166. DOI: 10.3791/54166
31. Spangenburg EE, Le Roith D, Ward CW, Bodine SC. **A functional insulin-like growth factor receptor is not necessary for load-induced skeletal muscle hypertrophy**. *J Physiol* (2008) **586** 283-91. DOI: 10.1113/jphysiol.2007.141507
32. Gignac PM, Kley NJ, Clarke JA, Colbert MW, Morhardt AC, Cerio D. **Diffusible iodine-based contrast-enhanced computed tomography (diceCT): an emerging tool for rapid, high-resolution, 3-D imaging of metazoan soft tissues**. *J Anat* (2016) **228** 889-909. DOI: 10.1111/joa.12449
33. Hogarth MW, Defour A, Lazarski C, Gallardo E, Diaz Manera J, Partridge TA. **Fibroadipogenic progenitors are responsible for muscle loss in limb girdle muscular dystrophy 2B**. *Nat Commun* (2019) **10** 2430. DOI: 10.1038/s41467-019-10438-z
34. Buras ED, Converso-Baran K, Davis CS, Akama T, Hikage F, Michele DE. **Fibro-adipogenic remodeling of the diaphragm in obesity-associated respiratory dysfunction**. *Diabetes* (2019) **68** 45-56. DOI: 10.2337/db18-0209
35. Uezumi A, Fukada S, Yamamoto N, Ikemoto-Uezumi M, Nakatani M, Morita M. **Identification and characterization of PDGFRα+ mesenchymal progenitors in human skeletal muscle**. *Cell Death Dis* (2014) **5** e1186. DOI: 10.1038/cddis.2014.161
36. Sun C, Berry WL, Olson LE. **PDGFRα controls the balance of stromal and adipogenic cells during adipose tissue organogenesis**. *Development* (2017) **144** 83-94. DOI: 10.1242/dev.135962
37. Dani C, Pfeifer A. **The complexity of Pdgfr signaling: regulation of adipose progenitor maintenance and adipocyte-myofibroblast transition**. *Stem Cell Investig* (2017) **4** 28. DOI: 10.21037/sci.2017.04.02
38. Liu L, Cheung TH, Charville GW, Rando TA. **Isolation of skeletal muscle stem cells by fluorescence-activated cell sorting**. *Nat Protoc* (2015) **10** 1612-24. DOI: 10.1038/nprot.2015.110
39. von Maltzahn J, Jones AE, Parks RJ, Rudnicki MA. **Pax7 is critical for the normal function of satellite cells in adult skeletal muscle**. *Proc Natl Acad Sci* (2013) **110** 16474-9. DOI: 10.1073/pnas.1307680110
40. Baubeta Fridh E, Andersson M, Thuresson M, Sigvant B, Kragsterman B, Johansson S. **Amputation rates, mortality, and pre-operative comorbidities in patients revascularised for intermittent claudication or critical limb ischaemia: a population based study**. *Eur J Vasc Endovasc Surg* (2017) **54** 480-6. DOI: 10.1016/j.ejvs.2017.07.005
41. Uccioli L, Meloni M, Izzo V, Giurato L, Merolla S, Gandini R. **Critical limb ischemia: current challenges and future prospects**. *Vasc Health Risk Manag* (2018) **14** 63-74. DOI: 10.2147/VHRM.S125065
42. Collinson DJ, Donnelly R. **Therapeutic angiogenesis in peripheral arterial disease: can biotechnology produce an effective collateral circulation?**. *Eur J Vasc Endovasc Surg* (2004) **28** 9-23. DOI: 10.1016/j.ejvs.2004.03.021
43. Kastora SL, Eley J, Gannon M, Melvin R, Munro E, Makris SA. **What went wrong with VEGF-A in peripheral arterial disease? A systematic review and biological insights on future therapeutics**. *J Vasc Res* (2022) **59** 381-93. DOI: 10.1159/000527079
44. Lepper C, Conway SJ, Fan CM. **Adult satellite cells and embryonic muscle progenitors have distinct genetic requirements**. *Nature* (2009) **460** 627-31. DOI: 10.1038/nature08209
45. Lepper C, Partridge TA, Fan CM. **An absolute requirement for Pax7-positive satellite cells in acute injury-induced skeletal muscle regeneration**. *Development* (2011) **138** 3639-46. DOI: 10.1242/dev.067595
46. Scholz D, Ziegelhoeffer T, Helisch A, Wagner S, Friedrich C, Podzuweit T. **Contribution of arteriogenesis and angiogenesis to postocclusive hindlimb perfusion in mice**. *J Mol Cell Cardiol* (2002) **34** 775-87. DOI: 10.1006/jmcc.2002.2013
47. Yang Y, Tang G, Yan J, Park B, Hoffman A, Tie G. **Cellular and molecular mechanism regulating blood flow recovery in acute versus gradual femoral artery occlusion are distinct in the mouse**. *J Vasc Surg* (2008) **48** 1546-58. DOI: 10.1016/j.jvs.2008.07.063
48. Yucel N, Chang AC, Day JW, Rosenthal N, Blau HM. **Humanizing the mdx mouse model of DMD: the long and the short of it**. *NPJ Regen Med* (2018) **3** 4. DOI: 10.1038/s41536-018-0045-4
49. Sarjeant K, Stephens JM. **Adipogenesis**. *Cold Spring Harb Perspect Biol* (2012) **4** a008417. DOI: 10.1101/cshperspect.a008417
50. Biferali B, Proietti D, Mozzetta C, Madaro L. **Fibro–Adipogenic progenitors cross-talk in skeletal muscle: the social network**. *Front Physiol* (2019) **10** 1074. DOI: 10.3389/fphys.2019.01074
51. Cappellari O, Benedetti S, Innocenzi A, Tedesco FS, Moreno-Fortuny A, Ugarte G. **Dll4 and PDGF-BB convert committed skeletal myoblasts to pericytes without erasing their myogenic memory**. *Dev Cell* (2013) **24** 586-99. DOI: 10.1016/j.devcel.2013.01.022
|
---
title: Identification of immune biomarkers associated with basement membranes in idiopathic
pulmonary fibrosis and their pan-cancer analysis
authors:
- Chenkun Fu
- Lina Chen
- Yiju Cheng
- Wenting Yang
- Honglan Zhu
- Xiao Wu
- Banruo Cai
journal: Frontiers in Genetics
year: 2023
pmcid: PMC10017543
doi: 10.3389/fgene.2023.1114601
license: CC BY 4.0
---
# Identification of immune biomarkers associated with basement membranes in idiopathic pulmonary fibrosis and their pan-cancer analysis
## Abstract
Idiopathic pulmonary fibrosis (IPF) is a chronic progressive interstitial lung disease of unknown etiology, characterized by diffuse alveolitis and alveolar structural damage. Due to the short median survival time and poor prognosis of IPF, it is particularly urgent to find new IPF biomarkers. Previous studies have shown that basement membranes (BMs) are associated with the development of IPF and tumor metastasis. However, there is still a lack of research on BMs-related genes in IPF. Therefore, we investigated the expression level of BMs genes in IPF and control groups, and explored their potential as biomarkers for IPF diagnosis. In this study, the GSE32537 and GSE53845 datasets were used as training sets, while the GSE24206, GSE10667 and GSE101286 datasets were used as validation sets. In the training set, seven immune biomarkers related to BMs were selected by differential expression analysis, machine learning algorithm (LASSO, SVM-RFE, Randomforest) and ssGSEA analysis. Further ROC analysis confirmed that seven BMs-related genes played an important role in IPF. Finally, four immune-related *Hub* genes (COL14A1, COL17A1, ITGA10, MMP7) were screened out. Then we created a logistic regression model of immune-related hub genes (IHGs) and used a nomogram to predict IPF risk. The nomogram model was evaluated to have good reliability and validity, and ROC analysis showed that the AUC value of IHGs was 0.941 in the training set and 0.917 in the validation set. Pan-cancer analysis showed that IHGs were associated with prognosis, immune cell infiltration, TME, and drug sensitivity in 33 cancers, suggesting that IHGs may be potential targets for intervention in human diseases including IPF and cancer.
## 1 Introduction
Idiopathic pulmonary fibrosis (IPF) is a chronic progressive interstitial lung disease of unknown etiology (Richeldi et al., 2017). Its pathological features are diffuse alveolitis and alveolar structural damage, eventually forming honeycomb lung (Wolters et al., 2014). The clinical symptoms of IPF include dry cough, fatigue, and progressive exertional dyspnea. IPF is a rare disease, but its incidence is increasing and is more common in elderly male patients (Maher et al., 2021; Kondoh et al., 2022). Although disease progression varies greatly among individuals, the median survival after diagnosis is less than 3–5 years. Currently, treatment options for patients with IPF remain limited. Anti-fibrotic drugs (Pirfenidone and Nintedanib) have been approved to treat IPF, but they only slow the decline in lung function in IPF patients and do not improve survival (King et al., 2014; Richeldi et al., 2014). These drugs can also cause gastrointestinal adverse reactions, which limits their widespread use to some extent (Bargagli et al., 2019). At present, lung transplantation remains the only effective treatment. Unfortunately, not all patients are suitable for transplant, and complications following the transplant place a huge burden on patients.
Basement membranes (BMs) are a specialized form of extracellular matrix found in various organs of the human body, providing structural support for epithelium, endothelium, muscle, adipocytes, schwann cells, and axons (Mak and Mei, 2017). BMs are mainly composed of laminin and type IV collagen, which are linked together by nicolysaccharide and heparan sulfate proteoglycans to form different BMs in various tissues (Kruegel and Miosge, 2010). BMs can direct cell polarity, differentiation, migration and survival. For example, BMs can control epithelial growth and differentiation during embryonic development (Kyprianou et al., 2020). In cancer, the breakdown of the basement membrane promotes metastasis (Banerjee et al., 2022). Variations in the BMs gene are closely associated with many human diseases (Horton and Barrett, 2021; Wilson, 2022). BMs can mark the pathway of cell migration and epithelialization during tissue repair (Vracko, 1974; Rousselle et al., 2019). Changes in BMs homeostasis can lead to abnormal ECM aggregation and fibrosis (Wilson, 2020). Alveolar BMs promote gas exchange between alveoli and capillaries and regulate the function of cytokines and growth factors (West and Mathieu-Costello, 1999). The integrity of BMs maintains the normal lung structure and is critical for restoring alveolar epithelial homeostasis after lung injury (Strieter and Mehrad, 2009). However, a loss of alveolar and capillary BMs integrity was observed in IPF, suggesting that BMs are involved in IPF genesis (Chen et al., 2016).
Increasing evidence supports the important role of immune response in IPF. On the one hand, damage to lung epithelial cells leads to the production of pro-inflammatory cytokines such as IL-1 and IL-6 in M1 alveolar macrophages (Huang et al., 2018). These cytokines play an important role in host resistance to pathogen invasion. Inhibition of TNF-α secretion can alleviate bleomycin-induced pulmonary fibrosis and collagen deposition (Matsuhira et al., 2020). On the other hand, under chronic inflammatory conditions, Th2 cells secrete cytokines to gradually transform pro-inflammatory M1 macrophages into pro-fibrotic M2 macrophages (Shapouri-Moghaddam et al., 2018). M2 macrophages secrete multiple chemokines and activate Wnt/β-catenin signaling pathways leading to fibroblast activation, myofibroblast differentiation and extracellular matrix remodeling (Wynn and Vannella, 2016). Inhibition of Wnt/β-catenin signaling attenuates M2 macrophage-induced myofibroblast differentiation and bleomycin-induced pulmonary fibrosis (Hou et al., 2018). Another study found that vaccination inhibited M2 macrophage production and fibrocyte recruitment in bleomycin-induced pulmonary fibrosis (Collins et al., 2012). Moreover, the importance of immune responses in IPF has been confirmed by genetic studies, such as DEP domain containing MTOR interacting protein (DEPTOR) increase the risk or susceptibility of IPF(Allen et al., 2020).
In this article, we aim to explore the immune markers associated with BMs in IPF and construct a nomogram model to predict the risk of IPF in patients. Increased evidence suggests that IPF is closely linked to cancer (Ballester et al., 2019; Tzouvelekis et al., 2019). However, little is known about the relationship between IPF and cancer. Therefore, we conducted an in-depth analysis of the role of BMs-related immune biomarkers in pan-cancer to explore the common pathogenesis of IPF and cancer, and to find potential therapeutic targets for patients with IPF and cancer. In Figure 1, you can see the workflow chart.
**FIGURE 1:** *The workflow chart of our study.*
## 2.1 Data download and processing
We obtained 222 BMs related genes from previous studies (Jayadev et al., 2022). We obtained datasets numbered GSE32537, GSE53845, GSE24206, GSE10667 and GSE101286 from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). The GSE32537, GSE53845 datasets were used as training set. Datasets GSE24206, GSE10667 and GSE101286 were used as validation sets for independent external validation. The main features of the five datasets were shown in Table 1. Datasets were merged after identity document transformation. The original data was normalized through R software “sva” package, followed by classification of the datasets into two categories: IPF and control groups. Finally, we evaluated the quality of the dataset using principal component analysis and plotted the PCA plot via R software “ggplot2” package.
**TABLE 1**
| Dataset | Platform | IPF | Normal | Publication years | Used for |
| --- | --- | --- | --- | --- | --- |
| GSE32537 | GPL6244 | 167 | 50 | 2013 | DEGs analysis |
| GSE53845 | GPL6480 | 40 | 8 | 2014 | DEGs analysis |
| GSE24206 | GPL570 | 17 | 6 | 2011 | Model validation |
| GSE10667 | GPL4133 | 31 | 15 | 2009 | Model validation |
| GSE101286 | GPL6947 | 12 | 3 | 2017 | Model validation |
## 2.2 Identification of differentially expressed genes in BMs
We screened differentially expressed genes (DEGs) through R software “limma” package, filter condition for $p \leq 0.05$ and | log2FC | > 1. A volcano map was drawn through the “ggplot2” package of R software to show DEGs. The “VennDiagram” package in R was used to obtain the intersection of BMs-related genes and DEGs, and the differentially expressed genes of basement membranes (BMDEGs) were obtained.
## 2.3 Protein–protein interaction (PPI) and enrichment analysis of BMDEGs
We visualized the PPI network of BMDEGs via the STRING database and performed GO, KEGG, and DO enrichment analyses of BMDEGs via the “ClusterProfiler” and “DOSE” packages of R software.
## 2.4 Three machine learning algorithms for screening disease candidate genes
We used LASSO, SVM-RFE and RandomForest machine learning algorithms to screen disease candidate genes. LASSO analysis was performed using 10-fold cross-validated penalty parameters via the “glmnet” package of R software. The minimal bimomial deviation was used to determine the optimal penalty parameter lambda. SVM-RFE algorithm detects the points with the minimum cross-validation error through the “e1071”, “kernlab” and “caret” packages in R software to screen disease candidate genes. The RandomForest algorithm uses the “randomforest” package of R software to screen disease candidate genes. The Venn diagram visualizes disease candidate genes obtained from the results of three machine learning algorithms.
## 2.5 Validation of disease candidate genes
To understand the specificity and sensitivity of disease candidate genes for IPF diagnosis, we draw the receiver operating characteristic (ROC) curve through the “pROC” package of R software. Results were presented in the form of area under the curve (AUC). If the AUC of candidate genes was greater than 0.6, we believed that it had diagnostic significance for IPF. Disease candidate gene expression in the IPF and control groups was shown in the box plot.
## 2.6 Analysis of immune infiltration
We performed correlation analysis of immunity through the “corrplot” package of R software and plotted the correlation heatmap. R software “ggpubr” and “reshape2” packages were used to analyze the differential expression of immune cells and immune functions in IPF and control groups. Spearman correlation analysis was conducted through the “psych” and “ggcorrplot” packages of the R software to analyze the correlation between disease candidate genes and immunity. The screening criteria for IPF immune-related hub genes (IHGs) were more than $\frac{1}{2}$ immune infiltration and the correlation coefficient was greater than 0.2.
## 2.7 Establishment and validation of IHGs risk model
We constructed the nomogram of IHGs and plotted the calibration curve to determine the reliability of the nomogram through the “rms” package of R software. ROC curves were plotted to assess the accuracy of IHGs in diagnosing IPF via the “ROCR” package of R software.
## 2.8 Differential analysis of IHGs in human cancer
We downloaded tumor transcriptome data, clinical data, immune subtype data, mutation data and stemness score (RNAss and DNAss) from the UCSC Xena database (https://xenabrowser.net/). The expression levels of IHGs in 33 tumor and adjacent samples were extracted using the “limma” package of R software. Then, we retained tumor types with more than 5 paracancer tissues, and analyzed the expression of IHGs in tumors and paracancer tissues through the “ggpubr” package of R software.
## 2.9 Survival analysis of IHGs in pan-cancer
COX regression analysis was conducted through the “survival” package of R software to determine whether IHGs expression was correlated with the survival time and survival status of cancer patients, and the results were presented in forest plots. Survival analysis was performed through the “survival” and “surminer” packages of R software to determine whether IHGs expression was linked with the prognosis of tumor patients.
## 2.10 Mutation analysis of IHGs in pan-cancer
Tumor mutation burden (TMB) and microsatellite instability (MSI) is specific indicators for predicting immunotherapy in cancer patients. Therefore, we used the “fmsb” package of R software to plot radar maps of TMB and MSI to determine the correlation between IHGs expression and 33 types of tumors.
## 2.11 Tumor microenvironment (TME) and tumor stemness analysis
We calculated the immune score, stromal score, and estimated score for each sample in the tumor using the “estimate” package of R software. The correlation between IHGs expression and purity in 33 tumors was analyzed by Spearman correlation analysis. The relationship between IHGs expression and tumor stemness was determined by the “corrplot” package of R software.
## 2.12 Immune analysis of IHGs in pan-cancer
Immune subtypes including C1(Wound Healing), C2(IFN γ Dominant), C3(Inflammatory), C4 (lymphocyte Depleted), C5 (M2 macrophages Dominant), C6 (TGF-β Dominant) subtypes. Previous studies have shown that among the six immune subtypes, C4 and C6 are associated with lower survival rates, while C3 and C5 are the opposite (Tamborero et al., 2018). The relationship between IHGs expression and immune subtypes was analyzed by the “limma”, “ggplot2” and “reshape2” packages of R software. The correlation between IHGs expression and immune checkpoints was completed by Pearson correlation analysis. Finally, we investigated the correlation between IHGs expression in tumors and 21 types of immune cells using TIMER2.0 database (http://timer.cistrome.org/).
## 2.13 Enrichment analysis
We used the GeneMANIA database (http://genemania.org/) to predict and visualize genes that function similar to IHGs. The Metascape website (https://metascape.org/) was used to analyze the functions in which genes may be involved.
## 2.14 Drug sensitivity analysis of IHGs in pan-cancer
We downloaded drug sensitivity data for 60 human cancers from the CellMiner websites (https://discover.nci.nih.gov/cellminer/) and screened 263 FDA-approved or clinical trial drugs for this study. The relationship between IHGs expression and drugs was analyzed by Pearson correlation analysis.
## 2.15 Statistical analysis
Statistical tests were performed using R software (version 4.1.3). For all statistical analyses, $p \leq 0.05$ was considered statistically significant (“***”, “**”, “*”, “ns” are “$p \leq 0.001$” “$p \leq 0.01$” “$p \leq 0.05$” “no significance”). Relevant scripts and supported data can be seen on the Github website (https://github.com/fuchenkun/Basement-membranes).
## 3.1 Screening for BMDEGs of IPF
We combined GSE32537, GSE53845 datasets and corrected the batch effects for subsequent analyses (Figures 2A, B). 253 DEGs were screened, of which 158 genes were over-expressed and 95 genes were under-expressed. The results of the DEGs were presented as a volcano map (Figure 2C). A Venn diagram was also created, which showed that 13 BMDEGs, of which 12 were upregulated and 1 downregulated (Figure 2D).
**FIGURE 2:** *Differential expression analysis and enrichment analysis. (A) PCA plot shows training set before batch effect. (B) PCA plot shows the training set after batch effect. (C) A volcanic map of DEGs. (D) The Venn diagram of BMs related genes and DEGs. (E) Protein interaction network of BMDEGs. GO function(F), KEGG pathway(G) and Disease enrichment analysis(H) of BMDEGs.*
## 3.2 Protein interaction network and enrichment analyses of BMDEGs
We constructed a protein interaction network for BMDEGs (Figure 2E). Then we performed enrichment analyses to better understand the functions, pathways, and diseases that BMDGEs might be involved in. As shown in Figure 2F, our results indicated that biological processes were mainly related to the structure of extracellular matrix, collagen catabolic processes, and responses to mechanical stimulus. The cellular components mainly involved extracellular matrix, basement membrane and endoplasmic reticulum. Extracellular matrix structural constituent, heparin binding, glycosaminoglycan binding, sulfur compound binding and extracellular matrix structural constituent conferring tensile strength were significantly enriched in molecular functions. The KEGG analysis revealed that BMDEGs tended to be enriched in the following terms: extracellular matrix organization, extracellular structure organization, external encapsulating structure organization, cell−substrate adhesion and response to mechanical stimulus (Figure 2G). Moreover, DO analysis found that BMDEGs were specifically enriched in IPF and were also associated with endocrine disorders, reproductive system diseases, and cancer (Figure 2H).
## 3.3 Screening and validation for IPF diagnostic markers
We further used machine learning algorithms to screen disease candidate genes from BMDEGs. LASSO regression analysis selected 11 genes (Figure 3A), the SVM-RFE algorithm identified eight genes (Figure 3B), and RandomForest screened 11 genes (Figure 3C). Finally, through the gene intersection obtained by the three algorithms, seven disease candidate genes (COL14A1, COL17A1, HMCN1, ITGA10, MMP7, OGN and ROBO2) were identified (Figure 3D). Then we analyzed the expression of seven disease candidate genes in the training group and validated them using external datasets. Validation group dataset eliminates batch effect for subsequent analysis (Supplementary Figure S1). As presented in Figure 3E, the boxplot showed that seven disease candidate genes were significantly upregulated in IPF groups and 1 candidate gene was significantly downregulated in IPF groups. We saw the same results in the validation dataset, but ROBO2 was not statistically significant (Figure 3F). We further performed ROC analysis to examine the diagnostic efficacy of seven disease candidate genes for IPF. The results suggest that the seven disease candidate genes have diagnostic value in distinguishing IPF groups from control groups: COL14A1 (AUC = 0.964), COL17A1 (AUC = 0.915), HMCN1 (AUC = 0.961), ITGA10 (AUC = 0.946), MMP7 (AUC = 0.937), OGN (AUC = 0.894) and ROBO2(AUC = 0.856) (Supplementary Table S1). Similarly, we evaluated the diagnostic efficacy of seven disease candidate genes for IPF in the validation group dataset using ROC analysis. The results indicated that AUCs of the disease candidate genes were COL14A1 (AUC = 0.881), COL17A1 (AUC = 0.949), HMCN1(AUC = 0.813), ITGA10 (AUC = 0.707), MMP7 (AUC = 0.910), OGN (AUC = 0.719) and ROBO2 (AUC = 0.600) (Supplementary Table S2). In conclusion, the AUC values of COL14A1, COL17A1, HMCN1, ITGA10, OGN and MMP7 in the training dataset and validation dataset were all greater than 0.7. These results suggest that the candidate genes are closely related to IPF and have the potential to be used as biomarkers of IPF and indicators to evaluate the efficacy of patients.
**FIGURE 3:** *Machine learning algorithms identify disease candidate genes. (A) LASSO model for screening candidate genes for disease. (B) SVM-RFE algorithm for screening disease candidate genes. (C) Random forest model to screen candidate genes for disease. (D) The Venn diagram shows the common disease candidate genes of LASSO, RandomForest and SVM-RFE algorithms. (E) Boxplot representing expression of disease candidate genes in training set. (F) Boxplot representing expression of disease candidate genes in the validation set.*
## 3.4 Immune infiltration analysis and IHGs screening
We used the ssGSEA algorithm to evaluate immune infiltration in 265 samples (Supplementary Figure S2). In the correlation analysis of immune cells, the positive correlation between Tfh cells and B-cell was the strongest, and the correlation coefficient was 0.8. The negative correlation between Tfh cells and NK cells was the strongest, and the correlation coefficient was −0.3 (Figure 4A). Interestingly, we did not observe a negative correlation for immune function, whereas there was a positive correlation ($r = 0.8$) between T-cell co inhibition and APC co inhibition (Figure 4B). For immune cells, the expression of aDCs, B-cell, DCs, iDCs, Mast cells, T helper cells, Tfh and Th1 cells were increased in IPF, while the expression of neutrophils, NK cells, pDCs, Th2 cells and Treg cells were decreased (Figure 4C). For immune function, Check point, HLA, Inflammation promoting, T-cell co stimulation and Parainflamation were over-expressed in IPF, while T-cell co inhibition and APC co inhibition were underexpressed in IPF (Figure 4D). Finally, four IHGs were screened by correlation analysis, including COL14A1, COL17A1, ITGA10 and MMP7 (Figure 4E). These results suggest that the activation of multiple immune cells and the coordination of immune functions are important in the pathogenesis of IPF.
**FIGURE 4:** *Immune infiltration analysis and establishment of IPF risk model. (A) Correlation heatmap of immune cells. (B) Immune function related heatmap. (C) Difference of immune cell expression between IPF and control groups. (D) Difference of immune function expression between IPF and control group. (E) Heatmap of correlation between IHGs expression and immune infiltration. (F) IHGs predicts the occurrence of IPF in training set. (G) IHGs predicts the occurrence of IPF the validation set.*
## 3.5 Construction and validation of IPF risk model
We created a logistic regression model of IHGs and used a nomogram to predict IPF risk (Figure 4F). The calibration curve used to evaluate the risk nomogram of IPF patients showed good consistency in this study. The results showed that the AUC of the training data set was 0.941, indicating that our model had good predictive ability. In order to further verify the prediction effect of our model, we used independent external validation dataset to verify (Figure 4G). The results showed that the calibration curve also showed good consistency in the training dataset. The AUC value in the validation dataset was 0.917, which also shows that our model had good predictive ability. Moreover, The C index also shows that our model had good predictive power. The C index was 0.941 ($95\%$ CI: 0.917–0.965) in the training dataset and 0.917 ($95\%$ CI: 0.840–0.994) in the validation dataset.
## 3.6 The expression level of IHGs in pan-cancer
To determine whether there are differences in the expression of IHGs in tumors, mRNA expression levels of IHGs in normal and tumor tissues were analyzed. As shown in Figure 5A, the expression of MMP7 in IHGs was relatively high, while the expression of ITGA10 was the lowest. The expression of IHGs in different cancer types is also quite different (Figure 5B). Overall, COL14A1, COL17A1, and ITGA10 tended to be downregulated in most tumors, while MMP7 tended to be upregulated in most tumors. Correlation analysis showed that ITGA10 was weakly positively correlated with COL14A1, and weakly negatively correlated with MMP7 and COL17A1 (Figure 5C). Although there is a correlation between IHGs, the correlation coefficient value is between −0.15 and 0.23, which proves that the correlation is weak or negligible. COL14A1 was highly expressed in 1 tumor and lowly expressed in 15 tumors (Figure 5D). COL17A1 was significantly upregulated in seven tumors, while significantly downregulated in eight tumors (Figure 5E). ITGA10 expression was increased in six tumors and decreased in eight tumors (Figure 5F). MMP7 expression was higher in 12 tumors and lower in four tumors (Figure 5G).
**FIGURE 5:** *Expression of IHGs in Human Cancer. (A) Boxplots of IHGs expression levels in cancer. (B) Heatmap of IHGs expression levels in different cancer types and adjacent tissues. (C) Positive (blue) and negative (red) correlations between IHGs. Expression of COL14A1
(D), COL17A1
(E), ITGA10
(F) and MMP7
(G) in different tumor types and adjacent tissues. (H) Forest plot shows the relationship between IHGs expression and OS in 33 tumors.*
## 3.7 Pan-cancer survival analysis of IHGs
Based on the results of differential analysis, we used forest maps and survival curves to further understand whether IHGs expression was linked with tumor prognosis (Figure 5H). Cox regression analysis revealed that increased COL14A1 expression was a negative factor affecting KIRP, LGG, BLCA, STAD and OV, while a positive factor affecting ACC. Increased expression of COL14A1 was related to shorter overall survival (OS) in BLCA, KIRP, LGG and UVM, whereas decreased expression of COL14A1 was related to shorter OS in ACC and LAML (Figures 6A–F). As shown as Figures 6G–I, Cox regression analysis revealed that COL17A1 over-expression was an adverse factor for PAAD and SKCM, but a favourable factor for BRCA. Overexpression of COL17A1 was linked with poorer OS in PAAD and SKCM, whereas increased COL17A1 expression predicted favorable OS in LGG. As seen as Figures 6J–R, Cox regression analysis found that the increased expression of ITGA10 was a negative factor for LGG, SARC and KICH, and a positive factor for SKCM and BRCA. High ITGA10 expression was related to shorter OS for KIRP, LGG, MESO, OV, SARC, STAD, THCA, and longer OS for BRCA and SKCM. According to Figures 6S–Y, Cox regression analysis revealed that high expression of MMP7 was an adverse factor for PAAD, ACC, LAML, KIRC, LIHC and SKCM, while high expression of MMP7 was a beneficial factor for BRCA. Survival analysis showed that patients with increased MMP7 expression in ACC, KIRC, LAML, LGG, LIHC, MESO, PAAD had shorter OS.
**FIGURE 6:** *Relationship between IHGs expression and prognosis of different tumors. OS survival curves for COL14A1 in six tumors: (A) ACC, (B) BLCA, (C) KIRP, (D) LAML, (E) LGG, (F) UVM. OS survival curves for COL17A1 in 3 tumors:(G) LGG, (H) PAAD, (I) SKCM. OS survival curves for ITGA10 in 9 tumors: (J) BRCA, (K) KIRP, (L) LGG, (M) MESO, (N) OV, (O) SARC, (P) SKCM, (Q) STAD,(R) THCA. OS survival curves for MMP7 in 7 tumors: (S) ACC, (T) KIRC, (U) LAML, (V) LGG, (W) LIHC, (X) MESO, (Y) PAAD.*
## 3.8 Correlation of IHGs expression with TME and tumor stemness
TME is closely related to tumorigenesis and tumor cells escaping the immune system. The therapeutic effect and clinical prognosis of tumor are also correlated with TME. Therefore, the correlation between IHGs expression and tumor purity was assessed to understand whether IHGs are involved in tumor immunity. Overall, IHGs expression was positively related to stromal scores, with COL14A1 having the strongest correlation with stromal score (Figure 7A). In terms of immune score, IHGs expression was positively correlated with CHOL, KICH, LIHC, and PCPG, etc., ( Figure 7B). This suggests that IHGs have similar effects in the TME. Furthermore, we also evaluated the correlation between IHGs expression and tumor stemness score to understand the effect of IHGs expression on tumor differentiation. We found that IHGs were negatively related to RNAss in most tumors (Figure 7C). In contrast, it was positively related to DNAss in KIRC, KIRP, THYM and UVM (Figure 7D). These results suggest that the higher the expression of IHGs, the weaker the stemness score and the higher the degree of tumor differentiation.
**FIGURE 7:** *Correlation of IHGs expression with TME, stemness score, immune subtypes and immune checkpoints. (A,B) IHGs expression was related to immune and stromal scores in 33 tumors. (C,D) IHGs expression was related to stemness score in 33 cancers. (E) IHGs expression was associated with immune subtypes. The correlation between COL14A1
(F), COL17A1
(G), ITGA10
(H) and MMP7
(I) and immune checkpoints in 33 cancers.*
## 3.9 Correlation of IHGs expression with immune subtype and immune checkpoints
Our results showed that COL14A1, COL17A1, and MMP7 were over-expressed in the C1, C2, and C6 subtypes, and ITGA10 was over-expressed in the C5 and C6 subtypes (Figure 7E). The high expression of COL14A1, COL17A1 and MMP7 was closely related to C1, C2 and C6 subtypes, indicating that these three genes may have carcinogenic effects. Moreover, we observed a significant correlation between IHGs expression and immune checkpoint genes in different tumor types. Specifically, the results revealed that COL14A1 was positively related to immune checkpoints in most tumors except MESO, OV, SARC, TGCT, THCA, THYM, UCEC, UCS, and UVM (Figure 7F). COL17A1 was significantly related to immune checkpoints in most tumors except ACC, CHOL, DLBC, UCS, and UVM (Figure 7G). ITGA10 was positively related to more than 30 immune checkpoint genes in COAD, ESCA, KICH, HNSC and LUSC (Figure 7H). In addition to CESC, CHOL, COAD, SARC, SKCM, and UVM, MMP7 was closely related to immune checkpoint genes (Figure 7I). These results indicate that IHGs expression is closely related to immune checkpoint genes, suggesting that IHGs may play a vital role in mediating tumor immune patterns.
## 3.10 Immune cell infiltration analysis of IHGs in pan-cancer
We analyzed the correlation between tumor infiltrating immune cells and IHGs expression by TIMER2.0 to understand whether IHGs participated in tumor immune infiltration. We found that COL14A1 was positively related to CAF, DCs, Endo, HSC, Macrophage, Monocyte and Tregs (Figure 8A). COL17A1 was negatively related to most immune cells in HNSC, LUSC and ESCA (Figure 8B). ITGA10 was positively related to CAF, Endo, HSC, neutrophils and Tregs in most tumors (Figure 8C). MMP7 was positively related to CAF, DCs, Macrophage and Monocyte in most tumors (Figure 8D). Compared with COL17A1, ITGA10 and MMP7, COL14A1 had a higher correlation coefficient with infiltrating cells. These results suggest a potential mechanism by which IHGs have different prognostic value in different tumors.
**FIGURE 8:** *Correlation of IHGs expression with immune cell infiltration, TMB and MSI. The correlation between COL14A1
(A), COL17A1
(B), ITGA10
(C) and MMP7
(D) and 21 immune cells in 33 cancers. The correlation between COL14A1
(E), COL17A1
(F), ITGA10
(G) and MMP7
(H) and TMB in 33 cancers. The correlation between COL14A1
(I), COL17A1
(J), ITGA10
(K) and MMP7
(L) and MSI in 33 cancers.*
## 3.11 TMB and MSI analysis of IHGs in human cancers
The more frequent the mutation of tumor cells, the more new antigens produced, making them more susceptible to immunotherapy (Gryfe et al., 2000; Samstein et al., 2019). Herein, we investigated the correlation between IHGs expression with TMB, MSI. We found that COL14A1/ITGA10/MMP7 was positively related to TMB in LGG, and negatively related to TMB in STAD, LUSC, LUAD, LIHC, HNSC and BLCA (Figures 8E–H). As for MSI, COL14A1/ITGA10/MMP7 expression was negatively correlated with STAD (Figures 8I–L). Although these correlations are important to guide immunotherapy in cancer patients, the correlation coefficients between IHGs expression and TMB and MSI did not exceed 0.6 in all tumor types, suggesting that IHGs are unlikely to affect tumorigenesis by participating in gene modification processes and are insufficient to independently predict patient response to immunotherapy.
## 3.12 Functional enrichiment analysis of IHGs in pan-cancer
We utilized GeneMANIA to screen genes associated with IHGs for comprehensive functional and pathway analysis of IHGs. Finally, we constructed an IHGs-centric PPI network consisting of 24 genes (Figure 9A). Metascape significantly enriched items include Extracellular matrix organization, ECM-receptor interaction, Degradation of the extracellular matrix, Integrin cell surface intercations, epidermis development, PID AJDISS 2PATHWAY, ECM proteoglycans, response to wounding, Proteoglycans in cancer, Wnt signaling pathway and pluripotency, Type I hemidesmosome assembly, Immunoregulatory interactions between a Lymphoid, appendage development and response to growth factor (Figure 9B).
**FIGURE 9:** *Functional enrichment analysis and drug sensitivity analysis of IHGs. (A) Construction of a PPI network with 24 genes centered on IHGs. (B) Enrichment analysis using Matescape. (C) Correlation analysis between IHGs and drug sensitivity of anticancer drugs in CellMiner.*
## 3.13 Drug response analysis of IHGs
We further explored whether the expression of IHGs has guiding significance for clinical medication. Among 263 drugs (FDA approved or clinical trials), we found that the sensitivity of 55 drugs was significantly correlated with IHGs expression levels. As presented in Figure 9C, we show the top 20 drugs most significantly associated with IHGs. The upregulated expression of IHGs was associated with increased sensitivity to Afatinib, Pentostatin, Gefitinib, kahalide f, Lapatinib, Vandetanib, Ibrutinib, Erlotinib, and okadaic acid, and decreased sensitivity to Pyrazoloacridine, AT−13387, Pazopanib, Vincristine, Lapachone, Etoposide, Lomustine, Paclitaxel, Epothilone B, and Pipamperone. Overall, we found that IHGs expression was associated with treatment response, suggesting that IHGs may be involved in tumor drug resistance.
## 4 Discussion
In this study, we screened 13 BMDEGs using bioinformatics methods, among which 12 genes were over-expressed and 1 gene was lowly expressed. Subsequent GO enrichment analysis showed that all BMDEGs were mainly related to extracellular matrix tissue and collagen catabolic process, while KEGG enrichment analysis showed a certain correlation with Notch signaling pathway. Based on three machine learning algorithms, we screened seven disease candidate genes. Using external datasets, it was confirmed that COL14A1, COL17A1, HMCN1, MMP7, OGN and ROBO2 were highly expressed in IPF, while ITGA10 was lowly expressed in IPF. ROC curve analysis further confirmed that all disease candidate genes have diagnostic value in IPF, suggesting that they may have potential application prospects in the treatment of IPF. The findings suggest their potential usage for diagnostic value in IPF since they are highly expressed in IPF. The findings will eventually lead to future studies about the potential role of their involvement in IPF, and if their roles are confirmed, the potential application prospects in the treatment of IPF are likely to evaluated. Finally, four IHGs (COL14A1, COL17A1, ITGA10, MMP7) were screened out. We then construct a logistic regression model of IHGs and use a nomogram to predict IPF risk. The AUC of the training set was 0.941, and that of the verification set was 0.917. It shows that our model has good predictive ability, and these four genes are potential biomarkers of IPF. Our study provides a theoretical basis for studying the role of BMs-associated immune biomarkers in the pathogenesis of IPF, and provides promising research suggestions for subsequent studies.
In pan-cancer, the expression of IHGs was significantly different across cancers compared to comparable normal tissues. COL14A1 was significantly downregulated in almost all cancer types. COL17A1, ITGA10 and MMP7 showed high intertumoral heterogeneity between tumor tissues and adjacent tissues. These results suggest that IHGs are a potential cancer biomarker.
Our study found that COL14A1 was an adverse factor for KIRP, LGG, BLCA, STAD and OV, and a favourable factor for ACC. However, there is a lack of relevant studies to support the effect of COL14A1 on the prognosis of these cancers, and more studies are needed to prove this. COL14A1 encodes the alpha chain of type XIV collagen, which is linked with mature collagen fibers (Schuppan et al., 1990). It has been reported that COL14A1 can affect arterial remodeling and participate in the occurrence of cardiovascular diseases (Weis-Müller et al., 2006; Guay et al., 2015). To our knowledge, there is a lack of literature on the role of COL14A1 in IPF. When compared cancer with or without metastasis, it seems that further decrease of COL14A1 has better outcome, but, this has to be tested in a larger scale to validate (Goto et al., 2015; Jiang et al., 2022). COL14A1 methylation is an unfavorable prognostic factor for renal cell carcinoma, and low COL14A1 expression seems to promote tumorigenesis of renal cell carcinoma (Morris et al., 2010). It seems that COL14A1 is increased in IPF, while it is decreased in cancer. The differential expression of COL14A1 may indicate the critical signaling that differentiates IPF - a disease with non-stopping fibroblast growth, from cancer-a disease with non-stopping malignant cell growth. Therefore, our studies suggest the critical and differentiating signaling may involve COL14A1, its function or signaling, and hopefully, that can be investigated by future studies.
COL17A1 is one of the triple-helix collagen genes encoding collagen XVII, a type II transmembrane protein found in basal epithelial cells that can affect cell growth and migration (Natsuga et al., 2019; Kozawa et al., 2021). COL17A1 also lacks relevant research in the field of IPF. Currently, research on COL17A1 has focused on cancer and skin diseases (Nishie, 2020). Our study found that COL17A1 over-expression was related to poor prognosis of SKCM and PAAD. Studies have shown that COL17A1 is over-expressed in a variety of cancers (Thangavelu et al., 2016; Huang et al., 2022). In contrast, another study found that upregulation of COL17A1 expression was related to better prognosis in breast cancer (Yodsurang et al., 2017). COL17A1 inhibits cancer cell migration and invasion by inactivating AKT/mTOR pathway, and its over-expression is linked with longer survival in patients with invasive breast cancer (Lothong et al., 2021).
ITGA10 is a type II collagen-binding integrin first isolated from chondrocytes (Camper et al., 1998). ITGA10 has the highest content in cartilage tissue and plays a crucial part in the formation of growth plates during bone development (Bengtsson et al., 2005). From the current overall research situation, the research on ITGA10 mainly focuses on cancer. Our study suggested that increased ITGA10 expression was linked with poor prognosis of SARC and longer survival of BRCA and SKCM. Previous studies have reported that ITGA10 promotes drug resistance and proliferation of osteosarcoma cells by the activation of PI3K/AKT signaling pathway (Li et al., 2021). Similarly, ITGA10 promotes myxfibrosarcoma survival and metastasis by activating TRIO/RAC and RICTOR signaling pathways, and antitumor effects were observed in mouse xenografts after ITGA10 inhibition (Okada et al., 2016). In addition, dysregulation and carcinogenic effects of ITGA10 had been observed in lung cancer, prostate cancer, and thyroid cancer (Mertens-Walker et al., 2015; Su et al., 2015; Saftencu et al., 2019). These findings suggest that ITGA10 acts as an oncogene. According to our knowledge, ITGA10 has not been reported on IPF, and more attention should be paid to IPF.
Different from COL14A1, COL17A1 and ITGA10, MMP7 had been supported by numerous literatures in the research field of IPF. MMP7 is the smallest member of the matrix metalloproteinase family. MMP7 plays a crucial part in the pathogenesis of fibrosis by degrading extracellular matrix proteins and activating multiple signaling molecules (Niu et al., 2019; Mahalanobish et al., 2020). MMP7 is a target gene of Wnt/β-catenin and highly expressed in proliferative epithelial cells of IPF (Zuo et al., 2002; Fujishima et al., 2010). Previous studies have identified MMP7 as a potential biomarker for IPF. For instance, MMP-7 was identified as a predictor of survival in a combined model incorporating clinical parameters and MUC5B genotype (Peljto et al., 2013; Biondini et al., 2021). Consistent with previous studies, we found increased MMP7 expression in IPF and negatively correlated with decreased FVC (Bauer et al., 2017). Some studies have shown that MMP7 combined with other biomarkers may improve the survival prediction of IPF patients (Song et al., 2013; Hamai et al., 2016). In addition, a phase II clinical study showed that MMP7 protein levels decreased in a dose-dependent manner after using JNK inhibitors, indicating that the use of MMP7 to track IPF progression has potential clinical benefits (van der Velden et al., 2016). However, a recent study found that there was no difference in the baseline concentration of MMP7 between IPF patients with or without disease progression, and short-term changes in its concentration could not reflect disease progression (Raghu et al., 2018; Khan et al., 2022). Another study indicated that MMP7 was over-expressed in patients with subclinical interstitial lung disease and under-expressed in patients with mature IPF compared to healthy controls (Drakopanagiotakis et al., 2018). It is suggested that MMP7 can be used as a potential marker for early detection of IPF. Besides, more and more evidences support MMP7 as an oncogene involved in tumor cell proliferation, migration and apoptosis (Sun et al., 2021; Van Doren, 2022). MMP7 is highly expressed in many cancers, and its expression is related to survival time and tumor stage (Lee et al., 2006; Liao et al., 2021). Knockdown of MMP7 gene can inhibit tumor proliferation, migration and reduce drug resistance (Sanli et al., 2013; Yuan et al., 2020). Therefore, MMP7 is expected to become a potential biomarker for evaluating tumor prognosis and a new target for tumor therapy.
GO enrichment analysis showed that BMDEGs were involved in the composition of basement membrane, extracellular matrix and endoplasmic reticulum cavity, and were related to extracellular matrix tissue, collagen metabolism and metallopeptidase activity. Alveolar epithelial damage and abnormal tissue repair are considered to be key factors in the development of IPF, which ultimately leads to the recruitment and activation of myofibroblasts to produce collagen-rich extracellular matrix. The deposition of extracellular matrix in IPF mainly involves matrix metalloproteinases (MMPs) and tissue inhibitors of metalloproteinases (TIMPs). More and more studies support the key role of MMPs in the pathogenesis of pulmonary fibrosis. Interestingly, some MMPs have pro-fibrotic effects, while others seem to play a protective role. In our study, we found that MMP7 was over-expressed in IPF patients, and previous studies have shown that it contributes to the progression and adverse consequences of IPF, while MMP19 seems to have a protective effect (Jara et al., 2015). Changes in the extracellular matrix of IPF can also affect the transcription of lung fibroblasts, resulting in abnormal translation of ECM proteins (Zolak and de Andrade, 2012). In addition, the degradation product of ECM (matrikines) also acts as a signal molecule and plays a central role in the fibrosis of IPF. Existing evidence suggests that ECM plays an important role in driving the circulation of pathogenic disease signals mediated by integrins, growth factors, matrikines, and MMPs (Hewlett et al., 2018). It is worth noting that the increased stiffness of ECM tissue is also a key driver of the fibrosis process. Compared with healthy lung scaffolds, collagen, proteoglycan and ECM glycoprotein in IPF scaffolds increased, but specific BMs proteins (such as laminin and collagen IV) decreased (Elowsson Rendin et al., 2019). New treatments for these ECM-driven processes are expected to bring benefits to IPF patients. Moreover, KEGG enrichment analysis showed that BMDEGs were related to Notch signaling pathway. It is reported that Notch signaling pathway plays a key role in the development, balance and regeneration of the respiratory system (Kiyokawa and Morimoto, 2020). The disorder of Notch signaling pathway is related to the occurrence of IPF, and the activation of Notch signaling can accelerate pulmonary fibrosis (Yang et al., 2022). Therefore, regulating the activation of Notch signaling pathway may be a new anti-fibrosis treatment strategy.
Although the pathogenesis of IPF remains unclear, a growing number of studies have implicated immune activation in its pathogenesis. Therefore, we used ssGSEA to further dissect the immune infiltration of the disease. In our results, we found that the expression of IHGs was correlated with different degrees of immune cell infiltration. The role of neutrophils in IPF remains unclear. On the one hand, inhibition of neutrophil chemokine CXCL8 or lack of neutrophil elastase can reduce bleomycin-induced pulmonary fibrosis (Gregory et al., 2015; Gschwandtner et al., 2017). Similarly, increased neutrophils are associated with decreased FVC and all-cause mortality in IPF patients (Achaiah et al., 2022). It has been reported that deletion of exon 18 of COL17A1 in mice leads to IL-17-related inflammatory responses in the skin and infiltration of eosinophils, neutrophils, T-cell and mast cells (Lindgren et al., 2023). XIV collagen is a neutrophil chemotactic factor that plays a role in neutrophil recruitment in rat inflammation (Nakagawa et al., 1999). On the other hand, the increase of neutrophils in BALF was not significantly related to the survival rate of IPF patients (Tabuena et al., 2005). Fibrosis caused mice infected with bacteria to show a higher mortality rate through the destruction of neutrophils (Warheit-Niemi et al., 2022). However, whether neutrophils have prognostic value in IPF is unclear, and further studies are needed to confirm the role of neutrophils in IPF. B-cell were increased in patients with IPF, which is consistent with our current findings. Activation of immune responses and increased infiltration of B-cell and macrophages are associated with IPF development (Xu et al., 2021). Previous studies have shown that B-cell and BLyS are elevated in patients with IPF and are inversely associated with patient outcome (Heukels et al., 2019). Consistent with increased B-cell activation, plasma IgA was elevated in IPF patients and inversely correlated with FVC (Heukels et al., 2019). These findings suggest that inhibition of B-cell activation has potential therapeutic value for IPF. In addition, our study showed a decrease in Tregs in IPF patients, which is consistent with previous findings. There is conflicting evidence supporting the role of Tregs in IPF. Tregs were originally thought to have anti-fibrotic effects. Tregs in peripheral blood and BALF were significantly reduced in IPF patients and were associated with decreased FVC (Kotsianidis et al., 2009). Subsequent studies have shown that tregs can promote fibrosis. The decrease of Tregs in peripheral blood and BALF of IPF patients is related to the degree of fibrosis (Peng et al., 2014) In the bleomycin-induced PF model, depletion of Tregs resulted in reduction of fibrosis, while induction or metastasis of Tregs resulted in worsening of fibrosis (Birjandi et al., 2016). Tregs may play different roles in different stages of fibrosis. Tregs play a pro-fibrotic role in the early stage of PF and a protective role in the late stage (Boveda-Ruiz et al., 2013). Tregs can promote collagen deposition and release of TGF-β in the early stage of PF (Lo Re et al., 2011). Tregs depletion attenuates PF by promoting Th17 response and regulating the shift of disturbed Th1/Th2 balance to Th1 dominance in lung tissue (Xiong et al., 2015). Consequently, we speculate that Tregs regulate different T-cell subsets at different stages of pulmonary fibrosis, which may explain the different roles of Tregs in pulmonary fibrosis. Existing studies have shown that inhibition of macrophage migration inhibitory factor (MIF) can downregulate the expression level of ITGA10 and reduce bleomycin-induced pulmonary inflammation and fibrosis in rats (Luo et al., 2021). The relationship between ITGA10 and MIF is still lacking relevant evidence and the future development of inhibitors targeting MIF may contribute to the treatment of pulmonary fibrosis. It has been reported that activated MMP7 is located on alveolar macrophages and proliferative epithelial cells (Fujishima et al., 2010). Future studies are needed to further confirm how IHGs participate in the pathological process of IPF by affecting immunity.
We further explored the correlation between IHGs expression and TME, immune subtypes and immune cell infiltration. Our study found that IHGs expression was linked with different levels of immune and stromal cell infiltration. Further analysis revealed that COL14A1 was positively correlated with CAF, DCs, Endo, HSC, Macrophage, Monocyte and Tregs. ITGA10 was positively correlated with CAF, Endo, HSC, Neutrophil and Tregs. MMP7 was positively correlated with CAF, DCs, Macrophage and Monocyte. The correlation between COL17A1 and immune cells was not prominent. In addition, we also found that COL14A1, COL17A1, and MMP7 were associated with more invasive immune subtypes, including C1, C2, and C6 subtypes. These results suggest that changes in immune and stromal cell composition make IHGs have different clinical features and immunotherapy responses. The relationship between IHGs expression and TME needs to be further studied at cellular and molecular levels.
We also explored the relationship between IHGs expression and immune checkpoint, tumor stemness score, TMB, and MSI. The results showed that IHGs expression was significantly related to RNAss and DNAss in most tumors. Previous studies have reported that higher stemness scores are linked with stronger tumor stem cell dedifferentiation and active biological processes, suggesting potential targets for chemotherapy drug development in cancer patients (Malta et al., 2018). IHGs expression level and immune checkpoint analysis showed that there was significant correlation between IHGs expression level and immune checkpoint in different tumors. Previous studies have shown that TMB is a good biomarker for predicting immunotherapy response in tumor patients (Goodman et al., 2017; Chan et al., 2019). MSI is also associated with prognosis, and high MSI indicates better prognosis (Ganesh et al., 2019). Our study suggested that IHGs expression was significantly related to TMB and MSI. These results reveal that immunotherapy may have potential benefits for cancer patients and may also help clinicians quickly identify patients who respond to immunotherapy. However, more studies at the molecular level are needed to fully explain the relationship between IHGs and immune response.
However, our research also has some shortcomings. Firstly, the biological mechanisms of COL14A1, COL17A1, ITGA10 and MMP7 in IPF and cancer are still unclear. Second, the results of this study need to be confirmed by relevant animals and human experiments. In the future research, we will continue to pay attention to the role of COL14A1, COL17A1, ITGA10 and MMP7 in IPF.
## 5 Conclusion
In summary, our study shows that BMs and immune disorders are closely associated with IPF. The IPF risk model based on IHGs showed that the high expression of COL14A1, COL17A1, ITGA10 and MMP7 was positively related to the risk of IPF. It was further confirmed that the AUC of the training set was 0.941 and that of the verification set was 0.917, indicating that COL14A1, COL17A1, ITGA10 and MMP7 were potential biomarkers for predicting the risk of IPF. Pan-cancer analysis showed that IHGs were related to prognosis, immune infiltration and drug sensitivity of cancer patients, and were expected to become new biomarkers for cancer patients. However, multicenter, large-scale and prospective studies are needed to confirm our results before COL14A1, COL17A1, ITGA10 and MMP7 are applied clinically.
## Data availability statement
Publicly available datasets were analyzed in this study. This data can be found here: The row data included in this study are available in GEO (https://www.ncbi.nlm.nih.gov/geo/, (accessed on 16 August 2022)) and UCSC Xena (https://xena.ucsc.edu/, (accessed on 24 June 2022)).
## Author contributions
YC designed the study, WY performed the analysis procedures, BC contributed to the revision of this article, HZ and XW contributed analysis tools, LC analyzed the results, and CF contributed to the writing of the manuscript. All authors have read and agreed to the published version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene.2023.1114601/full#supplementary-material
## References
1. Achaiah A., Rathnapala A., Pereira A., Bothwell H., Dwivedi K., Barker R.. **Neutrophil lymphocyte ratio as an indicator for disease progression in Idiopathic Pulmonary Fibrosis**. *BMJ Open Respir. Res.* (2022) **9** e001202. DOI: 10.1136/bmjresp-2022-001202
2. Allen R. J., Guillen-Guio B., Oldham J. M., Ma S. F., Dressen A., Paynton M. L.. **Genome-wide association study of susceptibility to idiopathic pulmonary fibrosis**. *Am. J. Respir. Crit. Care Med.* (2020) **201** 564-574. DOI: 10.1164/rccm.201905-1017OC
3. Ballester B., Milara J., Cortijo J.. **Idiopathic pulmonary fibrosis and lung cancer: Mechanisms and molecular targets**. *Int. J. Mol. Sci.* (2019) **20** 593. DOI: 10.3390/ijms20030593
4. Banerjee S., Lo W. C., Majumder P., Roy D., Ghorai M., Shaikh N. K.. **Multiple roles for basement membrane proteins in cancer progression and EMT**. *Eur. J. Cell Biol.* (2022) **101** 151220. DOI: 10.1016/j.ejcb.2022.151220
5. Bargagli E., Piccioli C., Rosi E., Torricelli E., Turi L., Piccioli E.. **Pirfenidone and Nintedanib in idiopathic pulmonary fibrosis: Real-life experience in an Italian referral centre**. *Pulmonology* (2019) **25** 149-153. DOI: 10.1016/j.pulmoe.2018.06.003
6. Bauer Y., White E. S., de Bernard S., Cornelisse P., Leconte I., Morganti A.. **MMP-7 is a predictive biomarker of disease progression in patients with idiopathic pulmonary fibrosis**. *ERJ Open Res.* (2017) **3** 00074. DOI: 10.1183/23120541.00074-2016
7. Bengtsson T., Aszodi A., Nicolae C., Hunziker E. B., Lundgren-Akerlund E., Fässler R.. **Loss of alpha10beta1 integrin expression leads to moderate dysfunction of growth plate chondrocytes**. *J. Cell Sci.* (2005) **118** 929-936. DOI: 10.1242/jcs.01678
8. Biondini D., Cocconcelli E., Bernardinello N., Lorenzoni G., Rigobello C., Lococo S.. **Prognostic role of MUC5B rs35705950 genotype in patients with idiopathic pulmonary fibrosis (IPF) on antifibrotic treatment**. *Respir. Res.* (2021) **22** 98. DOI: 10.1186/s12931-021-01694-z
9. Birjandi S. Z., Palchevskiy V., Xue Y. Y., Nunez S., Kern R., Weigt S. S.. **CD4(+)CD25(hi)Foxp3(+) cells exacerbate bleomycin-induced pulmonary fibrosis**. *Am. J. Pathol.* (2016) **186** 2008-2020. DOI: 10.1016/j.ajpath.2016.03.020
10. Boveda-Ruiz D., D'Alessandro-Gabazza C. N., Toda M., Takagi T., Naito M., Matsushima Y.. **Differential role of regulatory T cells in early and late stages of pulmonary fibrosis**. *Immunobiology* (2013) **218** 245-254. DOI: 10.1016/j.imbio.2012.05.020
11. Camper L., Hellman U., Lundgren-Akerlund E.. **Isolation, cloning, and sequence analysis of the integrin subunit alph a10, a beta1-associated collagen binding integrin expressed on chondro cytes**. *J. Biol. Chem.* (1998) **273** 20383-20389. DOI: 10.1074/jbc.273.32.20383
12. Chan T. A., Yarchoan M., Jaffee E., Swanton C., Quezada S. A., Stenzinger A.. **Development of tumor mutation burden as an immunotherapy biomarker: Utility for the oncology clinic**. *Ann. Oncol.* (2019) **30** 44-56. DOI: 10.1093/annonc/mdy495
13. Chen H., Qu J., Huang X., Kurundkar A., Zhu L., Yang N.. **Mechanosensing by the α6-integrin confers an invasive fibroblast phenotype and mediates lung fibrosis**. *Nat. Commun.* (2016) **7** 12564. DOI: 10.1038/ncomms12564
14. Collins S. L., Chan-Li Y., Hallowell R. W., Powell J. D., Horton M. R.. **Pulmonary vaccination as a novel treatment for lung fibrosis**. *PLoS One* (2012) **7** e31299. DOI: 10.1371/journal.pone.0031299
15. Drakopanagiotakis F., Wujak L., Wygrecka M., Markart P.. **Biomarkers in idiopathic pulmonary fibrosis**. *Matrix Biol.* (2018) **68-69** 404-421. DOI: 10.1016/j.matbio.2018.01.023
16. Elowsson Rendin L., Löfdahl A., Åhrman E., Müller C., Notermans T., Michaliková B.. **Matrisome properties of scaffolds direct fibroblasts in idiopathic pulmonary fibrosis**. *Int. J. Mol. Sci.* (2019) **20** 4013. DOI: 10.3390/ijms20164013
17. Fujishima S., Shiomi T., Yamashita S., Yogo Y., Nakano Y., Inoue T.. **Production and activation of matrix metalloproteinase 7 (matrilysin 1) in the lungs of patients with idiopathic pulmonary fibrosis**. *Arch. Pathol. Lab. Med.* (2010) **134** 1136-1142. DOI: 10.5858/2009-0144-OA.1
18. Ganesh K., Stadler Z. K., Cercek A., Mendelsohn R. B., Shia J., Segal N. H.. **Immunotherapy in colorectal cancer: Rationale, challenges and potential**. *Nat. Rev. Gastroenterol. Hepatol.* (2019) **16** 361-375. DOI: 10.1038/s41575-019-0126-x
19. Goodman A. M., Kato S., Bazhenova L., Patel S. P., Frampton G. M., Miller V.. **Tumor mutational burden as an independent predictor of response to immunotherapy in diverse cancers**. *Mol. Cancer Ther.* (2017) **16** 2598-2608. DOI: 10.1158/1535-7163.MCT-17-0386
20. Goto R., Nakamura Y., Takami T., Sanke T., Tozuka Z.. **Quantitative LC-MS/MS analysis of proteins involved in metastasis of B reast cancer**. *PloS one* (2015) **10** e0130760. DOI: 10.1371/journal.pone.0130760
21. Gregory A. D., Kliment C. R., Metz H. E., Kim K. H., Kargl J., Agostini B. A.. **Neutrophil elastase promotes myofibroblast differentiation in lung fibrosis**. *J. Leukoc. Biol.* (2015) **98** 143-152. DOI: 10.1189/jlb.3HI1014-493R
22. Gryfe R., Kim H., Hsieh E. T., Aronson M. D., Holowaty E. J., Bull S. B.. **Tumor microsatellite instability and clinical outcome in young patients with colorectal cancer**. *N. Engl. J. Med.* (2000) **342** 69-77. DOI: 10.1056/NEJM200001133420201
23. Gschwandtner M., Strutzmann E., Teixeira M. M., Anders H. J., Diedrichs-Möhring M., Gerlza T.. **Glycosaminoglycans are important mediators of neutrophilic inflammation**. *Cytokine* (2017) **91** 65-73. DOI: 10.1016/j.cyto.2016.12.008
24. Guay S. P., Brisson D., Mathieu P., Bossé Y., Gaudet D., Bouchard L.. **A study in familial hypercholesterolemia suggests reduced methylomic plasticity in men with coronary artery disease**. *Epigenomics* (2015) **7** 17-34. DOI: 10.2217/epi.14.64
25. Hamai K., Iwamoto H., Ishikawa N., Horimasu Y., Masuda T., Miyamoto S.. **Comparative study of circulating MMP-7, CCL18, KL-6, SP-A, and SP-D as disease markers of idiopathic pulmonary fibrosis**. *Dis. Markers* (2016) **2016** 4759040. DOI: 10.1155/2016/4759040
26. Heukels P., van Hulst J. A. C., van Nimwegen M., Boorsma C. E., Melgert B. N., von der Thusen J. H.. **Enhanced Bruton's tyrosine kinase in B-cells and autoreactive IgA in patients with idiopathic pulmonary fibrosis**. *Respir. Res.* (2019) **20** 232. DOI: 10.1186/s12931-019-1195-7
27. Hewlett J. C., Kropski J. A., Blackwell T. S.. **Idiopathic pulmonary fibrosis: Epithelial-mesenchymal interactions and emerging therapeutic targets**. *Matrix Biol.* (2018) **71-72** 112-127. DOI: 10.1016/j.matbio.2018.03.021
28. Horton W. B., Barrett E. J.. **Microvascular dysfunction in diabetes mellitus and cardiometabolic disease**. *Endocr. Rev.* (2021) **42** 29-55. DOI: 10.1210/endrev/bnaa025
29. Hou J., Shi J., Chen L., Lv Z., Chen X., Cao H.. **M2 macrophages promote myofibroblast differentiation of LR-MSCs and are associated with pulmonary fibrogenesis**. *Cell Commun. Signal* (2018) **16** 89. DOI: 10.1186/s12964-018-0300-8
30. Huang W. L., Wu S. F., Huang X., Zhou S.. **Integrated analysis of ECT2 and COL17A1 as potential biomarkers for pancreatic cancer**. *Dis. Markers* (2022) **2022** 9453549. DOI: 10.1155/2022/9453549
31. Huang X., Xiu H., Zhang S., Zhang G.. **The role of macrophages in the pathogenesis of ALI/ARDS**. *Mediat. Inflamm.* (2018) **2018** 1264913. DOI: 10.1155/2018/1264913
32. Jara P., Calyeca J., Romero Y., Plácido L., Yu G., Kaminski N.. **Matrix metalloproteinase (MMP)-19-deficient fibroblasts display a profibrotic phenotype**. *Am. J. Physiol. Lung Cell Mol. Physiol.* (2015) **308** L511-L522. DOI: 10.1152/ajplung.00043.2014
33. Jayadev R., Morais M., Ellingford J. M., Srinivasan S., Naylor R. W., Lawless C.. **A basement membrane discovery pipeline uncovers network complexity, regulators, and human disease associations**. *Sci. Adv.* (2022) **8** eabn2265. DOI: 10.1126/sciadv.abn2265
34. Jiang Y., Chen F., Ren X., Yang Y., Luo J., Yuan J.. **RNA-binding protein COL14A1, TNS1, NUSAP1 and YWHAE are valid biomarkers to predict peritoneal metastasis in gastric cancer**. *Front. Oncol.* (2022) **12** 830688. DOI: 10.3389/fonc.2022.830688
35. Khan F. A., Stewart I., Saini G., Robinson K. A., Jenkins R. G.. **A systematic review of blood biomarkers with individual participant data meta-analysis of matrix metalloproteinase-7 in idiopathic pulmonary fibrosis**. *Eur. Respir. J.* (2022) **59** 2101612. DOI: 10.1183/13993003.01612-2021
36. King T. E., Bradford W. Z., Castro-Bernardini S., Fagan E. A., Glaspole I., Glassberg M. K.. **A phase 3 trial of pirfenidone in patients with idiopathic pulmonary fibrosis**. *N. Engl. J. Med.* (2014) **370** 2083-2092. DOI: 10.1056/NEJMoa1402582
37. Kiyokawa H., Morimoto M.. **Notch signaling in the mammalian respiratory system, specifically the trachea and lungs, in development, homeostasis, regeneration, and disease**. *Dev. Growth Differ.* (2020) **62** 67-79. DOI: 10.1111/dgd.12628
38. Kondoh Y., Suda T., Hongo Y., Yoshida M., Hiroi S., Iwasaki K.. **Prevalence of idiopathic pulmonary fibrosis in Japan based on a claims database analysis**. *Respir. Res.* (2022) **23** 24. DOI: 10.1186/s12931-022-01938-6
39. Kotsianidis I., Nakou E., Bouchliou I., Tzouvelekis A., Spanoudakis E., Steiropoulos P.. **Global impairment of CD4+CD25+FOXP3+ regulatory T cells in idiopathic pulmonary fibrosis**. *Am. J. Respir. Crit. Care Med.* (2009) **179** 1121-1130. DOI: 10.1164/rccm.200812-1936OC
40. Kozawa K., Sekai M., Ohba K., Ito S., Sako H., Maruyama T.. **The CD44/COL17A1 pathway promotes the formation of multilayered, transformed epithelia**. *Curr. Biol.* (2021) **31** 3086-3097.e7. DOI: 10.1016/j.cub.2021.04.078
41. Kruegel J., Miosge N.. **Basement membrane components are key players in specialized extracellular matrices**. *Cell Mol. Life Sci.* (2010) **67** 2879-2895. DOI: 10.1007/s00018-010-0367-x
42. Kyprianou C., Christodoulou N., Hamilton R. S., Nahaboo W., Boomgaard D. S., Amadei G.. **Basement membrane remodelling regulates mouse embryogenesis**. *Nature* (2020) **582** 253-258. DOI: 10.1038/s41586-020-2264-2
43. Lee K. H., Shin S. J., Kim K. O., Kim M. K., Hyun M. S., Kim T. N.. **Relationship between E-cadherin, matrix metalloproteinase-7 gene expression and clinicopathological features in gastric carcinoma**. *Oncol. Rep.* (2006) **16** 823-830. DOI: 10.3892/or.16.4.823
44. Li H., Shen X., Ma M., Liu W., Yang W., Wang P.. **ZIP10 drives osteosarcoma proliferation and chemoresistance through ITGA10-mediated activation of the PI3K/AKT pathway**. *J. Exp. Clin. Cancer Res.* (2021) **40** 340. DOI: 10.1186/s13046-021-02146-8
45. Liao H. Y., Da C. M., Liao B., Zhang H. H.. **Roles of matrix metalloproteinase-7 (MMP-7) in cancer**. *Clin. Biochem.* (2021) **92** 9-18. DOI: 10.1016/j.clinbiochem.2021.03.003
46. Lindgren O., Le Menn G., Tuusa J., Chen Z. J., Tasanen K., Kokkonen N.. **Absence of NC14A domain of COLXVII/BP180 in mice results in IL-17‒associated skin inflammation**. *J. Invest. Dermatol* (2023) **143** 48-56.e7. DOI: 10.1016/j.jid.2022.07.019
47. Lo Re S., Lecocq M., Uwambayinema F., Yakoub Y., Delos M., Demoulin J. B.. **Platelet-derived growth factor-producing CD4+ Foxp3+ regulatory T lymphocytes promote lung fibrosis**. *Am. J. Respir. Crit. Care Med.* (2011) **184** 1270-1281. DOI: 10.1164/rccm.201103-0516OC
48. Lothong M., Sakares W., Rojsitthisak P., Tanikawa C., Matsuda K., Yodsurang V.. **Collagen XVII inhibits breast cancer cell proliferation and growth through deactivation of the AKT/mTOR signaling pathway**. *PLoS One* (2021) **16** e0255179. DOI: 10.1371/journal.pone.0255179
49. Luo Y., Yi H., Huang X., Lin G., Kuang Y., Guo Y.. **Inhibition of macrophage migration inhibitory factor (MIF) as a therapeutic target in bleomycin-induced pulmonary fibrosis rats**. *Am. J. Physiol. Lung Cell Mol. Physiol.* (2021) **321** L6-l16. DOI: 10.1152/ajplung.00288.2020
50. Mahalanobish S., Saha S., Dutta S., Sil P. C.. **Matrix metalloproteinase: An upcoming therapeutic approach for idiopathic pulmonary fibrosis**. *Pharmacol. Res.* (2020) **152** 104591. DOI: 10.1016/j.phrs.2019.104591
51. Maher T. M., Bendstrup E., Dron L., Langley J., Smith G., Khalid J. M.. **Global incidence and prevalence of idiopathic pulmonary fibrosis**. *Respir. Res.* (2021) **22** 197. DOI: 10.1186/s12931-021-01791-z
52. Mak K. M., Mei R.. **Basement membrane type IV collagen and laminin: An overview of their biology and value as fibrosis biomarkers of liver disease**. *Anat. Rec. Hob.* (2017) **300** 1371-1390. DOI: 10.1002/ar.23567
53. Malta T. M., Sokolov A., Gentles A. J., Burzykowski T., Poisson L., Weinstein J. N.. **Machine learning identifies stemness features associated with oncogenic dedifferentiation**. *Cell* (2018) **173** 338-354.e15. DOI: 10.1016/j.cell.2018.03.034
54. Matsuhira T., Nishiyama O., Tabata Y., Kaji C., Kubota-Ishida N., Chiba Y.. **A novel phosphodiesterase 4 inhibitor, AA6216, reduces macrophage activity and fibrosis in the lung**. *Eur. J. Pharmacol.* (2020) **885** 173508. DOI: 10.1016/j.ejphar.2020.173508
55. Mertens-Walker I., Fernandini B. C., Maharaj M. S. N., Rockstroh A., Nelson C. C., Herington A. C.. **The tumour-promoting receptor tyrosine kinase, EphB4, regulates expres sion of integrin-β8 in prostate cancer cells**. *BMC cancer* (2015) **15** 164. DOI: 10.1186/s12885-015-1164-6
56. Morris M. R., Ricketts C., Gentle D., Abdulrahman M., Clarke N., Brown M.. **Identification of candidate tumour suppressor genes frequently methylated in renal cell carcinoma**. *Oncogene* (2010) **29** 2104-2117. DOI: 10.1038/onc.2009.493
57. Nakagawa H., Takano K., Kuzumaki H.. **A 16-kDa fragment of collagen type XIV is a novel neutrophil chemotactic factor purified from rat granulation tissue**. *Biochem. Biophys. Res. Commun.* (1999) **256** 642-645. DOI: 10.1006/bbrc.1999.0393
58. Natsuga K., Watanabe M., Nishie W., Shimizu H.. **Life before and beyond blistering: The role of collagen XVII in epidermal physiology**. *Exp. Dermatol* (2019) **28** 1135-1141. DOI: 10.1111/exd.13550
59. Nishie W.. **Collagen XVII processing and blistering skin diseases**. *Acta Derm. Venereol.* (2020) **100** adv00054. DOI: 10.2340/00015555-3399
60. Niu J., Li X. M., Wang X., Liang C., Zhang Y. D., Li H. Y.. **DKK1 inhibits breast cancer cell migration and invasion through suppression of β-catenin/MMP7 signaling pathway**. *Cancer Cell Int.* (2019) **19** 168. DOI: 10.1186/s12935-019-0883-1
61. Okada T., Lee A. Y., Qin L. X., Agaram N., Mimae T., Shen Y.. **Integrin-α10 dependency identifies RAC and RICTOR as therapeutic targets in high-grade myxofibrosarcoma**. *Cancer Discov.* (2016) **6** 1148-1165. DOI: 10.1158/2159-8290.CD-15-1481
62. Peljto A. L., Zhang Y., Fingerlin T. E., Ma S. F., Garcia J. G., Richards T. J.. **Association between the MUC5B promoter polymorphism and survival in patients with idiopathic pulmonary fibrosis**. *Jama* (2013) **309** 2232-2239. DOI: 10.1001/jama.2013.5827
63. Peng X., Moore M. W., Peng H., Sun H., Gan Y., Homer R. J.. **CD4+CD25+FoxP3+ Regulatory Tregs inhibit fibrocyte recruitment and fibrosis via suppression of FGF-9 production in the TGF-β1 exposed murine lung**. *Front. Pharmacol.* (2014) **5** 80. DOI: 10.3389/fphar.2014.00080
64. Raghu G., Richeldi L., Jagerschmidt A., Martin V., Subramaniam A., Ozoux M. L.. **Idiopathic pulmonary fibrosis: Prospective, case-controlled study of natural history and circulating biomarkers**. *Chest* (2018) **154** 1359-1370. DOI: 10.1016/j.chest.2018.08.1083
65. Richeldi L., Collard H. R., Jones M. G.. **Idiopathic pulmonary fibrosis**. *Lancet* (2017) **389** 1941-1952. DOI: 10.1016/S0140-6736(17)30866-8
66. Richeldi L., du Bois R. M., Raghu G., Azuma A., Brown K. K., Costabel U.. **Efficacy and safety of nintedanib in idiopathic pulmonary fibrosis**. *N. Engl. J. Med.* (2014) **370** 2071-2082. DOI: 10.1056/NEJMoa1402584
67. Rousselle P., Montmasson M., Garnier C.. **Extracellular matrix contribution to skin wound re-epithelialization**. *Matrix Biol.* (2019) **75-76** 12-26. DOI: 10.1016/j.matbio.2018.01.002
68. Saftencu M., Braicu C., Cojocneanu R., Buse M., Irimie A., Piciu D.. **Gene expression patterns unveil new insights in papillary thyroid cancer**. *Med. Kaunas.* (2019) **55** 500. DOI: 10.3390/medicina55080500
69. Samstein R. M., Lee C. H., Shoushtari A. N., Hellmann M. D., Shen R., Janjigian Y. Y.. **Tumor mutational load predicts survival after immunotherapy across multiple cancer types**. *Nat. Genet.* (2019) **51** 202-206. DOI: 10.1038/s41588-018-0312-8
70. Sanli M., Akar E., Pehlivan S., Bakır K., Tuncozgur B., Isik A. F.. **The relationship of metalloproteinase gene polymorphisms and lung cancer**. *J. Surg. Res.* (2013) **183** 517-523. DOI: 10.1016/j.jss.2013.01.045
71. Schuppan D., Cantaluppi M. C., Becker J., Veit A., Bunte T., Troyer D.. **Undulin, an extracellular matrix glycoprotein associated with collagen fibrils**. *J. Biol. Chem.* (1990) **265** 8823-8832. DOI: 10.1016/s0021-9258(19)38962-8
72. Shapouri-Moghaddam A., Mohammadian S., Vazini H., Taghadosi M., Esmaeili S. A., Mardani F.. **Macrophage plasticity, polarization, and function in health and disease**. *J. Cell Physiol.* (2018) **233** 6425-6440. DOI: 10.1002/jcp.26429
73. Song J. W., Do K. H., Jang S. J., Colby T. V., Han S., Kim D. S.. **Blood biomarkers MMP-7 and SP-A: Predictors of outcome in idiopathic pulmonary fibrosis**. *Chest* (2013) **143** 1422-1429. DOI: 10.1378/chest.11-2735
74. Strieter R. M., Mehrad B.. **New mechanisms of pulmonary fibrosis**. *Chest* (2009) **136** 1364-1370. DOI: 10.1378/chest.09-0510
75. Su Y.-J., Lin W.-H., Chang Y.-W., Wei K.-C., Liang C.-L., Chen S.-C.. **Polarized cell migration induces cancer type-specific CD133/integrin/S rc/Akt/GSK3β/β-catenin signaling required for maintenance of cancer st em cell properties**. *Oncotarget* (2015) **6** 38029-38045. DOI: 10.18632/oncotarget.5703
76. Sun M., Chen Y., Liu X., Cui Y.. **LncRNACASC9 promotes proliferation, metastasis, and cell cycle inovarian carcinoma cells through cyclinG1/TP53/MMP7 signaling**. *Bioengineered* (2021) **12** 8006-8019. DOI: 10.1080/21655979.2021.1981795
77. Tabuena R. P., Nagai S., Tsutsumi T., Handa T., Minoru T., Mikuniya T.. **Cell profiles of bronchoalveolar lavage fluid as prognosticators of idiopathic pulmonary fibrosis/usual interstitial pneumonia among Japanese Patients**. *Respiration* (2005) **72** 490-498. DOI: 10.1159/000087673
78. Tamborero D., Rubio-Perez C., Muiños F., Sabarinathan R., Piulats J. M., Muntasell A.. **A pan-cancer landscape of interactions between solid tumors and infiltrating immune cell populations**. *Clin. Cancer Res.* (2018) **24** 3717-3728. DOI: 10.1158/1078-0432.CCR-17-3509
79. Thangavelu P. U., Krenács T., Dray E., Duijf P. H.. **In epithelial cancers, aberrant COL17A1 promoter methylation predicts its misexpression and increased invasion**. *Clin. Epigenetics* (2016) **8** 120. DOI: 10.1186/s13148-016-0290-6
80. Tzouvelekis A., Gomatou G., Bouros E., Trigidou R., Tzilas V., Bouros D.. **Common pathogenic mechanisms between idiopathic pulmonary fibrosis and lung cancer**. *Chest* (2019) **156** 383-391. DOI: 10.1016/j.chest.2019.04.114
81. van der Velden J. L., Ye Y., Nolin J. D., Hoffman S. M., Chapman D. G., Lahue K. G.. **JNK inhibition reduces lung remodeling and pulmonary fibrotic systemic markers**. *Clin. Transl. Med.* (2016) **5** 36. DOI: 10.1186/s40169-016-0117-2
82. Van Doren S. R.. **MMP-7 marks severe pancreatic cancer and alters tumor cell signaling by proteolytic release of ectodomains**. *Biochem. Soc. Trans.* (2022) **50** 839-851. DOI: 10.1042/BST20210640
83. Vracko R.. **Basal lamina scaffold-anatomy and significance for maintenance of orderly tissue structure**. *Am. J. Pathol.* (1974) **77** 314-346. PMID: 4614671
84. Warheit-Niemi H. I., Edwards S. J., SenGupta S., Parent C. A., Zhou X., O'Dwyer D. N.. **Fibrotic lung disease inhibits immune responses to staphylococcal pneumonia via impaired neutrophil and macrophage function**. *JCI Insight* (2022) **7** e152690. DOI: 10.1172/jci.insight.152690
85. Weis-Müller B. T., Modlich O., Drobinskaya I., Unay D., Huber R., Bojar H.. **Gene expression in acute stanford type A dissection: A comparative microarray study**. *J. Transl. Med.* (2006) **4** 29. DOI: 10.1186/1479-5876-4-29
86. West J. B., Mathieu-Costello O.. **Structure, strength, failure, and remodeling of the pulmonary blood-gas barrier**. *Annu. Rev. Physiol.* (1999) **61** 543-572. DOI: 10.1146/annurev.physiol.61.1.543
87. Wilson S. E.. **Corneal wound healing**. *Exp. Eye Res.* (2020) **197** 108089. DOI: 10.1016/j.exer.2020.108089
88. Wilson S. E.. **Fibrosis is a basement membrane-related disease in the cornea: Injury and defective regeneration of basement membranes may underlie fibrosis in other organs**. *Cells* (2022) **11** 309. DOI: 10.3390/cells11020309
89. Wolters P. J., Collard H. R., Jones K. D.. **Pathogenesis of idiopathic pulmonary fibrosis**. *Annu. Rev. Pathol.* (2014) **9** 157-179. DOI: 10.1146/annurev-pathol-012513-104706
90. Wynn T. A., Vannella K. M.. **Macrophages in tissue repair, regeneration, and fibrosis**. *Immunity* (2016) **44** 450-462. DOI: 10.1016/j.immuni.2016.02.015
91. Xiong S., Guo R., Yang Z., Xu L., Du L., Li R.. **Treg depletion attenuates irradiation-induced pulmonary fibrosis by reducing fibrocyte accumulation, inducing Th17 response, and shifting IFN-γ, IL-12/IL-4, IL-5 balance**. *Immunobiology* (2015) **220** 1284-1291. DOI: 10.1016/j.imbio.2015.07.001
92. Xu F., Tanabe N., Vasilescu D. M., McDonough J. E., Coxson H. O., Ikezoe K.. **The transition from normal lung anatomy to minimal and established fibrosis in idiopathic pulmonary fibrosis (IPF)**. *EBioMedicine* (2021) **66** 103325. DOI: 10.1016/j.ebiom.2021.103325
93. Yang D., Xu P., Su H., Zhong W., Xu J., Su Z.. **The histone methyltransferase DOT1L is a new epigenetic regulator of pulmonary fibrosis**. *Cell Death Dis.* (2022) **13** 60. DOI: 10.1038/s41419-021-04365-5
94. Yodsurang V., Tanikawa C., Miyamoto T., Lo P. H. Y., Hirata M., Matsuda K.. **Identification of a novel p53 target, COL17A1, that inhibits breast cancer cell migration and invasion**. *Oncotarget* (2017) **8** 55790-55803. DOI: 10.18632/oncotarget.18433
95. Yuan S., Lin L. S., Gan R. H., Huang L., Wu X. T., Zhao Y.. **Elevated matrix metalloproteinase 7 expression promotes the proliferation, motility and metastasis of tongue squamous cell carcinoma**. *BMC Cancer* (2020) **20** 33. DOI: 10.1186/s12885-020-6521-4
96. Zolak J. S., de Andrade J. A.. **Idiopathic pulmonary fibrosis**. *Immunol. Allergy Clin. North Am.* (2012) **32** 473-485. DOI: 10.1016/j.iac.2012.08.006
97. Zuo F., Kaminski N., Eugui E., Allard J., Yakhini Z., Ben-Dor A.. **Gene expression analysis reveals matrilysin as a key regulator of pulmonary fibrosis in mice and humans**. *Proc. Natl. Acad. Sci. U. S. A.* (2002) **99** 6292-6297. DOI: 10.1073/pnas.092134099
|
---
title: 'Endoscopic management of refractory leak and gastro-cutaneous fistula after
laparoscopic sleeve gastrectomy: a randomized controlled trial'
authors:
- Said Negm
- Bassam Mousa
- Ahmed Shafiq
- Mohamed Abozaid
- Ehab Abd Allah
- Adel Attia
- Taha AbdelKader
- Ahmed Farag
journal: Surgical Endoscopy
year: 2022
pmcid: PMC10017559
doi: 10.1007/s00464-022-09748-z
license: CC BY 4.0
---
# Endoscopic management of refractory leak and gastro-cutaneous fistula after laparoscopic sleeve gastrectomy: a randomized controlled trial
## Body
Laparoscopic sleeve gastrectomy (LSG) is one of the most performed surgical procedures for treatment of morbid obesity [1, 2]. Gastric leak is the highly feared complication following LSG and its observed incidence is 1–$2\%$ [3, 4], however other post-LSG complications, such as bleeding and stricture, are less frequently encountered with a median incidence of $1.2\%$ (range: 0.6–$1.6\%$) [5]. Compared to post-LSG gastric leak, the incidence of gastrointestinal (GI) leak following GI oncological surgeries is 8–$26\%$ and 3–$12\%$ in distal esophagectomy and total gastrectomy, respectively [6, 7], while the incidence of gastric leak after Roux-en-Y gastric bypass (RYGB) is 2–$8\%$ [8, 9]. Post-LSG gastric leak occurs through the anastomotic suture line of the sleeve greater curvature. Leaked luminal (gastric) contents may collect next to the anastomosis (leak) or exit through the skin or the drain (fistula) [10]. Post-LSG gastric leak may occur due to impaired healing at the sleeve greater curvature suture line resulting from increased intraluminal pressure associated with gastric sleeve twist, kink or stenosis, discrepancy between tissue thickness and staple height, impaired vascular supply and uncareful use of energy sources [11]. Gastric fistula and septic shock may follow post-LSG gastric leak [12]. According to its time-to-occurrence post-LSG, gastric leaks are classified as early (on or before 3rd postoperative day (POD)), intermediate (4th–7th PODs) and late (after 7th POD) [13]. Gastric leaks commonly occur between 5th and 6th PODs [13, 14]. The most common site ($86\%$) of post-LSG leaks is the proximal gastric sleeve particularly close to the gastroesophageal junction (angle of His), however leaks in the distal gastric sleeve occur in about $14\%$ [15]. The management of post-LSG leaks and gastro-cutaneous fistulae has not been well standardized yet [13]. It is possible to stabilize the patient and control the fistula, however, the control of leak is the most concerned issue that may pose difficulty, particularly if the leak is next to the esophagogastric junction. Patients with gastric leak who are hemodynamically unstable or in sepsis my require surgical intervention because the cost of conservative measures may be the patient’s life. Similarly, in many instances, post-LSG fistulae may not respond to conservative management, and intervention either surgically or endoscopic is usually mandated [16, 17]. Adoption of endoscopic techniques in the management of gastric leaks and gastro-cutaneous fistula has been tried in many studies [18]. Endoscopic management offers many advantages such as being less invasive, reducing septic shock and contamination, saving the time to take the proper decision, and resulting in better patient’s recovery [19, 20]. Endoscopic placement of a covered stent resulted in complete closure of post-LSG gastro-cutaneous fistula in 69–$100\%$ in early published series [21–24]. Repeat endoscopies for stent migration, retrosternal discomfort and reflux, and longer duration of external drainage were the common adverse effects associated with the use of stents in those early series [25]. Double-pigtail stent is commonly used to manage the gastric leak and permits internal drainage easing the perioperative management of gastric leak [26, 27]. A combination of covered stent and double-pigtail stent is a good option to manage gastric stenosis and associated gastric leak [26]. Endo-clips, which have been used for colonic perforation, are being used in management of post-LSG gastric leak but their role in chronic fistula is still controversial [28–30]. Lastly, endo-stitches have not been well evaluated as a primary tool for closure of post-LSG gastric leak or fistula, however it has numerous applications such as in sleeve gastroplasty, stent anchorage and closure of mucosal defects after endoscopic resections [31–35]. In this study, we compared the effectiveness of surgical versus different types of endoscopic intervention in management of post-LSG gastric leak and fistula.
## Abstract
### Background
Gastro-cutaneous fistula is a rare complication after laparoscopic sleeve gastrectomy (LSG) with incidence of occurrence 1–$2\%$. Most of gastro-cutaneous fistulae do not respond to conservative management and need intervention either surgically or endoscopically.
### Methods
This prospective randomized clinical study included referred patients who had LSG performed at our department or other centers, and complicated with post-LSG leak or gastro-cutaneous fistula between December/2019 and March/2021. Included patients were ASA *Physical status* I–II. Primary and secondary outcomes were recurrence of the fistula and mortality in each group after the intervention during the 18 months follow-up period, respectively.
### Results
Thirty patients were randomized into two groups: Surgery Group (SG, $$n = 15$$) and Endoscopy Group (EG, $$n = 15$$). Mean age of patients was 42.3 ± 8.7 and 42.6 ± 8.3 years-old in SG and EG, respectively. Females constituted $73.3\%$ and $80\%$ in SG and EG, respectively. Median time-to-gastric leak post LSG was six (range: 4–7) days in both groups. SG patients were surgically managed with primary repair of the gastric fistula and gastrojejunostomy in 13 patients or converting SG into Roux-en-Y gastric bypass in two patients, while EG patients were endoscopically managed with stitching, stenting, stenting and dilation, and clipping and dilation in 5, 4, 4 and 2 patients, respectively. Incidence of recurrent leak during 1st week was significantly higher in SG than EG ($p \leq 0.001$). No mortality reported in EG, while 2 patients died in SG ($$p \leq 0.48$$).
### Conclusion
Endoscopic intervention may offer a successful modality in managing post-LSG gastric leak and gastro-cutaneous fistula that do not respond to conservative measures in stable patients.
## Patients
We included all bariatric patients who developed gastric leak or fistula after LSG either performed at the Department of Surgery, Zagazig University Faculty of Medicine or referred to our department between December 2019 and March 2021. While all included patients were of the American Society of Anesthesiologists (ASA) *Physical status* I–II, patients with ASA status III–IV or those who demonstrated a satisfactory response to the conservative measures were excluded (Patients with physical status III and IV were managed according to their general, condition, clinical status and radiological and endoscopic findings either by conservative measures or surgery). This prospective randomized controlled clinical trial was approved by Zagazig University Faculty of Medicine Institutional Review Board (Approval Number: $\frac{11130}{2.12.2019}$) and performed in accordance with the code of ethics of the World Medical Association (Declaration of Helsinki) for studies involving human subjects. This study was retrospectively submitted in clinicaltrials.gov in May 2021 (NCT04879667). Written informed consent was obtained from all participants after explaining to them all the study procedures with its benefits and hazards.
Included patients were randomized at a 1:1 ratio to “Surgery Group, SG” or “Endoscopic Group” via the drawing of sealed envelopes containing computer-generated random numbers prepared by a third party before the start of the intervention. Sample size was calculated using open Epi program using the following data: confidence interval $95\%$, power of test $80\%$, ratio of unexposed/exposed 1, percent of patients with successful management of persistent gastric leak or fistula by surgical intervention $50\%$ and those with successful management by endoscopy $99\%$, odds ratio $99\%$, and risk ratio 2.
Primary and secondary outcomes were recurrence of the fistula and mortality in each group after the intervention during the 3 months follow-up period, respectively.
## Diagnosis
After full history taking and complete physical examination, post-LSG gastric leak was clinically suspected and then confirmed by laboratory investigations (complete blood picture, liver and kidney functions, coagulation profile), radiological imaging (chest X-ray, computed tomography (CT) with oral and I.V contrast) and upper GI endoscopy to assess the site, size and cause of the leak. We adopted a protocol of initial radiological or laparoscopic drainage according to the amount of intraperitoneal free fluid detected by CT scan, then endoscopically inserting a stent. If the leak did not satisfactorily respond to the initial measures within 6 weeks (recommended period of 5–8 weeks by the stent’s manufacturer and asa routine in our hospital for complete healing and easy extraction of the stent), a persistent gastric leak or fistula was considered, and the patient was evaluated for eligibility to be included in this study. The included eligible patients underwent another upper GI endoscopy to reconfirm site and size of the fistula, and CT abdomen with oral and I.V contrast to determine whether the fistula had a track (gastro-cutaneous fistula) or not (gastric leak).
## Intervention
Patients, randomized to the endoscopy group, underwent endoscopic stenting (fully covered self-expanded metallic stent, FCSEMS) in case of gastric leak, endoscopic Over-The-Scope Clipping (OTSC, Ovesco Endoscopy AG, Tubingen, Germany) in case of gastric-cutaneous fistula, endoscopic suturing (OverStitch [36], Apollo Endo-surgery, TX, United States) in case of large leak or fistula size regardless of the presence of track or not, and lastly, if there is distal sleeve pouch narrowing, we combined endoscopic OTSC or OverStitch with endoscopic balloon dilation (we had six cases were diagnosed with nonfunctional strictures due to fibrosis and successfully managed with balloon dilation there were no cases managed with strictureplasty). Patients, randomized to the SG, underwent either primary repair of the fistula and gastrojejunostomy or converting LSG into Roux-en-Y gastric bypass. Primary repair and gastrojejunostomy was utilized in cases of fistula in upper $\frac{1}{3}$ of the pouch, fistula of small size, large size sleeve pouch, old patients or patients without comorbidities. After primary closure of the fistula, a standard technique of gastrojejunostomy performed in antecolic orientation. The site of anastomosis was proximal to the site of the repaired fistula. The afferent loop is about 50 cm from duodenojejunal junction. while patients with fistula in middle and lower part of the pouch, fistula of large size, small size sleeve pouch, patients with good general conditions or patients with comorbidities were subjected to Roux en Y gastric bypass.
## Follow up after endoscopy and discharge from the hospital
All patients were clinically examined, and laboratory checked during the hospital stay. Any suspected gastric leak post repair mandated CT scan with oral and I.V contrast and upper GI endoscopy. After discharge, patients who had undergone balloon dilation as a part of their repair procedure, were endoscoped every 4 weeks to continue the dilation till relief of distal pouch narrowing. Patients were followed-up for 18 months post repair.
## Statistical analysis
Analysis of data was performed using SPSS (Statistical Package of Social Services) version 22. Quantitative variables were described as mean (± SD, standard deviation) and median (range) according to Shapiro test of normality. Qualitative variables were described as number and percentage. Chi-square test was used to compare qualitative variables between the two groups. Fisher exact test was used when one expected cell or more are less than five. Unpaired t-test was used to compare quantitative variables, in parametric data (SD < $30\%$ of the mean). Mann Whitney test was used instead of unpaired t-test in non-parametric data (SD > $30\%$ of the mean). The results were considered statistically significant when the significant probability was less than 0.05 ($P \leq 0.05$). P-value < 0.001 was considered highly statistically significant (HS), and P-value ≥ 0.05 was considered statistically insignificant (NS).
## Results
Of 67 (12 and 55 post-LSG leak or fistula patients with primary surgery performed in our department and other centers, respectively) patients who presented with post-LSG leak or fistula, 30 patients ($\frac{12}{30}$ and $\frac{18}{30}$ with primary LSG performed in our department and other centers, respectively) met the inclusion criteria for this study. The eligible 30 patients were randomized into two groups: SG and EG (Fig. 1). the other 37 patients were excluded due to: [1] 19 patients refused to participate to study after explaining to them the protocol of management and those were managed by surgery performed by a different team (resuscitation first in ICU then surgery either classic Roux en-Ygastric bypass, primary repair and gastrojejenostomy or drainage of any collection by interventional radiology then surgery later on according to the status of each patient).; [ 2] 18 patients did not met the inclusion criteria as there were ASA III and IV (they were ASA I and II then became ASA III and IV just before stent insertion at the start of our protocol of leak management), some of them also presented with septic shock and unstable general condition and were also managed like the 19 patients who refused to participate. Fig. 1Consort flow chart Mean age of patients with post-LSG gastric leak or fistula was 42.3 ± 8.7 and 42.6 ± 8.3 years-old in SG and EG, respectively. Females constituted $73.3\%$ ($\frac{11}{15}$) and $80\%$ ($\frac{12}{15}$) of patients in SG and EG, respectively (Table 1). Patients with diabetes mellitus were $13.3\%$ ($\frac{2}{15}$) and $20\%$ ($\frac{3}{15}$) of SG and EG, respectively (Table 1). Gastric fistula with epithelized track was recorded in $33.3\%$ ($\frac{5}{15}$) and $53.3\%$ ($\frac{8}{15}$) of SG and EG, respectively (Table 1) (this is a reference to a leak controlled by a drain). Median time-to-gastric leak post LSG was 6 (range: 4–7) days in both groups (Table 1). The fundus was the most common site of gastric fistula in $80\%$ ($\frac{12}{15}$) and $66.7\%$ ($\frac{10}{15}$) of SG and EG patients, respectively. However, the rest of patients in both groups experienced gastric fistula at the middle of pouch greater curvature (Table 1). There was no statistically significant difference regarding gastric leak or fistula diameter between the 2 groups ($$P \leq 0.18$$); being of small (< 1 cm) diameter in 5 ($33.3\%$) patients in each group; of moderate (1–2 cm) diameter in $60\%$ ($\frac{9}{15}$) and $33.3\%$ ($\frac{5}{15}$) of patients in SG and EG, respectively; of large (> 2 cm) diameter in $6.7\%$ ($\frac{1}{15}$) and $33.3\%$ ($\frac{5}{15}$) of patients in SG and EG, respectively (Table 1). SG patients were surgically managed with primary repair of the gastric fistula and gastrojejunostomy ($86.7\%$, $\frac{13}{15}$), or converting SG into Roux-en-Y gastric bypass ($13.3\%$, $\frac{2}{15}$), while EG patients were endoscopically managed with OverStitch ($33.3\%$, $\frac{5}{15}$), FCSEMS ($26.7\%$, $\frac{4}{15}$), dilation and FCSEMS ($26.7\%$, $\frac{4}{15}$), and dilation and OTSC ($13.3\%$, $\frac{2}{15}$) (Table 2).Table 1Characteristics of the patientsSurgery group ($$n = 15$$) Mean ± SD; n (%)Endoscopy group ($$n = 15$$) Mean ± SD; n (%)P-valueAge (years)42.3 ± 8.742.6 ± 8.30.42Gender Male4 (26.7)3 [20] Female11 (73.3)12 [80]Body mass index46.9 ± 3.745.7 ± 3.80.76Comorbidities0.9 Hypertension2 (13.3)2 (13.3) Diabetes mellitus2 (13.3)3 [20] Sleep apnea2 (13.3)3 [20] Osteoarthritis2 (13.3)2 (13.3) Infertility01 (6.7) No comorbidities7 (46.7)4 (26.7)Time of leak post sleeve, median (range) days6 (4–7)6 (4–7)0.78Site of fistula0.7 Fundus12 [80]10 (66.7) Middle of pouch sleeve3 [20]5 (33.3)Size of fistula0.2 Small (< 1 cm)5 (33.3)5 (33.3) Moderate (1–2 cm)9 [60]5 (33.3) Large (> 2 cm)1 (6.7)5 (33.3)Post-LSG0.46 Leak10 (66.7)7 (46.7) Fistula (gastro-cutaneous)5 (33.3)8 (53.3)Table 2Operative intervention and postoperative recurrenceSurgery group ($$n = 15$$) n (%)Endoscopy group ($$n = 15$$) n (%)P-valueType of intervention Surgical primary repair and gastrojejunostomy13 (86.7)0 < 0.001 Surgical RYGB2 (13.3)0 Endoscopic stenting and dilation04 (26.7) Endoscopic stenting alone04 (26.7) Endoscopic clipping alone00 Endoscopic suturing05 (33.3) Endoscopic clipping and dilation02 (13.3)*Recurrent fistula* within first week No1 (6.7)15 [100] < 0.001 Yes14 (93.3)0Highly significant P-value < 0.001 are given in bold The observed incidence of recurrent gastric leak during the first week post-repair was significantly higher in SG than EG ($P \leq 0.001$); being $93.3\%$ ($\frac{14}{15}$) and $0\%$ ($\frac{0}{15}$) in SG and EG, respectively (Table 2). In SG, recurrent gastro-cutaneous fistula with track and gastric fistula without track (leak) occurred in 9 ($60\%$) and 5 ($33.3\%$) patients, respectively (Table 2). SG patients who experienced recurrence ($\frac{14}{15}$) post repair needed endoscopic management while none in EG needed further endoscopic management of the fistula post-repair. During the 18 months follow-up period, EG demonstrated no cases of recurrent gastric fistula post-repair (Table 2).
No patients died in EG, while two patients died in SG and this difference was not statistically significant ($$P \leq 0.48$$) (Table 2).
## Discussion
*In* general, the most frightful complication after bariatric surgery is the anastomotic leak with an incidence of 0.8–$6\%$ [37–39]. During 30-day follow-up post-LSG, the gastric leak was $0.8\%$ [40]. In addition to identification of the site, the core principles in managing any GI fistula or leak is to drain the leaked contents and avoid further contamination by diverting the luminal contents or closure of the fistula or leak [36]. In hemodynamically stable patients, the first step to manage post-LSG gastric leak or fistula is bowel rest, percutaneous drainage, and adequate nutritional support. Failed conservative measures call for intervention whether surgically or endoscopically [41].
In our department, we perform about 500 LSG/year, and the incidence of post-LSG leaks or fistula is about $2.5\%$. Surgical management of gastro-cutaneous fistula after laparoscopic sleeve gastrectomy has increased incidence rates of morbidity and mortality. In this study, we experienced high failure rate of surgical intervention ($93.3\%$) within the first week post-repair (recurrence of fistula within one week in surgical group patients was due to long remaining sleeve pouch with axial rotation or marked narrowing of the pouch so, primary repair and gastrojejunostomy are usually associated with high recurrence rates of fistula because the main cause is not corrected, with axial rotation or marked narrowing of remaining pouch, the intra-gastric pressure increases and site of fistula opens again. This can be solved by classic Roux- en- y gastric bypass but not all patients are candidates for it, and also if the fistula occurred at upper third of the remaining pouch (most common site of fistula), the recurrence rate becomes high and classic Roux- en- y gastric bypass becomes difficult in case of upper third fistula due to marked adhesions and unhealthy remaining tissue at site of fistula). Immediate surgical intervention, with abdominal washout, irrigation, wide drainage and attempts for suturing of the leak if the tissues permit, may be preferred in unstable patients with early type leak [21].
On the other hand, endoscopic intervention has become the corner stone in managing the post-LSG gastric leak or gastro-cutaneous fistula with different modalities such as stenting, clipping, balloon dilatation and endo-suturing. In our study, OTSC, along with dilation, was used in two patients ($\frac{2}{15}$ of endoscopic group), and none demonstrated clip migration or development of post-OTSC stricture over 3 months follow-up period. We started deploying the clips perpendicular to the long axis of the defect. If needed, multiple clips were placed sequentially, starting at either edge of the defect towards the center. Standard clips were passed through-the-scope to achieve superficial tissue apposition engaging the mucosa and submucosa (with 1.2 mm-wide and 6 mm-long arms capable of an approximately 12 mm grasp) and were used in conjunction with thermal ablation or mechanical scraping of the tissue around the edges of the defect to achieve a more resilient seal. In a retrospective study, OTSC demonstrated a lower success rate ($50\%$) in managing GI fistulas in 30 patients ($\frac{25}{30}$ patients were post-bariatric surgery: 22 post-sleeve gastrectomy and three post-RYGB) with a median-time-to-OTSC delivery was 147 (range = 5–880) days [42]. Additionally, stricture post-OTSC developed after 30 days at the gastroesophageal junction in one patient who had post-sleeve gastric fistula in the previously mentioned study [42]. The authors, in the previously mentioned study, used an endoscopic cap with its diameter bigger than the defect and utilized “suction technique” that allowed better approximation of the edges with the inclusion of omentum or fat inside the clip, however, the authors did not recommend the use of graspers as it may reduce the endoscopic flexibility and suction applied at the cap [37, 38]. OTSC demonstrated a statistically significant successful closure rate for GI perforations and leaks (average $82\%$) compared to that of fistulas ($42.9\%$), and long-term success of OTSC as a primary than a rescue therapeutic option ($69\%$ vs. $46.9\%$, respectively, $$P \leq 0.004$$) for managing GI perforations and leaks, as well [39]. A systematic review concluded that OTS clips achieved successful closure rate of $51.5\%$ in GI fistulae and $66\%$ in GI anastomotic leaks [40].
In this study, we used a fully covered stent (Mega stent, Taewoong Medical Industries, Gyeonggi-do, South Korea) ultra large and long (length: 24 cm, diameter: 36 mm) stent. We did not experience any complication with Mega stent, particularly migration, thanks to the design of Mega stent that fits well for the post-sleeve anatomy with reduction of migration. It completely covers the whole sleeve pouch and its lower end rests in the duodenum [43]. The reported migration incidence of FCSEMS is twice that of partially covered stents ($26\%$ vs. $13\%$) [44]. A case series reported the success of using Mega stent for post-sleeve leaks [45]. Mega stent demonstrated $82\%$ success rate in closure of primary and secondary (after surgical repair) leaks following sleeve gastrectomy and RYGB, however, Mega stent use was combined with clips in selected cases in the previous study [46].
OverStitching is theoretically an optimum method of leak closure because it is the only true full thickness leak closure and performed endoscopically despite being a complex procedure. In this study, OverStitching was used in $33.3\%$ of patients who underwent endoscopic management. The procedure began with de-epithelialization of the edges of the leak using argon plasma coagulation before applying the OverStitching system. We did not experience post-OverStitch gastric leak. Granata et al. reported $77\%$ success rate of endoscopic management of gastric leak using direct stitches only [47]. Moreover, the same previous study demonstrated an increased success rate ($85\%$) of endoscopic suturing combined with FCSEMS and anchoring compared to direct stitches alone [47]. In managing GI fistulae and leaks using endoscopic suturing technique, Mukewar et al. reported a $100\%$ immediate success rate and $40\%$ sustained clinical success rate; noting that that gastro-gastric fistulae comprised almost half of the cases in that study [48].
This study has some limitations. The small sample size that may not give powerful statistical conclusions. Exclusion of patients with ASA status > II is another limitation. Regarding the de-epithelization of the edges of fistula either for OTSC or OverStitch, there was not a single method, and it was up to the endoscopist to use argon plasma laser or mechanical scrapping of the edges. Moreover, this study showed only 3 months follow-up period. The strength of the present study is being a randomized controlled trial and comparing different endoscopic interventions on one hand with the surgical intervention on the other hand.
## Conclusion
Endoscopic intervention can be a successful modality in managing post-LSG gastric leak and gastro-cutaneous fistula without the need to surgical intervention. No recurrence leak or fistula was noted after endoscopic clipping, stitching, or stenting. Further studies with large sample size and longer follow-up period are in demand to conclude strong and valid results.
## References
1. Cottam D, Qureshi FG, Mattar SG. **Laparoscopic sleeve gastrectomy as an initial weight-loss procedure for high-risk patients with morbid obesity**. *Surg Endosc* (2006) **20** 859-863. DOI: 10.1007/s00464-005-0134-5
2. Carubbi F, Ruscitti P, Pantano I. **Jejunoileal bypass as the main procedure in the onset of immune-related conditions: the model of BADAS**. *Expert Rev Clin Immunol* (2013) **9** 441-452. DOI: 10.1586/eci.13.26
3. Giuliani A, Romano L, Papale E. **Complications of postlaparoscopic sleeve gastric resection: review of surgical technique**. *Minerva Chir* (2019) **74** 213-217. DOI: 10.23736/S0026-4733.19.07883-0
4. Bellanger DE, Greenway FL. **Laparoscopic sleeve gastrectomy, 529 cases without a leak: short-term results and technical considerations**. *Obes Surg* (2011) **21** 146-150. DOI: 10.1007/s11695-010-0320-y
5. Khoursheed M, Al-Bader I, Mouzannar A. **Postoperative bleeding and leakage after sleeve gastrectomy: a single-center experience**. *Obes Surg* (2016) **26** 3007. DOI: 10.1007/s11695-016-2317-7
6. Blencowe NS, Strong S, McNair AG. **Reporting of short-term clinical outcomes after esophagectomy: a systematic review**. *Ann Surg* (2012) **255** 658-666. DOI: 10.1097/SLA.0b013e3182480a6a
7. Lang H, Piso P, Stukenborg C. **Management and results of proximal anastomotic leaks in a series of 1114 total gastrectomies for gastric carcinoma**. *Eur J Surg Oncol* (2000) **26** 168-171. DOI: 10.1053/ejso.1999.0764
8. Morales MP, Miedema BW, Scott JS. **Management of postsurgical leaks in the bariatric patient**. *Gastrointest Endosc Clin N Am* (2011) **21** 295-304. DOI: 10.1016/j.giec.2011.02.008
9. Gonzalez R, Sarr MG, Smith CD. **Diagnosis and contemporary management of anastomotic leaks after gastric bypass for obesity**. *J Am Coll Surg* (2007) **204** 47-55. DOI: 10.1016/j.jamcollsurg.2006.09.023
10. Bruce J, Russell EM, Mollison J. **The quality of measurement of surgical wound infection as the basis for monitoring: a systematic review**. *J Hosp Infect* (2001) **49** 99-108. DOI: 10.1053/jhin.2001.1045
11. Frattini F, Delpini R, Inversini D. **Gastric leaks after sleeve gastrectomy: focus on pathogenetic factors**. *Surg Technol Int* (2017) **31** 123-126. PMID: 29313318
12. Giuliani A, Romano L, Marchese M. **Gastric leak after laparoscopic sleeve gastrectomy: management with endoscopic double pigtail drainage. A systematic review**. *Surg Obes Relat Dis.* (2019) **15** 1414-1419. DOI: 10.1016/j.soard.2019.03.019
13. Rosenthal RJ, Diaz AA, Arvidsson D, Baker RS, Basso N. **International sleeve gastrectomy expert panel consensus statement: best practice guidelines based on experience of >12,000 cases**. *Surg Obes Relat Dis* (2012) **8** 8-19. DOI: 10.1016/j.soard.2011.10.019
14. Committee SG. **SAGES guideline for clinical application of laparoscopic bariatric surgery**. *Surg Obes Relat Dis* (2009) **5** 387-405. DOI: 10.1016/j.soard.2009.01.010
15. Walsh C, Karmali S. **Endoscopic management of bariatric complications: a review and update**. *World J Gastrointest Endosc* (2015) **7** 518-523. DOI: 10.4253/wjge.v7.i5.518
16. Gagner M, Buchwald JN. **Comparison of laparoscopic sleeve gastrectomy leak rates in four staple-line reinforcement options: a systematic review**. *Surg Obes Relat Dis* (2014) **10** 713-723. DOI: 10.1016/j.soard.2014.01.016
17. Gagner M. **Comment on: gastric leak after laparoscopic sleeve gastrectomy: management with endoscopic double pigtail drainage. A systematic review**. *Surg Obes Relat Dis.* (2019) **15** 1419. DOI: 10.1016/j.soard.2019.05.032
18. Nimeri A, Ibrahim M, Maasher A. **Management algorithm for leaks following laparoscopic sleeve gastrectomy**. *Obes Surg* (2016) **26** 21-25. DOI: 10.1007/s11695-015-1751-2
19. Parikh M, Issa R, McCrillis A. **Surgical strategies that may decrease leak after laparoscopic sleeve gastrectomy: a systematic review and meta-analysis of 9991 cases**. *Ann Surg* (2013) **257** 231-237. DOI: 10.1097/SLA.0b013e31826cc714
20. Al-Sabah S, Ladouceur M, Christou N. **Anastomotic leaks after bariatric surgery: it is the host response that matters**. *Surg Obes Relat Dis.* (2008) **4** 152-157. DOI: 10.1016/j.soard.2007.12.010
21. Sakran N, Goitein D, Raziel A, Beglaibter N, Grinbaum R. **Gastric leaks after sleeve gastrectomy: a multicenter experience with 2,834 patients**. *Surg Endosc.* (2013) **27** 240-5. DOI: 10.1007/s00464-012-2426-x
22. Moszkowicz D, Arienzo R, Khettab I. **Sleeve gastrectomy severe complications: is it always a reasonable surgical option?**. *Obes Surg* (2013) **23** 676-686. DOI: 10.1007/s11695-012-0860-4
23. Simon F, Siciliano I, Gillet A. **Gastric leak after laparoscopic sleeve gastrectomy: early covered self-expandable stent reduces healing time**. *Obes Surg* (2013) **23** 687-692. DOI: 10.1007/s11695-012-0861-3
24. Csendes A, Burdiles P, Burgos AM. **Conservative management of anastomotic leaks after 557 open gastric bypasses**. *Obes Surg* (2005) **15** 1252-1256. DOI: 10.1381/096089205774512410
25. Pequignot A, Fuks D, Verhaeghe P. **Is there a place for pigtail drains in the management of gastric leaks after laparoscopic sleeve gastrectomy?**. *Obes Surg* (2012) **22** 712-720. DOI: 10.1007/s11695-012-0597-0
26. Rebibo L, Bartoli E, Dhahri A. **Persistent gastric fistula after sleeve gastrectomy: an analysis of the time between discovery and reoperation**. *Surg Obes Relat Dis* (2016) **12** 84-93. DOI: 10.1016/j.soard.2015.04.012
27. Donatelli G, Dumont JL, Cereatti F. **Treatment of leaks following sleeve gastrectomy by endoscopic internal drainage (EID)**. *Obes Surg* (2015) **25** 1293-1301. DOI: 10.1007/s11695-015-1675-x
28. Diez-Redondo P, Blanco JI, Lorenzo-Pelayo S. **A novel system for endoscopic closure of iatrogenic colon perforations using the Ovesco(R) clip and omental patch**. *Rev Esp Enferm Dig* (2012) **104** 550-552. DOI: 10.4321/S1130-01082012001000009
29. Aly A, Lim HK. **The use of over the scope clip (OTSC) device for sleeve gastrectomy leak**. *J Gastrointest Surg* (2013) **17** 606-608. DOI: 10.1007/s11605-012-2062-8
30. Minami S, Gotoda T, Ono H. **Complete endoscopic closure of gastric perforation induced by endoscopic resection of early gastric cancer using endoclips can prevent surgery (with video)**. *Gastrointest Endosc* (2006) **63** 596-601. DOI: 10.1016/j.gie.2005.07.029
31. Abu Dayyeh BK, Rajan E, Gostout CJ. **Endoscopic sleeve gastroplasty: a potential endoscopic alternative to surgical sleeve gastrectomy for treatment of obesity**. *Gastrointest Endosc* (2013) **78** 530-535. DOI: 10.1016/j.gie.2013.04.197
32. Vargas EJ, Bazerbachi F, Rizk M. **Transoral outlet reduction with full thickness endoscopic suturing for weight regain after gastric bypass: a large multicenter international experience and meta-analysis**. *Surg Endosc* (2018) **32** 252-259. DOI: 10.1007/s00464-017-5671-1
33. Fujii LL, Bonin EA, Baron TH. **Utility of an endoscopic suturing system for prevention of covered luminal stent migration in the upper GI tract**. *Gastrointest Endosc* (2013) **78** 787-793. DOI: 10.1016/j.gie.2013.06.014
34. Kantsevoy SV, Bitner M, Mitrakov AA. **Endoscopic suturing closure of large mucosal defects after endoscopic submucosal dissection is technically feasible, fast, and eliminates the need for hospitalization (with videos)**. *Gastrointest Endosc* (2014) **79** 503-507. DOI: 10.1016/j.gie.2013.10.051
35. Sharaiha RZ, Kumta NA, DeFilippis EM. **A large multicenter experience with endoscopic suturing for management of gastrointestinal defects and stent anchorage in 122 patients: a retrospective review**. *J Clin Gastroenterol* (2016) **50** 388-392. DOI: 10.1097/MCG.0000000000000336
36. Ge PS, Thompson CC. **The use of the overstitch to close perforations and fistulas**. *Gastrointest Endosc Clin N Am* (2020) **30** 147-161. DOI: 10.1016/j.giec.2019.08.010
37. Donatelli G, Cereatti F, Dhumane P. **Closure of gastrointestinal defects with Ovesco clip: long-term results and clinical implications**. *Therap Adv Gastroenterol* (2016) **9** 713-721. DOI: 10.1177/1756283X16652325
38. Zhang XL, Qu JH, Sun G. **Feasibility study of secure closure of gastric fundus perforation using over-the-scope clips in a dog model**. *J Gastroenterol Hepatol* (2012) **27** 1200-1204. DOI: 10.1111/j.1440-1746.2012.07156.x
39. Haito-Chavez Y, Law JK, Kratt T. **International multicenter experience with an over-the-scope clipping device for endoscopic management of GI defects (with video)**. *Gastrointest Endosc* (2014) **80** 610-622. DOI: 10.1016/j.gie.2014.03.049
40. Kobara H, Mori H, Nishiyama N. **Over-the-scope clip system: a review of 1517 cases over 9 years**. *J Gastroenterol Hepatol* (2019) **34** 22-30. DOI: 10.1111/jgh.14402
41. Burgos AM, Braghetto I, Csendes A. **Gastric leak after laparoscopic-sleeve gastrectomy for obesity**. *Obes Surg* (2009) **19** 1672-1677. DOI: 10.1007/s11695-009-9884-9
42. Morrell DJ, Winder JS, Johri A, Docimo S. **Over-the-scope clip management of non-acute, full-thickness gastrointestinal defects**. *Surg Endosc* (2020) **34** 2690-2702. DOI: 10.1007/s00464-019-07030-3
43. Basha J, Appasani S, Sinha SK. **Mega stents: a new option for management of leaks following laparoscopic sleeve gastrectomy**. *Endoscopy.* (2014) **46** E49-E50. DOI: 10.1055/s-0033-1359120
44. van Boeckel PG, Sijbring A, Vleggaar FP. **Systematic review: temporary stent placement for benign rupture or anastomotic leak of the oesophagus**. *Aliment Pharmacol Ther* (2011) **33** 1292-1301. DOI: 10.1111/j.1365-2036.2011.04663.x
45. Galloro G, Magno L, Musella M. **A novel dedicated endoscopic stent for staple-line leaks after laparoscopic sleeve gastrectomy: a case series**. *Surg Obes Relat Dis* (2014) **10** 607-611. DOI: 10.1016/j.soard.2014.02.027
46. Shehab HM, Hakky SM, Gawdat KA. **An endoscopic strategy combining mega stents and over-the-scope clips for the management of post-bariatric surgery leaks and fistulas (with video)**. *Obes Surg* (2016) **26** 941-948. DOI: 10.1007/s11695-015-1857-6
47. Granata A, Amata M, Martino A. **Full-thickness gastric plication with overstitch endoscopic suturing device for postsurgical chronic gastroparesis**. *Endoscopy* (2020) **52** E235-E236. DOI: 10.1055/a-1076-0652
48. Mukewar S, Kumar N, Catalano M. **Safety and efficacy of fistula closure by endoscopic suturing: a multi-center study**. *Endoscopy* (2016) **48** 1023-1028. DOI: 10.1055/s-0042-114036
|
---
title: Combined effect of graded Thera-Band and scapular stabilization exercises on
shoulder adhesive capsulitis post-mastectomy
authors:
- Nancy H. Aboelnour
- FatmaAlzahraa H. Kamel
- Maged A. Basha
- Alshimaa R. Azab
- Islam M. Hewidy
- Mohamed Ezzat
- Noha M. Kamel
journal: Supportive Care in Cancer
year: 2023
pmcid: PMC10017571
doi: 10.1007/s00520-023-07641-6
license: CC BY 4.0
---
# Combined effect of graded Thera-Band and scapular stabilization exercises on shoulder adhesive capsulitis post-mastectomy
## Abstract
### Purpose
The main aim of the trial was to assess the combined impact of graded Thera-Band strengthening exercises and scapular stabilization exercises on shoulder pain, physical function, and quality of life (QoL) in post-mastectomy adhesive capsulitis (AC).
### Methods
Seventy females with unilateral post-mastectomy AC partook in the trial. Participants were subdivided equally into two groups at random. Both groups obtained the traditional physical therapy program; in addition, the intervention group received graded Thera-Band exercises for shoulder muscles and scapular stabilization exercises 5 days a week for 8 weeks. Range of motion (ROM) and muscle power of shoulder were assessed by digital goniometer and handheld dynamometer, respectively. Disability of the Arm, Shoulder, and Hand questionnaire (DASH) was utilized for assessment of shoulder function and visual analogue scale (VAS) for pain measurement while short-form (SF-36) for QoL assessment. All evaluation data was recorded prior to the trial and at the eighth week of interventions for both groups.
### Results
All participants achieved improvements in shoulder ROM, muscle power, pain, and all aspects of QoL; however, higher statistical improvements were reported in all measurements with respect to strengthening exercises group ($p \leq 0.001$).
### Conclusion
The addition of graded Thera-Band strengthening exercises and scapular stabilization exercises in post-mastectomy AC rehabilitation program has significant benefits in shoulder function and patients’ QoL.
Trial registration: This study is retrospectively registered at ClinicalTrials.gov NCT05311839.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00520-023-07641-6.
## Introduction
Breast cancer surgeries particularly mastectomy results in limited shoulder movement which can lead to arm, shoulder pain, and stiffness [1]. Females who underwent mastectomy have reported a significantly higher incidence of shoulder morbidity ($17\%$) [2]. Shoulder joint connective tissue fibrosis is common in post-mastectomy patients [3]. One of the most common symptoms of upper extremity morbidities is restriction of shoulder joint range of motion (ROM) which is linked to a lower quality of life (QoL) [4].
Adhesive capsulitis (AC) or “frozen shoulder” is an insidious inflammatory disorder characterized by a painful, progressive decrease in passive or active glenohumeral joint ROM caused by gradual fibrosis and subsequent contracture of the glenoid capsule [5]. AC frequently progresses through 3 different phases; the first one is the painful freezing phase, remains from 2 to 9 months, and is associated with emergence of sharp, diffuse shoulder discomfort that usually aggravates during night; the second phase is the frozen phase, lasts from 4 to 12 months, where pain starts to fade with a gradual decrease in glenohumeral joint ROM; and the last phase is the thawing phase where there is a gradual regain of ROM and takes from 5 months to 2 years to complete [6, 7].
AC can be treated either conservatively or surgically. Conservative treatment consists of a variety of exercise techniques and physical therapy modalities such as hot–cold therapy [8], transcutaneous electrical nerve stimulation (TENS), ultrasound (US), acupuncture [9, 10], and laser [11]. Active and passive ROM exercises, self-stretching, stretching exercises under the guidance of a physiotherapist, mobilization and manipulation techniques, resistance exercises, patient education, and home exercises are all part of the exercise program [12].
Exercise therapy might help in the reduction of pain and restoration of the range, coordination, and control of movement in patients with AC [13]. Graded resistance exercises are effective in decreasing fatigue levels and enhancing functional capacity and muscle strength [14]. Progressive strengthening exercise is extremely effective in reducing sarcopenia. Strength training improves muscle strength by increasing muscle mass and enhancing the recruitment and the firing rate of motor units [15]. Thera-Bands are a type of resistance exercise training tools which can provide a variable resistance and allow changes over ROM, thus preventing the risk of higher weight loading during strengthening exercises. In addition, elastic bands can provide efficient resistance and enhance muscle activation for promoting muscle strength in treatment of shoulder diseases. They are frequently used during exercise rehabilitation programs due to their simplicity, economic, and safety benefits [16–18].
The scapular stabilization exercise can be applied to patients with limited shoulder joint mobility and functional deterioration such as shoulder impingement syndrome, AC, and rotator cuff injury. As a result, rehabilitation exercises which strengthen the scapula’s stability can be extremely beneficial in rehabilitation treatment of patients with shoulder pain and problems [19].
Since the incidence of shoulder morbidity is six times higher following mastectomy compared to conservative therapy [20], which in turn influencing the patients’ ability on performing ADL, in addition to the lack of literature research on AC rehabilitation following mastectomy, the need of developing an efficient exercises program is necessary; hence, this experiment targeted to assess the impact of combining Thera-Band strengthening exercises with scapular stabilization training on AC following mastectomy in term of shoulder pain, function, and QoL.
## Design and setting
This study was a randomized controlled experimental trial and was approved by the Ethical Committee of Faculty of Physical Therapy, Cairo University. The study was retrospectively registered in the Clinical Trials Registry (No: NCT05311839), and it was carried out between January 2021 and December 2021 at the outpatient clinic of Faculty of Physical Therapy, Cairo University. All the procedures used were conducted following the ethical rules of the Declaration of Helsinki.
## Participants
Seventy females diagnosed with one-sided post-mastectomy AC participated in this study. All patients were diagnosed by a specialized orthopedist. The procedures of this trial were carried out at the outpatient clinic at the Faculty of Physical Therapy, Cairo University. Females who met the following inclusion criteria participated in the study; age ranged from 40 to 60 years, 2nd phase of AC, shoulder pain and stiffness for at least 3 months, and restriction in shoulder ROM (involving flexion, abduction, and internal/external rotation) less than $50\%$ when compared to the other shoulder. The exclusion criteria included shoulder or acromioclavicular joint osteoarthritis, bone diseases, infection, severe osteoporosis, tumors or metastasis, history of previous shoulder trauma or accidental injuries, previous history of dislocation, surgery on the specific shoulder, and any other shoulder problems as supraspinatus tendonitis and impingement, neurological diseases (parkinsonism, stroke, radiating pain to arm), recent shoulder fracture or wound, diabetes mellitus, rheumatoid arthritis, and severe psychiatric illness.
Prior to participating in the trial, all participants were educated about the trial’s purposes, benefits, and steps and signed a consent form. Small groups of one to four patients underwent the sessions at a time, supervised by specialized physical therapist. The study included two equal groups of participants who were allocated randomly. The intervention group received graded Thera-Band exercises for the restricted shoulder ROM (flexion, abduction, and internal/external rotation) and scapular stabilization exercises (60 min, 5 days per week for 8 weeks) plus a conventional physiotherapy program including hot packs, active ROM exercises, pendular exercises, wall climb exercises, mobilization exercises, and shoulder capsular stretching. The control group obtained only the conventional physiotherapy program (30–40 min, 5 days weekly for 8 weeks).
## Sample size and randomization
The estimated sample size for this trial was 32 participants for each group, as calculated using G*POWER statistical software (version 3.1.9.2) according to shoulder external rotation data from a pilot study conducted on 5 subjects per group. α = 0.05, power is $80\%$, effect size = 0.72, and allocation ratio N2/N1 = 1 was used in the calculations. For a possible drop out, the sample size was increased by $10\%$. Randomized assignment of patients was done equally either to intervention group ($$n = 35$$) or control group ($$n = 35$$). The randomization process was carried out by a physiotherapist who was not involved in the data collection processes. Each patient received a unique computer-generated random code using the GraphPad software© (1:1 simple randomization), concealed in a sealed envelope. The sealed envelopes were given to the physiotherapist before the treatment. The results were collected prior to the trial and at the 8th week of therapy by therapist who was blind to the distribution process.
## Treatment procedures
All patients within both groups received the conventional physiotherapy program, 5 days a week for 8 weeks, and the duration of treatment was approximately 30–40 min.
## The control group
The participants within this group received the conventional physiotherapy program including hot packs (5–10 min), passive mobilization exercises for glenohumeral (GH) joint, and scapulothoracic articulation. GH joint mobilizations (posterior gliding for increasing flexion and internal rotation, caudal glide for increasing abduction, and anterior gliding for increasing external rotation). Scapulothoracic mobilization was applied for improving the movements of scapula (protraction/retraction, elevation/depression, and rotation). For active ROM exercises and pendular exercises, the patient was instructed to lean forward and place the unaffected hand on a table. While keeping a straight back and a relaxed shoulder, softly sway the arm forth and backward; the exercise was repeated by shifting the arm from side to side and then in an orbicular movement (10 reps, 5 times per day), wall climb exercises (hold for 15–30 s at the peak, 10 reps, 5 times daily), and capsular stretching exercises (anterior, posterior and inferior capsular stretches) and sustaining for 15–30 s, 5 reps, 5 times/day [21, 22].
## The intervention group
The participants within this group received graded Thera-Band exercises and scapular stabilization exercises (60 min, 5 days per week for 8 weeks) plus a conventional physiotherapy program.
## Thera-Band exercises
Patients in the intervention group received graded Thera-Band exercises (Thera-Band®, Hygenic Corporation, Akron, OH, USA). For Thera-Band application for shoulder flexors and abductors, the patient was in a comfortable standing position with both feet firmly positioned on the Thera-Band. The patient was instructed to grasp the end of the Thera-Band and gradually flex and abduct the shoulder from the starting position, hold for (25 s), and then return to the starting position without bouncing. For shoulder internal rotation, the patient was asked to stand and hold the Thera-Band in the hand while the direction of resistive power away from the side at the level of elbow which is bent to 90°. The patient was instructed to internally rotate the arm by pulling across the front of the trunk. For shoulder external rotation, the patient was instructed to stand with Thera-Band beside the body at level of elbow and to flex her elbow to 90°, grasping the elastic band and rotating the arm laterally. All patients performed the above exercises for 30 min with 2–3 series of 10–15 reps of every exercise. All patients started strengthening exercises with the yellow color, progressed to the red, and then the green. When the patient could easily complete three sets of 10–15 reps, progression to the next color was considered [21–23].
## Scapular stabilization exercises
In addition to graded Thera-Band exercises, the patients in the intervention group performed the following scapular stabilization exercises for 30 min, 10 reps each:Scapular clock exercises: in the standing position, the patient was asked to place the arm on a wall with fully extended elbow, with the fingers directed towards the 12, 3, 6, and 9 o’clock positions. These exercises improved scapular elevation, protraction, depression, and retraction, respectively. Towel slide exercises: with slightly flexed elbow, the patient was instructed to place the hand on a towel on the wall and wash the wall in approximately a 12 inch backward and forward motion, moving from the extended arm, retracted scapula, and to a flexed arm and protracted scapula. Ball stabilization exercises: while standing close to the wall, the participant was asked to position her affected hand on the ball and keep the ball from moving as disturbance was applied in different directions. Lawnmower exercises: the patient stood with abducted legs, bent knees, and holding a weight in the hand for resistance. The patient was asked to pull using large amounts of lower extremity extension and trunk rotation to guide the shoulder movement. Serratus anterior punch: in standing position, the patient was instructed to perform alternative serratus anterior punches while holding the Thera-Band for resistance [24, 25].
## Outcome measurements
Physical function, pain, and QoL were among the outcome measures. The measures were taken before the trial and at the end of the eighth week of therapy. Shoulder range of motion, muscles strength, and the Disability of the Arm, Shoulder, and Hand questionnaire (DASH) are all used to measure physical function. Shoulder pain was assessed by visual analogue scale (VAS). The Medical Outcomes Study short-form questionnaire (SF-36) was used to assess QoL.
A digital goniometer with good reliability (r > 0.84) was utilized to quantify shoulder flexion and abduction ROM in supine with extended elbow, while internal/external rotations were measured in sitting with adducted shoulder and mid position of forearm. All measures were taken 3 times, and then the average was scored [26, 27].
A handheld dynamometer (J Tech Commender Muscle Tester, Salt Lake City, UT, USA), a valid tool ($r = 0.81$), was employed to test the muscle power of shoulder flexors, abductors, and internal/external rotators by measuring maximum isometric contraction in kg. Each record was done 3 times, and then the average was scored [28].
The DASH was designed to evaluate upper limb disorders and impairment and track changes and functional level over time. This questionnaire’s Arabic version is regarded as a plain, reliable ($r = 0.97$), and validated ($r = 0.94$) measurement instrument. The more the score, the more severe the symptoms [29].
The VAS is a reliable measure with an ICC of 0.97, consists of 10-cm line, and was utilized to quantify the pain severity within shoulder, where the score of zero means no pain, while a score of ten means significant pain [30].
The short-form SF-36, including physical function, role physical, general health, vitality, bodily pain, mental health, role emotional, and social function, was used for assessment of patients’ QoL. The Cronbach’s alpha coefficient was 0.94, while the inter-rater reliability was outstanding (ICC = 0.98) [31].
## Statistical analysis
An unpaired t test was employed to compare subjects’ characteristics between the groups. The chi-squared test was used to compare the allocation of the afflicted arm and adjunctive therapy between the groups. For ensuring normal data allocation, the Shapiro–Wilk test was utilized. For group homogeneity determination, Levene’s test for homogeneity of variances was utilized. The effect of therapy on VAS, DASH, shoulder ROM and strength, and QoL was investigated using a mixed model MANOVA. Post hoc test utilizing Bonferroni correction was performed for subsequent various comparisons. All statistical measurements had a significant level of p 0.05. For all statistical analysis, IBM SPSS (Chicago, IL, USA) version 25 for Windows (IBM SPSS, Chicago, IL, USA) was used.
## Participants’ characteristics
The patients’ flow diagram through the trial is illustrated in Fig. 1. Seventy patients participated in this study with no significant difference in subjects’ age, BMI, affected arm, and adjunctive therapy distribution between the groups ($p \leq 0.05$) (Table 1).Fig. 1The patients’ flow diagram through the trial Table 1Participants’ characteristics Intervention groupControl groupp valueAge, mean ± SD (years)49.54 ± 5.3350.57 ± 4.690.39BMI, mean ± SD (kg/m2)27.43 ± 1.4427.61 ± 1.070.54Time since surgery, mean ± SD (month)12.57 ± 1.6813.05 ± 1.590.21Affected arm, n (%) Right22 ($62.9\%$)19 ($54.3\%$)0.46 Left13 ($37.1\%$)16 ($45.7\%$)Adjunctive therapy, n (%) Radiotherapy22 ($62.9\%$)24 ($68.6\%$)0.61 Chemotherapy25 ($71.4\%$)21 ($60\%$)0.31 Hormonal10 ($28.6\%$)12 ($34.3\%$)0.61SD, standard deviation; p value, probability value
## Intervention effect on VAS, DASH, shoulder ROM, and strength and QoL
The interaction of intervention and time was significant (F [18,51] = 139.81, $$p \leq 0.001$$, = 0.98). The main effect of time was significant (F [18,51] = 1181.78, $$p \leq 0.001$$, = 0.99). The main effect of intervention was significant (F [18,51] = 35.59, $$p \leq 0.001$$, = 0.92).
## Comparisons in each group
The VAS and DASH of both groups were statistically lower following intervention contrast to pre-intervention (p 0.001) (Table 2). The post-therapy shoulder ROM (flexion, abduction, external and internal rotation) and shoulder strength (flexors, abductors, internal/external rotators) of both groups were statistically higher than the pre-therapy levels (p 0.001) (Tables 3 and 4). Both groups demonstrated a significant improvement in all aspects of their QoL after intervention compared to pre-intervention ($p \leq 0.001$) (Table 5).Table 2Mean VAS and DASH pre- and post-treatment of both groupsIntervention groupControl groupMean ± SDMean ± SDMDp valueVAS Pre-treatment6.88 ± 1.457.11 ± 1.56-0.230.52 Post-treatment2.54 ± 1.034.57 ± 1.39-2.030.001 MD (% of change)4.34 ($63.08\%$)2.54 ($35.72\%$)$$p \leq 0.001$$$p \leq 0.001$DASH Pre-treatment43.88 ± 6.5944.62 ± 6.09-0.740.62 Post-treatment18.85 ± 4.2230.17 ± 5.51-11.320.001 MD (% of change)25.03 ($57.04\%$)14.45 ($32.38\%$)$$p \leq 0.001$$$p \leq 0.001$SD, standard deviation; MD, mean difference; p value, probability valueTable 3Mean shoulder ROM pre- and post-treatment of both groupsROM (degrees)Intervention groupControl groupMean ± SDMean ± SDMDp valueFlexion Pre-treatment100.94 ± 10.7498.97 ± 11.21.970.45 Post-treatment162.37 ± 11.54135.22 ± 9.8827.150.001 MD (% of change) − 61.43 ($60.86\%$) − 36.25 ($36.63\%$)$$p \leq 0.001$$$p \leq 0.001$Abduction Pre-treatment80.4 ± 9.2279.8 ± 9.340.60.78 Post-treatment128.48 ± 9.16109.4 ± 10.2319.080.001 MD (% of change) − 48.08 (59.8) − 29.6 (37.09)$$p \leq 0.001$$$p \leq 0.001$External rotation Pre-treatment39.51 ± 6.5738.6 ± 5.660.910.53 Post-treatment68.77 ± 8.4655.68 ± 6.613.090.001 MD (% of change) − 29.26 ($74.06\%$) − 17.08 ($44.25\%$)$$p \leq 0.001$$$p \leq 0.001$Internal rotation Pre-treatment46.2 ± 4.8245.88 ± 4.850.320.78 Post-treatment75.4 ± 4.9861.4 ± 5.83140.001 MD (% of change) − 29.2 ($63.2\%$) − 15.52 ($33.83\%$)$$p \leq 0.001$$$p \leq 0.001$SD, standard deviation; MD, mean difference; p value, probability valueTable 4Mean shoulder strength pre- and post-treatment of both groupsStrength (kg)Intervention groupControl groupMean ± SDMean ± SDMDp valueFlexors Pre-treatment7.17 ± 1.156.97 ± 1.20.20.47 Post-treatment11.94 ± 1.478.91 ± 1.013.030.001 MD (% of change) − 4.77 ($66.53\%$) − 1.94 ($27.83\%$)$$p \leq 0.001$$$p \leq 0.001$Abductors Pre-treatment6.88 ± 1.056.57 ± 0.910.310.18 Post-treatment10.82 ± 1.298.34 ± 1.082.480.001 MD (% of change) − 3.94 (57.27) − 1.77 (26.94)$$p \leq 0.001$$$p \leq 0.001$External rotators Pre-treatment7.02 ± 1.096.65 ± 1.130.370.16 Post-treatment11.17 ± 1.28.28 ± 1.12.890.001 MD (% of change) − 4.15 ($59.12\%$) − 1.63 ($24.51\%$)$$p \leq 0.001$$$p \leq 0.001$Internal rotators Pre-treatment6.6 ± 1.116.2 ± 1.130.40.14 Post-treatment10.25 ± 0.957.8 ± 1.052.450.001 MD (% of change) − 3.65 ($55.3\%$) − 1.6 ($25.81\%$)$$p \leq 0.001$$$p \leq 0.001$SD, standard deviation; MD, mean difference; p value, probability valueTable 5Mean pre- and post-treatment of quality of life of both groupsIntervention groupControl groupMean ± SDMean ± SDMDp valuePhysical functioning Pre-treatment49.82 ± 3.8848.57 ± 3.941.250.18 Post-treatment73.05 ± 5.7260.6 ± 5.1212.450.001 MD (% of change) − 23.23 ($46.63\%$) − 12.03 ($24.77\%$)$$p \leq 0.001$$$p \leq 0.001$Role physical Pre-treatment47.57 ± 4.8146.71 ± 4.830.860.46 Post-treatment67.02 ± 5.8157.74 ± 4.539.280.001 MD (% of change) − 19.45 (40.89) − 11.03 (23.61)$$p \leq 0.001$$$p \leq 0.001$Bodily pain Pre-treatment46.48 ± 7.1545.2 ± 6.61.280.43 Post-treatment66.02 ± 8.6656.45 ± 6.579.570.001 MD (% of change) − 19.54 ($42.04\%$) − 11.25 ($24.89\%$)$$p \leq 0.001$$$p \leq 0.001$General health Pre-treatment44.17 ± 7.2843.4 ± 7.270.770.65 Post-treatment67.34 ± 5.8855.14 ± 6.0912.20.001 MD (% of change) − 23.17 ($52.46\%$) − 11.74 ($27.05\%$)$$p \leq 0.001$$$p \leq 0.001$Vitality Pre-treatment45.62 ± 7.2644.8 ± 6.950.820.62 Post-treatment62.57 ± 7.0852.88 ± 7.279.690.001 MD (% of change) − 16.95 ($37.15\%$) − 8.08 ($18.04\%$)$$p \leq 0.001$$$p \leq 0.001$Social health Pre-treatment54.68 ± 6.3953.82 ± 4.340.860.51 Post-treatment73.48 ± 5.6663.28 ± 4.9710.20.001 MD (% of change) − 18.8 ($34.38\%$) − 9.46 ($17.58\%$)$$p \leq 0.001$$$p \leq 0.001$Emotional health Pre-treatment46.68 ± 4.5246.31 ± 4.310.370.72 Post-treatment69.34 ± 5.5157.51 ± 4.8511.830.001 MD (% of change) − 22.66 ($48.54\%$) − 11.2 ($24.18\%$)$$p \leq 0.001$$$p \leq 0.001$Mental health Pre-treatment43.4 ± 7.2743.02 ± 6.610.380.82 Post-treatment67.97 ± 6.5554.17 ± 5.3613.80.001 MD (% of change) − 24.57 ($56.61\%$) − 11.15 ($25.92\%$)$$p \leq 0.001$$$p \leq 0.001$SD, standard deviation; MD, mean difference; p value, probability value
## Comparisons between both groups
Pre-therapy, statistical difference between the groups was not reported ($p \leq 0.05$). Following therapy, the intervention group showed statistical lowering in VAS and DASH than the control group ($p \leq 0.001$) (Table 2). Moreover, comparing to the control group, the intervention group had higher statistical improvements in shoulder ROM (flexion, abduction, external/internal rotation), strength (flexors, abductors, external and internal rotators) (Tables 3 and 4), and in all aspects of QoL after therapy ($p \leq 0.001$) (Table 5).
## Discussion
Pain and ROM restriction within AC are likely due to fascial constraints, muscle stiffness with trigger spots, in addition to capsular and ligamentous rigidity [32]. This study was a randomized retrospective-controlled trial that aimed to assess the impacts of shoulder strengthening and scapular stabilizing exercises on physical function, pain, and QoL in post-mastectomy patients with AC, where the control group received 30–40 min of traditional therapy and the intervention group with both interventions received > 60 min. The results illustrated higher statistical improvements of all parameters in favor to the strengthening exercises group. All patients in this trial achieved improvement in pain and shoulder joint ROM, as mobilization has been showed to lower pain through the neurophysiologic impacts of mobilization on peripheral mechanoreceptor activation and nociceptors inhibition; additionally, improvement of ROM could be explained by the effect of mobilization on shoulder AC as posterior–anterior glide was chosen to improve the outer ROM, while caudal glide was chosen to improve abduction. It is possible that the glenohumeral joint’s posterior–anterior and caudal glides improved capsular extensibility and lengthened soft tissues, which were restricting joint motion. As a result of the greater capsular extensibility, the glenohumeral joint may have had more ROM. These treatments are also hypothesized to boost proprioceptive and kinesthetic sensations within the joint, allowing participants to perform tasks within their new ROM; as a result, the individual’s ROM can be maintained. Another possible explanation is the impact of stretching exercises which have been shown to improve the extensibility of soft tissue through the creep response, modifying viscoelastic characteristics and hence increasing ROM. Individuals must perform tasks within their newly acquired range of motion in order to retain joint motion. This conclusion backs up prior studies indicating that mobilization and stretching exercises can help with AC [33–36].
Following mastectomy, the main cause of shoulder dysfunction is not only the glenohumeral joint disorder but also the adhesions in the axillary and pectoral areas between the pectoral muscles, subcutaneous tissue, and skin that may prevent complete extension of the pectoralis, resulting in limitation of both shoulder flexion and abduction [37, 38]. So, introducing progressive strengthening exercises during treatment could overcome the consequence muscles weakness caused by adhesions and encourage gaining more shoulder range, promote neuromuscular control, ameliorate general strength, and facilitate optimal strength ratios of the rotator cuff and scapular rotator muscles, and hence facilitate rapid recovery from AC, and that may illustrate the higher statistical improvement in all shoulder ROM within the resistance exercises group compared to the other group which in turn facilitate the activities of daily living (ADL) and improve the shoulder function. These results support and come in line with other studies [21–23, 39–42] and emphasize on the prominent role of strengthening exercises in AC rehabilitation, as Harishkumar et al. [ 2017] [21] added Thera-Band strengthening exercises to traditional care in AC cases and evaluated the shoulder function and ROM after 3 weeks of intervention which showed higher statistical improvements than receiving traditional care only, while Rawat et al. [ 2016] [22] studied the impacts of gradual resistance exercises of rotator cuff muscles beside receiving TENS and shoulder joint mobilization techniques for 1 month and reported significant ameliorations in shoulder functional level, pain, muscle strength, and ROM. Datar and Devi [23] studied the effects of Thera-Band strengthening exercises on shoulder dysfunction following mastectomy in term of muscle power, shoulder functional capacities, and ADL; evaluations were done after 2 months of intervention and demonstrated that Thera-Band exercises have significantly improved the outcome measures for strength and daily tasks involving upper extremity.
Shrug sign is one of the issues that may be exhibited during application of resistance exercises for shoulder abduction. One of the benefits of Thera-*Band is* the ability to assist in rotator cuff retraining during starting abduction, as the elastic band generates “upward and inward” resistance vector against which the patient must push in a “downward and outward” direction. This action activates the start of abduction additionally the rotator cuff’s depression and stability activities, which occur before and during abduction. According to anecdotal evidence, this exercise helps patients with the shrug sign diminish early upper trapezius activity during abduction [32].
In terms of muscle strength, the resistance training group demonstrated better statistical results, which could be illustrated by the fact that strengthening exercises cause significant physiological variations in skeletal muscles including contractile and/or non-contractile muscle compositions. Mechanical stress causes disruption of myofibers and extracellular matrix, which stimulates protein synthesis, resulting in muscle growth by increment the sarcomeres number, which leads to an increase in pinnation angle and fascicle length and so muscle expansion [43].
Several studies have documented alteration of scapular motion in the mastectomy side which in turn affect the shoulder motion pattern and contribute to incidence of frozen shoulder [44, 45]; scapular stabilizers’ weakness causes a disruption in the scapula–humeral rhythm, resulting in shoulder dysfunction and micro-damage to the shoulder muscles, capsule, and ligaments. The scapula must rotate vertically, tilt posteriorly, and rotate externally during overhead exercises; scapular stabilizers’ weakness causes an imbalance of force coupling between the trapezius, serratus anterior, and rhomboids, resulting in downward rotation, anterior tilting, and internal rotation of the scapula during arm abduction. This fatigue-induced weakness deficit may have a detrimental influence on scapular posture and allow for increased lateral scapular gliding during functional activities [25], so scapular stabilization is an important part of exercises therapy for optimizing scapular alignment during upper extremity movement and for providing direct control of the scapular posture, allowing for proper length–tension ratios in the shoulder muscles. Hence, adding scapular stabilization exercises in the rehabilitation program is very essential and coincided with various studies that support the role of these exercises in normally restoring both shoulder and scapular balance and motion [24, 25, 46–49] and that could be explained as one of the factors that assist in shoulder AC improvement, as Kirthika et al. [ 2015] [24] and Gulwani [2020] [25] added scapular stabilization exercises to conventional care in cases with 2nd phase of AC and evaluated the shoulder function and ROM after 2 weeks of intervention and concluded that those exercises are beneficial for enhancing shoulder ROM and functional abilities; also, Yatheendra et al. [ 2015] [46] evaluated the impacts of combined scapular stabilization exercises and mobilization techniques on patients with AC for 4 weeks; the study comes to the conclusion that both interventions are effective in reducing shoulder discomfort, enhancing ROM, and improving functional capacity in AC.
Moreover, the experimental group demonstrated greater improvement in ADL regarding to the DASH scores and additionally in QoL which may be attributed to the crucial role of strengthening exercises among the various forms of physical exercise programs due to the link between muscle impairment, pain, and dysfunction [50], as strengthening of shoulder and scapular stabilizers has significant impacts in reducing pain and improving shoulder ROM, functional capacity, and muscle power by regaining scapula–humeral rhythm in AC [32].
The study confirmed the importance of exercises therapy in AC management without reporting any adverse effects and presents the preliminary evidences for introducing strengthening exercises as an essential part in AC rehabilitation; however, some limitations must be considered when explaining these results; the most significant drawback of this experiment was the lack of scapular movement analysis which could provide better statistical results; other restrictions are the absence of blinding during the treatment sessions and absence of AC imaging assessments that could provide better prognosis; also the long-term effect of treatment was not examined due to the difficulty of following up after the trial, so future trials analyzing the scapular motion with patients’ follow-up are recommended; also it is critical to promote knowledge about the protection, early diagnosis, and quick treatment of shoulder problems following mastectomy in order to reduce women sufferance and financial costs, so trials should be conducted to evaluate early physical therapy intervention in prevention shoulder morbidity following mastectomy; moreover, evaluation of the impact of different approaches of exercises therapy with longer duration should be carried out.
## Conclusion
Rehabilitation exercises program including progressive strengthening of both scapular and shoulder muscles played a significant role in improvements of shoulder ROM and function which reflected on patients’ QoL and ADL, so emphasis on strengthening exercises during rehabilitation can be of great benefit in AC treatment.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOC 219 KB)
## References
1. Oliveira M, Gurgel M, Miranda M, Okubo M, Feijó L, Souza G. **Efficacy of shoulder exercises on locoregional complications in women undergoing radiotherapy for breast cancer: clinical trial**. *Braz J Phys Ther* (2009.0) **13** 136-143. DOI: 10.1590/S1413-35552009005000017
2. Blomqvist L, Stark B, Engler N, Malm M. **Evaluation of arm and shoulder mobility and strength after modified radical mastectomy and radiotherapy**. *Acta Oncol* (2004.0) **43** 280-283. DOI: 10.1080/02841860410026170
3. Dilaveri CA, Sandhu NP, Neal L, Neben-Wittich MA, Hieken TJ, Mac Bride MB. **Medical factors influencing decision making regarding radiation therapy for breast cancer**. *Int J Women’s Health* (2014.0) **6** 945-951. DOI: 10.2147/IJWH.S71591
4. Nesvold IL, Fossa SD, Holm I, Naume B, Dahl AA. **Arm/shoulder problems in breast cancer survivors are associated with reduced health and poorer physical quality of life**. *Acta Oncol* (2010.0) **49** 347-353. DOI: 10.3109/02841860903302905
5. Neviaser AS, Hannafin JA. **Adhesive capsulitis: a review of current treatment**. *Am J Sports Med* (2010.0) **38** 2346-2356. DOI: 10.1177/0363546509348048
6. 6.Prestgaard TA (n.d.) Frozen shoulder (adhesive capsulitis). In: UpToDate [online]. Available at: https://www.uptodate.com/contents/frozen-shoulder-adhesive-capsulitis. Accessed 1 Nov 2017
7. Dias R, Cutts S, Massoud S. **Frozen shoulder**. *BMJ* (2005.0) **331** 1453-1456. DOI: 10.1136/bmj.331.7530.1453
8. Leung MS, Cheing GL. **Effects of deep and superficial heating in the management of frozen shoulder**. *J Rehabil Med* (2008.0) **40** 145-150. DOI: 10.2340/16501977-0146
9. Cheing GL, So EM, Chao CY. **Effectiveness of electroacupuncture and interferential eloctrotherapy in the management of frozen shoulder**. *J Rehabil Med* (2008.0) **40** 166-170. DOI: 10.2340/16501977-0142
10. Ma T, Kao MJ, Lin IH, Chiu YL, Chien C, Ho TJ. **A study on the clinical effects of physical therapy and acupuncture to treat spontaneous frozen shoulder**. *Am J Chin Med* (2006.0) **34** 759-775. DOI: 10.1142/S0192415X06004272
11. Stergioulas A. **Low-power laser treatment in patients with frozen shoulder: preliminary results**. *Photomed Laser Surg* (2008.0) **26** 99-105. DOI: 10.1089/pho.2007.2138
12. Kelley MJ, McClure PW, Leggin BG. **Frozen shoulder: evidence and a proposed model guiding rehabilitation**. *J Orthop Sports Phys Ther* (2009.0) **39** 135-148. DOI: 10.2519/jospt.2009.2916
13. Hanchard NCA, Goodchild LM, Thompson J, O'Brien T, Davison D, Richardson C. **Evidence-based clinical guidelines for the diagnosis, assessment and physiotherapy management of contracted (frozen) shoulder: quick reference summary**. *Physiotherapy* (2012.0) **98** 117-120. DOI: 10.1016/j.physio.2012.01.001
14. Santos WD, Gentil P, de Moraes RF, Júnior JB, Campos MH, Lira CA. **Chronic effects of resistance training in breast cancer survivors**. *Biomed Res Int* (2017.0) **2017** 8367803. DOI: 10.1155/2017/8367803
15. Mayer F, Scharhag-Rosenberger F, Carlsohn A, Cassel M, Muller S, Scharhag J. **The intensity and effects of strength training in the elderly**. *DeutschesArzteblatt Int* (2011.0) **108** 359
16. Cho J, Lee K, Kim M, Hahn J, Lee W. **The effects of double oscillation exercise combined with elastic band exercise on scapular stabilizing muscle strength and thickness in healthy young individuals: a randomized controlled pilot trial**. *J Sports Sci Med* (2018.0) **17** 7-16. PMID: 29535573
17. Lins C, Castro A, Medina GIS, Azevedo E, Donato BS, Chagas MSS. **Alternative scapular stabilization exercises to target strength, endurance and function of shoulders in tetraplegia: a prospective non-controlled intervention study**. *J Spinal Cord Med* (2019.0) **42** 65-76. DOI: 10.1080/10790268.2017.1398943
18. Yoo IG, Yoo WG. **The effect of a new neck support tying method using thera-band on cervical rom and shoulder muscle pain after overhead work**. *J Phys Ther Sci* (2013.0) **25** 843-844. DOI: 10.1589/jpts.25.843
19. BasKurt Z, BasKurt F, Gelecek N, Ozkan MH. **The effectiveness of scapular stabilization exercise in the patients with subacromial impingement syndrome**. *J Back Musculoskelet Rehabil* (2011.0) **24** 173-179. DOI: 10.3233/BMR-2011-0291
20. Cheville AL, Tchou J. **Barriers to rehabilitation following surgery for primary breast cancer**. *J Surg Oncol* (2007.0) **95** 409-418. DOI: 10.1002/jso.20782
21. Harishkumar S, Kiruthika S, Arunachalam R, Kumerasan A. **To analyse the effect of theraband strengthening with conventional exercise on pain, function & range of motion in patients with adhesive capsulitis**. *Int J Pharma Bio Sci* (2017.0) **8** 214-227
22. Rawat P, Eapen C, Seema KP. **Effect of rotator cuff strengthening as an adjunct to standard care in subjects with adhesive capsulitis: a randomized controlled trial**. *J Hand Ther* (2017.0) **30** 235-241. DOI: 10.1016/j.jht.2016.10.007
23. Datar NA, Devi TP. **Effect of graded thera-band exercises on shoulder muscle strength and activities of daily life in modified radical mastectomy subjects**. *Biomed Pharmacol J* (2019.0) **12** 1345-1351. DOI: 10.13005/bpj/1763
24. Kirthika SV, Bhavani PB, Rajlakshmi V. **Effect of combining scapular stabilization techniques with conventional physiotherapy in improving range of motion and functional ability in subjects with phase II adhesive capsulitis of the shoulder joint**. *J Physiother Occup Ther* (2015.0) **1** 25-34
25. Gulwani AH. **A study to find out the effect of scapular stabilization exercises on shoulder ROM and functional outcome in diabetic patients with stage 2 adhesive capsulitis of the shoulder joint: an interventional study**. *Int J Sci Healthc Res* (2020.0) **5** 320-333
26. Sabari JS, Maltzev I, Lubarsky D, Liszkay E, Homel P. **Goniometric assessment of shoulder range of motion: comparison of testing in supine and sitting positions**. *Arch Phys Med Rehabil* (1998.0) **79** 647-651. DOI: 10.1016/S0003-9993(98)90038-7
27. Riddle DL, Rothstein JM, Lamb RL. **Goniometric reliability in a clinical setting. Shoulder measurements**. *Phys Ther* (1987.0) **67** 668. DOI: 10.1093/ptj/67.5.668
28. Roy JS, MacDermid JC, Orton B, Tran T, Faber KJ, Drosdowech D, Athwal GS. **The concurrent validity of a hand-held versus a stationary dynamometer in testing isometric shoulder strength**. *J Hand Ther* (2009.0) **22** 320-327. DOI: 10.1016/j.jht.2009.04.008
29. Alotaibi NM, Aljadi SH, Alrowayeh HN. **Reliability, validity and responsiveness of the Arabic version of the disability of arm, shoulder and hand (DASH-Arabic)**. *Disabil Rehabil* (2016.0) **38** 2469-2478. DOI: 10.3109/09638288.2015.1136846
30. Boonstra AM, Preuper HRS, Reneman MF, Posthumus JB, Stewart RE. **Reliability and validity of the visual analogue scale for disability in patients with chronic musculoskeletal pain**. *Int J Rehabil Res* (2008.0) **31** 165-169. DOI: 10.1097/MRR.0b013e3282fc0f93
31. Guermazi M, Allouch C, Yahia M, Huissa TB, Ghorbel S, Damak J, Mrad MF, Elleuch MH. **Translation in Arabic, adaptation and validation of the SF-36 health survey for use in Tunisia**. *Ann Phys Rehabil Med* (2012.0) **55** 388-403. DOI: 10.1016/j.rehab.2012.05.003
32. Page P, Labbe A. **Adhesive capsulitis: use the evidence to integrate your interventions**. *N Am J Sports Phys Ther* (2010.0) **5** 266-273. PMID: 21655385
33. Çelik D, Mutlu EK. **Does adding mobilization to stretching improve outcomes for people with frozen shoulder? A randomized controlled clinical trial**. *Clin Rehabil* (2016.0) **30** 786-794. DOI: 10.1177/0269215515597294
34. Russell S, Jariwala A, Conlon R, Selfe J, Richards J, Walton M. **A blinded, randomized, controlled trial assessing conservative management strategies for frozen shoulder**. *J Shoulder Elbow Surg* (2014.0) **23** 500-507. DOI: 10.1016/j.jse.2013.12.026
35. Dempsey AL, Mills T, Karsch RM, Branch TP. **Maximizing total end range time is safe and effective for the conservative treatment of frozen shoulder patients**. *Am J Phys Med Rehabil* (2011.0) **90** 738-745. DOI: 10.1097/PHM.0b013e318214ed0d
36. Favejee MM, Huisstede BM, Koes BW. **Frozen shoulder: the effectiveness of conservative and surgical interventions–systematic review**. *Br J Sports Med* (2011.0) **45** 49-56. DOI: 10.1136/bjsm.2010.071431
37. Kim KH, Yeo SM, Cheong IY, Kim Y, Jeon BJ, Hwang JH. **Early rehabilitation after total mastectomy and immediate reconstruction with tissue expander insertion in breast cancer patients: a retrospective case-control study**. *J Breast Cancer* (2019.0) **22** 472-483. DOI: 10.4048/jbc.2019.22.e40
38. Leonidou A, Woods DA. **A preliminary study of manipulation under anaesthesia for secondary frozen shoulder following breast cancer treatment**. *Ann R Coll Surg Engl* (2014.0) **96** 111-115. DOI: 10.1308/003588414X13824511649652
39. Chan HBY, Pua PY, How CH. **Physical therapy in the management of frozen shoulder**. *Singapore Med J* (2017.0) **58** 685-689. DOI: 10.11622/smedj.2017107
40. Ruivo RM, Donatelli R, Parraca JA. **A specific multi-approach intervention for adhesive capsulitis: a case page 3 of 7 report**. *J Excer Sports Orthop* (2017.0) **4** 1-7
41. Ok SJ. **Effects of resistance exercise using elastic band on range of motion, function and shoulder pain among patients with rotator cuff repair**. *Korean J Adult Nursing* (2016.0) **28** 491-500. DOI: 10.7475/kjan.2016.28.5.491
42. Zhu P, Liao B, Wang Z, Sun Z, Yang W, Cai Y. **Resistance band training after triamcinolone acetonide injection for subacromial bursitis: a randomized clinical trial**. *J Rehabil Med* (2021.0) **53** jrm00140. DOI: 10.2340/16501977-2752
43. Hedayatpour N, Falla D. **Physiological and neural adaptations to eccentric exercise: mechanisms and considerations for training**. *Biomed Res Int* (2015.0) **2015** 193741. DOI: 10.1155/2015/193741
44. Shamley D, Srinaganathan R, Oskrochi R, Lascurain-Aguirrebeña I, Sugden E. **Three-dimensional scapulothoracic motion following treatment for breast cancer**. *Breast Cancer Res Treat* (2009.0) **118** 315-322. DOI: 10.1007/s10549-008-0240-x
45. Crosbie J, Kilbreath SL, Dylke E, Refshauge KM, Nicholson LL, Beith JM. **Effects of mastectomy on shoulder and spinal kinematics during bilateral upper-limb movement**. *Phys Ther* (2010.0) **90** 679-691. DOI: 10.2522/ptj.20090104
46. Yatheendra kG, Sudhakar S, yashvanth A, Siva JN. **Effect of high-grade mobilisation techniques and scapular stabilization exercises in frozen shoulder**. *Int J Phys Educ Sports Health* (2015.0) **2** 80-83
47. Voight ML, Thomson BC. **The role of the scapula in the rehabilitation of shoulder injuries**. *J Athl Train* (2000.0) **35** 364-437. PMID: 16558649
48. Song MJ, Kang TW. **The effect of a four-week scapular stabilization exercise program using PNF technique on scapular symmetry and range of flexion motion, pain, function, and quality of life in post-mastectomy women with breast cancer**. *PNF Mov* (2021.0) **19** 19-29
49. Nam S, Kang T. **Effect of scapular stabilization exercise on shoulder joint range of motion, pain and functional level in women who underwent breast cancer resection**. *Korean J Orthop Phys Ther* (2017.0) **23** 69-74
50. Serra-Añó P, Pellicer-Chenoll M, García-Massó X, Morales J, Giner-Pascual M, González LM. **Effects of resistance training on strength, pain and shoulder functionality in paraplegics**. *Spinal Cord* (2012.0) **50** 827-831. DOI: 10.1038/sc.2012.32
|
---
title: 'Real-world patient characteristics and use of disease-modifying anti-rheumatic
drugs in patients with rheumatoid arthritis: a cross-national study'
authors:
- Ylenia Ingrasciotta
- Yinzhu Jin
- Saveria S. Foti
- Joan E. Landon
- Michele Tari
- Francesco Mattace-Raso
- Seoyoung C. Kim
- Gianluca Trifirò
journal: Clinical Rheumatology
year: 2022
pmcid: PMC10017582
doi: 10.1007/s10067-022-06478-4
license: CC BY 4.0
---
# Real-world patient characteristics and use of disease-modifying anti-rheumatic drugs in patients with rheumatoid arthritis: a cross-national study
## Abstract
### Introduction
Rheumatoid arthritis (RA) is associated with significant morbidity and economic burden. This study aimed to compare baseline characteristics and patterns of anti-inflammatory drug use and disease-modifying anti-rheumatic drug (DMARD) use among patients with RA in Southern *Italy versus* the United States.
### Method
Using Caserta Local Health Unit (Italy) and Optum’s de-identified Clinformatics® Data Mart (United States) claims databases, patients with ≥ 2 diagnosis codes for RA during the study period (Caserta: 2010–2018; Optum: 2010–2019) were identified. Baseline patient characteristics, as well as proportion of RA patients untreated/treated with NSAIDs/glucocorticoids/conventional DMARDs (csDMARDs)/biological/targeted synthetic DMARDs (b/tsDMARDs) during the first year of follow-up, and the proportion of RA patients with ≥ 1 switch/add-on between the first and the second year of follow-up, were calculated. These analyses were then stratified by age group (< 65; ≥ 65).
### Results
A total of 9227 RA patients from Caserta and 195,951 from Optum databases were identified (two-thirds were females). During the first year of follow-up, $45.9\%$ RA patients from *Optum versus* $79.9\%$ from Caserta were exclusively treated with NSAIDs/glucocorticoids; $17.2\%$ versus $11.3\%$ from Optum and Caserta, respectively, were treated with csDMARDs, mostly methotrexate or hydroxychloroquine in both cohorts. Compared to $0.6\%$ of RA patients from Caserta, $3.2\%$ of the Optum cohort received ≥ 1 b/tsDMARD dispensing. Moreover, 61,655 ($33.7\%$) patients from Optum cohort remained untreated compared to 748 ($8.3\%$) patients from the Caserta cohort. The subgroup analyses stratified by age showed that 42,989 ($39.8\%$) of elderly RA patients were untreated compared to 18,666 ($24.9\%$) young adult RA patients in Optum during the first year of follow-up. Moreover, a higher proportion of young adult RA patients was treated with b/tsDMARDs, with and without csDMARDs, compared to elderly RA patients (Optum<65: $6.4\%$; Optum≥65: $1.0\%$; P-value < 0.001; Caserta<65: $0.8\%$; Caserta≥65: $0.1\%$; P-value < 0.001). Among RA patients untreated during the first year after ID, $41.2\%$ and $48.4\%$ RA patients from Caserta and Optum, respectively, received NSAIDs, glucocorticoids, and cs/b/tsDMARDs within the second year of follow-up. Stratifying the analysis by age groups, $50.6\%$ of untreated young RA patients received study drug dispensing within the second year of follow-up, compared to only $36.7\%$ of elderly RA patients in Optum. Interestingly, more young adult RA patients treated with csDMARDs during the first year after ID received a therapy escalation to b/tsDMARD within the second year after ID in both cohorts, compared to elderly RA patients (Optum<65: $7.8\%$; Optum≥65: $1.8\%$; Caserta<65: $3.2\%$; Caserta≥65: $0.6\%$).
### Conclusions
Most of RA patients, with heterogeneous baseline characteristics in Optum and Caserta cohorts, were treated with anti-inflammatory/csDMARDs rather than bDMARDs/tsDMARDs during the first year post-diagnosis, especially in elderly RA patients, suggesting a need for better understanding and dealing with barriers in the use of these agents for RA patients.
Key Points• Substantial heterogeneity in baseline characteristics and access to bDMARD or tsDMARD drugs between RA patients from the United States and Italy exists.• Most of RA patients seem to be treated with anti-inflammatory/csDMARD drugs rather than bDMARD/tsDMARD drugs during the first year post-diagnosis.• RA treatment escalation is less frequent in old RA patients than in young adult RA patients.• An appropriate use of DMARDs should be considered to achieve RA disease remission or low disease activity.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s10067-022-06478-4.
## Introduction
Rheumatoid arthritis (RA) is a chronic systemic inflammatory disease that affects the joints, connective tissues, muscle, tendons, and fibrous tissue and is associated with significant morbidity and economic burden [1–4]. The estimated prevalence of RA worldwide varies between 0.3 and $1\%$ and is more common in women and in developed countries [5]. In the United States (US), RA affects approximately 1.3 million adults [6, 7]. In Italy, the RA prevalence is 0.3–$0.7\%$, confirming a higher prevalence in women than in men [8]. RA commonly affects patients aged 30–50 years old [9], and in patients aged above 60, the prevalence is equal to $2\%$ [10].
Elderly RA patients present frequently comorbidity such as cognitive impairment, depression, and frailty [11]. High incidence of comorbidities and drug-related adverse effects in elderly patients also raise therapeutic challenges for the disease management and to achieve a clinical remission of the disease [12, 13].
Evidence from the literature indicates that, despite available treatments, several unmet needs still exist with regard to RA management [14, 15]. Patients with RA experience substantial levels of pain and are not satisfied with their levels of physical functioning even with ongoing treatment [16]. Currently, the main therapeutic target for RA patients is achieving clinical remission, with low disease activity as the best possible alternative [17], to prevent functional impairment and disability [18, 19]. According to national and international guidelines and recommendations [17, 20, 21], several treatments for RA are available: glucocorticoids or non-steroidal anti-inflammatory drugs (NSAIDs), conventional disease-modifying anti-rheumatic drugs (csDMARDs), targeted synthetic DMARDs (tsDMARDs), and biological DMARDs (bDMARDs).
According to the disease severity, the use of these agents aims at controlling systemic inflammation to slow or prevent the disease progression. Methotrexate is considered the standard of care for RA; in patients with at least one contraindication such as severe hepatic or renal impairment, serious, acute, or chronic infections, and other contraindications or intolerance to methotrexate, leflunomide, or sulfasalazine could be considered as options in the first-line strategy of treatment. Moreover, if the treatment target is not achieved with the first csDMARD strategy, addition/switch to a tsDMARD or a bDMARD is recommended [17, 20, 21].
Over the past 20 years, the management of RA has radically changed. The choice of therapies, which were previously mostly based on csDMARDs, has expanded with the marketing of bDMARDs and, more recently, with the new class of tsDMARDs [22]. In particular, the introduction of bDMARDs has revolutionized treatments for RA, with a substantial positive effect on the quality of care of RA patients who suffer from moderate-to-severe disease or who have failed to improve with other medications [23]. However, due to the high cost of these drugs, heterogeneity in access to bDMARDs in RA patients across Europe has been observed [24].
In 2013, the European Medicine Agency (EMA) approved the first infliximab biosimilar, while the Food and Drug Administration (FDA) did in 2016. *In* general, biosimilars provide a 20–$30\%$ purchase cost reduction in comparison to the reference product, representing a valid cost containing strategy [25]; although health resources are limited, it is widely shared that innovative medicines should be made available to all citizens; as new biologic drugs are expensive, correct management strategies must be implemented.
Although the use of biologics has revolutionized the RA therapeutic landscape, leading to major changes in therapeutic targets, concerns about decreased efficacy due to immune senescence and a low benefit-risk profile in the elderly have led to a relative underutilization of biologics [26]. A rapidly ageing population and increasing rates of RA make the paucity of data in older adults with RA an increasingly important clinical issue.
Moreover, since the efficacy and safety of b/tsDMARDs have been thoroughly investigated in randomized clinical trials (RCTs) [27], real-world studies exploring the pattern of use of RA treatments in routine rheumatology practice considering unselected patients potentially representing the entire spectrum of disease severity are needed. The main objective of this study is to evaluate and compare the baseline characteristics and the pattern of real-world use of drugs (e.g., anti-inflammatory drugs and DMARDs) for the treatment of RA in Southern *Italy versus* the United States. The second aim of this study is to compare the pattern of real-world use of drugs for the treatment of RA young adult versus elderly RA patients in both countries.
## Data sources
This is a retrospective, cross-national cohort study. Data were extracted from Caserta Local Health Unit (LHU)-Italy and Optum’s de-identified Clinformatics® Data Mart-United States claims databases (DBs), covering 1.1 million and 53.3 million individuals, respectively, from January 2010 to September 2019 (Caserta: Jan 2010–Dec 2018). In particular, collected *Italian data* included demographics, outpatient pharmacy, hospital discharge database, requests for outpatient diagnostic tests and specialist’s visits, exemptions from healthcare service co-payment, and emergency department visit databases. All databases can be linked through an anonymous subject identifier. In addition, general practitioner’s prescriptions (from Arianna database) with related indication for use as well as electronic therapeutic plans (filled by the specialist and including information on drug prescribed, indication for use, drug dosages, and therapy duration) and results of diagnostic tests are collected in Caserta database. The Caserta LHU claims and General Practitioner Arianna databases have been shown to provide accurate and reliable information for pharmacoepidemiological research, as documented elsewhere [28–32]. In Caserta LHU DB, drug dispensing is coded using the Anatomical Therapeutic Chemical (ATC) classification system or specific Italian market authorization code (AIC), while indications for use and causes of hospitalizations are coded using the International Classification of Diseases, 9th revision, clinical modification (ICD9-CM). In Optum DB, drug dispensing is coded using generic names or J/Q codes if applicable, while indications for use and causes of hospitalizations are coded using ICD9-CM or ICD-10 codes.
Moreover, in Italy, biological drugs are fully reimbursed by the National Health Service (NHS) and for each biologic drug prescription, specialists have to fill a therapeutic plan, which indicates the exact drug name, number of dispensed packages, dosing regimen, and indication for use. Electronic therapeutic plans were available in the Caserta LHUs. These data can be linked through unique and anonymous patient identifiers to other claims databases, which contain several types of information, including causes of hospitalization and reasons for healthcare service co-payment exemptions.
Optum Clinformatics® Data Mart (CDM) is derived from a database of de-identified administrative health claims for members of large commercial and Medicare Advantage health plans. The database includes approximately 17–19 million annual covered lives, for a total of over 62 million unique lives over a 13-year period ($\frac{1}{2007}$ through $\frac{12}{2020}$). Clinformatics® Data *Mart is* statistically de-identified under the Expert Determination method consistent with HIPAA and managed according to Optum® customer data use agreements. CDM administrative claims submitted for payment by providers and pharmacies are verified, adjudicated, and de-identified prior to inclusion. This data, including patient-level enrollment information, is derived from claims submitted for all medical and pharmacy healthcare services with information related to healthcare costs and resource utilization. The population is geographically diverse, spanning all 50 states. Optum de-identified CDM contains longitudinal information on medical and pharmacy claims from a number of different managed care plans, including hospitalizations, outpatient visits, procedures, and pharmacy dispensing. All the medical/pharmacy claims through Optum insurance are recorded in the database as long as the patients were still enrolled in the insurance. As reported for Caserta LHU claims and General Practitioner Arianna databases, Optum Clinformatics® Data Mart has been shown to provide accurate and reliable information for pharmacoepidemiological research, as documented elsewhere [33–36].
## Study population
All patients aged ≥ 18 years with at least two RA diagnoses separated by ≥ 7 days but < 365 days were eligible for the study cohort. The date of the second RA diagnosis was defined as the index date (ID), and patients were required to have at least 1-year pre- and post-index continuous enrollment in their databases to ensure comprehensive availability of data on their healthcare use over this period [36–38]. In the Optum database, RA diagnoses were identified based on RA ICD-9 codes (714.xx) or ICD-10 codes (M05.xx, M06.xx, M08.xx, M12.xx) from inpatient or outpatient medical claims. In the Caserta database, RA diagnoses were identified based on RA ICD-9 codes (714.xx) from discharge diagnosis or emergency department visits or electronic therapeutic plans or from the General Practitioner database (i.e., Arianna database) which can be linked through anonymous subject identifier with claims databases. All patients with any csDMARD, bDMARD, or tsDMARD dispensing any time prior to the first RA diagnosis date were excluded. The identification criteria for the study cohort are shown in Online Resource 1.
## Exposure assessment
All the following drug classes were included: anti-inflammatory drugs (e.g., NSAIDs and glucocorticoids), csDMARDs (e.g., methotrexate, sulfasalazine, leflunomide, chloroquine, hydroxychloroquine, cyclosporine, azathioprine, auranofin, and sodium aurotiosulfate), bDMARDs, both originators and biosimilars (e.g., etanercept, adalimumab, infliximab, certolizumab pegol, golimumab, anakinra, abatacept, sarilumab, tocilizumab, and rituximab), and tsDMARDs (e.g., tofacitinib and baricitinib). Upadacitinib was not included because it was approved by EMA and by FDA in 2019. Online Resource 2 shows all the included drugs for this study.
## Data analysis
In each cohort, the following baseline patient characteristics were assessed: sex, age (categorized as follows: 18–44, 45–64, 65–79, ≥ 80, mean ± standard deviation) at ID, index year, geographic area of patients, comorbidities (e.g., hypertension, diabetes mellitus, chronic pulmonary disease, lipid metabolism disorders, chronic renal failure, liver disease, heart failure, ischemic heart disease, malignancy, smoking, obesity, psoriasis, and inflammatory bowel diseases) evaluated within 1 year prior to ID, number of unique prescription drugs based on generic names (categorized as 0, 1, 2, 3–5, 6–10, > 10) evaluated within 1 year prior to ID, and concomitant drugs (e.g., traditional NSAIDs, COX-2 inhibitors, opioids, antidepressant drugs, antihypertensive drugs, insulin and oral hypoglycemic agents, and lipid lowering agents) evaluated within 1 year prior to ID.
The proportion of RA patients treated or untreated within 1 year after ID in each cohort was calculated. Patients were categorized as follows:Untreated patients: patients without any study drug dispensing;Exclusive NSAID users: patients with at least one NSAID dispensing AND no dispensing of oral/parenteral glucocorticoids/bDMARD/csDMARD/tsDMARD;Glucocorticoid (± NSAID) users: patients with at least one oral/parenteral glucocorticoid dispensing AND no dispensing of csDMARD/bDMARD/tsDMARD;csDMARD (± glucocorticoid ± NSAID) users: patients with at least one csDMARD dispensing AND no dispensing of bDMARD/tsDMARD; orbDMARD/tsDMARD (± NSAID ± glucocorticoid ± csDMARD) users: patients with at least one bDMARD or tsDMARD dispensing.
Moreover, the proportion of each treatment type among RA patients, after excluding those who were never treated during the follow-up, was calculated. This analysis was then stratified by active substance, distinguishing between originator and biosimilar bDMARDs. Moreover, the proportion of RA patients with at least one switch/add-on between the first and the second year post-ID was calculated. Only RA patients with at least 2 years post-index continuous enrollment in the database were included.
## Subgroup analysis
Subgroup analyses of the proportion of RA patients untreated or treated within 1 year after ID in each cohort and of the proportion of RA patients with at least one switch/add-on between the first and the second year post-ID were conducted according to age (< 65; ≥ 65).
## Statistical analysis
Descriptive statistics were used for the aforementioned baseline variables. For comparisons between the two cohorts, a standardized mean difference (SMD) greater than 0.1 was considered as a sign of imbalance [39]. Statistical analyses were performed using SAS 9.2 (SAS Institute, Cary, NC, USA).
## Results
During the study period, 195,951 and 9227 subjects with a diagnosis of RA were identified from Optum and Caserta databases, respectively (Fig. 1). RA prevalence was higher in Caserta ($1.1\%$) than in Optum ($0.6\%$). Of these, more than two-thirds were female patients in both cohorts [Optum: $$n = 133$$,605 ($68.2\%$); Caserta: $$n = 6117$$ ($66.3\%$); SMD = 0.0408]. RA patients from Optum were older than those from Caserta (mean age ± SD: 66.8 ± 14.2 years in Optum vs. 57.1 ± 16.1 years in Caserta; SMD = 0.6788) (Table 1). In particular, 119,026 ($60.7\%$) and 3203 ($34.7\%$) RA patients were aged 65 years or over, in Optum and Caserta, respectively. Fig. 1Flow chart of the study cohort. Legend: LHU, Local Health Unit; csDMARDs, conventional disease-modifying anti-rheumatic drugs; bDMARDs, biological disease-modifying anti-rheumatic drugs; tsDMARDs, targeted synthetic disease-modifying anti-rheumatic drugs. * Data available until December 2018. °Data available until September 2019. Patients: (a) age ≥ 18 years; (b) ≥ 2 diagnoses of RA, separated by ≥ 7 days but < 365 days; (c) ≥ 1 year pre-index and 1-year post-index date continuous enrollment in their databasesTable 1Baseline characteristics of the study cohortOptumN = 195,951CasertaN = 9227SMD (d)Sex — N (%) Male62,346 (31.8)3110 (33.7)0.0408 Female133,605 (68.2)6117 (66.3) Mean age ± SD — year66.8 ± 14.257.1 ± 16.10.6788Age — N (%) 18–4418,377 (9.4)2184 (23.7)0.3921 45–6458,538 (29.9)3840 (41.6)0.2459 65–7984,433 (43.1)2506 (27.2)0.3377 ≥ 8034,593 (17.6)697 (7.5)0.3084Geographic area of patients — N (%) Northeast27,167 (13.9)-- South87,936 (44.9) Midwest39,045 (19.9) West41,803 (21.3)Index year — N (%) 201014,449 (7.4)489 (5.3)0.0861 201113,330 (6.8)1338 (14.5)0.2515 201213,330 (6.8)1623 (17.6)0.3345 201312,955 (6.6)1589 (17.2)0.3318 201412,518 (6.4)1349 (14.6)0.2699 201518,515 (9.4)899 (9.7)0.0102 201636,639 (18.7)1087 (11.8)0.1928 201741,071 (21.0)853 (9.3)0.3307 201833,144 (16.9)--Comorbidities — N (%)a Hypertension131,949 (67.3)4350 (47.1)0.4293 Diabetes mellitus56,861 (29.0)1117 (12.1)0.3589 Chronic pulmonary disease45,651 (23.3)1886 (20.4)0.0687 Hyperlipidemia115,589 (59.0)1656 (17.9)0.8431 Chronic renal failure17,921 (9.1)144 (1.6)0.2657 Liver diseases13,433 (6.9)126 (1.4)0.2209 Heart failure20,334 (10.4)192 (2.1)0.2768 Ischemic heart disease39,307 (20.1)938 (10.2)0.2495 Malignancy16,072 (8.2)455 (4.9)0.1213 Smoking35,639 (18.2)778 (8.4)0.2568 Obesity31,445 (16.0)199 (2.2)0.3837 Inflammatory bowel disease2452 (1.2)214 (2.3)0.099 Psoriasis3745 (1.9)181 (2.0)0.0073Previous use of any medications — mean ± SD9.7 ± 7.69.8 ± 6.30.0133Previous use of any medications — N (%)a 025,224 (12.9)53 (0.6)0.5056 16521 (3.3)166 (1.8)0.0952 26261 (3.2)404 (4.3)0.0579 3–525,862 (13.1)1907 (20.7)0.2038 6–1053,085 (27.1)3135 (34.0)0.1502 > 1078,998 (40.3)3562 (38.6)0.0347Concomitant drugs — N (%)a Traditional NSAIDs79,690 (40.7)7531 (81.6)0.8397 COX-2 inhibitors10,318 (5.3)2242 (24.3)0.8014 Opioids98,619 (50.3)1312 (14.2)0.7305 Antidepressant drugs49,347 (25.2)1205 (13.1)0.2812 Antihypertensives135,834 (69.3)4741 (51.4)0.3866 Insulin and oral hypoglycemic agents33,122 (16.9)1168 (12.7)0.1126 Lipid lowering agents73,127 (37.3)2197 (23.8)0.2806Legend: SMD, standardized mean difference; SD, standard deviation; NSAIDs, non-steroidal anti-inflammatory drugs. aEvaluated within 1 year prior to ID *In* general, compared to the Caserta cohort, a higher proportion of the Optum cohort had comorbidities at baseline ($80.0\%$ vs. $63.2\%$). Specifically, hypertension [Optum: $$n = 131$$,949 ($66.3\%$); Caserta: $$n = 4350$$ ($47.1\%$); SMD = 0.4294] and hyperlipidemia [Optum: $$n = 115$$,589 ($59.0\%$); Caserta: $$n = 1656$$ ($17.9\%$); SMD = 0.8431] were the two most common comorbidities in both cohorts. In both cohorts, less than $2\%$ of RA patients had other autoimmune disorders for which bDMARDs might be indicated (e.g., inflammatory bowel diseases and psoriasis). Interestingly, $40.2\%$ of patients from both cohorts had received more than 10 drugs during the 1-year period prior to the ID. Half of RA patients from Optum had received at least one dispensing for opioids, compared to $14\%$ of RA patients from Caserta (SMD = 0.7305). Contrarily, 7531 ($81.6\%$) and 2242 ($24.3\%$) in the Caserta cohort had received traditional NSAIDs and COX-2 inhibitors, respectively, versus 79,690 ($40.7\%$) and 10,318 ($5.3\%$) in the Optum cohort (SMDtraditional NSAIDs = 0.8397; SMDCOX-2 inhibitors = 0.8014).
## DMARD treatment patterns
During the first year of follow-up, one-third ($$n = 61$$,655; $33.7\%$) of RA patients from Optum were untreated with NSAIDs, glucocorticoids, or any DMARDs, compared to 748 ($8.3\%$) RA patients from Caserta (P-value < 0.001). Among treated patients, almost half (84,036; $45.9\%$) of RA patients from *Optum versus* more than two-thirds ($$n = 7199$$; $79.9\%$) from Caserta received NSAIDs/glucocorticoids dispensing (P-value < 0.001), but they did not receive specific RA treatments (e.g., csDMARDs, bDMARDs, or tsDMARDs); $17.2\%$ of patients from *Optum versus* $11.3\%$ of patients from Caserta were treated with csDMARDs (P-value < 0.001) (Fig. 2), mostly methotrexate or hydroxychloroquine in both cohorts. No sodium aurothiosulfate users were identified in both cohorts (Online Resource 3). Compared to $3.2\%$ of RA patients from Optum, only $0.6\%$ of RA patients from Caserta had at least one bDMARD/tsDMARD dispensing, with and without csDMARDs (P-value < 0.001) (Fig. 2). The most frequently used bDMARD was the adalimumab originator (Optum: $1.4\%$; Caserta: $0.2\%$; P-value < 0.001), followed by the etanercept originator (Optum: $1.1\%$; Caserta: $0.1\%$; P-value < 0.001). In both cohorts, no patients used anakinra, adalimumab biosimilars, or e rituximab biosimilars; no users of sarilumab were identified in Caserta (Online Resource 3). We found no tsDMARD users in *Caserta versus* 226 tsDMARD users (224 tofacitinib and 2 baricitinib) in Optum. Fig. 2Frequency (%) of treatment lines within the first year after ID. Legend: DMARD, disease-modifying anti-rheumatic drug; csDMARD, conventional synthetic disease-modifying anti-rheumatic drug; tsDMARD, targeted synthetic disease-modifying anti-rheumatic drug; bDMARD, biological disease-modifying anti-rheumatic drug The subgroup analysis stratified by age showed that 42,989 ($39.8\%$) of elderly RA patients were untreated compared to 18,666 ($24.9\%$) young adult RA patients in Optum (P-value < 0.001) (Fig. 3). Specifically, 14,851 ($13.7\%$) elderly RA patients versus 16,553 ($22.1\%$) young adult RA patients from Optum received csDMARDs during the first year after ID (P-value < 0.001). Concerning the use of csDMARDs from the Caserta cohort, no statistically significant differences were observed in the two age groups compared. Regarding the use of bDMARDs/tsDMARDs, a higher proportion of young adult RA patients was treated with bDMARDs/tsDMARDs, with and without csDMARDs, compared to elderly RA patients (Optum < 65: $6.4\%$; Optum ≥ 65: $1.0\%$; P-value < 0.001; Caserta < 65: $0.8\%$; Caserta ≥ 65: $0.1\%$; P-value < 0.001).Fig. 3Frequency (%) of treatment lines within the first year after ID, stratified by age group. Legend: DMARD, disease-modifying anti-rheumatic drug; csDMARD, conventional synthetic disease-modifying anti-rheumatic drug; tsDMARD, targeted synthetic disease-modifying anti-rheumatic drug; bDMARD, biological disease-modifying anti-rheumatic drug Among untreated RA patients during the first year after ID, $41.2\%$ from Optum and $48.4\%$ from Caserta received at least one study drugs dispensing within the second year of follow-up (P-value < 0.001) (Fig. 4).Fig. 4Proportion (%) of RA patients with at least one switch/add-on between the first and the second year after ID. Legend: DMARD, disease-modifying anti-rheumatic drug; csDMARD, conventional synthetic disease-modifying anti-rheumatic drug; tsDMARD, targeted synthetic disease-modifying anti-rheumatic drug; bDMARD, biological disease-modifying anti-rheumatic drug *In* general, almost two-thirds ($63.3\%$) of US elderly RA patients versus $49.4\%$ of young adult RA patients continued to be untreated between the first and the second year after ID (P-value < 0.001).
Stratifying the analysis by age groups, more than half ($50.6\%$) of untreated young RA patients during the first year after ID received study drug dispensing within the second year of follow-up, compared to only $36.7\%$ of elderly RA patients in Optum (P-value < 0.001). Among untreated patients from Caserta, no statistically significant differences were observed in the two compared age groups (P-value: 0.689) (Fig. 5). Interestingly, more young adult RA patients treated with csDMARDs during the first year after ID received a therapy escalation to b/tsDMARD within the second year after ID in both cohorts, compared to elderly RA patients (Optum < 65: $7.8\%$; Optum ≥ 65: $1.8\%$; P-value < 0.001; Caserta < 65: $3.2\%$; Caserta ≥ 65: $0.6\%$; P-value: 0.012).Fig. 5Proportion (%) of RA patients with at least one switch/add-on between the first and the second year after ID, stratified by age. Legend: DMARD, disease-modifying anti-rheumatic drug; csDMARD, conventional synthetic disease-modifying anti-rheumatic drug; tsDMARD, targeted synthetic disease-modifying anti-rheumatic drug; bDMARD, biological disease-modifying anti-rheumatic drug
## Discussion
This large retrospective cross-national population-based cohort study investigated the baseline characteristics and the pattern of use of different pharmacological treatment lines (anti-inflammatory drugs, csDMARDs, bDMARDs, and tsDMARDs) in patients with RA from the US and Italy over the 10-year study period. Our data about RA prevalence suggest that it was higher in Caserta than in Optum, but in line with prevalence reported in literature [5, 7, 8]. As expected, the distribution by sex showed a female/male ratio equal to 2:1 in both cohorts. *In* general, a higher proportion of RA patients from Optum had comorbidities at baseline, and they were older than RA patients from Caserta. Specifically, hypertension and hyperlipidemia were the two most common comorbidities, followed by obstructive pulmonary disease, in both cohorts. This is in line with a prospective Swedish study [40] as well as a cohort study using a commercial and Medicare claims database with national beneficiaries [36], showing that $47.1\%$ and $39.3\%$ of RA patients had history of hypertension, followed by $31.9\%$ patients with chronic obstructive pulmonary disease.
In both our cohorts, less than $2\%$ of RA patients had history of other autoimmune disorders for which bDMARDs might be indicated (e.g., inflammatory bowel diseases and psoriasis), as reported by Jin et al. [ 36]. This is also due by the exclusion of all RA patients with at least one csDMARD, bDMARD, or tsDMARD dispensing any time prior to the first RA diagnosis date.
On average, RA patients from both cohorts had received more than 10 drugs within 1 year prior to the ID. Half of RA patients from Optum had received at least one dispensing for opioids, compared to $14\%$ of RA patients from Caserta. It is known that abuse of opioids for the treatment of chronic pain is very common in the US. Recent years have seen an “opioid crisis” take place in the US, with widespread overuse and misuse of opioids, leading to a large number of overdose-related deaths [30, 41]. Zamora-Legoff et al., in a population-based study including RA patients from the Rochester Epidemiology Project (REP), a special record-linkage system that records all inpatient and outpatient encounters among the residents of Olmsted County, Minnesota, showed that over a third of RA patients used opioids, and in more than a tenth, the use was chronic [42]. Contrarily, our findings showed a higher use of traditional NSAIDs and COX-2 inhibitors at baseline among RA patients from Caserta than those in the US. The highest use of NSAIDs in Italy was confirmed by an Italian population-based study evaluating the clinical characteristics of elderly analgesic users in Caserta LHU and the frequency of potentially inappropriate analgesic use [30]. The study showed that, among 94,820 elderly persons receiving at least one analgesic drug, $36.6\%$ were incident NSAID users, while $13.2\%$ were incident weak opioid users and $8.1\%$ were incident strong opioid users. Specifically, $9.2\%$ of all elderly analgesic users were considered to have an inappropriate prescription for the NSAIDs (ketorolac or indomethacin) [30].
During the first year of follow-up, one-third of RA patients from Optum seem to be untreated with either NSAIDs, glucocorticoids, or any DMARDs, compared to $8\%$ of RA patients from Caserta. Specifically, almost $40\%$ of US elderly RA patients were untreated compared to $25\%$ of US young adult RA patients during the first year after ID, while no statistically significant differences were observed in the two age groups compared in the Caserta cohort. Moreover, our results showed that, overall, among untreated RA patients, almost half of patients from both study cohorts received at least one study drug dispensing within the second year of follow-up; however, almost two-thirds of US elderly RA patients versus half of young adult RA patients continued to be untreated between the first and the second year after ID. This is in line with a previous study, showing that more than $50\%$ of adults aged 45 years or older with some forms of arthritis remain untreated, despite many of them experiencing severe symptoms and poor physical function [43]. Nevertheless, an exploratory analysis showed that the proportion of untreated RA patients decreased to $6\%$ in Optum and $2\%$ in Caserta within 3 years after ID (data not shown). Regarding those treated, almost half of RA patients from *Optum versus* more than two-thirds of RA patients from Caserta received NSAIDs/glucocorticoids dispensing, but they did not receive RA-specific DMARD treatments. Among csDMARDs, mostly methotrexate and hydroxychloroquine were used in both cohorts. This is in line with national and international guidelines and recommendations [17, 20, 21]. Methotrexate remains the mainstay 1st-line DMARD in RA; not only is it an efficacious csDMARD by itself but it is also the basis for combination therapies, either with glucocorticoids or with other csDMARDs, bDMARDs, or tsDMARDs. The European Alliance of Associations for Rheumatology (EULAR) guidelines recommend that in patients with a contraindication to methotrexate (or early intolerance), leflunomide or sulfasalazine should be considered as part of the (first) treatment strategy [17, 20]. However, our results showed a low use of leflunomide and sulfasalazine in both countries, compared to hydroxychloroquine. However, EULAR guidelines state that antimalarials, and especially hydroxychloroquine, have a limited role, mainly reserved for patients with mild RA [17] given the only weak clinical and no structural efficacy of hydroxychloroquine [44].
According to the guidelines, bDMARDs/tsDMARDs represent a 2nd line of therapy usually reserved for patients who have failed or have contraindications to csDMARDs [17, 20, 21]. Although RA treatment has made major advances over the past few decades, especially with the introduction of biologics as a treatment option for RA patients, most of the patients in our study were found to be initially treated with anti-inflammatory drugs or csDMARDs rather than bDMARDs. This may be due to the patients in the study having had less severe RA or a state of low disease activity that warranted no treatment with biologic agents. It could also be that patients may still have been kept on csDMARDs despite not achieving remission or low disease activity as recommended in the RA guidelines [17, 20]. Given that claims databases do not collect clinical data on effectiveness or disease activity, we were not able to evaluate these hypotheses.
However, our results are confirmed by an Italian retrospective observational study using claims databases from Veneto, Marche, Abruzzo, Apulia, and Calabria Regions [45]. The mentioned study showed that, as a first treatment, $5\%$ of RA patient received bDMARDs versus $52\%$ were not treated with DMARDs and received no treatment at all or only NSAIDs/glucocorticoids versus $43\%$ of RA patients receiving csDMARDs ($83\%$ of csDMARD users continued with the same category of DMARDs during the follow-up).
Similar evidence from the US showed that only $2.6\%$ of RA patients initiated b/tsDMARD treatment within 1 year of diagnosis [46], confirming the low use of bDMARDs/tsDMARDs in our two cohorts, especially in elderly patients from US. A recent retrospective, cohort study using the US Corrona RA registry showed that $54\%$ of RA patients with persistent moderate-to-high disease activity after 6 months of treatment with a csDMARD drug did not receive their therapy escalation. Of the patients who completed a visit at 3–9 months after the index date, treatment advancement occurred in $29\%$ of the patients, with $71\%$ having no change. Dose escalation of the csDMARD, initiation of another csDMARD, and initiation of a bDMARD occurred in $13\%$, $8\%$, and $10\%$ of patients of the total population [47].
Our results showed that treatment escalation was less frequent in old RA patients than in young adult RA patients. Different studies have suggested that old RA patients may be less aggressively treated than they should be [10, 26, 48, 49]. The Ruban study reported that despite higher disease activity at diagnosis, elderly-onset RA (EORA) patients were less likely to receive combination DMARD therapies or biologic agents compared with young-onset RA (YORA) patients, even though these drugs (biologics in particular) have been shown to have similar efficacy in older and younger individuals [49]. Howard et al. showed that time to first biologic DMARD is strongly associated with age. The ≥ 75 s were more likely to be on less intensive therapies compared to the < 65 s (csDMARD monotherapy or steroid alone, versus csDMARD combination therapy or bDMARD).
This may in part be due to access, as public payers take longer than private payers to recognize criteria for use and issue approval of advanced therapeutic agents. Indeed, the access to bDMARDs/tsDMARDs still represents an insight. In Italy, although bDMARDs/tsDMARDs are fully reimbursed by the NHS, the access barrier is due to the guidelines, which recommend these high-cost treatments if the treatment target is not achieved with the csDMARD strategy. On the contrary, in the US, the access barrier to these high-cost treatments could be explained by the high median out-of-pocket cost (e.g., $ 40 for bDMARDs and $ 50 for tsDMARDs).
Our study showed that the most frequently used bDMARD was the adalimumab originator, followed by the etanercept originator. A very low proportion of RA patients received infliximab biosimilars, while no users of adalimumab biosimilar and rituximab biosimilar in both cohorts were identified. The first reimbursement approval by the Italian NHS was in July 2017 for rituximab biosimilar and August 2018 for adalimumab biosimilar. Concerning rituximab biosimilar dispensing, it may not be traced in Caserta DB because it was rarely used by Caserta LHU hospitals. Adalimumab biosimilar dispensing may not be traced in the Caserta database because the mean/median times lag between the Italian Drug Agency (AIFA) and Campania Drug Formulary Committee approval could reach some months. In the US, even though five adalimumab and two rituximab biosimilars have been approved by FDA, they were not marketed during the study period [50]. No users of anakinra in both cohorts as well as no users of sarilumab (Italian reimbursement at the end of 2018) in Caserta were identified during the study years. Anakinra was approved for the treatment of moderate‐severe RA but not generally used for RA anymore due to its lower effectiveness when compared to studies using other biologic therapies [51]. Concerning tsDMARDs (i.e., tofacitinib and baricitinib), less than $0.2\%$ of RA users from *Optum versus* no users in Caserta were identified because of recent reimbursement approval of these drugs.
The main strength of this population-based study is the large size and generalizability of the study cohort and the availability of the claims data from the US as well as a Local Health Unit from Southern Italy for the past decade. We acknowledge some limitations of our study, due to the descriptive nature of the analysis, based on data collected through administrative claims databases. However, real-world observational studies provide evidence on how specific drugs are used in the market and what impact they have in the long-term on the already limited health resources. This is in contrast with randomized controlled trials where data are limited to the experimental conditions of the trial design, and where results may not translate fully to the real-world [52–56]. Second, we cannot exclude a potential misclassification of RA patients from the US, thus resulting in a high proportion of untreated RA patients during the first year of follow-up. However, we defined our cohort selection based on previous studies [36–38] and we required all Optum patients to have continuous insurance enrollment during the study period to avoid misclassification due to insurance switching. Furthermore, the traceability of some pharmacy claims, such as NSAIDs/glucocorticoids, might not have been captured by the two databases because they are used as over-the-counter drugs or privately purchased; consequently, the proportion of untreated RA patients could be overestimated; an exploratory analysis was carried using a database provided by IMS Health on pharmacy sales data for all pharmacies in Caserta LHU in the years 2014–2018. Prescription data from IMS are aggregate prescription-level data through which it is possible to distinguish between units of drugs dispensed through the NHS and those purchased privately by citizens. This analysis showed that more than half of NSAIDs and glucocorticoids packages acquired in community pharmacies were bought privately and could not have been captured by the NHS administrative drug dispensing databases. On the contrary, csDMARDs, bDMARDs, and tsDMARDs were fully reimbursed and then traceable. Third, another limitation is represented by the lack of data in the administrative claims databases on clinical outcome measures, such as the effectiveness of treatment, disease severity, and other potential confounders, that could have influenced our results. Finally, our findings from Caserta may not be fully representative of those in the whole *Italian* general population. However, the applied methodology and the Caserta LHU claims database as well as the Arianna database have been shown to provide accurate and reliable information for pharmacoepidemiological research, as documented elsewhere [28–31].
## Conclusions
In conclusion, our study showed substantial heterogeneity in baseline characteristics and access to bDMARD or tsDMARD drugs between RA patients from the United States and Italy. Most RA patients in our study were treated with anti-inflammatory drugs or csDMARDs, especially elderly, rather than bDMARDs or tsDMARDs during the first year post-diagnosis, suggesting a need for better understanding and dealing with barriers in the use of these agents for diagnosed RA patients. In particular, regardless of age, appropriate use of DMARDs should be considered to achieve RA disease remission or low disease activity. With the increasing spectrum of therapeutic options and the new information on existing drugs, this study could be helpful to provide insights into the management of RA patients in clinical practice.
## Supplementary Information
Below is the link to the electronic supplementary material. Online Resource 1. Depiction of the study cohort identification criteria. Legend: Dx: RA diagnosis; MARD: Disease-Modifying Anti-Rheumatic Drug (JPG 193 KB)Online Resource 2. Study drugs approved for the treatment of RA. Legend: DMARD: Disease-Modifying Anti-Rheumatic Drug; csDMARD: Conventional Synthetic Disease-Modifying Anti-Rheumatic Drug; tsDMARD: Targeted Synthetic Disease-Modifying Anti-Rheumatic Drug; bDMARD: Biological Disease-Modifying Anti-Rheumatic Drug; AIC= Italian market authorization code; ATC= anatomical therapeutic chemical classification system (PDF 318 KB)Online Resource 3. Frequency (%) of different compound within the first year after ID. Legend: csDMARD: Conventional Synthetic Disease- Modifying Anti-Rheumatic Drug; tsDMARD: Targeted Synthetic Disease Modifying Anti-Rheumatic Drug; bDMARD: Biological Disease Modifying Anti-Rheumatic Drug. Note: Only compounds with proportions ≥$0.05\%$ were showed (PDF 188 KB)
## References
1. Helmick CG, Felson DT, Lawrence RC, Gabriel S, Hirsch R, Kwoh CK. **Estimates of the prevalence of arthritis and other rheumatic conditions in the United States. Part I**. *Arthritis Rheum* (2008.0) **58** 15-25. DOI: 10.1002/art.23177
2. Michaud K, Messer J, Choi HK, Wolfe F. **Direct medical costs and their predictors in patients with rheumatoid arthritis: a three-year study of 7,527 patients**. *Arthritis Rheum* (2003.0) **48** 2750-62. DOI: 10.1002/art.11439
3. Yi E, Ahuja A, Rajput T, Thomas AG, Park Y. **Clinical, economic, and humanistic burden associated with delayed diagnosis of axial spondyloarthritis: a systematic review**. *Rheumatol Ther* (2020.0) **7** 65-87. DOI: 10.1007/s40744-020-00194-8
4. Erol K, Gok K, Cengiz G, Kilic G, Kilic E, Ozgocmen S. **Extra-articular manifestations and burden of disease in patients with radiographic and non-radiographic axial spondyloarthritis**. *Acta Reumatol Port* (2018.0) **43** 32-39. PMID: 29342471
5. 5.World Health Organization (2021) Chronic rheumatic conditions. Available online: https://www.who.int/chp/topics/rheumatic/en/#:~:text=Rheumatic%20or%20musculoskeletal%20conditions%20comprise,and%20conditions%20resulting%20from%20trauma. Accessed on 10 May 2021
6. Myasoedova E, Crowson CS, Kremers HM, Therneau TMM, Gabriel SE. **Is the incidence of rheumatoid arthritis rising?: results from Olmsted County, Minnesota, 1955–2007**. *Arthritis Rheum* (2010.0) **62** 1576-1582. DOI: 10.1002/art.27425
7. Hunter TM, Boytsov NN, Zhang X, Schroeder K, Michaud KM, Araujo AB. **Prevalence of rheumatoid arthritis in the United States adult population in healthcare claims databases, 2004–2014 (2017)**. *Rheumatol Int* (2017.0) **37** 1551-1557. DOI: 10.1007/s00296-017-3726-1
8. Rossini M, Rossi E, Bernardi D, Viapiana O, Gatti D, Idolazzi L. **Prevalence and incidence of rheumatoid arthritis in Italy**. *Rheumatol Int* (2014.0) **34** 659-664. DOI: 10.1007/s00296-014-2974-6
9. Olofsson T, Petersson IF, Eriksson JK. **Predictors of work disability after start of anti-TNF therapy in a national cohort of Swedish patients with rheumatoid arthritis: does early anti-TNF therapy bring patients back to work?**. *Ann Rheum Dis* (2017.0) **76** 1245-1252. DOI: 10.1136/annrheumdis-2016-210239
10. Tutuncu Z, Kavanaugh A. **Rheumatic disease in the elderly: rheumatoid arthritis**. *Rheum Dis Clin North Am* (2007.0) **33** 57-70. DOI: 10.1016/j.rdc.2006.12.006
11. Xu X, Li QJ, Xia S, Wang MM, Ji W. **Tripterygium glycosides for treating late-onset rheumatoid arthritis: a systematic review and meta-analysis**. *Altern Ther Health Med* (2016.0) **22** 32-39. PMID: 27866179
12. Villa-Blanco JI, Calvo-Alen J. **Elderly onset rheumatoid arthritis differential diagnosis and choice of first-line and subsequent therapy**. *Drugs Aging* (2009.0) **26** 739-750. DOI: 10.2165/11316740-000000000-00000
13. Leon L, Gomez A, Vadillo C. **Severe adverse drug reactions to biological disease-modifying antirheumatic drugs in elderly patients with rheumatoid arthritis in clinical practice**. *Clin Exp Rheumatol* (2018.0) **36** 29-35. PMID: 28598787
14. Giacomelli R, Afeltra A, Alunno A. **International consensus: what else can we do to improve diagnosis and therapeutic strategies in patients affected by autoimmune rheumatic diseases (rheumatoid arthritis, spondyloarthritides, systemic sclerosis, systemic lupus erythematosus, antiphospholipid syndrome and Sjogren’s syndrome)?: the unmet needs and the clinical grey zone in autoimmune disease management**. *Autoimmun Rev* (2017.0) **16** 911-924. DOI: 10.1016/j.autrev.2017.07.012
15. Winthrop KL, Strand V, van der Heijde DM. **The unmet need in rheumatology: reports from the Targeted Therapies meeting, 2016**. *Clin Exp Rheumatol* (2016.0) **34** 69-76. PMID: 27586809
16. Taylor PC, Moore A, Vasilescu R, Alvir J, Tarallo M. **A structured literature review of the burden of illness and unmet needs in patients with rheumatoid arthritis: a current perspective**. *Rheumatol Int* (2016.0) **36** 685-695. DOI: 10.1007/s00296-015-3415-x
17. Smolen JS, Landewé RBM, Bijlsma JWJ, Burmester GR, Dougados M, Kerschbaumer A. **EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2019 update**. *Ann Rheum Dis* (2020.0) **79** 685-699. DOI: 10.1136/annrheumdis-2019-216655
18. Köhler BM, Günther J, Kaudewitz D, Lorenz HM. **Current therapeutic options in the treatment of rheumatoid arthritis**. *J Clin Med* (2019.0) **8** 938. DOI: 10.3390/jcm8070938
19. Van De Laar CJ, Oude Voshaar MAH, Fakhouri WKH, Zaremba-Pechmann L, De Leonardis F, De La Torre I, Van De Laar MAFJ. **Cost-effectiveness of a JAK1/JAK2 inhibitor vs a biologic disease-modifying antirheumatic drug (BDMARD) in a treat-to-target strategy for rheumatoid arthritis**. *Clinicoecon Outcomes Res* (2020.0) **12** 213-222. DOI: 10.2147/CEOR.S231558
20. Smolen JS, Landewé RBM, Bijlsma JWJ, Burmester GR, Chatzidionysiou K, Dougados M. **EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2016 update**. *Ann Rheum Dis* (2017.0) **76** 960-977. DOI: 10.1136/annrheumdis-2016-210715
21. Parisi S, Bortoluzzi A, Sebastiani GD, Conti F, Caporali R, Ughi N. **The Italian Society for Rheumatology clinical practice guidelines for rheumatoid arthritis**. *Reumatismo* (2019.0) **71** 22-49. DOI: 10.4081/reumatismo.2019.1202
22. Drosos AA, Pelechas E, Voulgari PV. **Treatment strategies are more important than drugs in the management of rheumatoid arthritis**. *Clin Rheumatol* (2020.0) **39** 1363-1368. DOI: 10.1007/s10067-020-05001-x
23. Scheinberg MA, Kay J. **The advent of biosimilar therapies in rheumatology—“O brave new world”**. *Nat Rev Rheumatol* (2012.0) **8** 430-436. DOI: 10.1038/nrrheum.2012.84
24. Putrik P, Ramiro S, Kvien TK, Sokka T, Pavlova M, Uhlig T. **Inequities in access to biologic and synthetic DMARDs across 46 European countries**. *Ann Rheum Dis* (2014.0) **73** 198-206. DOI: 10.1136/annrheumdis-2012-202603
25. Genazzani AA, Biggio G, Caputi AP, Del Tacca M, Drago F, Fantozzi R. **Biosimilar drugs: concerns and opportunities**. *BioDrugs* (2007.0) **21** 351-356. DOI: 10.2165/00063030-200721060-00003
26. Kobak S, Bes C. **An autumn tale: geriatric rheumatoid arthritis**. *Ther Adv Musculoskelet Dis* (2018.0) **10** 3-11. DOI: 10.1177/1759720X17740075
27. Angelini J, Talotta R, Roncato R, Fornasier G, Barbiero G, Dal Cin L, Brancati S, Scaglione F. **JAK-inhibitors for the treatment of rheumatoid arthritis: a focus on the present and an outlook on the future**. *Biomolecules* (2020.0) **10** 1002. DOI: 10.3390/biom10071002
28. Ingrasciotta Y, Sultana J, Giorgianni F, Caputi AP, Arcoraci V, Tari DU. **The burden of nephrotoxic drug prescriptions in patients with chronic kidney disease: a retrospective population-based study in Southern Italy**. *PLoS ONE* (2014.0) **9** e89072. DOI: 10.1371/journal.pone.0089072
29. Ingrasciotta Y, Sultana J, Giorgianni F, Fontana A, Santangelo A, Tari DU. **Association of individual non-steroidal anti-inflammatory drugs and chronic kidney disease: a population-based case control study**. *PLoS ONE* (2015.0) **10** e0122899. DOI: 10.1371/journal.pone.0122899
30. Ingrasciotta Y, Sultana J, Giorgianni F, Menditto E, Scuteri A, Tari M. **Analgesic drug use in elderly persons: a population-based study in Southern Italy**. *PLoS ONE* (2019.0) **14** e0222836. DOI: 10.1371/journal.pone.0222836
31. 31.Viola E, Trifirò G, Ingrasciotta Y, Sottosanti L, Tari M, Giorgianni F et al (2016) Adverse drug reactions associated with off-label use of ketorolac, with particular focus on elderly patients. An analysis of the Italian pharmacovigilance database and a population based study. Expert Opin Drug Saf 15(sup2):61–67. 10.1080/14740338.2016.1221401
32. Oppelt KA, Kuiper JG, Ingrasciotta Y, Ientile V, Herings RMC, Tari M. **Characteristics and absolute survival of metastatic colorectal cancer patients treated with biologics: a real-world data analysis from three European countries**. *Front Oncol* (2021.0) **11** 630456. DOI: 10.3389/fonc.2021.630456
33. Khosrow-Khavar F, Kim SC, Lee H, Lee SB, Desai RJ. **Tofacitinib and risk of cardiovascular outcomes: results from the Safety of TofAcitinib in Routine care patients with Rheumatoid Arthritis (STAR-RA) study**. *Ann Rheum Dis* (2022.0) **81** 798-804. DOI: 10.1136/annrheumdis-2021-221915
34. Desai RJ, Pawar A, Khosrow-Khavar F, Weinblatt ME, Kim SC. **Risk of venous thromboembolism associated with tofacitinib in patients with rheumatoid arthritis: a population-based cohort study**. *Rheumatology (Oxford)* (2021.0) **61** 121-130. DOI: 10.1093/rheumatology/keab294
35. Jin Y, Chen SK, Liu J, Kim SC. **Risk of incident type 2 diabetes mellitus among patients with rheumatoid arthritis: a population-based cohort study**. *Arthritis Care Res (Hoboken)* (2020.0) **72** 1248-1256. DOI: 10.1002/acr.24343
36. Jin Y, Desai RJ, Liu J, Choi NK, Kim SC. **Factors associated with initial or subsequent choice of biologic disease-modifying antirheumatic drugs for treatment of rheumatoid arthritis**. *Arthritis Res Ther* (2017.0) **19** 159. DOI: 10.1186/s13075-017-1366-1
37. Desai RJ, Solomon DH, Jin Y, Liu J, Kim SC. **Temporal trends in use of biologic DMARDs for rheumatoid arthritis in the United States: a cohort study of publicly and privately insured patients**. *J Manag Care Spec Pharm* (2017.0) **23** 809-814. DOI: 10.18553/jmcp.2017.23.8.809
38. Kim SY, Servi A, Polinski JM, Mogun H, Weinblatt ME, Katz JN. **Validation of rheumatoid arthritis diagnoses in health care utilization data**. *Arthritis Res Ther* (2011.0) **13** R32. DOI: 10.1186/ar3260
39. Austin PC. **Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples**. *Stat Med* (2009.0) **28** 3083-3107. DOI: 10.1002/sim.3697
40. Innala L, Sjöberg C, Möller B, Ljung L, Smedby T, Södergren A. **Comorbidity in patients with early rheumatoid arthritis—inflammation matters**. *Arthritis Res Ther* (2016.0) **18** 33. DOI: 10.1186/s13075-016-0928-y
41. Skolnick P. **The opioid epidemic: crisis and solutions**. *Annu Rev Pharmacol Toxicol* (2018.0) **58** 143-159. DOI: 10.1146/annurev-pharmtox-010617-052534
42. Zamora-Legoff JA, Achenbach SJ, Crowson CS, Krause ML, Davis JM, Matteson EL. **Opioid use in patients with rheumatoid arthritis 2005–2014: a population-based comparative study**. *Clin Rheumatol* (2016.0) **35** 1137-1144. DOI: 10.1007/s10067-016-3239-4
43. Theis KA, Brady TJ, Sacks JJ. **Where have all the patients gone? Profile of US adults who report doctor-diagnosed arthritis but are not being treated**. *J Clin Rheumatol* (2019.0) **25** 341-347. DOI: 10.1097/RHU.0000000000000896
44. Van der Heijde DM, Van Riel PL, Nuver-Zwart IH, Van de Putte LB. **Sulphasalazine versus hydroxychloroquine in rheumatoid arthritis: 3-year follow-up**. *Lancet* (1990.0) **335** 539. DOI: 10.1016/0140-6736(90)90771-v
45. Perrone V, Losi S, Rogai V, Antonelli S, Fakhouri W, Giovannitti M. **Treatment patterns and pharmacoutilization in patients affected by rheumatoid arthritis in Italian settings**. *Int J Environ Res Public Health* (2021.0) **18** 5679. DOI: 10.3390/ijerph18115679
46. 46.Bonafede M, Johnson BH, Shah N, Harrison DJ, Tang D, Stolshek BS (2018) Disease-modifying antirheumatic drug initiation among patients newly diagnosed with rheumatoid arthritis. Am J Manag Care 24(8 Spec No.):SP279–SP285
47. Harrold LR, Patel PA, Griffith J, Litman HJ, Feng H, Schlacher CA. **Assessing disease severity in bio-naïve patients with RA on treatment with csDMARDs: insights from the Corrona Registry**. *Clin Rheumatol* (2020.0) **39** 391-400. DOI: 10.1007/s10067-019-04727-7
48. Kato E, Sawada T, Tahara K. **The age at onset of rheumatoid arthritis is increasing in Japan: a nationwide database study**. *Int J Rheum Dis* (2017.0) **20** 839-845. DOI: 10.1111/1756-185X.12998
49. Ruban TN, Jacob B, Pope JE, Keystone EC, Bombardier C, Kuriya B. **The influence of age at disease onset on disease activity and disability: results from the Ontario Best Practices Research Initiative**. *Clin Rheumatol* (2016.0) **35** 759-763. DOI: 10.1007/s10067-015-3031-x
50. Gherghescu I, Delgado-Charro MB. **The biosimilar landscape: an overview of regulatory approvals by the EMA and FDA**. *Pharmaceutics* (2020.0) **13** 48. DOI: 10.3390/pharmaceutics13010048
51. 51.Mertens M, Singh JA (2009) Anakinra for rheumatoid arthritis. Cochrane Database Syst Rev (1):CD005121. 10.1002/14651858.CD005121.pub3
52. Saturni S, Bellini F, Braido F, Paggiaro P, Sanduzzi A, Scichilone N. **Randomized controlled trials and real life studies. Approaches and methodologies: a clinical point of view**. *Pulm Pharmacol Ther* (2014.0) **27** 129-38. DOI: 10.1016/j.pupt.2014.01.005
53. Garrison LP, Neumann PJ, Erickson P, Marshall D, Mullins CD. **Using real-world data for coverage and payment decisions: the ISPOR Real-World Data Task Force Report**. *Value Health* (2007.0) **10** 326-335. DOI: 10.1111/j.1524-4733.2007.00186.x
54. 54.Association of the British Pharmaceutical Industry (2011) Demonstrating value with real world data: a practical guide. Available online: http://www.abpi.org.uk/our-work/library/guidelines/Pages/real-world-data.aspx. Accessed on 19 Apr 2017
55. Nallamothu BK, Hayward RA, Bates ER. **Beyond the randomized clinical trial: the role of effectiveness studies in evaluating cardiovascular therapies**. *Circulation* (2008.0) **118** 1294-1303. DOI: 10.1161/CIRCULATIONAHA.107.703579
56. Fakhouri W, Lopez-Romero P, Antonelli S, Losi S, Rogai V, Buda S. **Treatment patterns, health care resource utilization and costs of rheumatoid arthritis patients in Italy: findings from a retrospective administrative database analysis**. *Open Access Rheumatol* (2018.0) **10** 103-111. DOI: 10.2147/OARRR.S164738
|
---
title: Diagnostic accuracy of ultrasound-based multimodal radiomics modeling for fibrosis
detection in chronic kidney disease
authors:
- Xin-Yue Ge
- Zhong-Kai Lan
- Qiao-Qing Lan
- Hua-Shan Lin
- Guo-Dong Wang
- Jing Chen
journal: European Radiology
year: 2022
pmcid: PMC10017610
doi: 10.1007/s00330-022-09268-3
license: CC BY 4.0
---
# Diagnostic accuracy of ultrasound-based multimodal radiomics modeling for fibrosis detection in chronic kidney disease
## Abstract
### Objectives
To predict kidney fibrosis in patients with chronic kidney disease using radiomics of two-dimensional ultrasound (B-mode) and Sound Touch Elastography (STE) images in combination with clinical features.
### Methods
The Mindray Resona 7 ultrasonic diagnostic apparatus with SC5-1U convex array probe (bandwidth frequency of 1–5 MHz) was used to perform two-dimensional ultrasound and STE software. The severity of cortical tubulointerstitial fibrosis was divided into three grades: mild interstitial fibrosis and tubular atrophy (IFTA), fibrotic area < $25\%$; moderate IFTA, fibrotic area 26–$50\%$; and severe IFTA, fibrotic area > $50\%$. After extracting radiomics from B-mode and STE images in these patients, we analyzed two classification schemes: mild versus moderate-to-severe IFTA, and mild-to-moderate versus severe IFTA. A nomogram was constructed based on multiple logistic regression analyses, combining clinical and radiomics. The performance of the nomogram for differentiation was evaluated using receiver operating characteristic (ROC), calibration, and decision curves.
### Results
A total of 150 patients undergoing kidney biopsy were enrolled (mild IFTA: $$n = 74$$; moderate IFTA: $$n = 33$$; severe IFTA: $$n = 43$$) and randomized into training ($$n = 105$$) and validation cohorts ($$n = 45$$). To differentiate between mild and moderate-to-severe IFTA, a nomogram incorporating STE radiomics, albumin, and estimated glomerular filtration (eGFR) rate achieved an area under the ROC curve (AUC) of 0.91 ($95\%$ confidence interval [CI]: 0.85–0.97) and 0.85 ($95\%$ CI: 0.77–0.98) in the training and validation cohorts, respectively. Between mild-to-moderate and severe IFTA, the nomogram incorporating B-mode and STE radiomics features, age, and eGFR achieved an AUC of 0.93 ($95\%$ CI: 0.89–0.98) and 0.83 ($95\%$ CI: 0.70–0.95) in the training and validation cohorts, respectively. Finally, we performed a decision curve analysis and found that the nomogram using both radiomics and clinical features exhibited better predictability than any other model (DeLong test, $p \leq 0.05$ for the training and validation cohorts).
### Conclusion
A nomogram based on two-dimensional ultrasound and STE radiomics and clinical features served as a non-invasive tool capable of differentiating kidney fibrosis of different severities.
### Key Points
• Radiomics calculated based on the ultrasound imaging may be used to predict the severities of kidney fibrosis.
• Radiomics may be used to identify clinical features associated with the progression of tubulointerstitial fibrosis in patients with CKD.
• Non-invasive ultrasound imaging-based radiomics method with accuracy aids in detecting renal fibrosis with different IFTA severities.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00330-022-09268-3.
## Introduction
Chronic kidney disease (CKD) describes a state of progressive structural and functional deterioration of the kidney, presenting as a reduced estimated glomerular filtration rate (eGFR). CKD can lead to end-stage kidney disease (ESKD) and is responsible for $9.1\%$ and $4.6\%$ of noncommunicable disease-related morbidity and mortality, respectively [1]. It is projected that CKD will become the fifth leading global cause of death by 2040 [2]. Consequently, timely diagnosis followed by early treatment initiation for those with CKD is crucial for optimizing their outcomes.
Interstitial fibrosis and tubular atrophy (IFTA) are tightly correlated with CKD severity and impact patients’ long-term prognosis. Moderate and severe IFTA, compared to mild IFTA, and global glomerulosclerosis are associated with more than a two- and three-fold increased risk of kidney function loss, respectively [3]. However, current methods for monitoring kidney fibrosis remain unsatisfactory. In clinical practice, eGFR is not always consistent with the degree of renal fibrosis. eGFR can be quite insensitive to subclinical kidney function impairment. Kidney biopsy is considered the gold standard for confirming CKD diagnosis and fibrosis grading [4–6]. However, kidney biopsy carries the risk of complications, and spatial sampling bias reduces the accuracy of pathological diagnosis; therefore, kidney biopsy has not been considered the preferred follow-up approach for patients with CKD [7, 8].
Morphological changes in the kidney cortex and volume mostly occur during ESKD. Under pathological examination, CKD is characterized by kidney fibrosis, or the pathological deposition of massive extracellular matrices related to an increasing number of fibroblasts [9, 10]. These changes are associated with subsequent scarring and sclerosis of kidney tissues, leading to kidney morphological alterations [11]. Ultrasound examinations can assess changes in speckling pattern and signal scattering, both of which variably correlate with changes in kidney morphology and rising parenchymal stiffness. However, distinguishing diseased kidneys from healthy ones using two-dimensional (2D) ultrasound can be difficult for sonographers. These limitations lead to the increased utility of radiomics. Radiomics are quantifications of medical images using statistical algorithms. The machine learning part is used for outcome prediction in subsequent steps. Radiomics aims to support diagnostic decisions through differentiating between different tissue types [12, 13]. Among radiomics, texture analysis is an emerging tool for quantitating the severity of kidney diseases. Radiomics has been applied to different imaging modalities for the identification and differentiation between kidney diseases, including kidney tumors, carcinomas [14–17], the discrimination of malignant and benign clinical T1 renal masses [18] and renal tumor histological subtypes [19], early kidney damage in patients with diabetes mellitus [20], the detection of kidney stones [21], and the differentiation between normal and diseased kidneys in those with CKD [22].
Based on the reasons outlined above, we combined radiomics data from 2D ultrasound and Sound Touch Elastography (STE) images, as well as clinical factors to construct models for application, followed by model verification. We tried to use a nomogram to predict the degree of IFTA among CKD patients without histopathological data. We aimed to provide a non-invasive diagnosis approach for CKD and used this approach to monitor the treatment responses and disease course of these patients.
## Ethics statement
The current study complied with the Declaration of Helsinki and was approved by the local ethics review board (KY2021146). We obtained written informed consent from each participant.
## Selection of study participants
The definition of CKD was made based on an eGFR < 60 mL/min/1.73 m2 for at least 3 months [23]. The inclusion criteria were CKD patients who had a clinical indication of kidney biopsy. The exclusion criteria were as follows: patients with any contraindications for kidney biopsy, asymmetric bilateral kidney atrophy, abnormal kidney structure, or poor resolution of kidney cortex and medulla on 2D ultrasound. Clinical and laboratory tests were collected from each patient within 2 days before they underwent kidney biopsy.
## Ultrasonography procedures
We used the Mindary Resona 7 Ultrasound System and SC5-1U convex array probe (bandwidth frequency of 1–5 MHz) (Mindray Bio-Medical Electronics Co., Ltd.) to perform 2D ultrasound and STE software. STE measurements were performed 5 times with uniform color fill, and the final standard deviation (SD) of the STE values was set at less than 2.0 as quality control. All examinations were performed by a sonographer with 8 years of experience, who was blinded to serological, imaging, and kidney biopsy pathological results.
## Kidney biopsy and pathological examination
Renal biopsy specimens within 3 days of renal ultrasound were obtained from patients with CKD. A renal needle biopsy was done to sample the lower pole parenchyma of the target kidney under ultrasound (US) guidance. To ensure that the selected US images matched the US biopsy location, the kidney puncture operation, 2D ultrasound, and STE examination were performed by the same sonographer. Two experienced pathologists scored the severity of glomerular sclerosis, tubulointerstitial injury, and vascular sclerosis based on the Banff scoring system and experiences from Farris et al [24, 25]. Any disagreement between pathologists was resolved by consensus. We used the Image-Pro Plus 6.0 software to evaluate the proportion of tubulointerstitial fibrotic areas. Patients with CKD were classified according to the Banff scoring system for kidney cortical fibrosis [26]. In this scoring system, the severity of cortical tubulointerstitial fibrosis was divided into three grades: mild IFTA, fibrotic area < $25\%$; moderate IFTA, fibrotic area 26–$50\%$; and severe IFTA, fibrotic area > $50\%$.
## Image segmentation
Images of Digital Imaging and Communications in Medicine (DICOM) format acquired during B-mode and STE examination were imported into ITK-snap software for manual image segmentation. We evaluated the region of interest (ROI) containing the kidney cortex but removed the kidney medulla and perirenal fat tissues during image curation. Any difference between the two interpreters was resolved by group discussions.
## Feature extraction and establishment of radiomics label
The DICOM images and ROIs obtained from ITK-SNAP software were imported into the AK software (Artificial Intelligence Kit, GE Healthcare) for extracting radiomics. The extracted features included first-order (histogram and morphologic features) alongside second-order parameters. The second-order parameters mainly involved Gray Level Co-Occurrence Matrix (GLCM), Gray Level Run Length Matrix (GLRLM), Gray Level Size Zone Matrix (GLZSM), Neighboring Gray Tone Difference Matrix (NGTDM), and Gray Level Dependence Matrix (GLDM). The ROI of all images was delineated by two sonographers. The inter-observer agreement was evaluated using interclass correlation coefficient (ICC) analysis, which was defined as good consistency for values between 0.75 and 1, fair consistency for values between 0.4 and 0.75, and poor for values under 0.4. ICC values higher than 0.75 were selected for further analysis. Patients were randomly divided into training and validation cohorts at a ratio of 7:3. We planned for two types of comparisons: mild versus moderate-to-severe IFTA and mild-to-moderate versus severe IFTA.
## Feature selection
Minimum redundancy maximum relevance (mRMR) was used to eliminate redundant and irrelevant features, retain the optimal ones, filter out the optimal feature subset through the least absolute shrinkage and selection operator (LASSO) algorithm, and build a final model. After determining the number of optimal features, we selected the most predictive feature subset and calculated the corresponding coefficients [27].
## Model construction and result validation
Feature extraction based on B-mode and STE images yielded a radiomics quality score (Rad-Score), which was the radiomics label calculated by the weighted summation of selected features by their coefficients. We used receiver operating characteristic curve (ROC) analysis to evaluate the performance of each constructed model. The Akaike information criterion of the clinical model was applied to determine the most appropriate clinical model. Multivariate logistic regression combining clinical features with the Rad-Score was conducted to establish a predictive model and generate a clinical nomogram. The usefulness of a nomogram lies in its ability to map prediction probability to points on a picture with a scale between 0 and 100. The total points accrued based on different types of features corresponded with the predicted probabilities of the index patient [28, 29] The predictive accuracy of each model was assessed by the area under the ROC curve (AUC) value for the training and validation cohorts. We tested the performance of the Knott diagram in the validation cohort. Using the logistic regression model established in the training cohort, we calculated the total score for each patient in the validation cohort and obtained the AUC and calibration curve. To estimate the prediction error of each model, we further tested the proposed model using the 1000-iteration bootstrap analysis for both the training and validation cohorts. We randomly selected $70\%$ of patients from the training or validation cohort and calculated the corresponding AUC values.
Comparisons between AUCs were made with the DeLong test. The calibration curves and Hosmer–Lemeshow test were used to investigate the performance of the nomogram. Finally, to evaluate the clinical practicability by quantifying the net benefits of the nomogram model in both the training and validation cohorts, the decision curve analysis (DCA) was performed based on clinical features and radiomics labels from B-model, STE-model, B plus STE model, and the combined models. DCA determines the clinical practicability of radiomics nomograms by quantifying the net benefits under different threshold probabilities in the validation set.
## Statistical analyses
SPSS (version 26.0; IBM), GraphPad Prism 8.0 (GraphPad Software), and R statistical software (version 4.0.2) were used for statistical tests. $p \leq 0.05$ was considered statistically significant.
## Basic clinical information from participants
The flowchart of patient selection is provided in Fig. 1. A total of 150 patients with CKD were identified in Table 1, along with their pathological diagnoses (Supplementary Table 1). The course of processing radiomics is shown in Fig. 2. We also illustrated how the ITK-SNAP software delineated the ROI of the punctured kidney cortex (Fig. 3). Table 2 shows the clinical characteristics of training and validation cohorts. Fig. 1The flowchart of patient selection processTable 1Patient’s characteristics at baselineCharacteristicAllMild IFTAModerate IFTASevere IFTAp valueNo. of patients150743343–Age (years)53 (38–64)50 (33–58)55 (48–66)57 (48–67)0.008Gender (female)50 ($33.3\%$)14 ($18.9\%$)12 ($35.3\%$)9 ($31.0\%$)0.321BMI (kg/m2)24.27 ± 4.2224.61 ± 4.224.08 ± 4.723.83 ± 3.890.673eGFR (mL/min/1.73 m2)48.02 ± 33.7068.99 ± 30.98538.49 ± 21.0619.25 ± 18.12< 0.001UACR268.19 (64.98–593.48)316.38 (84.1–743.85)83.4 (29.16–217.29)311.79 (119.25–546.62)0.007Renal length (mm)96.24 ± 10.7599.07 ± 8.6795.76 ± 13.1891.74 ± 10.580.038Renal width (mm)46.53 ± 7.3847.59 ± 8.6746.39 ± 6.1144.79 ± 10.390.64Cortical thickness (mm)11.53 ± 2.4112.16 ± 2.1210.91 ± 2.1310.93 ± 2.820.001ROI depth (mm)4.50 ± 2.794.52 ± 1.224.96 ± 1.164.13 ± 0.920.29Kidney elasticity (kPa)13.01 ± 3.4111.31 ± 2.4213.85 ± 3.6115.3 ± 3.19< 0.001β2-MG (mg/L)4.10 (2.45–7.68)2.87 (2.0–5.11)4.04 (3.16–6.62)8.42 (5.1–13.24)< 0.001Hemoglobin (g/dL)112.85 ± 29.69125.57 ± 28.74112.15 ± 27.4791.49 ± 19.06< 0.001BUN (mmol/L)8.10 (5.60–14.40)6.15 (4.9–8.65)9.1 (6.6–14.2)14.3 (9.3–21.3)< 0.001Blood albumin (g/L)33.9 (23.80–39.73)25.55 (17.95–38.5)37.9 (32.2–37.9)35.5 (29.2–38.7)0.085Blood glucose (mmol/L)5.62 (4.79–7.23)5.36 (4.67–6.76)5.59 (5.11–8.03)5.96 (4.79–7.59)0.10724h Ualb (mg/d)1740.90 (506.75–5056.88)2761.5 (775.8–7703.6)993.6 (316.95–1926.95)1879.8 (1183–4002)0.047BMI, body mass index; eGFR, estimated glomerular filtration rate; UACR, urinary albumin-to-creatinine ratio; β2-MG, β2 microgloulin; Scr, serum creatinine; 24h Ualb, 24h urinary protein; Data are presented as mean ± standard error, median (interquartile range), or count (percentage)Fig 2Radiomics flow chart of this study. We exported the collected images in DICOM format, used the ITK software to delineate ROIs, and performed image segmentation. We used the AK software to extract ultrasound radiomics, and built models based on the clinical characteristics of patients with CKD. Later, we performed model calibration and validation. Fig. 3Pathological examinations using Masson staining from kidney biopsy contrasting images from B-mode, and ITK-SNAP ROI in patients with different IFTA severities. A–C A 38-year-old female patient with systemic lupus erythematosus. Kidney biopsy showing focal proliferative lupus nephritis, III-(A), tubulo-interstitial atrophy was $5\%$ assessed by Masson staining, mild IFTA. D–F A 36-year-old female patient with chronic kidney disease, $40\%$ tubulo-interstitial atrophy assessed by Masson staining, moderate IFTA; G–I A 28-year-old male patient with chronic kidney disease. Tubulo-interstitial atrophy assessed by Masson staining was $70\%$, with severe IFTATable 2Clinical characteristics of the training and validation cohortsVariablesMild IFTA vs moderate-to-severe IFTAp valueMild-to-moderate IFTA vs severe IFTAp valueTraining cohort ($$n = 105$$)Validation cohort ($$n = 45$$)Training cohort ($$n = 105$$)Validation cohort ($$n = 45$$)Age (years)50.2 (17.3)49.3 (15.7)0.53651.1 (17.1)47.1 (15.9)0.104Gender Female35 (33.3)15 (33.3)134 (32.1)16 (36.4)0.61 Male70 (66.7)30 (66.7)72 (67.9)28 (63.6)BMI (kg/m2)23.8 (4.2)25.4 (4.2)0.0124.4 (4.3)24.1 (3.9)0.87eGFR (mL/min/1.73 m2)48.6 (34.6)46.7 (31.8)0.8544 (32.3)57.7 (35.3)0.04Renal length (mm)96 (10.8)96.8 (10.7)0.5996.1 (10.8)96.7 (10.7)0.61Renal width (mm)46.6 (7.8)46.4 (6.4)0.8246.6 (7.8)46.3 (6.2)0.95Cortical thickness (mm)11.5 (2.4)11.6 (2.5)0.6811.5 (2.5)11.5 (2.3)0.76ROI depth (mm)4.2 (1.1)5.3 (4.8)0.0424.3 (1.2)5 (4.8)0.46Kidney elasticity (kPa)13.1 (3.4)12.8 (3.4)0.5113.3 (3.6)12.3 (2.7)0.21β2-MG (mg/L)6.5 (7.8)6.2 (6.4)0.697 (8.1)5.1 (5.2)0.11Hemoglobin (g/dL)112.6 (29.9)113.4 (29.6)0.88110.2 (30.8)119.2 (25.9)0.06BUN (mmol/L)11.4 (8.3)10.4 (6.6)0.7211.9 (8.1)9.1 (6.9)0.002Blood albumin (g/L)31.4 (9.7)34.1 (16.8)0.3331.3 (10.5)34.3 (15.8)0.48Blood glucose (mmol/L)6.6 (3.8)6.8 (3.1)0.366.7 (3.8)6.6 (2.8)0.6224h Ualb (mg/d)3,255.1 (3,784.2)3,707 (4,012.9)0.413,583.4 (4,110.3)2,926.5 (3,113.8)0.56
## Feature selection, model construction, and results validation
We extracted 1156 radiomics from the B-mode and STE images for each participant, based on the result of reproducibility analysis by two sonographers, 739 radiomics had good consistency (ICC > 0.75), and retained 120 features after being filtered by the mRMR method. We also did texture feature selection based on the LASSO logistic regression (Supplementary Figure 1) and selected 36 radiomics after the procedure. These features were used to construct the radiomics signature (Fig. 4). The final formula for calculating Rad-*Scores is* shown in the Supplementary Materials. We compared the Rad-Scores between the training and the testing groups, as shown in Supplementary Figure 2. Fig. 4Radiomics signatures for B-mode and STE images. A Four features from B-mode images of mild versus moderate-to-severe IFTA; B Eleven features from STE images of mild versus moderate-to-severe IFTA; C Nine features from B-mode images of mild-to-moderate versus severe IFTA; D Twelve features from STE images of mild-to-moderate versus severe IFTA We further compared the results of B-model, STE model, clinical model and combined model between groups of different IFTA grades, accompanied by model validation, and prediction parameters were calculated using the Youden index (Table 3). We also showed the results using decision curve evaluation models in Supplementary Figure 3. The predicting models built on clinical features for identifying mild vs. moderate-to-severe IFTA and for mild-to-moderate vs. severe IFTA are shown in Supplementary Figure 4. Table 3Diagnostic performance of different model prediction for the assessment of IFTA in two training and validation groupsAUCACCACC LowerACC UpperSenSpePPVNPVMild IFTA vs moderate-to-severe IFTA B-model Train0.72 (0.63–0.82)0.670.570.760.550.800.760.61 Test0.71 (0.55–0.87)0.710.560.840.750.670.720.70 STE-model Train0.81 (0.73–0.89)0.750.660.830.950.530.700.90 Test0.73 (0.58–0.88)0.690.530.820.830.520.670.73 B-model + STE-model Train0.81 (0.74–0.89)0.750.660.830.880.610.720.81 Test0.75 (0.61–0.90)0.760.600.870.830.670.740.78 Clinical model Train0.88 (0.80–0.96)0.880.800.930.880.880.890.86 Test0.80 (0.66–0.94)0.780.630.890.960.570.720.93 Clinical model+ STE-model Train0.91 (0.85–0.97)0.850.760.910.790.920.920.79 Test0.85 (0.73–0.98)0.730.580.850.580.900.880.66Mild-to-moderate IFTA vs severe IFTA B-model Train0.80 (0.71–0.90)0.810.720.880.710.850.670.88 Test0.78 (0.65–0.92)0.700.550.830.500.780.460.81 STE-model Train0.81 (0.73–0.89)0.690.590.780.970.570.480.98 Test0.73 (0.58–0.88)0.610.450.760.750.560.390.86 B-model + STE-model Train0.93 (0.88–0.98)0.850.770.910.900.830.680.95 Test0.86 (0.75–0.97)0.800.650.900.920.750.580.96 Clinical model Train0.67 (0.55–0.79)0.680.580.770.680.680.470.84 Test0.55 (0.34–0.76)0.750.600.870.250.940.600.77 Clinical model+ B-model + STE-model Train0.93 (0.89–0.98)0.860.780.920.870.850.710.94 Test0.83 (0.70–0.95)0.800.650.900.750.810.600.90Note: ACC, accuracy; Sen, sensitivity; Spe, specificity; PPV, positive predictive value; NPV, negative predictive value
## Clinical features combined with ultrasound radiomics model performance and nomogram validation in analyses involving different IFTA group comparisons
During the validation of models comparing mild IFTA to moderate-to-severe IFTA, the clinical model established using serum albumin and eGFR achieved moderate prediction ability. Moderate prediction ability was also achieved using the STE radiomics model (Table 3). After adding the results of the STE radiomics model to the clinical model, the predictive performance of the combined model was significantly improved, with the nomogram shown in Fig. 5A, with AUCs of 0.91 ($95\%$ CI: 0.85–0.97) and 0.85 ($95\%$ CI: 0.77–0.98) for the training cohort and testing cohorts, respectively (DeLong test, $p \leq 0.05$) (Fig. 5B, C). The nomogram calibration curve showed good agreement between the predictions and observations in the two groups (Fig. 5D, E). The DCA of the nomogram is shown in Fig. 5F. The DCA based on the combined models (clinical and STE) showed greater benefits in the prediction of IFTA severity in the 20–$80\%$ threshold probabilities compared to the clinical and STE models. Fig. 5A Nomogram for clinical features (albumin and eGFR) of mild vs. moderate-to-severe IFTA combined with STE Rad-Scores. B, C Clinical characteristics of mild vs. moderate-to-severe IFTA combined with the ROC curves of STE model in the training and validation sets. D, E Calibration curves of the nomogram for clinical model of mild vs. moderate-to-severe IFTA combined with STE model in the training and validation cohorts. F Analysis of the cut curve of the histogram for comparison of mild with moderate-to-severe IFTA in the clinical model alone, STE model alone, and combined model. The Y-axis is net income. The blue line represents the decision curve of the STE model. The green line represents the clinical model curve, whereas the red line represents the decision curve of the STE model combined with clinical model of patients with CKD During the validation of models comparing mild-to-moderate IFTA to severe IFTA, the clinical model established based on age and eGFR achieved moderate prediction ability, with AUCs of 0.67 ($95\%$ CI: 0.55–0.79) and 0.55 ($95\%$ CI: 0.34–0.76) for the training and testing cohorts, respectively. Moderate prediction ability was also achieved using the B-mode radiomics model, with AUCs of 0.80 ($95\%$ CI: 0.71–0.90) and 0.78 ($95\%$ CI: 0.65–0.92) for the training and testing cohorts, respectively. Moderate prediction ability was similarly achieved using the STE radiomics model, with AUCs of 0.81 ($95\%$ CI: 0.73–0.89) and 0.73 ($95\%$ CI: 0.58–0.88) for the training and testing cohorts, respectively. Higher prediction ability was achieved using the B-mode plus STE radiomics model, with AUCs of 0.93 ($95\%$ CI: 0.88–0.98) and 0.86 ($95\%$ CI: 0.75–0.97) for the training and testing cohorts, respectively. Finally, models established using age and eGFR, B-mode, and STE radiomics data showed that the prediction ability of the combined model was high, with the nomogram shown in Fig. 6A. The AUCs of the training and the testing cohorts were 0.93 ($95\%$ CI: 0.89–0.98) and 0.83 ($95\%$ CI: 0.70–0.95), respectively (Fig. 6B, C). The AUCs of the combined model significantly differed from those of the clinical model, B model, or STE model (DeLong test, $p \leq 0.005$ for the training and validation cohorts). The nomogram calibration curves showed good agreement between predictions and observations in the two groups (Fig. 6D, E). The DCA of the nomogram is shown in Fig. 6F. Compared to other models, the combined nomogram model, showing the highest area under the curve, is the optimal decision making for maximal net benefit in classifying IFTA severity. Fig. 6A Nomogram for clinical features (age and eGFR) of mild-to-moderate IFTA vs. severe IFTA combined with B-mode Rad-Scores and STE Rad-Scores. B, C Clinical characteristics model of mild-to-moderate vs. severe IFTA combined with ROC curves of B-model and STE modelin the training set and validation set. D, E Calibration curves of this nomogram for clinical features model of mild-to-moderate vs. severe IFTA in combination with B-model and STE-model in the training and validation cohorts. F Analysis of the cutting curve of the nomogram for the clinical model alone, B model alone, STE model alone, and B+STE combined model and the clinical+B+STE combined model comparing mild-to-moderate with severe IFTA. The Y-axis is net income. The gray, green, blue, purple, and red lines represent the clinical model curve, B model curve, decision curve of the STE model, B+STE model curve, and decision curve of the clinical +B+STE model, respectively
## Discussion
The accurate and non-invasive classification of kidney fibrosis severities is crucial for clinical practice. Recently, researchers used machine learning based on elastography ultrasound images to gauge the severity of kidney fibrosis, with promising results [30]. The construction of a binary classification model is mostly used for comparing liver and kidney fibrosis severities [30–33]. In this study, we performed a binary classification by comparing one IFTA grade with the other grades as an approach. A combined model incorporating 2D ultrasound radiomics, STE radiomics, and clinical features for predicting IFTA severities was constructed and validated.
Among clinical features analyzed in this study, eGFR was an independent parameter as shown in different IFTA prediction models (Figs. 5 and 6), consistent with results from Zhu et al [30]. eGFR is an important indicator for estimating kidney function and assessing IFTA severity in patients with CKD [34]. eGFR is calculated based on a standardized formula using Scr, a laboratory index that is widely used for the clinical follow-up of these patients [35]. In the comparative model of mild and moderate-to-severe IFTA, the combined model based on clinical features (serum albumin and eGFR) and STE radiomics further improved the diagnostic performance. eGFR has limitations as an indicator for kidney fibrosis, since the levels of eGFR are frequently inconsistent with the degree of kidney fibrosis. Furthermore, eGFR is not sensitive to subclinical kidney damage [36]. In our training and testing cohorts, comparing mild-to-moderate IFTA to severe IFTA, the AUC of the clinical features (eGFR and age) model for discrimination was 0.67 ($95\%$ CI: 0.55–0.79) and 0.55 ($95\%$ CI: 0.34–0.76) in the training and testing cohorts, respectively, suggesting that clinical features model only performed worse than B-model or STE-model only or the combined model (Fig. 6B, C). Judging from the above arguments, we selected IFTA severity as the grouping variable and prediction model construction instead of eGFR. Therefore, a combined model consisting of clinical factors of eGFR and ultrasonography radiomics features can be helpful for achieving non-invasive monitoring of kidney fibrosis.
The main factors affecting STE elasticity measurements are anisotropy and the heterogeneity of kidney fibrosis. Other confounding factors for STE measurement include age and BMI [30]. In this study, age was used to construct a nomogram for predicting the comparison between mild and moderate-to-severe IFTA. Clinical model established by age and eGFR, and the combined model all achieved a fair predictive performance. In reality, kidneys become stiffened due to collagen deposition during ESKD, and STE measurement results will increase. However, with renal function further declining, kidneys may become softer due to poor blood perfusion, and the STE measurement results may decrease, whereas the kidney length becomes smaller on 2D ultrasound examination [37–39]. These factors likely lead to the emergence of a complex nonlinear relationship between 2D ultrasound measurements, STE measurements, and IFTA severity. In our study, a combined model built based on B-mode and STE results significantly improved the diagnostic performance of traditional ultrasound alone. Possible explanations for this finding include the ability of STE to capture the stiffness feature of patients’ kidneys, which is suitable for application during machine learning whose strength includes combining variables with nonlinear relationships and interactions [40]. Therefore, we used all variables including 2D ultrasound, STE radiomics, and clinical factors from these patients with CKD to model IFTA.
In the nomogram differentiating mild and moderate-to-severe IFTA models, STE radiomics and clinical factors were included, whereas 2D ultrasound radiomics were not. The reason is that the 2D ultrasound radiomics consist of data including the diameter of the kidney’s long axis and its cortical thickness. However, in patients with mild and moderate IFTA, changes in their kidney morphology remain minimal due to their early CKD stages [41]. In this study, there were no differences in kidney lengths and cortical thickness between different IFTA groups ($$p \leq 0.487$$ and $$p \leq 0.927$$ for the mild and moderate IFTA groups, respectively). During our construction of a comparison model between the mild and moderate-to-severe IFTA groups, we extracted 2D ultrasound image features from those with moderate-to-severe IFTA. Since the radiomics of moderate IFTA were included, the kidney morphological features that did not significantly differ between those with mild and moderate IFTA were extracted. The presence of redundant information might increase the probability of model overfitting, reducing model performance after constructing a joint model. However, when we compared between those with mild-to-moderate and severe IFTA, patients with ESKD and severe IFTA were more likely to have morphological kidney atrophy and cortical thinning. When we compared renal long-axis diameter and cortical thickness between severe and mild-to-moderate IFTA groups, there were differences between groups ($p \leq 0.01$). Therefore, the addition of radiomics including morphological differences of the kidneys in 2D ultrasound greatly increased the diagnostic performance of the combined model.
The combined model incorporating B-mode, STE, and clinical features can be applicable for IFTA detection for patients outside our training cohort, particularly during the follow-up of patients unable to receive a renal biopsy. The establishment of ultrasound radiomics model can be a great support for clinical ultrasound practice, and radiomics findings may assist in IFTA prediction in the future.
This study has some limitations. Patients selected were those with CKD and renal biopsy indications. The renal cortical tissues of patients with ESKD could be thin, precluding the derivation of histopathological results based on renal biopsy. The sample size of patients with severe IFTA was small, necessitating further expansion to reduce data redundancy during model construction, in order to facilitate the establishment of multi-classification models. In addition, this study was done based on data from one center, using a single-mode ultrasound diagnostic apparatus to collect ultrasound radiomics. Multi-center and different ultrasound modes may be needed to extract more 2D and ultrasound elasticity radiomics to construct a combined model and to test the generalizability of our established combined model. Finally, changes in 2D ultrasound and STE features and the course of CKD among these patients need to be further monitored and validated in the future.
## Conclusion
STE combined with 2D ultrasound examinations can improve the diagnostic performance of traditional ultrasound for tubulointerstitial fibrosis in patients with CKD. The radiomics nomograms constructed based on 2D ultrasound and STE imaging features in combination with clinical features are non-invasive tools with high accuracy in detecting renal fibrosis with different IFTA severities. This approach can be helpful for non-invasive monitoring of kidney fibrosis.
## References
1. 1.GBD Chronic Kidney Disease Collaboration (2020) Global, regional, and national burden of chronic kidney disease, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet 395:709–733
2. Foreman KJ, Marquez N, Dolgert A. **Forecasting life expectancy, years of life lost, and all-cause and cause-specific mortality for 250 causes of death: reference and alternative scenarios for 2016-40 for 195 countries and territories**. *Lancet* (2018.0) **392** 2052-2090. DOI: 10.1016/S0140-6736(18)31694-5
3. 3.Srivastava A, Palsson R, Kaze AD et al (2018) The prognostic value of histopathologic lesions in native kidney biopsy specimens: results from the Boston kidney biopsy cohort study. J Am Soc Nephrol 29:2213–2224
4. Floege J, Amann K. **Primary glomerulonephritides**. *Lancet* (2016.0) **387** 2036-2048. DOI: 10.1016/S0140-6736(16)00272-5
5. Hahn BH, McMahon MA, Wilkinson A. **American college of rheumatology guidelines for screening, treatment, and management of lupus nephritis**. *Arthritis Care Res (Hoboken)* (2012.0) **64** 797-808. DOI: 10.1002/acr.21664
6. Binda V, Moroni G, Messa P. **ANCA-associated vasculitis with renal involvement**. *J Nephrol* (2018.0) **31** 197-208. DOI: 10.1007/s40620-017-0412-z
7. Korbet SM, Volpini KC, Whittier WL. **Percutaneous renal biopsy of native kidneys: a single-center experience of 1,055 biopsies**. *Am J Nephrol* (2014.0) **39** 153-162. DOI: 10.1159/000358334
8. Franke M, Kramarczyk A, Taylan C, Maintz D, Hoppe B, Koerber F. **Ultrasound-guided percutaneous renal biopsy in 295 children and adolescents: role of ultrasound and analysis of complications**. *PLoS ONE* (2014.0) **9** e114737. DOI: 10.1371/journal.pone.0114737
9. Muñoz-Félix JM, González-Núñez M, Martínez-Salgado C, López-Novoa JM. **TGF-β/BMP proteins as therapeutic targets in renal fibrosis. Where have we arrived after 25 years of trials and tribulations?**. *Pharmacol Ther* (2015.0) **156** 44-58. DOI: 10.1016/j.pharmthera.2015.10.003
10. Klinkhammer BM, Goldschmeding R, Floege J, Boor P. **Treatment of renal fibrosis-turning challenges into opportunities**. *Adv Chronic Kidney Dis* (2017.0) **24** 117-129. DOI: 10.1053/j.ackd.2016.11.002
11. Berchtold L, Friedli I, Vallée JP, Moll S, Martin PY, De Seigneux S. **Diagnosis and assessment of renal fibrosis: the state of the art**. *Swiss Med Wkly* (2017.0) **147** w14442. PMID: 28634969
12. Kumar V, Gu Y, Basu S. **Radiomics: the process and the challenges**. *Magn Reson Imaging* (2012.0) **30** 1234-1248. DOI: 10.1016/j.mri.2012.06.010
13. Gillies RJ, Kinahan PE, Hricak H. **Radiomics: images are more than pictures, they are data**. *Radiology* (2016.0) **278** 563-577. DOI: 10.1148/radiol.2015151169
14. 14.Meng Xl, Shu J, Xia YW, Yang RW (2020) A CT-based radiomics approach for the differential diagnosis of sarcomatoid and clear cell renal cell carcinoma. Biomed Res Int 2020:7103647
15. Shin HJ, Kwak JY, Lee E. **Texture analysis to differentiate malignant renal tumors in children using gray-scale utrasonography images**. *Ultrasound Med Biol* (2019.0) **45** 2205-2212. DOI: 10.1016/j.ultrasmedbio.2019.03.017
16. Diaz de Leon A, Kapur P, Pedrosa I. **Radiomics in kidney cancer: MR Imaging**. *Magn Reson Imaging Clin N Am* (2019.0) **27** 1-13. DOI: 10.1016/j.mric.2018.08.005
17. Yu HS, Scalera J, Khalid M. **Texture analysis as a radiomic marker for differentiating renal tumors**. *Abdom Radiol (NY)* (2017.0) **42** 2470-2478. DOI: 10.1007/s00261-017-1144-1
18. Uhlig J, Biggemann L, Nietert MM. **Discriminating malignant and benign clinical T1 renal masses on computed tomography: a pragmatic radiomics and machine learning approach**. *Medicine (Baltimore)* (2020.0) **99** e19725. DOI: 10.1097/MD.0000000000019725
19. 19.Uhlig J, Leha A, Delonge LM et al (2020) Radiomic features and machine learning for the discrimination of renal tumor histological subtypes: a pragmatic study using clinical-routine computed tomography. Cancers (Basel) 12:3010
20. Deng Y, Yang BR, Luo JW, Du GX, Luo LP. **DTI-based radiomics signature for the detection of early diabetic kidney damage**. *Abdom Radiol (NY)* (2020.0) **45** 2526-2531. DOI: 10.1007/s00261-020-02576-6
21. De Perrot T, Hofmeister J, Burgermeister S. **Differentiating kidney stones from phleboliths in unenhanced low-dose computed tomography using radiomics and machine learning**. *Eur Radiol* (2019.0) **29** 4776-4782. DOI: 10.1007/s00330-019-6004-7
22. Bandara MS, Gurunayaka B, Lakraj G, Pallewatte A, Siribaddana S, Wansapura J. **Ultrasound based radiomics features of chronic kidney disease**. *Acad Radiol* (2022.0) **29** 229-235. DOI: 10.1016/j.acra.2021.01.006
23. 23.National Kidney Foundation (2002) K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Am J Kidney Dis 39:S1–S266
24. Mariani LH, Martini S, Barisoni L. **Interstitial fibrosis scored on whole-slide digital imaging of kidney biopsies is a predictor of outcome in proteinuric glomerulopathies**. *Nephrol Dial Transplant* (2018.0) **33** 310-318. DOI: 10.1093/ndt/gfw443
25. 25.Farris AB, Alpers CE (2014) What is the best way to measure renal fibrosis?: a pathologist’s perspective. Kidney Int Suppl 4:9–15
26. Solez K, Colvin RB, Racusen LC. **Banff 07 classification of renal allograft pathology: updates and future directions**. *Am J Transplant* (2008.0) **8** 753-760. DOI: 10.1111/j.1600-6143.2008.02159.x
27. Alshamlan H, Badr G, Alohali Y. **mRMR-ABC: a hybrid gene selection algorithm for cancer classification using microarray gene expression profiling**. *Biomed Res Int* (2015.0) **2015** 1-15. DOI: 10.1155/2015/604910
28. Iasonos A, Schrag D, Raj GV, Panageas KS. **How to build and interpret a nomogram for cancer prognosis**. *J Clin Oncol* (2008.0) **26** 1364-1370. DOI: 10.1200/JCO.2007.12.9791
29. Stephenson AJ, Scardino PT, Eastham JA. **Postoperative nomogram predicting the 10-year probability of prostate cancer recurrence after radical prostatectomy**. *J Clin Oncol* (2005.0) **23** 7005-7012. DOI: 10.1200/JCO.2005.01.867
30. 30.Zhu MY, Ma LY, Yang WQ et al (2021) Elastography ultrasound with machine learning improves the diagnostic performance of traditional ultrasound in predicting kidney fibrosis. J Formos Med Assoc. 10.1016/j.jfma.2021.08.011
31. Xue LY, Jiang ZY, Fu TT. **Transfer learning radiomics based on multimodal ultrasound imaging for staging liver fibrosis**. *Eur Radiol* (2020.0) **30** 2973-2983. DOI: 10.1007/s00330-019-06595-w
32. Pickhardt PJ, Graffy PM, Said A. **Multiparametric CT for noninvasive staging of hepatitis C virus-related liver fibrosis: correlation with the histopathologic fibrosis score**. *AJR Am J Roentgenol* (2019.0) **212** 547-553. DOI: 10.2214/AJR.18.20284
33. Park HJ, Lee SS, Park B. **Radiomics analysis of gadoxetic acid-enhanced MRI for staging liver fibrosis**. *Radiology* (2019.0) **290** 380-387. DOI: 10.1148/radiol.2018181197
34. Chen CJ, Pa TW, Hsu HH, Chien Hung L, Chen KS, Chen YC. **Prediction of chronic kidney disease stages by renal ultrasound imaging**. *Enterp Inf Syst* (2019.0) **14** 178-195. DOI: 10.1080/17517575.2019.1597386
35. Floege J, Barbour SJ, Cattran DC. **Management and treatment of glomerular diseases (part 1): conclusions from a kidney disease: improving Global Outcomes (KDIGO) Controversies Conference**. *Kidney Int* (2019.0) **95** 268-280. DOI: 10.1016/j.kint.2018.10.018
36. Ruiz-Ortega M, Rayego-Mateos S, Lamas S, Ortiz A, Rodrigues-Diez RR. **Targeting the progression of chronic kidney disease**. *Nat Rev Nephrol* (2020.0) **16** 269-288. DOI: 10.1038/s41581-019-0248-y
37. LeBleu VS, Taduri G, O’Connell J. **Origin and function of myofibroblasts in kidney fibrosis**. *Nat Med* (2013.0) **19** 1047-1053. DOI: 10.1038/nm.3218
38. Syversveen T, Brabrand K, Midtvedt K, Strøm EH, Hartmann A, Berstad AE. **Non-invasive assessment of renal allograft fibrosis by dynamic sonographic tissue perfusion measurement**. *Acta Radiol* (2011.0) **52** 920-926. DOI: 10.1258/ar.2011.110215
39. Warner L, Yin M, Glaser KJ. **Noninvasive In vivo assessment of renal tissue elasticity during graded renal ischemia using MR elastography**. *Invest Radiol* (2011.0) **46** 509-514. DOI: 10.1097/RLI.0b013e3182183a95
40. Sealfon RSG, Mariani LH, Kretzler M, Troyanskaya OG. **Machine learning, the kidney, and genotype-phenotype analysis**. *Kidney Int* (2020.0) **97** 1141-1149. DOI: 10.1016/j.kint.2020.02.028
41. Hoi S, Takata T, Sugihara T. **Predictive value of cortical thickness measured by ultrasonography for renal impairment: a longitudinal study in chronic kidney disease**. *J Clin Med* (2018.0) **7** 527. DOI: 10.3390/jcm7120527
|
---
title: Assessment of epicardial adipose tissue on virtual non-contrast images derived
from photon-counting detector coronary CTA datasets
authors:
- Franka Risch
- Florian Schwarz
- Franziska Braun
- Stefanie Bette
- Judith Becker
- Christian Scheurig-Muenkler
- Thomas J. Kroencke
- Josua A. Decker
journal: European Radiology
year: 2022
pmcid: PMC10017616
doi: 10.1007/s00330-022-09257-6
license: CC BY 4.0
---
# Assessment of epicardial adipose tissue on virtual non-contrast images derived from photon-counting detector coronary CTA datasets
## Abstract
### Objectives
To assess epicardial adipose tissue (EAT) volume and attenuation of different virtual non-contrast (VNC) reconstructions derived from coronary CTA (CCTA) datasets of a photon-counting detector (PCD) CT-system to replace true non-contrast (TNC) series.
### Methods
Consecutive patients ($$n = 42$$) with clinically indicated CCTA and coronary TNC were included. Two VNC series were reconstructed, using a conventional (VNCConv) and a novel calcium-preserving (VNCPC) algorithm. EAT was segmented on TNC, VNCConv, VNCPC, and CCTA (CTA-30) series using thresholds of −190 to −30 HU and an additional segmentation on the CCTA series with an upper threshold of 0 HU (CTA0). EAT volumes and their histograms were assessed for each series. Linear regression was used to correlate EAT volumes and the Euclidian distance for histograms. The paired t-test and the Wilcoxon signed-rank test were used to assess differences for parametric and non-parametric data.
### Results
EAT volumes from VNC and CCTA series showed significant differences compared to TNC (all $p \leq .05$), but excellent correlation (all R2 > 0.9). Measurements on the novel VNCPC series showed the best correlation (R2 = 0.99) and only minor absolute differences compared to TNC values. Mean volume differences were −$12\%$, −$3\%$, −$13\%$, and +$10\%$ for VNCConv, VNCPC, CTA-30, and CTA0 compared to TNC. Distribution of CT values on VNCPC showed less difference to TNC than on VNCConv (mean attenuation difference +$7\%$ vs. +$2\%$; Euclidean distance of histograms 0.029 vs. 0.016).
### Conclusions
VNCPC-reconstructions of PCD-CCTA datasets can be used to reliably assess EAT volume with a high accuracy and only minor differences in CT values compared to TNC. Substitution of TNC would significantly decrease patient’s radiation dose.
### Key points
• Measurement of epicardial adipose tissue (EAT) volume and attenuation are feasible on virtual non-contrast (VNC) series with excellent correlation to true non-contrast series (all R 2 >0.9).
• Differences in VNC algorithms have a significant impact on EAT volume and CT attenuation values.
• A novel VNC algorithm (VNC PC) enables reliable assessment of EAT volume and attenuation with superior accuracy compared to measurements on conventional VNC- and CCTA-series.
## Introduction
Epicardial adipose tissue (EAT) is the visceral fat located between the myocardial surface and the visceral layer of the pericardium [1]. Its extent and density are directly associated with the development and severity of a variety of cardiovascular and metabolic diseases, such as coronary artery disease, myocardial infarction, atrial fibrillation, or obesity-related insulin resistance [2–8].
EAT volume has been shown to be the most accurate measure to obtain EAT quantity, over thickness or area [7]. Echocardiography, cardiac magnetic resonance imaging (CMR), and cardiac computed tomography (CT) allow the non-invasive assessment of EAT quantity [9, 10]. However, echocardiography can only provide EAT thickness and CMR is time consuming with limited availability in clinical routine [11]. CT is already used for a wide range of cardiac examinations and provides highly reproducible, rapid EAT volume measurements on electrocardiographically (ECG) triggered true non-contrast (TNC) series [1]. Furthermore, not only the extent but also CT attenuation values within EAT volume were found to correlate with local and systemic inflammatory markers [12–14]. EAT volumetry is based on CT-value thresholds, varying from −250 to −190 HU and −50 to −30 HU, for the lower and upper threshold, respectively. By raising the upper threshold, EAT volumes can also be approximated on coronary CT angiography (CCTA) series [15]. Here it has been shown that an adjustment of the upper threshold from −30 to 0 HU on CCTA series provides more accurate EAT volumes compared to TNC values [16, 17].
The recent introduction of photon-counting detector CT (PCD-CT) systems with inherent spectral information on clinical routine scans now routinely enables several post-processing steps after data acquisition, including iodine removal from contrast-enhanced CT scans [18–21]. By now, two algorithms are available to create VNC series, conventional (VNCConv) and PureCalcium (VNCPC), that share a basic material differentiation into water and iodine. The VNCPC algorithm additionally performs a decomposition into iodine and calcium beforehand and was specifically designed to obtain full calcium contrast within the final image. Since none of the VNC algorithms specifically focus on decomposition into fat, adipose tissue is partly attributed to all base materials, and the attenuation values are expected to slightly differ from those of TNC [22]. The performance of the novel VNCPC algorithm on EAT quantification from CCTA scans has not yet been investigated.
In this study, we therefore sought to analyze VNC reconstructions derived from PCD-CCTA datasets for the assessment of EAT in comparison to reference TNC and CCTA series.
## Study population
The protocol for this retrospective single-center study was approved by the institutional review board (LMU Munich, project number 22-0456) with a waiver for written informed consent. Consecutive patients with a clinically indicated ECG-gated CT scan of the heart on a novel photon-counting detector CT (NAEOTOM Alpha, Siemens Healthineers) between $\frac{01}{2022}$ and $\frac{04}{2022}$ were included. Inclusion criteria were [1] age > 18 years, [2] pre-contrast TNC series for calcium scoring and contrast-enhanced CCTA series, and [3] availability of raw CT data for image reconstructions.
## Data acquisition
All patients received a pre-contrast scan for calcium scoring followed by a CCTA, at both 120 kV and a collimation of 144 × 0.4 mm. Reference tube current time product was adjusted by setting the image quality level to 19 for TNC and 60 for CTA. For the CTA, a triphasic contrast injection protocol with bolus tracking was used. In the first phase, 60 mL of nonionic iodinated contrast material (Iopromide 300 mgI/mL, Ultravist, Bayer) was injected followed by a $50\%$ diluted mixture of 30 mL contrast material and 30 mL normal saline solution and a saline chaser (25 mL). A flow of 5 mL/s was used in all three phases. By placing a region of interest in the descending aorta, bolus tracking was performed, and the scan was initiated 8 s after the enhancement reached 150 HU. If there was no clinical contraindication, 0.4 mg of nitroglycerin was administered sublingually 5 min prior to the scan and 5 mg of metoprolol was administered intravenously in patients with a heart rate of more than 70 bpm.
## Image reconstruction
All reconstructions were performed on a dedicated research workstation (ReconCT, Version 15.0.58331.0, Siemens Healthineers). For all patients, a TNC series based on the pre-contrast raw data, and a regular, a VNCConv, and VNCPC series based on the CTA were reconstructed, all at a virtual monochromatic level of 70 keV. For all reconstructions, a quantitative kernel Qr36 with a quantum iterative reconstruction algorithm with strength level 3 and a slice thickness/increment of $\frac{3.0}{1.5}$ mm was used. The VNC image series differ in the iodine removal algorithm. In both alternatives, a material decomposition into water and iodine is performed but the VNCPC algorithm takes some further steps beforehand to preserve the full calcium contrast in the final image. Emrich et al recently provided a detailed description of the VNCPC algorithm in [21].
## Image analysis
Image analyses were performed on a dedicated workstation (syngo.via version VB70A_CUT; Siemens Healthineers, using the CT Cardiac Risk Assessment application). For each patient and series, the fat volume in milliliter and the histogram of the attenuation values in HU within the semi-automatically segmented pericardial adipose tissue were measured. For all series, the lower threshold was set to −190 HU and the upper threshold to −30 HU [23–25]. To assess a potential underestimation of EAT volume on CTA series with a range of −190 to −30 HU (CTA-30), an additional measurement with an adapted upper threshold of 0 HU (CTA0) was performed [16, 17]. Figure 1 exemplarily shows a comparison of the segmentations, their volumes, and corresponding histograms. Only series with equal threshold range were considered in the analysis of the histograms, so CTA0 was excluded for reasons of inter-series comparability and similarity between CTA0 and CTA-30. Image noise was defined as standard deviation (SD) of CT values within the whole segmented EAT volume of the respective series. Fig. 1Demonstration of EAT segmentations, their volumes, and histograms. EAT = epicardial adipose tissue; CTA0 = CT angiography with an upper threshold of 0 HU; CTA-30 = CT angiography with an upper threshold of −30 HU; TNC = true non-contrast; VNCConv = conventional virtual non-contrast; VNCPC = PureCalcium virtual non-contrast
## Statistical analyses
Statistical analyses were performed using python (version 3.9.7). The Shapiro-Wilk test was used to test for normal distribution. The paired t-test and the Wilcoxon signed-rank test were used to assess differences for parametric and non-parametric data, respectively. In multiple comparisons, p values were adjusted using the Bonferroni method. Binary data are presented in frequencies (proportions) and continuous data with mean ± SD or as median with interquartile range (IQR) for parametric or non-parametric data, respectively. The coefficient of determination R2 was used to assess the accuracy of the linear regression predictions to approximate TNC measurements and serves as a correlation measure. Euclidean distance was used for quantitative comparison of the histograms, which is calculated as follows: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\left\Vert q-p\right\Vert}_2=\sqrt{\sum \limits_{$i = 1$}^n{\left({q}_i-{p}_i\right)}^2} $$\end{document}q−p2=∑$i = 1$nqi−pi2where q and p are the equal sized histograms with bin size 1 HU, n is the total number of bins (−190 to −30 HU = 161 bins), and i the respective bin at a certain CT-value. p values < 0.05 were considered to indicate statistically significant differences.
## Patient baseline characteristics
Sixty-six patients were primarily enrolled. Of these, 24 had to be excluded due to following reasons: missing non-contrast series ($$n = 12$$); missing CCTA series ($$n = 10$$); missing raw data ($$n = 2$$). The final study cohort comprised 42 patients (mean age 72 ± 10 years, 20 females). In non-contrast series, dose length product (DLP) and volumetric CT dose index (CTDIvol) were 34.3 (27.3–50.2) mGy∙cm and 1.7 (1.3–2.7) mGy. In CCTA, DLP and CTDIvol were 262.5 (95.4–503.5) mGy∙cm and 15.3 (5.3–33.5) mGy, respectively. The dose proportion of the pre-contrast scan corresponds to 12.9 (7.6–28.6)% and 13.1 (6.5–31.6)% of the total DLP and CTDIvol in all three phases. Table 1 summarizes the baseline study characteristics. Table 1Baseline study characteristicsTotal $$n = 42$$ClinicalAge, years72.0 ± 9.5Sex, female20 ($47.6\%$)CT radiation doseTNCCTACTDIvol, mGy1.7 (1.3–2.7)15.3 (5.4–33.5)DLP, mGy∙cm34.3 (27.3–50.2)262.5 (95.4–503.5)SSDE, mGy2.2 (1.9–3.3)22.0 (6.4–27.8)Effective mAs22 (18–26)37 (29.3–47)Values are mean ± standard deviation, median (interquartile range), or frequency (percentage). CT computed tomography, CTDIvol volumetric CT dose index, DLP dose length product, SSDE size-specific dose estimate
## EAT volume
Median EAT volume was measured 195.6 (122.6–268.4) mL on TNC series. Except for CTA0 measurements with a mean difference of +14.8 mL, corresponding to +10 % of the TNC volume, the volumes were significantly underestimated compared to TNC (Table 2 and Fig. 2). The mean differences were −26.9 mL and −29.1 mL in VNCConv and CTA-30, respectively, corresponding to −$12\%$ and −$13\%$ of the TNC volume. The most accurate measurement with the smallest difference in mean and standard deviation compared to volumes measured on TNC series was observed in VNCPC series with a mean difference of −5.7 mL, corresponding to a mean deviation of −$3\%$ to the TNC volume (Fig. 3). EAT volumes of CTA-30 and VNCConv did not significantly differ from each other (p value = 0.2). Table 2Epicardial adipose tissue volumes in mL on the respective image series and subgroup analyses including median differences in mL (and %), as well as the pairwise Wilcoxon p valueEAT volume, mL∆ EAT volume, mL p valueSeriesTNCVNCConvVNCPCCTA-30TNC195.6 (122.6–268.4)VNCConv177.6 (112.8–247.2)−26.9 (−$12\%$) < 0.001VNCPC189.5 (103.2–229.3)−5.7 (−$3\%$) 0.001−21.2 (−$12\%$) < 0.001CTA-30180.9 (103.2–229.2)−29.1 (−$13\%$) < 0.001−4.2 (−$1\%$) 0.2−23.4 (−$11\%$) < 0.001CTA0223.5 (131.6–306.6)+14.8 (+$10\%$) 0.001+40.5 (+$24\%$) < 0.001+20.5 (+$12\%$) < 0.001+43.9 (+$26\%$) < 0.001Volumes are median (IQR) and differences are mean (%). EAT epicardial adipose tissue, CTA0 CT angiography with an upper threshold of 0 HU, CTA-30 CT angiography with an upper threshold of − 30 HU, TNC true non-contrast, VNCConv conventional virtual non-contrast, VNCPC PureCalcium virtual non-contrastFig. 2Boxplot of the measured epicardial adipose tissue volume in mL. EAT = epicardial adipose tissue; CTA0 = CT angiography with an upper threshold of 0 HU; CTA-30 = CT angiography with an upper threshold of −30 HU; TNC = true non-contrast; VNCConv = conventional virtual non-contrast; VNCPC = PureCalcium virtual non-contrastFig. 3Mean difference plots between the EAT volumes in mL measured on TNC and the respective volumes measured on CTA and VNC. EAT = epicardial adipose tissue; CTA0 = CT angiography with an upper threshold of 0 HU; CTA-30 = CT angiography with an upper threshold of −30 HU; TNC = true non-contrast; VNCConv = conventional virtual non-contrast; VNCPC = PureCalcium virtual non-contrast.
In linear regression analyses, EAT volumes from all reconstructed series showed a strong positive correlation to the ground truth in TNC series (all R2 > 0.9). A near-perfect predictive accuracy was observed for EAT volumes measured on VNCPC series (R2 = 0.99) (Fig. 4). Fig. 4Linear regression plots between the EAT volumes in mL measured on TNC and the respective volumes measured on CTA and VNC. EAT = epicardial adipose tissue; CTA0 = CT angiography with an upper threshold of 0 HU; CTA-30 = CT angiography with an upper threshold of −30 HU; TNC = true non-contrast; VNCConv = conventional virtual non-contrast; VNCPC = PureCalcium virtual non-contrast EAT attenuation Mean attenuation within the EAT segmentation was −81.1 ± 5.8 HU, −75.4 ± 4.4 HU, −79.1 ± 5.9 HU, and −83.1 ± 8.3 HU for TNC, VNCConv, VNCPC, and CTA-30, respectively. Compared to TNC, CT values were significant higher on VNC series (+$6.6\%$ and +$2.3\%$ for VNCConv and VNCPC) and lower on CTA-30 series (−$2.1\%$). The noise level was 32.5 ± 2.0 HU, 31.0 ± 4.4 HU, 30.3 ± 2.4 HU, and 32.3 ± 3.6 HU for TNC, VNCConv, VNCPC, and CTA-30, respectively. Significant differences existed only between noise measured on VNCPC to TNC and CTA-30 (Table 3) (Fig. 5). Table 3Image noise as standard deviation of the CT values in HU, measured within the segmented epicardial adipose tissue volumes as well as p values of the pairwise t-testNoise, HUp valueSeriesVNCConvVNCPCCTA-30TNC32.5 ± 2.00.082< 0.0010.54VNCConv31.0 ± 4.40.130.32VNCPC30.3 ± 2.40.015CTA-3032.3 ± 3.6Values are mean ± standard deviation. CTA-30 CT angiography with an upper threshold of −30 HU, TNC true non-contrast, VNCConv conventional virtual non-contrast, VNCPC PureCalcium virtual non-contrastFig. 5A Boxplot of the mean CT values measured within the segmented EAT volumes. EAT = epicardial adipose tissue; CTA-30 = CT angiography with an upper threshold of −30 HU; TNC = true non-contrast; VNCConv = conventional virtual non-contrast; VNCPC = PureCalcium virtual non-contrast. B Boxplot of the standard deviation of CT values measured within the segmented EAT volumes. EAT = epicardial adipose tissue; CTA-30 = CT angiography with an upper threshold of −30 HU; TNC = true non-contrast; VNCConv = conventional virtual non-contrast; VNCPC = PureCalcium virtual non-contrast Figure 6 A shows the attenuation values within the segmented EAT volume divided by the total of voxel counts and averaged over all patients. The differences of the histograms represented by the Euclidean distance was greatest between TNC and VNCConv (0.029 ± 0.013) (Fig. 6B). Both distances, TNC-VNCPC and TNC-CTA-30, were significantly smaller (0.016 ± 0.007 and 0.017 ± 0.008, p’s <.05, for TNC-VNCPC and TNC-CTA-30, respectively) (Table 4). Fig. 6A Plots of the histograms divided by their total number of voxels and averaged over all patients for the respective image series. CTA-30 = CT angiography with an upper threshold of −30 HU; TNC = true non-contrast; VNCConv = conventional virtual non-contrast; VNCPC = PureCalcium virtual non-contrast. B Boxplots of the Euclidean distance between the histograms of TNC and the respective histograms of CTA-30 and VNC. CTA-30 = CT angiography with an upper threshold of −30 HU; TNC = true non-contrast; VNCConv = conventional virtual non-contrast; VNCPC = PureCalcium virtual non-contrastTable 4Euclidean distances between the normalized histograms of attenuation values within the epicardial adipose tissue volumes and p values of the pairwise t-testEuclidean distance, frequencyp valueSeries||VNCConv-TNC||2||VNCPC-TNC||2||VNCConv-TNC||20.029 ± 0.013||VNCPC-TNC||20.016 ± 0.007< 0.001||CTA-30-TNC||20.017 ± 0.0080.0020.54Values are mean ± standard deviation. CTA-30 CT angiography with an upper threshold of −30 HU, TNC true non-contrast, VNCConv conventional virtual non-contrast, VNCPC PureCalcium virtual non-contrast
## Discussion
This retrospective study evaluates the potential of substituting TNC series by VNC reconstructions derived from PCD-CCTA datasets for the quantification of EAT volume and its CT values. The main findings of this study are as follows: [1] VNC series derived from PCD-CT CCTA datasets enable consistent EAT volume measurements in comparison to reference TNC; [2] with TNC as ground truth, VNCPC shows superior and more consistent results for EAT volume compared to VNCConv, CTA-30, or CTA0; [3] the distribution of EAT attenuation values measured on VNC and CTA series significantly differs in comparison to TNC but the best agreement was observed for VNCPC.
Epicardial adipose tissue has gained attention as it has been associated with numerous pathologies. Correlations of EAT volume to atrial fibrillation, coronary artery disease, and sleep apnea syndrome have been reported as well as its ability to predict clinical coronary outcomes [2–8, 11]. CT can provide a rapid, reliable, and highly reproducible non-invasive assessment of EAT. Usually, cardiac CT already includes several series, of which the pre-contrast phase for calcium scoring is used to quantify EAT [1, 11]. The radiation exposure in CT acquisitions is a non-negligible disadvantage. To reduce radiation dose to a necessary minimum, there are a variety of approaches, one of which is to substitute the pre-contrast phase with a virtual non-contrast reconstruction based on the coronary CT angiography. With the introduction of a PCD-CT system that inherently provides spectral information for every scan, VNC series can be routinely reconstructed from every contrast-enhanced scan [26]. Studies have shown the suitability of VNC reconstructions for several applications, such as diagnosis of acute bleedings [27], coronary calcium quantification [21, 28], or in patients after endovascular aneurysm repair [29].
Our results show that EAT volume measurements for both the conventional and the novel VNC reconstructions have excellent correlation with the ground truth TNC, but also systematically underestimate. However, for VNCPC, the difference to TNC is negligibly small (−$3\%$). Further studies should be performed to investigate how this affects individual risk stratification by the application of specific volume thresholds. The underestimation can be attributed to the material differentiation into water and iodine which is performed to create VNC images. Since adipose tissue is partly split into both base materials, the CT values on the water image are systematically higher compared to TNC [22]. This effect can be seen especially in the positive shift of the VNCConv histogram. Nevertheless, many studies showed that VNC images mimic TNC very well for the vast majority of tissues examined. Sauter et al found an absolute difference of less than 10 HU for ROIs in aorta, liver, renal cortex, muscle, fluid, and also fat, measured on VNC images obtained from a dual layer detector CT system [30]. With photon-counting detector CT systems, similar results were found with a high quantitative and qualitative agreement of VNC and TNC [19, 31]. Although Choi et al observed an underestimation of fatty liver density on VNC, they did not find a significant diagnostic difference to TNC [32]. *In* general, the results of our study show that differences between the VNC algorithms have a measurable impact on EAT volume and attenuation, with a clearly superior assessment on VNCPC series.
Regarding EAT volumes obtained from CTA, an upper threshold of −30 HU resulted, as expected, in an underestimation compared to TNC. Xu et al found that an adapted upper threshold of −3 HU for measurements on CTAs result in statistical equivalent EAT volumes compared to TNC [17]. In this study, we tested an upper threshold of 0 HU for CTA (according to Marwan et al. [ 16]), and could not reproduce EAT volumes on TNC but overestimated them. One conceivable explanation could be that different contrast injection protocols lead to different CT value intervals between non-contrast and contrast scans. These intervals need to be analyzed individually and the threshold adjusted accordingly.
Using VNC or CTA for EAT volume measurement both pursue the same goal: to obviate the pre-contrast phase and thus reduce radiation dose, acquisition time, and cost. In our study, TNC on average accounted for $13\%$ of CTDIvol and DLP of the combined TNC and CCTA study, according to which a radiation dose reduction of approximately this percentage might be possible using the CTA or VNC approach. Processing of spectral CTA data promises the possibility for comprehensive diagnostic with minimal effort. The inherent enormous potential for many applications, such as monoenergetic imaging for artifact reduction, VNC series for calcium scoring, pure lumen for stenosis analysis, or iodine maps to measure iodine concentration, just to name a few, has already been evaluated for the most part in a number of studies [28, 33, 34]. This study shows that VNCPC reconstructions derived from PCD-CCTA datasets can reliably be used as a substitute for TNC to quantify EAT volume. In summary, the inherent spectral information obtained from PCD-CT scans should be used to the maximum extent to optimize each examination for the best possible diagnostic performance in each individual patient.
Of course, this study has its limitations: First, this study was carried out retrospectively and single-centered. Its findings must be confirmed by larger multi-centric studies. Second, only the two currently at our CT scanner available VNC algorithms were evaluated and future adjustments of the algorithms (e.g., by implementing the differentiation of water and fat) might lead to even more accurate results. Third, the possibility to adjust the upper threshold for quantifying the EAT volume on VNC or CCTA series was not fully exploited and might yield more consistent measurements.
In conclusion, novel VNCPC series derived from PCD-CCTA datasets can be used to assess EAT with consistent results with only minimal deviations to reference TNC and superior results compared to conventional VNC or CCTA series. Using VNCPC as a substitute for TNC might significantly reduce the applied radiation dose for the individual patient.
## References
1. Nagy E, Jermendy AL, Merkely B, Maurovich-Horvat P. **Clinical importance of epicardial adipose tissue**. *Arch Med Sci AMS* (2017.0) **13** 864-874. DOI: 10.5114/aoms.2016.63259
2. Ansaldo AM, Montecucco F, Sahebkar A. **Epicardial adipose tissue and cardiovascular diseases**. *Int J Cardiol* (2019.0) **278** 254-260. DOI: 10.1016/j.ijcard.2018.09.089
3. Brandt V, Bekeredjian R, Schoepf U. **Prognostic value of epicardial adipose tissue volume in combination with coronary plaque and flow assessment for the prediction of major adverse cardiac events**. *Eur J Radiol* (2022.0) **148** 110157. DOI: 10.1016/j.ejrad.2022.110157
4. Brandt V, Decker J, Schoepf UJ. **Additive value of epicardial adipose tissue quantification to coronary CT angiography–derived plaque characterization and CT fractional flow reserve for the prediction of lesion-specific ischemia**. *Eur Radiol* (2022.0) **32** 4243-4252. DOI: 10.1007/s00330-021-08481-w
5. Hatem SN, Sanders P. **Epicardial adipose tissue and atrial fibrillation**. *Cardiovasc Res* (2014.0) **102** 205-213. DOI: 10.1093/cvr/cvu045
6. Iacobellis G, Leonetti F. **Epicardial adipose tissue and insulin resistance in obese subjects**. *J Clin Endocrinol Metab* (2005.0) **90** 6300-6302. DOI: 10.1210/jc.2005-1087
7. Gorter PM, van Lindert ASR, de Vos AM. **Quantification of epicardial and peri-coronary fat using cardiac computed tomography; reproducibility and relation with obesity and metabolic syndrome in patients suspected of coronary artery disease**. *Atherosclerosis* (2008.0) **197** 896-903. DOI: 10.1016/j.atherosclerosis.2007.08.016
8. Goeller M, Achenbach S, Marwan M. **Epicardial adipose tissue density and volume are related to subclinical atherosclerosis, inflammation and major adverse cardiac events in asymptomatic subjects**. *J Cardiovasc Comput Tomogr* (2018.0) **12** 67-73. DOI: 10.1016/j.jcct.2017.11.007
9. van Woerden G, van Veldhuisen DJ, Gorter TM. **Importance of epicardial adipose tissue localization using cardiac magnetic resonance imaging in patients with heart failure with mid-range and preserved ejection fraction**. *Clin Cardiol* (2021.0) **44** 987-993. DOI: 10.1002/clc.23644
10. Parisi V, Petraglia L, Formisano R. **Validation of the echocardiographic assessment of epicardial adipose tissue thickness at the Rindfleisch fold for the prediction of coronary artery disease**. *Nutr Metab Cardiovasc Dis* (2020.0) **30** 99-105. DOI: 10.1016/j.numecd.2019.08.007
11. 11.Gaborit B, Sengenes C, Ancel P, et al (2017) Role of epicardial adipose tissue in health and disease: a matter of fat? In: Comprehensive physiology. John Wiley & Sons, Ltd, pp 1051–1082
12. Franssens BT, Nathoe HM, Leiner T. **Relation between cardiovascular disease risk factors and epicardial adipose tissue density on cardiac computed tomography in patients at high risk of cardiovascular events**. *Eur J Prev Cardiol* (2017.0) **24** 660-670. DOI: 10.1177/2047487316679524
13. Mahabadi AA, Balcer B, Dykun I. **Cardiac computed tomography-derived epicardial fat volume and attenuation independently distinguish patients with and without myocardial infarction**. *PLoS One* (2017.0) **12** e0183514. DOI: 10.1371/journal.pone.0183514
14. Monti CB, Capra D, Zanardo M. **CT-derived epicardial adipose tissue density: systematic review and meta-analysis**. *Eur J Radiol* (2021.0) **143** 109902. DOI: 10.1016/j.ejrad.2021.109902
15. Marwan M, Achenbach S. **Quantification of epicardial fat by computed tomography: why, when and how?**. *J Cardiovasc Comput Tomogr* (2013.0) **7** 3-10. DOI: 10.1016/j.jcct.2013.01.002
16. Marwan M, Koenig S, Schreiber K. **Quantification of epicardial adipose tissue by cardiac CT: influence of acquisition parameters and contrast enhancement**. *Eur J Radiol* (2019.0) **121** 108732. DOI: 10.1016/j.ejrad.2019.108732
17. Xu L, Xu Y, Coulden R. **Comparison of epicardial adipose tissue radiodensity threshold between contrast and non-contrast enhanced computed tomography scans: a cohort study of derivation and validation**. *Atherosclerosis* (2018.0) **275** 74-79. DOI: 10.1016/j.atherosclerosis.2018.05.013
18. Flohr T, Petersilka M, Henning A. **Photon-counting CT review**. *Phys Med* (2020.0) **79** 126-136. DOI: 10.1016/j.ejmp.2020.10.030
19. 19.Mergen V, Racine D, Jungblut L et al (2022) Virtual noncontrast abdominal imaging with photon-counting detector CT. Radiology. 10.1148/radiol.213260
20. Decker JA, Bette S, Scheurig-Muenkler C. **Virtual non-contrast reconstructions of photon-counting detector CT angiography datasets as substitutes for true non-contrast acquisitions in patients after EVAR—performance of a novel calcium-preserving reconstruction algorithm**. *Diagnostics* (2022.0) **12** 558. DOI: 10.3390/diagnostics12030558
21. 21.Emrich T, Aquino G, Schoepf U et al (2022) Coronary computed tomography angiography-based calcium scoring: in vitro and in vivo validation of a novel virtual noniodine reconstruction algorithm on a clinical, first-generation dual-source photon counting-detector system. Invest Radiol. 10.1097/RLI.0000000000000868
22. McCollough CH, Boedeker K, Cody D. **Principles and applications of multienergy CT: report of AAPM task group 291**. *Med Phys* (2020.0) **47** e881-e912. DOI: 10.1002/mp.14157
23. Nakazato R, Shmilovich H, Tamarappoo BK. **Interscan reproducibility of computer-aided epicardial and thoracic fat measurement from non-contrast cardiac CT**. *J Cardiovasc Comput Tomogr* (2011.0) **5** 172-179. DOI: 10.1016/j.jcct.2011.03.009
24. Wheeler GL, Shi R, Beck SR. **Pericardial and visceral adipose tissues measured volumetrically with computed tomography are highly associated in type 2 diabetic families**. *Invest Radiol* (2005.0) **40** 97-101. DOI: 10.1097/00004424-200502000-00007
25. Yoshizumi T, Nakamura T, Yamane M. **Abdominal fat: standardized technique for measurement at CT**. *Radiology* (1999.0) **211** 283-286. DOI: 10.1148/radiology.211.1.r99ap15283
26. Rajendran K, Petersilka M, Henning A. **First clinical photon-counting detector CT system: technical evaluation**. *Radiology* (2021.0) **303** 130-138. DOI: 10.1148/radiol.212579
27. Kahn J, Fehrenbach U, Böning G. **Spectral CT in patients with acute thoracoabdominal bleeding—a safe technique to improve diagnostic confidence and reduce dose?**. *Medicine (Baltimore)* (2019.0) **98** e16101. DOI: 10.1097/MD.0000000000016101
28. Schwarz F, Nance JW, Ruzsics B. **Quantification of coronary artery calcium on the basis of dual-energy coronary CT angiography**. *Radiology* (2012.0) **264** 700-707. DOI: 10.1148/radiol.12112455
29. Decker J, Bette S, Scheurig-Münkler C. **Virtual non-contrast reconstructions of photon-counting detector CT angiography datasets as substitutes for true non-contrast acquisitions in patients after EVAR—performance of a novel calcium-preserving reconstruction algorithm**. *Diagnostics* (2022.0) **12** 558. DOI: 10.3390/diagnostics12030558
30. Sauter AP, Muenzel D, Dangelmaier J. **Dual-layer spectral computed tomography: virtual non-contrast in comparison to true non-contrast images**. *Eur J Radiol* (2018.0) **104** 108-114. DOI: 10.1016/j.ejrad.2018.05.007
31. Niehoff JH, Woeltjen MM, Laukamp KR. **Virtual non-contrast versus true non-contrast computed tomography: initial experiences with a photon counting scanner approved for clinical use**. *Diagnostics* (2021.0) **11** 2377. DOI: 10.3390/diagnostics11122377
32. Choi MH, Lee YJ, Choi YJ, Pak S. **Dual-energy CT of the liver: true noncontrast vs. virtual noncontrast images derived from multiple phases for the diagnosis of fatty liver**. *Eur J Radiol* (2021.0) **140** 109741. DOI: 10.1016/j.ejrad.2021.109741
33. Wellenberg RHH, Boomsma MF, van Osch JAC. **Quantifying metal artefact reduction using virtual monochromatic dual-layer detector spectral CT imaging in unilateral and bilateral total hip prostheses**. *Eur J Radiol* (2017.0) **88** 61-70. DOI: 10.1016/j.ejrad.2017.01.002
34. 34.Sartoretti T, Mergen V, Jungblut L et al (2022) Liver iodine quantification with photon-counting detector CT: accuracy in an abdominal phantom and feasibility in patients. Acad Radiol. 10.1016/j.acra.2022.04.021
|
---
title: European Association for Endoscopic Surgery (EAES) consensus on Indocyanine
Green (ICG) fluorescence-guided surgery
authors:
- E. Cassinotti
- M. Al-Taher
- S. A. Antoniou
- A. Arezzo
- L. Baldari
- L. Boni
- M. A. Bonino
- N. D. Bouvy
- R. Brodie
- T. Carus
- M. Chand
- M. Diana
- M. M. M. Eussen
- N. Francis
- A. Guida
- P. Gontero
- C. M. Haney
- M. Jansen
- Y. Mintz
- S. Morales-Conde
- B. P. Muller-Stich
- K. Nakajima
- F. Nickel
- M. Oderda
- P. Parise
- R. Rosati
- M. P. Schijven
- G. Silecchia
- A. S. Soares
- S. Urakawa
- N. Vettoretto
journal: Surgical Endoscopy
year: 2023
pmcid: PMC10017637
doi: 10.1007/s00464-023-09928-5
license: CC BY 4.0
---
# European Association for Endoscopic Surgery (EAES) consensus on Indocyanine Green (ICG) fluorescence-guided surgery
## Body
In the last few years, with the growth and progressive spread of minimally invasive surgical techniques, several tools and instruments have been developed to enhance surgeons’ performance and patient safety and potentially decrease the risk of human errors [1]. Among these tools, such as high-definition visual systems like 4 K or 3D imaging, is Indocyanine Green (ICG) fluorescence-guided surgery (FGS), which is a modality of intraoperative imaging system that could significantly contribute to intraoperative anatomical navigation and improve decision-making during the surgical procedure [2, 3].
FGS is based on the ability of a dye (ICG) to emit a fluorescent signal when excited with a light source at a specific wavelength (near-infrared light spectrum of 700–900 nm). For several decades, clinical use of ICG has been reported in the assessment of hepatic blood flow, the assessment of choroidal blood flow and the measurement of cardiac output. ICG is rapidly and exclusively excreted into the bile. Due to its well-established clinical applications, relatively low cost, and extremely low toxic dose/reported allergic reactions, ICG is currently the most employed fluorophore in general surgery clinical settings [4–6]. As regards technology development, multiple near-infrared visual systems have already been developed, either using a laser beam or LED light sources, both for laparoscopic and robotic surgery, and it is to be expected that more systems will be developed and introduced in the near future. Furthermore, ICG fluorescence imaging has also been demonstrated to have a short learning curve, it does not require complex equipment in the operating room, and it is not time-consuming without interfering with the surgical workflow [7].
Since ICG fluorescence imaging is one of the most promising and rapidly developing technical innovations in surgery of the last decade, the clinical applications of this technology have expanded exponentially [4], including fluorescence cholangiography in laparoscopic cholecystectomy [8, 9], lymph node identification and mapping in oncologic surgery [10, 11] and bowel anastomotic perfusion assessment [12, 13].
In the last years, the number of studies published regarding ICG-FGS have rapidly grown, suggesting that this technology is safe, feasible and could represent a benefit for both surgeons by simplifying and guiding some procedures, and for patients, in terms of reducing post-operative complications. Nevertheless, there is still significant variability in clinical use and technical details of use, such as dose, concentration and timing of ICG administration. Additionally, there are issues regarding whether or not fluorescence-guided surgery could potentially be considered the standard of care in some surgical applications [14].
Therefore, the European Association of Endoscopic Surgery (EAES) sponsored this consensus development conference on the use of ICG Fluorescence-guided Surgery to critically review all available data on fluorescence imaging in abdominal surgery. The aim of this project was to provide consensus statements and to develop recommendations for the surgical community based on the available evidence and inputs of some of the most experienced European experts and opinion makers in this field.
## Abstract
### Background
In recent years, the use of Indocyanine Green (ICG) fluorescence-guided surgery during open and laparoscopic procedures has exponentially expanded across various clinical settings. The European Association of Endoscopic Surgery (EAES) initiated a consensus development conference on this topic with the aim of creating evidence-based statements and recommendations for the surgical community.
### Methods
An expert panel of surgeons has been selected and invited to participate to this project. Systematic reviews of the PubMed, Embase and Cochrane libraries were performed to identify evidence on potential benefits of ICG fluorescence-guided surgery on clinical practice and patient outcomes. Statements and recommendations were prepared and unanimously agreed by the panel; they were then submitted to all EAES members through a two-rounds online survey and results presented at the EAES annual congress, Barcelona, November 2021.
### Results
A total of 18,273 abstracts were screened with 117 articles included. 22 statements and 16 recommendations were generated and approved. In some areas, such as the use of ICG fluorescence-guided surgery during laparoscopic cholecystectomy, the perfusion assessment in colorectal surgery and the search for the sentinel lymph nodes in gynaecological malignancies, the large number of evidences in literature has allowed us to strongly recommend the use of ICG for a better anatomical definition and a reduction in post-operative complications.
### Conclusions
Overall, from the systematic literature review performed by the experts panel and the survey extended to all EAES members, ICG fluorescence-guided surgery could be considered a safe and effective technology. Future robust clinical research is required to specifically validate multiple organ-specific applications and the potential benefits of this technique on clinical outcomes.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00464-023-09928-5.
## Materials and methods
The objective of this Consensus on the use of ICG fluorescence-guided surgery was to provide evidence-based recommendations on the use of vision enhanced by ICG fluorescence compared with standard vision in different clinical settings. The scope of this project consisted of three main parts: (i) general topics, (ii) organ-specific data and (iii) ongoing trials. Within each of these topics, subcategories have been defined.
‘General topics’ included (a) cognitive load, (b) costs and cost-effectiveness.
‘Organ specific’ topics included (a) cholecystectomy, (b) perfusion assessment in colorectal surgery, (c) lymphatic mapping in colorectal surgery, (d) bariatric surgery, (e) spleen and adrenal surgery, (f) pancreatic surgery, (g) liver surgery, (h) perfusion assessment in Upper GI surgery, (i) lymphatic mapping in Upper GI surgery, (j) urology and (k) gynaecology.
## Research team and search strategy
An expert panel of surgeons functioned as the coordinating team (EC, AA, LB and NV); they formulated a list of questions related to each topic to be specifically addressed, which guided the literature research (Table 1).Table 1List of questions regarding the use of ICG-guided fluorescence surgery to be addressedGeneral topics1Does intraoperative ICG fluorescence introduce a higher cognitive load for the surgeon compared to standard laparoscopic systems?2What is the impact of intraoperative ICG fluorescence on costs?Organ-specific topics1Should ICG fluorescence versus standard surgery be used in laparoscopic cholecystectomy?1.1Does intraoperative ICG fluorescence provide better visualisation of biliary anatomy in laparoscopic cholecystectomy compared to standard light visualisation?1.2Does intraoperative ICG fluorescence reduce post-operative complications in laparoscopic cholecystectomy compared to standard light visualisation?2Should ICG fluorescence versus standard surgery be used in colorectal cancer surgery?2.1Does intraoperative ICG fluorescence angiography reduce morbidities in colorectal cancer surgery?2.2Should ICG fluorescence versus standard surgery be used for lymphadenectomy in colorectal cancer surgery?3Should ICG fluorescence versus standard surgery be used in bariatric surgery?4Should ICG fluorescence versus standard surgery be used in abdominal endocrine surgery?5Should ICG fluorescence versus standard surgery be used in pancreatic surgery?6Should ICG fluorescence versus standard surgery be used in liver surgery?6.1Does intraoperative ICG fluorescence reduce bile leaks in liver surgery?6.2Does intraoperative ICG fluorescence help in tumour identification and resection in liver surgery?7Should ICG fluorescence versus standard surgery be used in UpperGI cancer surgery?7.1Should ICG fluorescence versus standard surgery be used for lymphadenectomy in UpperGI cancer surgery?7.2Does intraoperative ICG fluorescence angiography reduce anastomotic leak in UpperGI cancer surgery8Should ICG fluorescence versus standard surgery be used in urologic surgery?9Should ICG fluorescence versus standard surgery be used in gynecologic surgery?
The coordinators invited 12 expert surgeons, members of the EAES research and technology committee with recognised expertise on the topic to join the panel of experts. Each was asked to nominate at least one young surgical researcher to participate. An international research team consisted of 12 young surgical researchers was formed to review and evaluate the existing literature on the use of ICG fluorescence-guided surgery. Each young researcher was mentored by an expert surgeon. The final list of topics was approved by the experts and subsequently divided among the teams.
All searches were performed in PubMed, Embase and Cochrane electronic libraries starting from September 2019 with no limitation regarding the year of publication or language. Due to the pandemic restrictions, the difficulty of the research group to meet in person, the literature search was extended and updated until November 2020. The composition of search strings has been discussed and approved by a librarian from the University of Milan. Search strings are provided in Supplementary Materials.
Study inclusion criteria were: Randomised Controlled Trials (RCTs), prospective and retrospective observational comparative studies. Case reports and non-comparative studies were excluded as well as studies in children under 12 years of age and papers not in the English language.
All search hits were screened by topic and reviewed by two team members for eligibility, based on title and abstract. If considered eligible, full-text articles were reviewed and summarised. In cases of disagreement, the coordinators acted as referees and made the final decision.
## Data extraction and appraisal of the methodological quality of the studies
Standardised data extraction forms were used across all topics. A uniform Excel database template for entering the data extracted from the selected papers was provided to all participants. Important outcome measures with respect to ICG fluorescence-guided surgery compared to standard light imaging were included in the template, such as operating time, conversion to open surgery, hospital stay, post-operative pain, adverse events, post-operative complications and mortality. The participants were encouraged to add any other outcome measure if necessary. The template contained predefined fields for noting important information for each study, like population characteristics, detailed information about the surgical procedure, the experience level of the surgeons, etc., and the results for each outcome, including the effect, size and statistical significance, where appropriate.
A PRISMA chart was completed for each literature search according to recommendations [15]. The methodological quality of included RCT was assessed using the Cochrane risk of bias score [16].
After data extraction, the teams worked out a presentation on their topic and a more comprehensive summary of their findings, including a flowchart of the selection of articles, a description of the population, a summary of the papers, conclusions, statements and recommendations.
If the number and quality of the studies included in the final analysis were considerate appropriate, metanalysis was conducted to answer PICO questions and prepare the related statements.
With each “Statement”, the level of evidence was given. The original Centre for Evidence-Based Medicine levels of evidence (LoE) system was used [17], which defines five levels, ranging from Level 1 (highest evidence) to Level 5 (lowest evidence). This tool allows for grading levels down on the basis of study quality, imprecision, inconsistency between studies, etc. It also allows to grade up in case of large or very large effect size.
With each ‘Recommendation’, the level of recommendation was given. These were graded as ‘strong’ or ‘weak’ or ‘no recommendation’ according to the GRADE system. GRADE is a systematic and explicit approach to judging the quality of evidence and strength of recommendations. GRADE specifically assesses methodological flaws, consistency of results across different studies, the generalizability of research results and treatment effectiveness. When data were considered sufficient, consensus statements were prepared by each team and scored with a grade of recommendation (GoR) [18–20].
## Consensus development process
A face-to-face first consensus meeting was held in Krakow on 20 January 2020 to present all findings and preliminary drafted consensus statements and recommendations, which were finalised during further virtual meetings. A modified Delphi method was used, as anonymity was not applicable in our situation [21, 22]. All statements and recommendations were shared with the proposed LoE and subjected to voting for agreement or disagreement. In the case of $100\%$ consensus, the statements and recommendations were accepted. Where there was a lack of Consensus, the research team responsible for that topic presented the underlying evidence and rationale for their statement. After discussion, further voting rounds were conducted until an agreement was reached.
All finalised recommendations and statements with LoE and GoR were planned to be presented at a dedicated session during the 28th EAES congress in Krakow 2020 to be voted by EAES delegates. Unfortunately, due to the Covid pandemic, the 2020 EAES congress was cancelled. As mentioned, the working group then organised further online meetings and updated the literature search up to November 2020. Previous results were revised based on updated literature. In April and June 2021 online survey among all EAES members was organised and consisted of two-rounds of voting, until an agreement greater than $75\%$ on each recommendation was obtained.
An online repository was created in order to obtain access to the full Consensus protocol, literature search strategies, PRISMA flow charts, and full text of articles included in the final analysis for each topic. PRISMA charts are provided in Supplementary Materials.
225 EAES members participated in the survey and voted for each recommendation: (a) Agree with the above-mentioned recommendation (b) Disagree or (c) Don’t know/No opinion.
For each topic analyzed, in addition to the statements and recommendations, a brief discussion on the results obtained has been reported.
## Results
The literature searches yielded 18,273 abstracts to be screened. In total, 117 articles were included and reviewed in detail to define 22 consensus statements and 16 recommendations. 227 EAES members completed first round online survey; a second-round survey, completed by 193 EAES members, was carried out only for recommendations that did not reach $75\%$ agreement.
## Cognitive load
The systematic literature search retrieved a total of 210 articles, which were independently screened by two experts (MJ and MS), with 10 cases of disagreement on inclusion, which were resolved during a discussion in the presence of a third independent collaborator. No articles were found eligible for inclusion.
Cognitive load and how it may be affected has not been studied when using fluorescence in laparoscopy to date. Studies about the use of fluorescence in laparoscopy report mostly validity and reliability and sometimes state that surgeons found fluorescence “easy to use”.
No studies reported experienced workload or used questionnaires like the NASA TLX to measure experienced cognitive load.
Recommendation With the currently available evidence, no recommendations regarding cognitive load can be provided.
## Cost-effectiveness
Due to the lack of articles specifically focussing on cost-effectiveness of Indocyanine Green fluorescence-guided surgery, no statements and recommendations were included in the EAES members survey for this topic.
Nevertheless, all members of the expert panel had recently participated and published a Health Technology Assessment (HTA) in order to investigate the impact of fluorescence surgery on costs and cost-effectiveness. In April 2020, Vettoretto et al., in cooperation with SICE (Italian Society of Endoscopic Surgery), designed a study where an HTA approach was implemented to investigate the economic, social, ethical, and organisational implications related to the adoption of ICG fluorescence-guided surgery compared to standard vision surgery [23].
With the support of a multidisciplinary team, qualitative and quantitative data were collected by means of literature evidence, validated questionnaires and self-reported interviews, considering the dimensions resulting from the EUnetHTA Core Model.
The multidisciplinary team included expert surgeons, healthcare economists, managerial engineers, HTA and methodology experts, statisticians and clinical engineers.
Systematic reviews were conducted to detect evidence in the literature concerning the use of ICG fluorescence-guided surgery in several clinical settings. To assess the costs of this technology, an activity-based costing analysis (ABC) was implemented to measure, record, and calculate both the cost and the performance of activities [24]. In colorectal, particularly in rectal surgery, stronger evidence (confirmed by the experts’ opinion and by real-world practice) supports a benefit in the use of ICG-FGS, with a significant reduction of complications, which could be translated into advantages in the length of hospitalisation. The final evaluation may depend on overcoming the phase of technological introduction, thus defining potential advantages and greater practicality of use.
The economic evaluation of the patient’s pathway was then integrated with cost-effectiveness and budget impact analyses. The cost-effectiveness evaluation was developed to define the technology presenting a better trade-off between the efficacy achieved and the costs absorbed. A budget impact analysis (BIA) was conducted to estimate the financial effects of both the use and the consequent spread of new healthcare technology in a setting with limited resources [25]. The budget impact analysis predicted a reduction in costs, thus freeing up some economic and organisational resources. From an economic point of view, results suggested the opportunity to achieve significant economic savings, ranging from 4 to $8\%$, even in a conservative scenario of analysis, showing that investments in this field could be feasible and sustainable.
Also, multiple HTA dimensions (safety, efficacy, equity etc.) were evaluated through specific qualitative questionnaires to gather clinicians’ perceptions regarding their ICG fluorescence-guided surgery use. Results showed that ICG could be the preferable solution from an effectiveness point of view (average value: 0.54 vs 2.14, p-value = 0.000). ICG Fluorescence would thus be favourable to patients’ reported outcomes, the detection rate, image quality, the visualisation of perfusion, the precision of the surgical technique, and the separation/discrimination between healthy and not healthy tissues. The use of ICG is perceived as improving the precision of the surgical technique, the identification of the blood vessels and the lymph node detection rate, allowing for better image quality compared with standard white light. Despite no static differences that emerged with regard to the safety aspect perceptions (average value: 0.81 vs 0.98, p-value > 0.05), ICG is related to a lower occurrence of surgical complications.
We refer to the consultation of the original study for further details on all aspects of this multidimensional HTA report [23].
## Cholecystectomy
Over the last several years, the use of ICG fluorescence-guided surgery during laparoscopic cholecystectomy has emerged as a new technology allowing real-time enhanced visualisation and identification of extra-hepatic biliary structures without the use of radiation. As a result, proper identification of vital structures and high-risk areas that should be observed until dissection enables the key landmarks to be localised, is facilitated. One advantage of performing fluorescent cholangiography (FC) routinely is the ability to recognise the common bile duct before dissection; this has proven to be useful not only in the normal course of the procedure but also serves as a precautionary measure in the presence of anatomical variations or in certain conditions (e. g., the presence of inflamed tissue/acute cholecystitis settings) posing an increased risk for iatrogenic injury [26].
A total of 936 records were identified. Following screening for eligibility, 37 articles (representing unique studies) were included for potential data extraction and assessment of the risk of bias. However, 28 studies were further excluded due to inadequate study design or incompletely reported outcome data. Nine studies containing 2763 patients met inclusion criteria: two RCTs, four prospective studies and three large retrospective studies conducted on a prospectively maintained database [27–35]. One of the RCTs was a single-centre non-inferiority trial comparing FC with standard x-ray intraoperative cholangiogram (IOC), while the other RCT was a multicentre trial on 639 patients comparing the efficacy of fluorescence-guided surgery with the use of conventional white light vision. Also, other prospective or retrospective case-match studies used white light as the comparator. Given the low incidence reported in the literature of bile duct injuries during laparoscopic cholecystectomy (0,5–$2\%$) [36], it is not possible to design a trial that uses this endpoint as the main outcome since, in order to demonstrate a possible statistical advantage in the use of FC, more than 4000 patients should be enrolled. Both RCTs used the rate of biliary structures visualisation as the primary endpoint: Dip et al. demonstrated that FC was statistically superior to white light in visualising extra-hepatic biliary ducts before surgical dissection of the Calot’s triangle; results were confirmed in the subgroup analysis of patients with BMI > 30 and surgery for acute cholecystitis; in the other RCT Lehrskov et al. showed that FC is not inferior to standard IOC in visualising critical junction between the cystic duct, common hepatic duct and common bile duct [33, 35].
The other studies analysed focussed on operative time, rate of conversion to open surgery and overall post-operative complications; no significant differences between the two techniques were reported in terms of complications, although they all showed that FC is a non-invasive adjunct to laparoscopic cholecystectomy, leading to improved patient outcomes with respect to operative times, decreased conversion to open procedures, and shorter length of hospitalisation [27–32, 34].
Statements i.Fluorescent cholangiography during laparoscopic cholecystectomy improves the identification of the extra-hepatic biliary anatomy before and after dissection of Calot’s triangle, compared with standard intraoperative imaging (LoE: high).ii. Fluorescent cholangiography during laparoscopic cholecystectomy may reduce operative time and conversion rate compared with standard intraoperative imaging (LoE: moderate/low)iii. Fluorescent cholangiography during laparoscopic cholecystectomy in obese patients may improve identification of the extra-hepatic biliary anatomy before and after dissection of Calot’s triangle, compared with standard intraoperative imaging (LoE: low)iv. Fluorescent cholangiography during laparoscopic cholecystectomy in case of acute cholecystitis may improve identification of the extra-hepatic biliary anatomy before and after dissection of Calot’s triangle, compared with standard intraoperative imaging (LoE: low) Recommendation We recommend the use of fluorescent cholangiography during laparoscopic cholecystectomy, whenever available, in order to improve the visualisation of the biliary structures.
Grade of recommendation: Strong.
This recommendation received a $75\%$ agreement on the first round of the online survey.
## ICG Fluorescence for perfusion assessment in colorectal surgery
Anastomotic leak is one of the most important complications following colorectal surgery. It is well established that one of the main reasons for an anatomic leak is insufficient tissue perfusion making the anastomosis not heal correctly. For this reason, in the last years, we have observed an increase in the number of centres using ICG to evaluate perfusion during colorectal anastomosis [37, 38].
Of 1612 papers screened in the systematic review in this field, we included in our qualitative analysis 54 trials, and from them, 30 were included in our quantitative analysis [39–68]. Twenty-five of them were retrospective studies, three prospective not RCTs and two RCTs. The risk of bias assessment did not highlight any significant bias, but for the definition of anastomotic leak.
The meta-analysis shows that the use of ICG is correlated with a reduction of events of anastomotic leakage, particularly in the rectum (RR = 0.32, IC $95\%$ 0.22–0.49, $p \leq 0.01$, I2 = $0\%$) (Fig. 1). Moreover, the literature research described a change of anastomotic line after ICG injection in $10.3\%$ of patients (10.2–$12.5\%$). The meta-analysis shows a reduction in the overall post-operative complications (RR = 0.67, IC $95\%$ 0.57–0.80, $p \leq 0.01$)). This is also true if we exclude from the list of complications the anastomotic leak (RR = 0.82, IC $95\%$ 0.69–0.98, $$p \leq 0.03$$). Moreover, the use of ICG to assess perfusion during colorectal surgery reduces the post-operative length of hospital stay (MD − 0.67, IC $95\%$: − 1.06–− 0.27, $p \leq 0.01$). The operative time does not increase when using ICG ($$p \leq 0.37$$). A protective stoma was performed in only $44\%$ of patients in the ICG group compared to $54\%$ of the control group ($$p \leq 0.45$$).Fig. 1Forest plot of ICG Fluorescence-guided surgery versus ICG- on anastomotic leakage in colorectal surgery Based on the literature research and meta-analytic results, we formulated the following statements and voted on the recommendations.
Statements i.The use of ICG fluorescence to assess perfusion during colorectal surgery significantly reduces the risk of anastomotic leak (LoE: high).ii. The use of ICG fluorescence in colorectal surgery can lead to a change in the resection line and/or refashioning the anastomosis (LoE: high).iii. The use of ICG fluorescence to assess tissue perfusion while performing laparoscopic or robotic colorectal anastomosis does not affect the operative time (LoE: high).iv. The use of ICG fluorescence to assess tissue perfusion in colorectal surgery reduces the length of hospital stay and overall morbidity (LoE: high).
Recommendations The use of ICG fluorescence in colorectal surgery to assess tissue perfusion is recommended in order to reduce the risk of anastomotic leak.
Grade of recommendation: Strong.
The use of ICG fluorescence in colorectal surgery to assess tissue perfusion is suggested in order to reduce overall morbidity.
Grade of recommendation: Weak.
## ICG Fluorescence for lymphatic mapping in colorectal surgery
Among possible clinical applications of ICG fluorescence-guided surgery, is nodal navigation and real-time lymphography for cancers [69]. Following submucosal, subserosal or intradermal injection, ICG disperses in lymph, binds to lipoproteins, and is drained via lymphatic pathways and nodes. The resulting ICG fluorescent lymphography is a tool that could be used to guide the surgeon in performing a more precise lymphadenectomy and resection, and it may be a better option for overall patient outcomes.
In colorectal surgery, ICG fluorescent lymphography has been reported for assessing lymphatic routes both to evaluate the presence and value of sentinel nodes and, especially in laparoscopic right colectomy and flexure cancers, to highlight a watershed area around main vascular branches and facilitate more precise mesenteric dissection [70].
The literature search identified 388 abstracts. After screening, 38 papers were assessed for eligibility, although many were excluded being case reports or pilot studies on few patients. 12 studies were included in the qualitative data analysis [71–82]. No RCTs were found. All studies were prospective studies (most of which were pilot or feasibility studies on a small cohort of patients), mainly conducted from 2016 onwards. Two studies were performed exclusively on right colon and flexure cancers, while the others included various colonic resections. ICG tracer injection was performed endoscopically in the submucosal peritumoral area in 4 series, while the other 8 authors injected the dye at the beginning of laparoscopic abdominal exploration in the subserosal layer.
The primary outcome was mainly the feasibility of ICG fluorescent lymphangiography for lymphatic mapping in colon cancer. In all cases, ICG lymphography resulted safe and feasible. Different rates of sensitivity and accuracy (positive and negative predictive value) of the technique have been reported in the studies. Seven studies focussed on sentinel node retrieval, in some cases with combined intraoperative histopathological analysis of the nodes, while eight studies evaluated lymphatic flow in the mesenteric area. A single study was conducted with a case-match comparison with a historical cohort of patients aiming to compare the overall number of lymph nodes removed with and without ICG fluorescence guidance: ICG lymphography has resulted in a higher number of nodes retrieved.
Particularly for this research topic, studies are very heterogeneous, both for surgical technique and outcomes analysed, and hardly comparable. Future studies are mandatory to optimise ICG fluorescence-guided lymphography in the colorectal cancer setting.
Statement ICG Fluorescent lymphatic mapping is safe and feasible to allow the identification of lymphatic anatomy during colectomy for cancer, although the clinical value is yet to be defined (LoE: low).
Recommendation We recommend further research to standardise the technique of fluorescent lymphatic mapping during colorectal surgery and to investigate its clinical benefit.
Grade of recommendation: Strong.
This recommendation received a $92\%$ agreement in the first round of the online survey.
## Bariatric surgery
The literature search identified 169 abstracts. Among 25 full texts assessed for eligibility, no RCT or prospective studies were found, and no studies could be included in the final analysis.
Regarding potential applications of ICG fluorescence-guided surgery in bariatrics, we report four retrospective studies conducted on a small cohort of patients; three of them dealt with the use of ICG fluorescence angiography to assess visceral perfusion in sleeve gastrectomy [83–85] and one pilot study reported experience with intra-operative leak test using a blend of methylene blue and indocyanine green during robotic gastric bypass surgery [86].
Hence, due to insufficient evidence, no statement could be made about ICG fluorescence in bariatric surgery.
Recommendation We recommend further research on the use of ICG fluorescence in bariatric surgery to assess its potential clinical benefits.
Grade of recommendation: Strong.
This recommendation received a $75\%$ agreement on the first round of the online survey.
## Spleen and adrenal surgery
The literature search identified 548 hits regarding spleen and adrenal surgery, although among them, 15 full texts were screened, and only 3 articles (case-series study and case reports) were assessed for eligibility. No RCT or prospective studies were found, and no studies could be included in the final analysis.
However, case reports and preliminary experiences showed multiple applications for implementation of ICG fluorescence imaging for surgery of the spleen and adrenal glands, such as clarification of spleen and adrenal vascular anatomy, margin identification in partial adrenalectomy and fluorescent angiography for spleen preservation in distal pancreatectomy [87–91].
Hence, due to insufficient evidence, no statement could be made for ICG fluorescence in spleen and adrenal surgery.
Recommendation We recommend further research on the use of ICG fluorescence in spleen and adrenal surgery to assess its potential clinical benefits.
Grade of recommendation: Strong.
This recommendation received $77\%$ agreement in the first round of the online survey.
## Pancreatic surgery
The literature search identified 1479 hits regarding pancreatic surgery, although among them, 24 full texts were screened, and only 1 article (retrospective study on 37 patients) was assessed for eligibility [92]. No RCT or prospective studies were found, and no studies could be included in the final analysis.
However, from our literature review, several case reports and preliminary experiences showed multiple potential applications of ICG fluorescence imaging to assist surgeons with real-time information in pancreatic surgery, such as tumour identification and tumour margin assessment, perfusion assessment of pancreatic and biliary anastomosis and identification of vascular anatomy [93–96].
Hence, due to insufficient evidence, no statement could be made about ICG fluorescence in pancreatic surgery.
Recommendation We recommend further research on the use of ICG fluorescence in pancreatic surgery to assess its potential clinical benefits.
Grade of recommendation: Strong.
This recommendation received an $81\%$ agreement in the first round of the online survey.
## Liver surgery
ICG fluorescence-guided surgery has gained popularity as intraoperative imaging modality in hepatobiliary surgery over the past decade, with a large number of studies conducted in Eastern countries, creating new interesting perspectives. Among multiple potential applications in this field, fluorescence imaging has proven to be helpful in identifying small subcapsular and superficial tumours but also to enhance deeper lesions identification and to obtain clear resection margins; it can also be used for visualizing extra-hepatic bile duct anatomy and hepatic segmental borders, increasing the accuracy and the easiness of open and minimally invasive hepatectomy especially for prevention of post-operative bile leaks [97, 98].
7536 abstracts were identified by literature search. Following elimination of duplicate records and elimination of articles meeting exclusion criteria, a total of 15 articles were included in final analysis, with only 1 RCT [99–113]. Eight studies compared conventional imaging (intraoperative ultrasound, IOUS) to ICG fluorescence imaging in identifying surface liver tumours. Overall ICG fluorescence was successfully able to identify superficial lesions, as small as 1–2 mm, that had previously not been identified preoperatively or with direct visualization. All studies agree that IOUS remains the gold standard, although some authors demonstrated that fluorescence imaging identified smaller lesions with higher accuracy than ultrasound and combining ICG with IOUS could significantly increase the sensitivity in locating superficial lesions. All authors also agree that ICG fluorescence for this application is highly reliable for tumors within 8–10 mm beneath the liver capsule, due to the limitations of infrared light to deeply penetrate into tissues.
As regards resection margins, nine studies analyzed the role of ICG fluorescence as a guidance during dissection: they showed that ICG can be especially beneficial in cases where liver tissue consistency is hardened secondary to other pathology, such as cirrhosis, making IOUS difficult and rendering tactile feedback unreliable; in fact, the lack of fluorescence in the normal tissue served as a guide for the dissection plane, allowing for higher R0 resection margin rates. Although two studies reported that ICG fluorescence technique might increase false positive rate of liver lesion detection due to the non-specific uptake of lesions which may include benign lesions [106, 108].
Two studies demonstrated the efficacy of the application of ICG intraoperatively for the identification of bile leakage following hepatic resection. The RCT by Kaibori et al. showed no post-operative bile leaks when evaluating with fluorescence while the standard leak test without fluorescence had $10\%$ leak rate. Marino et al. in their case matched study found that ICG fluorescence was able to identify bile leaks in $12\%$ of patients at the liver surface from resection; leaks were promptly sutured, and subsequently had no development of post-operative leaks [108, 114].
Statements i.ICG fluorescence-guided liver surgery can be useful for identifying more small superficial liver tumours (within 10 mm from the liver surface) compared to conventional imaging (LoE: Moderate).ii. ICG fluorescence-guided liver surgery can be useful to enhance the identification of deeper tumours during dissection (LoE: Low)iii. ICG fluorescence-guided liver surgery for primary liver tumours may help to achieve a better resection margin in comparison to intraoperative Ultrasound (IOUS) (LoE: Low).iv. ICG fluorescence-guided detection of liver lesions may result in a false positive rate of up to $25\%$ (LoE: Moderate).v. ICG fluorescence is useful for intraoperative detection and prevention of bile leaks from the cut liver surface when ICG is injected through the biliary tree (LoE: Strong).
Recommendations We recommend the use of ICG fluorescence in liver surgery to aid identification of bile leaks after liver resection.
Grade of recommendation: Weak.
This recommendation received a $71\%$ agreement on first round online survey.
We recommend the use of IOUS during liver surgery to complement the accuracy of ICG fluorescence newly detected lesions.
Grade of recommendation: Strong.
This recommendation received a $71\%$ agreement on first round online survey.
The use of ICG fluorescence in liver surgery may improve detection of superficial liver tumours.
Grade of recommendation: Strong.
This recommendation received a $72\%$ agreement on second-round online survey.
We recommend the use of ICG fluorescence in liver surgery may improve R0 resection rate for hepatic lesions.
Grade of recommendation: Weak.
This recommendation received a $62\%$ agreement on second-round online survey.
## ICG Fluorescence for perfusion assessment in Upper GI surgery
As well as for colorectal surgery, ensuring good visceral perfusion is probably the most important controllable factor in preventing anastomotic leakage. ICG fluorescent angiography has also been investigated for esophagectomy and gastrectomy, especially for perfusion assessment of the gastric conduit during esophagectomy, as the perfusion of the tube, especially in the proximal part, is solely based on the right gastroepiploic artery. ICG fluorescence might guide surgeons in estimating the blood supply of the gastric segment and identifying the optimal anastomotic site [114].
The literature search identified 188 abstracts. After screening, 30 papers were assessed for eligibility, although many were excluded being case reports or pilot studies on few patients. 9 studies were included in the final data analysis [115–123]. No RCTs were found, and 5 prospective and 4 retrospective studies were analysed. In all papers, ICG fluorescent angiography was applied to assess gastric conduit perfusion in minimally invasive Ivor-Lewis esophagectomies. Anastomotic leak (AL) rate has been evaluated as the primary outcome by all authors: five studies were designed as propensity score case-match comparison with historical series of standard esophagectomies; in all cases, AL rate was decreased in the ICG group compared to standard light vision (in 3 studies with statistical significance). These data were cross-referenced and confirmed by a late 2019 meta-analysis on the topic, where six trials that compared ICG fluorescence perfusion assessment with standard technique cases showed an AL rate risk reduction of $69\%$ [124]. Most studies reported a change of strategy on the planned anastomotic site, up to $25\%$ of cases, when ICG fluorescence was considered unsatisfactory; as regards perfusion evaluation, most recent studies also reported, as secondary outcomes, data on quantitative assessment of perfusion (especially in terms of evaluation of fluorescence intensity or time until acceptable subjective fluorescence was documented on the conduit).
Statement The use of ICG fluorescence to assess tissue perfusion may be effective in reducing the risk of a leak in esophago-gastric anastomosis (LoE: moderate/low).
Recommendations The use of ICG fluorescence is recommended to assess tissue perfusion in order to reduce the risk of anastomotic leak in esophago-gastric anastomosis.
Grade of recommendation: Weak.
This recommendation received a $72\%$ agreement in the first round of the online survey.
We recommend further research on the quantitative evaluation of ICG fluorescence in order to reduce subjective variability in perfusion assessment.
Grade of recommendation: Strong.
This recommendation received a $93\%$ agreement in the first round of the online survey.
## ICG Fluorescence for lymphatic mapping in Upper GI surgery
As already mentioned in addressing the role of ICG fluorescent lymphography for colorectal cancer, the possibility of real-time navigation of lymph nodes and lymphatic routes appears to be of great interest even more in Upper GI malignancies and might have significant clinical consequences. Especially for gastric cancer, several studies are available demonstrating that ICG is superior to both radioactive tracers and other probes used to date, showing high sensitivity in identifying not a single sentinel node but a group of lymph nodes and lymphatic channels that represents the first drainage stations from the tumour, which has been referred as the lymphatic basin. Over the years, the lymphatic basin concept has been investigated, especially in relation to early gastric cancer, focussing on ICG fluorescence lymphatic mapping aiming to customise surgical lymphadenectomy according to tumour T stage, patient condition and risk profile [125].
A total of 553 records were identified. Following screening for eligibility and according to inclusion/exclusion criteria established, 7 articles were included in the final analysis: 1 RCT, 4 prospective and 2 retrospective studies [126–132]. All studies evaluated ICG fluorescence lymphatic mapping for gastric cancer: in four studies, including the RCT, lymphography was performed in laparoscopic gastrectomy (both distal and D2 total gastrectomies), while in the other three prospective studies, surgical procedures were robotics. ICG injection in the peritumoral area was performed endoscopically in the submucosal layer in all studies, either intraoperatively or the day before surgery. The main outcome analysed was the number of removed lymph nodes; 5 studies, including the RCT conducted on 260 patients, demonstrated that ICG lymphatic mapping could noticeably improve lymphadenectomy (higher number of lymph nodes retrieved compared to white light standard imaging technique). No significant differences in post-operative complications were reported between the two techniques. Peri-operative outcomes were also reported as secondary outcomes in all papers where 2 studies demonstrated that ICG lymphography was significantly effective in reducing operative time and intraoperative blood loss compared to a standard light.
Statement During gastric cancer surgery, ICG fluorescent lymphatic mapping by endoscopic injection before surgery is safe and feasible and may lead to the identification and removal of a higher number of lymph nodes (LoE: moderate).
Recommendation The use of ICG fluorescent lymphatic mapping during gastric cancer surgery may be recommended to improve lymphadenectomy.
Grade of recommendation: Strong.
This recommendation received an $84\%$ agreement in the second-round of the online survey.
## Gynaecologic surgery
ICG fluorescence-guided imaging in gynaecologic surgery is used primarily for sentinel node dissection in endometrial and cervical cancer: indeed, accurate identification of sentinel lymph nodes in patients with cancer improves the detection of metastatic disease, and might decreases surgical morbidity. In this field, ICG lymphography has already proven to be a feasible, safe, time-efficient and reliable method for lymphatic mapping, with better bilateral detection rates; it would also avoid patients’ exposure to radioactive tracers, and for this reason, in some countries, ICG sentinel node mapping has already become the gold standard. Experience in vulvar cancer is more limited, with ICG used together with Tc-99 m as a dual tracer and alone in video endoscopic inguinal lymphadenectomy, while in early ovarian cancer, results are still preliminary but promising [133].
A total of 4260 records were identified. Following abstract screening for eligibility, 28 full texts were included for potential data extraction and assessment of the risk of bias. However, given the number and quality of studies found, 15 articles were finally included in the qualitative analysis: 2 RCTs and 13 prospective studies [134–148]. Both RCTs were comparing ICG versus methylene blue in sentinel nodes detection in cervical and uterine cancer; in particular, the FILM trial, published in the Lancet Oncology in 2018, was designed as a non-inferiority trial but ended up demonstrating that ICG mapping was superior to standard blue dye, being able to identify sentinel lymph nodes in a much larger proportion of patients, to detect at least one sentinel node and more effective in bilateral sentinel nodes identification.
It also has to be mentioned that in this setting the research and article screening was not conducted by a team of gynaecologists, however the analysis of the articles included and the creation of the statements was strongly based on a systematic review and consensus statement paper recently published on Annals of Surgical Oncology [133].
During the online survey, less than $60\%$ of EAES surgeon members showed agreement on this topic, while almost $40\%$ of them gave a “don’t know/no opinion” answer; for this reason, the expert panel decided not to run a second-round survey on this topic: since EAES members are mostly general/abdominal surgeons, we present hereby literature search results and expert’s discussion result, although no consensus was reached on Gynaecologic surgery setting.
Statements i.In surgery for endometrial, cervical and vulvar cancer, ICG fluorescent lymphatic mapping for sentinel node dissection and lymph nodes detection is safe and feasible (LoE: strong).ii. In surgery for endometrial, cervical and vulvar cancer, ICG fluorescent lymphatic mapping for sentinel node dissection and lymph nodes detection can be as effective as radioactive tracers and more effective than other dye tracers (LoE: strong) Recommendation We recommend the use of ICG fluorescence lymphatic mapping during surgery for endometrial and vulvar cancer.
Grade of recommendation: Strong.
This recommendation received a $50\%$ agreement on the first round of the online survey. No second-round survey has been performed for the above-mentioned reasons.
## Urologic surgery
As regards urologic surgery, ICG fluorescence imaging has been largely explored since this technology became available in robotic systems, which are widely employed in this surgical speciality. ICG fluorescence has been found to be useful during robotic partial nephrectomy in guiding selective/super-selective clamping of arteries, while differential fluorescence intensity may play a role in discerning between pathological and normal renal tissue resulting in the minimal renal parenchymal loss (only feasibility and preliminary studies available on this latter application). ICG guidance during robotic radical prostatectomy and cystectomy has been found to better-assist surgeons in identifying lymphatic drainage both for sentinel lymph node biopsy and for extended lymph node dissection, where several studies have shown, as for gastrointestinal and gynaecological tumours, a higher number of lymph nodes removed compared to the standard white light imaging [149].
A total of 394 records were identified. Following abstract screening for eligibility, 27 full texts were included for potential data extraction and assessment of the risk of bias. However, given the number and quality of studies found, 19 articles were finally included in qualitative analysis: 1 RCT, 13 prospective series, mainly with historic case-match comparison, and 5 large retrospective studies [150–168]. In eight studies, including the RCT, the object was robotic radical prostatectomy demonstrating that the use of ICG fluorescence imaging during extended pelvic lymph node dissection improves the identification of lymphatic drainage and tissue, resulting in a higher yield of lymph nodes compared to standard vision.
The other eleven studies analysed the role of fluorescence imaging in robotic partial nephrectomy, where ICG has been used to clarify vascular anatomy to perform selective clamping of the tumour-feeding vascular branches aiming to reduce ischemic renal trauma and potentially improve kidney function preservation. All studies reported that this procedure is safe and feasible and potentially leads to short-term renal functional outcomes.
As happened for the “gynaecology setting”, during the online survey, less than $60\%$ of EAES surgeon members showed agreement on this topic, while almost $40\%$ of them gave a “don’t know/no opinion” answer; for this reason, the expert panel decided not to run a second-round survey on this topic: since EAES members are mostly general/abdominal surgeons, we present hereby literature search results and expert’s discussion result, although no consensus was reached on Urologic surgery.
Also, for this topic, it is worth reporting that there is a large number of studies regarding ICG fluorescence-guided surgery applied to multiple fields and different surgical procedures and that in April 2020, it was published in the World Journal of Urology, an extensive systematic literature review to provide evidence-based expert recommendations on best practices in this field, to which we referred in our analysis [149, 169].
Statements i.ICG fluorescent lymphatic mapping for sentinel node dissection and lymph nodes detection during prostatectomy and cystectomy for cancer is safe and feasible (LoE: high).ii. ICG fluorescence lymphatic mapping in radical prostatectomy may lead to the identification and removal of a higher number of lymph nodes (LoE: moderate).iii. ICG fluorescence-guided robotic partial nephrectomy may offer better short-term renal functional outcomes by favouring selective clamping as compared to standard partial nephrectomy (LoE: low).iv. There is insufficient evidence to support the application of ICG fluorescence during robotic partial nephrectomy to differentiate renal tumours from normal kidney parenchyma (LoE: low).
Recommendations We recommend the use of ICG fluorescent lymphatic mapping during radical prostatectomy for the removal of a higher number of lymph nodes.
Grade of recommendation: Weak.
We recommend further research on the use of ICG fluorescence in urologic surgery to assess its potential clinical benefits.
Grade of recommendation: Strong.
This recommendation received a $50\%$ agreement on the first round of the online survey. No second-round survey has been performed for the above-mentioned reasons.
## Ongoing trials
At the time of writing, searching registries of privately and publicly funded clinical studies for the terms “minimally invasive surgery”, “laparoscopy”, “robotic” and “fluorescence” we found 17 ongoing trial registered on ClinicalTrials.gov: 6 in Europe, 3 in the United States, 2 in Asia, 1 in Turkey, one in South America, two in North America. As regards study design, two monocentric randomized controlled trials (RCTs), two multicentric RCTs, 13 monocentric observational trials are registered.
Four of them haven’t started recruiting yet. The remaining 13 trial are still recruiting (estimated studies completion date 2022–2024).
13 studies concern laparoscopic surgery, 4 the robotics. The main focus is oncological surgery (Upper GI, colorectal, prostate, hepatobiliary and lung cancer, peritoneal carcinomatosis, liver resection). Two non-oncological studies are focussed on hepatobiliary surgery and one more on minimally invasive general surgery.
Primary outcomes of the studies are: feasibility of ICG fluorescence imaging in laparoscopic and robotic surgery, the usefulness of ICG to guide lymphadenectomy in oncological surgery, enhanced anatomical visualization, primary tumour detection, localization of occult lesions, anastomotic leak prevention. Common secondary outcomes are: impact of ICG on perioperative complications, side effects after indocyanine green injection, surgical time, conversion rate, surgeon confidence, hospital stay.
## What is new in this Consensus paper?
This is the first Consensus on ICG fluorescence-guided surgery edited by the EAES. It covered the application of this technology to several different districts of interest, including urology and gynaecology. Compared to other guidelines available in the literature, this represents the literature-based opinion of a large group of endoscopic surgeons since the systematic analysis of the literature by the experts panel was followed by a two-rounds online survey extended to all EAES members.
## Implementation
The Consensus believes that it is feasible to successfully implement these recommendations into local practice and that the recommendations will be accepted by stakeholders. The main considerations regarding the implementation of this Consensus include costs and availability of the technology. In addition, some of the recommended techniques require specialized knowledge and skills. Finally, in order to achieve the full benefit of these recommendations, it is advised to standardize the techniques, for what it entails dose, concentration and route and timing of administrations of ICG depending on the different applications. The panel plans to survey physicians in the future in order to monitor and audit compliance with the recommendations put forth in this Consensus.
## Updating this Consensus
The EAES plans to repeat a comprehensive literature review in three years to reevaluate and identify new evidences. Particular attention will be paid to any future studies that specifically address the research recommendations proposed in this Consensus. A formal update will be generated when substantial literature is detected. When sufficient literature is available, the EAES will project to produce a to produce a structured guideline with summary evidence appraisal and a formal evidence-to-decision framework.
## Limitations of this Consensus
The main limitation of this *Consensus is* the low certainty of evidence for some of the key questions. In addition, being a Consensus, patients’ values were not actually obtained. On the contrary, the panel’s impression of their beliefs was used, based on experiences with patients. While the recommendations in this Consensus are based on the highest-level evidence meeting inclusion criteria, cost-effectiveness was not specifically addressed. Moreover, we were not able to take into account certain aspects of diversity, equity, and inclusion due to unavailability in the literature that was reviewed.
## Conclusions
The consensus conference proposed a wide number of recommended applications of ICG fluorescence-guided surgery aiming to patients’ benefit in different surgical specialties. These evidence-based recommendations are aimed to support safe diffusion of the technology. Whilst there are clear and strong evidence in certain areas to support its safety and effectiveness in improving clinical outcomes, further robust studies are required to improve the standardization of the techniques and to explore different possible applications.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (PDF 74 KB)Supplementary file2 (PDF 113 KB)Supplementary file3 (PDF 85 KB)Supplementary file4 (PDF 110 KB)Supplementary file5 (PDF 97 KB)Supplementary file6 (PDF 113 KB)Supplementary file7 (PDF 114 KB)Supplementary file8 (PDF 114 KB)Supplementary file9 (PDF 109 KB)Supplementary file10 (PDF 85 KB)Supplementary file11 (PDF 110 KB)Supplementary file12 (PDF 81 KB)Supplementary file13 (PDF 81 KB)Supplementary file14 (PDF 109 KB)Supplementary file15 (PDF 80 KB)Supplementary file16 (PDF 124 KB)Supplementary file17 (PDF 122 KB)Supplementary file18 (PDF 108 KB)Supplementary file19 (PDF 116 KB)Supplementary file20 (PDF 130 KB)Supplementary file21 (PDF 99 KB)Supplementary file22 (PDF 112 KB)Supplementary file23 (PDF 84 KB)Supplementary file24 (PDF 147 KB)
## References
1. Vettoretto N, Foglia E, Ferrario L, Arezzo A, Cirrocchi R, Cocorullo G. **Why laparoscopists may opt for three-dimensional view: a summary of the full HTA report on 3D versus 2D laparoscopy by S.I.C.E. (Società Italiana di Chirurgia Endoscopica e Nuove Tecnologie)**. *Surg Endosc* (2018.0) **32** 2986-2993. DOI: 10.1007/s00464-017-6006-y
2. Gioux S, Choi HS, Frangioni JV. **Image-guided surgery using invisible near-infrared light: fundamentals of clinical translation**. *Mol Imaging* (2010.0) **9** 237-255. DOI: 10.2310/7290.2010.00034
3. Diana M. **Enabling precision digestive surgery with fluorescence imaging**. *Transl Gastroenterol Hepatol* (2017.0) **2** 97. DOI: 10.21037/tgh.2017.11.06
4. Zelken JA, Tufaro AP. **Current trends and emerging future of indocyanine green usage in surgery and oncology: an update**. *Ann Surg Oncol* (2015.0) **22** S1271-S1283. DOI: 10.1245/s10434-015-4743-5
5. Fox I, Wood E. **Indocyanine green: physical and physiologic properties**. *Proc Staff Meet Mayo Clin* (1960.0) **7** 13
6. Alander JT, Kaartinen I, Laakso A, Patila T, Spillmann T, Tuchin VV. **A review of indocyanine green fluorescent imaging in surgery**. *Int J Biomed Imaging* (2012.0) **2012** 940585. DOI: 10.1155/2012/940585
7. Reinhart MB, Huntington CR, Blair LJ, Heniford BT, Augenstein VA. **Indocyanine green: historical context, current applications, and future considerations**. *Surg Innov* (2016.0) **23** 166-175. DOI: 10.1177/1553350615604053
8. Agnus V, Pesce A, Boni L, Van Den Bos J, Morales-Conde S, Paganini AM. **Fluorescence-based cholangiography: preliminary results from the IHU-IRCAD-EAES EURO-FIGS registry**. *Surg Endosc* (2020.0) **34** 3888-3896. DOI: 10.1007/s00464-019-07157-3
9. Vlek SL, van Dam DA, Rubinstein SM, de Lange-de Klerk ESM, Schoonmade LJ, Tuynman JB. **Biliary tract visualization using near-infrared imaging with indocyanine green during laparoscopic cholecystectomy: results of a systematic review**. *Surg Endosc* (2017.0) **31** 2731-2742. DOI: 10.1007/s00464-016-5318-7
10. Skubleny D, Dang JT, Skulsky S, Switzer N, Tian C, Shi X. **Diagnostic evaluation of sentinel lymph node biopsy using indocyanine green and infrared or fluorescent imaging in gastric cancer: a systematic review and meta-analysis**. *Surg Endosc* (2018.0) **32** 2620-2631. DOI: 10.1007/s00464-018-6100-9
11. Emile SH, Elfeki H, Shalaby M, Sakr A, Sileri P, Laurberg S. **Sensitivity and specificity of indocyanine green near-infrared fluorescence imaging in detection of metastatic lymph nodes in colorectal cancer: systematic review and meta-analysis**. *J Surg Oncol* (2017.0) **116** 730-740. DOI: 10.1002/jso.24701
12. van den Bos J, Al-Taher M, Schols RM, van Kuijk S, Bouvy ND, Stassen LPS. **Near-infrared fluorescence imaging for real-time intraoperative guidance in anastomotic colorectal surgery: a systematic review of literature**. *J Laparoendosc Adv Surg Tech* (2018.0) **28** 157-167. DOI: 10.1089/lap.2017.0231
13. Arezzo A, Bonino MA, Ris F, Boni L, Cassinotti E, Foo DCC. **Intraoperative use of fluorescence with indocyanine green reduces anastomotic leak rates in rectal cancer surgery: an individual participant data analysis**. *Surg Endosc* (2020.0) **34** 4281-4290. DOI: 10.1007/s00464-020-07735-w
14. Ds AV, Lin H, Henderson ER, Samkoe KS, Pogue BW. **Review of fluorescence guided surgery systems: identification of key performance capabilities beyond indocyanine green imaging**. *J Biomed Opt* (2016.0) **21** 80901. DOI: 10.1117/1.JBO.21.8.080901
15. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JP. **The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration**. *J Clin Epidemiol* (2009.0) **62** e1-e34. DOI: 10.1016/j.jclinepi.2009.06.006
16. Higgins JPT, Altman DG, Gotzsche PC, Jüni P, Moher D, Oxman AD. **The Cochrane collaboration’s tool for assessing risk of bias in randomised trials**. *BMJ* (2011.0) **343** d5928. DOI: 10.1136/bmj.d5928
17. 17.CEBM (2015) OCEBM Levels of Evidence | CEBM. http://www.cebm.net/ocebm-levels-of-evidence/
18. Guyatt GH, Oxman AD, Kunz R, Falck-Ytter Y, Vist GE, Liberati A, Schünemann HJ. **Going from evidence to recommendations**. *BMJ* (2008.0) **336** 1049-1051. DOI: 10.1136/bmj.39493.646875.AE
19. Goldet G, Howick J. **Understanding GRADE: an introduction**. *J Evid Based Med* (2013.0) **6** 50-54. DOI: 10.1111/jebm.12018
20. Atkins D, Best D, Briss PA, Eccles M, Falck-Ytter Y, Flottorp S. **Grading quality of evidence and strength of recommendations**. *BMJ* (2004.0) **328** 1490. DOI: 10.1136/bmj.328.7454.1490
21. Boulkedid R, Abdoul H, Loustau M, Sibony O, Alberti C. **Using and reporting the Delphi method for selecting healthcare quality indicators: a systematic review**. *PLoS ONE* (2011.0) **6** e20476. DOI: 10.1371/journal.pone.0020476
22. Hasson F, Keeney S, McKenna H. **Research guidelines for the Delphi survey technique**. *J Adv Nurs* (2000.0) **32** 1008-1015. PMID: 11095242
23. Vettoretto N, Foglia E, Ferrario L, Gerardi C, Molteni B, Nocco U. **Could fluorescence-guided surgery be an efficient and sustainable option? A SICE (Italian Society of Endoscopic Surgery) health technology assessment summary**. *Surg Endosc* (2020.0) **34** 3270-3284. DOI: 10.1007/s00464-020-07542-3
24. Raffish N. **Glossary of activity-based management**. *J Cost Manage* (1991.0) **5** 53-63
25. Mauskopf JA, Sullivan SD, Annemans L, Caro J, Daniel Mullins C, Nuijten M. **Principles of good practice for budget impact analysis: report of the ISPOR Task Force on good research practices— budget impact analysis**. *Value Health* (2007.0) **10** 336-344. DOI: 10.1111/j.1524-4733.2007.00187.x
26. Pesce A, Piccolo G, La Greca G, Fabbri N, Diana M, Feo CV. **Utility of fluorescent cholangiography during laparoscopic cholecystectomy: a systematic review**. *World J Gastroenterol* (2015.0) **21** 7877-7883. DOI: 10.3748/wjg.v21.i25.7877
27. Osayi SN, Wendling MR, Drosdeck JM, Chaudhry UI, Perry KA, Noria SF. **Near-infrared fluorescent cholangiography facilitates identification of biliary anatomy during laparoscopic cholecystectomy**. *Surg Endosc* (2015.0) **29** 368-375. DOI: 10.1007/s00464-014-3677-5
28. van Dam DA, Ankersmit M, van de Ven P, van Rijswijk AS, Tuynman JB, Meijerink WJHJ. **Comparing near-infrared imaging with indocyanine green to conventional imaging during laparoscopic cholecystectomy: a prospective crossover study**. *J Laparoendosc Adv Surg Tech* (2015.0) **25** 486-492. DOI: 10.1089/lap.2014.0248
29. Diana M, Soler L, Agnus V, D’Urso A, Vix M, Dallemagne B. **Prospective evaluation of precision multimodal gallbladder surgery navigation: virtual reality, near-infrared fluorescence, and x-ray-based intraoperative cholangiography**. *Ann Surg* (2017.0) **266** 890-897. DOI: 10.1097/SLA.0000000000002400
30. Sharma S, Huang R, Hui S, Smith MC, Chung PJ, Schwartzman A. **The utilization of fluorescent cholangiography during robotic cholecystectomy at an inner-city academic medical center**. *J Robot Surg* (2018.0) **12** 481-485. DOI: 10.1007/s11701-017-0769-y
31. Quaresima S, Balla A, Palmieri L, Seitaj A, Fingerhut A, Ursi P. **Routine near infra-red indocyanine green fluorescent cholangiography versus intraoperative cholangiography during laparoscopic cholecystectomy: a case-matched comparison**. *Surg Endosc* (2019.0) **34** 1959-1967. DOI: 10.1007/s00464-019-06970-0
32. Ambe PC, Plambeck J, Fernandez-Jesberg V, Zarras K. **The role of indocyanine green fluoroscopy for intraoperative bile duct visualization during laparoscopic cholecystectomy: an observational cohort study in 70 patients**. *Patient Saf Surg* (2019.0) **13** 2. DOI: 10.1186/s13037-019-0182-8
33. Dip F, LoMenzo E, Sarotto L, Phillips E, Todeschini H, Nahmod M. **Randomized trial of near infrared incisionless fluorescent cholangiography**. *Ann Surg* (2019.0) **270** 992-999. DOI: 10.1097/SLA.0000000000003178
34. Broderick RC, Lee AM, Cheverie JN, Zhao B, Blitzer RR, Patel RJ. **Fluorescent cholangiography significantly improves patient outcomes for laparoscopic cholecystectomy**. *Surg Endosc* (2021.0) **35** 5729-5739. DOI: 10.1007/s00464-020-08045-x
35. Lehrskov LL, Westen M, Larsen SS, Jensen AB, Kristensen BB, Bisgaard T. **Fluorescence or x-ray cholangiography in elective laparoscopic cholecystectomy: a randomized clinical trial**. *Br J Surg* (2020.0) **107** 655-661. DOI: 10.1002/bjs.11510
36. Pucher PH, Brunt LM, Davies N, Linsk A, Munshi A, Alejandro Rodriguez H. **Outcome trends and safety measures after 30 years of laparoscopic cholecystectomy: a systematic review and pooled data analysis**. *Surg Endosc* (2018.0) **32** 2175-2183. DOI: 10.1007/s00464-017-5974-2
37. Meyer J, Joshi H, Buchs NC, Ris F, Davies J. **Fluorescence angiography likely protects against anastomotic leak in colorectal surgery: a systematic review and meta-analysis of randomised controlled trials**. *Surg Endosc* (2022.0). DOI: 10.1007/s00464-022-09255-1
38. Blanco-Colino R, Espin-Basany E. **Intraoperative use of ICG fluorescence imaging to reduce the risk of anastomotic leakage in colorectal surgery: a systematic review and meta-analysis**. *Tech Coloproctol* (2018.0) **22** 15-23. DOI: 10.1007/s10151-017-1731-8
39. Alekseev M, Rybakov E, Shelygin Y, Chernyshov S, Zarodnyuk I. **A study investigating the perfusion of colorectal anastomoses using fluorescence angiography: results of the FLAG randomized trial**. *Colorectal Dis* (2020.0) **22** 1147-1153. DOI: 10.1111/codi.15037
40. Bonadio L, Iacuzzo C, Cosola D, Cipolat Mis T, Giudici F, Casagranda B. **Indocyanine green-enhanced fluorangiography (ICGf) in laparoscopic extraperitoneal rectal cancer resection**. *Updates Surg* (2020.0) **72** 477-482. DOI: 10.1007/s13304-020-00725-6
41. Boni L, Fingerhut A, Marzorati A, Rausei S, Dionigi G, Cassinotti E. **Indocyanine green fluorescence angiography during laparoscopic low anterior resection: results of a case-matched study**. *Surg Endosc* (2017.0) **31** 1836-1840. DOI: 10.1007/s00464-016-5181-6
42. Brescia A, Pezzatini M, Romeo G, Cinquepalmi M, Pindozzi F, Dall’Oglio A. **Indocyanine green fluorescence angiography: a new ERAS item**. *Updates Surg* (2018.0) **70** 427-432. DOI: 10.1007/s13304-018-0590-9
43. Buxey K, Lam F, Muhlmann M, Wong S. **Does indocyanine green improve the evaluation of perfusion during laparoscopic colorectal surgery with extracorporeal anastomosis?**. *ANZ J Surg* (2019.0) **89** E487-E491. DOI: 10.1111/ans.15320
44. Chang YK, Foo CC, Yip J, Wei R, Ng KK, Lo O. **The impact of indocyanine-green fluorescence angiogram on colorectal resection**. *Surgeon* (2019.0) **17** 270-276. DOI: 10.1016/j.surge.2018.08.006
45. De Nardi P, Elmore U, Maggi G, Maggiore R, Boni L, Cassinotti E. **Intraoperative angiography with indocyanine green to assess anastomosis perfusion in patients undergoing laparoscopic colorectal resection: results of a multicenter randomized controlled trial**. *Surg Endosc* (2020.0) **34** 53-60. DOI: 10.1007/s00464-019-06730-0
46. Dinallo AM, Kolarsick P, Boyan WP, Protyniak B, James A, Dressner RM. **Does routine use of indocyanine green fluorescence angiography prevent anastomotic leaks? A retrospective cohort analysis**. *Am J Surg* (2019.0) **218** 136-139. DOI: 10.1016/j.amjsurg.2018.10.027
47. Foo CC, Ng KK, Tsang J, Wei R, Chow F, Chan TY. **Colonic perfusion assessment with indocyanine-green fluorescence imaging in anterior resections: a propensity score-matched analysis**. *Tech Coloproctol* (2020.0) **24** 935-942. DOI: 10.1007/s10151-020-02232-7
48. Gröne J, Koch D, Kreis ME. **Impact of intraoperative microperfusion assessment with pinpoint perfusion imaging on surgical management of laparoscopic low rectal and anorectal anastomoses**. *Colorectal Dis* (2015.0) **17** 22-28. DOI: 10.1111/codi.13031
49. Hasegawa H, Tsukada Y, Wakabayashi M, Nomura S, Sasaki T, Nishizawa Y. **Impact of intraoperative indocyanine green fluorescence angiography on anastomotic leakage after laparoscopic sphincter-sparing surgery for malignant rectal tumors**. *Int J Colorectal Dis* (2020.0) **35** 471-480. DOI: 10.1007/s00384-019-03490-0
50. Hayami S, Matsuda K, Iwamoto H, Ueno M, Kawai M, Hirono S. **Visualization and quantification of anastomotic perfusion in colorectal surgery using near-infrared fluorescence**. *Tech Coloproctol* (2019.0) **23** 973-980. DOI: 10.1007/s10151-019-02089-5
51. Higashijima J, Shimada M, Yoshikawa K, Miyatani T, Tokunaga T, Nishi M. **Usefulness of blood flow evaluation by indocyanine green fluorescence system in laparoscopic anterior resection**. *J Med Invest* (2019.0) **66** 65-69. DOI: 10.2152/jmi.66.65
52. Impellizzeri HG, Pulvirenti A, Inama M, Bacchion M, Marrano E, Creciun M. **Near-infrared fluorescence angiography for colorectal surgery is associated with a reduction of anastomotic leak rate**. *Updates Surg* (2020.0) **72** 991-998. DOI: 10.1007/s13304-020-00758-x
53. Jafari MD, Wexner SD, Martz JE, McLemore EC, Margolin DA, Sherwinter DA. **Perfusion assessment in laparoscopic left-sided/anterior resection (PILLAR II): a multi-institutional study**. *J Am Coll Surg* (2015.0) **220** 82-92. DOI: 10.1016/j.jamcollsurg.2014.09.015
54. Jafari MD, Hong Lee K, Halabi WJ, Mills SD, Carmichael JC, Stamos MJ. **The use of indocyanine green fluorescence to assess anastomotic perfusion during robotic assisted laparoscopic rectal surgery**. *Surg Endosc* (2013.0) **27** 3003-3008. DOI: 10.1007/s00464-013-2832-8
55. Kawada K, Hasegawa S, Wada T, Takahashi R, Hisamori S, Hida K. **Evaluation of intestinal perfusion by ICG fluorescence imaging in laparoscopic colorectal surgery with DST anastomosis**. *Surg Endosc* (2017.0) **31** 1061-1069. DOI: 10.1007/s00464-016-5064-x
56. Kin C, Vo H, Welton L, Welton M. **Equivocal effect of intraoperative fluorescence angiography on colorectal anastomotic leaks**. *Dis Colon Rectum* (2015.0) **58** 582-587. DOI: 10.1097/DCR.0000000000000320
57. Kudszus S, Roesel C, Schachtrupp A, Höer JJ. **Intraoperative laser fluorescence angiography in colorectal surgery: a noninvasive analysis to reduce the rate of anastomotic leakage**. *Langenbecks Arch Surg* (2010.0) **395** 1025-1030. DOI: 10.1007/s00423-010-0699-x
58. Liot E, Assalino M, Buchs NC, Schiltz B, Douissard J, Morel P. **Does near-infrared (NIR) fluorescence angiography modify operative strategy during emergency procedures?**. *Surg Endosc* (2018.0) **32** 4351-4356. DOI: 10.1007/s00464-018-6226-9
59. Mizrahi I, Abu-Gazala M, Rickles AS, Fernandez LM, Petrucci A, Wolf J. **Indocyanine green fluorescence angiography during low anterior resection for low rectal cancer: results of a comparative cohort study**. *Tech Coloproctol* (2018.0) **22** 535-540. DOI: 10.1007/s10151-018-1832-z
60. Nishigori N, Koyama F, Nakagawa T, Nakamura S, Ueda T, Inoue T. **Visualization of lymph/blood flow in laparoscopic colorectal cancer surgery by ICG Fluorescence Imaging (Lap-IGFI)**. *Ann Surg Oncol* (2016.0) **23** S266-S274. DOI: 10.1245/s10434-015-4509-0
61. Ris F, Liot E, Buchs NC, Kraus R, Ismael G, Belfontali V. **Near-infrared anastomotic perfusion assessment network VOIR. Multicentre phase II trial of near-infrared imaging in elective colorectal surgery**. *Br J Surg* (2018.0) **105** 1359-1367. DOI: 10.1002/bjs.10844
62. Shapera E, Hsiung RW. **Assessment of anastomotic perfusion in left-sided robotic assisted colorectal resection by indocyanine green fluorescence angiography**. *Minim Invasive Surg* (2019.0) **2019** 3267217. DOI: 10.1155/2019/3267217
63. Skrovina M, Bencurik V, Martinek L, Machackova M, Bartos J, Andel P. **The significance of intraoperative fluorescence angiography in miniinvasive low rectal resections**. *Videosurgery Miniinv* (2020.0) **15** 43-48. DOI: 10.5114/wiitm.2019.84851
64. Spinelli A, Carvello M, Kotze PG, Maroli A, Montroni I, Montorsi M. **Ileal pouch-anal anastomosis with fluorescence angiography: a case-matched study**. *Colorectal Dis* (2019.0) **21** 827-832. DOI: 10.1111/codi.14611
65. Su H, Wu H, Bao M, Luo S, Wang X, Zhao C. **Indocyanine green fluorescence imaging to assess bowel perfusion during totally laparoscopic surgery for colon cancer**. *BMC Surg* (2020.0) **20** 102. DOI: 10.1186/s12893-020-00745-4
66. Wada T, Kawada K, Hoshino N, Inamoto S, Yoshitomi M, Hida K. **The effects of intraoperative ICG fluorescence angiography in laparoscopic low anterior resection: a propensity score-matched study**. *Int J Clin Oncol* (2019.0) **24** 394-402. DOI: 10.1007/s10147-018-1365-5
67. Watanabe J, Ishibe A, Suwa Y, Suwa H, Ota M, Kunisaki C. **Indocyanine green fluorescence imaging to reduce the risk of anastomotic leakage in laparoscopic low anterior resection for rectal cancer: a propensity score-matched cohort study**. *Surg Endosc* (2020.0) **34** 202-208. DOI: 10.1007/s00464-019-06751-9
68. Wojcik M, Doussot A, Manfredelli S, Duclos C, Paquette B, Turco C. **Intra-operative fluorescence angiography is reproducible and reduces the rate of anastomotic leak after colorectal resection for cancer: a prospective case-matched study**. *Colorectal Dis* (2020.0) **22** 1263-1270. DOI: 10.1111/codi.15076
69. Burghgraef TA, Zweep AL, Sikkenk DJ, van der Pas MHGM, Verheijen PM, Consten CJ. **In vivo sentinel lymph node identification using fluorescent tracer imaging in colon cancer. A systematic review and meta-analysis**. *Crit Rev Oncol Hematol* (2021.0) **158** 103149. DOI: 10.1016/j.critrevonc.2020.103149
70. Keller DS, Joshi HM, Rodriguez-Justo M, Walsh D, Coffey C, Chand M. **Using fluorescence lymphangiography to define the ileocolic mesentery: proof of concept for the watershed area using real-time imaging**. *Tech Coloproctol* (2017.0) **21** 757-760. DOI: 10.1007/s10151-017-1677-x
71. Andersen HS, Bjorne Bennedsen AL, Kobbelgaard Burgdorf S, Ravn Eriksen J, Eiholm S, Toxvaerd A. **In vivo and ex vivo sentinel node mapping does not identify the same lymph nodes in colon cancer**. *Int J Colorectal Dis* (2017.0) **32** 983-990. DOI: 10.1007/s00384-017-2777-9
72. Ankersmit M, Bonjer HJ, Hannink G, Schoonmade LJ, van der Pas MHGM, Meijerink WJHJ. **Near-infrared fluorescence imaging for sentinel lymph node identification in colon cancer: a prospective single-center study and systematic review with meta-analysis**. *Tech Coloproctol* (2019.0) **23** 1113-1126. DOI: 10.1007/s10151-019-02107-6
73. Carrara A, Motter M, Amabile D, Pellecchia L, Moscatelli P, Pertile R. **Predictive value of the sentinel lymph node procedure in the staging of non-metastatic colorectal cancer**. *Int J Colorectal Dis* (2020.0) **35** 1921-1928. DOI: 10.1007/s00384-020-03654-3
74. Chand M, Keller DS, Joshi HM, Devoto L, Rodriguez-Justo M, Cohen R. **Feasibility of fluorescence lymph node imaging in colon cancer: FLICC**. *Tech Coloproctol* (2018.0) **22** 271-277. DOI: 10.1007/s10151-018-1773-6
75. Currie AC, Brigic A, Thomas-Gibson S, Suzuki N, Moorghen M, Jenkins JT. **A pilot study to assess near infrared laparoscopy with indocyanine green (ICG) for intraoperative sentinel lymph node mapping in early colon cancer**. *Eur J Surg Oncol* (2017.0) **43** 2044-2051. DOI: 10.1016/j.ejso.2017.05.026
76. Hirche C, Mohr Z, Kneif S, Doniga S, Murawa D, Strik M. **Ultrastaging of colon cancer by sentinel node biopsy using fluorescence navigation with indocyanine green**. *Int J Colorectal Dis* (2012.0) **27** 319-324. DOI: 10.1007/s00384-011-1306-5
77. Nagata K, Endo S, Hidaka E, Tanaka JI, Kudo SE, Shiokawa A. **Laparoscopic sentinel node mapping for colorectal cancer using infrared ray laparoscopy**. *Anticancer Res* (2006.0) **26** 2307-2311. PMID: 16821607
78. Nishigori N, Koyama F, Nakagawa T, Nakamura S, Ueda T, Inoue T. **Visualization of lymph/blood flow in laparoscopic colorectal cancer surgery by ICG Fluorescence Imaging (Lap-IGFI)**. *Ann Surg Oncol* (2015.0) **23** S266-S274. DOI: 10.1245/s10434-015-4509-0
79. Park SY, Park JS, Kim HJ, Woo IT, Park IK, Choi GS. **Indocyanine green fluorescence imaging-guided laparoscopic surgery could achieve radical D3 dissection in patients with advanced right-sided colon cancer**. *Dis Colon Rectum* (2020.0) **63** 441-449. DOI: 10.1097/DCR.0000000000001597
80. Ushijima H, Kawamura J, Ueda K, Yane Y, Yoshioka Y, Daito K. **Visualization of lymphatic flow in laparoscopic colon cancer surgery using indocyanine green fluorescence imaging**. *Sci Rep* (2020.0) **10** 14274. DOI: 10.1038/s41598-020-71215-3
81. Watanabe J, Ota M, Suwa Y, Ishibe A, Masui H, Nagahori K. **Evaluation of lymph flow patterns in splenic flexural colon cancers using laparoscopic real-time indocyanine green fluorescence imaging**. *Int J Colorectal Dis* (2017.0) **32** 201-207. DOI: 10.1007/s00384-016-2669-4
82. Yeung TM, Wang LM, Colling R, Kraus R, Cahill R, Hompes R. **Intraoperative identification and analysis of lymph nodes at laparoscopic colorectal cancer surgery using fluorescence imaging combined with rapid OSNA pathological assessment**. *Surg Endosc* (2018.0) **32** 1073-1076. DOI: 10.1007/s00464-017-5644-4
83. Di Furia M, Romano L, Salvatorelli A, Brandolin D, Lomanto D, Cianca G. **Indocyanine green fluorescent angiography during laparoscopic sleeve gastrectomy: preliminary results**. *Obes Surg* (2019.0) **29** 3786-3790. DOI: 10.1007/s11695-019-04085-y
84. Ortega CB, Guerron AD, Yoo JS. **The use of fluorescence angiography during laparoscopic sleeve gastrectomy**. *JSLS* (2018.0) **22** e2018.00005. DOI: 10.4293/JSLS.2018.00005
85. Frattini F, Lavazza M, Mangano A, Amico F, Rausei S, Rovera F. **Indocyanine green-enhanced fluorescence in laparoscopic sleeve gastrectomy**. *Obes Surg* (2015.0) **25** 949-950. DOI: 10.1007/s11695-015-1640-8
86. Hagen ME, Diaper J, Douissard J, Jung MK, Buehler L, Aldenkortt F. **Early experience with intraoperative leak test using a blend of methylene blue and indocyanine green during robotic gastric bypass surgery**. *Obes Surg* (2019.0) **29** 949-952. DOI: 10.1007/s11695-018-03625-2
87. Colvin J, Zaidi N, Berber E. **The utility of indocyanine green fluorescence imaging during robotic adrenalectomy**. *J Surg Oncol* (2016.0) **114** 153-156. DOI: 10.1002/jso.24296
88. DeLong JC, Chakedis JM, Hosseini A. **Indocyanine green (ICG) fluorescence-guided laparoscopic adrenalectomy**. *J Surg Oncol* (2015.0) **112** 650-653. DOI: 10.1002/jso.24057
89. Sound S, Okoh AK, Bucak E. **Intraoperative tumor localization and tissue distinction during robotic adrenalectomy using indocyanine green fluorescence imaging: a feasibility study**. *Surg Endosc* (2016.0) **30** 657-662. DOI: 10.1007/s00464-015-4256-0
90. Kawasaki Y, Maemura K, Kurahara H, Mataki Y, Iino S, Sakoda M. **Usefulness of fluorescence vascular imaging for evaluating splenic perfusion**. *ANZ J Surg* (2018.0) **88** 1017-1021. DOI: 10.1111/ans.14364
91. Fujino H, Nagayama M, Kimura Y, Imamura M, Nobuoka T, Takemasa I. **Indocyanine green fluorescence imaging ensures perfusion of the remnant stomach during laparoscopic splenectomy in a patient after distal gastrectomy: a case report**. *Int J Surg Case Rep* (2021.0) **84** 106111. DOI: 10.1016/j.ijscr.2021.106111
92. Rho SY, Kim JS, Chong JU, Hwang HK, Yoon DS, Lee WJ. **Indocyanine green perfusion imaging-guided laparoscopic pancreaticoduodenectomy: potential application in retroperitoneal margin dissection**. *J Gastrointest Surg* (2018.0) **22** 1470-1474. DOI: 10.1007/s11605-018-3760-7
93. Kou HW, Yu MC, Chong SW, Hsu HY, Chou HH, Lee CW. **Successful localization and resection of small pancreatic cystic insulinoma using intraoperative near-infrared fluorescence imaging: a case report and literature review**. *Pancreas* (2020.0) **49** 1388-1392. DOI: 10.1097/MPA.0000000000001678
94. Newton AD, Predina JD, Shin MH, Frenzel-Sulyok LG, Vollmer CM, Drebin JA. **Intraoperative near-infrared imaging can identify neoplasms and aid in real-time margin assessment during pancreatic resection**. *Ann Surg* (2019.0) **270** 12-20. DOI: 10.1097/SLA.0000000000003201
95. Oba A, Inoue Y, Sato T, Ono Y, Mise Y, Ito H. **Impact of indocyanine green-fluorescence imaging on distal pancreatectomy with celiac axis resection combined with reconstruction of the left gastric artery**. *HPB (Oxford)* (2019.0) **21** 619-625. DOI: 10.1016/j.hpb.2018.09.023
96. Iguchi T, Iseda N, Hirose K, Ninomiya M, Honboh T, Maeda T. **Indocyanine green fluorescence to ensure perfusion in middle segment-preserving pancreatectomy: a case report**. *Surg Case Rep* (2021.0) **7** 262. DOI: 10.1186/s40792-021-01344-y
97. Rossi G, Tarasconi A, Baiocchi G, De Angelis GL, Gaiani F, Di Mario F. **Fluorescence guided surgery in liver tumors: applications and advantages**. *Acta Biomed* (2018.0) **89** 135-140. DOI: 10.23750/abm.v89i9-S.7974
98. Wang X, Teh CSC, Ishizawa T, Aoki T, Cavallucci D, Lee SY. **Consensus guidelines for the use of fluorescence imaging in hepatobiliary surgery**. *Ann Surg* (2021.0) **274** 97-106. DOI: 10.1097/SLA.0000000000004718
99. Abo T, Nanashima A, Tobinaga S, Hidaka S, Taura N, Takagi K. **Usefulness of intraoperative diagnosis of hepatic tumors located at the liver surface and hepatic segmental visualization using indocyanine green-photodynamic eye imaging**. *Eur J Surg Oncol* (2015.0) **41** 257. DOI: 10.1016/j.ejso.2014.09.008
100. Kose E, Kahramangil B, Aydin H, Donmez M, Takahashi H, Acevedo-Moreno LA. **A comparison of indocyanine green fluorescence and laparoscopic ultrasound for detection of liver tumors**. *HPB (oxford)* (2020.0) **22** 764-769. DOI: 10.1016/j.hpb.2019.10.005
101. Zhang YM, Shi R, Hou JC, Liu ZR, Cui ZL, Li Y. **Liver tumor boundaries identified intraoperatively using real-time indocyanine green fluorescence imaging**. *J Cancer Res Clin Oncol* (2017.0) **1** 51-58. DOI: 10.1007/s00432-016-2267-4
102. Boogerd LS, Handgraaf H, Lam HD, Huurman VA, Farina-Sarasqueta A, Frangioni JV. **Laparoscopic detection and resection of occult liver tumors of multiple cancer types using real-time near-infrared fluorescence guidance**. *Surg Endosc* (2017.0) **31** 952-961. DOI: 10.1007/s00464-016-5007-6
103. van der Vorst JR, Schaafsma B, Hutteman M, Verbrrk FP, Leifers GJ, Hartgrink HH. **Near-infrared fluorescence-guided resection of colorectal liver metastases**. *Cancer* (2013.0) **119** 3411-3418. DOI: 10.1002/cncr.28203
104. Takahashi H, Zaidi N, Berber E. **An intial report on the intraoperative use of indocyanine green fluorescence imaging in the surgical management of liver tumors**. *J Surg Oncol* (2016.0) **114** 625-629. DOI: 10.1002/jso.24363
105. Purich K, Dang J, Poonja A, Sun WYL, Bigam D, Birch D. **Intraoperative fluorescence imaging with indocyanine green in hepatic resection for malignancy: a systematic review and meta-analysis of diagnostic test accuracy studies**. *Surg Endosc* (2020.0) **34** 2891-2903. DOI: 10.1007/s00464-020-07543-2
106. Handgraaf HJM, Boogerd L, Hoppener DJ, Peloso A, Sibinga Mulder BG, Hoogstins CES. **Long-term follow-up after near-infrared fluorescence-guided resection of colorectal liver metastases: a retrospective multicenter analysis**. *Eur J Surg Oncol* (2017.0) **43** 1463-1471. DOI: 10.1016/j.ejso.2017.04.016
107. Marino MV, Di Saverio S, Podda M, Gomez Ruiz M, Gomez Fleitas M. **The application of indocyanine green fluorescence imaging during robotic liver resection: a case-matched study**. *World J Surg* (2019.0) **43** 2595-2606. DOI: 10.1007/s00268-019-05055-2
108. Zhou Y, Lin Y, Jin H, Hou B, Yu M, Yin Z. **Real-time navigation guidance using fusion indocyanine green fluorescence imaging in laparoscopic non-anatomical hepatectomy of hepatocellular carcinomas at segments 6,7, or 8 (with Videos)**. *Med Sci* (2019.0) **26** 1512-1217
109. Liu B, Liu T, Su M, Ma YQ, Zhang BF, Wang YF. **Improving the surgical effect for primary liver cancer with intraoperative fluorescence navigation compared with intraoperative ultrasound**. *Med Sci Monit* (2019.0) **8** 3406-3416. DOI: 10.12659/MSM.916423
110. Aoki T, Murakami M, Koizumi T, Matsuda K, Fujimori A, Kusano T. **Determination of the surgical margin in laparoscopic liver resections using infrared indocyanine green fluorescence**. *Langenbecks Arch Surg* (2018.0) **403** 671-680. DOI: 10.1007/s00423-018-1685-y
111. Chiow AKH, Rho S, Wee IJY, Lee LS, Choi GH. **Robotic ICG guided anatomical liver resection in a multi-center cohort: an evolution from "positive staining" into "negative staining" method**. *HPB (Oxford)* (2021.0) **23** 475-482. DOI: 10.1016/j.hpb.2020.08.005
112. Marino MV, Podda M, Fernandez CC, Ruiz MG, Fleitas MG. **The application of indocyanine green-fluorescence imaging during robotic-assisted liver resection for malignant tumor: a single arm feasibility cohort study**. *HPB (Oxford)* (2020.0) **22** 422-431. DOI: 10.1016/j.hpb.2019.07.013
113. Kaibori M, Ishizaki M, Matsui K, Kwon AH. **Intraoperative indocyanine green fluorescent imaging for prevention of bile leakage after hepatic resection**. *Surgery* (2011.0) **150** 91-98. DOI: 10.1016/j.surg.2011.02.011
114. Bertani C, Cassinotti E, Della Porta M, Pagani M, Boni L, Baldari L. **Indocyanine green—a potential to explore: narrative review**. *Ann Laparosc Endosc Surg* (2022.0) **7** 9. DOI: 10.21037/ales-21-5
115. Campbell C, Reames MK, Robinson M, Symanowski J, Salo JC. **Conduit vascular evaluation is associated with reduction in anastomotic leak after Esophagectomy**. *J Gastrointest Surg* (2015.0) **19** 806-812. DOI: 10.1007/s11605-015-2794-3
116. Karampinis I, Ronellenfitsch U, Mertens C, Gerken A, Hetjens S, Post S. **Indocyanine green tissue angiography affects anastomotic leakage after esophagectomy. A retrospective, case-control study**. *Int J Surg* (2017.0) **48** 210-214. DOI: 10.1016/j.ijsu.2017.11.001
117. Dalton BGA, Ali AA, Crandall M, Awad ZT. **Near infrared perfusion assessment of gastric conduit during minimally invasive Ivor Lewis esophagectomy**. *Am J Surg* (2018.0) **216** 524-527. DOI: 10.1016/j.amjsurg.2017.11.026
118. Noma K, Shirakawa Y, Kanaya N, Okada T, Maeda N, Ninomiya T. **Visualized evaluation of blood flow to the gastric conduit and complications in esophageal reconstruction**. *J Am Coll Surg* (2018.0) **226** 241-251. DOI: 10.1016/j.jamcollsurg.2017.11.007
119. Ohi M, Toiyama Y, Mohri Y, Saigusa S, Ichikawa T, Shimura T. **Prevalence of anastomotic leak and the impact of indocyanine green fluorescein imaging for evaluating blood flow in the gastric conduit following esophageal cancer surgery**. *Esophagus* (2017.0) **14** 351-359. DOI: 10.1007/s10388-017-0585-5
120. Kumagai Y, Hatano S, Sobajima J, Ishiguro T, Fukuchi M, Ishibashi KI. **Indocyanine green fluorescence angiography of the reconstructed gastric tube during Esophagectomy efficacy of the second rule**. *Dis Esophagus* (2018.0) **1** 31. DOI: 10.1093/dote/doy052
121. Ishige F, Nabeya Y, Hoshino I, Takayama W, Chiba S, Arimitsu H. **Quantitative assessment of the blood perfusion of the gastric conduit by indocyanine green imaging**. *J Surg Res* (2019.0) **234** 303-310. DOI: 10.1016/j.jss.2018.08.056
122. Talavera-Urquijo E, Parise P, Palucci M, Olivari G, Turi S, Cossu A. **Perfusion speed of indocyanine green in the stomach before tubulization is an objective and useful parameter to evaluate gastric microcirculation during Ivor-Lewis esophagectomy**. *Surg Endosc* (2020.0) **34** 5649-5659. DOI: 10.1007/s00464-020-07924-7
123. Slooter MD, de Bruin DM, Eshuis WJ, Veelo DP, van Dieren S, Gisbertz SS. **Quantitative fluorescence-guided perfusion assessment of the gastric conduit to predict anastomotic complications after esophagectomy**. *Dis Esophagus* (2021.0) **34** 1-8. DOI: 10.1093/dote/doaa100
124. Ladak F, Dang JT, Switzer N, Mocanu V, Tian C, Birch D. **Indocyanine green for the prevention of anastomotic leaks following esophagectomy: a meta-analysis**. *Surg Endosc* (2019.0) **33** 384-394. DOI: 10.1007/s00464-018-6503-7
125. He M, Jiang Z, Wang C, Hao Z, An J, Shen J. **Diagnostic value of near-infrared or fluorescent indocyanine green guided sentinel lymph node mapping in gastric cancer: a systematic review and meta-analysis**. *J Surg Oncol* (2018.0) **118** 1243-1256. DOI: 10.1002/jso.25285
126. Kwon IG, Son T, Kim HI, Hyung WJ. **Fluorescent lymphography-guided lymphadenectomy during robotic radical gastrectomy for gastric cancer**. *JAMA Surg* (2019.0) **154** 150-158. DOI: 10.1001/jamasurg.2018.4267
127. Cianchi F, Indennitate G, Paoli B, Ortolani M, Lami G, Manetti N. **The clinical value of fluorescent lymphography with indocyanine green during robotic surgery for gastric cancer: a matched cohort study**. *J Gastrointest Surg* (2020.0) **24** 2197-2203. DOI: 10.1007/s11605-019-04382-y
128. Lan YT, Huang KH, Chen PH, Liu CA, Lo SS, Wu CW. **A pilot study of lymph node mapping with indocyanine green in robotic gastrectomy for gastric cancer**. *SAGE Open Med* (2017.0) **5** 1-8. DOI: 10.1177/2050312117727444
129. Chen QY, Xie JW, Zhong Q, Wang JB, Lin JX, Lu J. **Safety and efficacy of indocyanine green tracer-guided lymph node dissection during laparoscopic radical gastrectomy in patients with gastric cancer: a randomized clinical trial**. *JAMA Surg* (2020.0) **155** 300-311. DOI: 10.1001/jamasurg.2019.6033
130. Park SH, Berlth F, Choi JH, Park JH, Suh YS, Kong SH. **Near-infrared fluorescence-guided surgery using indocyanine green facilitates secure infrapyloric lymph node dissection during laparoscopic distal gastrectomy**. *Surg Today* (2020.0) **50** 1187-1196. DOI: 10.1007/s00595-020-01993-w
131. Ma S, Zhang YM, Dou LZ, Liu H, Ma FH, Wang GQ. **Efficacy and feasibility of indocyanine green for mapping lymph nodes in advanced gastric cancer patients undergoing laparoscopic distal gastrectomy**. *J Gastrointest Surg* (2020.0) **24** 2306-2309. DOI: 10.1007/s11605-020-04706-3
132. Liu M, Xing J, Xu K, Yuan P, Cui M, Zhang C. **Application of near-infrared fluorescence imaging with indocyanine green in totally laparoscopic distal gastrectomy**. *J Gastric Cancer* (2020.0) **20** 290-299. DOI: 10.5230/jgc.2020.20.e25
133. Zapardiel I, Alvarez J, Barahona M, Barri P, Boldo A, Bresco P. **Utility of intraoperative fluorescence imaging in gynecologic surgery: systematic review and consensus statement**. *Ann Surg Oncol* (2021.0) **28** 3266-3278. DOI: 10.1245/s10434-020-09222-x
134. Backes FJ, Cohen D, Salani R, Cohn DE, O’Malley DM, Fanning E. **Prospective clinical trial of robotic sentinel lymph node assessment with isosulfane blue (ISB) and indocyanine green (ICG) in endometrial cancer and the impact of ultrastaging (NCT01818739)**. *Gynecol Oncol* (2019.0) **153** 496-499. DOI: 10.1016/j.ygyno.2019.03.252
135. Buda A, Crivellaro C, Elisei F, Di Martino G, Guerra L, De Ponti E. **Impact of indocyanine green for sentinel lymph node mapping in early stage endometrial and cervical cancer: comparison with conventional radiotracer (99m)Tc and/or blue dye**. *Ann Surg Oncol* (2016.0) **23** 2183-2191. DOI: 10.1245/s10434-015-5022-1
136. Di Martino G, Crivellaro C, De Ponti E, Bussi B, Papadia A, Zapardiel I. **Indocyanine green versus radiotracer with or without blue dye for sentinel lymph node mapping in stage >IB1 cervical cancer (>2 cm)**. *J Minim Invasive Gynecol* (2017.0) **24** 954-959. DOI: 10.1016/j.jmig.2017.05.011
137. Eriksson AG, Montovano M, Beavis A, Soslow RA, Zhou Q, Abu-Rustum NR. **Impact of obesity on sentinel lymph node mapping in patients with newly diagnosed uterine cancer undergoing robotic surgery**. *Ann Surg Oncol* (2016.0) **23** 2522-2528. DOI: 10.1245/s10434-016-5134-2
138. Frumovitz M, Plante M, Lee PS, Sandadi S, Lilja JF, Escobar PF. **Near-infrared fluorescence for detection of sentinel lymph nodes in women with cervical and uterine cancers (FILM): a randomised, phase 3, multicentre, non-inferiority trial**. *Lancet Oncol* (2018.0) **19** 1394-1403. DOI: 10.1016/S1470-2045(18)30448-0
139. Holloway RW, Ahmad S, Kendrick JE, Bigsby GE, Brudie LA, Ghurani GB. **A prospective cohort study comparing colorimetric and fluorescent imaging for sentinel lymph node mapping in endometrial cancer**. *Ann Surg Oncol* (2017.0) **24** 1972-1979. DOI: 10.1245/s10434-017-5825-3
140. How J, Gotlieb WH, Press JZ, Abitbol J, Pelmus M, Ferenczy A. **Comparing indocyanine green, technetium, and blue dye for sentinel lymph node mapping in endometrial cancer**. *Gynecol Oncol* (2015.0) **137** 436-442. DOI: 10.1016/j.ygyno.2015.04.004
141. Lier MCI, Vlek SL, Ankersmit M, van de Ven PM, Dekker JJML, Bleeker MCG. **Comparison of enhanced laparoscopic imaging techniques in endometriosis surgery: a diagnostic accuracy study**. *Surg Endosc* (2020.0) **34** 96-104. DOI: 10.1007/s00464-019-06736-8
142. Lührs O, Ekdahl L, Lönnerfors C, Geppert B, Persson J. **Combining Indocyanine Green and Tc**. *Gynecol Oncol* (2020.0) **156** 335-340. DOI: 10.1016/j.ygyno.2019.11.026
143. Martinelli F, Ditto A, Signorelli M, Bogani G, Chiappa V, Lorusso D. **Sentinel node mapping in endometrial cancer following Hysteroscopic injection of tracers: a single center evaluation over 200 cases**. *Gynecol Oncol* (2017.0) **146** 525-530. DOI: 10.1016/j.ygyno.2017.06.014
144. Papadia A, Zapardiel I, Bussi B, Ghezzi F, Ceccaroni M, De Ponti E. **Sentinel lymph node mapping in patients with stage I endometrial carcinoma: a focus on bilateral mapping identification by comparing radiotracer Tc99**. *J Cancer Res Clin Oncol* (2017.0) **143** 475-480. DOI: 10.1007/s00432-016-2297-y
145. Prader S, du Bois A, Harter P, Breit E, Schneider S, Baert T. **Sentinel lymph node mapping with fluorescent and radioactive tracers in vulvar cancer patients**. *Arch Gynecol Obstet* (2020.0) **301** 729-736. DOI: 10.1007/s00404-019-05415-2
146. Rossi EC, Kowalski LD, Scalici J, Cantrell L, Schuler K, Hanna RK. **A comparison of sentinel lymph node biopsy to lymphadenectomy for endometrial cancer staging (FIRES trial): a multicentre, prospective, cohort study**. *Lancet Oncol* (2017.0) **18** 384-392. DOI: 10.1016/S1470-2045(17)30068-2
147. Rozenholc A, Samouelian V, Warkus T, Gauthier P, Provencher D, Sauthier P. **Green versus blue: randomized controlled trial comparing indocyanine green with methylene blue for sentinel lymph node detection in endometrial cancer**. *Gynecol Oncol* (2019.0) **153** 500-504. DOI: 10.1016/j.ygyno.2019.03.103
148. Soergel P, Kirschke J, Klapdor R, Derlin T, Hillemanns P, Hertel H. **Sentinel lymphadenectomy in cervical cancer using near infrared fluorescence from indocyanine green combined with technetium-99m-nanocolloid**. *Lasers Surg Med* (2018.0) **50** 994-1001. DOI: 10.1002/lsm.22999
149. Pathak RA, Hemal AK. **Intraoperative ICG-fluorescence imaging for robotic-assisted urologic surgery: current status and review of literature**. *Int Urol Nephrol* (2019.0) **51** 765-771. DOI: 10.1007/s11255-019-02126-0
150. Angell JE, Khemees TA, Abaza R. **Optimization of near infrared fluorescence tumor localization during robotic partial nephrectomy**. *J Urol* (2013.0) **190** 1668-1673. DOI: 10.1016/j.juro.2013.04.072
151. Borofsky MS, Gill IS, Hemal AK, Marien TP, Jayaratna I, Krane LS. **Near-infrared fluorescence imaging to facilitate super-selective arterial clamping during zero-ischaemia robotic partial nephrectomy**. *BJU Int* (2013.0) **111** 604-610. DOI: 10.1111/j.1464-410X.2012.11490.x
152. Bjurlin MA, McClintock TR, Stifelman MD. **Near-infrared fluorescence imaging with intraoperative administration of indocyanine green for robotic partial nephrectomy**. *Curr Urol Rep* (2015.0) **16** 20. DOI: 10.1007/s11934-015-0495-9
153. Diana P, Buffi NM, Lughezzani G, Dell’Oglio P, Mazzone E, Porter J. **The role of intraoperative indocyanine green in robot-assisted partial nephrectomy: results from a large**. *Multi-inst Series Eur Urol* (2020.0) **78** 743-749. DOI: 10.1016/j.eururo.2020.05.040
154. Harke N, Schoen G, Schiefelbein F, Heinrich E. **Selective clamping under the usage of near-infrared fluorescence imaging with indocyanine green in robot-assisted partial nephrectomy: a single-surgeon matched-pair study**. *World J Urol* (2014.0) **32** 1259-1265. DOI: 10.1007/s00345-013-1202-4
155. Harke NN, Godes M, Wagner C, Addali M, Fangmeyer B, Urbanova K. **Fluorescence-supported lymphography and extended pelvic lymph node dissection in robot-assisted radical prostatectomy: a prospective, randomized trial**. *World J Urol* (2018.0) **36** 1817-1823. DOI: 10.1007/s00345-018-2330-7
156. KleinJan GH, van den Berg NS, Brouwer OR, de Jong J, Acar C, Wit EM. **Optimisation of fluorescence guidance during robot-assisted laparoscopic sentinel node biopsy for prostate cancer**. *Eur Urol* (2014.0) **66** 991-998. DOI: 10.1016/j.eururo.2014.07.014
157. Krane LS, Manny TB, Hemal AK. **Is near infrared fluorescence imaging using indocyanine green dye useful in robotic partial nephrectomy: a prospective comparative study of 94 patients**. *Urology* (2012.0) **80** 110-116. DOI: 10.1016/j.urology.2012.01.076
158. Lanchon C, Arnoux V, Fiard G, Descotes JL, Rambeaud JJ, Lefrancq JB. **Super-selective robot-assisted partial nephrectomy using near-infrared flurorescence versus early-unclamping of the renal artery: results of a prospective matched-pair analysis**. *Int Braz J Urol* (2018.0) **44** 53-62. DOI: 10.1590/S1677-5538
159. Manny TB, Patel M, Hemal AK. **Fluorescence-enhanced robotic radical prostatectomy using real-time lymphangiography and tissue marking with percutaneous injection of unconjugated indocyanine green: the initial clinical experience in 50 patients**. *Eur Urol* (2014.0) **65** 1162-1168. DOI: 10.1016/j.eururo.2013.11.017
160. Manny TB, Hemal AK. **Fluorescence-enhanced robotic radical cystectomy using unconjugated indocyanine green for pelvic lymphangiography, tumor marking, and mesenteric angiography: the initial clinical experience**. *Urology* (2014.0) **83** 824-829. DOI: 10.1016/j.urology.2013.11.042
161. Mattevi D, Luciani LG, Mantovani W, Cai T, Chiodini S, Vattovani V. **Fluorescence-guided selective arterial clamping during RAPN provides better early functional outcomes based on renal scan compared to standard clamping**. *J Robot Surg* (2019.0) **13** 391-396. DOI: 10.1007/s11701-018-0862-x
162. McClintock TR, Bjurlin MA, Wysock JS, Borofsky MS, Marien TP, Okoro C. **Can selective arterial clamping with fluorescence imaging preserve kidney function during robotic partial nephrectomy?**. *Urology* (2014.0) **84** 327-332. DOI: 10.1016/j.urology.2014.02.044
163. Polom W, Markuszewski M, Cytawa W, Czapiewski P, Lass P, Matuszewski M. **Fluorescent versus radioguided lymph node mapping in bladder cancer**. *Clin Genitourin Cancer* (2017.0) **15** e405-e409. DOI: 10.1016/j.clgc.2016.11.007
164. Ramírez-Backhaus M, Mira Moreno A, Gómez Ferrer A, Calatrava Fons A, Casanova J, Solsona Narbón E. **Indocyanine green guided pelvic lymph node dissection: an efficient technique to classify the lymph node status of patients with prostate cancer who underwent radical prostatectomy**. *J Urol* (2016.0) **196** 1429-1435. DOI: 10.1016/j.juro.2016.05.087
165. Sentell KT, Ferroni MC, Abaza R. **Near-infrared fluorescence imaging for intraoperative margin assessment during robot-assisted partial nephrectomy**. *BJU Int* (2020.0) **126** 259-264. DOI: 10.1111/bju.15089
166. Shimbo M, Endo F, Matsushita K, Hattori K. **Impact of indocyanine green-guided extended pelvic lymph node dissection during robot-assisted radical prostatectomy**. *Int J Urol* (2020.0) **27** 845-850. DOI: 10.1111/iju.14306
167. Simone G, Tuderti G, Anceschi U, Ferriero M, Costantini M, Minisola F. **"Ride the Green Light": indocyanine green-marked off-clamp robotic partial nephrectomy for totally endophytic renal masses**. *Eur Urol* (2019.0) **75** 1008-1014. DOI: 10.1016/j.eururo.2018.09.015
168. Soga N, Inoko A, Furusawa J, Ogura Y. **Evaluation to differentiate between tumor lesions and the parenchyma in partial nephrectomies for renal tumors based on quantitative fluorescence imaging using indocyanine green dye**. *Curr Urol* (2019.0) **13** 74-81. DOI: 10.1159/000499289
169. Veccia A, Antonelli A, Hampton LJ, Greco F, Perdonà S, Lima E. **Near-infrared fluorescence imaging with indocyanine green in robot-assisted partial nephrectomy: pooled analysis of comparative studies**. *Eur Urol Focus* (2020.0) **6** 505-512. DOI: 10.1016/j.euf.2019.03.005
|
---
title: Attention Biases for Eating Disorder-Related Stimuli Versus Social Stimuli
in Adolescents with Anorexia Nervosa – An Eye-Tracking Study
authors:
- Anca Sfärlea
- Anne Kathrin Radix
- Gerd Schulte-Körne
- Tanja Legenbauer
- Belinda Platt
journal: Research on Child and Adolescent Psychopathology
year: 2022
pmcid: PMC10017650
doi: 10.1007/s10802-022-00993-3
license: CC BY 4.0
---
# Attention Biases for Eating Disorder-Related Stimuli Versus Social Stimuli in Adolescents with Anorexia Nervosa – An Eye-Tracking Study
## Abstract
Anorexia nervosa (AN) is characterized by attention biases for eating disorder-related information as well as altered attentional processing of social information. However, little is known about the interplay between the altered attentional processing of these two types of information. The present study investigates attention biases for eating disorder-related information (pictures of bodies) versus social information (pictures of faces), in adolescents with AN. Attention biases were assessed via eye-tracking during a passive-viewing task in which female bodies and faces were presented simultaneously and thus competed directly for attention. Female adolescents (13–18 years) with AN ($$n = 28$$) were compared to a clinical comparison group (adolescents with major depression; $$n = 20$$) and a comparison group of adolescents with no mental illness ($$n = 24$$). All groups looked longer at bodies than at faces, i.e., showed attention biases for bodies in maintenance of attention. These biases were more pronounced in adolescents with AN than in both comparison groups, particularly for underweight bodies, at the expense of looking less at social stimuli. The results indicate “dual” attention biases in adolescents with AN (i.e., towards bodies and away from emotional faces) which could have a twofold negative impact on eating disorder psychopathology: increased attention to eating disorder-related information might directly influence eating disorder symptoms while less attention to social information might have an indirect influence through the amplification of interpersonal difficulties.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s10802-022-00993-3.
## Introduction
Anorexia nervosa (AN) is a severe mental disorder that mostly affects adolescent and young adult women with an onset age peak at 15.5 years (Solmi et al., 2022). It has a relatively poor long-term prognosis: Only between one and two thirds of patients recover fully while up to $30\%$ of cases take a chronic course (Eddy et al., 2017; Fichter et al., 2017; Herpertz-Dahlmann et al., 2018; Rydberg Dobrescu et al., 2020), resulting in the highest mortality rate of all mental disorders (Arcelus et al., 2011; Harris & Barraclough, 1998). This underlines the need for a deeper understanding of the mechanisms involved in the etiology and maintenance of the disorder.
AN is characterized by significantly low body weight, intense fear of gaining weight and body image disturbance (American Psychiatric Association, 2013). These core symptoms are accompanied by dysfunctional cognitions about food, weight, and shape (Vitousek & Hollon, 1990; Williamson et al., 1999, 2004), including attention biases for eating disorder (ED) -related information like pictures of food and bodies (Lee & Shafran, 2004; Ralph-Nearman et al., 2019; Stott et al., 2021). Attention biases are defined as automatic tendencies to preferentially attend to information that is consistent with one’s (maladaptive) cognitive schemata, for example disorder-related or negative information (Aspen et al., 2013). Importantly, biases for disorder-related information are proposed to not only represent an epiphenomenon of AN but to play a role in the development and maintenance of the ED (Aspen et al., 2013; Vitousek & Hollon, 1990; Williamson et al., 1999, 2004).
In addition to difficulties related to eating and body image, individuals with AN have also been found to show difficulties in social and emotional functioning (Caglar-Nazali et al., 2014; Mason et al., 2021; Oldershaw et al., 2011; Tauro et al., 2022) and these difficulties have also been proposed to contribute to the disorder’s development and maintenance (Treasure & Cardi, 2017; Treasure et al., 2012). For example, individuals with AN have been found to show alterations in the neurophysiological correlates of attentional processing of social stimuli such as faces (Fonville et al., 2014; Sfärlea et al., 2016), indicating, among other things, that they might show reduced selective attention for other people’s faces and perceive them as less intrinsically salient, i.e., less relevant for themselves (Sfärlea et al., 2016). Such alterations in the attentional processing of important social cues could contribute to individuals with AN increasingly isolating themselves from their family and peers, and this, in turn, might not only foster interpersonal problems but also exacerbate ED symptoms as a maladaptive reaction to these problems (Treasure et al., 2012).
Attention biases for social information, such as emotional faces, have been proposed as another factor contributing to socio-emotional difficulties in individuals with EDs (Harrison et al., 2010a). Research on such biases in individuals with AN have yielded mixed results: While some studies found stronger attention biases towards faces showing negative emotions in individuals with AN compared to those with no mental illness (Cardi et al., 2013; Harrison et al., 2010a, b), others found individuals with AN to turn their attention away from negative faces (Kim et al., 2014), and yet others found no differences in attention biases for emotional faces between individuals with AN and individuals with no mental illness (Bang et al., 2017; Cardi et al., 2015; Goddard & Treasure, 2013; Schneier et al., 2016). Most of these studies have investigated attention biases for angry faces as anger is suggested to play an important role in ED psychopathology (Ioannou & Fox, 2009). Attention biases towards angry faces have been suggested to reflect anger being particularly salient for individuals with AN (Harrison et al., 2010a) while attention biases away from angry faces, i.e., avoidance of angry faces, has been suggested to reflect anger being perceived as “toxic” and unacceptable by individuals with AN (Geller et al., 2000; Kim et al., 2014). Attention biases away from angry faces have been found to be associated with social communication difficulties in youth with autism spectrum disorders (García-Blanco et al., 2017) as they may involve behavioral avoidance of potentially aversive social interactions and thus impede the ability to adaptively solve interpersonal conflicts (García-Blanco et al., 2017).
To date, attention biases for disorder-relevant information and social information in AN have mostly been investigated as two different lines of research. Few studies have assessed attention biases for ED-related stimuli in direct comparison to social stimuli, i.e., have presented ED-related and social stimuli simultaneously so that these stimuli compete for attention. Investigating attention biases for disorder-relevant information versus social information might be a key to understanding how different factors involved in the maintenance of the disorder relate to each other. It could help to understand, for example, the altered processing of social cues in AN: Preoccupation with weight and shape, reflected in disorder-related information automatically attracting more attention than social information, might explain why social information like emotional faces seems to be less salient for individuals with AN (Sfärlea et al., 2016). The first study to examine attention to disorder-related versus social information (Watson et al., 2010) recorded eye-movements while participants viewed either pictures of faces or whole-body pictures including faces. When bodies were presented together with faces, weight-recovered women with AN spent less time looking at the faces compared to women without EDs, thus showing an attention bias towards bodies. Pinhas et al. [ 2014] also used eye-tracking and presented pictures of bodies and pictures of social interactions simultaneously. Adolescents with AN spent more time looking at body pictures than at pictures of positive social interactions, while this was not the case in adolescents with no mental illness. When thin and overweight body pictures were both presented alongside pictures of social interactions, adolescents with AN preferentially looked at the thin body pictures.
The aim of the present study was to add to this emerging body of literature by investigating attention biases for disorder-relevant information, i.e., pictures of bodies, versus social information, i.e., pictures of faces, in adolescents with AN. We focus on adolescents as adolescence is the most common time for the onset of AN (Somli et al., 2022) while at the same time difficulties in social functioning could have particularly adverse consequences in adolescence due to increasing social demands in this developmental period (Happé & Frith, 2014). When studying adolescents with AN, it has to be taken into account that a high proportion are also affected by comorbid mental disorders, especially depression (Bühren et al., 2014; Jaite et al., 2013). Adolescents with depression have been found to show attention biases for social information like faces (Lau & Waters, 2017; Platt et al., 2017) and it is possible that biases for body-related stimuli are also not a specific characteristic of individuals with EDs but are present transdiagnostically in various mental disorders. Previous studies have not addressed this possibility so we aimed to extend their findings by comparing adolescents with AN not only to a group of adolescents with no mental illness (“healthy” comparison; HC) but also to a clinical comparison group consisting of adolescents with major depression (MD).1 This allows to draw conclusions about the specificity of biases for AN.
Most studies investigating attention biases in individuals with AN have used emotional Stoop or modified Dot-Probe tasks (Aspen et al., 2013; Jiang & Vartanian, 2018; Lee & Shafran, 2004) which infer attention biases from differences in reaction times between different experimental conditions. The assessment of attention biases via reaction-time based measures, however, entails several limitations: i) they provide only a snapshot of attention at a single point in time (Armstrong & Olatunji, 2012), not accounting for the fact that attention consists of different distinct and consecutive subprocesses (Jiang & Vartanian, 2018; Kerr-Gaffney et al., 2019), ii) they only provide an indirect measure of attention and can be influenced by non-attention-related processes, such as slowed processing speed or response execution (Armstrong & Olatunji, 2012), and iii) both the Stroop (Dresler et al., 2012; Lee & Shafran, 2004) and the Dot-Probe task (Platt et al., 2022; Vervoort et al., 2021; Waechter et al., 2014) have been found to show very poor reliability and validity. An alternative to reaction-time based measures are eye-tracking paradigms which measure attention more directly as they allow to continuously record the course of visual attention over time. This has the advantage of capturing the dynamic nature of attention and being able to distinguish between different subprocesses as different eye-tracking parameters serve as indicators for different processes (e.g., Kerr-Gaffney et al., 2019). This is of particular interest when studying attention biases in AN as it has been suggested that individuals with AN might show first hypervigilance for and then avoidance of disorder-relevant information (Lee & Shafran, 2004). Furthermore, eye-tracking paradigms have been found to show superior psychometric properties in adults (Lazarov et al., 2019; Waechter et al., 2014) as well as adolescents (Platt et al., 2022).
Therefore, we chose a passive-viewing eye-tracking paradigm similar to that of Pinhas et al. [ 2014] in which we presented disorder-relevant information and social information simultaneously while recording eye movements. The paradigm included two types of trials: i) neutral trials in which normalweight bodies were presented alongside neutral faces and ii) emotional trials in which under- and overweight bodies were presented together with positive (happy) and negative (angry) faces. Angry faces were used as negative emotional faces in line with most previous studies investigating attention biases for emotional faces in AN (Bang et al., 2017; Goddard & Treasure, 2013; Harrison et al., 2010a, b; Kim et al., 2014; Schneier et al., 2016). We investigated two components of attention: initial orientation of attention, indicated by location of first fixation, as well as maintenance of attention, indicated by dwell time over the duration of the trial. Based on previous studies (Pinhas et al., 2014; Watson et al., 2010), we expected to find adolescents with AN to show attention biases for disorder-relevant stimuli even in the presence of socially relevant stimuli, compared to both comparison groups. Furthermore, we expected this bias to be particularly pronounced for underweight stimuli, i.e., we expected that when underweight and overweight body pictures are presented together with emotional faces, adolescents with AN would preferentially look at underweight bodies (in line with Pinhas et al., 2014).
## Participants
The study sample consisted of 722 adolescent females aged 13–18 years: $$n = 28$$ adolescents with AN, $$n = 20$$ adolescents with MD and $$n = 24$$ adolescents with no mental illness (HC group). Adolescents with AN were recruited from two University Departments of Child and Adolescent Psychiatry in Germany (inpatients and outpatients), while the adolescents with MD and adolescents with no mental illness were recruited and tested at just one of these sites. Adolescents with MD were all inpatients. Adolescents with no mental illness were recruited through local advertisements and schools. To rule out that differences between the group with AN and the comparison groups were driven by adolescents with AN being recruited and tested at different recruitment sites, we compared AN patients from the two sites. They differed in depression and ED symptoms but neither in other participant characteristics nor in any of our outcome variables.
Psychiatric diagnoses were assessed in all participants using a standardized, semi-structured clinical interview (Kinder-DIPS; Margraf et al., 2017; Schneider et al., 2017). The Kinder-DIPS is a well-established German diagnostic interview that allows diagnosis of a wide range of psychiatric axis I disorders according to DSM-5 (American Psychiatric Association, 2013). The interview was administered to the adolescent participants by trained interviewers, i.e., study staff with a Bachelor’s or Master’s degree in Psychology who received several hours of training and supervision before conducting and evaluating the interviews. In previous studies using this training procedure we found the interrater-reliability of the Kinder-DIPS to be very good (e.g., accordance rates of $100\%$ for current diagnoses of AN or MD in clinical groups as well as for no lifetime diagnoses in non-clinical groups; Lukas et al., 2022; Sfärlea et al., 2020), in line with Neuschwander et al. [ 2013], who reported accordance rates of at least $97\%$ for all diagnoses. However, interrater-reliability could not be assessed for the present study. Exclusion criteria for all participants were IQ < 80 (measured via the Zahlen-Verbingungs-Test; Oswald, 2016), neurological disorders, psychotic disorders, bipolar disorder, substance abuse, pregnancy, benzodiazepine intake, and non-corrected visual impairment. Depression symptoms were assessed with the German version of the Beck Depression Inventory-II (BDI-II; Hautzinger et al., 2006; available from 69 of the 72 included participants, Cronbach’s α = 0.96 in our sample.), anxiety symptoms were assessed with the German version of the trait version of the State-Trait Anxiety Inventory (STAI-T; Laux et al., 1981; available from 71 of the 72 included participants, Cronbach’s α = 0.96 in our sample), and eating psychopathology was assessed with the German version of the Eating Disorder Examination – Questionnaire (EDE-Q; Hilbert & Tuschen-Caffier, 2016; available from 71 of the 72 included participants, Cronbach’s α = 0.97 in our sample.). For participants with AN or MD, height and weight were obtained from their physicians while for participants with no mental illness, height and weight were measured in the laboratory.
Adolescents were included in the group with AN if they currently met criteria for AN according to DSM-5 (American Psychiatric Association, 2013) and had a body mass index (BMI) of < 18.5 and below the 25th age-corrected percentile (according to Kromeyer-Hauschild et al., 2001). Sixteen of the included participants with AN fulfilled criteria for at least one comorbid mental disorder, including MD ($$n = 12$$), anxiety disorders ($$n = 10$$), and obsessive-compulsive disorder ($$n = 3$$). Mean illness duration of AN (time since first onset) was 19.62 months (SD = 21.14; Mdn = 12.50; range 3–96).
Adolescents were included in the group with MD if they currently met criteria for an episode of MD according to DSM-5 (American Psychiatric Association, 2013) and reported no current symptoms or history of EDs. Within the group of adolescents with MD, 15 individuals met criteria for one or more comorbid mental disorders, including anxiety disorders ($$n = 13$$) and obsessive-compulsive disorder ($$n = 1$$).
Adolescents were included in the HC group if they did not meet criteria for any current or past axis I disorder as assessed by the Kinder-DIPS.
Participant characteristics are presented in Table 1. As expected, the three groups differed in BMI and BMI-percentile, with adolescents with AN having lower values than adolescents with MD and adolescents with no mental illness. Both clinical groups (adolescents with AN as well as adolescents with MD) reported more depression symptoms and ED pathology than the adolescents with no mental illness but did not differ from each other. Groups differed in trait anxiety scores, with adolescents with MD reporting the highest scores and adolescents with no mental illness reporting the lowest scores. This suggests that the group with MD had similar depression symptoms and even more pronounced anxiety symptoms than the group with AN, confirming its suitability as a clinical comparison group that takes into account both, depression and anxiety. Furthermore, groups differed also in age and IQ with the group with AN being slightly younger and having a higher IQ than the group with MD and the group of adolescents with no mental illness. Table 1Demographic and clinical characteristics of the sampleANMDHCANOVApost-hoc testsn = 28n = 20n = 24M (SD)M (SD)M (SD)Fpη2Age15.37 (1.36)16.37 (1.14)16.43 (1.56)4.890.0100.12AN < MD = HCIQ110.55 (10.75)102.60 (11.96)102.38 (11.86)4.250.0180.11AN > MD = HCBMI16.41 (1.36)23.95 (5.96)21.42 (3.26)25.66 < 0.0010.43AN < MD = HCBMI-percentile (age-corrected)6.68 (6.98)63.85 (31.09)52.04 (32.50)35.74 < 0.0010.51AN < MD = HCEating disorder symptoms (EDE-Q)2.96 (1.84)1.96 (1.63)0.92 (0.93)11.09 < 0.0010.25AN = MD > HCDepression symptoms (BDI-II)24.32 (16.14)32.39 (12.71)6.13 (5.54)24.31 < 0.0010.42MD = AN > HCAnxiety symptoms (STAI-T)52.61 (12.77)60.45 (12.16)32.48 (7.66)36.70 < 0.0010.52MD > AN > HCAN anorexia nervosa, MD major depression, HC “healthy” comparison, IQ intelligence quotient, BMI body mass index, EDE-Q Eating Disorder Examination - Questionnaire, BDI-II Beck Depression Inventory II, STAI-T trait version of the State-Trait Anxiety Inventory, M mean, SD standard deviation; post-hoc t-test remain significant after Bonferroni-Holm correction for multiple testing (Holm, 1979).
Four participants with AN and six participants with MD were receiving psychotropic medication. As psychotropic medication may influence eye-movements (e.g., Reilly et al., 2008; Wells et al., 2014) all analyses were repeated excluding these participants. The overall pattern of results remained the same, so findings based on the whole sample are reported.
## Procedure
The present study was conducted as part of a larger project on attention biases in AN (Radix et al., under review) which comprised three sessions in total. Prior to participation, written informed consent was obtained from all participants (and their parents/legal custodians for participants under 18 years of age) after a comprehensive explanation of the procedures. The task examined in the present study was administered in the first session of the project, whereas additional experimental tasks (which investigated the role of anxiety in triggering attention biases) were delivered in sessions two and three (see Radix et al., under review). The first session of the project began with the diagnostic interview. After that, photographs of the participants’ bodies that were to be used as stimuli in sessions two and three were taken. Then the present task was administered, which assessed attention biases by recording eye movements during passive viewing of ED-related information (pictures of bodies) versus social information (pictures of faces). Participants completed the questionnaire measures between sessions one and two. The study was approved by the ethics committee of the Medical Faculty of the Ruhr-University Bochum (15-5541-BR) as well as the ethics committee of the Medical Faculty of the LMU Munich (Project-No. 814-16). Participants received a reimbursement of €50 for participation in the whole project.
## Stimuli
Stimuli consisted of photographs of faces and bodies that were presented in grayscale on black background. Face stimuli were taken from the Karolinska Directed Emotional Faces database (KDEF; Lundqvist et al., 1998) and edited so that only the facial area was visible. Pictures of six female models displaying neutral, happy, and angry facial expressions (two models per emotion) were used. Body stimuli were taken from a set of standardized photographs of female bodies of different weight categories in underwear (Horndasch et al., 2015). Pictures of two normalweight (BMI 20.7–21.0), two underweight (BMI 17.0–17.6), and two overweight (BMI 25.2–29.6) bodies in front view were used.
## Experimental Task
Participants were seated in front of a 22-in. monitor (1680 × 1050 pixel resolution; viewing distance approximately 70 cm) on which the experiment was presented using E-Prime Version 2.0 (Psychology Software Tools Inc., 2013). Each trial began with a fixation cross that had to be fixated for 500 ms for the trial to start. Then a 2 × 2 stimulus array was presented for 12 s (cf. Pinhas et al., 2014). The task consisted of 24 neutral trials and 24 emotional trials (conform to the recommended minimum trial number for eye-tracking research; Orquin & Holmqvist, 2018) that were presented in random order. In neutral trials, the stimulus array consisted of two normalweight bodies and two neutral faces. In emotional trials, an underweight body, an overweight body, a happy face, and an angry face were presented (see Fig. 1 for example stimulus displays). In both trial types the position of pictures was randomly assigned to one of the quadrants with each picture category being presented in each quadrant exactly twelve (neutral trials) or six (emotional trials) times and each model being presented equally often. Pictures had a size of 275 × 375 pixels and were presented with a distance of 150 pixels from each other horizontally as well as vertically (equivalent to a size of approximately 7.8 × 10.6 cm / 6.4° × 8.6° visual angle and a distance of 4.2 cm / 3.4° visual angle; i.e., pictures were presented between 1.7° and 8.1° of visual angle horizontally and between 1.7° and 10.3° of visual angle vertically). Participants were instructed to fixate the fixation cross and then freely view the stimuli with the only requirement being that their attention had to remain on the screen. Fig. 1Examples of a neutral trial (left side) and an emotional (right side) trial in the eye-tracking paradigm. Stimuli were taken from Lundqvist et al. [ 1998] and Horndasch et al. [ 2015] After the viewing task participants evaluated the face and body stimuli (presented in random order) on the dimensions valence (ranging from 1 = very unpleasant to 9 = very pleasant) and arousal (ranging from 0 = not at all arousing to 9 = very arousing) using the 9-point Self-Assessment Mannequin scale (Lang, 1980). The results of this evaluation are presented in Supplement 2.
## Eye-Tracking
Eye movements during the experimental task were registered binocularly (but only data of the left eye were analyzed) at a sampling rate of 500 Hz with a monitor-integrated eye-tracking system that used infrared video-based tracking technology (RED500; SensoMotoric Instruments) and iView X software (SensoMotoric Instruments). Before the task started a 9-point calibration and validation procedure was conducted and calibration was accepted if the average error was less than 0.5° of visual angle. Eye movements were detected using a velocity based detection method implemented in SMI BeGaze 3.7 software (SensoMotoric Instruments) with saccades defined as events with a velocity above 75°/s for a minimum duration of 20 ms and fixations defined as events with lower velocities and a minimum duration of 60 ms (Dinkler et al., 2019; Fujiwara et al., 2017). To ensure adequate data quality, line graphs and scan paths of each trial were visually inspected and trials with excessive blinks or a considerable proportion of missing data were excluded. Only participants with at least 12 valid trials for both trial types were included in the final sample. On average, 23.03 neutral trials (SD = 2.22) and 23.11 emotional trials (SD = 2.14) were available per participant (not different between groups: Fs ≤ 1.99, ps > 0.1).
Two eye-tracking indices were examined: percentage of first fixations on each picture category as an indicator of initial orientation of attention and mean percentage of dwell time (defined as sum of durations of all fixations and saccades that hit the area of interest) on each picture category during the whole duration of the trial as an indicator of maintenance of attention. Reliabilities of these outcome measures were assessed by correlating scores based on odd versus even trials (split-half reliability; see, e.g., Platt et al., 2022, for a similar approach). Split half-reliabilities for percentages of first fixations were acceptable to good (Spearman-Brown-corrected reliability 0.77-0.88) while split half-reliabilities for percentages of dwell time were good to excellent (Spearman-Brown-corrected reliability 0.83-0.95).
## Data Analysis
Statistical data analysis was conducted with SPSS. The eye-tracking indices were analyzed separately for neutral and emotional trials using repeated-measures analyses of variance (ANOVAs) with the within-subjects factor PictureCategory (2 for neutral trials: face, body; 4 for emotional trials: happy face, angry face, underweight body, overweight body) and the between-subjects factor Group (3: AN, MD, HC). Significant effects were followed up by post-hoc ANOVAs and subsequent t-tests. Degrees of freedom were adjusted via the Greenhouse–Geisser correction when necessary, i.e., when the assumption of sphericity was violated. For all analyses, the significance level was set to $$p \leq 0.05$$ (two-tailed) and effect sizes are reported: ηp2 for ANOVAs (with ηp2 = 0.01 interpreted as a small effect, ηp2 = 0.06 interpreted as a medium effect, and ηp2 = 0.14 interpreted as a large effect; Cohen, 1988) and Cohen’s d for t-tests (with $d = 0.20$ interpreted as a small effect, $d = 0.50$ interpreted as a medium effect, and $d = 0.80$ interpreted as a large effect; Cohen, 1988).
## Initial Orientation of Attention
For neutral trials, the ANOVA on percentage of first fixations yielded no significant effects (Fs ≤ 1.94 ps > 0.1). For emotional trials, the ANOVA yielded a significant main effect of PictureCategory (F3,207 = 14.59, $p \leq 0.001$, ηp2 = 0.18), resulting from participants orienting their attention towards pictures of happy faces significantly more often than towards other pictures (ts71 ≥ 4.3, ps < 0.001), while the PictureCategory × Group interaction was non-significant (F < 1). See Fig. 2a, b and Supplementary Table 1 for descriptive results of the eye-tracking data. Fig. 2Results of the investigated eye-tracking indices. The top panels show percentage of first fixations on the different picture categories in the neutral trials (a) as well the emotional trials (b). The bottom panels show percentage of dwell time on the different pictures across the whole trial duration in neural trails (c) and emotional trials (d). Note that percentage of dwell time in neutral trials is the mean percentage across both pictures of a category, not the sum. Significant group differences are indicated as follows: * $p \leq 0.05$, ** $p \leq 0.01$, *** $p \leq 0.001.$ Error bars represent standard errors
## Maintenance of Attention
For neutral trials, the ANOVA on percentage of dwell time yielded a significant main effect of PictureCategory (F1,69 = 51.01, $p \leq 0.001$, ηp2 = 0.43) with longer dwell times on pictures of bodies compared to pictures of faces, while the main effect of Group was not significant (F2,69 = 2.31, $p \leq 0.1$). Furthermore, a significant PictureCategory × Group interaction (F2,69 = 3.22, $$p \leq 0.046$$, ηp2 = 0.09) emerged, which was followed up by one-way ANOVAs with the factor Group performed separately for faces and bodies and subsequent t-tests. For bodies, a significant effect of Group emerged (F2,69 = 5.33, $$p \leq 0.007$$, ηp2 = 0.13), due to adolescents with AN fixating longer on the bodies than adolescents with no mental illness (t50 = 3.23, $$p \leq 0.002$$, $d = 0.90$) and adolescents with MD (t46 = 2.02, $$p \leq 0.049$$, $d = 0.59$; this comparison would not be significant after correction for multiple testing; Holm, 1979). For faces, no effect of Group was found (F2,69 = 1.17, $p \leq 0.1$, ηp2 = 0.03).
For emotional trials, a significant main effect of PictureCategory (F1.89,130.20 = 76.93, $p \leq 0.001$, ηp2 = 0.53), resulting from significant differences between all categories (ts71 ≥ 3.67, ps < 0.001), and a significant PictureCategory × Group interaction (F3.77,130.20 = 5.75, $p \leq 0.001$, ηp2 = 0.14) emerged, while the main effect of Group was non-significant (F < 1). The interaction was followed up by one-way ANOVAs with the factor Group performed separately for each picture category and subsequent t-tests. A significant effect of Group emerged for angry faces (F2,69 = 5.31, $$p \leq 0.007$$, ηp2 = 0.13) as well as underweight bodies (F2,69 = 12.40, $p \leq 0.001$, ηp2 = 0.26), due to the adolescents with AN fixating longer on underweight bodies and shorter on angry faces than adolescents with no mental illness (underweight bodies: t50 = 4.86, $p \leq 0.001$, $d = 1.35$; angry faces: t50 = 3.17, $$p \leq 0.003$$, $d = 0.88$) and adolescents with MD (underweight bodies: t46 = 3.35, $$p \leq 0.002$$, $d = 0.98$; angry faces: t46 = 2.35, $$p \leq 0.023$$, $d = 0.69$). No effects emerged for happy faces (F2,69 = 2.41, $$p \leq 0.097$$, ηp2 = 0.07) as well as overweight bodies (F < 1). Results are presented in Fig. 2c, d as well as Supplementary Table 1.
In addition to the planned analyses that examined our a priori hypotheses, we conducted post-hoc exploratory analyses to determine whether groups differed in the course of attention deployment over time: We split the trials into four 3-s time intervals and conducted TimeWindow × PictureCategory × Group ANOVAs on percentage of dwell time. The results indicated that in the neutral trials, differences in dwell time resulted from the group with AN dwelling more on bodies particularly during the middle part of the trial (see Supplement 3 for detailed results).
## Additional Analyses
As participants in the three groups differed in age and IQ, we computed Pearson's correlations between these variables and percentage of dwell time on normalweight bodies, underweight bodies, and angry faces, to investigate if these participant characteristics may have accounted for the group differences we found. Correlations were computed separately within each group to rule out that emerging correlations would be artefacts of the group differences in both variables. Significant correlations between age and dwell time on underweight bodies (r = -0.47, $$p \leq 0.020$$) and angry faces ($r = 0.60$, $$p \leq 0.002$$) emerged only within the group of adolescents with no mental illness, indicating no systematic relationships.
## Discussion
The aim of the present study was to investigate attention biases for disorder-related stimuli (i.e., pictures of bodies) versus social stimuli (i.e., pictures of faces) in adolescents with AN compared to adolescents with MD or no mental illness using a passive-viewing eye-tracking paradigm where both types of stimuli were presented simultaneously and directly competed for attention. We found more pronounced biases in the maintenance of attention on body pictures in adolescents with AN than in the comparison groups, particularly for pictures of underweight bodies, at the expense of looking less at social stimuli.
Adolescents in all groups looked longer at bodies than at faces, i.e., showed attention biases for bodies in maintenance of attention in both, neutral and emotional trials. This is in line with results from another task in an overlapping sample (Radix et al., under review) and other studies finding attention biases for stimuli related to body shape in female adolescents as young as 14 years (Green & McKenna, 1993). It might reflect the high importance of body and shape for female adolescents in general (Wadden et al., 1991). In the emotional trials, all participants preferentially looked at underweight bodies. This is in line with Watson et al. [ 2010], who found that when pictures of underweight females were presented, both, women with and without AN, spent less time looking at the females’ faces. Our result also corresponds with findings from Radix et al. ( under review) and is in line with other studies that have found attention biases towards thin versus overweight bodies in young women (Glauert et al., 2010; Joseph et al., 2016). However, other studies have not reported such main effects of picture category but found only interactions of stimulus category and body dissatisfaction, i.e., different attention biases in women with high and low body dissatisfaction (Moussally et al., 2016; Tobin et al., 2018).
It has to be noted that there might also be another plausible explanation for the preferential looking at bodies across all groups and both trial types. Body and face stimuli differed in physical properties that might influence viewing times: Face stimuli had a higher luminance while body stimuli were more complex and included more details to explore and were presumably more interesting than the face stimuli. Therefore, the face stimuli might have become boring earlier as only two pictures per category were presented and repeated several times. The preference to look at underweight bodies in the emotional trials, however, cannot be explained by these physical properties of the stimuli as pictures of underweight and overweight bodies had similar complexity and luminance.
In line with our hypotheses, adolescents with AN looked longer at bodies than adolescents in the comparison groups, i.e., showed stronger attention biases for body pictures. This was the case in both, neutral as well as emotional trials, where the AN group’s preference for bodies was particularly pronounced for underweight body pictures. Finding more pronounced attention biases for bodies in adolescents with AN across both trial types together with the medium to large between-group effect sizes ($d = 0.59$–1.35) underline the robustness of this result. Importantly, adolescents with AN differed from both, adolescents with no mental illness as well as adolescents with MD, indicating that the more pronounced attention biases for body-related information are specific to their ED and cannot be attributed to comorbid depression, anxiety, or more general psychopathology.3 Our results are consistent with those of previous studies investigating attention biases for body-related in comparison to socially relevant stimuli that also found adolescents (Pinhas et al., 2014) and adults (Watson et al., 2010) with AN to show attention biases towards bodies and less attendance of socially relevant information. However, one study (Cornelissen et al., 2016) found women with AN to look more at the face region than women with no mental illness in a body-size estimation task where full body images including the head/face were presented. Our results are also in line with cognitive-behavioral theories predicting biases for body weight and shape related information in individuals with AN (Williamson et al., 1999, 2004). According to these theories, these biases could directly contribute to the maintenance of the disorder by reinforcing maladaptive schemata related to weight and shape which fuel ED psychopathology.
The particularly pronounced attention bias for underweight body pictures is also in line with our expectations and replicated the result of Pinhas et al. [ 2014]. It was also observed using a different paradigm in an overlapping sample (Radix et al., under review). Preferential looking at thin bodies has also been reported for other EDs: Blechert and colleagues [2009] found adult women with bulimia nervosa to fixate longer on low BMI bodies and shorter on high BMI bodies than women with no mental illness. By contrast, Rieger et al. [ 1998] found an attention bias away from words describing a thin physique in women with EDs. The preference for thin bodies and the dislike of overweight bodies in AN is also reflected in explicit evaluations of body pictures: most studies (including the present one, see Supplement 2) found adolescents and adults with AN to evaluate bodies with higher BMIs more negatively and bodies with lower BMIs more positively than females with no mental illness (Horndasch et al., 2015, 2018; von Wietersheim et al., 2012; but see also Watson et al., 2010). Underweight bodies are consistent with the goals of (most) individuals with AN and looking at them could motivate the maintenance of weight loss behaviors (Mento et al., 2021; Norris et al., 2006; Pinhas et al., 2014), so that biases for underweight bodies could have a particularly detrimental direct effect on ED psychopathology.
The aforementioned higher complexity of the body stimuli compared to the face stimuli could provide an alternative explanation as to why adolescents with AN showed a stronger preference for bodies vs. faces than the comparison groups: Adults (Lang et al., 2014) as well as adolescents (Lang & Tchanturia, 2014) with AN have been found to show weaker central coherence, i.e., superior attention to detail alongside poorer holistic processing, compared to individuals with no mental illness. Hence, the preferential attention for bodies (the stimuli characterized by more details), could reflect a bias for preferential processing of detail rather than preferential processing of stimuli related to weight and shape. Importantly, the physical properties of the stimuli cannot explain the particularly pronounced attention bias for underweight bodies in the emotional trials, in which underweight body pictures were presented alongside overweight body pictures. These two stimulus categories did not vary in complexity but only in the weight of the depicted bodies, hence, one would expect adolescents with AN to show increased dwell time also for overweight body pictures if effects were explained only by increased attention to detail in this group. Since this was not the case, the results seem to be at least to some extent modulated by picture content.
Importantly, no differences in initial orientation of attention were found between the three groups. Instead, participants in all groups showed preferential orientation towards happy faces in the emotional trials, which is in line with a meta-analysis showing that biases for positive stimuli occur in early attentional processing (Pool et al., 2016). Also, our exploratory analysis that examined the course of visual attention over time did not indicate more pronounced group differences in dwell time at the beginning of the trials. Thus, we found no evidence that adolescents with AN first show hypervigilance and then avoidance of body pictures. While initial orientation of attention is a more automatic and bottom-up controlled component of attention, maintenance of attention is more top-down controlled (Theeuwes et al., 2000), especially in such a task like the passive-viewing task employed in the present study, in which participants can deliberately choose where to look. Hence, our results suggest that adolescents with AN show more pronounced attention biases for disorder-relevant information (versus social information) in an internally-controlled and at least partially conscious aspect of attention, while not differing from the comparison groups in more automatic aspects of attention. This is similar to results from Kerr-Gaffney et al. [ 2021] who found individuals with AN to show initial orientation towards social stimuli but disengage from such stimuli more quickly than individuals with no mental illness.
Our results indicate that adolescents with AN do not only show attention biases for body stimuli when these are presented together with neutral stimuli but also in the presence of salient social information as pictures of emotional faces. In return for looking more at bodies, adolescents with AN attended less to faces, particularly to faces showing angry expressions. This is partly comparable with results in individuals with social anxiety disorder who have been found to show attention biases towards angry faces in early attentional processes (Armstrong & Olatunji, 2012; Duval et al., 2020; Liang et al., 2017) and avoidance of (angry) faces in later, more conscious aspects of attention (Garner et al., 2006), as well as gaze (Weeks et al., 2013) and behavior (Heuer et al., 2007). Importantly, this behavioral avoidance might further exacerbate symptoms of social anxiety (Kashdan et al., 2014), thereby acting as a maintenance factor. Less looking at faces and more looking at non-social cues (Frazier et al., 2017) as well as attention biases away from angry faces (García-Blanco et al., 2017; Ghosn et al., 2019) have also been found in individuals with autism spectrum disorders where they seem to be related to poorer social adjustment (Klin et al., 2002) and communication difficulties (García-Blanco et al., 2017). By attending less to social information, important social cues that are a key to successful communication and interaction are likely to be missed, which might contribute to the development and amplification of interpersonal difficulties (Kerr-Gaffney et al., 2019, 2021). The avoidance of angry faces, in particular, might involve behavioral avoidance of potentially aversive social interactions, thereby hampering the ability to adaptively solve interpersonal conflicts (García-Blanco et al., 2017). Through these mechanisms, the biases found in the present study could have an additional indirect negative effect on ED psychopathology. Of note, it has been suggested that different mechanisms may explain reduced attention to social cues in AN and autism spectrum disorders (Kerr-Gaffney et al., 2022). The present study might help to shed light on these different mechanisms as it indicates that increased attention to disorder-related characteristics of people (i.e., their bodies) might underlie the reduced attention to their faces in individuals with AN while probably other mechanisms (as for example attempts to reduce overstimulation or stress; Kerr-Gaffney et al., 2022) are responsible for the reduced attention to faces in individuals with autism spectrum disorders. However, in the present study pictures of faces were always presented alongside pictures of bodies (i.e., highly relevant disorder-related stimuli) so it remains unknown whether adolescents with AN would also look less at faces in the absence of disorder-related information (as found in adults with AN: Fujiwara et al., 2017; Kerr-Gaffney et al., 2021).
We designed the present study not only to gain insight in the preferential processing of disorder-related information in individuals with AN, but also hoping that the investigation of disorder-related information in relation to socially relevant information would help to understand the altered attentional processing of social cues in AN. The finding that adolescents with AN direct more attention to body related stimuli at the expense of directing less attention to emotional faces is indeed in line with our assumptions. However, previous studies investigating processing of social cues using event-related potentials (Hatch et al., 2010; Sfärlea et al., 2016) or functional magnetic resonance imaging (Fonville et al., 2014) suggested alterations in early, automatic stages of processing of emotional faces, whereas the present study did not find alterations in early, automatic attention allocation but in later, more conscious attentional processing. Hence, further investigation of the automatic and conscious processes involved in the processing of ED-related and socially relevant cues in AN are necessary to understand their interplay.
## Strengths and Limitations
The present study has several strengths. Some of them concern the sample: In contrast to previous studies, we compared adolescents with AN not only to adolescents with no mental illness but also to a clinical comparison group of adolescents with MD, allowing us to draw conclusions about biases being specific for adolescents with AN. To ensure diagnostic accuracy, all participants underwent an extensive diagnostic assessment using a standardized interview instead of relying on self-reported diagnoses. Although the size of our sample is modest, it is still considerably larger than in the previous studies investigating attention biases for ED-related versus social stimuli (Pinhas et al., 2014; Watson et al., 2010). Another strength is that we determined the reliability of our outcome measures which was acceptable to excellent for both eye-tracking indices, further underlining the robustness of our results.
Some limitations have to be noted as well. As mentioned earlier, the present study was conducted as part of a larger project on attention biases in adolescents with AN (Radix et al., under review) and photographs of the participants’ bodies that were to be used as stimuli in another task were taken in the same session as the task described in the present study was administered. This might have activated body related schemata (Labarge et al., 1998) which could have contributed to participants attending more to bodies. It remains unclear if the same effects would have been found if the task was administered completely independently from other procedures. However, as the procedure was the same for all participants, it is unlikely that it provides an explanation for the group differences we found.
Another limitation relates to our stimuli: we used standardized stimuli with body and face stimuli taken from validated databases (Horndasch et al., 2015; Lundqvist et al., 1998), presented in grayscale, and showing no distracting features like different haircuts or clothes, thereby eliminating as many confounding characteristics as possible. Still, pictures of faces and bodies varied in luminance and complexity and it cannot be ruled out that physical properties of the pictures in addition to picture content influenced our results. However, as pictures of faces and bodies are innately different in complexity this limitation does not only apply to our study but is a general limitation when stimuli of different categories compete for attention.
Further limitations concern our study sample: i) The three groups differed significantly in age and IQ, with participants in the AN group being significantly younger and having a higher IQ than those in the comparison groups. However, as neither age nor IQ were systematically related to our outcome measures, it is unlikely that group differences in dwell times can be attributed to variations in these participant characteristics. ii) We have only limited demographic data on the participants and cannot provide information on race/ethnicity and socio-economic status. It is therefore possible that our results apply only to a certain subgroup of AN patients, calling the generalizability of our study results into question. iii) The considerable overlap in psychopathology symptom scores between the two clinical groups, with not only the group with AN reporting high depression symptoms (which is expected) but also the group with MD reporting high ED psychopathology (see Table 1), calls the precision of the diagnostic assessment into question, even though we administered a well-established standardized interview. Unfortunately, the interrater-reliability for the diagnostic interview could not be determined in the present study as the diagnostic interviews were not audiotaped.
It also has to be mentioned that our study design which aimed to investigate attention biases for disorder-relevant information versus social information by presenting pictures of bodies alongside pictures of faces, is not only a strength of the present study as it taps into a gap in the literature, but also entails limitations: It does not allow to draw conclusions about attention biases for faces or bodies in the absence of the other category or to differentiate between biases towards underweight bodies and away from angry faces in adolescents with AN.
## Clinical Implications
We found adolescents with AN (in comparison to adolescents with MD or adolescents with no mental illness) to show more pronounced attention biases towards ED-related information, i.e., pictures or bodies, at the expense of looking less at socially relevant information, i.e., pictures of faces. These biases for disorder-related versus socially relevant information might be a starting-point for a “dual” cognitive bias modification approach: on the one hand, training individuals with AN to direct their attention away from ED-related information might have a direct positive influence on their ED symptoms, by reducing the impact of maladaptive weight and shape related schemata. On the other hand, training individuals with AN to direct their attention towards social stimuli might improve their socio-emotional functioning and reduce interpersonal difficulties, thereby indirectly positively influencing ED psychopathology.
Our findings could inform and be integrated into existing evidence-based treatments for adolescents with AN, such as family-based treatment or enhanced cognitive behavioral therapy (Dalle Grave et al., 2019; Le Grange et al., 2022), in multiple ways: For example, knowledge about attention biases might help adolescents with AN to identify and understand the mechanisms that maintain their ED psychopathology, which is a major goal of enhanced cognitive behavioral therapy (Dalle Grave et al., 2019). Therapists may measure their patients’ attention biases towards ED-related information and show them their individual eye-tracking trajectories, thereby illustrating an otherwise abstract process that reinforces dysfunctional cognitive schemata. This might even work without an eye-tracker: The therapist could present the patient multiple (sufficiently large) pictures on paper and videotape them while viewing these, so that the direction of gaze is visible and can be fed back to the patient. Attention bias modification trainings aiming to reduce this bias could, in a next step, help to disrupt the maintaining mechanism. Attention bias modification trainings to increase attention towards social stimuli, on the other hand, could be integrated in the additional module of enhanced cognitive behavioral therapy (that addresses interpersonal difficulties in patients for whom such difficulties were identified as maintaining factors; Dalle Grave & Calugi, 2020) or in Phase III of family-based treatment (which focuses on general issues of adolescent development; Lock & Le Grange, 2012). Training adolescents to direct their attention towards social stimuli including angry faces might promote adaptive solving of interpersonal conflicts, alleviate interpersonal difficulties and help adolescents to establish developmentally appropriate relationships not only within but also outside their families. Learning to tolerate others’ angry facial expression instead of avoiding them could also help to reduce experiential avoidance and to build distress tolerance, which are goals of emotion-focused treatments for individuals with AN (Sala et al., 2016).
## Conclusions
The present study contributes to the scarce literature investigating attention biases for disorder-related versus socially relevant information in AN. While all groups of adolescents showed attention biases for body stimuli, these biases were particularly pronounced in adolescents with AN. In turn, adolescents with AN looked less at pictures of faces. This might have a twofold negative impact on AN psychopathology: the increased attention to ED-related information might have a direct negative influence on ED symptoms while the less attending to social information might have an indirect negative influence through the exacerbation of interpersonal difficulties. Even though this might be a promising avenue for cognitive bias modification approaches, we clearly need further research to understand the interplay between biases in the processing of ED-related information and social information and how this altered processing contributes to ED psychopathology.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 36 KB)
## References
1. American Psychiatric AssociationDiagnostic and statistical manual of mental disorders20135American Psychiatric Association. *Diagnostic and statistical manual of mental disorders* (2013.0)
2. Arcelus J, Mitchell AJ, Wales J, Nielsen S. **Mortality rates in patients with anorexia nervosa and other eating disorders: A meta-analysis of 36 studies**. *Archives of General Psychiatry* (2011.0) **68** 724-731. DOI: 10.1001/archgenpsychiatry.2011.74
3. Armstrong T, Olatunji BO. **Eye tracking of attention in the affective disorders: A meta-analytic review and synthesis**. *Clinical Psychology Review* (2012.0) **32** 704-723. DOI: 10.1016/j.cpr.2012.09.004
4. Aspen V, Darcy AM, Lock J. **A review of attention biases in women with eating disorders**. *Cognition & Emotion* (2013.0) **27** 820-838. DOI: 10.1080/02699931.2012.749777
5. Bang L, Rø Ø, Endestad T. **Threat-detection and attentional bias to threat in women recovered from anorexia nervosa: Neural alterations in extrastriate and medial prefrontal cortices**. *European Eating Disorders Review* (2017.0) **25** 80-88. DOI: 10.1002/erv.2494
6. Blechert J, Nickert T, Caffier D, Tuschen-Caffier B. **Social comparison and its relation to body dissatisfaction in bulimia nervosa: Evidence from eye movements**. *Psychosomatic Medicine* (2009.0) **71** 907-912. DOI: 10.1097/PSY.0b013e3181b4434d
7. Bühren K, Schwarte R, Fluck F, Timmesfeld N, Krei M, Egberts K, Pfeiffer E, Fleischhaker C, Wewetzer C, Herpertz-Dahlmann B. **Comorbid psychiatric disorders in female adolescents with first-onset anorexia nervosa**. *European Eating Disorders Review* (2014.0) **22** 39-44. DOI: 10.1002/erv.2254
8. Caglar-Nazali HP, Corfield F, Cardi V, Ambwani S, Leppanen J, Olabintan O, Deriziotis S, Hadjimichalis A, Scognamiglio P, Eshkevari E, Micali N, Treasure J. **A systematic review and meta-analysis of ‘Systems for Social Processes’ in eating disorders**. *Neuroscience & Biobehavioral Reviews* (2014.0) **42** 55-92. DOI: 10.1016/j.neubiorev.2013.12.002
9. Cardi V, Corfield F, Leppanen J, Rhind C, Deriziotis S, Hadjimichalis A, Hibbs R, Micali N, Treasure J. **Emotional processing, recognition, empathy and evoked facial expression in eating disorders: An experimental study to map deficits in social cognition**. *PLoS ONE* (2015.0) **10** e0133827. DOI: 10.1371/journal.pone.0133827
10. Cardi V, Di Matteo R, Corfield F, Treasure J. **Social reward and rejection sensitivity in eating disorders: An investigation of attentional bias and early experiences**. *The World Journal of Biological Psychiatry* (2013.0) **14** 622-633. DOI: 10.3109/15622975.2012.665479
11. Cohen J. *Statistical power analysis for the behavioral sciences* (1988.0)
12. Cornelissen KK, Cornelissen PL, Hancock PJ, Tovée MJ. **Fixation patterns, not clinical diagnosis, predict body size over-estimation in eating disordered women and healthy controls**. *International Journal of Eating Disorders* (2016.0) **49** 507-518. DOI: 10.1002/eat.22505
13. Dalle Grave R, Calugi S. *Cognitive behavior therapy for adolescents with eating disorders* (2020.0)
14. Dalle Grave R, Eckhardt S, Calugi S, Le Grange D. **A conceptual comparison of family-based treatment and enhanced cognitive behavior therapy in the treatment of adolescents with eating disorders**. *Journal of Eating Disorders* (2019.0) **7** 42. DOI: 10.1186/s40337-019-0275-x
15. Dinkler L, Rydberg Dobrescu S, Råstam M, Gillberg IC, Gillberg C, Wentz E, Hadjikhani N. **Visual scanning during emotion recognition in long-term recovered anorexia nervosa: An eye-tracking study**. *International Journal of Eating Disorders* (2019.0) **52** 691-700. DOI: 10.1002/eat.23066
16. Dresler T, Ehlis A-C, Attar CH, Ernst L, Tupak SV, Hahn T, Warrings B, Markulin F, Spitzer C, Löwe B, Deckert J, Fallgatter AJ. **Reliability of the emotional Stroop task: An investigation of patients with panic disorder**. *Journal of Psychiatric Research* (2012.0) **46** 1243-1248. DOI: 10.1016/j.jpsychires.2012.06.006
17. Duval ER, Lovelace CT, Filion DL. **Attention allocation to subliminally presented affective faces in high and low social anxiety**. *International Journal of Psychophysiology* (2020.0) **153** 159-165. DOI: 10.1016/j.ijpsycho.2020.04.017
18. Eddy KT, Tabri N, Thomas JJ, Murray HB, Keshaviah A, Hastings E, Edkins K, Krishna M, Herzog DB, Keel PK, Franko DL. **Recovery from anorexia nervosa and bulimia nervosa at 22-year follow-up**. *The Journal of Clinical Psychiatry* (2017.0) **78** 184-189. DOI: 10.4088/JCP.15m10393
19. Essau CA. **Comorbidity of depressive disorders among adolescents in community and clinical settings**. *Psychiatry Research* (2008.0) **158** 35-42. DOI: 10.1016/j.psychres.2007.09.007
20. Fichter MM, Quadflieg N, Crosby RD, Koch S. **Long-term outcome of anorexia nervosa: Results from a large clinical longitudinal study**. *International Journal of Eating Disorders* (2017.0) **50** 1018-1030. DOI: 10.1002/eat.22736
21. Fonville, L., Giampietro, V., Surguladze, S., Williams, S., & Tchanturia, K. (2014). Increased BOLD signal in the fusiform gyrus during implicit emotion processing in anorexia nervosa. NeuroImage: Clinical, 4, 266–273. 10.1016/j.nicl.2013.12.002
22. Frazier TW, Strauss M, Klingemier EW, Zetzer EE, Hardan AY, Eng C, Youngstrom EA. **A meta-analysis of gaze differences to social and nonsocial information between individuals with and without autism**. *Journal of the American Academy of Child & Adolescent Psychiatry* (2017.0) **56** 546-555. DOI: 10.1016/j.jaac.2017.05.005
23. Fujiwara, E., Kube, V. L., Rochman, D., Macrae-Korobkov, A. K., Peynenburg, V., & The University of Alberta Hospital Eating Disorders Program. (2017). Visual attention to ambiguous emotional faces in eating disorders: Role of alexithymia. European Eating Disorders Review, 25, 451–460. 10.1002/erv.2535
24. García-Blanco A, López-Soler C, Vento M, García-Blanco MC, Gago B, Perea M. **Communication deficits and avoidance of angry faces in children with autism spectrum disorder**. *Research in Developmental Disabilities* (2017.0) **62** 218-226. DOI: 10.1016/j.ridd.2017.02.002
25. Garner M, Mogg K, Bradley BP. **Orienting and maintenance of gaze to facial expressions in social anxiety**. *Journal of Abnormal Psychology* (2006.0) **115** 760-770. DOI: 10.1037/0021-843X.115.4.760
26. Geller J, Cockell SJ, Hewitt PL, Goldner EM, Flett GL. **Inhibited expression of negative emotions and interpersonal orientation in anorexia nervosa**. *International Journal of Eating Disorders* (2000.0) **28** 8-19. DOI: 10.1002/1098-108X(200007)28:1<8::AID-EAT2>3.0.CO;2-U
27. Ghosn F, Perea M, Castelló J, Vázquez MÁ, Yáñez N, Marcos I, Sahuquillo R, Vento M, García-Blanco A. **Attentional patterns to emotional faces versus scenes in children with autism spectrum disorders**. *Journal of Autism and Developmental Disorders* (2019.0) **49** 1484-1492. DOI: 10.1007/s10803-018-3847-8
28. Glauert R, Rhodes G, Fink B, Grammer K. **Body dissatisfaction and attentional bias to thin bodies**. *International Journal of Eating Disorders* (2010.0) **43** 42-49. DOI: 10.1002/eat.20663
29. Goddard E, Treasure J. **Anxiety and social-emotional processing in eating disorders: Examination of family trios**. *Cognitive Therapy and Research* (2013.0) **37** 890-904. DOI: 10.1007/s10608-013-9535-2
30. Green MW, McKenna FP. **Developmental onset of eating related color-naming interference**. *International Journal of Eating Disorders* (1993.0) **13** 391-397. DOI: 10.1002/1098-108X(199305)13:4<391::AID-EAT2260130407>3.0.CO;2-U
31. Happé F, Frith U. **Annual research review: Towards a developmental neuroscience of atypical social cognition**. *Journal of Child Psychology and Psychiatry* (2014.0) **55** 553-577. DOI: 10.1111/jcpp.12162
32. Harris C, Barraclough B. **Excess mortality of mental disorder**. *The British Journal of Psychiatry* (1998.0) **173** 11-53. DOI: 10.1192/bjp.173.1.11
33. Harrison A, Sullivan S, Tchanturia K, Treasure J. **Emotional functioning in eating disorders: Attentional bias, emotion recognition and emotion regulation**. *Psychological Medicine* (2010.0) **40** 1887-1897. DOI: 10.1017/S0033291710000036
34. Harrison A, Tchanturia K, Treasure J. **Attentional bias, emotion recognition, and emotion regulation in anorexia: State or Trait?**. *Biological Psychiatry* (2010.0) **68** 755-761. DOI: 10.1016/j.biopsych.2010.04.037
35. Hatch A, Madden S, Kohn MR, Clarke S, Touyz S, Gordon E, Williams LM. **Emotion brain alterations in anorexia nervosa: A candidate biological marker and implications for treatment**. *Journal of Psychiatry & Neuroscience* (2010.0) **35** 267-274. DOI: 10.1503/jpn.090073
36. Hautzinger, M., Keller, F., & Kühner, C. (2006). BDI-II. Beck-Depressions-Inventar. Revision. Pearson.
37. Heuer K, Rinck M, Becker ES. **Avoidance of emotional facial expressions in social anxiety: The approach–avoidance task**. *Behaviour Research and Therapy* (2007.0) **45** 2990-3001. DOI: 10.1016/j.brat.2007.08.010
38. Herpertz-Dahlmann B, Dempfle A, Egberts KM, Kappel V, Konrad K, Vloet JA, Bühren K. **Outcome of childhood anorexia nervosa—The results of a five- to ten-year follow-up study**. *International Journal of Eating Disorders* (2018.0) **51** 295-304. DOI: 10.1002/eat.22840
39. Hilbert, A., & Tuschen-Caffier, B. (2016). Eating Disorder Examination - Questionnaire. Deutschsprachige Übersetzung, 2. Auflage.
40. Holm S. **A simple sequentially rejective multiple test procedure**. *Scandinavian Journal of Statistics* (1979.0) **6** 65-70
41. Horndasch S, Heinrich H, Kratz O, Mai S, Graap H, Moll GH. **Perception and evaluation of women’s bodies in adolescents and adults with anorexia nervosa**. *European Archives of Psychiatry and Clinical Neuroscience* (2015.0) **265** 677-687. DOI: 10.1007/s00406-015-0603-3
42. Horndasch S, Kratz O, Van Doren J, Graap H, Kramer R, Moll GH, Heinrich H. **Cue reactivity towards bodies in anorexia nervosa–common and differential effects in adolescents and adults**. *Psychological Medicine* (2018.0) **48** 508-518. DOI: 10.1017/S0033291717001994
43. Ioannou K, Fox JR. **Perception of threat from emotions and its role in poor emotional expression within eating pathology**. *Clinical Psychology & Psychotherapy* (2009.0) **16** 336-347. DOI: 10.1002/cpp.632
44. Jaite, C., Hoffmann, F., Glaeske, G., & Bachmann, C. J. (2013). Prevalence, comorbidities and outpatient treatment of anorexia and bulimia nervosa in German children and adolescents. Eating and Weight Disorders, 18, 157–165. 10.1007/s40519-013-0020-4
45. Jiang MY, Vartanian LR. **A review of existing measures of attentional biases in body image and eating disorders research**. *Australian Journal of Psychology* (2018.0) **70** 3-17. DOI: 10.1111/ajpy.12161
46. Joseph C, LoBue V, Rivera LM, Irving J, Savoy S, Shiffrar M. **An attentional bias for thin bodies and its relation to body dissatisfaction**. *Body Image* (2016.0) **19** 216-223. DOI: 10.1016/j.bodyim.2016.10.006
47. Kashdan TB, Goodman FR, Machell KA, Kleiman EM, Monfort SS, Ciarrochi J, Nezlek JB. **A contextual approach to experiential avoidance and social anxiety: Evidence from an experimental interaction and daily interactions of people with social anxiety disorder**. *Emotion* (2014.0) **14** 769-781. DOI: 10.1037/a0035935
48. Kerr-Gaffney J, Jones E, Mason L, Hayward H, Murphy D, Loth E, Tchanturia K. **Social attention in anorexia nervosa and autism spectrum disorder: Role of social motivation**. *Autism* (2022.0). DOI: 10.1177/13623613211060593
49. Kerr-Gaffney J, Mason L, Jones E, Hayward H, Harrison A, Murphy D, Tchanturia K. **Autistic traits mediate reductions in social attention in adults with anorexia nervosa**. *Journal of Autism and Developmental Disorders* (2021.0) **51** 2077-2090. DOI: 10.1007/s10803-020-04686-y
50. Kerr-Gaffney J, Harrison A, Tchanturia K. **Eye-tracking research in eating disorders: A systematic review**. *International Journal of Eating Disorders* (2019.0) **52** 3-27. DOI: 10.1002/eat.22998
51. Kim Y-R, Kim C-H, Park JH, Pyo J, Treasure J. **The impact of intranasal oxytocin on attention to social emotional stimuli in patients with anorexia nervosa: A double blind within-subject cross-over experiment**. *PLoS ONE* (2014.0) **9** e90721. DOI: 10.1371/journal.pone.0090721
52. Klin A, Jones W, Schultz R, Volkmar F, Cohen D. **Visual fixation patterns during viewing of naturalistic social situations as predictors of social competence in individuals with autism**. *Archives of General Psychiatry* (2002.0) **59** 809-816. DOI: 10.1001/archpsyc.59.9.809
53. Kromeyer-Hauschild K, Wabitsch M, Kunze D, Geller F, Geiß HC, Hesse V, von Hippel A, Jaeger U, Johnsen D. **Perzentile für den Body-mass-Index für das Kindes-und Jugendalter unter Heranziehung verschiedener deutscher Stichproben**. *Monatsschrift Kinderheilkunde* (2001.0) **149** 807-818. DOI: 10.1007/s001120170107
54. Labarge AS, Cash TF, Brown TA. **Use of a modified Stroop task to examine appearance-schematic information processing in college women**. *Cognitive Therapy and Research* (1998.0) **22** 179-190. DOI: 10.1023/A:1018732423464
55. Lang K, Lopez C, Stahl D, Tchanturia K, Treasure J. **Central coherence in eating disorders: An updated systematic review and meta-analysis**. *The World Journal of Biological Psychiatry* (2014.0) **15** 586-598. DOI: 10.3109/15622975.2014.909606
56. Lang K, Tchanturia K. **A systematic review of central coherence in young people with anorexia nervosa**. *Journal of Child & Adolescent Behaviour* (2014.0) **2** 140. DOI: 10.4172/2375-4494.1000140
57. Lang PJ, Sidowski JB, Johnson JH, Williams TA. **Behavioral treatment and bio-behavioral assessment: Computer applications**. *Technology in Mental Health Care Delivery Systems* (1980.0) 119-137
58. Lau JYF, Waters AM. **Annual research review: An expanded account of information-processing mechanisms in risk for child and adolescent anxiety and depression**. *Journal of Child Psychology and Psychiatry* (2017.0) **58** 387-407. DOI: 10.1111/jcpp.12653
59. Laux L, Glanzmann P, Schaffner P, Spielberger C. *STAI. Das State-Trait Angstinventar* (1981.0)
60. Lazarov A, Suarez-Jimenez B, Tamman A, Falzon L, Zhu X, Edmondson DE, Neria Y. **Attention to threat in posttraumatic stress disorder as indexed by eye-tracking indices: A systematic review**. *Psychological Medicine* (2019.0) **49** 705-726. DOI: 10.1017/S0033291718002313
61. Lee M, Shafran R. **Information processing biases in eating disorders**. *Clinical Psychology Review* (2004.0) **24** 215-238. DOI: 10.1016/j.cpr.2003.10.004
62. Le Grange D, Eckhardt S, Dalle Grave R, Crosby RD, Peterson CB, Keery H, Lesser J, Martell C. **Enhanced cognitive-behavior therapy and family-based treatment for adolescents with an eating disorder: A non-randomized effectiveness trial**. *Psychological Medicine* (2022.0). DOI: 10.1017/S0033291720004407
63. Liang C-W, Tsai J-L, Hsu W-Y. **Sustained visual attention for competing emotional stimuli in social anxiety: An eye tracking study**. *Journal of Behavior Therapy and Experimental Psychiatry* (2017.0) **54** 178-185. DOI: 10.1016/j.jbtep.2016.08.009
64. Lock J, Le Grange D. *Treatment manual for anorexia nervosa: A family-based approach* (2012.0)
65. Lukas L, Buhl C, Schulte-Körne G, Sfärlea A. **Family, friends, and feelings: The role of relationships to parents and peers and alexithymia in adolescents with anorexia nervosa**. *Journal of Eating Disorders* (2022.0) **10** 143. DOI: 10.1186/s40337-022-00661-3
66. Lundqvist, D., Flykt, A., & Öhman, A. (1998). The Karolinska Directed Emotional Faces - KDEF. Stockholm, Sweden: Department of Clinical Neuroscience, Psychology section, Karolinska Institutet.
67. Margraf, J., Cwik, J. C., Pflug, V., & Schneider, S. (2017). Structured clinical interviews for mental disorders across the life span: Psychometric quality and further developments of the DIPS open access interviews. [Strukturierte klinische Interviews zur Erfassung psychischer Störungen über die Lebensspanne: Gütekriterien und Weiterentwicklungen der DIPS-Verfahren.]. Zeitschrift für Klinische Psychologie und Psychotherapie,46, 176–186. 10.1026/1616-3443/a000430
68. Mason TB, Lesser EL, Dolgon-Krutolow AR, Wonderlich SA, Smith KE. **An updated transdiagnostic review of social cognition and eating disorder psychopathology**. *Journal of Psychiatric Research* (2021.0) **143** 602-627. DOI: 10.1016/j.jpsychires.2020.11.019
69. Mento C, Silvestri MC, Muscatello MRA, Rizzo A, Celebre L, Praticò M, Zoccali RA, Bruno A. **Psychological impact of pro-anorexia and pro-eating disorder websites on adolescent females: A systematic review**. *International Journal of Environmental Research and Public Health* (2021.0) **18** 2186. DOI: 10.3390/ijerph18042186
70. Moussally JM, Brosch T, Van der Linden M. *Time course of attentional biases toward body shapes: The ompact of body dissatisfaction.* (2016.0) **19** 159-168. DOI: 10.1016/j.bodyim.2016.09.006
71. Neuschwander M, In-Albon T, Adornetto C, Roth B, Schneider S. **Interrater-Reliabilität des diagnostischen Interviews bei psychischen Störungen im Kindes- und Jugendalter (Kinder-DIPS)**. *Zeitschrift für Kinder- und Jugendpsychiatrie und Psychotherapie* (2013.0) **41** 319-334. DOI: 10.1024/1422-4917/a000247
72. Norris ML, Boydell KM, Pinhas L, Katzman DK. **Ana and the Internet: A review of pro-anorexia websites**. *International Journal of Eating Disorders* (2006.0) **39** 443-447. DOI: 10.1002/eat.20305
73. Oldershaw A, Hambrook D, Stahl D, Tchanturia K, Treasure J, Schmidt U. **The socio-emotional processing stream in anorexia nervosa**. *Neuroscience & Biobehavioral Reviews* (2011.0) **35** 970-988. DOI: 10.1016/j.neubiorev.2010.11.001
74. Orquin JL, Holmqvist K. **Threats to the validity of eye-movement research in psychology**. *Behavior Research Methods* (2018.0) **50** 1645-1656. DOI: 10.3758/s13428-017-0998-z
75. Oswald WD. *ZVT Zahlen-Verbindungstest-Test. 3., überarbeitete und neu normierte Auflage* (2016.0)
76. Pinhas L, Fok K-H, Chen A, Lam E, Schachter R, Eizenman O, Grupp L, Eizenman M. **Attentional biases to body shape images in adolescents with anorexia nervosa: An exploratory eye-tracking study**. *Psychiatry Research* (2014.0) **220** 519-526. DOI: 10.1016/j.psychres.2014.08.006
77. Platt B, Sfärlea A, Buhl C, Loechner J, Neumüller J, Asperud Thomsen L, Starman-Wöhrle K, Salemink E, Schulte-Körne G. **An eye-tracking study of attention biases in children at high familial risk for depression and their parents with depression**. *Child Psychiatry & Human Development* (2022.0) **53** 89-108. DOI: 10.1007/s10578-020-01105-2
78. Platt B, Waters AM, Schulte-Koerne G, Engelmann L, Salemink E. **A review of cognitive biases in youth depression: Attention, interpretation and memory**. *Cognition and Emotion* (2017.0) **31** 462-483. DOI: 10.1080/02699931.2015.1127215
79. Pool E, Brosch T, Delplanque S, Sander D. **Attentional bias for positive emotional stimuli: A meta-analytic investigation**. *Psychological Bulletin* (2016.0) **142** 79-106. DOI: 10.1037/bul0000026
80. Psychology Software Tools Inc. (2013). E-Prime 2.0 [Computer software]. Sharpsburg, Pennsylvania, USA.
81. Radix, A. K., Sfärlea, A., Rinck, M., Becker E.S., Platt, B., Schulte-Körne, G., & Legenbauer, T. (under review). Watch out! - A path from anxiety to anorexia nervosa through biased attention.
82. Ralph-Nearman C, Achee M, Lapidus R, Stewart JL, Filik R. **A systematic and methodological review of attentional biases in eating disorders: Food, body, and perfectionism**. *Brain and Behavior* (2019.0) **9** e01458. DOI: 10.1002/brb3.1458
83. Reilly JL, Lencer R, Bishop JR, Keedy S, Sweeney JA. **Pharmacological treatment effects on eye movement control**. *Brain and Cognition* (2008.0) **68** 415-435. DOI: 10.1016/j.bandc.2008.08.026
84. Rieger E, Schotte DE, Touyz SW, Beumont PJV, Griffiths R, Russell J. **Attentional biases in eating disorders: A visual probe detection procedure**. *International Journal of Eating Disorders* (1998.0) **23** 199-205. DOI: 10.1002/(SICI)1098-108X(199803)23:2<199::AID-EAT10>3.0.CO;2-W
85. Rydberg Dobrescu S, Dinkler L, Gillberg C, Råstam M, Gillberg C, Wentz E. **Anorexia nervosa: 30-year outcome**. *The British Journal of Psychiatry* (2020.0) **216** 97-104. DOI: 10.1192/bjp.2019.113
86. Sala M, Heard A, Black EA. **Emotion-focused treatments for anorexia nervosa: A systematic review of the literature**. *Eating and Weight Disorders* (2016.0) **21** 147-164. DOI: 10.1007/s40519-016-0257-9
87. Schneider, S., Pflug, V., In-Albon, T., & Margraf, J. (2017). Kinder-DIPS Open Access: Diagnostisches Interview bei psychischen Störungen im Kindes-und Jugendalter 3., aktualisierte und erweiterte Auflage.10.13154/rub.101.90
88. Schneier FR, Kimeldorf MB, Choo TH, Steinglass JE, Wall MM, Fyer AJ, Simpson HB. **Attention bias in adults with anorexia nervosa, obsessive-compulsive disorder, and social anxiety disorder**. *Journal of Psychiatric Research* (2016.0) **79** 61-69. DOI: 10.1016/j.jpsychires.2016.04.009
89. SensoMotoric Instruments GmbH. (2018). BeGaze 3.7 [Computer software]. Teltow, Germany.
90. Sfärlea A, Buhl C, Loechner J, Neumüller J, Asperud Thomsen L, Starman K, Salemink E, Schulte-Körne G, Platt B. **“I am a total… loser” – the role of interpretation biases in youth depression**. *Journal of Abnormal Child Psychology* (2020.0) **48** 1337-1350. DOI: 10.1007/s10802-020-00670-3
91. Sfärlea A, Greimel E, Platt B, Bartling J, Schulte-Körne G, Dieler AC. **Alterations in neural processing of emotional faces in adolescent anorexia nervosa patients–an event-related potential study**. *Biological Psychology* (2016.0) **119** 141-155. DOI: 10.1016/j.biopsycho.2016.06.006
92. Solmi M, Radua J, Olivola M, Croce E, Soardo L, Salazar de Pablo G, Il Shin J, Kirkbride JB, Jones P, Kim JH, Kim JY, Carvalho AF, Seeman M, Correll CU, Fusar-Poli P. **Age at onset of mental disorders worldwide: Large-scale meta-analysis of 192 epidemiological studies**. *Molecular Psychiatry* (2022.0) **27** 281-295. DOI: 10.1038/s41380-021-01161-7
93. Stott N, Fox JRE, Williams MO. **Attentional bias in eating disorders: A meta-review**. *International Journal of Eating Disorders* (2021.0) **54** 1377-1399. DOI: 10.1002/eat.23560
94. Swanson SA, Crow SJ, Le Grange D, Swendsen J, Merikangas KR. **Prevalence and correlates of eating disorders in adolescents. Results from the national comorbidity survey replication adolescent supplement**. *Archives of General Psychiatry* (2011.0) **68** 714-723. DOI: 10.1001/archgenpsychiatry.2011.22
95. Tauro JL, Wearne TA, Belevski B, Filipčíková M, Francis HM. **Social cognition in female adults with anorexia nervosa: A systematic review**. *Neuroscience & Biobehavioral Reviews* (2022.0) **132** 197-210. DOI: 10.1016/j.neubiorev.2021.11.035
96. Theeuwes J, Atchley P, Kramer AF, Monsell S, Driver J. **On the time course of top-down and bottom-up control of visual attention**. *Control of Cognitive Processes: Attention and Performance XVIII* (2000.0) 105-124
97. Tobin LN, Sears CR, Zumbusch AS, von Ranson KM. **Attention to fat-and thin-related words in body-satisfied and body-dissatisfied women before and after thin model priming**. *PLoS ONE* (2018.0) **13** e0192914. DOI: 10.1371/journal.pone.0192914
98. Treasure J, Cardi V. **Anorexia nervosa, theory and treatment: Where are we 35 years on from Hilde Bruch's foundation lecture?**. *European Eating Disorders Review* (2017.0) **25** 139-147. DOI: 10.1002/erv.2511
99. Treasure J, Corfield F, Cardi V. **A three-phase model of the social emotional functioning in eating disorders**. *European Eating Disorders Review* (2012.0) **20** 431-438. DOI: 10.1002/erv.2181
100. Vervoort L, Braun M, De Schryver M, Naets T, Koster EH, Braet C. **A pictorial dot probe task to assess food-related attentional bias in youth with and without obesity: Overview of indices and evaluation of their reliability**. *Frontiers in Psychology* (2021.0) **12** 644512. DOI: 10.3389/fpsyg.2021.644512
101. Vitousek KB, Hollon SD. **The investigation of schematic content and processing in eating disorders**. *Cognitive Therapy and Research* (1990.0) **14** 191-214. DOI: 10.1007/BF01176209
102. von Wietersheim J, Kunzl F, Hoffmann H, Glaub J, Rottler E, Traue HC. **Selective attention of patients with anorexia nervosa while looking at pictures of their own body and the bodies of others, an exploratory study**. *Psychosomatic Medicine* (2012.0) **74** 107-113. DOI: 10.1097/PSY.0b013e31823ba787
103. Wadden TA, Brown G, Foster GD, Linowitz JR. **Salience of weight-related worries in adolescent males and females**. *International Journal of Eating Disorders* (1991.0) **10** 407-414. DOI: 10.1002/1098-108X(199107)10:4<407::AID-EAT2260100405>3.0.CO;2-V
104. Waechter S, Nelson AL, Wright C, Hyatt A, Oakman J. **Measuring attentional bias to threat: Reliability of dot probe and eye movement indices**. *Cognitive Therapy and Research* (2014.0) **38** 313-333. DOI: 10.1007/s10608-013-9588-2
105. Watson KK, Werling DM, Zucker NL, Platt ML. **Altered social reward and attention in anorexia nervosa**. *Frontiers in Psychology* (2010.0) **1** 36. DOI: 10.3389/fpsyg.2010.00036
106. Weeks JW, Howell AN, Goldin PR. **Gaze avoidance in social anxiety disorder**. *Depression and Anxiety* (2013.0) **30** 749-756. DOI: 10.1002/da.22146
107. Wells TT, Clerkin EM, Ellis AJ, Beevers CG. **Effect of antidepressant medication use on emotional information processing in major depression**. *American Journal of Psychiatry* (2014.0) **171** 195-200. DOI: 10.1176/appi.ajp.2013.12091243
108. Williamson DA, Muller SL, Reas DL, Thaw JM. **Cognitive bias in eating disorders: Implications for theory and treatment**. *Behavior Modification* (1999.0) **23** 556-577. DOI: 10.1177/0145445599234003
109. Williamson DA, White MA, York-Crowe E, Stewart TM. **Cognitive-behavioral theories of eating disorders**. *Behavior Modification* (2004.0) **28** 711-738. DOI: 10.1177/0145445503259853
|
---
title: Impact of catheter ablation and subsequent recurrence of atrial fibrillation
on glucose status in patients undergoing continuous glucose monitoring
authors:
- Masako Baba
- Kentaro Yoshida
- Akihiko Nogami
- Yuichi Hanaki
- Yasuaki Tsumagari
- Masayuki Hattori
- Hideyuki Hasebe
- Akito Shikama
- Hitoshi Iwasaki
- Noriyuki Takeyasu
- Masaki Ieda
journal: Scientific Reports
year: 2023
pmcid: PMC10017667
doi: 10.1038/s41598-023-31139-0
license: CC BY 4.0
---
# Impact of catheter ablation and subsequent recurrence of atrial fibrillation on glucose status in patients undergoing continuous glucose monitoring
## Abstract
Although glucose metabolism and atrial fibrillation (AF) have complex interrelationships, the impact of catheter ablation of AF on glucose status has not been well evaluated. Continuous glucose monitoring (CGM) with a FreeStyle Libre Pro (Abbott) was performed for 48 h pre-procedure, during the procedure, and for 72 h post-procedure in 58 non-diabetes mellitus (DM) patients with symptomatic AF and 20 patients with supraventricular or ventricular arrhythmias as a control group. All ablation procedures including pulmonary vein isolation were performed successfully. Glucose levels during procedures consistently increased in the AF and control groups (83.1 ± 16.1 to 110.0 ± 20.5 mg/dL and 83.3 ± 14.7 to 98.6 ± 16.3 mg/dL, respectively, $P \leq 0.001$ for both), and Δ glucose levels (max minus min/procedure) were greater in the AF group than control group ($P \leq 0.001$). There was a trend toward higher mean glucose levels at 72 h after the procedures compared with those before the procedures in both the AF and control groups (from 103.4 ± 15.6 to 106.1 ± 13.0 mg/dL, $$P \leq 0.063$$ and from 100.2 ± 17.1 to 102.9 ± 16.9 mg/dL, $$P \leq 0.052$$). An acute increase in glucose level at the time of early AF recurrence ($$n = 9$$, $15.5\%$) could be detected by simultaneous CGM and ECG monitoring (89.7 ± 18.0 to 108.3 ± 30.5 mg/dL, $$P \leq 0.001$$). In conclusion, although AF ablation caused a statistically significant increase in the glucose levels during the procedures, it did not result in a pathologically significant change after ablation in non-DM patients. Simultaneous post-procedure CGM and ECG monitoring alerted us to possible acute increases in glucose levels at the onset of AF recurrence.
## Introduction
Catheter ablation for atrial fibrillation (AF) has emerged as a well-established treatment option1,2. However, long-term maintenance of sinus rhythm is commonly difficult especially in patients with comorbidities such as metabolic syndrome. Although glucose metabolism disorder is one of the risk factors for occurrence of AF in the general population and for recurrence of AF after catheter ablation3,4, and substantial metabolic alternation in AF pathophysiology has been reported5–7, the impact of catheter ablation on the procedural glucose status has not been well evaluated. Myocardial damage by ablation lesion sets causes local or systemic inflammation8–10 and may modify autonomic nervous activities11,12. Non-cardiac factors, such as use of anesthetics, beta-antagonists, and heparin, ablation-related pain, and psychological stress, may also modify metabolism. These factors can potentially contribute to changes in glucose status and arrhythmia recurrence soon after AF ablation.
A comprehensive record of the changes in glucose levels can now be obtained using continuous glucose monitoring (CGM). CGM helps us to know how glucose status changes over time, and specific patterns of glycemic responses may reflect underlying physiology such as physiological and psychological stress, inflammation, and autonomic remodeling13. Also, simultaneous glucose and electrocardiographic (ECG) monitoring after ablation may document an acute change in the glucose level at the time of early recurrence of AF. These observations may give us insights into the multifactorial mechanisms underlying the clinically significant relation between glucose metabolism disorder and AF.
## Study design and subjects
This study prospectively enrolled 58 patients with symptomatic AF who underwent initial catheter ablation at Ibaraki Prefectural Central Hospital between April 2019 and February 2022. A separate group of 20 patients who underwent catheter ablation of arrhythmias other than AF served as a control group. Their arrhythmias comprised Wolff-Parkinson-White syndrome in 4, atrioventricular nodal reentrant tachycardia in 4, cavo-tricuspid isthmus-dependent atrial flutter in 3, and ventricular premature complexes (VPCs) in 9 patients. No patients in the control group had a history of AF. Excluded patients included those who were previously diagnosed as having diabetes mellitus (DM) by elevated HbA1c level > $6.5\%$ or were taking anti-diabetic medications, those with a history of ablation, and those with symptomatic heart failure. All anti-arrhythmic drugs and beta-blockers were discontinued at least five half-lives before ablation.
## Continuous glucose monitoring
Professional sensors for CGM (FreeStyle Libre Pro; Abbott GmbH & Co. KG) were used. The accuracy of the FreeStyle Libre Pro was reported to show an absolute difference between CGM and the plasma glucose level of $11.4\%$ in patients with diabetes14 and $10.5\%$ in individuals with normoglycemia15. The sensor, which consists of a thin needle placed in the subcutaneous tissue, measures the interstitial glucose level every 15 min for up to 14 days. Its small size allows free movement and performance of normal daily activity and causes less stress among the patients.
The sensor was inserted in the upper arm for at least 7 days in all patients. The CGM data were analyzed for 3 periods: for 48 h before ablation, during ablation, and for 72 h after ablation. The data recorded in the first 24 h were excluded from the analysis due to the need for stabilization between the sensor and the interstitial fluid after insertion of the device, and the 24 h before ablation were also excluded from the analysis because transesophageal echocardiography to exclude thrombus in the left atrial appendage was performed in this period. The period during ablation was defined as that from the femoral puncture to removal of the sheath. The period after ablation was defined as the 72 h beginning immediately after sheath removal. In addition, if AF recurred within 72 h after ablation, the glucose data from 15 min before the onset of AF to 90 min after its onset were analyzed.
## Catheter ablation
All patients fasted for at least 6 h before the procedure and were infused with lactated Ringer’s solution from 2 h before. Transesophageal echocardiography and cardiac computed tomography were also performed prior to the procedure to exclude left atrial thrombus and to reconstruct the left atrial anatomy. The electrophysiological study and catheter ablation were performed under conscious sedation with dexmedetomidine and fentanyl. Unfractionated heparin was administered to maintain an activated clotting time between 300 and 350 s. Surface electrocardiograms and intracardiac electrograms were continuously monitored and stored on an EP-WorkMate recording system (Abbott, Saint Paul, MN). Blood pressure was monitored continuously through a femoral arterial access line. A 6F 20-pole dual-site mapping catheter (BeeAT; Japan Lifeline Co., Ltd., Tokyo, Japan) was inserted through the subclavian vein and positioned in the coronary sinus, right atrium, and superior vena cava throughout the procedure. An intracardiac echocardiography catheter (AcuNav, Biosense Webster, Diamond Bar, CA) was advanced into the right atrium via the femoral approach to guide the transseptal puncture. Two long sheaths (SL0; AF Division, Abbott) were advanced into the left atrium. Pulmonary vein isolation was performed by point-by-point radiofrequency ablation with 3D electroanatomic mapping (CARTO 3 system, Biosense Webster). The radiofrequency current was delivered via a ThermoCool (Biosense Webster) catheter with power up to 35 W. The endpoint was the achievement of bidirectional conduction block between the left atrium and the pulmonary veins. A positive ganglionated plexus (GP) response during ablation was defined as transient ventricular asystole and atrioventricular block16. When non-pulmonary vein ectopies were reproducibly observed with and without continuous infusion of isoproterenol (1–4 µg/min), they were targeted for ablation. The patients with a clinical history of typical atrial flutter or induced flutter during the procedure underwent cavotricuspid isthmus ablation.
## Arrhythmias and biological assessments after ablation
After the procedure, patients remained hospitalized under continuous ECG monitoring for at least 3 days in the AF group and for at least one day in the control group to detect early recurrences of the targeted arrhythmias. For assessment of an acute inflammatory process after ablation, body temperature was measured every 6 h, and blood samples were taken to measure C-reactive protein (CRP) and troponin-T levels at one day after ablation. AF recurrence was defined as the appearance of AF or atrial tachycardia lasting > 30 s.
## Statistical analysis
Variables with a normal distribution are presented as the mean ± SD, and those with a skewed distribution are presented as the median (interquartile range). Categorical variables are expressed as numbers and percentages. For continuous variables, an unpaired Student t-test or the Mann–Whitney U test was used to test for differences between the two groups, whereas for categorical variables, the Fisher’s exact test was used. The changes in the glucose level during the procedure and at the onset of AF recurrence were analyzed by general linear model repeated measure. Logistic regression analysis was performed to determine independent predictors of early (within 72 h after ablation) recurrence of AF. Differences between pre- and post-ablation glucose levels were tested using a paired Student t-test. Statistical significance was set at $P \leq 0.05.$ All statistical analyses were performed using JMP version 12.0 (SAS Institute Inc., Cary, NC).
## Ethics approval and consent to participate
The study protocol was approved by the institutional review board of Ibaraki Prefectural Central Hospital, Kasama, Japan (IRB No. R1-4,) and complied with the principles of the Declaration of Helsinki. Written informed consent was obtained from all patients.
## Patient characteristics
Baseline characteristics of the patients are presented in Table 1. The mean age of the patients was 61.4 years, and $32.1\%$ were female. The age in the AF group was significantly higher and the body surface area (BSA) was greater than those in the control group. The baseline HbA1c was similar between the AF and control groups (5.9 ± $0.5\%$ vs 5.9 ± $0.5\%$, $$P \leq 0.764$$). In the AF group, $50\%$ of patients had paroxysmal AF.Table 1Baseline clinical characteristics. CharacteristicAll patients ($$n = 78$$)AF group ($$n = 58$$)Control group ($$n = 20$$)PFemale, n (%)25 (32.1)14 (24.1)11 (55.0)0.011Age (years)61.4 ± 11.363.1 ± 10.256.6 ± 13.20.026Body mass index (kg/m2)24.9 ± 3.525.3 ± 3.623.6 ± 2.90.055Body surface area (m2)1.8 ± 0.21.8 ± 0.21.7 ± 0.20.005CHADS2 score1.0 (0–1.0)1.0 (0–1.0)0.5 (0–1.0)0.354Paroxysmal AF, n (%)N/A29 (50.0)N/AN/ACoexisting conditions *Sick sinus* syndrome, n (%)11 (14.1)11 (19.0)0 [0]0.036 Hypertension, n (%)35 (44.9)27 (46.6)8 (40.0)0.612 Stroke, n (%)8 (10.3)6 (10.3)2 (10.0)0.965 *Sleep apnea* syndrome, n (%)8 (10.3)8 (13.8)0 [0]0.080 Structural heart disease, n (%)9 (11.5)6 (10.3)3 (15.0)0.574 Ischemic heart disease, n (%)5 (6.4)5 (8.6)0 [0]0.175Echocardiographic findings Ejection fraction (%)63.0 ± 9.063.3 ± 8.761.8 ± 10.40.573 LAVi (mL/m2)34.5 ± 13.236.1 ± 13.627.9 ± 9.40.035Laboratory data HbA1c (%)5.9 ± 0.55.9 ± 0.55.9 ± 0.50.764 eGFR (mL/min/1.73 m2)70.2 ± 15.370.2 ± 15.270.4 ± 15.90.957 BNP (pg/mL)39.0 (13.0–80.0)41.5 (15.0–82.3)26.0 (10.5–66.9)0.331CGM parameters before ablation (48 h) Mean glucose (mg/dL)102.5 ± 15.9103.4 ± 15.6100.2 ± 17.10.453AF atrial fibrillation, LAVi left atrial volume index, eGFR estimated glomerular filtration rate, BNP brain natriuretic peptide, CGM continuous glucose monitoring.
## Procedural characteristics
The total procedure time and radiofrequency ablation time were significantly longer in the AF group than those in the control group (255.0 ± 47.7 vs 166.7 ± 60.5 min, $P \leq 0.001$ and 57.2 [52.1–69.5] vs 9.1 [3.4–12.4] min, $P \leq 0.001$, respectively). There was no significant difference in the CRP level after ablation, but the troponin-T level and body temperature were higher in the AF group than those in the control group (troponin-T: 1.1 [0.9–1.3] vs 0.2 [0–0.3] ng/L, $P \leq 0.001$; body temperature: 37.3 ± 0.4 vs 36.9 ± 0.5 °C, $P \leq 0.001$, respectively) (Table 2). GP responses were observed in 21 patients ($36.2\%$) in the AF group. Table 2Procedural characteristics. CharacteristicAll patients ($$n = 78$$)AF group ($$n = 58$$)Control group ($$n = 20$$)PProcedure time (min)233.2 ± 63.5255.0 ± 47.7166.7 ± 60.5< 0.001Ablation time (min)53.6 (30.9–63.1)57.2 (52.1–69.5)9.1 (3.4–12.4)< 0.001Inflammatory parameters after ablation C-reactive protein (mg/dL)0.5 (0.3–0.7)0.5 (0.4–0.7)0.3 (0.1–0.7)0.087 Troponin-T (ng/L)1.0 (0.6–1.2)1.1 (0.9–1.3)0.2 (0–0.3)< 0.001 Body temperature (°C)37.2 ± 0.437.3 ± 0.436.9 ± 0.5< 0.001AF atrial fibrillation.
## Glucose profiles at baseline (48 h before ablation)
The mean glucose level at baseline (48 h before ablation) was comparable between the AF and control groups (103.4 ± 15.6 vs 100.2 ± 17.1 mg/dL, $$P \leq 0.453$$) (Table 1). In the AF group, the mean glucose level was similar between the patients with paroxysmal AF and those with persistent AF (103.1 ± 11.9 vs 103.7 ± 18.7 mg/dL, $$P \leq 0.905$$). Also, there was no difference in the mean glucose level between patients with sinus rhythm during the pre-ablation period ($$n = 35$$, $60.3\%$) and those with an AF rhythm ($$n = 23$$, $39.7\%$) (102.9 ± 11.6 vs 104.0 ± 20.3 mg/dL, $$P \leq 0.806$$).
## Glucose status during the procedures
The fluctuations in the glucose level during the procedure in the AF and control groups are presented in Fig. 1A,B, respectively. The glucose level increased consistently in both the AF and control groups (from 83.1 ± 16.1 to 110.0 ± 20.5 mg/dL and 83.3 ± 14.7 to 98.6 ± 16.3 mg/dL, $P \leq 0.001$, respectively). The Δ glucose level (maximum minus minimum) during the procedure in the AF group was higher than that in the control group (25.5 [17.0–36.0] vs 12.0 [10.5–18.0] mg/dL, $P \leq 0.001$) (Fig. 1C). In the AF group, there was no difference in the Δ glucose level during the procedure between patients with and without GP responses (26.0 [16.8–37.8] vs 25.0 [16.5–33.8] mg/dL, $$P \leq 0.704$$) (Fig. 1D).Figure 1(A,B) Interstitial glucose levels during and after the procedure in the AF group and the control group. ( C) The Δ glucose level (maximum minus minimal) during the procedure in the AF and control groups. Time zero was defined as the time of femoral puncture. The mean time from femoral puncture to the start of ablation was 75.5 ± 18.8 min in the AF group and 66.8 ± 28.6 min in the control group. The mean time from femoral puncture to sheath removal was 224.9 ± 39.7 min in the AF group and 154.5 ± 66.0 min in the control group. ( D) The Δ glucose level in patients with and without GP responses. AF atrial fibrillation, GP ganglionated plexus.
## Glucose profiles 72 h after ablation
There was a trend toward higher mean glucose levels at 72 h after the procedures than those before the procedures in both the AF and control groups (from 103.4 ± 15.6 to 106.1 ± 13.0 mg/dL, $$P \leq 0.063$$ and from 100.2 ± 17.1 to 102.9 ± 16.9 mg/dL, $$P \leq 0.052$$) (Fig. 2A). In the AF group, there was no increase in the glucose level after ablation in patients without a GP response (103.4 ± 17.8 to 104.9 ± 13.9 mg/dL, $$P \leq 0.442$$), although patients with a GP response had a significant increase (103.3 ± 10.7 to 108.3 ± 11.1 mg/dL, $$P \leq 0.017$$) (Fig. 2B).Figure 2(A) The mean absolute change in the glucose level before and after ablation. ( B) In the AF group, the mean glucose level was separately assessed in patients with and without GP responses. ABL ablation, AF atrial fibrillation, GP ganglionated plexus.
## Characteristics of early recurrence of AF
There was no recurrence of arrhythmia in the control group. Among the 58 patients who underwent AF ablation, 9 patients ($15.5\%$) experienced AF recurrence within 72 h after ablation. The median time to AF recurrence was 31.75 h (range 9.75–63 h). The characteristics of the patients with and without AF recurrences are shown in Table 3.Table 3Baseline clinical characteristics in patients with and without AF recurrence within 72 h after ablation. CharacteristicAF occurring within 72 h after ablationP(+)$$n = 9$$(−)$$n = 49$$Female, n (%)3 (33.3)11 (22.4)0.483Age (years)62.8 ± 12.063.1 ± 10.00.940Body mass index (kg/m2)22.4 ± 2.025.9 ± 3.60.007Body surface area (m2)1.6 ± 0.21.8 ± 0.20.012CHADS2 score0 (0–2.0)1.0 (0–1.0)0.513Paroxysmal AF, n (%)3 (33.3)26 (53.1)0.277Duration of AF (months)48 (17–96)25 (8–51)0.308Coexisting conditions *Sick sinus* syndrome, n (%)4 (44.4)7 (14.3)0.034 Hypertension, n (%)2 (22.2)25 (51.0)0.111 Stroke, n (%)1 (11.1)5 (10.2)0.935 *Sleep apnea* syndrome, n (%)1 (11.1)7 (14.3)0.800 Structural heart disease, n (%)2 (22.2)4 (8.2)0.203 Ischemic heart disease, n (%)0 [0]5 (10.2)0.316Echocardiographic findings Ejection fraction (%)65.4 ± 2.962.9 ± 9.40.429 LAVi (mL/m2)39.0 ± 13.435.6 ± 13.60.491Laboratory data HbA1c (%)5.9 ± 0.55.8 ± 0.50.830 eGFR (mL/min/1.73 m2)64.6 ± 16.071.2 ± 15.00.235 BNP (pg/mL)82.3 (25.8–103.3)39.9 (14.5–73.3)0.128CGM parameter: before ablation period (48 h) Mean glucose (mg/dL)97.3 ± 14.4104.4 ± 15.70.235AF atrial fibrillation, LAVi left atrial volume index, eGFR estimated glomerular filtration rate, BNP brain natriuretic peptide, CGM continuous glucose monitoring.
There was no difference in the mean glucose level before ablation between patients with and without AF recurrence. The body mass index (BMI) and BSA were smaller in the patients with AF recurrence than in those without (22.4 ± 2.0 vs 25.9 ± 3.6 kg/m2, $$P \leq 0.007$$ and 1.6 ± 0.2 vs 1.8 ± 0.2 m2, $$P \leq 0.012$$, respectively). Sick sinus syndrome was more common in the patients with AF recurrence ($44.4\%$ vs $14.3\%$, $$P \leq 0.034$$). No significant differences were noted in the distribution of paroxysmal versus persistent AF or in echocardiographic parameters between the two groups. The ablation time in patients with AF recurrence was significantly longer than that in those without recurrence (73.2 ± 16.0 vs 58.7 ± 15.8 min, $$P \leq 0.015$$). Troponin-T level and body temperature after ablation were higher in the patients with AF recurrence than in those without (1.5 [1.0–1.7] vs 1.1 [0.8–1.2] ng/L, $$P \leq 0.065$$ and 37.6 ± 0.4 °C vs 37.3 ± 0.3 °C, $$P \leq 0.020$$, respectively) (Table 4). Multivariate analysis showed no independent predictors for early AF recurrence (Supplemental Table).Table 4Procedural characteristics in patients with and without AF recurrence within 72 h after ablation. CharacteristicAF occurring within 72 h after ablationP(+)$$n = 9$$(−)$$n = 49$$Procedure time (min)277.2 ± 35.3250.7 ± 48.90.127Ablation time (min)73.2 ± 16.058.7 ± 15.80.015CTI block, n (%)2 (22.2)9 (18.4)0.786SVCI, n (%)2 (22.2)14 (28.6)0.695Focal atrial, n (%)4 (44.8)9 (18.4)0.085GP response during procedure, n (%)5 (55.6)16 (32.7)0.189Inflammatory parameters after AF ablation C-reactive protein (mg/dL)0.4 (0.2–0.9)0.5 (0.4–0.7)0.361 Troponin-T (ng/L)1.5 (1.0–1.7)1.1 (0.8–1.2)0.065 Body temperature (°C)37.6 ± 0.437.3 ± 0.30.020AF atrial fibrillation, CTI cavotricuspid isthmus, SVCI superior vena cava isolation, GP ganglionated plexus.
## Glucose status at the onset of early recurrence
To clarify whether higher glucose level is a cause or effect of AF recurrence, we focused on the acute change in the glucose level at the time of AF recurrence. Nine patients experienced early AF recurrence within 72 h after ablation. To eliminate the influence of diet in the analysis of glycemic variability at the time of AF recurrence, we excluded 3 patients in whom AF recurrence occurred after a meal. In the remaining 6 patients, the onset of AF recurrence occurred up to 90 min before the meal. In this subgroup in which the effect of dietary glycemic change could be excluded, the glucose level increased significantly from 15 min before AF onset to 90 min after AF onset (from 89.7 ± 18.0 to 108.3 ± 30.5 mg/dL, $$P \leq 0.001$$) (Fig. 3), whereas there was no significant increase in the glucose level during fasting in patients without early recurrence of AF ($$n = 49$$) (data not shown).Figure 3Interstitial glucose level immediately after the onset of early recurrence of atrial fibrillation (AF).
## Discussion
To our best knowledge, this is the first study to evaluate both the periprocedural glucose status of non-DM patients undergoing AF ablation and the glucose status at the onset of early AF recurrence.
The potentially important and novel findings of our study are as follows:The glucose levels gradually and consistently increased during the ablation procedure in both the AF and control groups, but the extent of the increase was greater in the AF group than that in the control group. No adverse events such hypo- or hyperglycemia occurred during the procedures. There was a trend toward higher mean glucose levels at 72 h after the procedures than that before the procedures in both the AF and control groups. In the AF group, the presence of GP responses was associated with the increase in the glucose level after ablation. During the 72-h period after ablation, AF recurred in 9 ($16\%$) of the 58 patients. Acute inflammation such as pericarditis characterized by an increase in fever and troponin-T elevation may be associated with these recurrences. Notably, an immediate increase in the glucose level was observed at the time of early AF recurrence in these patients.
The glucose level during the procedures was increased despite fasting in both groups. AF ablation affected the glucose level more significantly than did other procedures, paroxysmal supraventricular tachycardia, and VPCs. One possible mechanism is the increase in the sympathetic activities due to mental and physical stress particularly because of the conscious sedation used in this series. The longer procedure time, longer ablation time, pain typically occurring during left atrial posterior wall ablation, and ganglionated plexus ablation may possibly explain the greater increase in sympathetic activities in the AF group. Another mechanism might be a weakness in glucose tolerance that was originally present in the patients with AF. It is widely known that even patients without previous evidence of DM can have transient hyperglycemia due to external factors and environments, namely “stress hyperglycemia”17. The development of stress hyperglycemia is caused by a highly complex interplay of counter-regulatory hormones such as catecholamines, growth hormone, cortisol, and cytokines18,19. Although patients with DM were excluded from this study, subclinical impairment of glucose tolerance could be present in the AF patients. The AF group was older than the control group, and impaired glucose tolerance is also one of the common risk factors for AF development in the general population20. It is well known that systemic inflammation causes an elevation in blood glucose21,22. However, neither a high fever of > 38 °C nor a clinically significant elevation in the CRP level was observed in this patient series, and it is unclear whether local (atrial) inflammation associated with pericarditis would lead to elevation of the glucose level.
Diabetes leads to increased morbidity and length of stay of surgical patients. One of the reasons is hypo- and hyperglycemia. Hypoglycemia sometimes manifests as drowsiness, which may be wrongly attributed to sedation. NHS Diabetes guidelines for the perioperative management of the adult patient with diabetes recommend glucose monitoring for patients undergoing general anesthesia if the patient receives insulin and the procedure is longer than 1–2 h23,24. However, there are no data, no recommendations, and no guidelines regarding glucose monitoring during catheter ablation despite the long procedure time of over 2 h and the use of conscious sedation or general anesthesia. Although hypo- and hyperglycemias could occur in theory during ablation of AF, the present results first provide electrophysiologists with actual evidence of little risk from these factors and little need for routine glucose measurements during the procedures in non-diabetes patients. However, because only non-diabetes patients were included in the present study, the presence or absence of hypo- and hyper-glycemia in diabetes-patients should be of great interest to electrophysiologists to improve safety of the procedures. In this regard, the present study may provide future direction for further studies investigating glucose status during ablation procedures.
In this study, there was a trend toward higher mean glucose levels at 72 h after the procedures than that before the procedures in both the AF and control groups. Although this was a small change (% increase of ~ $3\%$), and at least in non-DM patients, the effect of ablation may be transient and limited to the period during the procedure, this study implied that autonomic change by GP ablation may affect glucose metabolism after ablation in patients with AF.
One more important result was the acute increase in the glucose level following AF onset. There is no data, to our knowledge, regarding the acute effect of AF onset on the glucose level in humans because continuous monitoring of both ECG and glucose level has been difficult to perform. The patients with immediate AF recurrence were characterized by small BSA and BMI, longer ablation and radiofrequency times, and a greater increase in body temperature and CRP and troponin-T levels after ablation, suggesting the higher likelihood of acute inflammation probably associated with pericarditis8–10. Symptoms such as palpitations and chest discomfort, anxiety, and hemodynamic change in addition to inflammation and autonomic imbalance could also be reasons for the glucose elevation. These results may provide an important message to physicians treating patients with DM and AF in that unexplained hyperglycemia might occur when AF events are frequent especially when the AF is asymptomatic. The simultaneous application of ECG monitoring and CGM over a longer-term period is currently difficult, but it may provide further insights into glucose metabolism as a cause and effect of AF, i.e., a vicious circle, in future studies.
Several limitations should be acknowledged. First, this is a single-center study, and the number of patients is relatively small. Second, calorie intake at home before ablation could not completely controlled. However, eating habits were reviewed when instruction on CGM was provided in the outpatient clinic, and calorie intake during hospital admission was consistent (1800 kcal). Third, because the present study is the first, to our knowledge, to evaluate the glucose level during ablation procedures and, therefore, focused on patients who are estimated to have higher risk for hypo- and hyperglycemia, i.e., patients with AF, glucose levels during other types of procedures such as percutaneous coronary artery intervention and pacemaker implantation, should also be of great interest to cardiologists. These additional evaluations may contribute to further clarification of the mechanisms involved in the interplay between AF and DM. Finally, patients with DM were excluded from the study because this was a first study to evaluate the pure effect of AF ablation on glucose metabolism. Further studies including those on patients with DM may provide insights into the pathological interplay between glucose metabolism, the development of AF, and AF ablation.
In conclusion, AF ablation resulted in a significant increase in the glucose level during the procedures, but it did not cause a pathologically significant change early (within 72 h) after ablation in non-DM patients. Simultaneous CGM and ECG monitoring post-procedure alerted us to the acute increase in the glucose levels at the onset of AF recurrence, which suggested multifactorial contributions to glucose metabolism such as inflammation, autonomic imbalance, mental stress, and hemodynamic impairment.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-31139-0.
## References
1. Haïssaguerre M. **Spontaneous initiation of atrial fibrillation by ectopic beats originating in the pulmonary veins**. *N. Engl. J. Med.* (1998) **339** 659-666. DOI: 10.1056/NEJM199809033391003
2. Parameswaran R, Al-Kaisey AM, Kalman JM. **Catheter ablation for atrial fibrillation: Current indications and evolving technologies**. *Nat. Rev. Cardiol.* (2021) **18** 210-225. DOI: 10.1038/s41569-020-00451-x
3. Zheng Y. **Meta-analysis of metabolic syndrome and its individual components with risk of atrial fibrillation in different populations**. *BMC Cardiovasc. Disord.* (2021) **21** 90. DOI: 10.1186/s12872-021-01858-1
4. Aune D. **Diabetes mellitus, blood glucose and the risk of atrial fibrillation: A systematic review and meta-analysis of cohort studies**. *J. Diabetes Complicat.* (2018) **32** 501-511. DOI: 10.1016/j.jdiacomp.2018.02.004
5. Mayr M. **Combined metabolomic and proteomic analysis of human atrial fibrillation**. *J. Am. Coll. Cardiol.* (2008) **51** 585-594. DOI: 10.1016/j.jacc.2007.09.055
6. De Souza AI. **Proteomic and metabolomic analysis of atrial profibrillatory remodelling in congestive heart failure**. *J. Mol. Cell. Cardiol.* (2010) **49** 851-863. DOI: 10.1016/j.yjmcc.2010.07.008
7. Harada M, Melka J, Sobue Y, Nattel S. **Metabolic considerations in atrial fibrillation—Mechanistic insights and therapeutic opportunities**. *Circ. J.* (2017) **81** 1749-1757. DOI: 10.1253/circj.CJ-17-1058
8. Richter B. **Markers of oxidative stress after ablation of atrial fibrillation are associated with inflammation, delivered radiofrequency energy and early recurrence of atrial fibrillation**. *Clin. Res. Cardiol.* (2012) **101** 217-225. DOI: 10.1007/s00392-011-0383-3
9. Koyama T. **Comparison of characteristics and significance of immediate versus early versus no recurrence of atrial fibrillation after catheter ablation**. *Am. J. Cardiol.* (2009) **103** 1249-1254. DOI: 10.1016/j.amjcard.2009.01.010
10. Chang SL. **Characteristics and significance of very early recurrence of atrial fibrillation after catheter ablation**. *J. Cardiovasc. Electrophysiol.* (2011) **22** 1193-1198. DOI: 10.1111/j.1540-8167.2011.02095.x
11. Hsieh MH. **Alterations of heart rate variability after radiofrequency catheter ablation of focal atrial fibrillation originating from pulmonary veins**. *Circulation* (1999) **100** 2237-2243. DOI: 10.1161/01.CIR.100.22.2237
12. Bauer A. **Effects of circumferential or segmental pulmonary vein ablation for paroxysmal atrial fibrillation on cardiac autonomic function**. *Heart Rhythm* (2006) **3** 1428-1435. DOI: 10.1016/j.hrthm.2006.08.025
13. Hall H. **Glucotypes reveal new patterns of glucose dysregulation**. *PLoS Biol.* (2018) **16** e2005143. DOI: 10.1371/journal.pbio.2005143
14. Bailey T, Bode BW, Christiansen MP, Klaff LJ, Alva S. **The performance and usability of a factory-calibrated flash glucose monitoring system**. *DiabetesTtechnol. Ther.* (2015) **17** 787-794
15. Fechner E, Opt Eyndt C, Mulder T, Mensink RP. **Diet-induced differences in estimated plasma glucose concentrations in healthy, non-diabetic adults are detected by continuous glucose monitoring-a randomized crossover trial**. *Nutr. Res.* (2020) **80** 36-43. DOI: 10.1016/j.nutres.2020.06.001
16. Po SS, Nakagawa H, Jackman WM. **Localization of left atrial ganglionated plexi in patients with atrial fibrillation**. *J. Cardiovasc. Electrophysiol.* (2009) **20** 1186-1189. DOI: 10.1111/j.1540-8167.2009.01515.x
17. Dungan KM, Braithwaite SS, Preiser JC. **Stress hyperglycaemia**. *Lancet* (2009) **373** 1798-1807. DOI: 10.1016/S0140-6736(09)60553-5
18. Mizock BA. **Alterations in fuel metabolism in critical illness: hyperglycemia**. *Best Pract. Res. Clin. Endocrinol. Metab.* (2001) **15** 533-551. DOI: 10.1053/beem.2001.0168
19. Barth E. **Glucose metabolism and catecholamines**. *Crit. Care Med.* (2007) **35** S508-518. DOI: 10.1097/01.CCM.0000278047.06965.20
20. Yang S. **Risk of atrial fibrillation in relation to the time course of type 2 diabetes mellitus and fasting blood glucose**. *Am. J. Cardiol.* (2019) **124** 1881-1888. DOI: 10.1016/j.amjcard.2019.09.009
21. Pradhan AD, Manson JE, Rifai N, Buring JE, Ridker PM. **C-reactive protein, interleukin 6, and risk of developing type 2 diabetes mellitus**. *JAMA* (2001) **286** 327-334. DOI: 10.1001/jama.286.3.327
22. Barlow J, Solomon TPJ, Affourtit C. **Pro-inflammatory cytokines attenuate glucose-stimulated insulin secretion from INS-1E insulinoma cells by restricting mitochondrial pyruvate oxidation capacity—Novel mechanistic insight from real-time analysis of oxidative phosphorylation**. *PLoS ONE* (2018) **13** e0199505. DOI: 10.1371/journal.pone.0199505
23. Dhatariya K. **NHS Diabetes guideline for the perioperative management of the adult patient with diabetes**. *Diabet. Med.* (2012) **29** 420-433. DOI: 10.1111/j.1464-5491.2012.03582.x
24. Joshi GP. **Society for Ambulatory Anesthesia consensus statement on perioperative blood glucose management in diabetic patients undergoing ambulatory surgery**. *Anesth. Analg.* (2010) **111** 1378-1387. DOI: 10.1213/ANE.0b013e3181f9c288
|
---
title: Acetate attenuates hyperoxaluria-induced kidney injury by inhibiting macrophage
infiltration via the miR-493-3p/MIF axis
authors:
- Wei Zhu
- Chengjie Wu
- Zhen Zhou
- Guangyuan Zhang
- Lianmin Luo
- Yang Liu
- Zhicong Huang
- Guoyao Ai
- Zhijian Zhao
- Wen Zhong
- Yongda Liu
- Guohua Zeng
journal: Communications Biology
year: 2023
pmcid: PMC10017675
doi: 10.1038/s42003-023-04649-w
license: CC BY 4.0
---
# Acetate attenuates hyperoxaluria-induced kidney injury by inhibiting macrophage infiltration via the miR-493-3p/MIF axis
## Abstract
Hyperoxaluria is well known to cause renal injury and end-stage kidney disease. Previous studies suggested that acetate treatment may improve the renal function in hyperoxaluria rat model. However, its underlying mechanisms remain largely unknown. Using an ethylene glycol (EG)-induced hyperoxaluria rat model, we find the oral administration of $5\%$ acetate reduced the elevated serum creatinine, urea, and protected against hyperoxaluria-induced renal injury and fibrosis with less infiltrated macrophages in the kidney. Treatment of acetate in renal tubular epithelial cells in vitro decrease the macrophages recruitment which might have reduced the oxalate-induced renal tubular cells injury. Mechanism dissection suggests that acetate enhanced acetylation of Histone H3 in renal tubular cells and promoted expression of miR-493-3p by increasing H3K9 and H3K27 acetylation at its promoter region. The miR-493-3p can suppress the expression of macrophage migration inhibitory factor (MIF), thus inhibiting the macrophages recruitment and reduced oxalate-induced renal tubular cells injury. Importantly, results from the in vivo rat model also demonstrate that the effects of acetate against renal injury were weakened after blocking the miR-493-3p by antagomir treatment. Together, these results suggest that acetate treatment ameliorates the hyperoxaluria-induced renal injury via inhibiting macrophages infiltration with change of the miR-493-3p/MIF signals. Acetate could be a new therapeutic approach for the treatment of oxalate nephropathy.
Acetate promotes miR-493-3p expression, which in turn suppresses the expression of macrophage migration inhibitory factor, leading to decreased macrophages recruitment and reduced hyperoxaluria-induced renal injury in mice.
## Introduction
Hyperoxaluria results from either inherited disorders of glyoxylate metabolism leading to hepatic oxalate overproduction (primary hyperoxaluria) or increased intestinal oxalate absorption (secondary hyperoxaluria)1. Hyperoxaluria can cause not only nephrolithiasis and nephrocalcinosis, but also renal tubular damage, interstitial inflammation and fibrosis, and eventually end-stage renal disease2,3. Currently, available interventions aimed at the reduction of oxalate production include FDA-approved siRNA suppressing the expression of glycolate oxidase as well as pyridoxine in patients with primary hyperoxaluria4,5 and the use of an oxalate-reduced diet and calcium supplementation in patients with enteric hyperoxaluria6. No therapies are yet known that blunt the effect of hyperoxaluria-induced inflammation and fibrosis in the kidney associated with renal failure.
Short-chain fatty acids (SCFAs) are end products from the fermentation of dietary fibers by the intestinal microbiota7,8. The most abundant SCFA is acetate. Recently, albeit limited, studies have attempted to use acetate therapeutically in animal and cell models of kidney injuries, such as ischemia-reperfusion-induced acute kidney injury, and diabetic nephropathy9,10. In our previous study, we first found exogenous acetate could improve renal function in rat models of hyperoxaluria11,12. This is independent on a decreased in urinary oxalate excretion and calcium oxalate crystals deposition in the kidney. However, how acetate ameliorates hyperoxaluria-induced renal injury and its underlying mechanisms remain incompletely understood.
Because hyperoxaluria nephropathy has an important inflammatory component yet acetate has anti-inflammatory properties, we investigated whether acetate treatment could protect rats from hyperoxaluria-induced kidney injury. Furthermore, we investigated whether this protection could involve direct modulation of the inflammatory process and or ameliorating of the macrophages infiltration in the hyperoxaluria rat model.
## Acetate-treatment ameliorates hyperoxaluria-induced renal injury and fibrosis
We examined the effect of acetate on renal injury and fibrosis using an ethylene glycol (EG) induced hyperoxaluria. 8-week-old Sprague-Dawley rats received $1\%$ EG in drinking water for 4 weeks to induce hyperoxaluria. In the meantime, rats were treated with $5\%$ acetate (2 ml/kg) or distilled H2O everyday by gavage. The results showed that acetate treatment diminished levels of serum creatinine and urea, and renal weight in hyperoxaluria rats while acetate treatment did not influence the urine oxalate levels (Fig. 1a, b). In addition, the increase in the percentage of necrotic tubules in the hyperoxaluria rats was significantly recovered after acetate treatment (Fig. 1c).Fig. 1Acetate treatment protects against hyperoxaluria-induced renal injury.a Detection of 24-h oxalate excretion in urine samples of each group of rats. b Serum creatinine and BUN levels, kidney weight, and renal damage degrade for each group. c Representative histologic kidney images of PAS, TUNEL, 8-ohdg, αSMA and Masson’s trichrome stain (MTS). Renal damage was evaluated by scoring percentage of necrotic tubules in PAS sections. d Immunostaining of IL-1β, TNFα, CD68 and CD86 in kidney sections. e Relative transcript levels of genes in renal inflammation, fibrosis in the kidney tissue of each group of rats were measured using quantitative real-time PCR (q-PCR). f The expressions of renal inflammation factors and fibrosis-related protein in the kidney tissue were measured using Western Blot. $$n = 6$$ for each group. For (c) and (d), quantitations are at the right. Ctrl, control. EG, ethylene glycol. Ac, acetate. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
We further evaluated renal apoptosis, fibrosis, and inflammation by immunohistochemistry (IHC) and quantitative real-time PCR (Q-PCR). By IHC analysis, acetate treatment significantly reduced the level of renal tubular epithelial cell apoptosis and oxidative stress (Fig. 1c, TUNEL assay and 8-ohdg expression). The acetate treatment also decreased the number of myofibroblasts, the fibrotic area and interstitial collagen deposition in hyperoxaluria rats, which was shown by the anti-α-smooth muscle actin (αSMA) staining and Masson’s trichrome staining (MTS) (Fig. 1c). The inflammatory cytokines (IL-1β and TNFα) were also diminished by acetate treatment. In addition, a low frequency of infiltrating macrophages (CD68+ and CD86+) was observed in acetate-treated rats (Fig. 1d). Q-PCR indicated a significant reduction in the expression of renal fibrosis-related genes (fn-1, αSMA, Col1a1, Col3a1, Ctgf, fsp1, KIM) and inflammatory cytokines (Tgfα, Tgfβ, Il-1b, Il-4, Il-6, CD44, CD68) in the kidney of the acetate-treatment group (Fig. 1e). Results from Western blot also showed a reduction in the expression of renal fibrosis-related proteins (αSMA, Fn-1, Col3a1) and inflammatory factors (IL-1β, IL-6, TGF-β) in the kidney of acetate-treatment group (Fig. 1f).
In summary, results from Fig. 1a–f demonstrated that acetate treatment ameliorated the progression of renal injury, fibrosis, macrophages infiltration, and local inflammation in the hyperoxaluria rats.
## Acetate decreases oxalate-induced renal tubular cell injury via inhibiting macrophages infiltration
As recent studies indicate that macrophage infiltration may play a key role in the progression of hyperoxaluria nephropathy, we were interested in testing the impact of macrophages on oxalate-induced renal cells injury13. We developed an in vitro system composed of renal tubular cells and macrophages to simulate hyperoxaluria conditions (see outline in Fig. 2a). Human renal epithelial HK-2 cells were co-cultured with phorbol 12-myristate 13-acetate (PMA)-induced THP-1 macrophages (MΦs). Oxalate was exposed to HK-2 cells after 48-h co-culture and the renal cell injury (cytotoxicity) was evaluated via measuring lactate dehydrogenase (LDH) activity. We found that HK-2 cells co-cultured with THP-1 cells in Transwell systems significantly increased the oxalate-induced renal cell injury (Fig. 2b). In addition, coculturing with THP-1 cells elicited remarkable upregulation of the proinflammatory cytokines Ccl-2, Ccl-3, Ccl-4, Ccl-5 and Il-1b in HK-2 cells after exposure to oxalate condition (Fig. 2c). Similar results were observed when we replaced the HK-2 cells with mouse cortical collecting duct M-1 cells as well as replaced the THP-1 macrophages with mouse RAW264.7 macrophages (Fig. 2b, c). These data showed that macrophages could promote oxalate-induced renal inflammation and cells injury. Fig. 2Acetate decreases oxalate-induced renal tubular cell injury via inhibiting macrophage infiltration.a In co-culture system 4 groups were established, including Ctrl (control)—renal tubular cells cultivated in normal culture medium, Ox (Oxalate)—renal tubular cells cultivated in culture medium containing 0.5 mM oxalate, Ox+MΦ—renal tubular cells co-cultured with macrophages in culture medium containing 0.5 mM oxalate, MΦ—renal tubular cells co-cultured with macrophages in normal culture medium. b Lactate dehydrogenase (LDH) release measurement in renal tubular cells co-cultured with macrophages in 0.5 mM oxalate condition. c The Q-PCR analysis of inflammation-related gene expression in renal tubular cells co-cultured with macrophages in 0.5 mM oxalate condition. d Experimental outline for macrophages recruitment assay. The conditioned medium (CM) were collected from renal tubular cells treated with oxalate and/or acetate for 24 h. For macrophage recruitment assay, 1 × 105/well MΦs were added in the upper chambers, and the CM were placed into the lower chambers of transwell plates. e Macrophage recruitment to the normal culture medium without renal tubular cells. f The macrophages recruitment to CM from renal tubular cells treated with/without acetate were shown. $$n = 3$$ for each group *$P \leq 0.05$, **$P \leq 0.01.$ # $P \leq 0.05$ compared with Ox group.
Next we tested the impact of the acetate on the macrophages infiltration to renal epithelial cells. We added acetate in HK-2 cells and examined the impact on the THP-1 macrophages recruitment to the renal epithelial HK-2 cells (see outline in Fig. 2d). Results showed that neither acetate nor oxalate affected the THP-1 macrophage recruitment in cell-free culture medium (Fig. 2e). When HK-2 cells exposure to oxalate, the THP-1 macrophage recruitment to the HK-2 conditioned medium (CM) increased significantly using the Transwell migration system. Conversely, adding acetate to the HK-2 cells inhibited the HK-2 cells CM capacity to more recruit the THP-1 macrophages. Similar results were also observed when we placed the THP-1 macrophages/HK-2 cells with RAW264.7 macrophages/M-1 cells (Fig. 2f).
Together, results from Fig. 2a–f demonstrated that acetate treatment in renal tubular cells inhibit the macrophages infiltration which could promote oxalate-induced renal inflammation and cells injury.
## Mechanism dissection of how acetate can alter the macrophages recruitment: via suppressing the macrophages migration inhibitory factor (MIF) expression
To dissect the mechanism of how treatment of acetate in renal epithelial cells can inhibit the macrophages infiltration, we applied Western blot-based cytokine array analysis to screen inflammatory cytokines in HK-2 cells CM that potentially involved macrophages recruitment. The results revealed that the expression of MIF was altered most significantly after acetate treatment (Fig. 3a). In addition, we also confirmed the MIF protein expression of renal epithelial cells was also decreased after treatment of acetate via ELISA assay (Fig. 3b). Importantly, the MIF expression in the kidney was decreased after treatment of acetate in hyperoxaluria rats (Fig. 3c). We therefore decided to further study the impact of MIF on the acetate-altered macrophages recruitment. Fig. 3Acetate inhibits macrophage recruitment via modulating the MIF signals.a Cytokine assay of different CM from HK-2 cells. CM of HK-2 treated with/without acetate were collected after 24 h incubation. MIF showed the most obvious decrease in CM from HK-2 treated with acetate (yellow squares). b The level of MIF in the CM of renal tubular cells was detected by ELISA. c Immunostaining of MIF in kidney sections of each group of rats. $$n = 6$$ for each group. d Treatment of recombinant MIF in renal epithelial cells could partially reverse the acetate-decreased MΦs recruitment. * $P \leq 0.05$, **$P \leq 0.01.$
Using an interruption approach, we confirmed that treatment of recombinant MIF in renal epithelial cells could partially reverse the acetate-decreased macrophages recruitment, suggesting that treatment of acetate in renal epithelial cells may function via suppressing the MIF expression in the CM to inhibit MΦs recruitment (Fig. 3d). Similar results were also observed when we placed the THP-1 macrophages/HK-2 cells with RAW264.7 macrophages/M-1 cells (Fig. 3d).
Together, results from Fig. 3a–d suggest that treatment of acetate in renal epithelial cells may function via downregulating the MIF signals to inhibit the macrophages recruitment.
## Mechanism dissection of how acetate suppresses MIF protein expression: via upregulating the miR-493-3p
The finding that acetate down-regulated MIF expression at the protein level but not at the mRNA level both in the animal model (Fig. 4a) and cultured cells (Fig. 4b) suggested that MIF expression is regulated at the post-transcriptional level, involving mechanisms such as differential miRNA expression. To directly test this hypothesis, we examined the expression of miRNAs that potentially regulated MIF based on the search of online databases (DIANA-miRGen, MicroCosm Targets, RNA22) and published literature12,14. Results suggested that 5 miRNAs (miR-363-5p, miR-493-3p, miR-629-3p, miR-1293, miR-1537-3p) were likely candidates that were up-regulated by acetate in both in vivo (Fig. 4c) and in vitro models (Fig. 4d). We further assayed the consequences on MIF expression after directly transfection of these 5 miRNAs mimetics into HK-2 and M-1 cells, and results suggested that miR-493-3p was the best candidate for further study since altering this miRNA significantly suppress MIF expression (Fig. 4e, f).Fig. 4Acetate modulates MIF via upregulation of miR-493-3p in renal tubular cells.a q-PCR analysis of MIF mRNA expression in kidney from rats. b q-PCR analysis of MIF mRNA expression after 2 mM sodium acetate and/or 0.5 mM oxalate treatment for 24 h in HK-2 or M-1 cells. c 26 potential miRNAs candidates were screened by q-PCR assay in kidney from rats. d q-PCR analysis of 5 miRNA expressions after 2 mM sodium acetate and/or 0.5 mM oxalate treatment for 24 h in HK-2 or M-1 cells. e HK-2 or M-1 cells were transfected with 5 candidate miRNAs mimic or a negative control (NC). MIF expression was analyzed 48 h by Western blot. GAPDH serves as a loading control. f The protein expression levels of MIF in the CM of HK-2 and M-1 cells after transfection of 5 candidate miRNAs were assessed by ELISA. g HK-2 and M-1 cells were transfected with the miR-493-3p inhibitor or NC. 24 h later cells were treated with 0.5 mM oxalate and/or 2 mM sodium acetate. MIF expression was analyzed 24 h later by Western blot. h Macrophages recruitment to the CM from HK-2 cells (upper) and M-1 cells (lower) with four groups (0.5 mM oxalate, 0.5 mM oxalate + 2 mM sodium acetate, 0.5 mM oxalate + 2 mM sodium acetate + miR-493-3p inhibitor, 0.5 mM oxalate + miR-493-3p inhibitor). i Macrophages recruitment to the CM from HK-2 cells (upper) and M-1 cells (lower) with two groups (0.5 mM oxalate, 0.5 mM oxalate + miR-493-3p mimetic). j Co-transfection of MIF 3′UTR constructs containing wild type and mutant seed regions with miR-493-3p into HEK-293 cells and luciferase assay was applied to detect the luciferase activity. Ctrl, control. EG, ethylene glycol. Ac, acetate. Ox, oxalate. Data are from 6 rats in each group. n.s, not significant, *$P \leq 0.05$, **$P \leq 0.01$, ****$P \leq 0.0001.$ # $P \leq 0.05$ compared with Ox group.
As expected, results from the interruption approach via transient transfection with miR-493-3p antisense inhibitor led to partially reverse acetate-suppressed MIF expression in HK-2 and M-1 cells (Fig. 4g). The consequences of such reversion may then lead to partially reverse the acetate-decreased MΦs recruitment (Fig. 4h). In addition, transfection of miR-493-3p mimetic mimicked acetate effects in decreasing MΦs recruitment in HK-2 and M-1 cells (Fig. 4i).
Next, to directly test that the acetate-induced miR-493-3p could suppress the MIF expression, we searched and identified the potential binding sites of miR-493-3p in the 3′UTR of MIF mRNA, and generated a luciferase reporter construct bearing the 3′UTR of MIF gene using a dual-luciferase reporter as well as a mutated version at the predicted target sites. The luciferase assay results revealed that miR-493-3p could suppress luciferase expression of the wild-type MIF 3′UTR construct, but not the mutant MIF 3′UTR construct, thus miR-493-3p could directly target MIF 3′UTR to suppress its expression (Fig. 4j).
Together, results from Fig. 4a–j suggested that acetate can suppress MIF protein expression via upregulating the miR-493-3p expression in the renal epithelial cells.
## Mechanism dissection of how acetate promotes miR-493-3p expression: via restoring histone acetylation
To dissect the molecular mechanisms of how acetate promoted the expression of miR-493-3p, we first focused on histone acetylation since previous studies showed that acetate could function as an epigenetic metabolite to regulate gene expression12,15. We treated HK-2 and M-1 cells with acetate under hyper-oxalate conditions and found that acetate counteracted the decline of histone acetylation from hyper-oxalate treatment. Of particular interest, acetate induced a significant increase of H3K9 and H3K27 acetylation levels, but not H3K36 and H3K56 acetylation levels (Fig. 5a), indicating that acetate rescued oxalate-reduced histone acetylation with particular specificity. In addition, using IHC assays we found that EG treatment in rats significantly weakens renal H3K9ac and H3K27ac signal, and acetate treatment reversed this decrease (Fig. 5b).Fig. 5Acetate activated miR-493-3p through epigenetic regulation.a Acetate rescued hyper-oxalate-reduced H3K9 and H3K27 acetylation levels. HK-2 or M-1 cells were treated with or without 2 mM sodium acetate under hyper-oxalate condition (0.5 mM) for 24 h. The histone acetylation levels were determined by Western blot. Total H3 served as a loading control. b IHC staining of H3K9ac and H3K27ac in kidney tissues from each group rats (amplification × 200). $$n = 6$$ for each group. c ChIP-qPCR assays showing histone acetylation enrich at miR-493-3p promoter region in HK-2 cells treated with or without 2 mM sodium acetate under normal or hyper-oxalate condition (0.5 mM) for 24 h. Rabbit IgG was included as negative control. For (b), quantitations are at the right. Ctrl, control. EG, ethylene glycol. Ac, acetate. Ox, oxalate. n.s, not significant, *$P \leq 0.05$, **$P \leq 0.01$,***$P \leq 0.001.$
To link the genome-wide histone acetylation change with locus-specific transcription of miR-493-3p, we carried out chromatin immunoprecipitation (ChIP)-qPCR assays for histones located at presumptive promoters of miR-493-3p and found that the acetylation levels (H3K9ac and H3K27ac) at miR-493-3p promoter were repressed after oxalate treatment, which were derepressed by acetate treatment in HK-2 cells (Fig. 5c).
Collectively, these data support the notion that acetate promotes the expression of miR-493-3p through epigenetic regulations.
## In vivo miR-493-3p is critical for the acetate effects in decreasing hyperoxaluria-induced kidney injury and fibrosis
To directly test whether the acetate-induced miR-493-3p is mediating the effect of acetate in decreasing kidney injury and fibrosis in hyperoxaluria rat model, we performed an experiment in which chemically modified antisense oligonucleotides12,16 specific to miR-493-3p (Antagomir-493-3p) was injected intraperitoneally (30 nmol/kg/week) into rats that had received 2 ml/kg/day $5\%$ acetate by gavage and $1\%$ EG in drinking water (see detail in Fig. 6a). The results reveled that treatment of antagomir-493-3p could partly reverse the acetate effects in diminishing levels of serum creatinine and urea, and renal weight in hyperoxaluria rats (Fig. 6b). As expected, antagomir-493-3p administration could also partly reverse the acetate effecting in improving renal apoptosis, fibrosis, and inflammation using IHC (Fig. 6c, d). In addition, IHC staining also showed that adding antagomir-493-3p led to reverse the effect of acetate-suppressed MIF expression (Fig. 6d). Finally, we assayed the infiltration of macrophages in renal tissues, and found acetate-treated rats had lower renal expression of CD68 and CD86 (two widely used marker of rat macrophages). As expected, adding antagomir-493-3p could then lead to partially reverse the acetate-suppressed macrophages infiltration (Fig. 6d).Fig. 6Antagomir-493-3p treatment attenuated acetate effects of regulating macrophages recruitment and decreasing renal injury.a A diagram describing the injection schedule for antagomir-493-3p; EG, ethylene glycol; i.p., intraperitoneal injection. b Renal function was quantified by serum creatinine and BUN levels, and kidney weight. $$n = 6$$ for each group. c Representative histologic kidney images of PAS, TUNEL, αSMA and Masson’s trichrome stain (MTS). d Immunostaining of IL-1β, TNFα, CD68, CD86, and MIF in kidney sections. $$n = 6$$ for each group. For (c) and (d), quantitations are at the right. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
These in vivo results from Fig. 6a–d confirmed that acetate diminished hyperoxaluria-induced kidney injury and fibrosis at least via upregulating the expression of miR-493-3p.
## Discussion
Hyperoxaluria can potential cause devastating consequences that can present as early as infancy or in the sixth decade of life and if not addressed appropriately, can cause significant morbidity and mortality including acute kidney injury and chronic kidney disease2,3,17. Macrophages-mediated inflammation plays an important role in regulating kidney injury and tissue fibrogenesis13. In the present study, we found acetate, which has anti-inflammatory properties, could attenuates hyperoxaluria-induced kidney injury and fibrosis via inhibiting macrophages infiltrating (Fig. 7). To our knowledge, this is the first study demonstrating the protective role of acetate in oxalate nephropathy. Other studies have observed in a reduction in kidney injury in other models after acetate treatment9,10,18.Fig. 7A scheme of acetate effect. The scheme diagram summarizes the pathway described: acetate enhances the H3K9 and H3K27 acetylation levels at miR-493-3p promoter region, which downregulate MIF expression, and decrease macrophages infiltration to attenuate the hyperoxaluria-induced renal injury. The figure was created by Figdraw.
Inflammation, apoptosis, and fibrosis are hallmarks of the hyperoxaluria rate model, and acetate treatment inhibited both processes. This is an expected result as acetate has previously been found to possess anti-inflammatory properties10. Infiltrating macrophages have long been known to be master players in inflammatory kidney diseases and to be associated with the development of kidney fibrosis and thereby failure of kidney function19. In our study, we found coculturing renal tubular cells with macrophages could increase the inflammatory cytokines levels and aggravate the oxalate-induced renal tubular cells injury. Increased CCL chemokines and IL-1β expression were detected in renal tubular cells coculturing with macrophages. IL-1β is strongly associated with the severity of tubulointerstitial lesions and renal impairment20. Some in vitro studies have demonstrated that IL-1β can exert potentially pro-fibrotic effects such as stimulation of proliferation and extracellular matrix production21.
The changes in inflammatory cytokines levels induced by coculturing renal tubular cells with macrophages without direct contact indicate that the crosstalk between macrophages and renal tubular cells is mediated by soluble factors released by these cells. In our study, we found MIF from renal tubular cells was elevated significantly after oxalate addition, and was suppressed by acetate treatment. MIF is an upstream proinflammatory cytokines and functions to initiate the inflammatory cascade response, and activates macrophages and T cells22. MIF was initially identified for its ability to inhibit the random migration of macrophages in vitro. However, recent evidence showed that MIF has pleiotropic effects on cell migration and chemotaxis23,24. Gregory et al. reported that MIF can induce macrophages recruitment through CCL2 and its receptor CCR225. Hoi et al. found that renal macrophages recruitment and glomerular injury were significantly reduced in MIF knockout mice model, suggesting that MIF as a critical effector of organ injury in systemic lupus erythematosus26.
Here, we first confirmed that MIF derived from renal tubular cells could play a key role in altering the renal macrophages and renal injury. We further proved that acetate could modulate MIF expression in renal epithelial cells, suggesting that altering the MIF expression via treating acetate to diminish the oxalate-induced renal injury is possible.
Acetate is an SCFA and has been reported to be readily absorbed in the intestines, transported into the blood stream, and easily incorporated in tissues8,27. Several studies reported that treatment of acetate could reduce kidney damage in different kidney injury animal models10. More evidence showed that acetate could through epigenetic regulation of histone acetylation in addition to its potential binding to G-protein membrane receptors (GPR41 and GPR43)15,28. Our previous study demonstrated that acetate could influence urinary compositions by regulating histone acetylation12. Constant with previous findings, in this study we found that acetate could influence macrophages infiltration to reduce oxalate-induced renal injury. It does so likely through regulating histone acetylation at H3K9 and H3K27 with a consequent activation of transcription of miR-493-3p, which in turn can suppress the expression of MIF, a key regulator of renal macrophages infiltration.
In summary, our study demonstrated that the treatment of acetate could function by altering the macrophages infiltration to influence oxalate-induced renal injury and fibrosis via altering the miR-493-3p/MIF signaling, which may provide clinicians a novel therapy to oxalate nephropathy.
## Animal studies
All rat experiments were performed under protocols approved by the Institutional Animal Care and Use Committee of the Guangzhou Medical University (Guangzhou, China).
## Development of hyperoxaluria rat model
Male Sprague-Dawley rats, aged 6-8 weeks, were purchased from Guangdong Laboratory Animal Center. Rats were housed in polypropylene cages, and had access to food and water ad libitum. We established the hyperoxaluria rat model following the reported protocol11,12,29. Rats were given free access to food and drinking water containing $1\%$ (v/v) EG for a period of 4 weeks. The rats were placed in metabolic cages for urine collection 1 day before sacrificing. Whole blood was collected and transferred to serum separator tubes and centrifuged to isolate the serum for further analysis. Serum BUN and creatinine were determined by Unicel DxC 600 synchronic biochemical detecting system. Urine oxalate was measured using ion exchange chromatography (Metrohm, Switzerland).
## Acetate treatment
8-week-old rats were divided into four groups. In the control group, animals were given tap water as their drinking water and 2 ml/kg ddH2O by gavage for 4 weeks. The EG group animals were exposed to $1\%$ EG in their drinking water and 2 ml/kg ddH2O by gavage for 4 weeks. In the acetate group, rats were orally administered $5\%$ (v/v) acetic acid dissolved with ddH2O daily for 4 weeks.
## Antagomir treatment
The miR-493-3p antagomir (5′-CUGGCACACAGUAGACCUUCA-3′) and negative control antagomir were synthesized by Guangzhou RiboBio (Ribobio, China). Each antagomir was dissolved by autoclaved PBS according to the manufacturer’s guidelines followed by ip injection to rats at dose of 30 nmol/kg body wt−1.
## Histology and morphometric analyses
Paraffin-embedded kidney pieces were cut into 5 μm sections and mounted on glass slides. The sections were deparaffinized with xylene, stained with MTS and periodic acid-Schiff (PAS). Tubular damage (epithelial necrosis) in PAS-stained sections was scored as follows: 0, normal; 1, <$10\%$; 2, 10–$25\%$; 3, 26–$75\%$; 4, >$75\%$. Tubular necrosis was defined as the loss of proximal tubular brush border blebbing of apical membranes, or intraluminal aggregation of cells and proteins30. At least 5 fields (magnification, ×200) were reviewed for each slide.
## Immunohistochemistry analysis (IHC)
Kidney tissue were fixed in $10\%$ formaldehyde in PBS, embedded in paraffin, and cut into 5 um sections and used for histology and IHC staining with specific primary antibodies against 8-ohdg, α-SMA, IL-1β, TNFα, CD68, CD86, MIF, H3K9ac and H3K29ac. To enhance antigen exposure, the slides were treated with 10 mM sodium citrate (pH = 6 0) at 98 °C for 15 min for antigen retrieval. The slides were incubated with endogenous peroxidase blocking solution and then were incubated with the primary antibody at 4 °C overnight. After rinsing with PBS, the slides were incubated for 45 min with biotin-conjugated secondary antibody, washed, and then incubated with enzyme conjugate horseradish peroxidase (HRP) streptavidin. Freshly prepared DAB (Zymed, South San Francisco, CA) was used as a substrate to detect HRP. Finally, slides were counterstained with hematoxylin and mounted with aqueous mounting media. The German immunoreactive score (IRS) (0 − 12) was calculated by multiplying the percentage of immunoreactive kidney epithelial cells ($0\%$ = 0; 1–$10\%$ = 1, 11–$50\%$ = 2, 51–$80\%$ = 3; and 81–$100\%$ = 4) by staining intensity (negative = 0; weak = 1; moderate = 2; and strong = 3)31. The CD68, CD86, H3K9ac or H3K27ac- positive cells were determined using light microscopy. The antibodies used in this study are listed in Supplementary Table 1.
## TUNEL assay
TUNEL was performed using the in situ Apoptosis Detection Kit (S7100-KIT; EMD milipore, CA, USA)11. Briefly, the paraffin-embedded sections were dewaxed. The sections were incubated in $0.3\%$ H2O2 at room temperature to eliminate the endogenous peroxidase activity. Proteinase K was applied to the sections for 15 min at room temperature. TdT enzyme was applied to the sections and incubated in a humidified chamber for 1 h at 37 °C to allow extension of the nicked ends of the DNA fragments with digoxigenin-dUTP. Color was developed using $0.05\%$ DAB with $0.006\%$ H2O2 as substrate. For negative controls, distilled water was used instead of TdT enzyme.
## RNA extraction and quantitative real-time PCR (Q-PCR) analysis
Total RNA was extracted by TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. RNAs (1 μg) were subjected to reverse transcription using Superscript III transcriptase (Invitrogen). Q-PCR was conducted using a Bio-Rad CFX96 system with SYBR Green to determine the mRNA expression level of a gene of interest. RNA expression levels were normalized to the expression of GAPDH. Primers used are in Supplementary Table 2.
For miRNA detection, 2 μg of total RNAs were subjected to reverse transcription using All-in-OneTM miRNA First-strand cDNA Synthesis Kit. Q-PCR was conducted using an All-in-OneTM miRNA qRT-PCR Detection Kits. Expression levels were normalized to the expression of 5 S rRNA or U6 snRNA.
## Cell lines and co-culture experiments
The human proximal tubular epithelial HK-2 cells, human monocyte THP-1 cells, human embryonic kidney cell line HEK-293T, mouse macrophage RAW264.7 cells, and mouse cortical collecting duct M-1 cells were purchased from the American Type Culture Collection (ATCC) (Rockville, MD). The HK-2, RAW264.7, and M-1 cells were maintained in Dulbecco’s modified Eagle’s media with $10\%$ fetal bovine serum (FBS) and $1\%$ penicillin/streptomycin. The THP-1 cells were cultured in RPMI-1640 media supplemented with $10\%$ FBS. The THP-1 cells were differentiated to macrophages by treating with 100 ng/ml PMA for 3 days before being used in experiments. 6-well Transwell plates (3 μm) were used for co-culture experiments (Corning Inc., Corning, NY).
## Synthesis and transfection of miRNA mimics and inhibitors
miRNA mimics and miRNA inhibitor were designed and synthesized by Guangzhou RiboBio (Ribobio, China). miRNA inhibitor was all nucleotides with 2′-O-methyl modification. 24 h prior to transection, cells were placed onto a 6-well plate at 40–$60\%$ confluence. Transfection was performed with riboFECTTM CP Reagent (Ribobio, China) according to the manufacturer’s protocol. The medium was replaced 24 h after transfection with new culture medium.
## Renal tubular cells exposure to oxalate and/or acetate and collection of the CM
Oxalate (Sigma) stock solution (10 mM) in PBS was diluted in medium to achieve final concentration of 0.5 mM. Sodium acetate (Sigma) stock solution (10 mM) in PBS was diluted in medium to achieve final concentration of 2 mM. HK-2 and M-1 cells were placed in 6-well culture dishes incubated overnight. The next day, the cell medium was changed to normal medium with 0.5 mM oxalate, and/or 2 mM sodium acetate. After 24 h of treatment, cells were collected for Q-PCR or western blot experiments, and the CM were collected for further experiments.
## Macrophages recruitment assay
Chambers with 5.0 μm polycarbonate filter inserted in 24-well plates were used in the quantitative cell migration assays (Corning Inc., Corning, NY). In all, 1 × 105 PMA-differentiated-THP-1 macrophages or mouse RAW264.7 macrophages were plated onto the upper chambers, and the lower chambers were filled with the CM from HK-2 or M-1 cells. After 18- to 20-h incubation, the non-migrated cells in the upper chamber were removed and cells that migrated into the membrane were fixed with methanol, stained with crystal violet, and photographed under an inverted microscope. Cell numbers were counted in five randomly chosen microscopic fields per membrane. All experiments were performed in triplicate wells for each condition.
## Western blot
Total protein was extracted by RIPA buffer containing $1\%$ protease inhibitors (Amresco, Cochran, CA). Proteins (30-50 μg) were separated on $10\%$ SDS/PAGE gel and then transferred onto PVDF membranes (Millipore). After blocking the membranes, they were incubated with appropriate dilutions (1:1000) of specific primary antibodies. The quantification was carried out by subtracting background from the band intensity of western blots by using Image J software.
## Human cytokine antibody array and ELISA
CM was collected from HK-2. Relative amounts of cytokine levels were determined using Human Cytokine Array kit (ARY005B, R&D systems) according to the manufacturer’s instructions. CM collected from culture cells were also used for detection of MIF by MIF ELISA kits (BOSTER) according to the manufacturer’s instructions.
## ChIP-qPCR assay
ChIP-qPCR assays were performed using a commercial kit (PierceTM Agrose ChIP Kit) according to the manufacturer’s instructions. Briefly, 1 × 107 HK-2 cells were cross-linked with $1\%$ paraformaldehyde, lysis and sonicated 13–15 times on ice until chromatin was 100–800 bps in size, with the center being ~300 bp. Solubilized chromatin was immunoprecipitated with ChIP grade antibodies for H3K9 acetyl, H3K27 acetyl or rabbit IgG (negative control). The DNA fragments were detected by qPCR. Histone acetylation marks were mapped at promoter spanning −2 to 2 kb of miR-493-3p. Primers spanning the regions with peaks were adopted for ChIP-qPCR analysis. Primers used are in Supplementary Table 2.
## Luciferase reporter assay
Wild-type (WT) human MIF 3′UTR and mutated MIF 3′UTR (with a mutated sequence on the miR-493-3p binding site) were amplified from a human cDNA library. The 3ʹ-UTR of MIF was constructed into psiCheck2 (Promega, Madison, WI, USA) by the Gibson assembly method. HEK293T cells were co-transfected with 25 ng/ml of either the luciferase reporter with WT or mutated 3′UTR, and 100 pmol of either miRNA mimics or miRNA negative control (NC). 48 h after co-transfection, a Dual-Luciferase Reporter Assay (Promega, USA) was carried out according to the manufacturer’s protocol.
## Statistics and reproducibility
All experiments were repeated independently, and statistical methods are described in the figure legends. p‐values were determined by unpaired Student’s t test using commercially available software (Prism 8) unless special methods were mentioned. $p \leq 0.05$ was considered statistically significant.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Peer Review File Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary The online version contains supplementary material available at 10.1038/s42003-023-04649-w.
## Peer review information
Communications Biology thanks Yongji Yan and the other, anonymous, reviewer for their contribution to the peer review of this work. Primary Handling Editor: Zhijuan Qiu. Peer reviewer reports are available.
## References
1. Demoulin N. **Pathophysiology and management of hyperoxaluria and oxalate nephropathy: a review**. *Am. J. Kidney Dis.* (2022.0) **79** 717-727. DOI: 10.1053/j.ajkd.2021.07.018
2. Leumann E, Hoppe B. **The primary hyperoxalurias**. *J. Am. Soc. Nephrol.* (2001.0) **12** 1986-1993. DOI: 10.1681/ASN.V1291986
3. Nazzal L, Puri S, Goldfarb DS. **Enteric hyperoxaluria: an important cause of end-stage kidney disease**. *Nephrol. Dial. Transplant.* (2016.0) **31** 375-382. DOI: 10.1093/ndt/gfv005
4. 4.Dutta, C. et al. Inhibition of Glycolate Oxidase With Dicer-substrate siRNA Reduces Calcium Oxalate Deposition in a Mouse Model of Primary Hyperoxaluria Type 1. Mol. Ther. (2016). Available at: http://www.nature.com/doifinder/10.1038/mt.2016.4, Accessed 17 February 2016.
5. Garrelfs SF. **Lumasiran, an RNAi therapeutic for primary hyperoxaluria type 1**. *N. Engl. J. Med.* (2021.0) **384** 1216-1226. DOI: 10.1056/NEJMoa2021712
6. Knauf F. **NALP3-mediated inflammation is a principal cause of progressive renal failure in oxalate nephropathy**. *Kidney Int.* (2013.0) **84** 895-901. DOI: 10.1038/ki.2013.207
7. Kasubuchi M. **Dietary gut microbial metabolites, short-chain fatty acids, and host metabolic regulation**. *Nutrients* (2015.0) **7** 2839-2849. DOI: 10.3390/nu7042839
8. Miyamoto J. **The role of short-chain fatty acid on blood pressure regulation**. *Curr. Opin. Nephrol. Hypertens.* (2016.0) **25** 379-383. DOI: 10.1097/MNH.0000000000000246
9. Huang W. **Short-chain fatty acids ameliorate diabetic nephropathy via GPR43-mediated inhibition of oxidative stress and NF-**. *Oxid. Med. Cell. Longev.* (2020.0) **2020** 1-21.. DOI: 10.1155/2020/8706898
10. Andrade-Oliveira V. **Gut bacteria products prevent AKI induced by ischemia-reperfusion**. *J. Am. Soc. Nephrol.* (2015.0) **26** 1877-1888. DOI: 10.1681/ASN.2014030288
11. 11.Zhu, W. et al. Alteration of the gut microbiota by vinegar is associated with amelioration of hyperoxaluria-induced kidney injury. Food Funct. 2020. Available at: https://pubs.rsc.org/en/content/articlelanding/2020/fo/c9fo02172h, Accessed 4 March 2020.
12. Zhu W. **Dietary vinegar prevents kidney stone recurrence via epigenetic regulations**. *EBioMedicine* (2019.0) **45** 231-250. DOI: 10.1016/j.ebiom.2019.06.004
13. Anders H-J. **The macrophage phenotype and inflammasome component NLRP3 contributes to nephrocalcinosis-related chronic kidney disease independent from IL-1–mediated tissue injury**. *Kidney Int.* (2018.0) **93** 656-669. DOI: 10.1016/j.kint.2017.09.022
14. Liu N. **MiR-451 inhibits cell growth and invasion by targeting MIF and is associated with survival in nasopharyngeal carcinoma**. *Mol. Cancer* (2013.0) **12** 123. DOI: 10.1186/1476-4598-12-123
15. Gao X. **Acetate functions as an epigenetic metabolite to promote lipid synthesis under hypoxia**. *Nat. Commun.* (2016.0) **7** 11960. DOI: 10.1038/ncomms11960
16. Zhou S. **In vivo therapeutic success of MicroRNA-155 antagomir in a mouse model of lupus alveolar hemorrhage: in vivo therapeutic success of miR-155 antagomir in pristane-induced DAH**. *Arthritis Rheumatol.* (2016.0) **68** 953-964. DOI: 10.1002/art.39485
17. Green ML, Hatch M, Freel RW. **Ethylene glycol induces hyperoxaluria without metabolic acidosis in rats**. *Am. J. Physiol.—Ren. Physiol.* (2005.0) **289** F536-F543. DOI: 10.1152/ajprenal.00025.2005
18. Al-Harbi NO. **Short chain fatty acid, acetate ameliorates sepsis-induced acute kidney injury by inhibition of NADPH oxidase signaling in T cells**. *Int. Immunopharmacol.* (2018.0) **58** 24-31. DOI: 10.1016/j.intimp.2018.02.023
19. 19.Wang, X. et al. The role of macrophages in kidney fibrosis. Front. Physiol.12 (2021). Available at: https://www.frontiersin.org/articles/10.3389/fphys.2021.705838, Accessed 18 July 2022.
20. Nikolic-Paterson DJ. **Interleukin-1 in renal fibrosis**. *Kidney Int. Suppl.* (1996.0) **54** S88-S90. PMID: 8731202
21. Vesey DA. **Interleukin-1β induces human proximal tubule cell injury, α-smooth muscle actin expression and fibronectin production1**. *Kidney Int.* (2002.0) **62** 31-40. DOI: 10.1046/j.1523-1755.2002.00401.x
22. Lan HY. **Role of macrophage migration inhibition factor in kidney disease**. *Nephron Exp. Nephrol.* (2008.0) **109** e79-e83. DOI: 10.1159/000145463
23. Bernhagen J. **MIF is a noncognate ligand of CXC chemokine receptors in inflammatory and atherogenic cell recruitment**. *Nat. Med.* (2007.0) **13** 587-596. DOI: 10.1038/nm1567
24. Hermanowski-Vosatka A. **Enzymatically inactive macrophage migration inhibitory factor inhibits monocyte chemotaxis and random migration**. *Biochemistry* (1999.0) **38** 12841-12849. DOI: 10.1021/bi991352p
25. Gregory JL. **Macrophage Migration Inhibitory Factor Induces Macrophage Recruitment via CC Chemokine Ligand 2**. *J. Immunol.* (2006.0) **177** 8072-8079. DOI: 10.4049/jimmunol.177.11.8072
26. Hoi AY. **Macrophage migration inhibitory factor deficiency attenuates macrophage recruitment, glomerulonephritis, and lethality in MRL/lpr**. *mice. J. Immunol. Baltim. Md 1950* (2006.0) **177** 5687-5696
27. Watson AJ. **Acetate absorption in the normal and secreting rat jejunum**. *Gut* (1990.0) **31** 170-174. DOI: 10.1136/gut.31.2.170
28. Le Poul E. **Functional characterization of human receptors for short chain fatty acids and their role in polymorphonuclear cell activation**. *J. Biol. Chem.* (2003.0) **278** 25481-25489. DOI: 10.1074/jbc.M301403200
29. Zhu W. **Loss of the androgen receptor suppresses intrarenal calcium oxalate crystals deposition via altering macrophage recruitment/M2 polarization with change of the miR-185-5p/CSF-1 signals**. *Cell Death Dis.* (2019.0) **10** 275. DOI: 10.1038/s41419-019-1358-y
30. Choi DE. **Pretreatment of sildenafil attenuates ischemia-reperfusion renal injury in rats**. *Am. J. Physiol. Ren. Physiol.* (2009.0) **297** F362-F370. DOI: 10.1152/ajprenal.90609.2008
31. 31.Remmele, W. & Stegner, H. E. Recommendation for uniform definition of an immunoreactive score (IRS) for immunohistochemical estrogen receptor detection (ER-ICA) in breast cancer tissue. Pathologe8, 138–140 (1987).
|
---
title: A rare human variant that disrupts GPR10 signalling causes weight gain in mice
authors:
- Fleur Talbot
- Claire H. Feetham
- Jacek Mokrosiński
- Katherine Lawler
- Julia M. Keogh
- Elana Henning
- Edson Mendes de Oliveira
- Vikram Ayinampudi
- Sadia Saeed
- Amélie Bonnefond
- Mohammed Arslan
- Giles S. H. Yeo
- Philippe Froguel
- David A. Bechtold
- Antony Adamson
- Neil Humphreys
- Inês Barroso
- Simon M. Luckman
- I. Sadaf Farooqi
journal: Nature Communications
year: 2023
pmcid: PMC10017677
doi: 10.1038/s41467-023-36966-3
license: CC BY 4.0
---
# A rare human variant that disrupts GPR10 signalling causes weight gain in mice
## Abstract
Disruption of brain-expressed G protein-coupled receptor-10 (GPR10) causes obesity in animals. Here, we identify multiple rare variants in GPR10 in people with severe obesity and in normal weight controls. These variants impair ligand binding and G protein-dependent signalling in cells. Transgenic mice harbouring a loss of function GPR10 variant found in an individual with obesity, gain excessive weight due to decreased energy expenditure rather than increased food intake. This evidence supports a role for GPR10 in human energy homeostasis. Therapeutic targeting of GPR10 may represent an effective weight-loss strategy.
The brain-expressed receptor GPR10 is involved in energy homeostasis in mice. Here the authors identify rare loss of function variants in GPR10 in people with severe obesity and showed that one of these variants causes obesity when modelled in mice, suggesting that future studies could explore GPR10 as a potential target for weight-loss therapy.
## Introduction
There is a substantial unmet need for new weight-loss treatments to reduce the morbidity and mortality associated with obesity1. *Human* genetic studies have established that some of the molecules known to regulate energy homeostasis in rodents contribute to human physiology, validating these molecules and pathways as therapeutic targets2.
G protein-coupled receptor 10 (GPR10) is a centrally expressed G protein-coupled receptor (GPCR) which acts as the cognate receptor for prolactin-releasing peptide (PrRP), an evolutionarily conserved RF-amide peptide3–5. PrRP reduces food intake and increases energy expenditure when administered centrally in rodents. Moreover, PrRP-containing neurons in the dorsomedial nucleus of the hypothalamus (DMH) have been shown to play a key role in mediating the thermogenic effects of the adipocyte-derived hormone leptin6,7. There has been interest in targeting GPR10 for weight-loss therapy. Indeed, systemically administered palmitoylated PrRP analogues reduce body weight in mouse models of diet-induced obesity, potentially through central mechanisms8,9.
In this study, we identified multiple rare variants in PRLHR (Prolactin Releasing Hormone Receptor), also known as GPR10, in cases with severe obesity and ancestry-matched controls, which we showed in cellular studies, impaired ligand binding and G protein-dependent signalling. To test the physiological consequence of human variants on body weight, we generated mice harbouring a functional GPR10 variant found in an individual with severe obesity. We found that transgenic mice gained excessive weight due to decreased energy expenditure rather than increased food intake. Cumulatively, these studies indicate that therapeutic targeting of GPR10 may be an effective weight-loss strategy.
## Characterisation of rare variants in GPR10
To investigate the potential contribution of genetic variation in GPR10 to severe human obesity, we interrogated exome sequencing and targeted resequencing data on 2548 European ancestry individuals with severe, early-onset obesity recruited to the Genetics of Obesity study (GOOS; www.goos.org.uk) (mean Body Mass Index (BMI) standard deviation score>3; age of onset <10 years); mutations in known obesity genes had been excluded. We also studied 1117 ancestry-matched controls analysed using the same methods10. We identified 15 rare heterozygous variants in GPR10 in 17 unrelated individuals with obesity (A266D was found in three unrelated individuals); we also found five rare variants in controls (Odds Ratio [$95\%$ confidence intervals] =1.5 [0.5–5.2], $$p \leq 0.5$$, Fisher’s exact test) (Fig. 1a, b; Supplementary Table 1). As a limited number of family members were available for study, the mode of inheritance is unclear (six probands inherited the GPR10 variant from a parent with overweight/obesity; one proband inherited the variant from a parent who was normal weight. Variant carriers had a number of other clinical features including anxiety, impaired memory and impaired pain sensation (Table 1).Fig. 1Functional characterisation of rare variants in GPR10 identified in cases with severe obesity and controls.a Rare variants identified in individuals with severe early onset obesity (magenta) and in controls (blue) shown on a schematic of the GPR10 protein and b on a structural model. For residues located in α-helices, a generic residue number according to a structure-based GPCRdb numbering scheme shown in subscript18. Wild-type (WT) and mutant forms of GPR10 were studied in cells alongside a low frequency variant (P305L); mock transfected cells served as a negative control. Effects on ligand (PrRP-31)-induced binding c and d receptor-mediated activation of Gαq/11-regulated inositol triphosphate signalling in transiently transfected COS-7 cells. Data shown as sum curves from 3-9 experiments normalized to WT maximal binding/response±SEM. e COS-7 cells were transiently co-transfected with WT and varying amounts of mutant GPR10 to investigate dominant negative effects on PrRP-31-induced Gαq/11-regulated inositol triphosphate signalling. Mean±SEM of 4 experiments are shown; differences between means for 2.5 ng WT alone and with additional 2.5 ng mutant GPR10 were compared with Students t-test, ns – not significant; **$p \leq 0.01$; ***$p \leq 0.001$ (based on data shown in Supplementary Fig. 1, panels d-i). Source data are provided as a Source Data file. Table 1Phenotypes seen in carriers of rare variants in GPR10VariantAge (years)Weight (kg)Height (cm) [SDS]BMI (kg/m2) [SDS]Temp (o C)Heart rate (bpm)BP (mmHg)Glucose (mmol/l)Insulin (pmol/l)Neurobehavioural conditionsA121D*13.7107.0167.5 [0.9]38.1 [3.6]N/AN/AN/AN/AN/AV162L*14.978.9153.4 [−1.8]33.5 [3.1]N/AN/AN/A4.6100P193S*18.4109.0169.9 [1.0]37.7 [3.3]$\frac{36.454111}{564.2116}$Night terrors/sleep walks, anxiety, depression, self-harm, impaired memoryP193S$\frac{52.988.7180.727.236.271127}{744.623}$Anxiety, depression, impaired memory, impaired pain sensationH206N*4.755.2121.0 [3.0]37.7 [6.5]N/AN/A$\frac{85}{554.8}$N/AR220L*15.7129.2155.6 [−1.2]53.4 [4.4]$\frac{36.485103}{5514.2}$#69Aggression, low moodR220L$\frac{54.291.8173.730.436.066107}{618.4}$#50Anxiety, low mood, emotional lability, impaired memoryR220L$\frac{22.580.0160.231.236.17599}{554.7}$R222S*$\frac{24.791.4166.732.936.845124}{674.7}$#38Anxiety, depression, autism-likeR222S*16.2119.6161.9 [−0.2]45.6 [4.0]N/AN/AN/A4.1293L242F*10.178.5144.1 [0.8]37.8 [3.9]N/AN/AN/A4.1N/AA266D*$\frac{26.0116.4161.144.936.958113}{644.599}$Impaired memoryA266D8.628.0124.5 [−1.1]18.1 [0.9]$\frac{36.773100}{594.245}$Hyperactive, impaired memory, impaired pain sensationA266D*8.768.0132.3 [0.2]38.9 [4.3]$\frac{37.081134}{784.3103}$Aggression, fearless, emotional labilityA266D9.864.4144.6 [1.2]30.8 [3.3]$\frac{36.679125}{694.3120}$A266D*11.4102.3145.3 [−0.2]48.5 [4.3]N/AN/AN/A4.7225D340N*12.495.7161.2 [1.3]36.8 [3.5]N/AN/AN/A5.1193L350S*17.7105.8170.3 [1.1]36.5 [3.2]$\frac{36.872119}{574133}$Impaired memoryL350S$\frac{48.471.0170.724.436.551130}{683.67}$Data for individuals with severe obesity with GPR10 variants where this was available. BMI Body Mass Index; age and gender-adjusted height and BMI Standard Deviation Scores (SDS) included for children up to age 18 years; fasting plasma insulin (0–60 pmol/l); * denotes proband, #denotes type 2 diabetes and/or treated with metformin, oC temperature (temp) in degrees Celsius, heart rate in beats per minute (bpm) and blood pressure (BP) in millimetres of mercury (mmHg) were measured in the rested fasted state, N/A indicates data not available.
To test whether the rare variants in GPR10 might have functional consequences, we studied all 20 variants identified in cases with obesity and/or controls in cells transiently transfected with constructs encoding wild-type (WT) or mutant human GPR10 (Fig. 1a–e; Supplementary Table 2). We also studied a previously reported low frequency variant in GPR10, P305L (Minor Allele Frequency, MAF = $5\%$). Whilst this variant has not been associated with BMI, it has been associated with reduced systolic and diastolic blood pressure in a previous population based study11. None of the GPR10 mutants reduced cell surface expression measured by enzyme-linked immunosorbent assay (ELISA) (Supplementary Fig. 1a). In competitive radioligand binding assays, [125I]-labelled PrRP-31 did not bind to 6 of 20 GPR10 mutants. For some of these mutants, cell surface expression was not affected. As a number lie in the extracellular domain of the receptor, these amino acid substitutions may impair the conformation of the ligand binding pocket. For a further 7 of 20 mutants, the maximal amount of radioligand bound was significantly reduced (Fig. 1c), but none of the mutants affected binding affinity (IC50; Supplementary Table 2). This discrepancy may be partially explained by experimental methods (amplified signal detection in cell surface ELISA vs ratiometric measurements of receptor density in competition radioligand binding assays). Saturation binding studies may provide further insights into the mechanisms underlying these observations.
We next measured canonical Gαq/11-coupled signalling by quantifying ligand-induced inositol triphosphate turnover. Ten of 20 GPR10 mutants (7 of 15 in cases; 3 of 5 in controls) caused a loss of function (LoF) in this assay (Fig. 1d). In HEK293 cells stably expressing WT GPR10, we showed that GPR10 activation inhibits basal and forskolin-induced cyclic adenosine monophosphate (cAMP) accumulation (Supplementary Fig. 1b) consistent with Gαi/o–coupled signalling. We found that twelve of twenty GPR10 mutants (8 of 15 in cases, 4 of 5 in controls) impaired Gαi/o–coupled signalling measured using an inositol triphosphate turnover assay in the presence of GαΔ6qi4myr chimeric G-protein12 (Supplementary Fig. 1c). In these assays, the proportion of variants that exhibit LoF did not differ significantly between cases and controls ($p \leq 0.05$).
LoF missense GPR10 variants found in heterozygous form may impact the phenotype through haplo-insufficiency or by having a dominant negative effect, where the mutated receptor inhibits signalling by the WT receptor. In preliminary studies, we studied seven mutants that significantly impaired Gαq/11-coupled signalling, in COS-7 cells co-transfected with varying doses of WT and mutant GPR10. Two of the four complete LoF variants found in cases (A121D, P237R), but none of the three LoF variants found in controls (Fig. 1e and Supplementary Fig. 1d–i) exhibited dose-dependent inhibition of signalling by WT GPR10. Further studies will be needed to examine the potential structural and molecular mechanisms which may underlie these effects. Cumulatively, we found that the majority of GPR10 variants found in cases with severe obesity (11 of 15) and in controls (4 of 5), caused LoF in one or more cellular assay.
In a further study, we interrogated 581 exomes from unrelated probands with severe obesity from consanguineous families of Pakistani origin. In this cohort we identified a homozygous variant in the gene encoding GPR10 (NM_004248.3: c.665_670dup / p.R222_Q223dup) in a 5-year old girl with hyperphagia and weight gain from the age of 2.5 years, resulting in severe obesity (BMI SDS: 5.5). On a follow-up examination at the age of 8.5 years, the proband was reported to suffer from extended periods of anxiety and depression lasting for as many as 10 days (there was no history of developmental delay). Both parents were heterozygous for the variant which is found at MAF of 0.003512 in South Asian exomes (gnomAD v2.1.1). We characterised the function of this mutation in cells; the variant resulted in a partial loss of function (Fig. 1d).
## Gpr10-/- knockout mice show increased weight gain on chow diet
Given their rarity, it is difficult to establish whether LoF human variants affecting GPR10 can cause obesity. We therefore set out to test directly whether a rare human variant in GPR10 found in an individual with obesity could cause weight gain in mice. We first generated Gpr10-/- null mice with a 54-base pair targeted deletion within the coding region and found that on a chow diet, both male and female mice exhibited increased weight gain compared with littermate controls (Fig. 2a, b). Male mice were significantly obese at 14-weeks age; female mice at 25-weeks of age. Importantly, whereas no difference in food intake was seen (Fig. 2c), pre-obese male Gpr10-/- null mice exhibited reduced energy expenditure, measured as a difference in oxygen consumption (VO2) by indirect calorimetry (Fig. 2d, e); this effect persisted in older Gpr10-/- null mice (Supplementary Fig. 2) compared with wild-type controls. These findings are in keeping with previous reports of obesity in Gpr10-/- mice and the OLETF strain of rats, which carry a mutation in Gpr1013–15.Fig. 2Obesity in Gpr10-/- knockout mice.a, b Body weight curves for male and female homozygous wild-type Gpr10+/+ (grey squares) and Gpr10-/- null mice (red circles) (males, $$n = 21$$-25 per group; females, $$n = 6$$ per group). Data are presented as mean±SEM and analysed with a mixed-effects analysis followed by Sidak’s multiple-comparison post hoc test. c Daily food intake (g) in 6 week old homozygous wild-type Gpr10+/+ (grey squares; $$n = 5$$) and Gpr10-/- null mice (red circles; $$n = 6$$) fed on normal chow. Data are presented as mean±SEM and comparisons made using two-tailed Student’s t-test. d, e Oxygen consumption (VO2, ml/h) for male wild-type Gpr10+/+ (grey squares) and Gpr10-/- null mice (red circles) at 6 weeks across the light-dark cycle ($$n = 5$$ in each group). Data are presented as mean±SEM and comparisons made using two-way ANOVA followed by Bonferroni’s multiple-comparison post hoc tests. Source data are provided as a *Source data* file.
## Knock-in Gpr10P193S mice develop obesity
We then generated a knock-in mouse model of the human GPR10 variant (Gpr10P193S) (Fig. 3 and Supplementary Fig. 3). This mutation affects a residue that is highly conserved across species and among Family A GPCRs16, is located at the extracellular end of transmembrane domain 4 and is predicted to affect the ligand binding pocket. As detailed above, GPR10P193S shows impaired PrRP-31 binding (Fig. 1c) and complete loss of receptor activation in cells (Fig. 1d and Supplementary Fig. 1c).Fig. 3Obesity in Gpr10P193S/P193S mutant mice.a, b Body weight curves of wild-type (+/+; grey squares), heterozygous (+/P193S; purple triangles) and homozygous Gpr10P193S/P193S (pink circles) male ($$n = 9$$, 18 and 7 respectively) and female mice ($$n = 8$$, 8 and 5 respectively) maintained on $60\%$ high energy diet from 8 weeks. Data are presented as mean± SEM and were analysed with a mixed-effects analysis followed by Sidak’s multiple-comparison post hoc test. c Daily food intake (g) in wild-type (+/+; grey squares), heterozygous (+/P193S; purple triangles) and homozygous Gpr10P193S/P193S (pink circles) male mice aged 8 weeks on normal chow ($$n = 12$$, 13 and 13 respectively). d, e Oxygen consumption (VO2, ml/hr) across the light-dark cycle in male wild-type (+/+; grey squares), heterozygous (+/P193S; purple triangles) and homozygous Gpr10P193S/P193S (pink circles) mice on standard chow at 8 weeks ($$n = 9$$, 9 and 11 respectively). Data are presented as mean ± SEM and comparisons made using two-way ANOVA followed by Bonferroni’s multiple-comparison post hoc tests. Source data are provided as a *Source data* file.
We found that Gpr10P193S homozygous mutant mice were of a similar weight to wild-type littermates at the age of 8 weeks (Fig. 3a, b), before they were placed on a high-energy diet. Both male and female mutant mice exhibited greater weight gain than wild-type littermates (Fig. 3a, b), with a distinct gene dose effect between homozygous (Gpr10P193S/P193S), heterozygous (Gpr10+/P193S) and wild-type (Gpr10+/+) littermates. At 16 weeks, the difference in weight between Gpr10P193S/P193S and wild-type littermates was comparable to that between Gpr10-/- and Gpr10+/+ mice (~5 g). As with Gpr10-/- mice, pre-obese male Gpr10P193S mice did not exhibit a difference in daily food intake (Fig. 3c), but did exhibit lower energy expenditure when compared with weight-matched littermates (Fig. 3d, e). Cumulatively, these findings support the importance of GPR10 in the regulation of energy expenditure and weight regulation.
## Common variants at the GPR10 locus
It is challenging to establish whether GPR10 variants that do not display Mendelian inheritance are associated with severe obesity, as studies in tens of thousands of cases with severe obesity and controls would be needed17 and cohorts of this scale do not exist. To investigate whether common variants in the region of GPR10 (PRLHR) are associated with anthropometric or metabolic traits, we searched for GWAS and PheWAS associations and putative prioritisation of GPR10 as a causal gene using the Open Targets Genetics platform (Supplementary Tables 3–6). We found that GPR10 is prioritised at loci associated with percentage body fat (lead variant rs4752183, $$p \leq 2$$ × 10−8; GPR10 locus-to-gene Score=0.7) and other anthropometric traits (Supplementary Table 3). The low frequency coding variant (P305L, rs8192524-A; allele frequency $5\%$) is nominally associated with higher percentage body fat and other anthropometric traits in single-variant analyses of UK Biobank genotypes (Supplementary Table 4) and exomes (Supplementary Table 5). Gene-based analyses combining low frequency and rare missense variants in GPR10 tend towards an association with higher mean body fat percentage ($$p \leq 0.02$$, beta=0.001, Burden test; $$p \leq 0.007$$, SKAT-O; SAIGE-GENE mixed model, UK Biobank 280,000 Non-Finnish European exomes; https://genebass.org) and other anthropometric traits (Supplementary Fig. 4a, b; Supplementary Table 6).
An important related question is whether rare variants in GPR10 observed in the population contribute to severe obesity (BMI > 40 kg/m2). To address this question, we interrogated rare variants (MAF < $0.1\%$) in GPR10 in the subset of approximately 150,000 unrelated European ancestry exomes from UK Biobank (Methods). We observed a nominal association between rare coding variants in GPR10 (MAF < $0.1\%$, 139 variants) and severe obesity (BMI > 40 kg/m2, $$n = 2725$$/152837 people) ($$p \leq 0.03$$, Robust SKAT-O; $$p \leq 0.04$$, Robust Burden test) with odds ratio >1 ($$n = 18$$/629 carriers, $\frac{2707}{152208}$ non-carriers, OR ($95\%$ CI) = 1.63 (0.96,−2.6)). We did not observe an association with continuous BMI ($$p \leq 0.6$$, SKAT-O) nor with BMI dichotomised at the upper quartile (BMI > 29.8 kg/m2; $$p \leq 0.8$$, Robust SKAT-O). Five rare variants had nominal single-variant association with severe obesity (BMI > 40 kg/m2; $p \leq 0.05$,) all of which had odds ratio >1 and were located in transmembrane domain 4/extracellular loop 2a (TM4/EL2a) (V196G, R209S, A177P) or C-terminal domain (D340N, I357R; both are residues where variants were identified in our study of people with severe early-onset obesity) (Supplementary Fig. 4). Region-based case-control tests for severe obesity (BMI > 40 kg/m2) suggest an association with rare coding variants (MAF < $0.1\%$) in EL2a/TM4 (OR ($95\%$ CI) = 5.7 (2.0–13), $$p \leq 0.001$$, Fisher’s exact; $$n = 6$$/64 carriers, $\frac{2719}{152773}$ non-carriers; $$p \leq 0.001$$, Robust SKAT-O) (Supplementary Table 7 and Supplementary Fig. 4d). The variant P193S, identified in the GOOS cohort and shown to cause obesity when studied in mice, lies at the EL2a boundary with transmembrane domain 4 (Supplementary Fig. 4e). These findings are intriguing and suggest that a subset of variants in the gene encoding GPR10 contribute to human weight regulation.
## Discussion
In conclusion, our data in humans and in mice, suggest that targeting GPR10 represents a potential therapeutic strategy for obesity. Further studies are needed to explore the potential consequences of targeting GPR10. An outstanding question relates to the interaction between GPR10 and the physiological response to stress. PrRP neurons in the medulla oblongata and/or in the dorsomedial hypothalamus are activated by a variety of stressful stimuli5. Central administration of PrRP activates corticotropin-releasing hormone neurons and oxytocin neurons in the hypothalamus and facilitates adrenocorticotropic hormone and oxytocin release into the systemic circulation18,19. Experiments in rodents suggest that PrRP has inhibitory effects on the neuroendocrine response to stress20. Here we show that human phenotypes associated with loss of function GPR10 mutations include anxiety, depression, impaired memory and impaired pain sensation (Table 1). While gaps in understanding remain, these observations suggest that GPR10 agonism may have beneficial effects on anxiety, depression and memory and may alter the perception of pain. As seen with other targets for weight loss therapy e.g. the Serotonin 2c receptor and the Cannabinoid 1 receptor21,22, the overlap between neural pathways that regulate weight and those that modulate other behaviours and cognitive phenotypes presents potential challenges for drug development. The transgenic mouse model of a human LoF GPR10 mutation generated here will enable further exploration of the role of GPR10 in coupling the stress response to changes in energy homeostasis.
## Study design and approval
*All* genetic and clinical studies were approved by the Multi-Regional Ethics Committee and the Cambridge Local Research Ethics Committee (MREC $\frac{97}{21}$ and REC number $\frac{03}{103}$) and conducted in accordance with the principles of the Declaration of Helsinki. All participants, or their legal guardian for those aged under 16, provided written consent for all assessments; participants under the age of 16 provided oral assent. GPR10 variant carriers were identified as part of genetic studies of individuals recruited to the Genetics of Obesity Study (GOOS), a cohort of 8000 individuals with severe early-onset obesity; age of obesity onset less than 10 years19. Severe obesity in GOOS was defined as a Body Mass Index (BMI, weight in kilograms divided by the square of the height in meters) standard deviation score ≥3 (standard deviation scores calculated according to the United Kingdom reference population). The details of the genetic analysis have been published previously10. Participants did not receive compensation for taking part in this study.
## Exomes from consanguineous families
We searched for rare GPR10 homozygous variants in 581 exomes from unrelated probands (0.2–22 years of age) with severe, early onset obesity from consanguineous families of Pakistani origin. The details regarding the cohort and the genetic investigation have been published before20.
## UK Biobank 200 K exomes
This research has been conducted using the UK Biobank Resource under Application Number 53821. UKB analysis was based on interim UKB exomes releases, and future work will analyse a larger UKB exome release. We used the UK Biobank pVCF variant file (chromosome 10, block 34) from the OQFE exome pipeline (UK Biobank Field 23156; $$n = 200$$,629 exomes available to us (https://ukbiobank.dnanexus.com/)). Relatedness was obtained from the UK Biobank Genetic Data resource (ukbgene rel) and one person was excluded from each related pair among all the OQFE exomes (kinship ≥ 0.0442, KING, third-degree kinship or closer; subsetted the pairwise kinship results from ‘ukbgene rel’ to retain the pairs contained within OQFE exomes and excluded the individuals in column ‘ID2’; excluded $$n = 15$$,547 people). The resulting unrelated OQFE exomes were taken forward for analysis ($$n = 185$$,082). Exomes were further restricted to *European* genetic grouping (Field 22006) ($$n = 153$$,352).
We used Body mass index (BMI, kg/m2) obtained from the UK Biobank initial assessment visit (UK Biobank Field 21001, Instance 0) and this value was available to us for $$n = 184$$,$\frac{294}{185}$,082 unrelated exomes and 152,$\frac{837}{153}$,352 unrelated European exomes. The reported BMIs for selected individuals were checked for plausibility (ie. not obviously result of inaccurate derivation or recording of BMI) by inspecting other relevant measurements: height (Field 50), weight (Field 21002) and waist circumference (Field 48) at initial assessment, and related longitudinal measurements from repeat assessments where available.
Variant annotation was performed using Ensembl Variant Effect Predictor (VEP) (Ensembl release 102) and consequences defined with respect to Ensembl canonical transcript ENST00000239032. Variant filtering by minor allele frequency (MAF) was based on the maximum reported MAF among unrelated OQFE exomes, each gnomAD exome subpopulation (VEP field “gnomAD_exomes_POPMAX_AF”) and each 1000 Genomes subpopulation (VEP). We performed gene-based association tests for rare (MAF < $0.1\%$) non-synonymous exonic or splicing GPR10 variants and severe obesity (BMI > 40 kg/m2, case:control ratio=1:53), for BMI dichotomised at the upper quartile of the unrelated exomes (BMI > 29.8 kg/m2, case: control ratio=1:3) and for continuous BMI.
Gene-based burden and SKAT-O tests were performed using the SKATBinary_Robust function for dichotomised BMI (or the SKAT function for continuous BMI) in R package SKAT v2.0.1 (method = ”Burden” or “SKATO” with default settings)21. The null models were calculated using SKAT_Null_Model(y~X) where the covariates matrix (X) contained age (Field 21003.0.0), sex (Field 31.0.0), ten genetic principal components (Fields 22009.0.1-10) and sequencing batch (UKB 50 K or 150 K exomes). Single-variant p-values are reported using the gene-based SKATBinary_Robust output values for p.value_singlevariant. Odds ratios were calculated for the number of variant carriers at the relevant rare variant or region using Fisher’s Exact test. Region-based tests for severe obesity (BMI > 40 kg/m2) as a binary trait were performed as for gene-based tests after restricting to the regions defined in Supplementary Table 7; burden and SKAT-O p-values should be interpreted with caution due to the small number of expected cases in the specified regions.
## UKBB 300 K exomes
This research been conducted using the https://genebass.org resource of summary statistics generated from the UK Biobank resource (under application 26041 and 48511). Briefly, genebass.org provides summary statistics for ~280,000 non-Finnish European (NFE) exomes using the SAIGE-GENE mixed model framework, including single-variant tests and gene-based burden (mean) and SKAT-O tests. We obtained summary statistics from https://genebass.org (accessed 6th Aug 2021) for BMI (UKB Field 21001), for anthropometry traits measured at UKB Assessment Centres, and for selected Early Life traits (comparison of height or body size at age 10).
## Open Targets Genetics GWAS/PheWAS loci and single-variant summary statistics
We used the Open Targets Genetics platform22 to obtain genome-wide significant GWAS/PheWAS loci for which a “locus-to-gene” (L2G) score was reported for GPR10 (https://genetics.opentargets.org/gene/ENSG00000119973, accessed 19th Aug 2021) and inspected whether GPR10 was prioritised as a putatively causal gene at loci associated with anthropometric traits or triglyceride levels (https://genetics.opentargets.org/study-locus/[studyIdentifier]/[variantIdentifier]).
We obtained single-variant summary statistics for GPR10 coding variant P305L (variant identifier 10-118594331-G-A, hg38) for anthropometric traits nominally associated with this variant ($P \leq 0.05$) in the UK Biobank (NEALE2/Neale v2, http://www.nealelab.is/uk-biobank/) or NHGRI-EBI GWAS Catalog23 (accessed 19th Aug 2021).
## cDNA constructs and site-directed mutagenesis
GPR10 WT cDNA was cloned into pCMV-Tag2B with an N-terminal FLAG tag and variant constructs were generated using QuickChange II XL site-directed mutagenesis kit (Agilent) according to the manufacturer’s recommendations. N-terminally cMyc-tagged GPR10 WT construct was generated by substitution of the FLAG tag in the above described pCMV-Tag2B construct using Q5® site-directed mutagenesis kit (New England Biolabs, Inc.) according to the supplier’s protocol. All constructs were verified by Sanger sequencing.
## Cell culture and transfection
COS7 cells were kindly provided by Professor Alan Tunnacliffe (Department of Chemical Engineering and Biotechnology, University of Cambridge, UK) and HEK293 cells were kindly provided by Professor Dario Alessi (MRC Protein Phosphorylation and Ubiquitylation Unit, University of Dundee, UK). HEK293 cells were authenticated via GENETICA Genotypes Analysis in May 2019, showing $97\%$ match when compared to the reference profile ATCC sequence. COS-7 cells were not authenticated. Prior to the experiments, COS-7 and HEK293 cells were tested negative for mycoplasma contamination using MycoAlert enzymatic assay (LT07-703, Lonza) and MycoProbe Mycoplasma Detection kit (CUL001B, R&D Systems), respectively. COS-7 and HEK293 cells were maintained in low and high glucose Dulbecco’s modified eagle medium (Sigma-Aldrich, D6046; Gibco, 31966), respectively, supplemented with $10\%$ fetal bovine serum (Gibco, 10270, South America origin), $1\%$ GlutaMAXTM (100X) (Gibco, 35050), and 100 units/mL penicillin and 100 μg/mL streptomycin (Sigma-Aldrich, P0781) and cultured at 37 °C in humidified air containing $5\%$ CO2. Cells were transfected with respective cDNA constructs using Lipofectamine 2000™ (Gibco, 11668) in serum-free Opti-MEM medium (Gibco, 31985) according to the manufacturer’s protocols. For cell surface ELISA, competitive radioligand binding and inositol triphosphate turnover assays, COS-7 cells were seeded in poly-D-lysine (Sigma-Aldrich, P7886) coated 96-well plates at 20,000 cells/well density. Cells were transiently transfected using Lipofectamine 2000 (Invitrogen, 11668) according to manufacturer’s recommendations. In experiments using varying amounts of cDNA to assess the dominant negative effect of GPR10 mutants, the total amount of cDNA used in each transfection was maintained at 10 ng/well by topping up with empty vector.
## Generation of GPR10 stably expressing cell line
HEK293 cells stably expressing WT GPR10 were generated by calcium precipitation transfection with the cDNA construct described above that contains Neomycin resistance gene allowing selection in eukaryotic cells. Briefly, 20 µg GPR10 WT cDNA diluted in Tris-EDTA (TE) buffer supplemented with 250 uM CaCl2 was precipitated into an equal volume of 2x concentrated HEPES-buffered saline (HBS buffer) and added after 45 min dropwise onto cells scarcely seeded (i.e. approximately 30–$50\%$ confluence) in 10 cm diameter Petri dishes. After 24 h, cell culture medium was replaced with fresh media supplemented with 250 ug/ml Geneticin (G418, Sigma-Aldrich, G8168; effective dose was established in a killing curve experiment prior to generation of the stably expressing cells). Selection media was changed regularly every 1-2 days to assure the effective selection of the cells in which WT GPR10 and Neomycin resistance gene bearing plasmid were successfully integrated into the genome. These cells were then pooled and subsequently used for the cAMP accumulation assay described below.
## Cell surface expression ELISA
Cells transfected with N-terminal FLAG- or cMyc-tagged GPR10 constructs were tested for cell surface localization of the receptor with ELISA. A day after transfection, cells were fixed with $3.7\%$ paraformaldehyde for 15 min at room temperature and washed three times with PBS. Subsequently, non-specific binding sites were blocked with $3\%$ non-fat dry milk in 50 mM Tris-PBS pH 7.4 (blocking buffer) for 1 h at room temperature. Next, cells were incubated with a either a mouse monoclonal anti-FLAG antibody (Sigma-Aldrich, F1804) or anti-cMyc antibody (Millipore, CBL434, dilution 1:1000 in blocking buffer) overnight at 4 °C followed by triple washing with PBS and incubation with goat anti-mouse IgG(H + L)-HRP conjugate (Bio-Rad Laboratories, 172-1011) (1:1250 in $1.5\%$ non-fat dry milk in 50 mM Tris-PBS pH 7.4) for 2 h at room temperature. Finally, plates were washed three times with PBS and a high performance chromogenic substrate, 3,3´,5,5´̵tetramethylbenzidine (TMB CORE+, Bio-Rad Laboratories, BUF062) was used to detect HRP activity. The reaction was terminated with 0.5 M H2SO4 and absorbance by the colour reaction product at 450 nm was determined using Tecan Infinite M1000 PRO microplate reader.
## Competition Radioligand binding assay
The effect of GPR10 mutants on PrRP-31 binding was assessed in a competition radioligand binding assay in intact, transiently transfected COS-7 cells. Cells were cultured in white 96-well plates and transiently transfected with 5 ng cDNA/well of GPR10 construct or empty vector (negative control) one day after seeding. Binding assays were performed approximately 24 h post transfection. For all following steps, cells were kept at 4 °C (on wet ice) and reagents were ice-cold to assure no receptor internalisation was induced by the agonist, causing undesired intracellular accumulation of radiolabeled tracer. Firstly, cells were washed (200 µl/well) and incubated with binding assay buffer (50 µl/well, 200 mM HEPES, pH 7.4 supplemented with 119 mM NaCl, 4.7 mM KCl, 5 mM MgCl2, 5.5 mM glucose, 1 mg/ml BSA). Varying doses of unlabelled PrRP-31 (1 pM – 1 µM) were added to the cells (5 µl/well), swiftly followed by dispensing 50 µl / well 125I-labelled PrRP-31 (Phoenix Pharmaceuticals Inc., T-008-50; 5 µl/ml dilution) and cells were incubated for 5 h at 4 °C. After washing twice with ice-cold binding assay buffer (200 µl/well), 25 µl/well 0.1 M NaOH was dispensed, cells were shaken for 2 min at 1000 rpm, followed by addition of 100 µl/well scintillation fluid, MicroScint-20 (Perkin Elmer, 6013621) and subsequent shaking for 5 min at 1000 rpm. Activity of 125I-PrRP-31 bound was quantified after at least 3 h settle time using a TopCount 9012 Microplate Liquid Scintillation Counter (Packard) through quantification of β-emission.
## cAMP accumulation assay
Measurement of ligand-induced cAMP generation in GPR10 WT stably expressing cells was done using HitHunter® cAMP assay (DiscoverX, 90-0075SM) according to manufacturer’s protocol with modifications. 20,000 cells/well were seeded in a white poly-D-lysine coated 96-well plate and cultured overnight. The following day, cells were washed with PBS and incubated in 30 μL PBS supplemented with 1 mM 3-isobutyl-1-methylxanthine (IBMX, Cayman Chemical, 13347) for 30 min prior to stimulation with an agonist. Cells were stimulated with serial dilutions of PrRP-31 (1.5 μL/well, 20x dilution, Bachem, 4028740) for further 30 min at 37 °C. Intracellular cAMP detection was carried out directly after the ligand stimulation. 10 μL anti-cAMP antibody followed by 40 μL chemiluminescent substrate/lysis buffer/enzyme donor-cAMP complex mix prepared according to manufacturer’s protocol were added and plates were incubated shaking for 1 h at ambient temperature. Finally, 40 μL enzyme acceptor was dispensed and chemiluminescent signal was quantified after 4-5 h in a TopCount 9012 Microplate Counter (Packard).
## Inositol triphosphate turnover assay
COS-7 cells were transiently transfected with 5 ng/well GPR10 WT or variants cDNA construct, and in case of assays assessing Gαi/o-specific signalling only, co-transfected with additional 7.5 ng/well of GαΔ6qi4myr construct. Following the transfection, cells were cultured overnight in full media supplemented with 5 μl/ml [3H]-myo-inositol (Perkin Elmer, NET115600). After initial wash with Hank’s balanced salt solution (HBSS, Gibco, 14025), 100 μl/well of HBSS containing 10 mM LiCl (Sigma Aldrich, L9650) was added followed by stimulation with 5 μL/well 20x PrRP-31 stock solution for 75 min at 37 °C. Next, assay buffer was aspirated and cell were lysed with 50 μl/well 10 mM formic acid (Sigma Aldrich, F0507) while incubated on ice for at least 30 min. 20 μl lysate was transferred to a solid white 96-well plate containing 80 μl/well 12.5 mg/ml yttrium silicate poly-lysine-coated scintillation proximity beads (Perkin Elmer, RPNQ0010) suspension in water. Plates were sealed with a TopSeal-A PLUS (Perkin Elmer, 6050185), shaken vigorously at high speed for approximately 5 min and relative amount of radiolabeled IP1 was quantified after 8 h settle time using TopCount 9012 Microplate Counter (Packard).
## Analysis of in vitro data
All results from cell-based assays were analysed using GraphPad Prism 8 (GraphPad Software). Radioligand binding and signalling assays (i.e. cAMP accumulation and inositol triphosphate turnover assays) were performed in triplicates and were independently replicated at least three times. Results of each experiment were normalized to basal signal for mock transfected cells and maximal efficacy of PrRP-31 for GPR10 WT in a given assay as specified in a graph. Presented dose-response curves were plotted from the merged normalized data analysed with 3-parameter non-linear regression equation.
## Structure prediction and visualisation
Structural model of GPR10 WT was generated with a protein structure prediction service Robetta (http://www.robetta.org/) using Rosetta Comparative Modelling approach24. Generated structure was visualized and images were rendered using Open-Source PyMOL 1.8.x (https://www.lfd.uci.edu/~gohlke/).
## Generation of Gpr10P193S mutant
We used CRISPR-Cas9 to generate the Gpr10P193S mutation on a C57BL6/J background. Gpr10 (Prlhr) is found on mouse chromosome 19. The sgRNA TGGTAGGTGTGCACCGCGGC-CGG targets the mutation site directly, and adheres to our stringent criteria for off target predictions (guides with mismatch (MM) of 0, 1 or 2 for elsewhere in the genome were discounted, and MM3 were tolerated if predicted off targets were not exonic) according to http://www.sanger.ac.uk/htgt/wge/). The sgRNA sequence was purchased as crRNA oligos, which were annealed with tracrRNA (both oligos supplied IDT; Coralville, USA) in sterile, RNase-free injection buffer (TrisHCl 1 mM, pH 7.5, EDTA 0.1 mM) by combining 2.5 µg crRNA with 5 µg tracrRNA and heating to 95 oC, which was allowed to slowly cool to room temperature.
The following ssODN repair template, with capital letters indicating intended base pair changes was synthesised (IDT): 5’ctgaggctcagcgcctacgcggtgctgggcatctgggctctatctgcagtgctggcgctgTcggcTgcggtgcacacctaccatgtggagctcaagccccacgacgtgagcctctgcgag3’.
This repair template was designed to convert P193 CCG codon to a Serine TCG codon (Supplementary Fig. 4a). A second silent C > T mutation results in loss of an EagI restriction site for screening purposes.
CRISPR reagents (final concentrations; sgRNA 20 ng/µl, Cas9 protein 20 ng/µl, ssODN 50 ng/µl) were directly microinjected into C57BL6/J (Envigo, Bicester, UK) zygote pronuclei using standard protocols. Zygotes were cultured overnight and the resulting two-cell embryos surgically implanted into the oviduct of day 0.5 post coitum pseudopregnant mice. Potential founder mice were identified by extraction of genomic DNA from ear clips, followed by PCR using primers that flank the homology arms and sgRNA site (Geno F1 ttcacactcaccacaatcgc and Geno R1 tcactgacacccgtacgtaa). WT and HDR sequences both produce a 304 bp band, with the WT allele susceptible to EagI digest (note, NHEJ could also result in loss of EagI digest). Several candidate mice were identified (Supplementary Fig. 4b) and three were taken forward for sequencing. The product was amplified using high fidelity Phusion polymerase (NEB), gel extracted and subcloned into pCRblunt (Invitrogen, UK). Colonies were mini-prepped and Sanger sequenced with M13 Forward and Reverse primers, and aligned to predicted knock-in sequence (Supplementary Fig. 4c). Positive pups were bred with a WT C57BL6/J to confirm germline transmission and a colony established.
## Mouse phenotyping studies
All procedures were conducted in accordance with the United Kingdom Animals (Scientific Procedures) Act, 1986 (ASPA). All animal experiments were performed according to U.K. Home Office licensing laws and approved by the local Animal Welfare and Ethical Review Board (University of Manchester, UK). In all experiments mice were group housed unless otherwise stated in a 12:12 h light: dark cycle, at room temperature (22 oC; 50–$55\%$ humidity). Mice were provided with ad libitum access to standard rodent chow unless otherwise stated (#801151 RM1-P; Special Diet Services, Essex, UK) and water.
The Gpr10-/- null mouse (C57BL6/J background) was a kind gift from Dr Alain Stricker-Krongrad (Millennium Pharmaceuticals, Cambridge, USA). The Gpr10P193S mutant was generated by the University of Manchester Genome Editing Unit, as described below.
For body growth curves Gpr10-/- null mice (male $$n = 21$$; female $$n = 6$$) and Gpr10+/+ wild-type mice (male $$n = 25$$; female $$n = 6$$) were maintained on standard chow and weighed weekly. Homozygous Gpr10P193S/P193S (male $$n = 7$$; female $$n = 5$$), heterozygous Gpr10+/P193S (male $$n = 18$$; female $$n = 8$$) and wild-type Gpr10+/+ (male $$n = 9$$; female $$n = 8$$) mice were switched from standard chow to ad libitum access to $60\%$ high energy diet (HED; #824054 RM $60\%$ energy from fat; Special Diet Services, Essex, UK) at 8 weeks of age and weighed weekly for body growth curves.
In separate daily food intake studies, pre-obese Gpr10-/- null and Gpr10+/+ wild-type mice (6 week old male $$n = 5$$ in each group; 8 week old female $$n = 6$$ in each group), and 8 week old homozygous Gpr10P193S/P193S (male $$n = 11$$), heterozygous Gpr10+/P193S (male $$n = 9$$) and wild-type Gpr10+/+ (male $$n = 9$$) were housed individually in standard cages with ad libitum access to standard chow. After at least a one week period of acclimatisation to single housing, daily food intake was measured over a minimum of 3 days at the same time each day and averaged to provide mean daily food take weight±SEM. Following initial food intake studies, Gpr10-/- null and Gpr10+/+ wild-type mice (male $$n = 4$$ in each group; female $$n = 6$$ in each group) and homozygous Gpr10P193S/P193S (male $$n = 11$$), heterozygous Gpr10+/P193S (male $$n = 9$$) and wild-type Gpr10+/+ (male $$n = 9$$) were then acclimated to indirect calorimetric cages (Columbus Instruments, Columbus, OH, USA) for 2-3 days prior to study and data were collected for a minimum of 3 days during which time oxygen consumption (VO2 ml/hr) was measured every 8 min using Oxymax® software (Columbus Instruments, Columbus, OH, USA). Cages were not equipped with running wheels, and environmental enrichment was limited to bedding material. During this time all mice had ad libitum access to standard chow and water. Data were averaged every 30 min for continuous plots and day and night data averaged over the 3-day study period. Mice were then returned to their home cages and daily food intake and oxygen consumption were recorded for male and female Gpr10-/- null and Gpr10+/+ wild-type mice during the same study at later stages; at 10, 14 and 18 weeks for male mice ($$n = 5$$ in each group), and at 25 weeks for female mice (Gpr10+/+ $$n = 6$$ and Gpr10-/- null $$n = 5$$ respectively).
## Statistical analysis of in vivo data
Data are presented as means±SEM. Statistical analyses were performed using Prism statistical package (GraphPad Software Inc, San Diego, USA). Two-tailed Student’s t-tests or two-way ANOVA followed by Bonferroni’s multiple-comparison post hoc tests were used to compare two or three groups, respectively. Body-weight data were analysed with a mixed-effects analysis followed by Sidak’s multiple-comparison post hoc test.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 -7 Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-36966-3.
## Source data
Source Data
## Peer review information
Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work.
## References
1. Berrington de Gonzalez A. **Body-mass index and mortality among 1.46 million white adults**. *N. Engl. J. Med.* (2010) **363** 2211-2219. DOI: 10.1056/NEJMoa1000367
2. van der Klaauw AA, Farooqi IS. **The hunger genes: pathways to obesity**. *Cell* (2015) **161** 119-132. DOI: 10.1016/j.cell.2015.03.008
3. Hinuma S. **A prolactin-releasing peptide in the brain**. *Nature* (1998) **393** 272-276. DOI: 10.1038/30515
4. Lawrence CB, Celsi F, Brennand J, Luckman SM. **Alternative role for prolactin-releasing peptide in the regulation of food intake**. *Nat. Neurosci.* (2000) **3** 645-646. DOI: 10.1038/76597
5. Dodd GT, Luckman SM. **Physiological Roles of GPR10 and PrRP Signaling**. *Front Endocrinol. (Lausanne)* (2013) **4** 20. DOI: 10.3389/fendo.2013.00020
6. Dodd GT. **The thermogenic effect of leptin is dependent on a distinct population of prolactin-releasing peptide neurons in the dorsomedial hypothalamus**. *Cell Metab.* (2014) **20** 639-649. DOI: 10.1016/j.cmet.2014.07.022
7. Bechtold DA, Luckman SM. **Prolactin-releasing Peptide mediates cholecystokinin-induced satiety in mice**. *Endocrinology* (2006) **147** 4723-4729. DOI: 10.1210/en.2006-0753
8. Prazienkova V. **Impact of novel palmitoylated prolactin-releasing peptide analogs on metabolic changes in mice with diet-induced obesity**. *PloS one.* (2017) **12** e0183449. DOI: 10.1371/journal.pone.0183449
9. Maletinska L. **Novel lipidized analogs of prolactin-releasing peptide have prolonged half-lives and exert anti-obesity effects after peripheral administration**. *Int J. Obes. (Lond.)* (2015) **39** 986-993. DOI: 10.1038/ijo.2015.28
10. Hendricks AE. **Rare Variant Analysis of Human and Rodent Obesity Genes in Individuals with Severe Childhood Obesity**. *Sci. Rep.* (2017) **7** 4394. DOI: 10.1038/s41598-017-03054-8
11. Bhattacharyya S. **Association of polymorphisms in GPR10, the gene encoding the prolactin-releasing peptide receptor with blood pressure, but not obesity, in a U.K. Caucasian population**. *Diabetes* (2003) **52** 1296-1299. DOI: 10.2337/diabetes.52.5.1296
12. Kostenis E. **Is Galpha16 the optimal tool for fishing ligands of orphan G-protein-coupled receptors?**. *Trends Pharmacol. Sci.* (2001) **22** 560-564. DOI: 10.1016/S0165-6147(00)01810-1
13. Gu W, Geddes BJ, Zhang C, Foley KP, Stricker-Krongrad A. **The prolactin-releasing peptide receptor (GPR10) regulates body weight homeostasis in mice**. *J. Mol. Neurosci.* (2004) **22** 93-103. DOI: 10.1385/JMN:22:1-2:93
14. Bjursell M, Lenneras M, Goransson M, Elmgren A, Bohlooly YM. **GPR10 deficiency in mice results in altered energy expenditure and obesity**. *Biochem Biophys. Res. Commun.* (2007) **363** 633-638. DOI: 10.1016/j.bbrc.2007.09.016
15. Watanabe TK. **Mutated G-protein-coupled receptor GPR10 is responsible for the hyperphagia/dyslipidaemia/obesity locus of Dmo1 in the OLETF rat**. *Clin. Exp. Pharm. Physiol.* (2005) **32** 355-366. DOI: 10.1111/j.1440-1681.2005.04196.x
16. Mirzadegan T, Benko G, Filipek S, Palczewski K. **Sequence analyses of G-protein-coupled receptors: similarities to rhodopsin**. *Biochemistry* (2003) **42** 2759-2767. DOI: 10.1021/bi027224+
17. Zuk O. **Searching for missing heritability: designing rare variant association studies**. *Proc. Natl Acad. Sci.* (2014) **111** E455-E464. DOI: 10.1073/pnas.1322563111
18. Isberg V. **Generic GPCR residue numbers - aligning topology maps while minding the gaps**. *Trends Pharmacol. Sci.* (2015) **36** 22-31. DOI: 10.1016/j.tips.2014.11.001
19. Farooqi IS. **Clinical spectrum of obesity and mutations in the melanocortin 4 receptor gene**. *N. Engl. J. Med* (2003) **348** 1085-1095. DOI: 10.1056/NEJMoa022050
20. Saeed S. **Genetic Causes of Severe Childhood Obesity: A Remarkably High Prevalence in an Inbred Population of Pakistan**. *Diabetes* (2020) **69** 1424-1438. DOI: 10.2337/db19-1238
21. Zhao Z. **UK Biobank Whole-Exome Sequence Binary Phenome Analysis with Robust Region-Based Rare-Variant Test**. *Am. J. Hum. Genet.* (2020) **106** 3-12. DOI: 10.1016/j.ajhg.2019.11.012
22. Ghoussaini M. **Open Targets Genetics: systematic identification of trait-associated genes using large-scale genetics and functional genomics**. *Nucleic Acids Res.* (2021) **49** D1311-D1320. DOI: 10.1093/nar/gkaa840
23. Buniello A. **The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019**. *Nucl Acids Res.* (2019) **47** D1005-D1012. DOI: 10.1093/nar/gky1120
24. Song Y. **High-resolution comparative modeling with RosettaCM**. *Structure* (2013) **21** 1735-1742. DOI: 10.1016/j.str.2013.08.005
|
---
title: Vascular comorbidity is associated with decreased cognitive functioning in
inflammatory bowel disease
authors:
- Ronak Patel
- Ruth Ann Marrie
- Charles N. Bernstein
- James M. Bolton
- Lesley A. Graff
- James J. Marriott
- Chase R. Figley
- Jennifer Kornelsen
- Erin L. Mazerolle
- Md Nasir Uddin
- John D. Fisk
- James Bolton
- James Bolton
- Lesley Graff
- Jennifer Kornelsen
- Erin Mazerolle
- Ronak Patel
- Teresa D. Figley
- Carl A. Helmick
journal: Scientific Reports
year: 2023
pmcid: PMC10017678
doi: 10.1038/s41598-023-31160-3
license: CC BY 4.0
---
# Vascular comorbidity is associated with decreased cognitive functioning in inflammatory bowel disease
## Abstract
Reports of cognitive impairment in inflammatory bowel disease (IBD) have been mixed. IBD and cardiovascular disease are often co-morbid, yet it remains unknown whether vascular comorbidity confers a risk for decreased cognitive functioning, as observed in other populations. Participants with IBD were recruited from a longitudinal study of immune-mediated disease. Participants were administered a standardized neuropsychological test protocol, evaluating information processing speed, verbal learning and memory, visual learning and memory, and verbal fluency/executive function. Cognitive test scores were standardized using local regression-based norms, adjusting for age, sex, and education. Vascular risk was calculated using a modified Framingham Risk Score (FRS). We tested the association between FRS and cognitive test scores using a quantile regression model, adjusting for IBD type. Of 84 IBD participants, 54 had ulcerative colitis and 30 had Crohn’s disease; mean (SD) age was 53.36 (13.95) years, and a high proportion were females ($$n = 58$$). As the risk score (FRS) increased, participants demonstrated lower performance in information processing speed (β = − 0.12; $95\%$ CI − 0.24, − 0.006) and verbal learning (β = − 0.14; $95\%$ CI − 0.28, − 0.01) at the 50th percentile. After adjusting for IBD type and disease activity, higher FRS remained associated with lower information processing speed (β = − 0.14; $95\%$ CI − 0.27, − 0.065). Vascular comorbidity is associated with lower cognitive functioning in persons with IBD, particularly in the area of information processing speed. These findings suggest that prevention, identification, and treatment of vascular comorbidity in IBD may play a critical role for improving functional outcomes in IBD.
## Introduction
Inflammatory bowel disease (IBD), including Crohn’s disease (CD) and ulcerative colitis (UC) is a complex chronic immune-mediated disease whereby the gastrointestinal tract becomes inflamed and ulcerated1. Canada is among countries with the highest incidence rate of IBD in the world2. In 2018, the number of Canadians living with IBD was approximately 270,000 and this number is predicted to rise to 403,000 by 20302. Globally, epidemiological studies have demonstrated a growing prevalence and burden of IBD around the world3.
The last decade has seen an increased focus on understanding the role of the gut-brain axis in human disease4. This bidirectional communication pathway between the gut and brain involves a complex interplay between neural, immune, and endocrine systems which is thought to have a critical role in the pathophysiology of IBD5,6 Stress-inducing conditions are now known to lead to the activation of peripheral and neuronal cell pathways that influence neuroinflammatory mechanisms. Chronic neuroinflammation can lead to neuronal cell death7. Increasing attention has been given to how these neuronal changes might affect cognitive functioning in IBD, although data regarding cognitive functioning in persons with IBD remain relatively limited8.
Early studies examining the link between IBD and cognition found that persons with IBD exhibited a relative and selective deficit in verbal intellectual functioning (i.e., the ability to reason and solve problems using language) as compared to their own performance-based IQ and to healthy controls9,10. However, subsequent studies have not reported this selective deficit in verbal intellectual functioning. Some studies have failed to find any changes in cognitive functioning in persons with IBD11, while others have failed to find differences in cognitive functioning between persons with IBD and controls when variables such as concurrent mood disorder and level of education were considered5. Several comorbid conditions have been linked to IBD, with a growing number of studies demonstrating that patients with IBD have an increased risk of developing cardiovascular disease12 Both conditions reflect chronic inflammatory processes and share certain pathophysiological mechanisms that may influence each other13–15. In the general population, cardiovascular disease, including high blood pressure and diabetes, is associated with cognitive decline and is known to affect cognitive functions such as information processing speed (i.e., how quickly information is processed), executive functioning (i.e., higher-order functions such as planning, organization, reasoning, and problem-solving), and learning and memory (i.e., the ability to retain new information)16. However, to our knowledge, no study to date has directly examined the effect of comorbid cardiovascular disease and related risk factors on cognitive functioning in persons with IBD.
One measure that simultaneously accounts for multiple cardiovascular conditions and risk factors, while also considering hypertension treatment status and gender, is the Framingham Risk Score (FRS)17. Higher FRS scores have been associated with decreased brain volume in population-specific samples such as older adults18,19 and more recent findings that reveal lower FRS scores in type 2 diabetes may be associated with better cognitive performance, as compared to those with higher FRS scores20. Given the increased co-occurrence of cardiovascular disease in IBD12, and inconsistencies in findings of prior studies of cognition in IBD, we aimed to evaluate the association between FRS and cognitive performance in a sample of persons with IBD. We hypothesized that higher FRS in individuals with IBD would be associated with poorer cognitive functioning.
## Participants
The source population was participants in a longitudinal study of the effects of psychiatric comorbidity on chronic immune-mediated inflammatory diseases (the ‘IMID’ study) that recruited 247 persons with clinically confirmed IBD21. The present study enrolled a subgroup of 84 IBD study participants aged ≥ 18 years, with adequate knowledge of the English language. This subgroup sample reflected a group of participants who were serially approached for participation on the basis of whether they were still participating in the IMID study and had an upcoming annual visit between September 2016 and July 2017, during which funding for the current study was available. Exclusion criteria included comorbid brain tumors or neurodegenerative disorders or contraindications to magnetic resonance imaging, The University of Manitoba Health Research Ethics Board approved the study and all participants provided written informed consent. This research was performed in accordance with relevant guidelines/regulations set out by the University of Manitoba Health Research Ethics Board. As described in detail elsewhere21, participants completed validated questionnaires, and underwent standardized clinical assessments conducted by trained personnel. Participants nor the public were involved in the design, conduct, reporting, or dissemination of our research.
## Sociodemographic information and health behaviors
Participants reported sex, date of birth, race and ethnicity (white and non-white), highest level of education attained, annual household income, and marital status. Highest level of education completed was reported as elementary school, junior high school, high school diploma/General Education Diploma (GED), college, technical/trade, university bachelor’s degree, university master’s degree, university doctorate or other. Annual household income was reported as < $15,000, $15,000–29,999, $30,000–49,999, $50,000–100,000, > $100,000 or ‘I do not wish to answer’. Participants who reported ever smoking ≥ 100 cigarettes were classified as smokers22. Current smoking status was reported as not at all, some days, or every day. Body mass index (BMI, kg/m2) was calculated based on height and weight measured at the study visit.
## Clinical characteristics
We extracted age at symptom onset, age at IBD diagnosis, and type of IBD diagnosis (ulcerative colitis or Crohn’s disease) and current disease-modifying pharmacological agents used (if any) from medical records. Disease-modifying pharmacological agents included thiopurines, methotrexate, TNF-alpha antagonists, ustekinumab, vedoliuzmab, corticosteroids and 5-aminosalicylates (5-ASA). All other medications were recorded by patient interview. We used the Montreal Classification to classify IBD disease course23.To assess whether or not there was likely active intestinal inflammation, stool was collected for calprotectin measurement, where a value of ≥ 250 mcg/g of stool was considered active disease24.
We used a validated comorbidity questionnaire to capture the number of physical comorbidities that participants had25; these comorbid conditions could include chronic lung disease, cancer (breast, colon, lung, skin, and other), migraine, thyroid disease, lupus, osteoarthritis, osteoporosis, fibromyalgia, kidney disease, peptic ulcer, liver disease, and epilepsy. We report physical comorbidities as a count (0, 1, ≥ 2).
## Cognition
Our neuropsychological test battery consisted of well-validated measures assessing the major domains of information processing speed (the rate at which information is processed), verbal learning and memory (the ability to learn and retain information that we hear), visual learning and memory (the ability to learn and retain information that we see), and verbal fluency/executive ability (the ability to generate words utilizing different search strategies). We assessed information processing speed using the oral version of the Symbol Digit Modalities Test26 (SDMT) verbal learning and memory using the California Verbal Learning Test27 (CVLT-II; Trial 1–5 total recall score), visual learning and memory using the Brief Visuospatial Memory Test-Revised28 (BVMT-R; summed recall score for all three learning trials), and language and executive abilities using tests of verbal fluency29 (letter and animal categories). Neuropsychological tests were administered and scored by research assistants who underwent a comprehensive training protocol led and supervised by a board-certified clinical neuropsychologist (first author R.P.).
## Vascular comorbidity
we focused on hypertension and diabetes because of their associations with disability in other clinical populations and their effects on cognition in the general population30. Participants reported these comorbidities using a validated questionnaire25. They reported if a physician had diagnosed the comorbidity, and if yes, the year of diagnosis and whether the condition was currently treated.
We augmented the information provided by questionnaire with additional assessments. At the in-person study visit, concurrent with participants’ cognitive assessment, blood pressure was measured once in the seated position using an automatic blood pressure machine. We collected a serum sample to measure hemoglobin A1c (HbA1c). We classified participants as hypertensive31 (i.e., any of: self-reported or physician-diagnosed hypertension, use of hypertensive medications, measured BP > $\frac{140}{90}$) or not. We classified participants as having diabetes (i.e., any of: self-reported or physician-diagnosed diabetes, use of medications for diabetes, HbA1c > $6.5\%$32) or not. We also classified participants as having hyperlipidemia (self-reported or physician-diagnosed hyperlipidemia or use of lipid-lowering medications) or not. As few participants with any of these conditions reported being untreated, we did not pursue analyses stratifying these conditions by treatment status.
We then used this information to calculate vascular risk for each participant. The Framingham Risk Score (FRS) is a sex-specific weighted index, and incorporates information regarding age, smoking history, hypertension, and diabetes, as well as either lipid status or body mass index17. Points are added for factors that increase cardiovascular risk such as a history of smoking and current cardiovascular conditions, while negative points are assigned for factors that are protective17. Points for hypertension incorporate measured systolic blood pressure and treatment status (treated or not treated). Because we did not have serum lipid measurements, we used the version of the FRS that relied on body mass index. Age was excluded from the FRS to ensure we did not confound the effect of age on cognition with the effects of vascular risk factors19.
## Psychiatric comorbidity
Participants reported symptoms of depression and anxiety using the Hospital Anxiety and Depression.
Scale33 (HADS), which is validated for use in IBD populations34. The HADS includes 7 items each that assess symptoms of depression (HADS-D) and anxiety (HADS-A) respectively, with total scores ranging from 0 to 21. Participants were classified as to whether they had clinically meaningful symptoms of depression (HADS-D score) or anxiety (HADS-A score). The literature varies regarding the optimal cut-point for the HADS in IBD. Therefore, we employed the more specific cut-point of ≥ 11, which indicates clinically meaningful symptoms of depression and anxiety35.
## Statistical analyses
We used descriptive statistics to characterize the study population, including mean (standard deviation), median (interquartile range) and frequency (percent). For the cognitive data, raw test scores were converted to age, sex and education-adjusted z-scores using locally derived regression-based norms36. Z-scores of ≤ − 1.5 were classified as impaired. The Wechsler Test of Adult Reading (WTAR), was included to provide an age-, sex-, education-, and ethnicity-adjusted Full-Scale IQ estimate of premorbid intelligence and was used to characterize the sample37. We examined Spearman correlations ($95\%$ confidence intervals [CI]) between the FRS and cognitive test results. We tested the association between the FRS and cognition using quantile regression. Quantile regression allows the evaluation of a relationship of an independent variable across the full range of a continuous dependent variable rather than its conditional mean and does not require distributional assumptions such as normality or homoscedasticity38. In the absence of any a priori information to guide the choice of quantile, we examined the 50th percentile (quantile). The primary outcome (dependent variable) was information processing speed (SDMT), while the independent variable of interest was FRS score (continuous). Secondary analyses were conducted with other cognitive variables of interest (e.g., verbal learning and memory [CVLT], and visual learning and memory [BVMT-R]). We adjusted for IBD type and active disease but did not adjust for psychiatric comorbidity (i.e., HADS-A and HADS-D scores) due to the small number of participants with elevated anxiety/depression scores (see Table 1). To account for the multiple comparisons introduced by using multiple cognitive tests in the regression analysis we applied a Benjamini–Hochberg correction for multiple comparisons for the secondary outcomes, with a false discovery rate of 0.05. Statistical analyses were completed using SAS V9.4 (SAS Institute Inc., Cary, NC).Table 1Descriptive statistics of the entire IBD sample, ulcerative colitis and Chron's disease (CD) subgroups; Mean (SD).CharacteristicAll ($$n = 84$$)UC ($$n = 54$$)CD ($$n = 30$$)Parametric P-value*Non-Parametric P-value**Age53.36 (13.95)52.61 (13.8)54.71 (14.34)0.510.45Gender, n (%)Male26 (31.0)15 (27.78)11 (36.67)0.400.46Female58 (69.0)39 (72.22)19 (63.33)Race, n (%)Non-White9 (10.7)4 (7.41)5 (16.67)0.190.27White75 (89.3)50 (92.59)25 (83.33)Education, n (%)≤ High School/GED14 (17.5)7 (13.73)7 (24.14)0.240.36> High School66 (82.5)44 (86.27)22 (75.86)Income, n (%)< 50,00017 (20.2)12 (22.22)5 (16.67)0.580.66> 50,00062 (73.8)38 (70.37)24 [80]Declined to report5 (6.0)4 (7.41)1 (3.33)Current Smoker, n (%)No73 (86.9)47 (87.04)26 (86.67)0.961.00Yes11 (13.1)7 (12.96)4 (13.33)Other comorbidities, n (%)024 (28.6)15 (27.78)9 [30]0.830.84120 (23.8)12 (22.22)8 (26.67)240 (47.6)27 [50]13 (43.33)Current BMI27.84 (5.92)26.99 (5.32)29.39 (6.69)0.0750.13Any disease modifying pharmacotherapy, n (%)No25 (29.8)15 (27.78)10 (33.33)0.590.63Yes59 (70.2)39 (72.22)20 (66.67)HADS-D < 11No81 (96.40)53 (98.15)28 (93.33)0.260.29HADS-D ≥ 11Yes3 (3.60)1 (1.85)2 (6.67)HADS-A < 11No72 (85.70)48 (88.89)24 (80.00)0.260.33HADS-A ≥ 11Yes12 (14.3)6 (11.11)6 (20.00)Active diseasea, n (%)No47 (58.8)31 (59.6)16 (57.1)0.83Yes33 (41.2)21 (40.4)12 (42.9)UC Involvement, n (%)E1: Ulcerative Proctitis–3 (10.34)0.590.63E2: Left-sided disease–14 (48.28)E3: Extensive ulcerative colitis–12 (41.38)CD Location, n (%)L1: Terminal Ilium Only20 (37.74)–0.590.63L2: Colon Only6 (11.32)–L3: Small Bowel and Colon27 (50.94)–CD Behavior, n (%)B1: Inflammatory19 (35.19)–0.590.63B2: Structuring18 (33.33)–B3: Fistulizing17 (31.48)–CD: Upper GI involvement, n (%)–1.2CD Perianal disease, n (%)–8.3a = 4 missing.*The parametric P-value is calculated by ANOVA for numerical covariates and chi-square test for the categorical covariates.**The non-parametric P-value is calculated by the Kruskal–Wallis test for numerical covariates and Fisher's exact test for categorical variables.
## Results
Out of the 84 participants enrolled, 54 had a diagnosis of UC and 30 had a diagnosis of CD. The sample overall was predominantly white females with the equivalent of a high school education or greater. Over half were currently taking a disease-modifying pharmacological therapy/agent (see Table 1). Generally, participants had relatively intact cognitive functioning with above average overall estimated intellectual functioning (Table 2). *Participants* generally performed within the average range across all cognitive domains; a similar pattern was observed when scores were separated by IBD subtype (Table 3). Correlational analyses revealed higher vascular comorbidity, as measured by higher FRS, was associated with poorer performance in the areas of information processing speed, verbal learning, visual memory, and verbal fluency (Table 4). In a quantile regression model, higher FRS was associated with lower information processing speed and verbal learning at the 50th percentile (Table 5). After adjustment by IBD subtype and active disease, higher FRS remained associated with lower cognitive functioning for information processing speed (P-value = 0.041). The magnitude of the association of higher FRS with lower verbal learning remained similar, with broader confidence intervals observed, but the association was not statistically significant. These results generally remained unchanged when excluding those participants with elevated mood and anxiety symptoms (supplemental table e1).Table 2Cognitive functioning results (z-scores) for full IBD sample. FSIQ estimate, mean (SD)111.02 (6.80)SDMT, Man mean (SD)0.23 (1.24)CVLT verbal learning, mean (SD)0.57(1.19))CVLT delayed recall, mean (SD)0.64 (1.06)BVMT-R total learning, mean (SD)− 0.05 (0.88)BVMT-R delayed recall, mean (SD)0.37 (1.10)LNS, mean (SD)0.01 (1.03)Verbal fluency, mean (SD)− 0.11 (1.03)FSIQ Full Scale Intellectual Quotient, SDMT Symbol Digits Modality Test, CVLT California Verbal Learning Test, 2nd Ed., BVMT-R Brief Visualspatial, Test – Revised, LNS Letter Number Sequencing Test. Table 3Cognitive functioning results for UC vs. CD.UCCDFSIQ estimate, mean (SD)111.30 (6.34)110.52 (7.67)SDMT, Man mean (SD)0.29 (1.30)0.12 (1.12)CVLT verbal learning, mean (SD)0.70 (1.13)0.33 (1.28)CVLT delayed recall, mean (SD)0.71 (0.99)0.51 (1.18)BVMT-R total learning, mean (SD)0.01 (0.85)− 0.16 (0.93)BVMT-R delayed recall, mean (SD)0.10 (1.09)− 0.28 (1.09)LNS, mean (SD)0.04 (0.95)0.20 (1.18)Verbal fluency, mean (SD)0.20 (0.99)0.04 (1.12)FSIQ Full Scale Intellectual Quotient, SDMT Symbol Digits Modality Test, CVLT California Verbal Learning Test, 2nd Ed., BVMT-R Brief Visualspatial, Test – Revised, LNS Letter Number Sequencing Test. Table 4Spearman correlation coefficients ($95\%$ confidence intervals) between FRS and cognitive functioning z-scores.r-valueP-valueSDMT− 0.28 (− 0.46, − 0.06)0.01CVLT-II verbal learning− 0.25 (− 0.44, − 0.04)0.02CVLT-II delayed recall− 0.21 (− 0.40, 0.01)0.056BVMT-R total learning− 0.25 (− 0.44, − 0.04)0.02BVMT-R delayed recall− 0.26 (− 0.44, − 0.04)0.02LNS− 0.09 (− 0.29, 0.13)0.42Verbal fluency− 0.22 (− 0.41, − 0.01)0.04Significant values are in bold. SDMT Symbol Digit Modalities Test, CVLT-II California Verbal Learning Test-II, BVMT-R Brief Visuospatial Memory Test-Revised, LNS Letter-Number Sequencing Test. Table 5Unadjusted and adjusted regression coeffecients ($95\%$ confidence intervals) for the association between FRS and cognitive functioning. QuantileSDMTCVLT-IICVLT-II LDBMVT-RBMVT-R DRLNSFluencyUnadjustedBeta ($95\%$CI)0.5− 0.12 (− 0.24,-0.01)− 0.14 (− 0.27, − 0.01)− 0.023 (− 0.13, 0.08)− 0.091 (− 0.16, − 0.02)− 0.43 (− 0.11, 0.03)− 0.026 (− 0.14, 0.09)− 0.075 (− 0.15,0.001)$$P \leq 0.033$$$P \leq 0.035$P = 0.48P = 0.054P = 0.34P = 0.68P = 0.077Adjusted for IBD type and disease activity0.5− 0.14 (− 0.27, 0.065)− 0.11 (− 0.22, − 0.008)− 0.030 (− 0.15, 0.094)− 0.073 (− 0.15, 0.003)− 0.068 (− 0.14, 0.004)− 0.01 (− 0.14, 0.12)− 0.083 (− 0.17, 0.004)$$P \leq 0.041$$$P \leq 0.072$P = 0.63P = 0.055P = 0.059P = 0.87P = 0.058Significant values are in bold. Sign using wald test as are the other ones sign using LR test. SDMT Symbol Digit Modalities Test, CVLT-II California Verbal Learning Test-II, LD Long Delay, BVMT-R Brief Visuospatial Memory Test-Revised, DR Delayed Recall, LNSLetter-Number Sequencing Test.
## Discussion
In this cross-sectional study, we examined the association between vascular risk and cognitive function in a sample of persons with clinically confirmed IBD. Findings of altered cognitive function in persons with IBD have been mixed, and to date no study has directly examined the effect of vascular comorbidity on cognitive function in IBD. However, our findings demonstrate an association between increased vascular risk and decreased cognitive function in IBD. We found that higher vascular comorbidity was correlated with poorer performance in information processing speed, verbal learning, visual memory, and verbal fluency. The main results from our quantile regression analyses revealed that higher vascular comorbidity was predictive of lower information processing speed, and this remained true even after adjusting for IBD type and disease activity.
A growing body of literature suggests IBD is associated with cognitive impairment. In a recent systematic review and meta-analysis8, it was shown that persons with IBD, in disease remission, exhibited deficits in overall executive functioning including moderate deficits in working memory, as compared to healthy controls. However, the authors failed to find any differences in learning and recall. Our study reveals that cardiovascular disease may be an important factor mediating the relationship between IBD and cognitive impairment. In non-IBD populations, vascular conditions such as diabetes and hypertension have been associated with poorer cognitive outcomes including an increased risk of developing dementia39,40. Meta-analytic findings reveal type 2 diabetes is associated with impairments in motor functioning, executive functioning, processing speed, and verbal and visual memory41. Similarly, meta-analytic findings (across 12 studies and 4,076 individuals) have demonstrated significant associations between increasing blood pressure and reductions in episodic memory in older adults who are free of clinical dementia or stroke39. In our quantile regression model, we found that increased FRS in IBD is associated with decreased information processing speed, a domain frequently affected by cardiovascular disease. Deficits in cognition such as information processing speed have been associated with increasing levels of disability and reduced functioning in everyday life including decreased occupational functioning42.
Several purported pathophysiological mechanisms may underlie decreased cognitive function in IBD and co-occurring cardiovascular disease. Such postulated mechanisms include altered metabolic functioning interfering with neurogenesis in regions such as the hippocampus (a region critical for the facilitation of learning and memory), the expression of pro-inflammatory cytokines leading to neuronal damage, and oxidative stress leading to chronic neuroinflammation and neurodegeneration13–15. Pathophysiological changes may also be evident as alterations in brain structure. Indeed, we recently demonstrated that higher vascular comorbidity as indexed by FRS was associated with lower brain volume at baseline, and with greater brain volume loss over time, in persons with multiple sclerosis19. Future studies should examine whether vascular comorbidity is associated with similar reductions in brain volume and decreased cognitive function over time in IBD. Systematic investigation of determining effects of vascular morbidity on cognitive functioning has been a critical gap in the IBD literature. Our use of FRS as a summary score of vascular risk reduced the number of comparisons, thereby avoiding the challenges associated with having small numbers of participants affected by a specific comorbidity while also accounting for the frequent co-occurrence of comorbidities associated with increased vascular risk. Nevertheless, there are several limitations to our study. We did not include a non-IBD control group with increased vascular risk, and therefore, we were unable to directly test whether there is an additive or synergistic interaction between IBD and vascular risk on cognitive functioning. Inclusion of a group of individuals with vascular risk factors without IBD would allow us to directly examine to what extent the magnitude of the observed effects differs between persons with and without IBD. Our overall sample size was modest, with more individuals in the UC versus CD group, though after adjusting for IBD type and disease activity, the association between increased FRS and lower information processing speed, remained unchanged. As outlined in Table 5, the confidence intervals associated with each of the regression coefficients generally increased in size, most likely reflecting the smaller sample size of each subgroup. We were unable to examine the effects of psychiatric comorbidity, as mental health concerns were only modestly elevated in this sample, but this allowed us to more readily isolate the effects of vascular morbidity on cognition in IBD. Nevertheless, future studies should aim to recruit larger samples to systematically investigate the influence of these additional comorbid factors on cognition in IBD. Hyperlipidemia was not used to calculate our primary exposure of interest (FRS), but the FRS based on BMI performs similarly. In the current study, blood pressure was measured once in the seated position; however, it is important to consider is that FRS does not solely depend on blood pressure measurement. Studies have shown that ambulatory blood pressure monitoring does not substantially improve risk prediction with the FRS over the average of two measurements43.Thus, the impact of misclassification based on a single measurement is likely to be small and biased toward the null. Nevertheless, future studies may benefit from using an average across multiple measurements. We were also unable to evaluate the effects of specific therapies or treatments used to manage vascular risk factors, but the FRS incorporates information regarding the severity of risk factors and treatment status for hypertension. This should be the subject of future work.
Our findings demonstrate that higher vascular risk is associated with lower cognitive function in persons with IBD, specifically with respect to information processing speed and learning and memory. These findings suggest that early prevention, identification, and treatment of vascular conditions in IBD may be essential considerations in the overall clinical management of the disease course as it may lead to improved functional outcomes and an overall increase in quality of life. Future studies should examine the role that treatments specifically targeted at reducing vascular risk have on mitigating cognitive decline in IBD. Future studies of cognition in IBD should also consider the potential roles that vascular risk factors and other comorbid conditions may have on current cognitive functioning as well as changes in cognitive functioning over time.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-31160-3.
## References
1. Bernstein CN, Blanchard JF, Rawsthorne P. **The prevalence of extraintestinal diseases in inflammatory bowel disease: A population-based study**. *Am. J. Gastroenterol.* (2001.0) **96** 1116-1122. DOI: 10.1111/j.1572-0241.2001.03756.x
2. Kaplan GG, Bernstein CN, Coward S. **The impact of inflammatory bowel disease in Canada 2018: Epidemiology**. *J. Can. Assoc. Gastroenterol.* (2018.0) **2** 58
3. Alatab S, Sepanlou SG, Ikuta K. **The global, regional, and national burden of inflammatory bowel disease in 195 countries and territories, 1990–2017: A systematic analysis for the global burden of disease study 2017**. *Lancet Gastroenterol. Hepatol.* (2020.0) **5** 17-30. DOI: 10.1016/S2468-1253(19)30333-4
4. Morais LH, Schreiber HL, Mazmanian SK. **The gut microbiota–brain axis in behaviour and brain disorders**. *Nat. Rev. Microbiol.* (2020.0) **19** 241-255. DOI: 10.1038/s41579-020-00460-0
5. Berrill JW, Gallacher J, Hood K. **An observational study of cognitive function in patients with irritable bowel syndrome and inflammatory bowel disease**. *Neurogastroenterol. Motility* (2013.0) **25** 58. DOI: 10.1111/nmo.12219
6. Bonaz BL, Bernstein CN. **Brain-gut interactions in inflammatory bowel disease**. *Gastroenterology.* (2013.0) **144** 36-49. DOI: 10.1053/j.gastro.2012.10.003
7. Kempuraj D, Thangavel R, Selvakumar GP. **Brain and peripheral atypical inflammatory mediators potentiate neuroinflammation and neurodegeneration**. *Front. Cell. Neurosci.* (2017.0) **5** 11
8. Hopkins CWP, Powell N, Norton C. **Cognitive impairment in adult inflammatory bowel disease: A systematic review and meta-analysis**. *J. Acad. Consultation-Liaison Psychiat.* (2021.0) **62** 387-403. DOI: 10.1016/j.psym.2020.10.002
9. Attree EA, Dancey CP, Keeling D. **Cognitive function in people with chronic illness: Inflammatory bowel disease and irritable bowel syndrome**. *Appl. Neuropsychol.* (2003.0) **10** 96-104. DOI: 10.1207/S15324826AN1002_05
10. Dancey CP, Attree EA, Stuart G. **Words fail me: The verbal IQ deficit in inflammatory bowel disease and irritable bowel syndrome**. *Inflamm. Bowel Dis.* (2009.0) **15** 852-857. DOI: 10.1002/ibd.20837
11. Castaneda AE. **Cognitive functioning and depressive symptoms in adolescents with inflammatory bowel disease**. *World J. Gastroenterol.* (2013.0) **19** 1611. DOI: 10.3748/wjg.v19.i10.1611
12. Bernstein CN, Nugent Z, Shaffer S. **Comorbidity before and after a diagnosis of inflammatory bowel disease**. *Aliment. Pharmacol. Therap.* (2021.0) **5** 89
13. Schicho R, Marsche G, Storr M. **Cardiovascular complications in inflammatory bowel disease**. *Curr. Drug Targets* (2015.0) **16** 181-188. DOI: 10.2174/1389450116666150202161500
14. Biondi RB, Salmazo PS, Bazan SG. **Cardiovascular risk in individuals with inflammatory bowel disease**. *Clin. Exp. Gastroenterol.* (2020.0) **13** 107-113. DOI: 10.2147/CEG.S243478
15. Argollo M, Gilardi D, Peyrin-Biroulet C. **Comorbidities in inflammatory bowel disease: A call for action**. *Lancet Gastroenterol. Hepatol.* (2019.0) **4** 643-654. DOI: 10.1016/S2468-1253(19)30173-6
16. Vasquez BP, Zakzanis KK. **The neuropsychological profile of vascular cognitive impairment not demented: A meta-analysis**. *J. Neuropsychol.* (2015.0) **9** 109-136. DOI: 10.1111/jnp.12039
17. D’Agostino RB, Vasan RS, Pencina MJ. **General Cardiovascular Risk Profile for use in primary care**. *Circulation* (2008.0) **117** 743-753. DOI: 10.1161/CIRCULATIONAHA.107.699579
18. Rondina JM, Squarzoni P, Souza-Duran FL. **Framingham coronary heart disease risk score can be predicted from structural brain images in elderly subjects**. *Front. Aging Neurosci.* (2014.0) **9** 6
19. Marrie RA, Patel R, Figley CR. **Higher Framingham risk scores are associated with greater loss of brain volume over time in multiple sclerosis**. *Multiple Sclerosis Relat. Disord.* (2021.0) **54** 103088. DOI: 10.1016/j.msard.2021.103088
20. 20.Restrepo C, Werden E, Singleton R, et al. Comparison of brain atrophy and cognitive performance in individuals with low and high cardiovascular risk: Data from the Diabetes and Dementia (D2) Study. Alzheimer’s and Dementia 2020.
21. Marrie RA, Graff L, Walker JR. **Effects of psychiatric comorbidity in immune-mediated inflammatory disease: Protocol for a prospective study**. *JMIR Res. Protocols* (2018.0) **7** 58. DOI: 10.2196/resprot.8794
22. Grant BF, Hasin DS, Chou SP. **Nicotine dependence and psychiatric disorders in the United States**. *Arch. Gen. Psychiatry* (2004.0) **61** 1107. DOI: 10.1001/archpsyc.61.11.1107
23. Silverberg MS, Satsangi J, Ahmad T. **Toward an integrated clinical, molecular and serological classification of inflammatory bowel disease: Report of a Working Party of the 2005 Montreal World Congress of Gastroenterology**. *Can. J. Gastroenterol.* (2005.0) **19** 5-36. DOI: 10.1155/2005/269076
24. Lin JF, Chen JM, Zuo JH. **Meta-analysis: fecal calprotectin for assessment of inflammatory bowel disease activity**. *Inflamm Bowel Dis* (2014.0) **20** 1407-1415. DOI: 10.1097/MIB.0000000000000057
25. Horton M, Rudick RA, Hara-Cleaver C, Marrie RA. **Validation of a self-report comorbidity questionnaire for multiple sclerosis**. *Neuroepidemiology* (2010.0) **35** 83-90. DOI: 10.1159/000311013
26. 26.Smith A. Symbol digit modalities test: Manual. Los Angeles: Western Psychological Services, 1982.
27. Delis DC, Kramer JH, Kaplan E, Ober BA. *California Verbal Learning Test, second edition, adult version: Manual* (2000.0)
28. 28.Benedict RHB, & Brandt J. Hopkins Verbal Learning Test-Revised / Brief Visuospatial Memory Test-Revised: Professional Manual Supplement. Odessa: Psychological Assessment Resources, 1997.
29. Strauss E, Sherman EMS, Spreen O. *A Compendium of Neuropsychological Tests: Administration, Norms, and Commentary* (2006.0)
30. van den Berg E, Kloppenborg RP, Kessels RPC. **Type 2 diabetes mellitus, hypertension, dyslipidemia and obesity: A systematic comparison of their impact on cognition**. *Biochimica Et Biophysica Acta Mol. Basis Dis.* (2009.0) **1792** 470-481. DOI: 10.1016/j.bbadis.2008.09.004
31. Weinstein G, Maillard P, Himali JJ. **Glucose indices are associated with cognitive and structural brain measures in Young Adults**. *Neurology* (2015.0) **84** 2329-2337. DOI: 10.1212/WNL.0000000000001655
32. 32.American Diabetes Association. Standards of medical care in diabetes-2011. Diabetes care 2011; 34 Suppl 1(Suppl 1):S11–S61.
33. Zigmond AS, Snaith RP. **The hospital anxiety and depression scale**. *Acta Psychiatr. Scand.* (1983.0) **67** 361-370. DOI: 10.1111/j.1600-0447.1983.tb09716.x
34. Bernstein CN, Zhang L, Lix LM. **The validity and reliability of screening measures for depression and anxiety disorders in inflammatory bowel disease**. *Inflamm. Bowel Dis.* (2018.0) **24** 1867-1875. DOI: 10.1093/ibd/izy068
35. Marrie RA, Graff LA, Fisk JD. **The relationship between symptoms of depression and anxiety and disease activity in IBD over time**. *Inflamm. Bowel Dis.* (2021.0) **27** 1285-1293. DOI: 10.1093/ibd/izaa349
36. Marrie RA, Whitehouse CE, Patel R. **Performance of regression-based norms for cognitive functioning of persons with multiple sclerosis in an independent sample**. *Front. Neurol.* (2021.0) **9** 11
37. Wechsler D. *Wechsler Test of Adult Reading: WTAR* (2001.0)
38. Koenker R, Hallock KF. **Quantile regression**. *J. Econ. Pers.* (2001.0) **15** 143-156. DOI: 10.1257/jep.15.4.143
39. Gifford KA, Badaracco M, Liu D. **Blood pressure and cognition among older adults: A meta-analysis**. *Arch. Clin. Neuropsychol.* (2013.0) **28** 649-664. DOI: 10.1093/arclin/act046
40. Zilliox LA, Chadrasekaran K, Kwan JY. **Diabetes and cognitive impairment**. *Curr. Diabetes Rep.* (2016.0) **16** 49. DOI: 10.1007/s11892-016-0775-x
41. Palta P, Schneider ALC, Biessels GJ. **Magnitude of cognitive dysfunction in adults with type 2 diabetes: A meta-analysis of six cognitive domains and the most frequently reported neuropsychological tests within domains**. *J. Int. Neuropsychol. Soc.* (2014.0) **20** 278-291. DOI: 10.1017/S1355617713001483
42. Kavaliunas A, Tinghög P, Friberg E, Olsson T. **Cognitive function predicts work disability among multiple sclerosis patients**. *Multiple Sclerosis J.* (2019.0) **5** 2055217318822134. DOI: 10.1177/2055217318822134
43. Bell KJL, Beller E, Sundström J. **Ambulatory blood pressure adds little to Framingham Risk Score for the primary prevention of cardiovascular disease in older men: Secondary analysis of observational study data**. *BMJ Open.* (2014.0) **4** e006044. DOI: 10.1136/bmjopen-2014-006044
|
---
title: 'Comprehensive prognostic effects of systemic inflammation and Insulin resistance
in women with breast cancer with different BMI: a prospective multicenter cohort'
authors:
- Guo-Tian Ruan
- Hai-Lun Xie
- Chun-Lei Hu
- Chen-An Liu
- He-Yang Zhang
- Qi Zhang
- Zi-Wen Wang
- Xi Zhang
- Yi-Zhong Ge
- Shi-Qi Lin
- Meng Tang
- Meng-Meng Song
- Xiao-Wei Zhang
- Xiao-Yue Liu
- Kang-Ping Zhang
- Ming Yang
- Kai-Ying Yu
- Kun-Hua Wang
- Wen Hu
- Li Deng
- Ming-Hua Cong
- Han-Ping Shi
journal: Scientific Reports
year: 2023
pmcid: PMC10017691
doi: 10.1038/s41598-023-31450-w
license: CC BY 4.0
---
# Comprehensive prognostic effects of systemic inflammation and Insulin resistance in women with breast cancer with different BMI: a prospective multicenter cohort
## Abstract
To investigate the prognostic value of systemic inflammation and insulin resistance in women with breast cancer with different body mass index (BMI). This multicenter, prospective study included 514 women with breast cancer. Multivariate survival analysis showed that patients with high C-reactive protein (CRP), high CRP to albumin ratio (CAR), high lymphocyte to CRP ratio (LCR), high low-density lipoprotein cholesterol to high-density lipoprotein cholesterol ratio (LHR), and high triglyceride to high-density lipoprotein cholesterol ratio (TG/HDL-c) were significantly associated with worse prognosis. The mortality rate of patients with both high CAR and high LHR or both low LCR and high LHR were 3.91-fold or 3.89-fold higher than patients with both low CAR and low LHR or both high LCR and low LHR, respectively. Furthermore, the combination of LCR and LHR significantly predicted survival in patients within the high BMI group. The CRP, CAR, LCR, LHR, and TG/HDL-c were associated with poor survival in women with breast cancer. The combination of CAR and LHR or LCR and LHR could better predict the prognostic outcomes of women with breast cancer, while the combination of LCR and LHR could better predict the prognosis of those patients with overweight or obese patients.
## Introduction
The 2022 cancer statistics for the United States show that from 2014 to 2018, female breast cancer incidence continued to increase (by $0.5\%$ annually), and the number of new female breast cancer cases was 287,850 ($31\%$), ranking as the most prevalent new cancer in women. Female breast cancer had the second highest mortality rate with 43,250 ($15\%$) deaths1. In China, the incidence of female breast cancer still the highest among women2. Inflammation and insulin resistance (IR) play important roles in a variety of chronic diseases, including cancer3. Cancer is generally considered an inflammatory disease, and systemic inflammation is often a hallmark of cancer and a major driver of metabolic alterations in cancer patients4,5. The production of acute-phase proteins, such as C-reactive protein (CRP), is considered an accurate measure of systemic inflammation and pro-inflammatory cytokine activity6. Glucose intolerance is the earliest identified metabolic abnormality in cancer patients7, resulting in a type II diabetic state with IR8. Glicksman et al. showed that about $37\%$ of cancer patients have a diabetic glucose tolerance curve9. The characteristics of IR in cancer patients are distinct from those in type II diabetic patients, in which normal fasting blood glucose is associated with high, normal, or low insulin levels10, manifested by increased hepatic glucose production and gluconeogenesis, possibly due to intracellular gluconeogenesis11. The redistribution of glucose to supply energy needs can lead to hypoglycemia, which in turn, leads to an increase in compensatory hormonal signaling or glucagon.
In recent years, obesity has become the most common metabolic disease worldwide, and its incidence has rapidly increased12. Unfortunately, obesity is fast becoming an epidemic in developed and many developing countries13. Overweight or obesity is associated with an increased risk of recurrence or death in patients with breast cancer14,15. Some obesity-related cancers, such as those of the breast and internal organs, occur in or near fat depots. This suggests that altered fat biology, typically found in the context of elevated BMI, locally contributes to the development of several cancers16. Obesity-induced inflammation or inflammatory disturbances are a major feature of adipose tissue dysfunction17. In fact, adipose tissue is not only a storehouse of excess energy in the form of triacylglycerols (TAGs), but is also an active endocrine organ secreting different peptides called adipocytokines18. The production and expression of inflammatory adipocytokines, such as interleukin (IL)-6, tumor necrosis factor α (TNF-α), and monocyte chemoattractant protein 1, are increased in obese and insulin-resistant subjects19. Compared with lean people, adipose tissue in obese subjects was inflamed by inflammatory macrophages20. Macrophages are important and key contributors to adipocyte inflammation21. Inflammatory macrophages typically accumulate within adipose tissue, and this accumulation leads to localized inflammation. This local inflammation leads to multiple metabolic disturbances, including atherosclerosis and systemic inflammation22. In addition, CRP, another inflammatory marker, is elevated in the serum of individuals with higher BMI23. IR is a common pathological condition in obese patients with impaired insulin action in adipose tissue. During IR, insulin is significantly increased in the circulation to avoid hyperglycemia24. Therefore, insulin is included in the study as a hormone, and insulin levels are often increased in the setting of obesity24. This hyperinsulinemia is associated with BMI25.
Some reports showed that CRP alone26–28 or in combination with other inflammatory markers, such as the CRP to albumin ratio (CAR)29,30, were associated with poor prognosis in breast cancer. Recently, Lymphocyte to C-reactive Protein Ratio (LCR) has been reported to be related to cancer prognosis31–33, but there is no relevant report on the relationship between LCR and breast cancer prognosis. Studies showed that elevated insulin levels and hyperinsulinemia are associated with poor prognosis in breast cancer patients34,35. Some simple and feasible IR surrogate indicators reported earlier have attracted attention relative to the homeostasis model assessment of IR (HOMA-IR)36. These IR indicators included fasting triglyceride glucose (TyG) index37,38, low-density lipoprotein cholesterol to high-density lipoprotein cholesterol ratio (LDL-c/HDL-c, LHR)39, triglyceride to high-density lipoprotein cholesterol ratio (TG/HDL-c)38, and total cholesterol to high-density lipoprotein cholesterol ratio (TC/ HDL-c)38. Thus, in this study, we aimed to select the optimal IR index in breast cancer, and select the best combination of inflammation index and IR index in breast cancer patients with different body mass index (BMI). Finally, we selected the best combination of inflammation and IR indicators for combined survival analysis and selected the best combination of indicators to predict the survival of patients with breast cancer with different BMI. This study aimed to analyze the prognostic value of systemic inflammation and IR markers in women with breast cancer, as well as their distribution and ability to predict survival in different BMI subgroups.
## Baseline characteristics
After excluding 3 male breast cancer cases and 27 missing TNM stage data, a total of 514 women with breast cancer were included in our study. The detailed flow chart is showed in Fig. 1. Their mean age was 53.72 ± 10.87 years, and the population’s mean BMI was 24.36 kg/m2. Comparing the baseline differences between patients in different BMI groups (low BMI group, BMI < 24 kg/m2 vs. high BMI group, BMI ≥ 24 kg/m2), the age (54.65 vs. 52.75, $$P \leq 0.047$$), BMI (27.23 vs. 21.34, $P \leq 0.001$), CRP (3.02 vs. 2.78, $$P \leq 0.020$$), CAR (0.07 vs. 0.06, $$P \leq 0.029$$), TyG (4.55 vs. 4.51, $$P \leq 0.008$$), LHR (2.40 vs. 2.09, $P \leq 0.001$), TG/HDL-c (1.72 vs. 1.34, $P \leq 0.001$), and TC/HDL-c (4.15 vs. 3.61, $P \leq 0.001$) were all higher in the patients in high BMI group than those in low BMI group. However, the LCR (6408.0 vs. 5254.9, $$P \leq 0.029$$) was higher in the patients in the low BMI group than those in high BMI group. Table 1 shows the baseline characteristics of the women with breast cancer. The median follow-up time for patients was 43.1 (40.7–49.6) months, and the 5-year overall mortality rate was 70 ($18\%$), resulting in 41.4 mortality events per 1000 patient-year. Figure 1Flowchart of patient selection for this study. Table 1Baseline characteristics. VariablesOverallBMI < 24 (kg/m2)BMI ≥ 24(kg/m2)P value($$n = 514$$)($$n = 251$$)($$n = 263$$)Age (mean (SD))53.72 (10.87)52.75 (11.40)54.65 (10.28)0.047BMI (mean (SD))24.36 (3.80)21.34 (2.02)27.23 (2.70)< 0.001Tumor stage (%)0.312 I–II283 (55.1)132 (52.6)151 (57.4) III–IV231 (44.9)119 (47.4)112 (42.6)Surgery (%)420 (81.7)195 (77.7)225 (85.6)0.028Radiotherapy (%)27 (5.3)14 (5.6)13 (4.9)0.901Chemotherapy (%)328 (63.8)149 (59.4)179 (68.1)0.050Immunotherapy (%)43 (8.4)20 (8.0)23 (8.7)0.874KPS (mean (SD))90.60 (10.29)89.28 (11.57)91.86 (8.73)0.004Tumor metastasis (%)42 (8.2)19 (7.6)23 (8.7)0.745Family history of cancer (%)100 (19.5)40 (15.9)60 (22.8)0.063Diabetes (%)39 (7.6)19 (7.6)20 (7.6)1.000Hypertension (%)78 (15.2)26 (10.4)52 (19.8)0.004CHD (%)25 (4.9)8 (3.2)17 (6.5)0.128Lymphocyte, *109/L (mean (SD))1.62 (1.14)1.64 (1.48)1.60 (0.66)0.643Albumin (mean (SD))40.60 (4.80)40.60 (5.37)40.59 (4.20)0.975CRP (median (IQR))3.01 (3.61)2.78 (2.81)3.02 (3.64)0.020LCR (median (IQR))5806.5 (12,301.7)6408.0 (13,677.6)5254.9 (10,204.5)0.031CAR (median (IQR))0.07 (0.09)0.06 (0.07)0.07(0.09)0.029Glucose (mean (SD))5.85 (1.95)5.77 (2.14)5.93 (1.74)0.343TC (mean (SD))4.88 (1.55)4.76 (1.57)5.00 (1.52)0.081TG (mean (SD))1.79 (1.03)1.61 (1.03)1.96 (1.00)< 0.001HDL-c (mean (SD))1.32 (0.33)1.39 (0.36)1.25 (0.29)< 0.001LDL-c (mean (SD))2.83 (0.79)2.76 (0.83)2.89 (0.75)0.051TyG (mean (SD))4.53 (0.19)4.51 (0.20)4.55 (0.18)0.008LHR (mean (SD))2.25 (0.73)2.09 (0.71)2.40 (0.71)< 0.001TG/HDL-c (mean (SD))1.54 (1.18)1.34 (1.23)1.72 (1.10)< 0.001TC/HDL-c (mean (SD))3.89 (1.44)3.61 (1.33)4.15 (1.49)< 0.001TSF (mean (SD))22.09 (7.91)18.97 (7.20)25.06 (7.42)< 0.001SD standard deviation, IQR interquartile range, BMI body mass index, KPS karnofsky performance status, CHD coronary heart disease, CRP C-reactive protein, CAR C-reactive protein to albumin ratio, LCR lymphocyte to C-reactive protein ratio, TC total cholesterol, TG triglyceride, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol, TyG triglyceride-glucose index, LHR LDL-c/HDL-c ratio.
## Differences in the distribution of inflammation and IR markers in different BMI subgroups
The distribution curves for systemic inflammation-related indicators in different BMI subgroups showed that the CRP, CAR, and LCR values in the high BMI group were significantly higher than those in the low BMI group (All $P \leq 0.05$) (Fig. 2A–C). Similarly, we analyzed the differences in the distribution of different IR indicators in different BMI subgroups and found that TyG, LHR, TG/HDL-c, and TC/HDL-c were all highly distributed in the high BMI group compared with the low BMI subgroup patients (All $P \leq 0.05$) (Fig. 2D–G).Figure 2The distribution of systemic inflammatory indicators and IR makers stratified by BMI in women with breast cancer. ( A) CRP; (B) CAR; (C) LCR; (D) TyG; (E) LHR; (F) TG/HDL-c; (G) TC/HDL-c. Notes IR insulin resistance, BMI body mass index, CRP C-reactive protein, LCR lymphocyte to C-reactive protein ratio, CAR C-reactive protein to albumin ratio, TyG triglyceride-glucose index, LHR LDL-c/HDL-c ratio, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol.
## Prognostic AUC curves and survival analysis correlated with systemic inflammatory markers and IR markers
To select the optimal inflammatory index and IR index in female breast cancer, we drew the prognostic area under the curve (AUC) curves of inflammatory index and IR index, respectively. The results showed that the predictive ability of LCR and CAR was better than that of CRP among different inflammatory indicators, while among different IR indicators, TyG showed the worst predictive ability of prognosis, compared with LHR, TG/HDL-c, and TC/HDL-c (Fig. 3).Figure 3The prognostic AUC curves of systemic inflammatory indicators and IR makers in female breast cancer. ( A) Systemic inflammatory indicators of CRP, CAR, and LCR; (B) IR makers of TyG, LHR, TG/HDL-c, and TC/HDL-c. Notes AUC area under the curve, IR insulin resistance, CRP C-reactive protein, LCR lymphocyte to C-reactive protein ratio, CAR C-reactive protein to albumin ratio, TyG triglyceride-glucose index, LHR LDL-c/HDL-c ratio, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol.
The survival curves of CRP, CAR, and LCR in women with breast cancer showed that patients with high CRP, high CAR, or high LCR had a worse prognosis than patients with low CRP($$P \leq 0.0025$$), low CAR ($P \leq 0.001$), or low LCR ($P \leq 0.001$), respectively (Fig. 4A–C). In addition, survival curves showed that compared with patients with low TyG, low LHR, or low TG/HDL-c, patients with high TyG ($$P \leq 0.03$$), high LHR ($$P \leq 0.017$$), or high TG/HDL-c ($$P \leq 0.018$$) had worse prognosis, respectively. However, there was no significant difference in survival between patients with low TC/HDL-c or high TC/HDL-c ($$P \leq 0.085$$) (Fig. 5A–D).Figure 4The Kaplan–Meier survival curves of systemic inflammatory indicators in women with breast cancer. ( A) CRP; (B) CAR; (C) LCR. Notes CRP C-reactive protein, LCR lymphocyte to C-reactive protein ratio, CAR C-reactive protein to albumin ratio. Figure 5The Kaplan–Meier survival curves of IR makers in women with breast cancer. ( A) TyG; (B) LHR; (C) TG/HDL-c; (D) TC/HDL-c. Notes IR insulin resistance, TyG triglyceride-glucose index, LHR LDL-c/HDL-c ratio, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol.
Multivariate survival analysis of systemic inflammatory indicators in women with breast cancer indicated that patients with high CRP [model 4: HR ($95\%$ CI) = 2.21 (1.24–3.94), $$P \leq 0.007$$] had a shorter OS than patients with low CRP, patients with high CAR [model 4: HR ($95\%$ CI) = 2.56 (1.46–4.47), $$P \leq 0.001$$] had a shorter OS than those with low CAR, and patients with high LCR [model 4: HR ($95\%$ CI) = 2.43 (1.47–4.02), $$P \leq 0.001$$] had a shorter OS than patients with low LCR (Table 2).Table 2Univariate and multivariate analysis. VariablesOS (model 0)OS (model 1)OS (model 2)OS (model 3)OS (model 4)Crude HR ($95\%$ CI)Crude PAdjusted HR ($95\%$ CI)Adjusted PAdjusted HR ($95\%$ CI)Adjusted PAdjusted HR ($95\%$ CI)Adjusted PAdjusted HR ($95\%$ CI)Adjusted PCRP As continues1.12 (0.91–1.37)0.2881.12 (0.91–1.37)0.2901.15 (0.92–1.44)0.2311.20 (0.96–1.50)0.1151.20 (0.96–1.51)0.109As binary CRP ≤ 10RefRefRefRefRef CRP > 103.37 (1.93–5.88) < 0.0012.99 (1.73–5.16) < 0.0011.8 (1.03–3.14)0.0392.21 (1.24–3.94)0.0072.21 (1.24–3.94)0.007As tertiles T1 (< 0.170)RefRefRefRefRef T2 (0.170–0.528)0.98 (0.52–1.84)0.9460.96 (0.51–1.82)0.9060.77 (0.41–1.45)0.4160.77 (0.40–1.48)0.4410.77 (0.40–1.48)0.439 T3 (> 0.528)1.84 (1.07–3.19)0.0291.81 (1.04–3.15)0.0371.18 (0.67–2.06)0.5651.25 (0.70–2.23)0.4571.26 (0.71–2.26)0.433 P for trend0.0240.0290.4780.3770.358 CARRefRefRefRefRef As continues1.13 (0.91–1.42)0.2751.13 (0.91–1.42)0.2731.14 (0.90–1.45)0.2851.18 (0.93–1.50)0.1791.18 (0.93–1.51)0.169As binary CAR ≤ 0.24 CAR > 0.243.2 (1.89–5.42)< 0.0013.21 (1.9–5.43)< 0.0012.05 (1.2–3.52)0.0092.56 (1.46–4.47)0.0012.56 (1.46–4.47)0.001As tertiles T1 (< 0.034)RefRefRefRefRef T2 (0.034–0.087)1.07 (0.57–2.00)0.841.05 (0.55–1.97)0.890.78 (0.41–1.46)0.4330.78 (0.41–1.50)0.4610.78 (0.41–1.50)0.462 T3 (> 0.087)1.98 (1.13–3.45)0.0161.94 (1.11–3.41)0.0211.15 (0.65–2.04)0.6351.27 (0.70–2.30)0.4321.28 (0.71–2.32)0.415 P for trend0.0130.0160.5260.3460.332LCR As continues1.10 (0.52–2.30)0.8081.12 (0.54–2.35)0.7621.45 (0.66–3.16)0.3571.48 (0.69–3.15)0.3151.46 (0.69–3.12)0.324As binary LCR ≤ 2321.9RefRefRefRefRef LCR > 2321.93.46 (2.17–5.52)< 0.0013.44 (2.15–5.50)< 0.0012.03 (1.25–3.32)0.0042.44 (1.47–4.03)0.0012.43 (1.47–4.02)0.001As tertiles T1 (> 10,608.11)RefRefRefRefRef T2 (4000–10,608.11)1.09 (0.57–2.1)0.7941.08 (0.56–2.08)0.8190.9 (0.47–1.73)0.7450.9 (0.46–1.78)0.7650.91 (0.46–1.79)0.782 T3 (< 4000)2.43 (1.38–4.29)0.0022.4 (1.36–4.26)0.0031.34 (0.75–2.4)0.3221.47 (0.81–2.7)0.2091.47 (0.81–2.7)0.209 P for trend0.0010.0020.2610.1590.161TyG As continues1.08 (0.85–1.37)0.5341.07 (0.84–1.36)0.5871.03 (0.81–1.31)0.840.97 (0.75–1.25)0.7890.97 (0.75–1.25)0.805As binary TyG ≤ 4.72RefRefRefRefRef TyG > 4.721.89 (1.06–3.40)0.0321.86 (1.03–3.35)0.041.56 (0.85–2.87)0.1491.38 (0.71–2.68)0.3431.42 (0.73–2.78)0.302As tertiles T1 (< 4.459)RefRefRefRefRef T2 (4.459–4.584)0.81 (0.45–1.48)0.4970.79 (0.43–1.45)0.4541.01 (0.55–1.84)0.9761.00 (0.54–1.86)0.9931.00 (0.54–1.87)0.993 T3 (> 4.584)1.15 (0.67–2.00)0.6071.12 (0.64–1.95)0.6831.03 (0.59–1.8)0.9070.94 (0.51–1.74)0.8350.94 (0.51–1.74)0.850 P for trend0.5710.6360.9060.8330.848LHR As continues1.18 (0.94–1.47)0.1591.16 (0.92–1.47)0.1961.05 (0.83–1.33)0.7081.04 (0.82–1.33)0.7351.04 (0.82–1.33)0.740 As binary LHR ≤ 3.20RefRefRefRefRef LHR > 3.202.18 (1.17–4.06)0.0142.16 (1.16–4.02)0.0162.02 (1.08–3.79)0.0282.42 (1.27–4.63)0.0082.40 (1.25–4.61)0.008As tertiles T1 (< 1.925)RefRefRefRefRef T2 (1.925–2.554)0.91 (0.51–1.63)0.7520.90 (0.50–1.61)0.7190.84 (0.47–1.51)0.5520.89 (0.49–1.62)0.6960.87 (0.47–1.59)0.649 T3 (> 2.554)1.13 (0.65–1.98)0.6591.10 (0.62–1.93)0.7510.93 (0.53–1.64)0.7960.86 (0.48–1.54)0.6120.86 (0.48–1.54)0.607 P for trend0.6460.7380.8130.6170.616TG/HDL As continues0.97 (0.77–1.21)0.7710.95 (0.76–1.20)0.6930.99 (0.77–1.28)0.9520.91 (0.71–1.17)0.4780.92 (0.71–1.18)0.495As binary TG/HDL-c ≤ 0.60RefRefRefRefRef TG/HDL-c > 0.604.31 (1.36–13.69)0.0134.25 (1.33–13.56)0.0153.84 (1.20–12.32)0.243.50 (1.08–11.32)0.3703.51 (1.08–11.36)0.036As tertiles T1 (< 0.912)RefRefRefRefRef T2 (0.912–1.621)1.64 (0.91–2.96)0.0981.61 (0.88–2.94)0.1231.24 (0.68–2.29)0.4841.31 (0.71–2.44)0.3881.33 (0.71–2.47)0.372 T3 (> 1.621)1.33 (0.73–2.44)0.3561.30 (0.69–2.43)0.4151.13 (0.60–2.10)0.7051.10 (0.58–2.09)0.7691.12 (0.59–2.14)0.731 P for trend0.3930.4840.7570.8190.781TC/HDL As Continues1.14 (0.90–1.46)0.2791.13 (0.88–1.45)0.3381.10 (0.83–1.46)0.5081.04 (0.79–1.36)0.7821.04 (0.79–1.36)0.785As binary TC/HDL-c ≤ 3.81RefRefRefRefRef TC/HDL-c > 3.811.75 (1.09–2.82)0.021.73 (1.07–2.81)0.0251.38 (0.85–2.23)0.1951.38 (0.84–2.26)0.2041.40 (0.85–2.30)0.185As tertiles T1 (< 3.335)RefRefRefRefRef T2 (3.335–4.158)1.13 (0.63–2.03)0.6821.11 (0.61–2.00)0.7351.03 (0.57–1.86)0.9251.02 (0.56–1.86)0.9471.02 (0.56–1.87)0.942 T3 (> 4.158)1.27 (0.72–2.24)0.4141.22 (0.68–2.20)0.4981.08 (0.61–1.93)0.7871.04 (0.57–1.88)0.9091.05 (0.58–1.90)0.885 P for trend0.4130.4970.7850.9090.885CRP C-reactive protein, CAR CRP/Albumin ratio, LCR lymphocyte/CRP ratio, TyG fasting triglyceride glucose index, TG triglyceride, TC total cholesterol, LHR LDL-c/HDL-c ratio, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol, HR hazards ratio, CI confidence interval, BMI body mass index, KPS karnofsky performance status, TSF triceps skinfold thickness.aModel 0: Unadjusted.bModel 1: Adjusted for BMI.cModel 2: Adjusted for age, BMI and tumor stage.dModel 3: Adjusted for age, tumor stage, BMI, KPS, surgery, chemotherapy, radiotherapy, immunotherapy, family history of cancer, tumor metastasis, diabetes, hypertension, and coronary heart disease.eModel 4: Adjusted for age, tumor stage, BMI, KPS, surgery, chemotherapy, radiotherapy, immunotherapy, family history of cancer, tumor metastasis, diabetes, hypertension, coronary heart disease, and TSF.
Multivariate survival analysis of the IR index in women with breast cancer indicated that patients with high LHR [model 4: HR ($95\%$ CI) = 2.40 (1.25–4.61), $$P \leq 0.008$$] had a shorter OS than those with low LHR and patients with high TG/ HDL-c [model 4: HR ($95\%$ CI) = 3.51 (1.08–11.36), $$P \leq 0.036$$] had a shorter OS than those with low TG/HDL-c. However, TyG [model 4: HR ($95\%$ CI) = 1.42 (0.73–2.78), $$P \leq 0.302$$] and TC/HDL-c [model 4: HR ($95\%$ CI) = 1.40 (0.85–2.30), $$P \leq 0.185$$] were not significant survival predictors in women with breast cancer (Table 2).
## Survival analysis stratified by different BMI groups
We analyzed the prognostic value of systemic inflammatory markers and IR markers in different BMI subgroups. In the BMI < 24 kg/m2 subgroup, we observed that all markers did not show significant prognostic value (All $P \leq 0.05$). In the BMI ≥ 24 kg/m2 subgroup, patients with high CRP [Adjusted HR ($95\%$ CI) = 2.39 (1.00–5.71), $$P \leq 0.049$$], high CAR [Adjusted HR ($95\%$ CI) = 2.85 (1.23–6.60), $$P \leq 0.014$$], high LCR [Adjusted HR ($95\%$ CI) = 4.32 (2.06–9.06), $P \leq 0.001$], high TyG [Adjusted HR ($95\%$ CI) = 2.87 (1.20–6.85), $$P \leq 0.017$$], or high LHR [Adjusted HR ($95\%$ CI) = 2.91 (1.20–7.06), $$P \leq 0.018$$] predicted worse prognoses, while TG/HDL-c and TC/HDL-c did not show significant prognostic value (All $P \leq 0.05$) (Table 3).Table 3Survival analysis stratified by different BMI groups. VariablesBMI < 24 (kg/m2)*P valueBMI > 24 (kg/m2)*P valueCRP CRP ≤ 1011 CRP > 101.81 (0.78–4.23)0.1702.39 (1.00–5.71)0.049CAR CAR ≤ 0.24 CAR > 0.242.06 (0.90–4.73)0.0892.85 (1.23–6.60)0.014LCR LCR ≤ 2321.9 LCR > 2321.91.44 (0.67–3.12)0.3504.32 (2.06–9.06)< 0.001TyG TyG ≤ 4.7211 TyG > 4.720.55 (0.16–1.94)0.3502.87 (1.20–6.85)0.017LHR LHR ≤ 3.2011 LHR > 3.202.06 (0.70–6.10)0.1922.91 (1.20–7.06)0.018TGH TG/HDL-c ≤ 0.6011 TG/HDL-c > 0.603.78 (0.87–16.44)0.0772.61 (0.34–19.92)0.355TCH TC/HDL-c ≤ 3.8111 TC/HDL-c > 3.810.62 (0.26–1.48)0.2851.58 (0.77–3.25)0.211CRP C-reactive protein, LHR LDL-c/HDL-c ratio, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol, HR hazards ratio, CI confidence interval, BMI body mass index, KPS karnofsky performance status, TSF triceps skinfold thickness.*Adjusted for age, tumor stage, KPS, surgery, chemotherapy, radiotherapy, immunotherapy, family history of cancer, tumor metastasis, diabetes, hypertension, coronary heart disease, and TSF.
## Combined analysis of prognostic systemic inflammatory indicators and IR index
We performed a combined survival analysis with the prognostic systemic inflammatory index and IR index. In all patients, CAR combined with LHR or LCR combined with LHR predicted a longer OS in women with breast cancer. The prognosis of patients in the low CAR and high LHR or high CAR and low LHR group [Adjusted HR ($95\%$ CI) = 2.21 (1.27–3.87), $$P \leq 0.005$$] and the high CAR and high LHR group [Adjusted HR ($95\%$ CI) = 3.91 (1.56–9.81), $$P \leq 0.004$$] was worse than in patients in the low CAR and low LHR groups. The prognosis of patients in the high LCR and high LHR or low LCR and low LHR group [Adjusted HR ($95\%$ CI) = 2.30 (1.36–3.87), $$P \leq 0.002$$] and the low LCR and high LHR group [Adjusted HR ($95\%$ CI) = 3.89 (1.65–9.21), $$P \leq 0.002$$] was worse than in patients in the high LCR and low LHR groups. However, no prognostic information was generated by the other combinations. In addition, when we performed a combined survival analysis in different BMI subgroups, we only observed a significant survival difference in the combined analysis of LCR and LHR in the high BMI subgroup. The prognosis of patients in the high LCR and high LHR or low LCR and low LHR group [Adjusted HR ($95\%$ CI) = 3.61 (1.69–7.69), $$P \leq 0.001$$] and the low LCR and high LHR group [Adjusted HR ($95\%$ CI) = 7.79 (2.42–25.11), $$P \leq 0.001$$] was worse than that of the patients in the low LCR and low LHR groups (Table 4).Table 4Combined survival analysis of prognostic systemic and IR indicators. VariablesAll patients#BMI < 24 (kg/m2)*BMI ≥ 24 (kg/m2)*Adjusted HR ($95\%$ CI)P valueAdjusted HR ($95\%$ CI)P valueAdjusted HR ($95\%$ CI)P valueCRP and LHR Low CRP and low LHR111 Low CRP and high LHR or high CRP and low LHR2.48 (1.44–4.27)0.0012.23 (0.95–5.27)0.0662.42 (1.15–5.09)0.020 High CRP and high LHR2.56 (0.88–7.51)0.0861.93 (0.39–9.54)0.4194.75 (0.97–23.3)0.055 P for trend0.0030.1390.005CAR and LHR Low CAR and low LHR111 Low CAR and high LHR or high CAR and low LHR2.21 (1.27–3.87)0.0052.05 (0.85–4.94)0.1112.13 (0.99–4.58)0.054 High CAR and high LHR3.91 (1.56–9.81)0.0042.68 (0.66–10.88)0.1688.80 (2.19–35.31)0.002 P for trend< 0.0010.0560.001LCR and LHR High LCR and low LHR111 Low LCR and low LHR or high LCR and high LHR2.30 (1.36–3.87)0.0021.44 (0.64–3.25)0.3823.61 (1.69–7.69)0.001 Low LCR and high LHR3.89 (1.65–9.21)0.0022.32 (0.59–9.14)0.2317.79 (2.42–25.11)0.001 P for trend< 0.0010.175< 0.001CRP and TG/HDL-c Low CRP and Low TG/HDL-c111 Low CRP and high TG/HDL-c or high CRP and low TG/HDL-c2.64 (0.81–8.62)0.1092.55 (0.57–11.52)0.2232.30 (0.3–17.71)0.424 High CRP and high TG/HDL-c5.91 (1.69–20.61)0.0055.33 (1.07–26.69)0.0425.44 (0.62–47.42)0.126 P for trend0.0010.0210.032CAR and TG/HDL-c Low CAR and low TG/HDL-c111 Low CAR and high TG/HDL-c or high CAR and low TG/HDL-c2.52 (0.77–8.25)0.1272.46 (0.54–11.13)0.2422.24 (0.29–17.32)0.438 High CAR and high TG/HDL-c6.50 (1.88–22.44)0.0035.84 (1.18–28.92)0.0316.32 (0.73–54.51)0.094 P for trend< 0.0010.0110.011LCR and TG/HDL-c High LCR and low TG/HDL-c111 Low LCR and low TG/HDL-c or high LCR and high TG/HDL-c2.18 (0.66–7.17)0.2002.34 (0.51–10.7)0.2731.56 (0.2–12.24)0.673 Low LCR and high TG/HDL-c5.55 (1.64–18.72)0.0063.79 (0.79–18.11)0.0956.69 (0.84–53.17)0.072 P for trend< 0.0010.062< 0.001CRP C-reactive protein, LHR LDL-c/HDL-c ratio, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol, HR hazards ratio, CI confidence interval, BMI body mass index, KPS karnofsky performance status, TSF triceps skinfold thickness.*Adjusted for age, tumor stage, KPS, surgery, chemotherapy, radiotherapy, immunotherapy, family history of cancer, tumor metastasis, diabetes, hypertension, coronary heart disease, and TSF.
## Discussion
In this study, we found that the levels of inflammation (CRP, CAR, and LCR) and IR (TyG, LHR, TG/HDL-c, and TC/HDL-c) in breast cancer patients with BMI ≥ 24 kg/m2 were significantly higher than those in patients with BMI < 24 kg/m2. In other words, inflammation and IR levels in overweight or obese patients are high. The expression of adipocytokines in human adipose tissue and their corresponding circulating concentrations are influenced by human fat mass. In obese patients, there was also a positive correlation between adipocyte TNF-α expression and plasma TNF-α concentration with BMI40. Plasma IL-6 and CRP concentrations were also positively correlated with BMI23,41. Obesity is a common cause of chronic inflammation, and white adipose tissue (WAT) in obese patients is infiltrated by immune cells, including macrophages and lymphocytes, at the systemic and tissue level. WAT inflammation is associated with increased circulating levels of CRP and IL-616. Obesity is a well-established risk factor for IR and type II diabetes. With the rising prevalence of obesity, an increasing number of patients at cancer diagnosis are overweight or obese and have impaired glycemic control. Obesity and excess adipose tissue lead to increased production of free fatty acids, leptin, and cytokines, and these metabolic abnormalities are associated with decreased physical activity and increased triglycerides, leading to hyperinsulinemia and IR42.
We also examined the relationship between systemic inflammation and IR and breast cancer survival. Studies have shown an increased risk of breast cancer in obese postmenopausal women, and it has been hypothesized that circulating estrogen levels may be elevated in obese postmenopausal women43. Therefore, considering these potential interference factors, we separated patients into different age groups and adjusted the survival analyses to reduce the interference caused by estrogen levels in different age groups. We found that elevated systemic inflammatory markers (CRP, CAR, and LCR) were all significantly associated with reduced OS in breast cancer patients. Similarly, we observed a significant association between increased IR markers (LHR and TG/ HDL-c) and decreased OS in breast cancer patients.
Pierce et al. analyzed the prognostic value of inflammatory markers in women with stage 0 to IIIA breast cancer in a multicenter, prospective, cohort study and found that CRP was associated with poor prognosis in women with breast cancer compared with the highest and lowest tertiles [HR 2.27; $95\%$ CI 1.27–4.08]26. Gunter et al. and Albuquerque et al. also found that CRP levels were positively associated with breast cancer risk27,28. Zhou et al. used propensity score matching to estimate the prognostic role of CAR in non-metastatic breast cancer patients and found that elevated CAR levels were associated with increased age, postmenopausal status, and a higher risk of recurrence or death in breast cancer patients. Elevated CAR was an independent risk factor for long-term prognosis, predicting decreased disease-free survival [HR 2.225; $$P \leq 0.024$$] and OS [HR 9.189; $$P \leq 0.003$$] of breast cancer patients29. Chen et al. found that preoperative CAR could be an important independent prognostic marker for HER2-negative, luminal breast cancer, and elevated CAR was associated with poorer disease-free survival and cancer-specific survival30. In this study, for the first time, we found that LCR could be used as an independent prognostic marker in breast cancer patients. Previous studies reported that LCR is associated with poor prognosis in other tumors, such as colorectal cancer33, gastric cancer32, and hepatocellular carcinoma44. As for IR prognostic indicators, previous studies have shown that LHR is associated with poor prognosis in colorectal cancer45,46 and gastric cancer47. Dai et al. analyzed the relationship between TG/ HDL-c and prognosis in triple-negative breast cancer patients and found that patients with high TG/ HDL-c was associated with poor OS [HR: 1.935; $95\%$ CI 1.032–3.629]48. Similar results showed that TG/ HDL-c was associated with poor prognosis in other cancers, including in endometrial cancer49 and gastric cancer50.
We observed markers of inflammation and IR in different BMI subgroups and found that LCR could predict survival in different BMI subgroups. And CRP, CAR, TyG, LHR predict the prognosis of patients within the high BMI subgroup. The results of the combined survival analyses showed that the inflammatory insulin combination of LCR&LHR and CAR&LHR could differentiate the prognosis of breast cancer patients. Especially, LCR&LHR could also significantly differentiate the prognosis of patients in the high BMI subgroup. Furthermore, the observation that breast WAT inflammation predicts a poorer clinical course in breast cancer patients is consistent with earlier reports showing that TNF-α, IL-1beta, IL-6 and CRP promote tumor growth in a mouse model of obesity and elevated levels of IL-6 and CRP were associated with the development and progression with female breast cancer16. Inflammation and IR are closely related. The IR state that develops with increased obesity is associated with activation of inflammatory responses in different organ sites, including adipose tissue, liver, and skeletal muscle, which increases secretion and systemic levels of proinflammatory cytokines51. Some adipocytokines help regulate insulin action and are associated with IR syndromes52. Leptin interferes with insulin signaling, and in type II diabetes, plasma leptin levels correlate with the degree of IR, a relationship independent of BMI and body fat mass53,54. Thus, the IR syndrome is associated with hyperleptinemia and hyperinsulinemia55, which allows endocrine hyperactivity of these proteins at target sites, including mammary epithelial tissue and vascular endothelial cells. Adipose tissue TNF-α expression was also positively correlated with plasma insulin concentrations56, and increased adipocyte secretion of TNF-α was associated with decreased insulin sensitivity in obese individuals41. In abdominal obesity, high circulating TNF-α levels are associated with hyperinsulinemia and IR57. IR is also associated with human adipose tissue-derived IL-641. Adipose tissue has biological activities that regulate appetite, inflammation, insulin sensitivity, fat metabolism, and energy balance58. Excessive adipose tissue will lead to the production of inflammatory cytokines and the upregulation of nuclear factor-κB, leading to increased nitric oxide and reactive oxygen species, resulting in IR, excess glucose, and increased free fatty acid, thereby further spreading inflammation59.
Our study has several strengths. First, this is a prospective, cohort study of women with breast cancer based on a multi-medical center trial to analyze the prognosis of different systemic inflammation and IR markers. Second, our study analyzed the inflammation and IR levels in different BMI subgroups and examined high-inflammation and high-IR status in overweight or obese female breast cancer patients to identify the best markers of inflammation and IR. Our study also has some limitations. First, we only collected a fasting blood sample and thus, cross-sectional data. Longitudinal data is needed for a patient's observation of inflammation and IR. Second, although we consider the effect of hormonal levels in patients and make prognostic adjustments for different ages, we still need to collect relevant data. Third, different pathological types of breast cancer may cause heterogeneity, and the results of more pathological types need to be included. Fourth, our IR-related metric is only a surrogate metric, and we cannot deny its simplicity and feasibility, but the assessment of patients' IR status still needs to be done.
## Conclusion
In conclusion, our data showed that higher CRP, CAR, LCR, LHR, and TG/ HDL-c were associated with increased risk in women with breast cancer. Elevated BMI showed the higher inflammation and IR levels in women with breast cancer. The combination of CAR and LHR or LCR and LHR could significantly predict the prognosis of women with breast cancer, while the combination of LCR and LHR can significantly predict prognosis in those patients with overweight or obese patients.
## Study population
The data collected in this study from women with breast cancer were obtained from a prospective, multi-medical center-based cancer population study in China between 2013 and 2021. The hospitals included Fujian Cancer Hospital, Bethune First Hospital of Jilin University, Zhejiang Cancer Hospital, First Affiliated Hospital of Sun Yat-Sen University, Chongqing Daping Hospital, and Chongqing Third People's Hospital. The inclusion criteria for this study were: 1. Female patients aged not less than 18 years; 2. Pathologically diagnosed with breast cancer; and 3. Clearly conscious and able to communicate autonomously. There were no strict exclusion criteria. The current study complied with the Declaration of Helsinki, was approved by the Human Research Committees at the various medical centers, and all participants provided informed consent.
## Anthropometric and laboratory measurements
At the start of the study, participants' demographic information, medical and family history, and quality of life assessment were collected through questionnaires administered by trained investigators. All research centers, which participated in your study, had the same standards of biomarkers laboratory testing. Baseline clinical characteristics collected from patients included age, body mass index (BMI), comorbidities (diabetes, yes/no; hypertension, yes/no; and coronary heart disease, yes/no), tumor-related information (family history of cancer, tumor stage, surgery, yes/no; radiation therapy, yes/no; chemotherapy, yes/no; immunotherapy, yes/no; and tumor metastasis, yes/no), Karnofsky performance status (KPS), triceps skinfold thickness (TSF), and laboratory test indicators [C-reactive protein (CRP), fasting blood glucose (FBG), triglyceride (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-c), and high-density lipoprotein cholesterol (HDL-c)].
The patient's body measurements were obtained by clinicians or nurses, height and weight were measured while the patients were wearing light hospital gowns and socks, and TSF was obtained by taking the average of three measurements with a skinfold caliper. BMI was defined as the ratio of weight (kg) to height squared (m2). Blood samples from patients were collected for analysis in the laboratory within 48 h prior to admission after patients had fasted for at least 8 h prior to sample collection. The index of CAR and LCR were calculated by: CRP/albumin and Lymphocyte/CRP, respectively. The TyG index was defined as Ln [TC (mg/dl) * FBG (mg/dl)]/2. The ratios LDL-c/HDL-c (LHR), TG/HDL-c, and TC/HDL-c were defined as: LDL-c/HDL-c, TG/HDL-c, and TC/HDL-c, respectively.
## Outcomes
Overall survival (OS), representing the study endpoint, was calculated from the date of diagnosis of cancer until death or last follow-up. Follow-up of patients was completed by follow-up staff.
## Statistical analyses
Data are shown as percentages, mean ± standard deviation, or median ± interquartile interval. Baseline characteristics of obese and nonobese populations were compared using the chi-square test and Fisher's exact test for categorical variables and a t-test for continuous normal distribution variables (Wilcoxon test for non-parametric variables). Cutoff values were generated by largest selected rank statistical analysis method for continuous data (see Supplementary Fig. S1 online).
The prognostic AUC curves were performed to selcet the optimal inflammation index and IR index. The survival curves were calculated using the Kaplan–Meier method, and the level of significance was assessed using the log-rank test. Associations between prognostic factors and OS were examined using multivariable Cox proportional hazards regression models, and results were reported as hazard ratios (HRs) and $95\%$ confidence intervals ($95\%$ CIs). We assessed confounding covariates by adding each covariate sequentially to the base model. Model 0: unadjusted; model 1: adjusted for BMI; model 2: adjusted for age, tumor stage, and BMI; model 3: adjusted for age, tumor stage, BMI, KPS, surgery, chemotherapy, radiotherapy, immunotherapy, family history of cancer, tumor metastasis, diabetes, hypertension, and coronary heart disease.; model 4: adjusted for age, tumor stage, BMI, KPS, surgery, chemotherapy, radiotherapy, immunotherapy, family history of cancer, tumor metastasis, diabetes, hypertension, coronary heart disease, and TSF.
All P values were two-sided. P values less than 0.05 were considered statistically significant. All statistical analyses were performed using the R software version 4.1.1.
## Ethics approval
This study followed the Helsinki declaration. All participants signed an informed consent form, and this study was approved by the Institutional Review Board of each hospital (Registration number: ChiCTR1800020329).
## Supplementary Information
Supplementary Figure S1. The online version contains supplementary material available at 10.1038/s41598-023-31450-w.
## References
1. Siegel RL, Miller KD, Fuchs HE, Jemal A. **Cancer statistics, 2022**. *CA Cancer J. Clin.* (2022) **72** 7-33. DOI: 10.3322/caac.21708
2. Chen W. **Cancer statistics in China, 2015**. *CA Cancer J. Clin.* (2016) **66** 115-132. DOI: 10.3322/caac.21338
3. Lee DY. **Impact of systemic inflammation on the relationship between insulin resistance and all-cause and cancer-related mortality**. *Metabolism* (2018) **81** 52-62. DOI: 10.1016/j.metabol.2017.11.014
4. Korniluk A, Koper O, Kemona H, Dymicka-Piekarska V. **From inflammation to cancer**. *Ir. J. Med. Sci.* (2017) **186** 57-62. DOI: 10.1007/s11845-016-1464-0
5. Argiles JM, Lopez-Soriano FJ, Busquets S. **Counteracting inflammation: A promising therapy in cachexia**. *Crit. Rev. Oncog.* (2012) **17** 253-262. DOI: 10.1615/CritRevOncog.v17.i3.30
6. Fearon KC. **Pancreatic cancer as a model: Inflammatory mediators, acute-phase response, and cancer cachexia**. *World J. Surg.* (1999) **23** 584-588. DOI: 10.1007/PL00012351
7. Argiles JM, Lopez-Soriano FJ. **Insulin and cancer (Review)**. *Int. J. Oncol.* (2001) **18** 683-687. PMID: 11251161
8. Tayek JA. **A review of cancer cachexia and abnormal glucose metabolism in humans with cancer**. *J. Am. Coll. Nutr.* (1992) **11** 445-456. DOI: 10.1080/07315724.1992.10718249
9. Glicksman AS, Rawson RW. **Diabetes and altered carbohydrate metabolism in patients with cancer**. *Cancer* (1956) **9** 1127-1134. DOI: 10.1002/1097-0142(195611/12)9:6<1127::AID-CNCR2820090610>3.0.CO;2-4
10. Dev R, Bruera E, Dalal S. **Insulin resistance and body composition in cancer patients**. *Ann. Oncol.* (2018) **29** ii18-ii26. DOI: 10.1093/annonc/mdx815
11. Fonseca G, Farkas J, Dora E, von Haehling S, Lainscak M. **Cancer cachexia and related metabolic dysfunction**. *Int. J. Mol. Sci.* (2020) **21** 2321. DOI: 10.3390/ijms21072321
12. Formiguera X, Canton A. **Obesity: Epidemiology and clinical aspects**. *Best Pract. Res. Clin. Gastroenterol.* (2004) **18** 1125-1146. DOI: 10.1016/S1521-6918(04)00091-5
13. Finucane MM. **National, regional, and global trends in body-mass index since 1980: Systematic analysis of health examination surveys and epidemiological studies with 960 country-years and 9.1 million participants**. *Lancet.* (2011) **377** 557-567. DOI: 10.1016/S0140-6736(10)62037-5
14. Duggan C. **Associations of insulin resistance and adiponectin with mortality in women with breast cancer**. *J. Clin. Oncol.* (2011) **29** 32-39. DOI: 10.1200/JCO.2009.26.4473
15. Majed B, Moreau T, Asselain B. **Overweight, obesity and breast cancer prognosis: Optimal body size indicator cut-points**. *Breast Cancer Res. Treat.* (2009) **115** 193-203. DOI: 10.1007/s10549-008-0065-7
16. Iyengar NM, Gucalp A, Dannenberg AJ, Hudis CA. **Obesity and cancer mechanisms: Tumor microenvironment and inflammation**. *J. Clin. Oncol.* (2016) **34** 4270-4276. DOI: 10.1200/JCO.2016.67.4283
17. van Kruijsdijk RC, van der Wall E, Visseren FL. **Obesity and cancer: The role of dysfunctional adipose tissue**. *Cancer Epidemiol. Biomark. Prev.* (2009) **18** 2569-2578. DOI: 10.1158/1055-9965.EPI-09-0372
18. Kershaw EE, Flier JS. **Adipose tissue as an endocrine organ**. *J. Clin. Endocrinol. Metab.* (2004) **89** 2548-2556. DOI: 10.1210/jc.2004-0395
19. Sartipy P, Loskutoff DJ. **Monocyte chemoattractant protein 1 in obesity and insulin resistance**. *Proc. Natl. Acad. Sci. USA* (2003) **100** 7265-7270. DOI: 10.1073/pnas.1133870100
20. Xu H. **Chronic inflammation in fat plays a crucial role in the development of obesity-related insulin resistance**. *J. Clin. Invest.* (2003) **112** 1821-1830. DOI: 10.1172/JCI200319451
21. Cancello R. **Reduction of macrophage infiltration and chemoattractant gene expression changes in white adipose tissue of morbidly obese subjects after surgery-induced weight loss**. *Diabetes* (2005) **54** 2277-2286. DOI: 10.2337/diabetes.54.8.2277
22. Wang Z, Nakayama T. **Inflammation, a link between obesity and cardiovascular disease**. *Mediat. Inflamm.* (2010) **2010** 535918. DOI: 10.1155/2010/535918
23. Amin MN. **How the association between obesity and inflammation may lead to insulin resistance and cancer**. *Diabetes Metab. Syndr.* (2019) **13** 1213-1224. DOI: 10.1016/j.dsx.2019.01.041
24. Guo Q. **IGF-I CA19 repeat polymorphisms and cancer risk: A meta-analysis**. *Int. J. Clin. Exp. Med.* (2015) **8** 20596-20602. PMID: 26884978
25. Leung KC, Doyle N, Ballesteros M, Waters MJ, Ho KK. **Insulin regulation of human hepatic growth hormone receptors: Divergent effects on biosynthesis and surface translocation**. *J. Clin. Endocrinol. Metab.* (2000) **85** 4712-4720. PMID: 11134133
26. Pierce BL. **Elevated biomarkers of inflammation are associated with reduced survival among breast cancer patients**. *J. Clin. Oncol.* (2009) **27** 3437-3444. DOI: 10.1200/JCO.2008.18.9068
27. Albuquerque KV. **Pre-treatment serum levels of tumour markers in metastatic breast cancer: A prospective assessment of their role in predicting response to therapy and survival**. *Eur. J. Surg. Oncol.* (1995) **21** 504-509. DOI: 10.1016/S0748-7983(95)96935-7
28. Gunter MJ. **Circulating adipokines and inflammatory markers and postmenopausal breast cancer risk**. *J. Natl. Cancer Inst.* (2015) **107** djv169. DOI: 10.1093/jnci/djv169
29. Zhou L. **A retrospective propensity score matched study of the preoperative C-reactive protein to albumin ratio and prognosis in patients with resectable non-metastatic breast cancer**. *Med. Sci. Monit.* (2019) **25** 4342-4352. DOI: 10.12659/MSM.913684
30. Chen F. **Prognostic significance of neutrophil-to-lymphocyte ratio and C-reactive protein/albumin ratio in luminal breast cancers with HER2-negativity**. *Front. Oncol.* (2022) **12** 845935. DOI: 10.3389/fonc.2022.845935
31. Zhang X. **The nutrition-inflammation marker enhances prognostic value to ECOG performance status in overweight or obese patients with cancer**. *JPEN J. Parenter Enter. Nutr.* (2022) **47** 109-119. DOI: 10.1002/jpen.2407
32. Okugawa Y. **Lymphocyte-to-C-reactive protein ratio and score are clinically feasible nutrition-inflammation markers of outcome in patients with gastric cancer**. *Clin. Nutr.* (2020) **39** 1209-1217. DOI: 10.1016/j.clnu.2019.05.009
33. Okugawa Y. **Lymphocyte-C-reactive protein ratio as promising new marker for predicting surgical and oncological outcomes in colorectal cancer**. *Ann. Surg.* (2020) **272** 342-351. DOI: 10.1097/SLA.0000000000003239
34. Goodwin PJ. **Fasting insulin and outcome in early-stage breast cancer: Results of a prospective cohort study**. *J. Clin. Oncol.* (2002) **20** 42-51. DOI: 10.1200/JCO.2002.20.1.42
35. Pasanisi P. **Metabolic syndrome as a prognostic factor for breast cancer recurrences**. *Int. J. Cancer.* (2006) **119** 236-238. DOI: 10.1002/ijc.21812
36. DeFronzo RA, Tobin JD, Andres R. **Glucose clamp technique: A method for quantifying insulin secretion and resistance**. *Am. J. Physiol.* (1979) **237** E214-223. PMID: 382871
37. Fritz J. **The triglyceride-glucose index as a measure of insulin resistance and risk of obesity-related cancers**. *Int. J. Epidemiol.* (2020) **49** 193-204. DOI: 10.1093/ije/dyz053
38. Kheirollahi A. **Evaluation of lipid ratios and triglyceride-glucose index as risk markers of insulin resistance in Iranian polycystic ovary syndrome women**. *Lipids Health Dis.* (2020) **19** 235. DOI: 10.1186/s12944-020-01410-8
39. Zhou M. **The triglyceride to high-density lipoprotein cholesterol (TG/HDL-C) ratio as a predictor of insulin resistance but not of beta cell function in a Chinese population with different glucose tolerance status**. *Lipids Health Dis.* (2016) **15** 104. DOI: 10.1186/s12944-016-0270-z
40. Rose DP, Komninou D, Stephenson GD. **Obesity, adipocytokines, and insulin resistance in breast cancer**. *Obes. Rev.* (2004) **5** 153-165. DOI: 10.1111/j.1467-789X.2004.00142.x
41. Kern PA, Ranganathan S, Li C, Wood L, Ranganathan G. **Adipose tissue tumor necrosis factor and interleukin-6 expression in human obesity and insulin resistance**. *Am. J. Physiol. Endocrinol. Metab.* (2001) **280** E745-751. DOI: 10.1152/ajpendo.2001.280.5.E745
42. Gallagher EJ, LeRoith D. **Insulin, insulin resistance, obesity, and cancer**. *Curr. Diabetes Rep.* (2010) **10** 93-100. DOI: 10.1007/s11892-010-0101-y
43. Lorincz AM, Sukumar S. **Molecular links between obesity and breast cancer**. *Endocr. Relat. Cancer.* (2006) **13** 279-292. DOI: 10.1677/erc.1.00729
44. Ni HH. **Combining pre- and postoperative lymphocyte-C-Reactive protein ratios can better predict hepatocellular carcinoma prognosis after partial hepatectomy**. *J. Inflamm. Res.* (2022) **15** 2229-2241. DOI: 10.2147/JIR.S359498
45. Liu YL, Qian HX, Qin L, Zhou XJ, Zhang B. **Serum LDL-C and LDL-C/HDL-C ratio are positively correlated to lymph node stages in males with colorectal cancer**. *Hepatogastroenterology.* (2011) **58** 383-387. PMID: 21661400
46. Notarnicola M. **Serum lipid profile in colorectal cancer patients with and without synchronous distant metastases**. *Oncology* (2005) **68** 371-374. DOI: 10.1159/000086977
47. Ma MZ, Yuan SQ, Chen YM, Zhou ZW. **Preoperative apolipoprotein B/apolipoprotein A1 ratio: A novel prognostic factor for gastric cancer**. *Onco Targets Ther.* (2018) **11** 2169-2176. DOI: 10.2147/OTT.S156690
48. Dai D. **Pretreatment TG/HDL-C ratio is superior to triacylglycerol level as an independent prognostic factor for the survival of triple negative breast cancer patients**. *J. Cancer.* (2016) **7** 1747-1754. DOI: 10.7150/jca.15776
49. Luo YZ. **Pretreatment triglycerides-to-high density lipoprotein cholesterol ratio in postmenopausal women with endometrial cancer**. *Kaohsiung J. Med. Sci.* (2019) **35** 303-309. DOI: 10.1002/kjm2.12033
50. Sun H. **Triglyceride-to-high density lipoprotein cholesterol ratio predicts clinical outcomes in patients with gastric cancer**. *J. Cancer.* (2019) **10** 6829-6836. DOI: 10.7150/jca.35939
51. McNelis JC, Olefsky JM. **Macrophages, immunity, and metabolic disease**. *Immunity* (2014) **41** 36-48. DOI: 10.1016/j.immuni.2014.05.010
52. Matsuzawa Y, Funahashi T, Nakamura T. **Molecular mechanism of metabolic syndrome X: Contribution of adipocytokines adipocyte-derived bioactive substances**. *Ann. N. Y. Acad. Sci.* (1999) **892** 146-154. DOI: 10.1111/j.1749-6632.1999.tb07793.x
53. Fischer S. **Insulin-resistant patients with type 2 diabetes mellitus have higher serum leptin levels independently of body fat mass**. *Acta Diabetol.* (2002) **39** 105-110. DOI: 10.1007/s005920200027
54. Wauters M. **Leptin levels in type 2 diabetes: Associations with measures of insulin resistance and insulin secretion**. *Horm. Metab. Res.* (2003) **35** 92-96. DOI: 10.1055/s-2003-39054
55. Leyva F. **Hyperleptinemia as a component of a metabolic syndrome of cardiovascular risk**. *Arterioscler. Thromb. Vasc. Biol.* (1998) **18** 928-933. DOI: 10.1161/01.ATV.18.6.928
56. Hotamisligil GS, Arner P, Caro JF, Atkinson RL, Spiegelman BM. **Increased adipose tissue expression of tumor necrosis factor-alpha in human obesity and insulin resistance**. *J. Clin. Invest.* (1995) **95** 2409-2415. DOI: 10.1172/JCI117936
57. Aldhahi W, Hamdy O. **Adipokines, inflammation, and the endothelium in diabetes**. *Curr. Diabates Rep.* (2003) **3** 293-298. DOI: 10.1007/s11892-003-0020-2
58. Ibrahim MM. **Subcutaneous and visceral adipose tissue: Structural and functional differences**. *Obes. Rev.* (2010) **11** 11-18. DOI: 10.1111/j.1467-789X.2009.00623.x
59. Sonnenberg GE, Krakower GR, Kissebah AH. **A novel pathway to the manifestations of metabolic syndrome**. *Obes. Res.* (2004) **12** 180-186. DOI: 10.1038/oby.2004.24
|
---
title: Mycophenolic acid directly protects podocytes by preserving the actin cytoskeleton
and increasing cell survival
authors:
- Seif El Din Abo Zed
- Agnes Hackl
- Katrin Bohl
- Lena Ebert
- Emilia Kieckhöfer
- Carsten Müller
- Kerstin Becker
- Gregor Fink
- Kai-Dietrich Nüsken
- Eva Nüsken
- Roman-Ulrich Müller
- Bernhard Schermer
- Lutz T. Weber
journal: Scientific Reports
year: 2023
pmcid: PMC10017704
doi: 10.1038/s41598-023-31326-z
license: CC BY 4.0
---
# Mycophenolic acid directly protects podocytes by preserving the actin cytoskeleton and increasing cell survival
## Abstract
Mycophenolate Mofetil (MMF) has an established role as a therapeutic agent in childhood nephrotic syndrome. While other immunosuppressants have been shown to positively affect podocytes, direct effects of MMF on podocytes remain largely unknown. The present study examines the effects of MMF’s active component Mycophenolic Acid (MPA) on the transcriptome of podocytes and investigates its biological significance. We performed transcriptomics in cultured murine podocytes exposed to MPA to generate hypotheses on podocyte-specific effects of MPA. Accordingly, we further analyzed biological MPA effects on actin cytoskeleton morphology after treatment with bovine serum albumin (BSA) by immunofluorescence staining, as well as on cell survival following exposure to TNF-α and cycloheximide by neutral red assay. MPA treatment significantly (adjusted $p \leq 0.05$) affected expression of 351 genes in podocytes. Gene Ontology term enrichment analysis particularly clustered terms related to actin and inflammation-related cell death. Indeed, quantification of the actin cytoskeleton of BSA treated podocytes revealed a significant increase of thickness and number of actin filaments after treatment with MPA. Further, MPA significantly reduced TNFα and cycloheximide induced cell death. MPA has a substantial effect on the transcriptome of podocytes in vitro, particularly including functional clusters related to non-immune cell dependent mechanisms. This may provide a molecular basis for direct beneficial effects of MPA on the structural integrity and survival of podocytes under pro-inflammatory conditions.
## Introduction
Mycophenolic acid (MPA) is the active component of the prodrug mycophenolate mofetil (MMF). MPA acts as a highly selective and potent inhibitor of the inosine-5′-monophosphate dehydrogenase (IMPDH), leading to a depletion of the pool of guanosine triphosphate (GTP) and deoxy GTP1,2. It has been shown that MPA binds the type II isoform of IMPDH, expressed in activated B- and T-lymphocytes, with an affinity five times higher than the constitutively expressed type I isoform, leading to an effective blockage of lymphocyte proliferation while having only minor effects on the multiplication of other cell types3,4. This mechanism of action, which results in a comparably low adverse risk profile, has led to a wide use of MMF in the field of organ transplantation, where it emerged as an effective drug for prevention of acute rejection of, inter alia, kidney allografts5. Further, MMF has been a vital part of systemic lupus erythematosus therapy, especially in patients with solid organ manifestations, such as lupus nephritis6. MMF is also recommended as a glucocorticoid sparing agent for steroid sensitive nephrotic syndrome in the recently published KDIGO 2021 Clinical Practice Guideline for the Management of Glomerular Diseases7. Its use in treating the most prevalent cause of glomerular disease in children, the idiopathic nephrotic syndrome (INS), has been constantly expanding8. Despite extensive research the pathomechanism of INS has not yet been fully elucidated. Traditionally, INS has been viewed as a T-lymphocyte derived disease, but successful treatment with the anti-CD20 antibody rituximab has amended this viewpoint9–12. Importantly, there has been increasing evidence for a central role of podocytes in the pathogenesis of INS. Podocytes are terminally differentiated cells, whose foot processes (FP), interdigitate with those of neighboring podocytes. This connection constitutes the outer layer of the glomerular filtration barrier, the slit diaphragm (SD)13. During disease the actin cytoskeleton undergoes dynamic changes, thereby inducing FP effacement, which leads to a reduced compressive force on the glomerular membrane and consequently, to the loss of albumin and other proteins into the urine14–16. Other immunosuppressants used in treating INS such as glucocorticoids, cyclosporine A and rituximab positively affect the actin cytoskeleton of podocytes as part of their antiproteinuric effect17–19. Research into the effect of MPA on kidney cells is scarce. In both in vitro and in vivo experiments, it has been shown that MPA inhibits the proliferation of mesangial cells, through the depletion of the guanosine pool20,21. With regard to podocytes, MPA showed a preservative effect on nephrin expression in adriamycin-induced nephritis and diabetes mellitus models22,23. In the latter model, MPA was able to reduce podocyte apoptosis through a reduction of Bax and cleaved caspase-3 (CC3) protein levels. Transcriptomic analysis of murine kidneys in a lupus nephritis model showed an increase in actin associated terms after treatment with MMF24. In the same study, MMF reduced the activation of Rac1 of cultured podocytes compared to controls. To improve our insight into the effect of MPA on podocytes, we were prompted to further investigate direct, non-immune cell mediated responses through which MPA could favorably affect these cells. We therefore performed a transcriptomic analysis of MPA treated podocytes, followed by an investigation of the functional impact of the discovered changes in mRNA levels.
## Cell culture and MPA treatment
A conditionally immortalized murine podocyte cell line (kindly provided by Dr. P. Mundel) was cultivated on type I collagen (A1064401—Invitrogen) coated culture dishes and kept in RPMI-1640 Medium + GlutaMAX (#61870036, Gibco), supplemented with $10\%$ fetal bovine serum (FBS, 10270106—Gibco), $1\%$ HEPES solution (H0887—Sigma) and $1\%$ sodium pyruvate (S8636—Sigma) at 33 °C in the presence of recombinant mouse Interferon-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma$$\end{document}γ(INF-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma$$\end{document}γ, #315–05, PeproTech) to allow proliferation at permissive conditions. This was followed by incubation at 37 °C without additional INF-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma$$\end{document}γ for 10 days to induce differentiation at the non-permissive condition, as described previously25. Differentiation was confirmed by synaptopodin staining (data not shown). Cells were regularly checked for Mycoplasma infection. After differentiation, cells were treated with MPA (M3536—Sigma-Aldrich) with a concentration of 10 mg/L for 2 h, followed by a concentration of 4 mg/L for 22 h. MPA concentrations were chosen based on the pharmacokinetics of MPA plasma concentrations measured in children with idiopathic nephrotic syndrome26. Cells of the control group received the respective amount of methanol as vehicle (Fig. 1A).Figure 1Treatment with MPA leads to changes in the transcriptome of podocytes. ( A) Conditionally immortalized podocytes were differentiated treated with either Vehicle or Mycophenolic Acid (MPA) for 24 h and subjected to RNAseq analysis. ( B) Volcano Plot visualizing the effects of MPA on the transcriptome of podocytes. Blue, adjusted p-value (padj) < 0.05; red, padj > 0.05. MPA treatment resulted in 130 significantly downregulated and 221 significantly upregulated genes. ( C) Genes were sorted into a heatmap according to their z-score transformation value, resulting in four clusters. Cluster 1, red, downregulated genes. Cluster 2, blue, highly downregulated genes. Cluster 3, green, upregulated genes. Cluster 4, orange, highly upregulated genes. ( B) Clusters were subjected to a Gene Ontology (GO)-term enrichment analysis. Important significantly enriched terms are summarized for each cluster.
## RNA extraction, cDNA library construction and sequencing
Cells were treated according to treatment protocol A (Fig. 1A). Total RNA was extracted using an RNA-Extraction Kit by Zymo Research (R2052direct-zol RNA-miniprep) according to the manufacturers’ instruction, mRNA sequencing was performed at the Cologne Center for Genomics. Libraries were prepared using the Illumina® Stranded TruSeq® RNA sample preparation Kit. Library preparation started with 2 µg total RNA. After poly-A selection (using poly-T oligo-attached magnetic beads), mRNA was purified and fragmented using divalent cations under elevated temperature. The RNA fragments underwent reverse transcription using random primers. This was followed by second strand cDNA synthesis with DNA Polymerase I and RNase H. After end repair and A-tailing, indexing adapters were ligated. The products were then purified and amplified (14 PCR cycles) to create the final cDNA libraries. After library validation and quantification (Agilent Tape Station), equimolar amounts of library were pooled. The pool was quantified by using the Peqlab KAPA Library Quantification Kit and the Applied Biosystems 7900HT Sequence Detection System. The pool was sequenced on an Illumina NovaSeq6000 sequencing instrument with a PE100 protocol27,28.
## RNAseq data mining (GO-term)
The reads were trimmed with Trimmomatic (version 0.36) using default parameters29. Additionally, up to ten low quality bases in the beginning of the reads were removed. The trimmed reads were mapped to the GRCm39 mouse reference genome with STAR version 2.6.0 using default parameters30. The analysis of differential gene expression was performed using the R package DeSeq2 (version 1.26.0)31,32. For visualization purposes the batch effect was removed using the method removeBatchEffect from the R package limma33. The enrichment analysis was done using the R package topGO34,35. For the heatmap and clustering the R package pheatmap was used36. The Z-transformed gene expression of genes with an adjusted p-value (padj) < 0.05 (calculated by DeSeq2) in the comparison MPA vs control was visualized and clustered in the heatmap.
## Immunofluorescence and quantification of the actin cytoskeleton
Undifferentiated podocytes were seeded on coverslips and differentiated as described above. After differentiation, cells were incubated in medium with fatty-acid free, low endotoxin bovine serum albumin (BSA, A8806 –Sigma-Aldrich) at a concentration of 50 mg/ml for 48 h (BSA + vehicle [veh]). We used the study design of Yoshida et al.37 as orientation but adjusted the concentration of BSA to reach a robust injury model, where Type C and Type D cells prevail. The treatment group received additional MPA treatment in the second half of the 48 h (BSA + MPA). An additional group was treated only with MPA (MPA). Cells without BSA exposition and MPA treatment served as controls (Fig. 2A). The treated cells were then fixed with $4\%$ paraformaldehyde prior to blocking and permeabilization with $5\%$ Normal Donkey Serum and $0.1\%$ Triton X-100 in phosphate buffered saline solution (PBS) for 30 min. Cells stained for synaptopodin were incubated with a synaptopodin antibody (S9442—Sigma-Aldrich, 1:500) overnight, followed by incubation with the appropriate secondary antibody at room temperature (RT) for 60 min (Alexa Fluor® 488 AffiniPure Donkey Anti-Rabbit IgG (H + L), 1:250). Cells stained for actin were incubated with fluorescently conjugated phalloidin at RT for 60 min (647P1-33—Dyomics, 1:50). The coverslips were then mounted with Prolong Gold and DAPI. Images were acquired with a Zeiss Meta 710 Confocal Laser Scanning Microscope and processed with ImageJ. For quantitative analysis of actin fiber assembly patterns, a previously described scoring system was adapted, and images were scored by an observer blinded to cell treatment38. At least 90 cells were analyzed per sample. The four groups describing the different actin fiber assembly patterns were defined as follows: Type A: $90\%$ of cell area filled with thick cables; type B: at least two thick cables running under the nucleus, the rest of the cell area filled with fine cables; type C: no thick cables but some cables present; and type D: no cables visible in the central area of the cell (Fig. 2B).Figure 2MPA protects the actin cytoskeleton from BSA induced injury. ( A) Cells were treated with bovine serum albumin (BSA) for 48 h and received either MPA or Vehicle for the second 24 h. The actin cytoskeleton was visualized using immunofluorescence staining. ( B) Quantification of different cell types after treatment with bovine serum albumin (BSA) and Mycophenolic Acid (MPA). Grey = Type A: > $90\%$ of cell area filled with thick cables; green = type B: at least two thick cables running under the nucleus, with the rest of the cell area filled with fine cables; yellow = type C: no thick cables but some cables present; red = type D: no cables visible in the central area of the cell. BSA treatment decreases healthy Type A and B cells, while increasing less healthy Type C and D cells. Additional MPA treatment reverses this effect in all categories, increasing Type A and B cells compared to BSA only, while decreasing Type C and D cells. The treatment with only MPA resulted in a similar distribution of cell types as in controls. At least 90 cells were analyzed per sample. ( C) Quantification of Type D cells for each treatment. Treatment with BSA significantly increases the number of cableless D cells. In contrast additional treatment with MPA significantly decreases the number of D cells compared to treatment with only BSA **$p \leq 0.01.$ ( D) Representative confocal images of the podocytes after each respective treatment. Green, phalloidin. Blue, DAPI.
## Measurement of cell viability
Cells were seeded and differentiated in a 96-well culture dish. After differentiation, the treatment group was pretreated with MPA for 24 h. Afterwards cells were treated with recombinant mouse Tumor Necrosis Factor-α (TNF-α; Recombinant Mouse TNF-alpha [aa 80–235] Protein, #410-MT, R&D Systems, 20 ng/ml), Cycloheximide (CHX; C4859, Sigma-Aldrich, 5 µg/ml) and MPA for an additional 24 h (TCM). We used TNF-α in combination with the synergistically acting CHX in order to reach a high susceptibility to apoptosis39,40. Other treatment groups included cells receiving only TNF-α and CHX (TC), as well as cells receiving TNF-α, CHX and a pan-caspase inhibitor, Emricasan (SEL-S7775, Biozol) (TCE)41. Cells receiving only the vehicles of all respective agents served as control. An additional control group received only MPA for 48 h (M). The treatment schedules and concentrations of all agents are visualized in Fig. 3A. During the last 2 h of treatment, neutral red was added to each well. Next, cells were washed, and the dye was extracted out of viable cells with a destaining solution ($50\%$ EtOH, $49\%$ VE H2O, $1\%$ acetic acid). The absorbance of each well was measured through a spectrophotometer at 540 nm. The amount of viable cells was calculated as a ratio of the absorbance of each well to the absorbance of their respective control. Each condition was conducted with triplets of three different passages. Figure 3MPA increases cell viability during TNF-α and CHX induced cell death. ( A) Cells were treated with either Vehicle or MPA (M) for 48 h, or Emricasan (E) for 24 h. All cells received Tumor Necrosis Factor-α (T) and Cycloheximide (C) for 24 h. An additional control group received only MPA for 48 h (M). This resulted in 5 treatment groups (Control, TC, TCM, TCE and M), which were analyzed with a cell viability assay. ( B) Quantification of cell viability after treatment with combinations of Tumor Necrosis Factor-α (TNF-α; T), Cycloheximide (CHX; C), Mycophenolic Acid (MPA; M) and Emricasan (E). Control cells received only vehicles. Cell viability of each sample was normalized against the control of their respective passage. TC-treatment resulted in a drastic decrease of cell viability. Additional MPA treatment significantly increases cell viability compared to TC only (36.8 vs. $62.9\%$). Emricasan was able to prevent cell death in a very high degree ($91\%$). Treatment with only MPA showed no signs for proliferative activity of the drug. ** $p \leq 0.01$ (C) Western Blot Analysis of cleaved Caspase-3 levels. Lane 1, Control. Lane 2, TC. Lane 3, TCM. Treatment with TC results in a steep increase of cleaved Caspase-3 levels. TCM treated cells display a reduction of cleaved Caspase-3 compared to TC treated cells. Fl, full length. DMSO, Dimethyl sulfoxide. MtOH, Methanol. Gapdh, Glyceraldehyde 3-phosphate dehydrogenase. The displayed blot was cropped for better visualization. The boxes exemplify different regions of the same gel. The uncropped blots can be found as Supplementary Fig. 1.
## Western blots related to cell viability
Cells of the treatment groups Control, TC and TCM were treated as described above (Fig. 3A). Treated cells were scraped off the dishes with ice cold PBS and cell pellets were isolated through centrifugation at 5.000 rpm for 5 min, before being snap frozen in liquid nitrogen. The cell pellets were resuspended in RIPA Buffer with additional Na3VO4 (20 µl/ml) and phenylmethylsulfonylfluoride (4,4 µl/ml) and then subject to the BCA Method to determine the amount of total protein. 15 µg of total protein were loaded per lane on $10\%$ or $12\%$ SDS-Page gels. The separated proteins were transferred onto polyvinylidene difluoride membranes, which were blocked in $5\%$ BSA for 1 h. Next, the membranes were probed with primary antibodies overnight at 4 °C. After washing, the membranes were incubated with the secondary antibody at RT for 1 h. The signals were visualized with SuperSignal West Femto Maximum Sensitivity Substrate (#34095, Thermo Scientific) and bands were quantified using Fiji-ImageJ Version $\frac{2.1.0}{1.53}$c42. Protein expression was quantified as the ratio of a specific band to glyceraldehyde-3-phosphate dehydrogenase (GAPDH).
The following primary antibodies were used in this study: cleaved Caspase-3 (Asp175) Antibody #9661 (Cell Signaling Technology), 1:1.000; GAPDH (D16H11) XP® Rabbit mAb#5174 (Cell Signaling Technology), 1:1.000. A horseradish peroxidase-conjugated goat antibody against rabbit (Cy™3 AffiniPure Donkey Anti-Rabbit IgG (H + L), AB_2307443, Jackson Immuno Research) was used as a secondary antibody (1:30.000).
## Statistical analysis
A two-way ANOVA test was used to compare the quantification of the actin cytoskeleton. A t-test with Welch’s correction was used to compare the results of the cell viability assay. Results with $p \leq 0.05$ were considered statistically significant. Results are reported as mean \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± SEM.
## RNA sequencing reveals a significant direct effect of MPA on the transcriptome of podocytes
To gain a profound overview about the MPA-induced processes, we conducted an RNA sequencing (RNAseq) of MPA-treated podocytes. The experimental design is shown in Fig. 1A. We identified 351 genes that met the criterion of an adjusted $p \leq 0.05$ in order to be considered significantly affected by MPA treatment compared to controls.
A volcano plot visualizes the distribution of gene transcripts, showing that of the 351 differentially expressed genes 130 were significantly reduced and 221 were significantly increased (Fig. 1B). Both isoforms of IMPDH were highly expressed in all samples, with a consistently higher read count for IMPDH II (Suppl. Data). Interestingly, IMPDH II showed a significant, compensatory upregulation upon MPA treatment (log2FC 0.26, padj = 0.0002).
In order to better reveal inherent relations between these genes, we organized our data by the z-score transformed gene expression of each gene (Fig. 1C). Comparison of the data demonstrates a considerable consistency of expression changes inside each condition. *Four* gene meta-clusters were identified: down-regulated (Cluster 1), highly down-regulated (Cluster 2), up-regulated (Cluster 3) and highly up-regulated (Cluster 4).
To clarify molecular functions and biological processes affected by MPA-regulated genes and to address their relevance in podocytes, we next performed enrichment analysis for each cluster using Gene Ontology (GO). Figure 1C also depicts a selection of significant terms associated with each cluster. Complete result tables of enrichment are provided as supplementary information (Supplementary Dataset File 1–4). Looking at the results of the enrichment analysis we could outline two major groups of terms that were of particular interest.
## MPA treatment increases actin associated terms and genes
Looking at the significant terms of the clusters with genes induced upon MPA treatment (cluster 3 and cluster 4), we noticed an accumulation of terms related to the cytoskeletal structure of the cell. On the one hand, we had an increase of terms narrowing down the cellular components affected by the differentially expressed genes, like actin cytoskeleton (fold enrichment [FE] 1.7, $$p \leq 0.01$$), stress fiber (FE 2.2, $$p \leq 0.03$$), or actomyosin (FE 2.0, $$p \leq 0.04$$) (Fig. 1C). On the other hand, we could also see an increase in terms associated with the dynamic changes of those cellular components, like actin filament bundle assembly (FE 2.5, $$p \leq 0.01$$) and regulation of Rho protein- and small GTPase mediated signal transduction (FE 3.2 and 3.0, $$p \leq 0.004$$ and 0.03 respectively). Genes annotated to the terms of this subgroup include SYNPO2L (log2FC 0.80, padj = 466E-10), an actin associated protein from the synaptopodin family, and ITGB1 (log2FC 0.17, padj = 0.0006), a protein crucial for cytoskeletal organization and signal transduction in podocytes. Further, we could see an upregulation of guanine nucleotide exchange factors (e.g., ARHGEF2), downstream effectors of RhoA (e.g., ROCK2) and other proteins that influence the activity of smallGTPases (e.g., AMOT) (Table 1).Table 1Genes associated to the actin cytoskeleton. NameProtein nameAdjusted p-valueLog2-fold changeSYNPO2LSynaptopodin-2-like protein4.66E-100.80DRR1Actin-associatedprotein FAM107A0.02430.39TXNRD1Thioredoxin reductase 12.38E-080.25MYL12BMyosin regulatory light chain 12B8.11E-070.23CD157ADP-ribosyl cyclase/cyclic ADP-ribose hydrolase 21.03E-060.22ALDOAFructose-bisphosphatealdolase A1.21E-070.21ITGB1Integrin beta-10.00060.17ARHGEF9Rho guanine nucleotide exchange factor 90.00860.31ARHGEF2Rho guanine nucleotide exchange factor 20.01640.15ROCK2Rho-associatedproteinkinase 20.02570.14AMOTAngiomotin0.03350.14Selection of DEG’s annotated to significantly enriched GO-terms associated to inflammation and cell death. The portrayed genes were selected due to possible biological significance for podocyte health. Adjusted p-value was considered significant at $p \leq 0.05.$
## MPA treatment influences terms related to inflammatory pathways and cell death
Another major group that came to our attention includes terms related to inflammatory pathways and cell death. Here, we could see an increase in terms describing the downregulation of inflammatory pathways, especially the p38-Mitogen activated protein kinase (p38MAPK) and the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) pathway. The negative regulation of p38MAPK cascade (FE 5.3, $$p \leq 0.001$$) presented itself as the GO-term with the largest fold enrichment (Fig. 1C). Notably, we found an increase in gene transcripts of the dual specificity protein phosphatase (DUSP) family, namely DUSP1, DUSP4 and DUSP6 (log2FC 0.23, 0.39, 0,31 and padj = 0.0004, 9.33E-05, 0.007 respectively). Treatment with MPA also increases the transcription of ubiquitin carboxyl-terminal hydrolase CYLD (log2FC 0.12, padj = 0.04), which is known to suppress the NF-κB, p38MAPK and c-Jun N-terminal kinase (JNK) cascades. Simultaneously, terms concerning the activation of these pathways through INF-χ and transforming growth factor-β (TGF-β) were significantly downregulated and therefore appeared in Cluster 1. One example of a downregulated transcript is THBS1 (log2FC − 0.16, padj = 0.006), which, if translated, can lead to an activation of TNF-α and TGF-β pathways. The transcript with the largest negative fold change after MPA treatment is PLK-1 (log2FC − 1.03, padj = 0.0004), which translates into a serine-threonine kinase with important functions in cell cycle regulation. In addition to the regulation of terms concerning inflammatory pathways, we could also see a surge of terms mentioning the negative regulation of cell death pathways, such as apoptosis and necroptosis. Mentionable examples of genes regulated by MPA and annotated to these functions are CCN-I, PEA15A, and DDX3X (Table 2).Table 2Genes associated to inflammation and cell death. NameAdjusted p-valueLog2-fold changeDUSP1Dual specificityproteinphosphatase 10.00040.23DUSP4Dual specificityproteinphosphatase 49.33E-050.39DUSP6Dual specificityproteinphosphatase 60.01760.21AK4Adenylatekinase 40.00130.45SLC25A4ADP/ATP translocase 12.10E-100.32CYLDUbiquitin carboxyl-terminal hydrolase CYLD0.04420.12TAX1BP1Tax1-binding protein 10.03980.12PLK-1Serine/threonine-protein kinase PLK10.0004− 1.03CDK9Cyclin-dependentkinase 90.0136− 0.19ARRB1Beta-arrestin-10.0183− 0.20CXCL12Stromalcell-derivedfactor 11.42E-06− 0.26LRP5Low-density lipoprotein receptor-related protein 50.0346− 0.17THBS1Thrombospondin-10.0056− 0.6Selection of DEG’s annotated to significantly enriched GO-terms associated to the actin cytoskeleton. The portrayed genes were selected due to possible biological significance for podocyte health. Adjusted p-value was considered significant at $p \leq 0.05.$
## MPA treatment is able to protect the actin cytoskeleton from BSA induced stress
To validate the biological significance of changes in the actin cytoskeleton related genes, we designed an experimental setup intended to damage actin filaments in a non-immunological manner and with resemblance to the pathophysiology of the nephrotic syndrome (Fig. 2A). We then stained the cells with Phalloidin, which interacts highly selectively with F-actin, to visualize the cytoskeletal structure of each cell (Fig. 2D). After 48 h of incubation with BSA, cells showed a disorganization of the actin cytoskeleton and an increased appearance of vacuoles. The quantification of the cytoskeleton revealed a substantial decrease of cells with thick cables (Type A and B) after treatment with BSA compared to controls. Simultaneously, the number of cells with thin or no cables (Type C and D) increased remarkably compared to control cells (Fig. 2B). Interestingly, the BSA + MPA group displayed a vastly different appearance, compared to the BSA + veh group. The number of Type A and B cells increased significantly compared to cells treated only with BSA ($$p \leq 0.004$$ and $$p \leq 0.001$$ respectively, Fig. 2B). Concomitantly, the share of non-stress fiber containing Type D cells decreased in a significant manner ($$p \leq 0.004$$, Fig. 2C), displaying an approximation of the actin cytoskeleton pattern of BSA + MPA cells, to that of control cells. The treatment with only MPA resulted in a similar distribution of cell types as in controls.
## MPA treatment reduces TNF-α and CHX induced cell death
Next, we examined consequences of the transcriptomic changes regarding inflammation and cell death in a functional environment. In view of the pathways that appear to be affected by MPA treatment, we chose TNF-α and CHX as appropriate, synergistically acting stressing agents, with a potent ability to induce cell death. The experimental design is shown in Fig. 3A. Using the neutral red uptake assay, we discovered that treatment with TNF-α and CHX reduced cell viability to only $36.8\%$ (Fig. 3B). If simultaneously treated with MPA, no significant difference in cell survival was notable (data not shown). In contrast, cells with a treatment schedule including 24 h of pretreatment with MPA, showed a significantly reduced cell death rate, with $62.9\%$ of cells viable after incubation with TNF-α and CHX (Fig. 3B, $$p \leq 0.008$$). Emricasan, an inhibitor of pan-caspases, was able to strongly diminish the cell damaging activity of the two agents, leaving $91\%$ of cells viable after treatment. A proliferative effect due to MPA treatment was ruled out using an MPA-only group (Fig. 3B). To further validate our results and to elucidate the role of caspase-inhibition through MPA, we decided to perform a Western *Blot analysis* of CC3 levels. As expected, TC treatment strongly initiated Caspase-3 cleavage. More importantly, we found that additional MPA treatment decreased the amount of CC3 (Fig. 3C). A quantification of the western blot results can be found in the supplementary information (Supp. Fig. 2).
## Discussion
To our knowledge, we are the first to use an unbiased transcriptomic approach in order to investigate how MPA affects podocytes in a non-immunologic environment in a favorable manner. Interestingly, the result of the transcriptomic analysis showed that treatment with MPA for 24 h increased gene transcription of IMPDH II significantly (Fig. 1B), implying an effective exposure to MPA. This result is in line with past research, which had also demonstrated that murine podocytes express IMPDH transcripts and that IMPDH activity can be inhibited in cell culture43.
Hierarchical clustering identified 4 clusters, which were then analyzed for functional implications by conducting a GO-Term Enrichment Analysis. Upon closer examination we discovered that there was an accumulation of enriched terms associated to actin, especially in clusters of upregulated genes, such as actin cytoskeleton and regulation of small GTPase mediated signal transduction (Fig. 1C). Actin is the predominant cytoskeletal structure in FPs and as reviewed previously its reorganization or dysfunction can be a leading cause for proteinuria44–46. Since stabilization of the actin cytoskeleton has been demonstrated as a direct effect of cyclosporine A on podocytes19 we tested if MPA has a similar ability to directly act on the podocyte actin cytoskeleton. Therefore, we designed an experimental setup in which BSA was used as a stressing agent. In line with previous literature37, BSA treatment markedly decreased the amount and thickness of stress fibers. Notably, MPA was able to protect the cells from the albumin overload induced injury. Treatment with MPA reconstituted the number of actin rich Type A and B cells, while showing a significant decrease of Type D cells compared to the BSA + veh group (Fig. 2B,C).
This result strongly supported our concept of a stabilizing effect of MPA on the actin cytoskeleton. Next, we went back to our transcriptome data and looked at the actin cytoskeleton stabilization associated genes in more detail. The first interesting candidate was SYNPO2L (Table 1). SYNPO2L is a member of the synaptopodin family, which has been shown to stabilize the actin cytoskeleton in numerous studies38,47–49. If overexpressed SYNPO2L itself can activate the actin signaling pathway, increasing proteins like RhoA and Actn2 while inducing stress fiber formation50. Since synaptopodin is lost in childhood nephrotic syndrome51, an increase in a member of the synaptopodin family, like SYNPO2L, could potentially compensate for the loss of the actin supporting protein. Another interesting gene regulated through MPA is ITGB1, a major adhesion molecule in podocytes, which is downregulated under mechanical stress as well as under albumin overload, both being leading pathological mechanisms in INS52,53. The decrease of ITGB1 can result in a reduced podocyte adhesion and an increase in Caspase-3 activity54. Lee et al. were able to show that pyrintegrin, a beta1-integrin-agonist, is able to protect murine podocytes from effacement and subsequently mice from proteinuria55. In our study we discovered that MPA increases ITGB1 gene transcription, thereby possibly making use of its stabilizing effect on podocytes.
Further examination of the RNAseq-data revealed another group of terms, including negative regulation of p38MAPK and negative regulation of apoptotic process in the upregulated clusters and a negative regulation of TGF-β receptor signaling pathway, that are linked to cell death and inflammation (Fig. 1C). The p38MAPK-pathway is a well-known pro-inflammatory and pro-apoptotic pathway, that is among other things activated during endoplasmic reticulum stress (ERS)56,57. ERS, TGF-β and p38MAPK have all been shown to be upregulated through albumin overload in podocytes in vitro37,58. Koshikawa et al. were able to demonstrate that p38MAPK-activation is increased in rodent kidney disease models, as well as in human glomerulopathies, including MCGN and FSGS59. In line with these findings, the inhibition of TGF-β and p38MAPK proved to be protective for podocytes in vitro and in vivo37,59,60. Considering our transcriptomic analysis together with the aforementioned studies, we tested if MPA has the ability to protect podocytes from inflammation and subsequent cell death and examined this hypothesis using a cell viability assay. Previous work had shown that nephrotic syndrome in mice and albumin exposure to podocytes in vitro increase TNF-RNA61, which in turn can induce podocyte injury and glomerular disease progression62,63. We therefore treated our cells with a routinely used combination of TNF-α and CHX, which showed a massively detrimental impact on podocyte survival. Nonetheless we were able to demonstrate a significantly higher survival of cells pre-treated with MPA compared to those not treated with MPA ($62.9\%$ vs. $36.8\%$, Fig. 3B). Thus, our data impressively demonstrate that TNF-α -induced cell death can effectively be mitigated when cells are pretreated with MPA. On the contrary, treatment with MPA after stimulation with TNF-α was not successful in preventing cell death. Subsequently, we demonstrated that MPA treatment reduces Caspase-3 cleavage induced by TC treatment (Fig. 3C). Taken together, we were able to show that MPA attenuates TNF-α induced cell death in cultured podocytes. Mechanistically, this can at least partially be related to a reduction of Caspase-3 activation.
Strikingly, our transcriptomic results showed the upregulation of three members of the DUSP family after MPA treatment, namely DUSP1, 4 and 6, which are negative regulators of MAPK-pathways including the p38- and JNK-pathway (Table 2)64. DUSP4 and 6 have both been found to be decreased in diabetical nephropathy models65,66. If overexpressed in podocyte culture, DUSP4 decreases the activation of p38MAPK, JNK and Caspase-365. Similarly, overexpression of DUSP6 protects podocytes from synaptopodin and nephrin loss, while reducing high-glucose induced apoptosis66.
Other mentionable genes affected by MPA that could potentially play a role in the treatment of proteinuric diseases include PLK-1 and CXCL12, which were both downregulated through MPA treatment (Table 2). PLK-1 is a key regulator of the cell cycle and cell division and its inhibition has been shown to reduce proteinuria in a diabetic nephropathy model67,68. CXCL12, also known as stromal cell derived factor, is a homeostatic chemokine, whose production by podocytes has been shown to contribute to podocyte loss and albuminuria in type 2 diabetes69. Its blockage ameliorates proteinuria and increases podocyte numbers in models of both diabetical nephropathy and adriamycin-induced nephropathy70,71.
Taken together, MPA does indeed protect the actin cytoskeleton and can prevent the cell death of podocytes in cell culture. Although our results have the usual limitations of cultured podocytes regarding cell type specificity and environmental surrounding, the cell culture model is highly valuable to study effects beyond those of immune cells on podocytes, which are confounding the results in almost all in vivo models. Nevertheless, future studies will have to examine the translatability of our findings into animal models, as well as compare our findings to transcriptomic analysis of human samples. Understanding the mechanisms of action of clinically used immunosuppressive drugs in INS will help to further adapt therapeutic regimes to fully exploit its therapeutic potential. In conclusion, this study provides compelling evidence for a direct favorable effect of MPA on podocytes, while providing potential responsible pathways.
## Supplementary Information
Supplementary Information 1.Supplementary Information 2.Supplementary Information 3.Supplementary Information 4.Supplementary Information 5. The online version contains supplementary material available at 10.1038/s41598-023-31326-z.
## References
1. Allison A. **Mechanisms of action of mycophenolate mofetil**. *Lupus* (2005.0) **14** 2-8. DOI: 10.1191/0961203305lu2109oa
2. 2.Allison, A. C., & Eugui, E. M. Mycophenolate Mofetil and Its Mechanisms of Action. Vol 47. (2000). www.elsevier.comrlocaterimmpharm.
3. Eugui EM, Mirkovich A, Allison AC. **Lymphocyte-selective antiproliferative and immunosuppressive effects of mycophenolic acid in mice**. *Scand. J. Immunol.* (1991.0) **33** 175-183. DOI: 10.1111/j.1365-3083.1991.tb03747.x
4. Carr SF, Papp E, Wu JC, Natsumeda Y. **Characterization of human type I and type II IMP dehydrogenases**. *J. Biol. Chem.* (1993.0) **268** 27286-27290. DOI: 10.1016/s0021-9258(19)74247-1
5. Sollinger H. **Mycophenolate Mofetil for the prevention of acute rejection in primary cadaveric renal allograft recipients**. *Transplantation* (1995.0) **60** 225-232. DOI: 10.1097/00007890-199508000-00003
6. Kuhn A, Bonsmann G, Anders HJ, Herzer P, Tenbrock K, Schneider M. **The diagnosis and treatment of systemic lupus erythematosus**. *Dtsch. Arztebl. Int.* (2015.0) **112** 423-432. DOI: 10.3238/arztebl.2015.0423
7. Rovin BH, Adler SG, Barratt J. **KDIGO 2021 clinical practice guideline for the management of glomerular diseases**. *Kidney Int.* (2021.0) **100** S1-S276. DOI: 10.1016/j.kint.2021.05.021
8. Noone DG, Iijima K, Parekh R. **Idiopathic nephrotic syndrome in children**. *Lancet* (2018.0) **392** 61-74. DOI: 10.1016/S0140-6736(18)30536-1
9. Shalhoub RJ. **Pathogenesis of lipoid nephrosis: A disorder of T-cell function**. *Lancet* (1974.0) **304** 556-560. DOI: 10.1016/S0140-6736(74)91880-7
10. Ravani P, Rossi R, Bonanni A. **Rituximab in children with steroid-dependent nephrotic syndrome: A multicenter, open-label, noninferiority, randomized controlled trial**. *J. Am. Soc. Nephrol.* (2015.0) **26** 2259-2266. DOI: 10.1681/ASN.2014080799
11. Iijima K, Sako DM, Nozu K. **Rituximab for childhood-onset, complicated, frequently relapsing nephrotic syndrome or steroid-dependent nephrotic syndrome: A multicentre, double-blind, randomised, placebo-controlled trial**. *Lancet* (2014.0) **384** 1273-1281. DOI: 10.1016/S0140-6736(14)60541-9
12. Ahn YH, Kim SH, Han KH. **Efficacy and safety of rituximab in childhood- onset, difficult-to-treat nephrotic syndrome**. *Medicine* (2018.0) **97** 1-9. DOI: 10.1097/MD.0000000000013157
13. Tryggvason K, Wartiovaara J. **Molecular basis of glomerular permselectivity**. *Curr. Opin. Nephrol. Hypertens.* (2001.0) **10** 543-549. DOI: 10.1097/00041552-200107000-00009
14. Garg P. **A review of podocyte biology**. *Am. J. Nephrol.* (2018.0) **47** 3-13. DOI: 10.1159/000481633
15. Benzing T. **Signaling at the slit diaphragm**. *J. Am. Soc. Nephrol.* (2004.0) **15** 1382-1391. DOI: 10.1097/01.ASN.0000130167.30769.55
16. Butt L, Unnersjö-Jess D, Höhne M. **A molecular mechanism explaining albuminuria in kidney disease**. *Nat. Metab.* (2020.0) **2** 461-474. DOI: 10.1038/s42255-020-0204-y
17. 17.Fornoni, A., Sageshima, J., & Wei, C., et al.Rituximab Targets Podocytes in Recurrent Focal Segmental Glomerulosclerosis. (2011). www.ScienceTranslationalMedicine.org.
18. Hosseiniyan Khatibi SM, Ardalan M, Abediazar S, Zununi Vahed S. **The impact of steroids on the injured podocytes in nephrotic syndrome**. *J. Steroid Biochem. Mol. Biol.* (2020.0). DOI: 10.1016/j.jsbmb.2019.105490
19. Faul C, Donnelly M, Merscher-Gomez S. **The actin cytoskeleton of kidney podocytes is a direct target of the antiproteinuric effect of cyclosporine A**. *Nat Med.* (2008.0) **14** 931-938. DOI: 10.1038/nm.1857
20. Ziswiler R, Steinmann-Niggli K, Kappeler A, Daniel C, Marti HP. **Mycophenolic acid: A new approach to the therapy of experimental mesangial proliferative glomerulonephritis**. *J. Am. Soc. Nephrol.* (1998.0) **9** 2055-2066. DOI: 10.1681/asn.v9112055
21. Hauser IA, Renders L, Radeke HH, Sterzel RB, Goppelt-Struebe M. **Mycophenolate mofetil inhibits rat and human mesangial cell proliferation by guanosine depletion**. *Nephrol. Dial. Transplant.* (1999.0) **14** 58-63. DOI: 10.1093/ndt/14.1.58
22. 22.Lv, W., et al.Mycophenolate Mofetil Inhibits Hypertrophy and Apoptosis of Podocyte in Vivo and in Vitro. Vol 8. www.ijcem.com/. (2015).
23. Nakhoul F, Ramadan R, Khankin E. **Glomerular abundance of nephrin and podocin in experimental nephrotic syndrome: Different effects of antiproteinuric therapies**. *Am. J. Physiol. Ren. Physiol.* (2005.0). DOI: 10.1152/ajprenal.00451.2004
24. Fu J, Wang Z, Lee K. **Transcriptomic analysis uncovers novel synergistic mechanisms in combination therapy for lupus nephritis**. *Kidney Int.* (2018.0) **93** 416-429. DOI: 10.1016/j.kint.2017.08.031
25. Shankland SJ, Pippin JW, Reiser J, Mundel P. **Podocytes in culture: Past, present, and future**. *Kidney Int.* (2007.0) **72** 26-36. DOI: 10.1038/sj.ki.5002291
26. Benz MR, Ehren R, Kleinert D. **Generation and validation of a limited sampling strategy to monitor mycophenolic acid exposure in children with nephrotic syndrome**. *Ther. Drug Monit.* (2019.0) **41** 696-702. DOI: 10.1097/FTD.0000000000000671
27. Peifer M, Hertwig F, Roels F. **Telomerase activation by genomic rearrangements in high-risk neuroblastoma**. *Nature* (2015.0) **526** 700-704. DOI: 10.1038/nature14980
28. George J, Lim JS, Jang SJ. **Comprehensive genomic profiles of small cell lung cancer**. *Nature* (2015.0) **524** 47-53. DOI: 10.1038/nature14664
29. Bolger AM, Lohse M, Usadel B. **Trimmomatic: A flexible trimmer for Illumina sequence data**. *Bioinformatics* (2014.0) **30** 2114-2120. DOI: 10.1093/bioinformatics/btu170
30. Dobin A, Davis CA, Schlesinger F. **STAR: Ultrafast universal RNA-seq aligner**. *Bioinformatics* (2013.0) **29** 15-21. DOI: 10.1093/bioinformatics/bts635
31. Anders S, Huber W. **Differential expression and sequence-specific interaction of karyopherin α with nuclear localization sequences**. *J. Biol. Chem.* (1997.0) **272** 4310-4315. DOI: 10.1074/jbc.272.7.4310
32. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol.* (2014.0) **15** 1-21. DOI: 10.1186/s13059-014-0550-8
33. Ritchie ME, Phipson B, Wu D. **Limma powers differential expression analyses for RNA-sequencing and microarray studies**. *Nucleic Acids Res.* (2015.0) **43** e47. DOI: 10.1093/nar/gkv007
34. 34.Alexa, A., & Rahnenführer, J. topGO: Enrichment analysis for Gene Ontology. R Packag version 2460. http://bioconductor.org/packages/release/bioc/html/topGO.html. (2021).
35. Alexa A, Rahnenführer J, Lengauer T. **Improved scoring of functional groups from gene expression data by decorrelating GO graph structure**. *Bioinformatics* (2006.0) **22** 1600-1607. DOI: 10.1093/bioinformatics/btl140
36. 36.Kolde, R. Pheatmap: Pretty heatmaps. R Pacakage Version. ;61(926). https://cran.r-project.org/package=pheatmap (2019).
37. Yoshida S, Nagase M, Shibata S, Fujita T. **Podocyte injury induced by albumin overload in vivo and in vitro: Involvement of TGF-beta and p38 MAPK**. *Nephron Exp. Nephrol.* (2008.0). DOI: 10.1159/000124236
38. Ning L, Suleiman HY, Miner JH. **Synaptopodin is dispensable for normal podocyte homeostasis but is protective in the context of acute podocyte injury**. *J. Am. Soc. Nephrol.* (2020.0). DOI: 10.1681/asn.2020050572
39. Nolop KB, Ryan US. **Enhancement of tumor necrosis factor-induced endothelial cell injury by cycloheximide**. *Am. J. Physiol. Lung Cell Mol. Physiol.* (1990.0). DOI: 10.1152/ajplung.1990.259.2.l123
40. Kieckhöfer E, Slaats GG, Ebert LK. **Primary cilia suppress Ripk3-mediated necroptosis**. *Cell Death Discov.* (2022.0) **8** 1-12. DOI: 10.1038/s41420-022-01272-2
41. Hoglen NC, Chen L, Fisher CD, Hirakawa BP, Groessl T, Contreras PC. **tetrafluoro-phenoxy ) -pentanoic Acid ): A liver-targeted caspase inhibitor**. *J. Cell. Mol. Med.* (2004.0) **309** 634-640
42. Schindelin J, Arganda-Carreras I, Frise E. **Fiji: An open-source platform for biological-image analysis**. *Nat. Methods* (2012.0) **9** 676-682. DOI: 10.1038/nmeth.2019
43. Nakajo A, Khoshnoodi J, Takenaka H. **Mizoribine corrects defective nephrin biogenesis by restoring intracellular energy balance**. *J. Am. Soc. Nephrol.* (2007.0) **18** 2554-2564. DOI: 10.1681/ASN.2006070732
44. Drenckhahn D, Franke RP. **Ultrastructural organization of contractile and cytoskeletal proteins in glomerular podocytes of chicken, rat, and man**. *Lab Invest.* (1988.0) **59** 673-682. PMID: 3141719
45. Greka A, Mundel P. **Cell biology and pathology of podocytes**. *Annu. Rev. Physiol.* (2012.0) **74** 299-323. DOI: 10.1146/annurev-physiol-020911-153238
46. Perico L, Conti S, Benigni A, Remuzzi G. **Podocyte-actin dynamics in health and disease**. *Nat. Rev. Nephrol.* (2016.0) **12** 692-710. DOI: 10.1038/nrneph.2016.127
47. Yu SMW, Nissaisorakarn P, Husain I, Jim B. **Proteinuric kidney diseases: A podocyte’s slit diaphragm and cytoskeleton approach**. *Front. Med.* (2018.0). DOI: 10.3389/fmed.2018.00221
48. Asanuma K, Yanagida-Asanuma E, Faul C, Tomino Y, Kim K, Mundel P. **Synaptopodin orchestrates actin organization and cell motility via regulation of RhoA signalling**. *Nat. Cell Biol.* (2006.0) **8** 485-491. DOI: 10.1038/ncb1400
49. Chalovich JM, Schroeter MM. **Synaptopodin family of natively unfolded, actin binding proteins: Physical properties and potential biological functions**. *Biophys. Rev.* (2010.0) **2** 181-189. DOI: 10.1007/s12551-010-0040-5
50. Van Eldik W, Den Adel B, Monshouwer-Kloots J. **Z-disc protein CHAPb induces cardiomyopathy and contractile dysfunction in the postnatal heart**. *PLoS ONE* (2017.0) **12** 1-22. DOI: 10.1371/journal.pone.0189139
51. Srivastava T, Garola RE, Whiting JM, Alon US. **Synaptopodin expression in idiopathic nephrotic syndrome of childhood**. *Kidney Int.* (2001.0) **59** 118-125. DOI: 10.1046/j.1523-1755.2001.00472.x
52. Li D, Lu Z, Jia J, Zheng Z, Lin S. **Changes in microRNAs associated with podocytic adhesion damage under mechanical stress**. *JRAAS J. Renin-Angiotensin-Aldosterone Syst.* (2013.0) **14** 97-102. DOI: 10.1177/1470320312460071
53. Cheng YC, Chen CA, Chang JM, Chen HC. **Albumin overload down-regulates integrin-β1 through reactive oxygen species-endoplasmic reticulum stress pathway in podocytes**. *J. Biochem.* (2014.0) **158** 101-108. DOI: 10.1093/jb/mvv020
54. Dessapt C, Baradez MO, Hayward A. **Mechanical forces and TGFβ1 reduce podocyte adhesion through α3β1 integrin downregulation**. *Nephrol. Dial. Transpl.* (2009.0) **24** 2645-2655. DOI: 10.1093/ndt/gfp204
55. Lee HW, Khan SQ, Faridi MH. **A podocyte-based automated screening assay identifies protective small molecules**. *J. Am. Soc. Nephrol.* (2015.0) **26** 2741-2752. DOI: 10.1681/ASN.2014090859
56. 56.Zarubin, T., Han, J. Activation and Signaling of the P38 MAP Kinase Pathway. Vol 15. www.cell-research.com7C (2005).
57. Xu C, Bailly-maitre B, Reed JC, Xu C, Bailly-maitre B, Reed JC. **Endoplasmic reticulum stress: Cell life and death decisions Find the latest version: Review series Endoplasmic reticulum stress: Cell life and death decisions**. *J. Clin. Invest.* (2005.0) **115** 2656-2664. DOI: 10.1172/JCI26373.2656
58. Gonçalves GL, Costa-Pessoa JM, Thieme K, Lins BB, Oliveira-Souza M. **Intracellular albumin overload elicits endoplasmic reticulum stress and PKC-delta/p38 MAPK pathway activation to induce podocyte apoptosis**. *Sci. Rep.* (2018.0). DOI: 10.1038/s41598-018-36933-9
59. Koshikawa M, Mukoyama M, Mori K. **Role of p38 mitogen-activated protein kinase activation in podocyte injury and proteinuria in experimental nephrotic syndrome**. *J. Am. Soc. Nephrol.* (2005.0) **16** 2690-2701. DOI: 10.1681/ASN.2004121084
60. Pengal R, Guess AJ, Agrawal S. **Inhibition of the protein kinase MK-2 protects podocytes from nephrotic syndrome-related injury**. *Am. J. Physiol. Ren. Physiol.* (2011.0) **301** 509-519. DOI: 10.1152/ajprenal.00661.2010.-While
61. Okamura K, Dummer P, Kopp J. **Endocytosis of albumin by podocytes elicits an inflammatory response and induces apoptotic cell death**. *PLoS One* (2013.0). DOI: 10.1371/journal.pone.0054817
62. Pedigo CE, Ducasa GM, Leclercq F. **Local TNF causes NFATc1-dependent cholesterol-mediated podocyte injury**. *J. Clin. Invest.* (2016.0) **126** 3336-3350. DOI: 10.1172/JCI85939
63. Chen A, Feng Y, Lai H. **Soluble RARRES1 induces podocyte apoptosis to promote glomerular disease progression**. *J. Clin. Invest.* (2020.0) **130** 5523-5535. DOI: 10.1172/JCI140155
64. Huang CY, Tan TH. **DUSPs, to MAP kinases and beyond**. *Cell Biosci.* (2012.0). DOI: 10.1186/2045-3701-2-24
65. Denhez B, Rousseau M, Dancosst DA. **Diabetes-induced DUSP4 reduction promotes podocyte dysfunction and progression of diabetic nephropathy**. *Diabetes* (2019.0) **68** 1026-1039. DOI: 10.2337/db18-0837
66. Chen L, Wang Y, Luan H, Ma G, Zhang H, Chen G. **DUSP6 protects murine podocytes from high glucose-induced inflammation and apoptosis**. *Mol. Med. Rep.* (2020.0) **22** 2273-2282. DOI: 10.3892/mmr.2020.11317
67. Barr FA, Silljé HHW, Nigg EA. **Polo-like kinases and the orchestration of cell division**. *Nat. Rev. Mol. Cell Biol.* (2004.0) **5** 429-440. DOI: 10.1038/nrm1401
68. Zhang L, Wang Z, Liu R. **Connectivity mapping identifies BI-2536 as a potential drug to treat diabetic kidney disease**. *Diabetes* (2021.0) **70** 589-602. DOI: 10.2337/db20-0580
69. Sayyed SG, Hägele H, Kulkarni OP. **Podocytes produce homeostatic chemokine stromal cell-derived factor-1/CXCL12, which contributes to glomerulosclerosis, podocyte loss and albuminuria in a mouse model of type 2 diabetes**. *Diabetologia* (2009.0) **52** 2445-2454. DOI: 10.1007/s00125-009-1493-6
70. Darisipudi MN, Kulkarni OP, Sayyed SG. **Dual blockade of the homeostatic chemokine CXCL12 and the proinflammatory chemokine CCL2 has additive protective effects on diabetic kidney disease**. *Am. J. Pathol.* (2011.0) **179** 116-124. DOI: 10.1016/j.ajpath.2011.03.004
71. Romoli S, Angelotti ML, Antonelli G. **CXCL12 blockade preferentially regenerates lost podocytes in cortical nephrons by targeting an intrinsic podocyte-progenitor feedback mechanism**. *Kidney Int.* (2018.0) **94** 1111-1126. DOI: 10.1016/j.kint.2018.08.013
|
---
title: An integrated single cell and spatial transcriptomic map of human white adipose
tissue
authors:
- Lucas Massier
- Jutta Jalkanen
- Merve Elmastas
- Jiawei Zhong
- Tongtong Wang
- Pamela A. Nono Nankam
- Scott Frendo-Cumbo
- Jesper Bäckdahl
- Narmadha Subramanian
- Takuya Sekine
- Alastair G. Kerr
- Ben T. P. Tseng
- Jurga Laurencikiene
- Marcus Buggert
- Magda Lourda
- Karolina Kublickiene
- Nayanika Bhalla
- Alma Andersson
- Armand Valsesia
- Arne Astrup
- Ellen E. Blaak
- Patrik L. Ståhl
- Nathalie Viguerie
- Dominique Langin
- Christian Wolfrum
- Matthias Blüher
- Mikael Rydén
- Niklas Mejhert
journal: Nature Communications
year: 2023
pmcid: PMC10017705
doi: 10.1038/s41467-023-36983-2
license: CC BY 4.0
---
# An integrated single cell and spatial transcriptomic map of human white adipose tissue
## Abstract
To date, single-cell studies of human white adipose tissue (WAT) have been based on small cohort sizes and no cellular consensus nomenclature exists. Herein, we performed a comprehensive meta-analysis of publicly available and newly generated single-cell, single-nucleus, and spatial transcriptomic results from human subcutaneous, omental, and perivascular WAT. Our high-resolution map is built on data from ten studies and allowed us to robustly identify >60 subpopulations of adipocytes, fibroblast and adipogenic progenitors, vascular, and immune cells. Using these results, we deconvolved spatial and bulk transcriptomic data from nine additional cohorts to provide spatial and clinical dimensions to the map. This identified cell-cell interactions as well as relationships between specific cell subtypes and insulin resistance, dyslipidemia, adipocyte volume, and lipolysis upon long-term weight changes. Altogether, our meta-map provides a rich resource defining the cellular and microarchitectural landscape of human WAT and describes the associations between specific cell types and metabolic states.
Single-cell studies of human white adipose tissue (WAT) provide insights into the specialized cell types in the tissue. Here the authors combine publicly available and newly generated high-resolution and bulk transcriptomic results from multiple human datasets to provide a comprehensive cellular map of white adipose tissue.
## Introduction
White adipose tissue (WAT) is a uniquely plastic organ that can expand or shrink in response to caloric supply and demand. The ability to function across pronounced variations in tissue mass is governed by a plethora of resident and recruited cell types1. Disturbed WAT remodeling leads to changes in the cell composition of the tissue, which in turn increases the risk of developing insulin resistance, type 2 diabetes, and other cardiometabolic complications2. Defining the cellular landscape and microarchitecture of human WAT in health and disease is therefore of considerable clinical relevance.
To determine cell composition, single-cell technologies have been applied to WAT obtained from different depots. Together, these studies have identified novel specialized cells in adipose tissue such as: (i) adipogenic precursor cells with anti-adipogenic effects3–5, (ii) lipid-associated macrophages (LAMs) with central roles in metabolic health6 and (iii) adipocyte subtypes with distinct sensitivities to insulin7 or thermogenic effects8. However, a caveat with most single-cell studies is that issues related to data production restrict sequencing depths and the number of samples that can be processed. Consequently, most published reports in the adipose field are based on small cohort sizes and even the largest ones have, so far, included fewer than 15 individuals9,10. This, together with qualitative differences between technical platforms and bioinformatic approaches, limits the generalizability of the observed findings.
To address this, we performed a comprehensive meta-analysis of newly generated and publicly available data where we integrated single-cell (scSeq) and single-nucleus (snSeq) RNA sequencing results. Based on this, we created a cellular meta-map which we used to deconvolve spatial and bulk transcriptomic data from women and men spanning over a very broad range in age, BMI, and metabolic states. By mining this rich resource, we provide a nomenclature of cells residing in human WAT, their localization and how they relate to metabolic health.
## Human WAT contains four major cell classes
To define the cellular landscape of human WAT, we retrieved scSeq, snSeq and spatial transcriptomic (STx) results from ten published reports comprising samples from subcutaneous, visceral, and perivascular WAT6–15. We combined this with unpublished data from four additional cohorts (Massier et al. # 1–4), resulting in a total of 17 datasets across studies and depots (Supplementary Table 1). As displayed in Fig. 1a, together these included 401,320 quality-filtered cells/nuclei (hereafter referred to as objects) obtained from 103 samples of 83 donors spanning over a broad range in age (22–77 years) and body mass index (BMI, 17–55 kg/m2).Fig. 1A meta-map to define human WAT composition.a For each included cohort, number (n) and gender of subjects, age (years) and body mass index (BMI) ranges (min–max) as well as number of objects (cells/nuclei), method (single-nucleus Seq [snSeq], single-cell Seq [scSeq] or spatial transcriptomics [STx]) and Jaccard index are displayed. Massier et al. # 1–4 refer to data generated for the present meta-analysis and gray bars (n/a) indicate that no information was obtained. Boxes for age and BMI represent a range (min–max), boxplots are presented as interquartile range plus median and Tukey whiskers. Summarizing statistics are displayed in the right panels as mean ± S.D. b Network displaying nodes (subclusters from each study) and edges (marker gene overlap). Data were distributed into four major classes and named based on prominent marker genes. Node sizes are reflecting cluster proportions. The displayed network does not include results from Hildreth et al.11, as data from this study overlapped poorly with the others. c, d Cell class proportions comparing c) methods (snSeq vs. scSeq) and d depots (subcutaneous [sc] WAT vs. omental [om] vs. perivascular [pv]). Note that adipocytes are only available by snSeq. Data are shown as mean ± S.D. Statistical differences were calculated by two-sided Mann–Whitney U test between sc ($$n = 10$$) and om ($$n = 4$$) WAT or scSeq ($$n = 6$$) and snSeq ($$n = 8$$). Because of fewer cases, no statistics were calculated for adipocytes and pvWAT. e K-nearest-neighbor batch-effect test (kBET) and adjusted Rand index (ARI) for raw or integrated data using the indicated methods displayed according to method, depot, and cohort. BBKNN batch balanced k-nearest neighbors, scVI single-cell variational inference, rPCA reciprocal principal component analysis. Source data are provided as a *Source data* file.
To create a meta-map, we first defined cell clusters by processing all datasets individually using Seurat v4.116 (Supplementary Fig. 1a and Supplementary Data 1). Based on these results, we calculated a Jaccard index comparing cell cluster marker genes between studies. We found that all data, except the one from Hildreth et al.11, largely overlapped (Fig. 1a). This demonstrates that: (i) most cell types were present across the different cohorts/depots and (ii) snSeq, scSeq, and STx on average capture similar trends. Because STx is a spot-based approach that requires specific tools for data deconvolution17, we analyzed the Bäckdahl et al.7 study separately. For the remaining snSeq/scSeq datasets, we performed a high-level topological analysis, which revealed that clusters identified in the individual studies separated into four major cell classes: adipocytes, fibroblast and adipogenic progenitors (FAPs), vascular and immune cells (Fig. 1b). To validate our findings, we applied CellTypist18 to the immune cells which confirmed our annotations for this class (Supplementary Fig. 1b).
Our comprehensive classification allowed us to estimate the cellular composition of WAT (Fig. 1c, d). For this, we first used snSeq data, as this method, in contrast to scSeq, captures fat cells. After quantifying the number of adipocytes in each dataset, we next compared the proportions of the other cell classes between methods. We found that FAPs constituted the largest class of ~$40\%$ of the total cell population according to both snSeq and scSeq. This was followed by adipocytes and immune cells, which were present in a 1:1 ratio and together constituted ~$40\%$ of the cells. For adipocytes, this is in line with what has previously been reported in the literature19. About $15\%$ of the objects were vascular cells, but the recovery of this cell class was strongly influenced by methods where the proportion was approximately three times higher with snSeq compared to scSeq. This could be due to pre-selection bias introduced by enzymatic digestion as has been shown before for endothelial cells in mouse kidney20. Comparisons across depots showed that there were no significant differences in cell class abundance. As expected, scSeq studies where cells had been enriched for either FAPs (CD45−)13 or immune cells (CD45+)14 prior to sequencing displayed markedly higher proportions of these cell types compared to the rest of the datasets, supporting the validity of our classification (Supplementary Fig. 1a). Of note, for perivascular WAT, snSeq data were only available from one cohort implying that the generalizability of the results from this depot needs to be further validated.
## Cell class-based analyses successfully integrates data across studies
As much of the biological variability overlapped between cohorts, we performed data integration across studies. In contrast to the label-centric analysis presented in Fig. 1b, this approach facilitates the identification of rare cell types which risk being omitted in more sparsely sampled datasets. Many tools are available for integrative analyses and all of them introduce different biases where a trade-off between overfitting vs. insufficient integration must be considered21,22. In a first step, we applied single-cell annotation using variational inference (scANVI) as prior knowledge of cell type annotations can improve integration results23. In line with the findings in Fig. 1a, b, data from Hildreth et al. could not be integrated with scANVI and this dataset was therefore excluded from all subsequent analyses (Supplementary Fig. 1c). After removing Hildreth et al. the scANVI results mirrored the high-level topological analysis shown in Fig. 1b (Supplementary Fig. 1d). We therefore split the data into these four classes and tested different integrative frameworks. We included reciprocal principal component analysis (rPCA)16, batch balanced k-nearest neighbors (BBKNN)24, Harmony25 and single-cell variational inference (scVI)26, as all of them have benchmarked well in prior systematic comparisons27,28. To evaluate these methods across the main confounders (techniques, depots and cohorts), we calculated the adjusted Rand index (ARI)29, k-nearest-neighbor batch-effect test (kBET)30 and Local Inverse Simpson Index (LISI)25,31. In comparison to using all objects as input, we obtained enhanced integration results when splitting the data into immune cells, vascular cells, FAPs, and adipocytes. Our results showed that (i) Harmony performed well for FAP integration, (ii) scVI and rPCA provided the best results across all scores for methods, cohorts, and depots, and (iii) kBET showed overall the lowest rejection rates for scVI (Fig. 1e, Supplementary Fig. 2a, b). This in combination with a previous report suggesting that scVI works best for complex data27, prompted us to apply this method to all four cell classes separately. Given that the included datasets varied in object number, we capped them to contain similar object counts to reduce bias and enhance integration (Supplementary Fig. 2c, d). Based on this, we included an optimal number of objects for each cell class which resulted in cell clusters represented by most studies and methods (Supplementary Fig. 2d, e). With these criteria, we created WAT annotation models which are publicly available and can be applied in future studies (see Methods).
## Immune cells with different origins and activation states are present in human WAT
The immune cell class integrated well across techniques, studies, and depots (Fig. 1e, Supplementary Fig. 2b). As displayed in Supplementary Figures 3a, b, we annotated two major groups containing distinct subpopulations of: (i) T, natural killer (NK), and NKT cells ($28.8\%$) and (ii) monocytes, macrophages, and dendritic cells ($67.3\%$) as well as three minor populations including mast ($2.94\%$), B ($0.87\%$), and plasma B cells ($0.16\%$).
The large number of objects allowed us to further dissect these groups separately to provide a high-resolution map of the different cell populations. In the major lymphocyte group, we identified 11 different clusters which were denoted lyC0-10, in order of abundance from high to low (Fig. 2a, b, Supplementary Table 2 and Supplementary Data 2). These included subtypes of (i) Th1-polarized (lyC0), tissue-resident memory (TRM, lyC01), naive/early differentiated (lyC04), and naive/regulatory (TREG, lyC08) CD4+ T cells (ii) CD8+ T cells including early- (lyC02) and late-differentiated (lyC03) cells, (iii) NKT cells (lyC05), as well as (iv) CD16+ (lyC06) and CD16− (lyC09) NK cells. Multiple cell classes, including TREG and TRM, could be further subdivided based on their differentiation states (Fig. 2a, lower panel). We confirmed these data by flow cytometry (Supplementary Fig. 3c) using a panel of antibodies identified in the transcriptomic analyses (see Methods). Comparisons across depots revealed that all identified T and NK cell populations were present in subcutaneous, omental, and perivascular WAT albeit with slightly different proportions (Fig. 2b). For example, Th1 CD4+ T cells (lyC0) were more abundant in omental while late-differentiated CD8+ T cells (lyC03) enriched in subcutaneous WAT (Fig. 2b). These observations were confirmed by deconvolution of bulk transcriptomic data in two independent cohorts32,33 (Fig. 2c).Fig. 2Analyses of the WAT immune cell panorama reveal novel subtypes.a Nomenclature (upper panel) and expression patterns of selected marker genes (lower panel) for T, NKT, and NK cells (lyC0-10). b Proportions (%) of T, NKT, and NK cells (lyC0-10) in subcutaneous (sc), omental (om), and perivascular (pv) WAT depots. c The enrichment of lyC0 in omental and lyC03 in subcutaneous WAT was supported by deconvolution of bulk transcriptomic data from Arner et al.32 (left panel) and Krieg et al.33 (right panel). p values were calculated by two-sided Wilcoxon signed-rank test. d Same visualization as a, but for monocytes and macrophages. e Selected marker gene expression profiles for omental-enriched myC08 and myC12. f Same visualizations as b, but for monocytes and macrophages. g Same as c, but for myC07, myC08, and myC12. DC dendritic cells, diff differentiated, LAM lipid-associated macrophages, MMe metabolic-regulated macrophages, Mo monocytes, Mox redox-regulatory metabolic macrophages, NK natural killer cells, NKT natural killer T cells, Th helper T cells, TREG regulatory T cells, TRM tissue-resident memory T cells. Source data are provided as a *Source data* file.
We next classified the myeloid group. This resulted in 16 different clusters (myC0-15) including macrophages, non-classical (myC05) and classical (myC14) monocytes, as well as class 2 dendritic cells (myC03) (Fig. 2d, Supplementary Table 2 and Supplementary Data 2). In-depth analyses of macrophages revealed several M2-like subpopulations (myC0-01, myC04, myC07-09, myC11-13) out of which two (myC08 and −12) have not been previously reported in WAT (Fig. 2d, e). We also identified mixed M1/M2-like (myC06), lipid-associated (LAM, myC02), metabolic-regulated (MMe, myC10) and redox-regulatory metabolic (Mox, myC15) macrophages (Fig. 2d). Similar to lymphoid cells, our flow-cytometry-based follow-up studies confirmed the presence of several of the identified myeloid subpopulations in subcutaneous WAT (Supplementary Fig. 3d). Analyses over the three depots revealed that LAMs (myC02) were present in both subcutaneous and omental WAT but were virtually absent in the perivascular depot, while Mox (myC15) were only present in the subcutaneous region (Fig. 2f). The omental region was enriched for different types of M2-like cells (myC07-08 and myC12), observations which were confirmed by deconvolution of bulk transcriptomic data (Fig. 2g). In addition to these larger groups, we also analyzed B and mast cells (Supplementary Fig. 3e, f, Supplementary Data 2). B cells separated into activated (bC01, characterized by upregulation of genes related to regulation of T cell-mediated cytotoxicity) and non-activated states (bC0). While mast cells could not be clearly subdivided, there were trends in the data indicating that both MCT (tryptase-positive) and MCTC (positive for chymase and carboxypeptidase) cells were present.
## WAT contains vascular cells with angio- and adipogenic expression profiles
For vascular cells, our data integrated well between techniques, studies, and depots (Fig. 1e, Supplementary Fig. 2b). We identified 12 distinct cell populations (vC0-11), which were broadly split into four major groups including several blood (vC0-03, vC05, vC08-11), and lymphatic (vC06) endothelial cells as well as vascular smooth muscle cells (vC07) and pericytes (vC04) (Fig. 3a, Supplementary Table 2 and Supplementary Data 2). Based on nomenclatures from several scSeq atlases, blood endothelial cells were further classified into the classical capillary (vC0), venous (vC01 and vC11), mixed capillary/venous (vC03), and arterial (vC02) subpopulations (Fig. 3b). Three subtypes (vC08-10) separated from the major endothelial cell populations. Top marker genes for these revealed that vC08 displayed a considerable overlap with the herein-identified myC08 (Jaccard index: 0.33), indicating that they may represent cells with similar function or intermediate cell states. Two of the shared marker genes were TIMP1 and ITLN1 which encode secreted proteins inhibiting neovascularization34,35, suggesting that these cells may exert anti-angiogenic effects (Fig. 3c). vC09 was enriched for genes previously described in “early endothelial progenitor cells” (e.g., TYROBP, FCER1G), a hematopoietic cell type that promotes angiogenesis via paracrine mechanisms (Fig. 3d)36,37. Thus, human WAT may contain vascular cells that either stimulate or inhibit vascularization. One vascular subtype (vC05) expressed multiple marker genes for committed preadipocytes (e.g., CXCL14, APOD, and CFD) and was the only vascular population that expressed PDGFRA (Fig. 3e). In mice, all peri-aortic adventitial fibroblasts are PDGFRA+ and give rise to perivascular adipocytes15. It is therefore possible that vC05 represents an intermediate cell state between endothelial and adipocyte precursor cells. Admittedly, this notion needs further functional studies. Comparisons between depots revealed that all cell types, including vC05, were present in the three regions. However, lymphatic (vC06) and TIMP1-expressing (vC08) endothelial cells were enriched in omental and perivascular WAT (Fig. 3f), observations which were confirmed by deconvolution of bulk transcriptomic data (Fig. 3g).Fig. 3The vascular cell class contains mixed and intermediate cell states.a, b Nomenclature and visualizations of selected marker genes for vascular cells (vC0-11), including a UMAPs and b violin plots. c–e Multiple UMAPs of marker genes for three subtypes of blood ECs (vC08, vC09, and vC05, respectively). f The proportions (%) of different vascular subtypes in subcutaneous (sc), omental (om), and perivascular (pv) WAT depots. g The proportion of vC06 and vC08 in sc and om depots was supported by deconvolution of bulk transcriptomic data from Arner et al.32. ( upper panel) and Krieg et al.33 (lower panel). p values were calculated by two-sided Wilcoxon signed-rank test. EC endothelial cells, VSMC vascular smooth muscle cells. Source data are provided as a *Source data* file.
## Subcutaneous FAPs include subpopulations distinguished by degree of commitment
Although FAPs include fibroblasts and stem cells at different stages of commitment, an established nomenclature for this class is still lacking. This, in combination with their pronounced heterogeneity and depot-specific contribution to WAT expansion (at least in mice38), prompted us to analyze this cell class depot-by-depot. The subcutaneous FAPs separated into 17 clusters (sfC0-16) distributed into four smaller (sfC03, −11, −13, and −16) and one very large ($86.5\%$ of all FAP objects) group (Fig. 4a and Supplementary Data 2). The former included mesothelial-like cells (sfC11), CD74+ stromal cells (sfC13) reported to have antifibrotic properties39, and late committed preadipocytes (sfC16). One cluster (sfC03) enriched for cell adhesion molecules and was the only FAP that did not express CD34, PDGFRA, or PDGFRB (Fig. 4b).Fig. 4FAPs display different levels of commitment in human WAT.a Nomenclature and proportions for subcutaneous FAPs (sfC0-16) including a UMAP with selected marker genes (left panel) and a stacked bar chart displaying the proportion (%) of different subtypes in subcutaneous white adipose tissue (WAT) (right panel). b, c Selected FAP marker gene expression profiles displayed in UMAPs. d Pseudo-time trajectory analysis initiated from the CD55/PI16-enriched cell cluster (sfC02). Two main trajectories were discovered: route 1 (upper) and route 2 (lower). e–g CD55+ positive human adipose-derived stem cells were e analyzed by flow-cytometry and f imaged before/after adipogenic induction in vitro. Nuclei are stained by Hoechst (blue) and lipid droplets by BODIPY (green). Experiment was repeated three times with similar results. Scale bar is 20 μm. g deconvolution of bulk RNAseq data from these cells shows how the expression of marker genes for FAP subtypes in panel a vary during adipogenesis (colors are matched in a and g). h Heat map displaying similarities (Jaccard index) of gene expression profiles between inguinal WAT from mice (P1.1 to P4) and subcutaneous FAPs (sfC0-16). Note that human cells displaying overlap with mouse FAPs are only found in route 1. Explanatory legend is visualized in j. i Same as a, but for omental FAPs (ofC0-14). j Same as in h; but for comparisons of subcutaneous and omental human FAPs. k Flow cytometric analysis of CD55+, APOD+, CD74+, and EZR+ FAPs from stromal vascular cells (CD45−, CD31− and CD34+) of subcutaneous and omental WAT, respectively. Percentages represent the frequency of all gated live single cells in the representative sample. APC adipose precursor cell, CPA committed preadipocytes, FAP fibroblast and adipogenic progenitor cells, MSL mesothelial-like cells. Source data are provided as a *Source data* file.
In contrast to the smaller clusters, the larger group included multiple extracellular matrix- and CD55/PI16-expressing cells (Fig. 4a, c, Supplementary Fig. 4a). The latter marks a universal fibroblast population that can differentiate into more specialized cells40, including adipocytes13. Therefore, we used them as a starting point in pseudo-time analyses which identified two distinct trajectories (routes 1–2) branching into opposite directions at the first step (Fig. 4d). To test if these trajectories were recapitulated during de novo adipocyte formation, we deconvolved bulk RNAseq data from human CD55+ cells undergoing adipogenesis in vitro41 (Fig. 4e, f). Our results show that the CD55/PI16 gene cluster (sfC02) was highly expressed prior to differentiation and was markedly downregulated upon adipogenic induction (Fig. 4g). This was followed by a transient upregulation of route 1-localized clusters (sfC04-05 and sfC07-08), which thus constitute transient cell states. Route 1 ended in sfC0, a cell population enriched for genes that were upregulated at late stages of differentiation (e.g., APOD and CFD) and most likely represent committed preadipocytes (Fig. 4a, d, g). In contrast to route 1, the cell populations in route 2 (sfC01, sfC06, sfC09-10, sfC12, and sfC14-15) could not be detected during in vitro adipogenesis. These results were recapitulated in three additional cell models41–43 (Supplementary Fig. 4b), suggesting that under standard in vitro conditions, CD55+ adipocyte precursor cells cannot recapitulate the full FAP heterogeneity observed in subcutaneous WAT.
To improve our FAP annotations and determine inter-species similarities, we systematically compared our findings with scSeq data obtained from mouse inguinal WAT (which is frequently used as a proxy for human subcutaneous WAT)3,5,44. We found that mouse early adipocyte progenitors (P1-1 and P1-2) overlapped with CD55/PI16-expressing adipocyte precursors (sfC02), while more committed preadipocytes (P2-1 and P2-2) resembled sfC0 (Fig. 4h). *The* gene expression profile of the P3 population, which has been defined as FAPs with anti-adipogenic properties (AREG), showed overlap with one of the transient cell clusters along route 1 (sfC04). There were no overlaps with any of the route 2-localized FAPs indicating that clusters along this trajectory are not overlapping with the Lin-/SCA+ cells pre-selected in the studies by Schwalie et al.3 and Dong et al.5. However, DPP4, which is an established marker of pro-adipogenic FAPs in mouse inguinal WAT44, projected onto route 2 and was enriched in sfC06 and −10 (Supplementary Fig. 4c). This demonstrates that specific FAPs in this trajectory overlap with murine PDGFRB+ cells with adipogenic capacity.
## Omental and subcutaneous FAP signatures largely overlap
The omental FAPs separated into 15 clusters (ofC0-14) (Fig. 4i, Supplementary Data 2). To annotate these, we compared their signatures to subcutaneous FAPs. We found that several cell populations shared overlapping marker genes between depots (Fig. 4j, Supplementary Fig. 4c-d). Thus, CD55/PI16-expressing adipose precursors (sfC02) matched ofC07, and APOD/CFD-expressing committed preadipocytes (sfC0) corresponded to ofC02. All FAPs in route 2 showed strong similarities to three omental clusters (ofC05, ofC09, and ofC14). CD74 was highly enriched in sfC13 and ofC10 (Supplementary Fig. 4d), but the overall Jaccard index between these two cell types was modest. The mesothelial-like cell signature in subcutaneous WAT (sfC11) was recapitulated in multiple omental FAPs (ofC03, ofC06, and ofC11-12). Of note, DPP4 was enriched in these omental clusters, but not in the corresponding subcutaneous cells, indicating that this gene marks different subsets of FAPs in the two depots (Supplementary Fig. 4c), a notion previously suggested in mice44. Flow cytometric analysis of subtype-specific surface markers in the stromal vascular fraction of WAT biopsies from subcutaneous and omental WAT confirmed the presence of sf02/ofC07 (CD55+), sfC0/ofC02 (APOD+), sfC13/ofC10 (CD74+) and mesothelial-like cells (EZR+) in both depots (Fig. 4k, Supplementary Fig. 4d). We did not find any expression profiles similar to the fibro-inflammatory progenitors previously described in mouse epigonadal WAT4 (Supplementary Fig. 4e). Data from the perivascular depot separated into eight clusters (pf0-pf07) (Supplementary Fig. 4f). While pfC0 resembled ofC07 (adipose precursors) and ofC12, pfC03 was linked to ofC02 (committed preadipocytes). The remaining clusters did not display strong similarities to any of the other FAP subtypes identified in subcutaneous WAT.
## Adipocytes display inconsistent heterogeneity between studies
In contrast to scSeq, snSeq allows for analyses of mature adipocytes. In the present study, we analyzed snSeq datasets from subcutaneous ($$n = 4$$), omental ($$n = 2$$), and perivascular ($$n = 1$$) fat depots. As with the other cell classes, we jointly analyzed objects classified as fat cells first. However, this resulted in poor ARI, kBET, and LISI indices (Fig. 1e, Supplementary Fig. 2b), suggesting that complete data integration was not possible to achieve between studies and depots (Supplementary Fig. 5a). As a next step, we analyzed the depots separately. This, however, did not improve our integration as most clusters separated according to studies rather than biological similarities between datasets (Supplementary Fig. 5b).
A possible reason for incomplete data integration is that there are few common features between studies, even in the low-dimensional space21. To test if data harmonization was influenced by a limited degree of overlap, we next analyzed the studies individually. We focused on data produced by Emont et al.9 and our own results (Massier et al. # 1) generated in subcutaneous WAT as they contain the largest number of subjects and objects. We selected the top 50 marker genes for all clusters and found that adipocytes displayed lower fold-changes compared to the other cell classes (Fig. 5a, left panel). This indicates that the degree of cell heterogeneity is less pronounced in adipocytes. However, there may still be consistencies between the studies. We, therefore, transferred cell type classifications between datasets. In comparison with the other cell types where marker genes displayed clear enrichments and overlap between studies, we found a low overlap and a limited set of reproducible adipocyte marker genes across studies (Fig. 5a, right panel). These included a cluster of genes encoding proteins involved in lipid metabolism e.g., ABCD2, ACACB, CD36, DGAT2, GPAM, HACD2, and LPL (Fig. 5b).Fig. 5Adipocyte snSeq data display inconsistent marker gene overlaps in WAT.a Single-nucleus sequencing (snSeq) data from Emont et al. (# 1) and Massier et al. (# 1) were analyzed according to top marker genes for adipocytes, FAPs, vascular, and immune cells, respectively. In comparison with other cell classes, adipocyte marker genes displayed lower fold-changes (left panel) and a limited overlap (right panel). b Representative examples of adipocyte marker genes in subcutaneous WAT displaying overlap between the indicated studies. c Dendrograms of snSeq, bulk RNAseq of isolated mature adipocytes (from the FANTOM5 atlas46,81 and Harms et al.45) and spatial transcriptomics7 (STx) (upper panel). Comparisons of scSeq, snSeq, bulk RNAseq, and STx data of subcutaneous white adipose tissue from the same individual (lower panel). d Heatmap of adipocyte marker genes with a >50-fold enrichment in adipocytes vs. other tissues included in the FANTOM 5 atlas. Results are shown for each study as well as for the combined snSeq data after integration. e Same as in d, but for genes with discordant expression (∣Δz-score∣>5) comparing STx and snSeq data. Source data are provided as a *Source data* file.
Because of the modest heterogeneity and low reproducibility across studies of adipocytes, we benchmarked the snSeq data and our recently published STx-based results7 against rRNA-depleted45 and cap-trapped41 bulk RNAseq of isolated human mature fat cells. This revealed that the STx data were positioned close in space to the bulk results, while all snSeq datasets clustered together with considerably fewer similarities to the other samples (Fig. 5c, upper panel). To test the reproducibility of these results, we generated snSeq, STx, and bulk RNAseq from the same individual and repeated the analyses. This confirmed that STx correlated better with bulk RNAseq data than snSeq (Fig. 5c, lower panel). We characterized this further by using the FANTOM5 expression atlas46, where we identified 218 transcripts that were enriched at least 50-fold in adipocytes compared to other cell types (Supplementary Fig. 5c). *These* genes included multiple established adipocyte markers and we analyzed their expression levels in data from the three platforms (Fig. 5d). We observed marked variations in the expression between methods. By filtering based on genes that displayed a high discordance (>5 |Δ z-scores|) between snSeq and STx, we found that several STx adipocyte subpopulation marker genes were less expressed or not detected by snSeq, but highly abundant in bulk RNA and STx data (Fig. 5e, Supplementary Fig. 5d). These included marker genes of previously described7 adipocyte subtypes (Fig. 5e). Conversely, PPARG, WDPCP, and PDE3B were higher in snSeq. These types of platform-specific biases may contribute to the low reproducibility and heterogeneity scores as has been shown previously for adipocytes47. In subsequent analyses, we, therefore, pooled the adipocyte snSeq data into one class.
## Multiple FAP-relayed signals target M2-like macrophage subpopulations
Having defined cell types present in WAT through our meta-map, we inferred their communication routes using CellChat48. We first summarized the expression of ligands vs. receptors across the different cell types and next identified specific cell-cell interactions via ligand-receptor patterns in the subcutaneous and omental depots (Fig. 6a, b, Supplementary Fig. 6a). Our analysis suggested that FAPs were mainly relaying information, which in turn was primarily received by M2 macrophages. Clustering of pathways by functionality allowed us to link these FAP-myeloid communication routes to for example complement, chemerin, and IL16 signaling (Supplementary Fig. 6b). Another striking finding was that Mox (myC15), which are only present in subcutaneous WAT and characterized by exceptionally high expression levels of various pro-inflammatory cyto- and chemokines (Fig. 6c), both relayed and received CCL, CXCL and TNF signals. Thus, these cells received CCL5 input from different CD8+ T cells (in particular lyC02-03) via CCR1, and signaled to vascular cells (vC01 and vC11) via ligands (CCL2, CXCL2, CXCL8) recognized by ACKR1 (Fig. 6d). For TNF, Mox signaled to adipocytes and myeloid cells via TNFRSF1A, and to CD8+ T cells/NK cells (lyC03, −06, −09) and endothelial cells (vC01-03) via TNFRSF1B (Fig. 6d). The validity of these in silico analyses was supported by the observation that, the KIT receptor was only present in mast cells49 while periostin was solely relayed from pericytes50 (Supplementary Fig. 6c).Fig. 6FAPs send multiple signals which are received by M2-like macrophages.a Incoming and outgoing interaction strengths for FAPs/adipocyte, vascular, myeloid, and lymphoid cells in subcutaneous (upper panels) and omental (lower panels) WAT. Selected cell types are indicated. b Subcutaneous (upper) and omental (lower) cell-cell communication predictions for clusters identified in Supplementary Fig. 6a. Lines indicate interactions between cell types where the strength is proportional to the line width and the color defines the sending subpopulation. c Violin plots for selected marker genes enriched in the Mox cluster (myC15). d Contribution of each ligand-receptor (L-R) pair to the overall cell-cell communication strength. Blue bars are strongly contributed by Mox (myC15). e Predicted cell-cell interactions for the indicated L-R pairs. Line widths and colors indicate signaling strengths and sending subpopulations, respectively. Cell types are in numerical order as shown in the left-most panel. Note that the myeloid and lymphoid clusters ends with mast cells and B cells, respectively. Source data are provided as a *Source data* file.
## Spatial distribution of FAPs suggest distinct adipogenic niches
To add a spatial dimension to our meta-map, we re-analyzed previously generated STx data of subcutaneous human WAT from ten individuals7 using six deconvolution tools: DestVI, stereoscope, Tangram, RCTD, cell2location and SPOTlight. Although all six frameworks provided similar results (Supplementary Fig. 7a), we opted to use cell2location as it has been shown to be a robust deconvolution tool17. In contrast to FAPs, lymphocytes, and adipocytes, we found that vascular and myeloid cells were concentrated in specific areas of the tissue (Fig. 7a). For the latter, these regions included classical monocytes (myC14) and M1-like macrophages (LAMs [myC02], Mmes [myC10] and Mox [myC15]) but were devoid of M2-like macrophages (Fig. 7a). We confirmed clustering of LAMs in specific areas of WAT by immunofluorescence (Fig. 7b), thereby confirming previous findings in mice6. To complement these studies, we systematically identified within-spot colocalization patterns. For this, we correlated deconvolution scores for each cell type across all spots and identified three clusters (Fig. 7c, Supplementary fig. 7b). These included endothelial cells as well as two groups of FAPs. Further analyses showed that one FAP subtype (sfC12) was enriched in areas close to endothelial cells (Fig. 7d). This contrasted with the other FAPs, which displayed strongly negative relationship to vascular cells. In addition, another FAP subtype (sfC08) was found close to LAMs (myC02) (Fig. 7d). These associations were present in multiple individuals (Supplementary Fig. 7c) and were confirmed by immunofluorescence (Fig. 7e). Altogether, this suggests that FAPs have specific tissue distributions, possibly to form different types of adipogenic niches. Fig. 7WAT contains niches populated by specific sets of cells.a Representative sections from two subjects displaying areas densely populated by myeloid cells (left panels). The indicated regions are magnified where the hematoxylin & eosin stain is shown in the middle and the Visium slide myeloid score is shown. Deconvolution scores for myeloid subpopulations in the inlay regions are shown in the right panels. Boxplots are presented as interquartile range plus median and Tukey whiskers; scale bar is 100 µm. b Representative immunostaining of human subcutaneous white adipose tissue incubated with antibodies targeting LAM marker proteins TREM2 and CD9, respectively. Nuclei were stained with Hoechst. The experiment was repeated three times with similar results. Scale bar is 100 μm. c Pair-wise correlation heatmap displaying within-spot associations between cellular subpopulations. Full heatmap is shown in Supplementary Fig. 7b. d Representative sections displaying the distributions of selected subpopulations of FAPs (sfC08 and −12), myeloid (myC02), and vascular cells (vC01). Scale bar is 500 µm. e Representative immunostaining of human subcutaneous white adipose tissue incubated with antibodies targeting the sfC12 marker protein SLIT2 as well as the endothelial protein CD31. Nuclei were stained with Hoechst. The experiment was repeated seven times with similar results. Scale bar is 50 μm in the merged panel and 10 μm in the inlay. Source data are provided as a *Source data* file.
## Two clusters of cells are reciprocally associated with metabolic health
A major weakness of most transcriptional studies with single-cell resolution is that the clinical relevance of the identified cell types is difficult to determine. To address this limitation, we used single-cell marker genes to deconvolve subcutaneous WAT bulk transcriptomic data from eight independent studies comprising a total of 864 individuals. As displayed in Fig. 8a and further detailed in Supplementary Table 3, together these cohorts included both adult men and women with a broad range in age, BMI, waist-to-hip ratio, circulating triglycerides, HDL-cholesterol, and leptin levels as well as insulin sensitivity estimated by HOMAIR. We also retrieved data on fat cell volume and lipolysis (basal and isoprenaline-stimulated), which are two measures linked to metabolic health51. By clustering the correlations between cell types and the mentioned parameters, we identified major trends in the data. Thus, while cluster A included six cell types that associated positively with a metabolically beneficial profile, cluster B contained 15 cell types that correlated negatively with the same parameters (Fig. 8b). Cluster C included 37 cell types which displayed weaker links to metabolic states. Of note, age was not associated with any specific cell type (Fig. 8b).Fig. 8Deconvolution of transcriptomic data reveals cluster-specific clinical associations.a Bulk transcriptomic data from eight cohorts were retrieved, the distribution in age, BMI, and HOMA-IR are shown in the left panel. Summary statistics are detailed in the right panels. b Heatmap displaying the association between individual cell types (denoted by numbers and color according to the classification in Figs. 2–4) with: anthropometric measures, HOMA-IR, circulating levels of HDL-cholesterol, triglycerides, and leptin (all in the fasted state), fat cell volume as well as adipocyte lipolysis (basal, isoprenaline-stimulated and isoprena-stimulated/basal). Three main clusters (A–C) were identified where cluster A and B are magnified in the right panel. c Representative Forest plots displaying the associations between individual measures and cell types. Data are shown as correlations with $95\%$ confidence intervals for each study and summarized using both common and random effects models. For all displayed data, p values were <0.0001. d Stability of clusters A and B were determined in the two indicated cohorts where WAT bulk transcriptomes were generated before and two years following bariatric surgery. e Effects of weight loss induced by bariatric surgery in two cohorts. Panels display deconvolution scores for the indicated cell subpopulations. p values were calculated by two-sided paired sample t test ($$n = 15$$; Petrus et al.54 and $$n = 37$$; Kerr et al.53) and boxplots are presented as interquartile range plus median and Tukey whiskers with individual, paired data points. Source data are provided as a *Source data* file.
In more detailed analyses, we found that cluster B was overrepresented by immune cells, e.g., LAMs (myC02), Mmes (myC10), DC2 (myC03), six out of nine M2-like macrophages (myC0-01, myC07-08, myC11 and myC13), early differentiated CD8+ (lyC02) and CD16+ NK cells (lyC06) (Fig. 8b, c). These results are in line with previous data demonstrating that specific immune cell subtypes are enriched in states of insulin resistance and obesity1. In contrast to cluster B, cluster A enriched for FAPs and vascular cells (Fig. 8b, c). The latter included blood endothelial capillary cells (vC0) and could be due to capillary rarefaction, a phenomenon characterized by reduced capillary beds previously reported in WAT from people with obesity52. For FAPs, we observed that CD55/PI16-expressing adipose precursors (sfC02) were the only cells present in cluster B while some intermediate/late states (sfC04, −14, and −16) were found in cluster A. This suggests that adipogenesis may be impacted by metabolic health at several different levels. To reveal associations between adipocytes and clinical measures, we used STx marker genes as this platform reflects the transcriptional profiles of adipocytes more closely than snSeq. In concordance with previous results obtained in a small cohort7, we found that AdipoPLIN correlated negatively with BMI, insulin resistance and circulating leptin levels while AdipoLEP was positively associated with all these measures (Supplementary Fig. 8a). In contrast, AdipoSAA displayed weak correlations with all investigated parameters.
Finally, to test if clusters A and B were impacted by weight changes induced by bariatric surgery, we retrieved WAT transcriptomic data from two of the studies where subjects were followed two and five years post-operatively ($$n = 52$$)53,54. Our results revealed that in both cohorts, clusters A and B were stable and normalized by weight loss (Fig. 8d, Supplementary Fig. 8b). One exception was blood endothelial capillary cells (vC0), which remained unaltered by weight loss (Fig. 8e). These data53 also allowed us to assess effects of weight regain comparing follow-ups at two and five years. As displayed in Supplementary Figure 8c, subdividing subjects into tertiles based on long-term weight regain or stability showed that except for vC0, several cell subtypes followed the changes in body weight. Altogether, these observations suggest that the cellular landscape of WAT is dynamic and mirrors alterations in fat mass.
## Discussion
Our meta-analysis integrates existing and newly generated single-cell data with bulk sequencing of in vitro adipogenesis and intact WAT from large clinical cohorts. By overlapping this with spatial transcriptomics and data from human and murine single-cell resources generated in different organs, we provide a meta-map of cell types and their spatial organization in human WAT. Altogether, this allowed us to define >60 distinct cell types including immune cells with diverse activation states, intermediate vascular cell types with hybrid transcriptional profiles, and FAPs displaying distinct tissue localization and different levels of adipogenic commitment.
To create a cellular nomenclature across adipose depots, we included data from subcutaneous, omental, and perivascular WAT in our meta-analysis. Based on depot comparisons, we found that although proportions differed, most cell subpopulations were present in all three regions. This was also true for FAPs, even though they have been suggested to contribute to depot-specific differences in tissue growth. For example, both adipose precursors and committed preadipocytes were found in all three sites. However, within subcutaneous WAT, the FAPs displayed distinct localizations where some were found close to vessels and others were adjacent to specific macrophages. The latter is of interest given that our ligand-receptor analyses suggested that FAPs relayed multiple signals to myeloid cells. Apart from these quantitative and microarchitectural aspects, we also observed that a limited number of cell types were unique to either region, including Mox and a few M2-like macrophage subtypes. Although we corroborated these results by deconvolving bulk transcriptomic data from paired samples of subcutaneous and omental WAT, they need to be validated in additional cohorts and the function of these cells needs to be determined. In addition, the spatial analyses presented herein were only performed in subcutaneous WAT and should be followed-up in other depots.
In our meta-analysis, adipocytes comprised ~$20\%$ of the WAT cell population and displayed a transcriptional fingerprint that was distinct from FAPs, immune and vascular cells. However, in contrast to these cell classes, adipocyte data were exclusively generated using snSeq and the results displayed less pronounced and reproducible heterogeneity between studies. A possible reason for this lack of consistency is that snSeq preferentially detects nascent and long transcripts47. This is further supported by our benchmarking of technical platforms where snSeq data, in comparison to STx results, associated poorly with the transcriptional signature of isolated fat cells. In fact, numerous adipocyte subtype marker genes identified by STx (e.g., LEP, PLIN4, SAA1, RBP4)7 were not, or very weakly, detected by snSeq. We therefore conclude that combined analyses using different technical platforms are required to confidently identify adipocyte subtypes.
Although we have combined studies to obtain data at the single-cell level from approximately 100 samples, we have not determined the influence of age, anthropometric measures, and disease states on these results. Instead, we created a cartography of cells present in WAT and used this framework to deconvolve bulk transcriptomic results from over 860 samples. This expands previous efforts55,56 and allowed us to link the identified cell populations to multiple clinical and WAT parameters. More specifically, we show that CD55/PI16-expressing adipose precursors as well as a large set of immune cells, including LAMs and Mmes, were enriched in individuals with markers of a pernicious metabolic phenotype, i.e., subjects with large fat cell volume, high waist-to-hip ratio, high HOMA-IR and impaired lipid mobilization. Conversely, a group of intermediate FAPs and capillary endothelial cells associate negatively with the same parameters. Additional analyses in cohorts before and after bariatric surgery, revealed that multiple cell clusters enriched in people with obesity and insulin resistance are normalized upon weight loss. Together, these data extend previous results on WAT expansion by confidently identifying specific cell types linked to increased inflammation as well as attenuated adipogenesis and vascularization. Nevertheless, a limitation with the present work is that we did not investigate longitudinal data following other types of interventions including, life-style-related changes, and/or pharmacological treatments.
Taken together, in this meta-analysis of 17 datasets and >800 bulk transcriptional profiles from eight clinical studies, we have comprehensively defined the cellular composition of human WAT in health and metabolic disease. Thus, by jointly analyzing data from multiple types of studies, we have created a framework that is easily accessible and includes additional tools for WAT analyses such as new models in CellTypist. We, therefore, provide a rich resource to facilitate future studies of specific WAT-resident cell types in relation to aspects not investigated herein, such as ethnic differences and the impact of therapeutic interventions.
## Inclusion and Ethics
This study was performed in agreement with the Declaration of Helsinki. Studies of cohorts presented for the first time herein (Massier et al. # 1–4), were approved by the regional ethics boards in Stockholm (clinical trials identifiers: NCT01785134 and NCT01727245) and Leipzig (approval numbers: 159-12-21052012 and $\frac{004}{21}$-ek) and explained in detail to each participant who gave informed written consent. For retrospective analyses of published data, the studies have been approved by the respective ethical boards where informed written consent was obtained from all participants. For data detailed in Fig. 8d, e, primary outcomes for both clinical trials (NCT01785134 (DEOSH) and NCT01727245 (NEFA)) have been described at clinicaltrials.gov and previously published7,53,57–59, and the studies have been completed.
## Sample collection and preparation
Samples of cohort 1 and 3 were collected in Stockholm (Sweden) and processed by T.W. in Zürich (Switzerland). Adipocyte nuclei were isolated following a modified nuclear isolation protocol60. In total, 50 mg of fresh or frozen WAT was first minced into 1–3 mm pieces and then homogenized on ice in $0.1\%$ CHAPS in CST buffer supplemented with 0.2 U/μl RNAase inhibitor (RI) using a Dounce homogenizer. After homogenization, samples were left on ice for five minutes following which PBS supplemented with BSA and 0.2U/μl RI was added to obtain a final concentration of $1\%$ BSA. The lysates were filtered through 40-μm cell strainers and centrifuged at 500 × g for five minutes at 4 °C. The nuclei pellets were resuspended with $1\%$ BSA in PBS supplemented with 0.2 U/μl RI and centrifuged again at 50 × g for five minutes at 4 °C. This step was repeated once more. After the final resuspension, nuclei were filtered through 20-μm cell strainers and loaded directly on a 10X Chip G. 10X-libraries were prepared with the Chromium Single-Cell v3.1 reagent kit following the manufacturer’s protocol (10X Genomics). Suspensions containing around 1200 nuclei per μl were loaded on Chip G followed by reverse transcription to obtain cDNA, which subsequently was amplified and used for library construction. After preparation, the libraries were sequenced on a NovaSeq 6000 platform (Illumina). For data analysis, the human genome assembly GRCh38.p13 was used. Mapping was performed using 10x Genomics Cell Ranger (v6.0.2). CellBender (v0.2.0)61 was used on ‘raw_feature_bc_matrix’ to remove empty droplets and ambient RNA; scDblFinder (v1.5.11)62 was applied to exclude potential doublets. Downstream analyses were performed as all other included studies (see below).
Samples of cohort 2 and 4 were collected and processed in Leipzig (Germany) by P.A.N.N. Paired samples of omental and subcutaneous WAT were obtained from female patients with obesity undergoing different bariatric surgery procedures (Roux-en-Y Gastric Bypass and sleeve gastrectomy). After collection, samples were washed using PBS, placed on ice until the end of the surgery procedure, snap-frozen in liquid nitrogen and stored at −80 °C for later use. Sequencing analyses were performed on isolated nuclei as detailed above. Single-nuclei RNAseq. libraries were generated using the Chromium Single Cell 3′ v3 assay (10× Genomics) and sequenced with NovaSeq 6000 S4 flow cell platform (Illumina). Raw reads were aligned to the human genome (hg38) and cells were called using 10x Genomics Cell Ranger (v.6.0.1).
## Collection of publicly available datasets
Peer-reviewed WAT datasets containing either snSeq, scSeq or STx with publicly available results published until 31.03.2022 were included in the present meta-analysis (Supplementary Table 1). If not present in the published articles, the corresponding authors were contacted by email to obtain information regarding sample numbers and gender as well as ranges of age and BMI.
## Re-analysis of publicly available datasets
Publicly available datasets were re-analyzed with Seurat v4.1.016 in R v4.1.263 (pipeline available via GitHub [https://github.com/lmassier/hWAT_singlecell]). Mitochondrial and hemoglobin genes, as well as further confounding transcripts, including MALAT2 and NEAT1, were removed prior to analysis. All data were normalized using sctransform64 and corrected for subject effects using Harmony v0.1.025 before performing independent component analysis65. Clusters were determined using FindNeighbors and FindClusters in Seurat after generating Uniform Manifold Approximation and Projection (UMAP) data projections using RunUMAP66. Spatial data were analyzed as described recently7.
## Cluster classification and annotation
Cluster classification for individual cohorts was assisted by a supervised network analysis where each cell cluster was represented as a node. Nodes were connected by edges by calculating overlap percentages of positively enriched marker genes (FDR-adjusted p value <0.05). Minimum requirements for edge connections were >$15\%$ genes overlapped in one of the two nodes and >$5\%$ in both nodes. Based on this, Jaccard similarity scores were calculated, and edges were created based on the five highest values. The network was built in R v4.1.2. using igraph v1.2.11 and visualized in Cytoscape 3.7.1 using Rcy3 v2.14.167. Integrated cluster annotations were performed manually using multiple reference datasets (Supplementary Table 2).
## Data integration and bechmarking
We evaluated the following integration tools: rPCA provided with Seurat16, Harmony25, BBKNN as well as scVI. In addition to identification/clustering of prominent marker genes in the integrated data, benchmarking included calculations of ARI coefficients, LISI scores, kBET acceptance rates (1- rejection rate) for integration across methods, depots and cohorts. ARI was calculated using the adj.rand.index function in pdfCluster v1.0-3 by supplying factors of either depots, methods or cohorts in addition to the identified clusters. Average LISI scores were estimated using lisi v1.0 compute_lisi command adding embeddings of the UMAPs along the meta data. kBET scores were calculated using kBET v0.99.6 and the respective integrated reductions (e.g., Harmony or scVI) in the Seurat object. Data from different depots or methods within the same study were treated as independent cohorts, which was used as batch variable to integrate over, thereby correcting for differences in methods and sequencing platforms. Of note, all cohorts except for Acosta et al12. ( <$0.5\%$ of analyzed cells), were sequenced in 3’ direction, thereby facilitating integration (Supplementary Table 1). Guided by our testing, we opted to integrate data individually for adipocytes, FAPs, immune and vascular cells. Based on these results, scVI integration using the 2000 most variable features (using VariableFeatures in Seurat) was applied in the final analysis. Seurat objects were transcribed into anndata objects using the sceasy v0.0.6 function convertFormat and scVI was run using default settings in R with reticulate v1.24 (setup_anndata, SCVI, train, get_latent_representation). Subsequent subcluster analyses were performed based on different depots (omental, perivascular, and subcutaneous WAT) or lineages (e.g., myeloid and lymphoid cells). When comparing similarity between individual data sets using Jaccard index, marker genes were selected based on a log2 fold-change > 0.5, and adjusted p value <0.05.
## Deconvolution
Deconvolution of bulk transcriptomic data of human WAT was performed using BisqueRNA v1.0.568 using the marker gene approach with a minimum gene count of six. To validate depot differences, two cohorts32,33 with available data from omental and subcutaneous WAT were used. We also retrieved sequencing as well as clinical phenotype data from six additional published datasets53,54,69–72 (Supplementary Table 3). Deconvolved data were compared in a meta-approach using Hmisc v4.6–0 to calculate Spearman correlation and meta v5.2–0 to calculate and visualize summarized results using both common and random models73.
## Flow-cytometry staining and analysis of data
The procedures for preparing cells for flow-cytometry have been described in detail elsewhere74. In brief, stromal vascular fractions (SVF) were thawed and stained with different antibody combinations prior to analysis with a flow cytometer. Washing steps were performed with wash buffer (PBS supplemented with $0.5\%$ BSA [#A4503, Sigma-Aldrich] and 2 mM EDTA [#E7889, Sigma-Aldrich]) and the cells were centrifuged at 200 × g for 10 minutes. To remove red blood cells, samples were incubated in red blood cell lysis buffer (15.5 mM NH4Cl, 0.57 mM K2HPO4, 0.01 mM EDTA × 2 H2O in PBS) for 6 minutes and subsequently washed with wash buffer. Samples were divided into aliquots containing approximately one million cells and subsequently stained with either (i) a lymphocyte antibody cocktail, (ii) an antibody panel for myeloid cells and fibroblasts combined, or (iii) a separate antibody panel for FAPs. The antibody-fluorochrome conjugates used are listed in Supplementary Table 4. Staining with the lymphocyte antibody cocktail was performed by resuspending the cells first in stain buffer ($1\%$ FBS, 2 mM EDTA in PBS) with CCR7-APC-Cy7 and incubating for ten minutes at 37 °C. Subsequently, the surface marker antibodies were added to the cell suspension and the incubation was carried out at room temperature for 20 minutes. The cells were washed once with stain buffer, centrifuged at 400 × g for five minutes, and resuspended into 1× fixation/permeabilization solution (#00-5223-56 and #00-5123-43, eBioscience). After incubating the cell suspensions for 30 minutes at room temperature in the dark, they were washed with permeabilization buffer (#00-8333-56, eBioscience), centrifuged at 400 × g for 5 minutes, and resuspended with an intracellular staining cocktail in 1× permeabilization buffer for 30 minutes at room temperature in the dark. Lastly, the cells were washed with permeabilization buffer and fixed with $1\%$ PFA (#22023-20 ml, Biotium) for 15 minutes prior to analysis with a BD Symphony analyzer equipped with 355, 405, 488, 561 and 640 nm lasers and DIVA software (BD Biosciences). Fixable live/dead Aqua stain (#L34957, Invitrogen) was included in the surface stain cocktail and used for dead cell exclusion. Gating was performed as outlined in Supplementary Figure 9a. The cells stained with the myeloid panel were incubated in the antibody cocktail for 30 minutes at 4 °C in the dark, washed once with wash buffer, resuspended into flow buffer ($0.1\%$ BSA, 2 mM EDTA in PBS), and analyzed immediately with the flow cytometer with a previously set compensation and gating setup. Fixable live/dead yellow dye (#L34968, Invitrogen) was used to exclude dead cells. The gating setup preceding the UMAP analysis is outlined in Supplementary Figure 9b. Cells stained with the FAPs panel were stained in a similar fashion as with the myeloid panel. 7-AAD (#559925, BD Biosciences) was used as live/dead exclusion dye and fluorescence minus one (FMO) controls were used for the gating (Supplementary Fig. 9c). The results were analyzed with FlowJo Software v10.7.1 and v10.8.0 (BD Biosciences). Dimensionality reduction for analyzing the myeloid cells was performed using the UMAP FlowJo plugin v3.1. For this analysis, 1267 myeloid cells from each individual were exported into a new file, barcoded and concatenated. FlowJo Phenograph plugin v3 was applied for unsupervised clustering, with the optimal k-nearest neighbors implemented automatically.
## Lipid droplet staining before and after adipogenesis
For imaging of CD55+ cells before and after adipogenesis, cells were fixed in $4\%$ PFA for 15 minutes at room temperature and washed twice with PBS. Lipid droplets and nuclei were stained with PBS containing BODIPY $\frac{493}{503}$ (1:2500, ThermoFisher) and Hoechst 33342 (1:5000, #ab228551, Abcam) for 15 minutes. Cells were then washed four times with PBS and images were acquired using CREST V3 confocal system (Crest Optics) mounted on an inverted Nikon Ti2 microscope equipped with a Prime BSIexpress sCMOS camera (pixel size 6.5 μm) from Photometrics. A Nikon 20x/0.75 air objective was used to acquire images.
## Immunostaining of LAMs (myC02)
For immunofluorescence, WAT samples were fixed in cold $4\%$ paraformaldehyde (PFA) for 24 hours, embedded in paraffin and then sliced into 6 μm thick sections. Antigens were retrieved by heating up the sections for 20 minutes in 10 mM citrate buffer pH 6.0 (tri-sodium citrate in distilled water) using a microwave. The samples were subsequently washed three times with PBS containing $0.3\%$ Triton X-100 and blocked for one hour at room temperature in PBS containing $0.1\%$ BSA, $0.1\%$ Triton X-100, 50 mM glycine, and $0.05\%$ Tween. Primary antibodies (anti-TREM2 [1:100, #13483-1-AP, Proteintech] and anti-CD9 [1:500, #60232-1-Ig, Proteintech]) were diluted in $0.1\%$ BSA, $0.1\%$ Triton-X-100, 10 mM glycine and $0.05\%$ Tween in PBS and incubated with the sections overnight at 4 °C. This was followed by three wash steps with PBS containing $0.3\%$ Triton X-100 and incubations for 10 minutes at room temperature. After this, slides were incubated with secondary antibodies (donkey α-rabbit conjugated with Alexa Fluor 594 [1:200, #A-21207, Thermofisher] and donkey α-mouse conjugated with Alexa Flour 488 [1:200, # A-21202, Thermofisher]) diluted in $0.1\%$ Tween in PBS for an hour at room temperature. Following three wash steps with PBS containing $0.3\%$ Triton-X-100, 100 µL Sudan Black B in $70\%$ ethanol was added to each section and the samples were incubated for two minutes at room temperature. The slides were thereafter rinsed in PBS and incubated with Hoechst 33342 (1:10000) diluted in PBS for ten minutes to stain nuclei. Prior to mounting in DAKO Fluorescence mounting media (S302380-2, Agilent Technologies), the samples were washed with PBS and swirled in distilled water.
## Immunostaining of FAPs (sfC12) and endothelial cells
Subcutaneous WAT blocks embedded in optimal cutting temperature compound were sliced into 16 μm thick sections, which were then fixed in $4\%$ PFA for five minutes at room temperature and washed twice with PBS. After this, glycine was added (100 mM final concentration) for ten minutes and slides were blocked for one hour in PBS containing $1\%$ BSA, $0.3\%$ Triton X-100, and $10\%$ normal donkey serum. Subsequently, the slides were incubated overnight in 4 °C with primary antibodies (anti-SLIT2 [1:200, #20217-1-AP, Proteintech] and anti-CD31 1:100, #M082329-2, Agilent Technologies]) in incubation buffer containing $5\%$ normal donkey serum, $1\%$ BSA and $0.3\%$Triton X-100 in PBS. Slides were washed three times with PBS containing $0.3\%$Triton X-100 for five minutes. They were thereafter incubated with secondary antibodies (donkey α-rabbit conjugated with Alexa Fluor 594 [1:200, #A-21207, Thermofisher] and donkey α-mouse conjugated with Alexa Flour 488 [1:200, #A-21202, Thermofisher]) in incubation buffer for 1 hour at room temperature. Additional washing steps were performed with PBS supplemented with $0.3\%$Triton X-100. Then, Hoechst 33342 (1:10,000) diluted in PBS was added for 10 minutes to stain nuclei. Prior to mounting in DAKO Fluorescence mounting media, the samples were washed with PBS and swirled in distilled water.
For immunostaining of both myC02 and sfC12/endothelial cells, images were acquired using the NIS Elements software, a CSU-X1 spinning disk confocal (Yokogawa) mounted on an inverted TiE microscope (Nikon) equipped with a ×1.2 magnification lens and a Kinetix back-illuminated sCMOS camera (pixel size 6.5 μm QE$95\%$) (Photometrics). A Nikon ×$\frac{20}{0.75}$ air objective was used to acquire images.
## Bulk RNA sequencing of in vitro adipogenesis
To annotate FAP clusters, we retrieved bulk sequencing data of human subcutaneous adipocyte precursor cells undergoing adipogenesis from four model systems: (i) adipose-derived stem cells41, (ii) primary SVF-derived cells from subcutaneous WAT41, (iii) Simpson-Golabi-Behmel (SGBS) syndrome cells42, and (iv) human multipotent adipose-derived stem (hMADS) cells43. Scores were generated for each FAP cluster at each timepoint of adipogenesis by calculating expression fold-changes of top 30 marker genes over the background of the respective bulk datasets.
## Comparing adipocyte-enriched transcripts between platforms
To identify adipocyte-enriched genes unrelated to single-cell approaches, we retrieved bulk RNAseq data of human isolated subcutaneous adipocytes from the FANTOM5 database46. This was compared to pseudobulk data, which was retrieved using AverageExpression in Seurat, of both individual snSeq and STx studies, as well as the integrated snSeq data. We calculated transcript fold-changes in these samples compared to the complete dataset. In total, 218 genes with >50-fold enrichment in fat cells compared to all other samples were considered to be adipocyte-enriched. We compared expression levels of these genes between snSeq and STx platforms after z-score normalizing each dataset. As an extra control, we included an additional bulk RNAseq dataset of human isolated subcutaneous adipocytes45. By filtering for genes with a |Δz-score| >5, we identified genes with the strongest differences between snSeq and STx.
## Spatial deconvolution
We performed cell2location v0.175 for deconvolution of STx by using a subset of annotated scRNA-seq data as input. We subsampled the single-cell data set for each cell type according to following criteria: (i) *If a* cell type had ≤1500 cells, select all cells; (ii) if a cell type had >1500 cells, randomly select 1500 cells. We compared cell2location with five additional spatial deconvolution methods: stereoscope v0.2.076, SPOTlight v0.1.077, RCTD v2.0.078, Tangram v1.0.279, DestVI v0.16.280 to validate the robustness of deconvolution results. Default parameter settings were used for deconvolution analysis.
We used spot-wise Pearson correlation between the estimated cell type abundances to quantify cell type colocalization pattern. The Pearson correlations were computed across all spots for each pair of cell types and each subject. High positive correlation indicated that two cell types exhibited similar spatial distributions, while negative correlation suggested distinct spatial distributions between two cell types.
## Cell-cell communication analysis
Cell-cell communication analysis was performed using CellChat v1.4.048 based on the curated ligand-receptor interaction database (CellChatDB). In brief, subcutaneous and omental normalized gene expression matrices were provided as input to CellChat, respectively. The total numbers of interactions and interaction strengths were computed by the computeCommunProb function, and the communication probabilities for each cell signaling pathway were calculated by computeCommunProbPathway function.
## Statistics
As detailed under each subheading above, statistics were performed in R v4.1.2 or GraphPad Prism v9. Data distributions were tested by Kolmogorov-Smirnov and Shapiro-Wilk tests and parametric vs. non-parametric tests were used accordingly. Spearman’s rank correlation was used to assess relationship of two continuous variables. For analyses requiring family-wise error rate corrections, p value <0.05 after correcting for false discovery rate using Benjamini-Hochberg was considered significant.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary Information Peer Review File Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-36983-2.
## Source data
Source Data
## Peer review information
: Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.
## References
1. Sakers A, De Siqueira MK, Seale P, Villanueva CJ. **Adipose-tissue plasticity in health and disease**. *Cell* (2022.0) **185** 419-446. DOI: 10.1016/j.cell.2021.12.016
2. Shao M. **De novo adipocyte differentiation from Pdgfrbeta(+) preadipocytes protects against pathologic visceral adipose expansion in obesity**. *Nat. Commun.* (2018.0) **9** 890. DOI: 10.1038/s41467-018-03196-x
3. Schwalie PC. **A stromal cell population that inhibits adipogenesis in mammalian fat depots**. *Nature* (2018.0) **559** 103-108. DOI: 10.1038/s41586-018-0226-8
4. Hepler C. **Identification of functionally distinct fibro-inflammatory and adipogenic stromal subpopulations in visceral adipose tissue of adult mice**. *Elife* (2018.0) **7** e39636. DOI: 10.7554/eLife.39636
5. Dong H. **Identification of a regulatory pathway inhibiting adipogenesis via RSPO2**. *Nat. Metab.* (2022.0) **4** 90-105. DOI: 10.1038/s42255-021-00509-1
6. Jaitin DA. **Lipid-associated macrophages control metabolic homeostasis in a Trem2-dependent manner**. *Cell* (2019.0) **178** 686-698.e614. DOI: 10.1016/j.cell.2019.05.054
7. Backdahl J. **Spatial mapping reveals human adipocyte subpopulations with distinct sensitivities to insulin**. *Cell Metab.* (2021.0) **33** 1869-1882.e1866. DOI: 10.1016/j.cmet.2021.07.018
8. Sun W. **snRNA-seq reveals a subpopulation of adipocytes that regulates thermogenesis**. *Nature* (2020.0) **587** 98-102. DOI: 10.1038/s41586-020-2856-x
9. Emont MP. **A single-cell atlas of human and mouse white adipose tissue**. *Nature* (2022.0) **603** 926-933. DOI: 10.1038/s41586-022-04518-2
10. Vijay J. **Single-cell analysis of human adipose tissue identifies depot and disease specific cell types**. *Nat. Metab.* (2020.0) **2** 97-109. DOI: 10.1038/s42255-019-0152-6
11. Hildreth AD. **Single-cell sequencing of human white adipose tissue identifies new cell states in health and obesity**. *Nat. Immunol.* (2021.0) **22** 639-653. DOI: 10.1038/s41590-021-00922-4
12. Acosta JR. **Single cell transcriptomics suggest that human adipocyte progenitor cells constitute a homogeneous cell population**. *Stem Cell Res. Ther.* (2017.0) **8** 250. DOI: 10.1186/s13287-017-0701-4
13. Merrick D. **Identification of a mesenchymal progenitor cell hierarchy in adipose tissue**. *Science* (2019.0) **364** eaav2501. DOI: 10.1126/science.aav2501
14. Karunakaran D. **RIPK1 gene variants associate with obesity in humans and can be therapeutically silenced to reduce obesity in mice**. *Nat. Metab.* (2020.0) **2** 1113-1125. DOI: 10.1038/s42255-020-00279-2
15. Angueira AR. **Defining the lineage of thermogenic perivascular adipose tissue**. *Nat. Metab.* (2021.0) **3** 469-484. DOI: 10.1038/s42255-021-00380-0
16. Hao Y. **Integrated analysis of multimodal single-cell data**. *Cell* (2021.0) **184** 3573-3587.e3529. DOI: 10.1016/j.cell.2021.04.048
17. Li B. **Benchmarking spatial and single-cell transcriptomics integration methods for transcript distribution prediction and cell type deconvolution**. *Nat. Methods* (2022.0) **19** 662-670. DOI: 10.1038/s41592-022-01480-9
18. Dominguez Conde C. **Cross-tissue immune cell analysis reveals tissue-specific features in humans**. *Science* (2022.0) **376** eabl5197. DOI: 10.1126/science.abl5197
19. Eto H. **Characterization of structure and cellular components of aspirated and excised adipose tissue**. *Plast. Reconstr. Surg.* (2009.0) **124** 1087-1097. DOI: 10.1097/PRS.0b013e3181b5a3f1
20. Denisenko E. **Systematic assessment of tissue dissociation and storage biases in single-cell and single-nucleus RNA-seq workflows**. *Genome Biol.* (2020.0) **21** 130. DOI: 10.1186/s13059-020-02048-6
21. Forcato M, Romano O, Bicciato S. **Computational methods for the integrative analysis of single-cell data**. *Brief Bioinform.* (2021.0) **22** 20-29. DOI: 10.1093/bib/bbaa042
22. Peng M, Li Y, Wamsley B, Wei Y, Roeder K. **Integration and transfer learning of single-cell transcriptomes via cFIT**. *Proc. Natl. Acad. Sci. USA* (2021.0) **118** e2024383118. DOI: 10.1073/pnas.2024383118
23. Xu C. **Probabilistic harmonization and annotation of single-cell transcriptomics data with deep generative models**. *Mol. Syst. Biol.* (2021.0) **17** e9620. DOI: 10.15252/msb.20209620
24. Polanski K. **BBKNN: fast batch alignment of single cell transcriptomes**. *Bioinformatics* (2020.0) **36** 964-965. DOI: 10.1093/bioinformatics/btz625
25. Korsunsky I. **Fast, sensitive and accurate integration of single-cell data with Harmony**. *Nat. Methods* (2019.0) **16** 1289-1296. DOI: 10.1038/s41592-019-0619-0
26. Gayoso A. **A Python library for probabilistic analysis of single-cell omics data**. *Nat. Biotechnol.* (2022.0) **40** 163-166. DOI: 10.1038/s41587-021-01206-w
27. Luecken MD. **Benchmarking atlas-level data integration in single-cell genomics**. *Nat. Methods* (2022.0) **19** 41-50. DOI: 10.1038/s41592-021-01336-8
28. Tran HTN. **A benchmark of batch-effect correction methods for single-cell RNA sequencing data**. *Genome Biol.* (2020.0) **21** 12. DOI: 10.1186/s13059-019-1850-9
29. Hubert L, Arabie P. **Comparing partitions**. *J. Classif.* (1985.0) **2** 193-218. DOI: 10.1007/BF01908075
30. Buttner M, Miao Z, Wolf FA, Teichmann SA, Theis FJ. **A test metric for assessing single-cell RNA-seq batch correction**. *Nat. Methods* (2019.0) **16** 43-49. DOI: 10.1038/s41592-018-0254-1
31. Simpson EH. **Measurement of diversity**. *Nature* (1949.0) **163** 688-688. DOI: 10.1038/163688a0
32. Arner P. **The epigenetic signature of systemic insulin resistance in obese women**. *Diabetologia* (2016.0) **59** 2393-2405. DOI: 10.1007/s00125-016-4074-5
33. Krieg L. **Multiomics reveal unique signatures of human epiploic adipose tissue related to systemic insulin resistance**. *Gut* (2021.0) **71** 2179-2193. DOI: 10.1136/gutjnl-2021-324603
34. Chen L. **ITLN1 inhibits tumor neovascularization and myeloid derived suppressor cells accumulation in colorectal carcinoma**. *Oncogene* (2021.0) **40** 5925-5937. DOI: 10.1038/s41388-021-01965-5
35. Johnson MD. **Inhibition of angiogenesis by tissue inhibitor of metalloproteinase**. *J. Cell Physiol.* (1994.0) **160** 194-202. DOI: 10.1002/jcp.1041600122
36. Medina RJ. **Molecular analysis of endothelial progenitor cell (EPC) subtypes reveals two distinct cell populations with different identities**. *BMC Med. Genomics* (2010.0) **3** 18. DOI: 10.1186/1755-8794-3-18
37. Keighron C, Lyons CJ, Creane M, O’Brien T, Liew A. **Recent advances in endothelial progenitor cells toward their use in clinical translation**. *Front. Med. (Lausanne)* (2018.0) **5** 354. DOI: 10.3389/fmed.2018.00354
38. Wang QA, Tao C, Gupta RK, Scherer PE. **Tracking adipogenesis during white adipose tissue development, expansion and regeneration**. *Nat. Med.* (2013.0) **19** 1338-1344. DOI: 10.1038/nm.3324
39. Borrelli MR. **The antifibrotic adipose-derived stromal cell: grafted fat enriched with CD74+ adipose-derived stromal cells reduces chronic radiation-induced skin fibrosis**. *Stem Cells Transl. Med.* (2020.0) **9** 1401-1413. DOI: 10.1002/sctm.19-0317
40. Buechler MB. **Cross-tissue organization of the fibroblast lineage**. *Nature* (2021.0) **593** 575-579. DOI: 10.1038/s41586-021-03549-5
41. Ehrlund A. **Transcriptional dynamics during human adipogenesis and its link to adipose morphology and distribution**. *Diabetes* (2017.0) **66** 218-230. DOI: 10.2337/db16-0631
42. Khan A. **SNEV(hPrp19/hPso4) regulates adipogenesis of human adipose stromal cells**. *Stem Cell Rep.* (2017.0) **8** 21-29. DOI: 10.1016/j.stemcr.2016.12.001
43. Tini G. **DNA methylation during human adipogenesis and the impact of fructose**. *Genes Nutr.* (2020.0) **15** 21. DOI: 10.1186/s12263-020-00680-2
44. Shao M. **Pathologic HIF1alpha signaling drives adipose progenitor dysfunction in obesity**. *Cell Stem Cell* (2021.0) **28** 685-701.e687. DOI: 10.1016/j.stem.2020.12.008
45. Harms MJ. **Mature human white adipocytes cultured under membranes maintain identity, function, and can transdifferentiate into brown-like adipocytes**. *Cell Rep.* (2019.0) **27** 213-225.e215. DOI: 10.1016/j.celrep.2019.03.026
46. Forrest AR. **A promoter-level mammalian expression atlas**. *Nature* (2014.0) **507** 462-470. DOI: 10.1038/nature13182
47. Gupta A. **Characterization of transcript enrichment and detection bias in single-nucleus RNA-seq for mapping of distinct human adipocyte lineages**. *Genome Res.* (2022.0) **32** 242-257. DOI: 10.1101/gr.275509.121
48. Jin S. **Inference and analysis of cell-cell communication using CellChat**. *Nat. Commun* (2021.0) **12** 1088. DOI: 10.1038/s41467-021-21246-9
49. Timokhina I, Kissel H, Stella G, Besmer P. **Kit signaling through PI 3-kinase and Src kinase pathways: an essential role for Rac1 and JNK activation in mast cell proliferation**. *EMBO J.* (1998.0) **17** 6250-6262. DOI: 10.1093/emboj/17.21.6250
50. Huizer K. **Periostin is expressed by pericytes and is crucial for angiogenesis in glioma**. *J. Neuropathol. Exp. Neurol.* (2020.0) **79** 863-872. DOI: 10.1093/jnen/nlaa067
51. Rosen ED, Spiegelman BM. **What we talk about when we talk about fat**. *Cell* (2014.0) **156** 20-44. DOI: 10.1016/j.cell.2013.12.012
52. Pasarica M. **Reduced adipose tissue oxygenation in human obesity: evidence for rarefaction, macrophage chemotaxis, and inflammation without an angiogenic response**. *Diabetes* (2009.0) **58** 718-725. DOI: 10.2337/db08-1098
53. Kerr AG, Andersson DP, Ryden M, Arner P, Dahlman I. **Long-term changes in adipose tissue gene expression following bariatric surgery**. *J. Intern. Med.* (2020.0) **288** 219-233. DOI: 10.1111/joim.13066
54. Petrus P. **Transforming growth factor-beta3 regulates adipocyte number in subcutaneous white adipose tissue**. *Cell Rep.* (2018.0) **25** 551-560.e555. DOI: 10.1016/j.celrep.2018.09.069
55. Lenz M, Arts ICW, Peeters RLM, de Kok TM, Ertaylan G. **Adipose tissue in health and disease through the lens of its building blocks**. *Sci. Rep.* (2020.0) **10** 10433. DOI: 10.1038/s41598-020-67177-1
56. Norreen-Thorsen M. **A human adipose tissue cell-type transcriptome atlas**. *Cell Rep.* (2022.0) **40** 111046. DOI: 10.1016/j.celrep.2022.111046
57. Hoffstedt J. **Long-term protective changes in adipose tissue after gastric bypass**. *Diabetes Care* (2017.0) **40** 77-84. DOI: 10.2337/dc16-1072
58. Mileti E. **Human white adipose tissue displays selective insulin resistance in the obese state**. *Diabetes* (2021.0) **70** 1486-1497. DOI: 10.2337/db21-0001
59. Ryden M. **The adipose transcriptional response to insulin is determined by obesity, not insulin sensitivity**. *Cell Rep.* (2016.0) **16** 2317-2326. DOI: 10.1016/j.celrep.2016.07.070
60. Slyper M. **A single-cell and single-nucleus RNA-Seq toolbox for fresh and frozen human tumors**. *Nat. Med.* (2020.0) **26** 792-802. DOI: 10.1038/s41591-020-0844-1
61. 61.Fleming, S. J., Marioni, J. C. & Babadi, M. CellBender remove-background: a deep generative model for unsupervised removal of background noise from scRNA-seq datasets. bioRxiv, https://www.biorxiv.org/content/10.1101/791699v1 (2019).
62. Germain PL, Lun A, Garcia Meixide C, Macnair W, Robinson MD. **Doublet identification in single-cell sequencing data using scDblFinder**. *F1000Research* (2022.0) **10** 979. DOI: 10.12688/f1000research.73600.2
63. 63.Team, R. C. R. A Language and Environment for Statistical Computing, (Vienna, Austria, 2018).
64. Hafemeister C, Satija R. **Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression**. *Genome Biol.* (2019.0) **20** 296. DOI: 10.1186/s13059-019-1874-1
65. Hyvarinen A, Oja E. **Independent component analysis: algorithms and applications**. *Neural Netw.* (2000.0) **13** 411-430. DOI: 10.1016/S0893-6080(00)00026-5
66. 66.Becht, E. et al. Dimensionality reduction for visualizing single-cell data using UMAP. Nat. Biotechnol.37, 38–44 (2018).
67. Gustavsen JA, Pai S, Isserlin R, Demchak B, Pico AR. **RCy3: network biology using cytoscape from within R**. *F1000Res* (2019.0) **8** 1774. DOI: 10.12688/f1000research.20887.2
68. Jew B. **Accurate estimation of cell composition in bulk expression through robust integration of single-cell information**. *Nat. Commun* (2020.0) **11** 1971. DOI: 10.1038/s41467-020-15816-6
69. Arner E. **Adipose tissue microRNAs as regulators of CCL2 production in human obesity**. *Diabetes* (2012.0) **61** 1986-1993. DOI: 10.2337/db11-1508
70. Arner P, Andersson DP, Backdahl J, Dahlman I, Ryden M. **Weight gain and impaired glucose metabolism in women are predicted by inefficient subcutaneous fat cell lipolysis**. *Cell Metab.* (2018.0) **28** 45-54.e43. DOI: 10.1016/j.cmet.2018.05.004
71. Imbert A. **Network analyses reveal negative link between changes in adipose tissue GDF15 and BMI during dietary-induced weight loss**. *J. Clin. Endocrinol. Metab.* (2022.0) **107** e130-e142. DOI: 10.1210/clinem/dgab621
72. Armenise C. **Transcriptome profiling from adipose tissue during a low-calorie diet reveals predictors of weight and glycemic outcomes in obese, nondiabetic subjects**. *Am. J. Clin. Nutr.* (2017.0) **106** 736-746. DOI: 10.3945/ajcn.117.156216
73. 73.Schwarzer, G., Carpenter, J. R. & Rücker, G. Meta-Analysis with R (Springer Cham, 2015).
74. Acosta JR. **Increased fat cell size: a major phenotype of subcutaneous white adipose tissue in non-obese individuals with type 2 diabetes**. *Diabetologia* (2016.0) **59** 560-570. DOI: 10.1007/s00125-015-3810-6
75. Kleshchevnikov V. **Cell2location maps fine-grained cell types in spatial transcriptomics**. *Nat. Biotechnol.* (2022.0) **40** 661-671. DOI: 10.1038/s41587-021-01139-4
76. Andersson A. **Single-cell and spatial transcriptomics enables probabilistic inference of cell type topography**. *Commun. Biol.* (2020.0) **3** 565. DOI: 10.1038/s42003-020-01247-y
77. Elosua-Bayes M, Nieto P, Mereu E, Gut I, Heyn H. **SPOTlight: seeded NMF regression to deconvolute spatial transcriptomics spots with single-cell transcriptomes**. *Nucleic Acids Res.* (2021.0) **49** e50. DOI: 10.1093/nar/gkab043
78. Cable DM. **Robust decomposition of cell type mixtures in spatial transcriptomics**. *Nat. Biotechnol.* (2022.0) **40** 517-526. DOI: 10.1038/s41587-021-00830-w
79. Biancalani T. **Deep learning and alignment of spatially resolved single-cell transcriptomes with Tangram**. *Nat. Methods* (2021.0) **18** 1352-1362. DOI: 10.1038/s41592-021-01264-7
80. Lopez R. **DestVI identifies continuums of cell types in spatial transcriptomics data**. *Nat. Biotechnol.* (2022.0) **40** 1360-1369. DOI: 10.1038/s41587-022-01272-8
81. Noguchi S. **FANTOM5 CAGE profiles of human and mouse samples**. *Sci. Data* (2017.0) **4** 170112. DOI: 10.1038/sdata.2017.112
|
---
title: Structural basis of selective cannabinoid CB2 receptor activation
authors:
- Xiaoting Li
- Hao Chang
- Jara Bouma
- Laura V. de Paus
- Partha Mukhopadhyay
- Janos Paloczi
- Mohammed Mustafa
- Cas van der Horst
- Sanjay Sunil Kumar
- Lijie Wu
- Yanan Yu
- Richard J. B. H. N. van den Berg
- Antonius P. A. Janssen
- Aron Lichtman
- Zhi-Jie Liu
- Pal Pacher
- Mario van der Stelt
- Laura H. Heitman
- Tian Hua
journal: Nature Communications
year: 2023
pmcid: PMC10017709
doi: 10.1038/s41467-023-37112-9
license: CC BY 4.0
---
# Structural basis of selective cannabinoid CB2 receptor activation
## Abstract
Cannabinoid CB2 receptor (CB2R) agonists are investigated as therapeutic agents in the clinic. However, their molecular mode-of-action is not fully understood. Here, we report the discovery of LEI-102, a CB2R agonist, used in conjunction with three other CBR ligands (APD371, HU308, and CP55,940) to investigate the selective CB2R activation by binding kinetics, site-directed mutagenesis, and cryo-EM studies. We identify key residues for CB2R activation. Highly lipophilic HU308 and the endocannabinoids, but not the more polar LEI-102, APD371, and CP55,940, reach the binding pocket through a membrane channel in TM1-TM7. Favorable physico-chemical properties of LEI-102 enable oral efficacy in a chemotherapy-induced nephropathy model. This study delineates the molecular mechanism of CB2R activation by selective agonists and highlights the role of lipophilicity in CB2R engagement. This may have implications for GPCR drug design and sheds light on their activation by endogenous ligands.
Cannabinoid CB2 receptor (CB2R) agonists are investigated as therapeutic agents in the clinic. Here, authors report the discovery of LEI-102, a CB2R agonist, used in conjunction with three other CBR ligands (APD371, HU308, and CP55,940) to investigate selective CB2R activation.
## Introduction
Preparations of the plant *Cannabis sativa* have been used for centuries in the treatment of various diseases, including cancer and neuropathic pain1. The synthetic version of its psychoactive constituent, Δ9-tetrahydrocannabinol (THC, Fig. 1), is in FDA approved drugs Marinol® or Syndros® (dronabinol). The extracted version of THC is one of the active constituents of oromucosal spray Sativex® (nabiximols). These drugs are primarily used for the treatment of chemotherapy-induced nausea, enhancement of appetite in cachexic AIDS-patients, and to alleviate the spasticity and pain associated with multiple sclerosis2–6. However, THC-based therapies are associated with clinically undesired psychotropic and cardiovascular adverse effects and challenging pharmacokinetic properties due to their high lipophilicity that may limit their therapeutic efficacy7–10.Fig. 1Chemical structures. Shown are the main constituent of *Cannabis sativa* Δ9-tetrahydrocannabinol (THC), and the two major endocannabinoids anandamide (AEA) and 2-arachidonoylglycerol (2-AG), as well as non-selective CBR agonist CP55,940 and CB2R agonists HU308, APD371, LEI-101, and LEI-102.
THC exerts its therapeutic effects mostly via the G protein-coupled receptors (GPCRs) cannabinoid CB1 and CB2 receptors (CB1R and CB2R), which have $68\%$ sequence identity in their seven transmembrane (TM) domains11. Both receptors are activated by the endogenous signaling lipids anandamide (AEA) and 2-arachidonoylglycerol (2-AG) (Fig. 1), the two main endocannabinoids. The CB1R, which is the most abundantly expressed GPCR in the central nervous system (CNS) is responsible for the psychotropic side effects of THC12–14. It plays a role in memory, learning, neurogenesis, neuronal migration, and synaptogenesis. Furthermore, its presence in many organ tissues belies more non-neurological functions15. The CB2R is mainly found on the cells of the immune system and is upregulated under pathophysiological conditions16,17. Its activation in general is associated with anti-inflammatory responses in tissue injury of the liver, heart, kidney, colon, and brain as determined in various preclinical models18–22. Based on preclinical studies, it is thought that selective CB2R agonists may retain and exceed certain therapeutic properties of THC without inducing psychotropic side effects23.
Various academic and industrial groups have developed selective CB2R ligands24. HU308 (Fig. 1) was the first selective CB2R agonist to be reported that displayed anti-inflammatory and analgesic properties in mouse models without inducing CNS-side effects18. However, poor physico-chemical properties (e.g. low solubility, high lipophilicity) of HU308, which has a calculated logarithm of octanol-water partition coefficient (cLogP) of 8.025, and its analogs prevented the successful clinical translation of this class of cannabinoid-based drugs.
A next generation of CB2R ligands was developed with improved drug-like properties. For instance, Olorinab® (APD371, Fig. 1) is the most polar CB2R agonist reported to date with a cLogP of −0.426. A phase 2a small-scale safety and tolerability trial in 14 patients with chronic abdominal pain associated with Crohn’s disease showed mild-to-moderate adverse events and an improvement in abdominal pain scores27. We have previously disclosed pyridinylbenzylimidazolidine-2,4-dione derivatives as selective CB2R agonists and studied their affinity, target binding kinetics and potency as a function of their lipophilicity, which resulted in the discovery of the orally available and peripherally restricted selective CB2R agonist LEI-101 (Fig. 1)28–30. It is intriguing that the CB2R binding pocket tolerates a wide array of ligands with very different scaffolds and hydrophobicity. For example, HU308 has a 2-billion-fold higher lipophilicity than APD371. Despite the tremendous progress in the field of CB2R drug discovery, we still do not have any molecular understanding on how these CB2R agonists selectively activate CB2R over CB1R.
Recently, three-dimensional structures of the CB1R and CB2R have been elucidated in both the active and inactive states by crystallography or cryo-electron microscopy (cryo-EM) and the binding modes of diverse ligands and their activation mechanism were reported31–35. Remarkably, those structures revealed that CB1R and CB2R possess a highly similar, lipophilic orthosteric agonist binding pocket, which makes it challenging to explain the selective activation of CB2R. To date, no structural studies with selective CB2R agonists have been reported that could aid in understanding the molecular basis of CB2R selectivity.
Here, we present the discovery of LEI-102, a potent and selective CB2R agonist with good physico-chemical and biological properties. LEI-102 is used in conjunction with CB2R selective agonists APD371 and HU308, and non-selective agonist CP55,940 to investigate the activation mechanism of CB2R. For this study, we combine ligand-target binding kinetics, site-directed mutagenesis, and cryo-EM methods. We find that CB2R has a distinct activation mechanism compared to CB1R. Additionally, we find that the physico-chemical properties of the ligands influence their entry pathway into the receptor. Highly lipophilic ligands, such as HU308 and the endocannabinoids, may reach the binding pocket through the membrane, whereas more polar ligands, such as LEI-102, APD371 and CP55,940, enter the receptor via an alternative route. Furthermore, we show that the favorable physico-chemical properties of LEI-102 and CB2R selectivity underscore its promising in vivo efficacy via oral administration in a chemotherapy-induced nephropathy model without inducing CNS-mediated side effects. Together, these studies enhance our insights into how certain physico-chemical properties of ligands translate to in vivo activity and changes their engagement to GPCRs.
## LEI-102 as a high affinity and potent CB2R-selective agonist
To obtain a selective CB2R agonist with beneficial physico-chemical properties, LEI-102, a pyridinylbenzylimidazolidine-2,4-dione derivative, was designed and synthesized (Supplementary Fig. 1). LEI-102 combined an isobutyl substituent on the imidazolidine with an aminotetrahydropyran to replace the cyclopropyl and thiomorpholine 1,1-dioxide in LEI-101, respectively30. LEI-102 had a cLogP of 2.1 as calculated by ChemDraw 19.0 (Supplementary Table 1). The inhibitory constant (pKi), potency (pEC50) and intrinsic activity (Emax) of LEI-102 were determined in [3H]RO6957022 displacement assays on stably expressing CB2R membranes and [35S]GTPγS G protein activation assays using HEK293T membranes transiently expressing recombinant hCB2R or hCB1R, respectively (Supplementary Table 2). APD371, HU308, CP55,940 and the endocannabinoids AEA and 2-AG were also explored. LEI-102 had a high binding affinity for CB2R (pKi = 8.0 ± 0.1) and was more potent than the selective CB2R agonists APD371 and HU308. LEI-102 did not bind CB1R, thereby showing at least 1000-fold selectivity (Supplementary Table 3). In G protein activation assays, LEI-102 activated the receptor as a partial agonist (Emax 76 ± $1\%$) with a pEC50 value of 6.9 ± 0.2 (Supplementary Table 2).
## Distinct target binding kinetic profiles of CB2R agonists
To quantify the ligand-target binding kinetic parameters of the agonists in more detail, we performed displacement and competition association assays with [3H]RO6957022 on membranes stably expressing hCB2R (Supplementary Table 2). The equilibrium Ki and kinetic KD values were well correlated, validating the competition association assay. First, we determined the dissociation rate constants (koff) of all agonists and converted these into a residence time (RT). LEI-102 had a RT of 16 min, which was around half that of APD371 (45 min) and CP55,940 (32 min), whereas HU308 had the longest RT at the receptor of 71 min (Supplementary Table 2). Endocannabinoids 2-AG and AEA had the shortest RT, both approximately 7 min. Of note, we found that the association rate constants (kon) varied greatly between the different agonists, ranking from fast to slow engagement CP55,940 > LEI-102 > 2-AG > APD371 > HU308 = AEA. The calculated engagement time (ET) to CB2R at 1 µM of each agonist further emphasized that CP55,940 arrived at CB2R within one second, whereas APD371, LEI-102, and 2-AG needed between 16 and 40 s to reach the CB2R binding site. Interestingly, HU308 and AEA took 143 and 152 s to bind CB2R, respectively. In view of the distinct target-binding kinetic profiles of the four synthetic CB2R agonists, we decided to elucidate their binding poses in CB2R using cryo-EM method.
## Overall similar structural comparison of CB2R-Gi in complex with different agonists
To obtain the stable complex sample of CB2R-Gi bound with LEI-102, APD371, HU308, or CP55,940, a similar procedure was used as for our previous AM12033-CB2R-Gi complex preparation (PDB: 6KPF). Single particle analysis of the cryo-EM samples yielded a normal global map for CB2R-LEI-102-Gi-scFv16, CB2R-APD371-Gi-scFv16, CB2R-HU308-Gi-scFv16, and CB2R-CP55,940-Gi-scFv16, complex, at 2.9, 3.0, 3.0, and 2.9 Å, respectively (Fig. 2, Supplementary Table 4, and Supplementary Figs. 2–5). The ligand, receptor and G protein in the isolated complex were clearly visible in the cryo-EM maps (Fig. 2 and Supplementary Fig. 6). The overall structures of the four complexes were comparable, with root mean square deviation (RMSD) of the Cα atoms of the receptors around 0.35 Å. The ligand binding interfaces of the four CB2R and Gi complexes were similar to each other, and to those of the previous AM12033-CB2R-Gi or WIN55212-2-CB2R-Gi complex structures. Fig. 2Cryo-EM structures of CB2R-Gi complexes. Cryo-EM density maps of a LEI-102 (Dark green), b APD371 (Sky blue), c HU308 (Olive), and d CP55,940 (Teal) bound CB2R in complex with Gαi (Slate), Gβ (Salmon), Gγ (Pale green), scFv16 (Violet purple). e–l Overall structures of CB2R-Gi complexes and enlarged view of orthosteric pocket of f LEI-102, h APD371, j HU308, and l CP55,940 using the same color codes as (a–d), with agonists shown as cornflower blue (LEI-102), orange (APD371), dark salmon (HU308) and purple (CP55,940) sticks, respectively.
## The binding mode of LEI-102 in CB2R
A clear electron density in the orthosteric ligand binding pocket in the LEI-102-CB2R-Gi complex resulted in the unambiguously defined binding pose of LEI-102 (Supplementary Fig. 6a). LEI-102 predominantly interacted with the residues in the binding pocket via hydrophobic interactions (Fig. 3a and Supplementary Fig. 7a). The isobutyl substituent of LEI-102 showed interactions with residues S902.60 (Ballesteros-Weinstein numbering in superscript), F1063.25, K1093.28, and I1103.29 in CB2R. The imidazolidine-2,4-dione formed a π-π interaction with F942.64 and showed further hydrophobic interactions with F1063.25 and P184ECL2. The benzyl formed an aromatic interaction with F183ECL2, and hydrophobic interactions with F872.57 and S2857.39. The phenyl ring in the core of LEI-102 formed a cation–π interaction with F183ECL2 and T-shaped π-π interaction with F2817.35. The pyridine had hydrophobic contacts with F1173.36 and W2586.48. The aminotetrahydropyran sidechain protruded into the long channel and formed hydrophobic interactions with residues I1103.29, T1143.33, I186ECL2, Y1905.39, L1915.40, W1945.43, and M2656.55. Additionally, a hydrogen bond was formed with T1143.33 (Supplementary Fig. 7a).Fig. 3Key interactions between CB2R and agonists. Key residues involved in a LEI-102, b APD371, c HU308, and d CP55,940 binding in CB2R-Gi complex structures. The amino acids involved in interactions are shown as sticks, hydrogen bonds are highlighted with yellow dashed lines. Same color codes as in Fig. 2.
## The binding mode of APD371 in CB2R
APD371 mainly formed hydrophobic and aromatic interactions with residues from ECL2, TM2, TM3, TM5, TM6, and TM7 (Fig. 3b and Supplementary Fig. 7b). The carbonyl group of APD371 formed a putative hydrogen bond with S2857.39 and a hydrophobic interaction with F872.57. The pyrazole and pyrazine cores of APD371 formed aromatic interactions with F183ECL2. Furthermore, the pyrazine core formed hydrophobic contacts with T1143.33, I186ECL2, L1915.40, and W1945.43. The (S)-1-hydroxy-3,3-dimethylbutyl head formed hydrophobic contacts with residues M26N-terminus, S902.60, F942.64, F1063.25, I1103.29 and V1133.32. The cyclopropyl group formed hydrophobic contacts with F1173.36, W1945.43, W2586.48, and V2616.51.
## The binding mode of HU308 in CB2R
The interactions between HU308 and CB2R were hydrophobic, including residues from ECL2, TM2, TM3, TM5, TM6, and TM7 (Fig. 3c and Supplementary Fig. 7c). The phenyl of 2,6-dimethoxyphenyl core formed hydrophobic interactions with F872.57, F183ECL2, and S2857.39, the C2-methoxy formed hydrophobic contacts with A2827.36 and S2857.39, and the C6-methoxy formed hydrophobic contacts with I1103.29, V1133.32, and T1143.33, respectively. The dimethylheptyl chain of HU308 extended into the long channel and formed hydrophobic interactions with residues from ECL2 (F183ECL2), TM3 (T1143.33, F1173.36) and TM5 (W1945.43). The 1,1-dimethyl formed hydrophobic interactions with residues F872.57, F1173.36, F2817.35, and S2857.39. The bicyclic head of HU308 formed hydrophobic interactions with M26N-terminus, F1063.25, I1103.29, S902.60, F942.64, P184ECL2, and the 2-methanol formed a hydrophobic interaction with F942.64 (Supplementary Fig. 7c).
## The binding mode of CP55,940 in CB2R
CP55,940 adopted an L-shape conformation in the orthosteric binding pocket (Fig. 3d and Supplementary Fig. 6d). The cyclohexanol group formed hydrophobic interactions with F942.64, L182ECL2, F183ECL2, and P184ECL2. The hydroxyl group established a hydrogen bond with L182ECL2 and the hydroxypropyl formed hydrophobic contacts with F872.57, S902.60, F912.61, I1103.29, and V1133.32. The phenol core formed hydrophobic interactions with F872.57, F183ECL2, F2817.35, and S2857.39, and its hydroxyl additionally formed a hydrogen bond with S2857.39. The dimethyl formed hydrophobic interactions with F183ECL2, F2817.35, M2656.55, F872.57, F1173.36, and C2887.42. The dimethylheptyl alkyl chain of CP55,940 extended into the long channel and formed hydrophobic interactions with residues I1103.29, F183ECL2, I186ECL2, W1945.43, T1143.33 and F1173.36 (Supplementary Fig. 7d).
## LEI-102 and APD371 require H952.65 for G protein activation in CB2R
To study the mechanism of CB2R activation, five residues in the binding pocket were further characterized based on the complex structures (Fig. 3). Six CB2R mutants were created, i.e. four residues (S2857.39, H952.65, I1103.29, and F1173.36) were replaced by alanine, as these are conserved between CB2R and CB1R, and two others (I1103.29, V2616.51) were substituted by the hCB1R reciprocal residue leucine. All mutants were sufficiently expressed at the cell surface as determined with an ELISA (Supplementary Fig. 8, Supplementary Table 5). To characterize the binding mechanisms of LEI-102, APD371, HU308, and CP55,940, their responses were investigated by [3H]CP55,940 displacement and [35S]GTPγS binding assays. Of note, in the [3H]CP55,940 displacement assay, only the CB2R-I1103.29L mutant showed a sufficient binding window (data not shown). This prevented the affinity determination of the four agonists on other mutant receptors. Five mutant receptors, except CB2R-F1173.36A, were still active in the [35S]GTPγS functional assay, thereby allowing us to study the receptor activation mechanism (Fig. 4a–d and Supplementary Table 6). All four synthetic agonists were unable to activate CB2R-F1173.36A, which indicated an important role of this residue in the activation of CB2R.Fig. 4Characterization of G protein activation of wild type (WT) and mutant CB2R by synthetic agonists and endocannabinoids. Dose-response curves for G protein activation of WT and mutants that are located in the CB2R binding pocket by a LEI-102, b APD371, c HU308, and d CP55,940. e–j Dose response curves for G protein activation of WT and mutants that are proposed to be involved in ligand entry of CB2R via either the ECL2 or membrane access by e LEI-102, f APD371, g HU308, h CP55,940, i AEA and j 2-AG. a–j *The maximum* activation level of WT CB2R was set to $100\%$ while the basal levels were set to $0\%$. Data are presented as mean ± SEM of at least three individual experiments performed in duplicate (specific n values are given in Supplementary Table 6). Source data are provided as a Source Data file.
The potency of LEI-102 was significantly increased at the CB2R-I1103.29L mutant to a pEC50 value of 7.8 ± 0.1 in the G protein activation assay, while the binding affinity remained similar to wild type (WT) receptor (Fig. 4a and Supplementary Tables 6 and 7). Three mutations CB2R-I1103.29A, CB2R-S2857.39A and CB2R-V2616.51L had no significant effect on the potency of LEI-102 in the functional assay. In contrast, the potency on mutant receptor CB2R-H952.65A was significantly reduced for LEI-102. No gain in binding affinity for the swap mutant in CB1R-L3596.51V was found with LEI-102 (Supplementary Tables 3 and 8).
APD371 acted as a full CB2R agonist with a pEC50 value of 7.9 ± 0.1 and a higher maximal activation compared to that of CP55,940 in the functional assay (Supplementary Table 2). Mutant receptor CB2R-I1103.29L did not affect the G protein response of APD371 (Fig. 4b and Supplementary Table 6), while the binding affinity was significantly reduced to a pKi of 7.1 ± 0.0 (Supplementary Table 7). APD371 potency was not affected by mutant receptors CB2R-I1103.29A or CB2R-S2857.39A. The responses of APD371 for CB2R-H952.65A and CB2R-V2616.51L were significantly impacted with 158-fold and 10-fold drop in potency, respectively (Supplementary Table 6).
Thus, we uncovered a crucial role for CB2R-H952.65 in G protein activation of CB2R by LEI-102 and APD371. Furthermore, LEI-102 activation was increased for the CB2R-I1103.29L mutant, while APD371 activation relied on CB2R-V2616.51.
## An important role for S2857.39 and V2616.51 in CB2R activation by HU308 and CP55,940
The potency and affinity of HU308 on CB2R were not affected by the CB2R-I1103.29L swap mutant (Fig. 4c and Supplementary Tables 6, 7). In addition, activation of mutant receptors CB2R-I1103.29A and CB2R-H952.65A by HU308 was not affected with pEC50 values of 6.4 ± 0.5 and 6.6 ± 0.6, respectively. The maximum activation level of mutant receptor CB2R-S2857.39A was unaffected compared to WT receptor, but a significant 15-fold loss in potency was observed. Lastly, CB2R-V2616.51L had a significant loss of potency, i.e. more than 120-fold lower (Fig. 4c and Supplementary Table 6).
Similar to HU308, the potency of CP55,940 on CB2R was not affected by the CB2R-I1103.29 mutations compared to WT in the G protein activation assay, nor was its binding affinity for CB2R-I1103.29L (Fig. 4d and Supplementary Tables 6, 7). In response to CP55,940, mutant receptors CB2R-S2857.39A and CB2R-V2616.51L were significantly affected with decreased pEC50 values of 6.7 ± 0.1 and <5, respectively. Moreover, the potency of CP55,940 was significantly affected on the CB2R-H952.65A with a 40-fold decrease compared to WT receptor (Fig. 4d and Supplementary Table 6). No gain in potency or affinity was observed for the swap mutant CB1R-L3596.51V for either HU308 or CP55,940 (Supplementary Tables 3 and 8).
Taken together, this showed that CB2R-S2857.39 and CB2R-V2616.51 were crucial for HU308 and CP55,940 to activate the G protein at CB2R, where CP55,940 additionally required an interaction with CB2R-H952.65.
## HU308 and endocannabinoids gain access via membrane entry
Our detailed ligand-target binding kinetic analysis revealed that the highly lipophilic HU308 and anandamide had a very slow on-rate compared to the other ligands. Since it has previously been postulated that ligands of lipid receptors may gain access to the binding pocket via a membrane channel, we examined two potential ligand entry pathways at CB2R, i.e. either via ECL2 or via a membrane channel in TM1 and TM7. To this end, four additional mutant receptors were created. Three residues in the ECL2 of CB2R, which were different from CB1R, were mutated towards the reciprocal CB1R residues, i.e. CB2R-L185ECL2H, CB2R-L182ECL2I, and CB2R-E181ECL2D. In the fourth mutant receptor, four residues in TM1 and TM7 that align the potential membrane channel in CB2R were mutated to the reciprocal CB1R residues and combined as a quadruple mutant, i.e. CB2R-K2797.33T, CB2R-K331.32Q, CB2R-V361.35I and CB2R-C401.39S (termed “CB2R-QuadrupleTM1,7”). Next, we tested all four synthetic agonists and the two endocannabinoids on these four CB2R mutant receptors in [3H]CP55,940 and [35S]GTPγS assays. Only CB2R-L185ECL2H and CB2R-QuadrupleTM1,7 were evaluated in the [3H]CP55,940 displacement assays due to insufficient binding window for the other two mutant receptors (data not shown). The binding affinities of the agonists were not affected for mutant receptors CB2R-L185ECL2H and CB2R-QuadrupleTM1,7 (Supplementary Table 7). Interestingly, the potencies of LEI-102, APD371, and CP55,940 in the functional assay were not significantly affected for any of the mutant receptors, whereas HU308 and the endocannabinoids were less potent on CB2R-L182ECL2I (Fig. 4e–j and Supplementary Table 6). Additionally, the endocannabinoids showed a decreased potency on CB2R-L181ECL2D, but not on CB2R-L185ECL2H. Of note, HU308 and both endocannabinoids completely lost their ability to activate CB2R in the CB2R-QuadrupleTM1,7 mutant, suggesting that this may be an important access point to the receptor binding pocket for these agonists (Fig. 4g, i, j).
## LEI-102 attenuates cisplatin-induced nephrotoxicity without CB1R-mediated side effects
In view of the excellent physico-chemical properties of LEI-102 and its selective CB2R agonist profile, we investigated the compound in a well-established in vivo model of kidney inflammation and injury induced by cisplatin. In this model, CB2R activation is associated with protective effects29. Cisplatin (25 mg/kg, i.p.) induced marked elevations of serum creatinine and blood urea nitrogen levels (functional markers of kidney injury) 72 h following cisplatin injection in wild type mice compared with vehicle-treated control animals. LEI-102 showed a dose-dependent attenuation of the functional markers of cisplatin-induced kidney injury both when administered p.o. ( orally) or i.p. ( Fig. 5a). Renal dysfunction was also accompanied by morphological damage to the kidney tubules determined by histological examination following PAS staining. LEI-102 (10 mg/kg) significantly decreased tubular injury as determined by this staining (Fig. 5b). Marked increases in oxidative and nitrative stress markers (4-HNE and 3-nitrotyrosine) were observed in kidneys of cisplatin-treated mice determined by immunostaining and quantitative ELISA. Furthermore, LEI-102 (10 mg/kg by i.p. or p.o.) decreased lipid peroxidation and protein nitration (Fig. 5c, d). Additionally, the pro-inflammatory cytokines TNFα and IL1β that were elevated due to the cisplatin-induced injury were attenuated in LEI-102 treated mice (Fig. 5e). Importantly, the protective effects of LEI-102 against cisplatin-induced renal dysfunction and tubular damage (Fig. 5f), histopathological injury (Fig. 5g) and markers of oxidative-nitrative stress (Fig. 5h, i) were abolished in CB2R knockout mice, which had enhanced kidney injury/dysfunction compared to their wild types. Fig. 5CB2R agonist LEI-102 attenuates cisplatin-induced renal dysfunction, oxidative stress, and inflammation in a CB2R-dependent manner.a Cisplatin-induced renal dysfunction 72 h after administration to mice as evidenced by increased serum levels of blood urea nitrogen (BUN) and creatinine (CREA), which were attenuated by CB2R agonist LEI-102 in a dose-dependent manner when administered either i.p. or p.o. (* $p \leq 0.001$ vs. vehicle group, #$p \leq 0.001$ vs. cisplatin group). b Periodic Acid-Schiff (PAS) staining in representative kidney sections from cisplatin treatment samples showing protein cast, vacuolation, and desquamation of epithelial cells in the renal tubules which are attenuated with LEI-102. Tubular damage score from kidney sections is shown (*$p \leq 0.001$ vs. vehicle group, #$p \leq 0.001$ vs. cisplatin group). c The cisplatin-induced nitrative and oxidative stress (nitrotyrosine staining (top row) and HNE staining (bottom row)) in representative kidney sections were also attenuated by LEI-102. This was confirmed by quantitative determination of protein nitration and HNE adducts formation by ELISA (d) (*$p \leq 0.001$ vs. vehicle group, #$p \leq 0.001$ vs. cisplatin group). e The cisplatin-induced kidney pro-inflammatory cytokine expressions were also attenuated by the CB2R agonist. (* $p \leq 0.001$ vs. vehicle group, #$p \leq 0.05$ vs. cisplatin group). The protective effects of LEI-102 on cisplatin-induced kidney dysfunction (BUN and CREA) and tubular injury (tubular damage score) (f) (*$p \leq 0.001$ vs. vehicle WT or KO group, #$p \leq 0.05$ vs. cisplatin WT group), histopathological injury (g), nitrative (h) and oxidative stress (i) were abolished in CB2R knockout mice. All results are means ± SEM of $$n = 6$$/group for panels a, b, d, e, f Closed and open symbols are used for male and female mice respectively (4 males and 2 females/group). In panels a, b, d, and e one-way ANOVA followed by Tukey’s post hoc test for multiple comparisons were used, in panel f unpaired two-tailed t-test was used. The analysis was conducted using GraphPad Prism 6 software. $p \leq 0.05$ was considered statistically significant (the exact p values are indicated in the supplemental data).
To determine whether LEI-102 maintained its selectivity for CB2R over CB1R in vivo, LEI-102 was tested in the mouse tetrad assay for CB1R activity18. In this assay, four consecutive behavioral tests, related to anti-nociception, hypothermia, catalepsy, and spontaneous activity, were performed 120 min after administration of the agonist. LEI-102 (25 mg/kg, p.o.) did not produce any effects in the tetrad assay as compared with vehicle. There were no effects on nociceptive behavior assessed in tail withdrawal test nor on body temperature (Fig. 6 upper row). No effect was found on locomotor behavior (Fig. 6 lower row) in case of distance traveled, time spent mobile, or running speed of mice. Nor was catalepsy observed following administration of LEI-102. These results indicated that LEI-102 (or one of its metabolites) did not produce CB1R-mediated CNS-side effects at doses up to 25 mg/kg (p.o.). Hence, the CB2R agonist LEI-102 maintained its selectivity over CB1R in vivo. Fig. 6The CB2R agonist LEI‐102 does not induce cannabimimetic CB1R-mediated effects (Tetrad assay) in vivo. LEI‐102 (25 mg/kg, p.o.) did not affect nociceptive behavior assessed in tail withdrawal test and body temperature, as compared with mice receiving vehicle (upper row). No effects on locomotor behavior were found (lower row). Results are means ± SEM; $$n = 8$$ per group.
## Discussion
So far, several crystal and cryo-EM CBR structures have been resolved in which non-selective agonists adopt a nearly identical binding position in the orthosteric pocket, regardless of the receptor34–38. In this study, we aimed to generate a better understanding of the binding and activation mechanism of CB2R-selective agonists. Therefore, we combined ligand-target binding kinetics, site-directed mutagenesis, and cryo-EM studies to investigate the activation mechanism of CB2R for the introduced CB2R selective agonist LEI-102 supplemented with agonists APD371, HU308, and CP55,940 on a molecular level. Furthermore, we investigated potential hotspots for CB2R/CB1R selectivity by creating swap mutants and discovered a ligand entry pathway for CBR agonists and endocannabinoids.
First, our data revealed a crucial role for CB2R-F1173.36 as replacement by alanine resulted in a complete loss of G protein activation by all tested agonists (Fig. 4a–d and Supplementary Table 6). It has been shown that the CB1R counterpart F2003.36 plays an important regulatory role in activation as part of the “twin toggle switch” with CB1R-W3566.4839. In contrast, CB2R-W2586.48 has been described to be solely responsible for activation as a toggle switch without the help of CB2R-F1173.36 in structural studies, since the conformation of CB2R-F1173.36 in agonist-bound structures is comparable to the conformation in the antagonist-bound CB2R structure as well as the CB1R agonist-bound structures33,36. Our mutation data further supports this hypothesis, as we do not see the same constitutive activity pattern (Supplementary Table 6) as observed by McAllister et al. for the reciprocal CB1R-F2203.36 excluding CB2R-F1173.36 from a suppressive function39. Together, this data provides evidence for a different, but important, role for F1173.36 in CB2R activation.
In CB1R, water-mediated interactions between CB1R-H1782.65, CB1R-S3837.39, and bound ligands have previously been shown with in silico modeling40,41. The importance of CB1R-S3837.39 for classical synthetic cannabinoids such as AM11542, AM841, and CP55,940 was further emphasized in CB1R-S3837.39A mutants36. This is in line with the observation that removal or methylation of the phenolic OH on classical cannabinoids, such as in L-759656, JWH-133, and HU308, always affords selectivity over CB1R18,42. Non-classical agonists, such as WIN55,212-2, do not form a hydrogen bond with CB1R-S3837.39 and consequently are not affected by an alanine mutation43. This translates to our results that CP55,940 and HU308 are more affected by the CB2R-S2857.39A mutation than LEI-102 and APD371 (Fig. 4a–d and Supplementary Table 3). The decrease in activation is at least 30-fold smaller for CB2R than CB1R36. The elucidated cryo-EM structures of our four agonists did not show direct interactions with CB2R-H952.65, though we cannot rule out its role in stabilizing the surrounding residues. The large effect seen on G protein activation of CB2R-H952.65A by LEI-102, APD371, and CP55,940 (Fig. 4a–d and Supplementary Table 6) must therefore stem from an indirect interaction, supporting the polar network hypothesis between CB2R-H952.65 and CB2R-S2857.39 in CB2R.
Residues at position 6.51 have previously been described to be involved in the binding sites of µ, δ, and κ opioid receptors, the dopamine D2 receptor, and adenosine receptors, and could play a role in ligand binding selectivity between different subtypes44–46. In our studies, introduction of the bulkier CB1R leucine on this position in CB2R-V2616.51L reduced the G protein activation by APD371, HU308, and CP55,940, while LEI-102 could still be accommodated in the binding pocket (Fig. 4a–d and Supplementary Table 6). Furthermore, with the swap mutant CB1R-L3595.61V we found a trend in partial recovery of displacement of [3H]CP55,940 by the CB2R selective agonists LEI-102, HU308, and APD371, although not significant (Supplementary Table 3). This supports a role of this residue in the selectivity of agonists in CB2R.
The ECL2 has frequently been implicated to be important for GPCR activation and some GPCRs even use their ECL2 as a ligand to auto-activation47. There are distinct differences between the conformations of ECL2 in CB1R and CB2R. In antagonist-bound CB1R crystal structures, the ECL2 dips into the binding pocket, interacting with the ligand and inducing the inactive conformation31,32. The inactive state of CB2R, however, does not expand like CB1R and instead the ECL2 acts more as a lid on the binding pocket in active and inactive CB2R, akin to active CB1R33. A key distinction seen in the CB1R crystal structures with AM6538 and taranabant, is the ionic lock formed by CB1R-E100N-terminus (CB2R-L17) and CB1R‑H270ECL2 (CB2R-L185)31,32. We observed improved binding of [3H]CP55,940 for LEI-102 and HU308 with the CB1R-H270ECL2L mutation, while the non-selective agonists showed no change (Supplementary Table 3). Through the loss of this ionic lock, selectivity over CB1R is partially lost, showing that the expulsion of ECL2 upon ligand entry may play an important role in selectivity.
In recent years, computational studies have suggested that lipophilic ligands for various GPCRs, such as the opsin receptor, sphingosine-1-phosphate receptor 1 (S1P1) and cannabinoid receptors, might gain access to the binding pocket through lateral diffusion via a membrane channel between TM1 and TM732,41,48–51. We experimentally examined this membrane entry pathway by creating a CB2R quadruple mutant (K331.32Q, V361.35I, C401.39S, and K2797.33T) for which we observed a significant loss of potency and a corresponding trend in reduced affinity, although not significant, for HU308 and the endocannabinoids (Fig. 4e−j, Supplementary Table 6 and 7). These compounds are more lipophilic than LEI-102 and APD371, making them more suitable to traverse the membrane to enter between TM1 and TM7. Notably, HU308 and anandamide also showed a substantially longer ET in our assays compared to the other agonists (Supplementary Table 2). This might suggest a possible relationship between a slower association and membrane channel entry at the CB2R. Likewise, for a peptide GPCR a trend in reduced association rate was found with increasing lipophilicity52. Nevertheless, this is in contrast with the mechanism at the α2-adrenoceptor at which lipophilic compounds had a faster association rate53. This shows the diversity in drug-target binding kinetics as receptor-specific properties and thus the importance of investigating these mechanisms for individual receptors54.
The discovery of a membrane access channel for endocannabinoids on the CB2R is also intriguing from a physiological perspective. Endocannabinoids are produced on demand and act as autocrine or paracrine effectors in the immune system regulating the migration of CB2R-expressing immune cells17. Our results suggest that endocannabinoids first have to travel through the plasma membrane via lateral diffusion to reach the receptor. This may suggest that the trafficking and cellular uptake of endocannabinoids could be regulated through extracellular or intracellular vesicles that merge with the plasma membrane. Regardless of the exact mechanism of endocannabinoid trafficking, this study provides experimental evidence of a membrane channel located between TM1 and TM7 in CB2R that is being used by the endocannabinoids to enter the receptor.
The ligands of the CB2R, such as the phytocannabinoids and endocannabinoids, are typically very lipophilic, which comes at a cost of reduced solubility, increased off-target activity, and poor pharmacokinetic properties10,25. Thus, balancing lipophilicity of a drug candidate is an important goal in medicinal chemistry. The first generation of experimental drugs targeting the CB2R mimicked the plant-based cannabinoids. Consequently, they were highly lipophilic and suffered from poor clinical translation10. *New* generations of CB2R agonists have optimized physico-chemical properties. For instance, LEI-102 and APD371 are orders of magnitude more hydrophilic than HU308. Remarkably, they can bind the same binding pocket in CB2R as HU308. Our data revealed that LEI-102 and APD371 do not enter the receptor via the membrane channel like HU308, but gain access most likely via the extracellular space. LEI-102 and APD371 also form a specific (indirect) polar interaction network with H952.65 to activate CB2R, which is not observed for HU308. This flexibility of the CB2R binding pocket to be activated by a diverse set of chemotypes allows to select for a chemotype with more drug-like properties. This notion is supported by the oral efficacy of LEI-102 in the chemotherapy-induced nephropathy model and lack of CNS-adverse side effects (Figs. 5 and 6).
Targeting CB2R with agonists is a promising avenue for the treatment of autoimmune diseases, neuroinflammation, and various forms of tissue injury/inflammation/fibrosis in the liver, heart, brain, and kidney17. In this study, we show that LEI-102 protects against cisplatin-induced nephropathy in a CB2R-dependent manner by attenuating kidney inflammation and injury (Fig. 5). We also show that CB2R knockout mice develop more severe nephropathy compared to their wild types suggesting a protective role of endocannabinoid-CB2R signaling during kidney injury. These results are consistent with protective effects of CB2R agonists in various models of kidney injury/diseases and deleterious effect of CB2R deletion in these models29,55–63.
In conclusion, we have discovered LEI-102 as a selective CB2R agonist that is efficacious in attenuating tissue injury in chemotherapy-induced nephropathy model without inducing CNS-mediated side effects. Using LEI-102 and five other CBR agonists, we have shown that the physicochemical properties determine not only pharmacokinetic properties of ligands, but also how they engage with their target. Altogether, we elucidated several important molecular mechanisms for selective engagement and activation of the CB2R, which may have implications for drug design and lipid signaling at GPCRs in general.
## General materials for functional assays
Monoclonal M2 mouse anti-FLAG primary antibody (#F3165) was purchased from Sigma-Aldrich (Zwijndrecht, the Netherlands), while secondary goat anti-mouse HRP-conjugated antibody (#115-035-003) was bought from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Bicinchoninic acid (BCA) ad BCA protein assay reagent was obtained from Pierce Chemical Company (Rockford, IL, USA). [ 3H]RO6957022 (specific activity 82.83 Ci mmol−1) was custom synthesized at F. Hoffman-La Roche Ltd (Basel, Switzerland). [ 35S]GTPγS (specific activity 1250 Ci mmol−1 #NEG030H250UC), [3H]CP55,940 (specific activity 108.5 Ci mmol−1 #NET1051250UC) and GF/C filter plates (#6055690) were purchased from PerkinElmer (Waltham, MA, USA). CP55,940 (#C1112), AM630 (#SML0327) and DL-dithiotreitol (DTT, #646563) were obtained from Sigma-Aldrich, HU308 (#H800010) was from LKT Laboratories (St. Paul, MN, USA), APD371 was provided by F. Hoffmann-La Roche Ltd, anandamide (AEA, #1339), 2-Arachidonylglycerol (2-AG, #1298) and phenylmethylsulfonyl fluoride (PMSF, #4486) were purchased from Tocris Bioscience (Bristol, UK) and GDP (#J61646) was from Thermo Fisher Scientific (Waltham, MA, USA). All buffers and solutions were prepared using Millipore water (deionized using a MilliQ A10 Biocel with a 0.22 µm filter) and analytical grade reagents and solvents. Buffers are prepared at room temperature (RT) and stored at 4 °C, unless stated otherwise.
## Cell lines
Spodoptera frugiperda (Sf 9) cells were used for CB2R-Gi co-expression for cryo-EM studies. Sf 9 cells were grown in ESF 921 medium (Expression systems) at 27 °C and 125 rpm. For transfections, human embryonic kidney 293T (HEK293T; female, ATCC #CRL-3216) cells were grown as monolayers in culture medium i.e. Dulbecco’s Modified Eagle’s Medium (Sigma-Aldrich #6546), supplemented with $10\%$ fetal calf serum (Sigma-Aldrich #F7524), 2 mM L-glutamine (Sigma-Aldrich #G8541), 100 IU/mL penicillin and 100 µg/mL streptomycin (Duchefa Biochemie #P0142 and #S0148) under a humidified atmosphere at 37 °C with $5\%$ CO2. Subculture was done twice a week at 80–$90\%$ confluence on 10 cm ø plates by trypsinization. CHO cells stably expressing hCB2R (CHOK1_hCB2bgal; PathHunter EA Parental Cell line, female, DiscoverX #93-0706C2) were cultured in Ham’s F12 Nutrient Mixture (Sigma-Aldrich #4888) supplemented with $10\%$ fetal calf serum, 2 mM L-glutamine, 100 IU/mL penicillin, 100 µg/mL streptomycin, 300 µg/mL hygromycin (Bio-Connect #ANT-HG-5) and 800 µg/mL G418 (Bio-Connect #SC-29065B) in a humidified atmosphere at 37 °C with $5\%$ CO2. Cells were subcultured twice a week when reaching 80–$90\%$ confluence on 10 or 15 cm ø plates by trypsinization.
## Synthesis of LEI-102
All reagents and solvents were purchased from commercial sources and were of analytical grade (Sigma-Aldrich, BroadPharm®). Reagents and solvents were not further purified before use. All moisture sensitive reactions were performed under inert atmosphere. Solvents were dried using 4 Å molecular sieves prior to use when anhydrous conditions were required. Water used in reactions was always demineralized. Analytical Thin-layer Chromatography (TLC) was routinely performed to monitor the progression of a reaction and was conducted on Merck Silica gel 60 F254 plates. Reaction compounds on the TLC plates were visualized by UV irradiation (λ254) and/or spraying with potassium permanganate solution (K2CO3 (40 g), KMnO4 (6 g), and H2O (600 mL)), ninhydrin solution (ninhydrin (1.5 g), n-butanol (100 mL) and acetic acid (3.0 mL)) or molybdenum solution ((NH4)6MO7 · 4 H2O (25 g/L) and (NH4)4Ce(SO4)4 · H2O (10 g/L) in sulfuric acid ($10\%$)) followed by heating as appropriate. Purification by flash column chromatography was performed using Screening Devices B.V. silica gel 60 (40–63 µm, pore diameter of 60 Å). Solutions were concentrated using a *Heidolph laborata* W8 4000 efficient rotary evaporator with a *Laboport vacuum* pump.
Analytical purity was determined with Liquid Chromatography-Mass Spectrometry (LC-MS) using a Finnigan LCQ Advantage MAX apparatus with electrospray ionization (ESI), equipped with a Phenomenex Gemini 3 μm NX-C18 110 Å column (50 × 4.6 mm), measuring absorbance at 254 nm using a Waters 2998 PDA UV detector and the m/z ratio by using an Acquity Single Quad (Q1) detector. Injection was with the Finnigan Surveyor Autosampler Plus and pumped through the column with the Finnigan Surveyor LC pump plus to be analyzed with the Finnigan Surveyor PDA plus detector. Samples were analyzed using eluent gradient $10\%$ → $90\%$ ACN in MilliQ water (+$0.1\%$ TFA (v/v)).
For purification by mass guided preparative High-Performance Liquid Chromatography (Prep-HPLC) the Waters AutoPurification HPLC/MS apparatus was used with a Gemini prep column 5 μm 18 C 110 Å (150 × 21.2 mm), Waters 2767 Sample manager, Waters 2545 Binary gradient module, Waters SFO System fluidics organizer, Waters 515 HPLC pump M, Waters 515 HPLC pump L attached to a Waters SQ detector Acquity Ultra performance LC.
1H, 13C, 1H-COSY and HSQC Nuclear Magnetic Resonance (NMR) spectra were recorded on a Bruker AV 300 ($\frac{300}{75}$ MHz), AV 400 ($\frac{400}{100}$ MHz) or AV 500 ($\frac{500}{125}$ MHz) spectrometer at ambient temperature using CDCl3 as solvent. Chemical shifts (δ) are referenced in parts per million (ppm) with tetramethylsilane (TMS) or CDCl3 resonance as the internal standard peak (CDCl3/TMS, δ 0.00 for 1H (TMS), δ 77.16 for 13C (CDCl3)). Multiplicity is reported as s = singlet, d = doublet, dd = doublet of doublet, t = triplet, q = quartet, p = quintet, m = multiplet. Coupling-constants (J) are reported in Hertz (Hz) (Supplementary Fig. 1)
## (6-bromo-3-fluoropyridin-2-yl)methanol (2)
To a solution of 6-bromo-3-fluoro-2-methylpyridine (1, 10.7 g, 56.3 mmol, 1 eq) under an inert atmosphere at 0 °C in DCM (370 mL) was added portion-wise m-CPBA (23.6 g, 70–$75\%$, 100 mmol, 1.8 eq). The reaction mixture was stirred at room temperature (rt) for 4 days. Sat. NaHCO3 and sat. Na2S2O3 was added (1:1, v/v) and the layers were separated. The aqueous layer was extracted thrice with DCM. The combined organic layer was dried over MgSO4, filtered, and concentrated under reduced pressure. To the residue was added TFAA (17 mL, 122 mmol, 2.2 eq) at 0 °C. After 15 min, the temperature was increased to 55 °C for 3 h. The mixture was concentrated under reduced pressure, redissolved in DCM and sat. Na2CO3 was added. The layers were separated and the organic layer was washed with sat. NaHCO3. The solvent was evaporated and the crude was dissolved in THF:MeOH (20:1, v/v) and K2CO3 (18.2 g, 132 mmol, 2.3 eq) was added. After 17 h H2O was added and the layers were separated. The aqueous layer was extracted thrice with EtOAc. The combined organic layers were dried over MgSO4, filtered, and the solvent evaporated under reduced pressure. The crude was purified with flash column chromatography (10–$20\%$ EtOAc in pentane) to yield 5.79 g (19.7 mmol, $35\%$) of a white solid. 1H‑NMR (500 MHz, CDCl3) δ 7.42 (ddt, $J = 8.5$, 3.5, 0.7 Hz, 1H), 7.29 (t, $J = 8.5$ Hz, 1H), 4.80 (d, $J = 3.3$ Hz, 2H). 13C-NMR (126 MHz, CDCl3) δ 156.10 (d, $J = 256.2$ Hz), 148.74 (d, $J = 19.1$ Hz), 135.01 (d, $J = 2.9$ Hz), 128.17 (d, $J = 4.2$ Hz), 126.09 (d, $J = 19.8$ Hz), 59.07.
## (6-bromo-3-fluoropyridin-2-yl)methyl methanesulfonate (3)
To a cooled (0 °C) mixture of (6-bromo-3-fluoropyridin-2-yl)methanol (1.6 g, 7.8 mmol, 1 eq) and Et3N (2.5 mL, 17.9 mmol, 2.3 eq) in dry THF (40 mL) was added dropwise MsCl (1.0 mL, 12.9 mmol, 1.7 eq). After stirring at rt for 1 h the solution was concentrated under reduced pressure. DCM and H2O were added and the layers were separated. The aqueous layer was extracted thrice with DCM. The combined organic layers were washed with brine, dried over MgSO4, filtered, and the solvent evaporated under reduced pressure to yield 1.65 g (5.8 mmol, $75\%$) of an yellow solid. 1H‑NMR (500 MHz, CDCl3) δ 7.52 (dd, $J = 8.6$, 3.5 Hz, 1H), 7.37 (t, $J = 8.5$ Hz, 1H), 5.33 (d, $J = 2.1$ Hz, 2H), 3.13 (s, 3H). 13C‑NMR (126 MHz, CDCl3) δ 157.82 (d, $J = 261.3$ Hz), 142.15 (d, $J = 16.0$ Hz), 130.74 (d, $J = 4.4$ Hz), 127.06 (d, $J = 20.4$ Hz), 65.50 (d, $J = 1.6$ Hz), 38.39.
## N-((6-bromo-3-fluoropyridin-2-yl)methyl)tetrahydro-2H-pyran-4-amine (4)
(6-Bromo-3-fluoropyridin-2-yl)methyl methanesulfonate (1.49 g, 5.3 mmol, 1 eq), K2CO3 (1.6 g, 11.6 mmol, 2.2 eq) and tetrahydro-2H-pyran-4-amine (0.66 mL, 6.7 mmol, 1.3 eq) were suspended in acetonitrile and stirred at 50 °C for 6 h, then an additional 3 days at rt. After dilution with DCM and H2O the layers were separated. The aqueous layer was extracted thrice with DCM. The combined organic layers were dried over MgSO4, filtered, and the solution evaporated under reduced pressure. The crude was purified with flash column chromatography (20–$100\%$ EtOAc in pentane) to yield 1.01 g (3.5 mmol, $67\%$) as a yellow oil. 1H‑NMR (300 MHz, CDCl3) δ 7.40 (dd, $J = 8.6$, 3.6 Hz, 1H), 7.35–7.26 (m, 1H), 4.08–3.95 (m, 4H), 3.42 (td, $J = 11.6$, 2.2 Hz, 2H), 2.74 (tt, $J = 10.5$, 4.1 Hz 1H), 1.89 (ddd, $J = 12.7$, 4.5, 2.3 Hz, 2H), 1.52 (dtd, $J = 13.1$, 11.0, 4.5 Hz, 2H). 13C‑NMR (75 MHz, CDCl3) δ 157.12 (d, $J = 255.9$ Hz), 149.21 (d, $J = 17.0$ Hz), 127.83 (d, $J = 4.2$ Hz), 125.97 (d, $J = 21.2$ Hz), 66.76, 53.64, 44.90, 33.59.
## 2-((4-bromobenzyl)amino)acetamide (6)
To a mixture of 4-bromobenzaldehyde (5, 9.2 g (49.7 mmol, 1.1 eq) and 2‑aminoacetamide hydrochloride (5.06 g, 45.8 mmol, 1.0 eq) in MeOH:H2O (170 mL, 5:1, v/v) was added NaOH (2.06 g, 51.5 mmol, 1.1 eq) and left to stir at rt overnight. NaBH4 (3.6 g, 95.2 mmol, 2.1 eq) was added and the solution was stirred overnight at rt. The solution was acidified to pH 3 with 2 M HCl, then neutralized with sat. aqueous NaHCO3. Methanol was evaporated under reduced pressure and the resulting slurry was filtered to yield 11.0 g (45.2 mmol, $91\%$) of a white solid. 1H‑NMR (300 MHz, methanol-d4) δ 7.69–7.59 (m, 2H), 7.47–7.38 (m, 2H), 4.22 (s, 2H), 3.81 (s, 2H).
## 1-(4-bromobenzyl)imidazolidine-2,4-dione (7)
To a suspension of 2-((4-bromobenzyl)-amino)acetamide (10.0 g, 40.1 mmol, 1,0 eq) in acetonitrile (300 mL) were added CDI (13.86 g, 85.5 mmol, 2.1 eq) and DMAP (10.2 g, 83.5 mmol, 2.1 eq). The mixture was heated to 60 °C under inert atmosphere for 70 h. HCl (1 M, 250 mL) was added and the aqueous layer extracted thrice with EtOAc. The combined organic layers were washed with H2O and brine, dried over MgSO4, filtered, and the solvent evaporated under reduced pressure. The crude was purified with flash column chromatography with dry loading over Celite (5-$10\%$ acetone in DCM) to yield 3.95 g (14.7 mmol, $37\%$) of a yellow solid. 1H‑NMR (300 MHz, CDCl3) δ 7.83 (bs, 1H), 7.56 – 7.45 (m, 2H), 7.20 – 7.10 (m, 2H), 4.49 (s, 2H), 3.79 (s, 2H). 13C‑NMR (75 MHz, CDCl3) δ 132.41, 129.95, 77.58, 77.16, 76.74, 50.36, 46.01.
## 1-(4-bromobenzyl)-3-isobutylimidazolidine-2,4-dione (8)
To solution of 1-(4-bromobenzyl)imidazolidine-2,4-dione (2.00 g, 7.4 mmol, 1,0 eq) in anhydrous DMF (18 mL) were subsequently added K2CO3 (3.08 g, 22.3 mmol, 3,0 eq) and 1-bromo-2-methylpropane (1.62 mL, 14.9 mmol, 2,0 eq) and the mixture was stirred for 20 h at rt. The mixture was filtered and the filtrate diluted with diethyl ether and washed thrice with water (3 × 50 mL). The combined organic layers were washed with brine, dried (MgSO4), filtered, and concentrated under reduced pressure. The crude was purified with flash column chromatography (10–$40\%$ EtOAc in pentane) to yield 2.12 g (6.52 mmol, $88\%$) of a white solid. LCMS (LCQ Fleet, 10–$90\%$): tr = 7.00 min, m/z: 325.17 [M + H]+, 327.08 [M + H]+ (Br). 1H-NMR (300 MHz, CDCl3) δ 7.47 (d, $J = 8.3$ Hz, 2H), 7.14 (d, $J = 8.3$ Hz, 2H), 4.52 (s, 2H), 3.74 (s, 2H), 3.33 (d, $J = 7.4$ Hz, 2H), 2.15–2.04 (m, 1H), 0.91 (d, $J = 6.8$ Hz, 6H). 13C NMR (75 MHz, CDCl3) δ 169.57, 156.78, 134.41, 131.79, 129.48, 121.77, 60.01, 48.61, 45.98, 45.71, 28.57, 19.70.
## 3-isobutyl-1-(4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)benzyl)imidazolidine-2,4-dione (9)
A mixture of 1-(4-bromobenzyl)-3-isobutylimidazolidine-2,4-dione (0.50 g, 1.54 mmol, 1 eq), KOAc (0.66 g, 6.76 mmol, 4.4 eq) and bis(pinacolato)diboron (0.59 g, 2.31 mmol, 1.5 eq) in DMF (10 mL) was sonicated for 15 min under argon flow. Subsequently, Pd(dppf)Cl2 (0.07 g, 0.09 mmol, 0.06 eq) was added and the mixture was stirred at 75 °C for 20 h. The mixture was cooled to rt, diluted with EtOAc (100 mL) and water (10 mL) and the layers were separated. The water layer was extracted thrice with EtOAc (3 × 20 mL). The combined organic layers were extracted with sat. aqueous NaHCO3, water and brine, dried (MgSO4), filtered, and concentrated under reduced pressure. The raw product was co-evaporated with CHCl3 and used in the next step without further purification.
## 1-(4-(5-fluoro-6-(((tetrahydro-2H-pyran-4-yl)amino)methyl)pyridin-2-yl)benzyl)-3-isobutylimidazolidine-2,4-dione (LEI-102)
To a degassed mixture of N-((6-bromo-3-fluoropyridin-2-yl)methyl)tetrahydro-2H-pyran-4-amine (4, 0.29 g, 1.0 mmol, 1,0 eq), 3-isobutyl-1-(4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)benzyl)imidazolidine-2,4-dione (9, 0.56 g, ~1.5 mmol, crude) and K2CO3 (1.29 g, 6.0 mmol, 6,0 eq) in toluene:ethanol (10 mL, 4:1, v/v) was added under argon atmosphere Pd(PPh3)4 (0.18 g, 0.10 mmol, 0.1 eq). The resulting mixture was stirred for 18 h at 75 °C, subsequently cooled to rt, and filtered. The filtrate was diluted with EtOAc and washed with water and brine, dried (MgSO4), filtered, and concentrated under reduced pressure. The crude was purified with flash column chromatography (0‑$20\%$ MeOH in EtOAc) to yield 0.24 g of a white solid (0.53 mmol, $53\%$). Further purification with preparative HPLC resulted in a yield of 0.204 g (0.45 mmol, $45\%$). LCMS (LCQ Advantage, 10–$90\%$): tr = 5.32 min, m/z: 455.27 [M + H]+, 908.93 [2 M + H]+. HRMS (ESI+) m/z: calcd. for C25H32FN4O3 [M + H], 455.245; found, 455.245. 1H NMR (400 MHz, CD3CN) δ 8.05 (d, $J = 8.3$ Hz, 2H), 7.86 (dd, $J = 8.7$, 3.6 Hz, 1H), 7.61 (t, $J = 9.0$ Hz, 1H), 7.34 (d, $J = 8.1$ Hz, 2H), 4.53 (s, 2H), 4.45 (s, 2H), 3.93 (dd, $J = 11.4$, 4.4 Hz, 2H), 3.77 (s, 1H), 3.47 (tt, $J = 11.8$, 3.8 Hz, 2H), 3.30 (td, $J = 11.9$, 1.9 Hz, 2H), 3.24 (d, $J = 7.3$ Hz, 2H), 2.05 (br d, $J = 13.3$ Hz, 2H), 1.99 (dt, $J = 13.2$, 6.6 Hz, 1H), 1.83 (qd, $J = 12.1$, 4.5 Hz, 2H), 0.88 (d, $J = 6.7$ Hz, 6H). 13C NMR (100 MHz, CD3CN) δ 171.54, 157.25 (d, $J = 226.6$ Hz), 156.12, 153.03 (d, $J = 4.5$ Hz), 140.51 (d, $J = 16.1$ Hz), 138.83, 137.70, 129.13, 128.22, 125.56 (d, $J = 18.8$ Hz), 122.81 (d, $J = 4.3$ Hz), 118.38, 66.55, 55.55, 50.30, 46.81 (d, $J = 7.9$ Hz), 42.95, 30.02, 28.32, 20.32.
## Constructs
The N-BRIL fused wild type (WT) human CB2R construction and co-expression of G protein for cryo-EM study were performed using the similar procedure as described before34. In brief, the WT human CB2R was modified to contain a fusion protein BRIL to improve the protein expression and thermostability, along with a 10×His-tag and a FLAG-tag at the N-terminal. The CB2R, Gαi1 and Gβ1γ2 subunits were cloned into the pFastBac vector separately using cloning kits.
## Expression and purification of CB2R-Gi-Scfv16 complexes
Methods of complex expression and purification in the current study have been described previously34. The CB2R and Gi heterotrimer were co-expressed in Sf 9 insect cells using the Bac-to-Bac Baculovirus Expression System (Invitrogen). Cells were infected with three separate virus preparations for CB2R, Gαi1 and Gβ1γ2 at a ratio of 1:2:2 at a cell density of 2.5 × 106 cells/mL. After 48 h, the cell culture was collected by centrifugation and the cell pellets were stored at −80 °C until use. The cell pellets were thawed and lysed in the hypotonic buffer of 10 mM HEPES (pH 7.5), 10 mM MgCl2, 20 mM KCl with EDTA-free complete protease inhibitor cocktail tablets (Roche, #5056489001). The CB2R-Gi complex was formed in membranes by addition of 25 μM agonist (LEI-102, APD371, HU308, and CP55,940, respectively) and 2 units of apyrase (NEB, #M0398S) in the presence 500 µg scFv16. The lysate was incubated for overnight at 4 °C and discard the supernatant by centrifugation at 186,000 × g for 30 min. Subsequently, the solubilization buffer containing 50 mM HEPES (pH 7.5), 100 mM NaCl, $0.75\%$ (w/v) lauryl maltose neopentyl glycol (LMNG, Anatrace, #4216588), $0.15\%$ (w/v) cholesterol hemisucinate (CHS, Sigma-Aldrich, #C6512) supplemented with 25 μM agonist and 2 units of apyrase (NEB) were added to solubilize complexes for 2 h at 4 °C. Insoluble material was removed by centrifugation at 186,000 × g for 30 min and the supernatant was immobilized by batch binding to TALON IMAC resin (Clontech, #635507) including 20 mM imidazole over 6 h at 4 °C. Then, the resin was packed and washed with 15 column volumes (CVs) of washing buffer I containing 25 mM HEPES (pH 7.5), 100 mM NaCl, $10\%$ (v/v) glycerol, $0.1\%$ (w/v) LMNG, $0.02\%$ (w/v) CHS, 30 mM imidazole and 20 μM agonist, and 15 CVs of washing buffer II containing 25 mM HEPES (pH 7.5), 100 mM NaCl, $10\%$ (v/v) glycerol, $0.03\%$ (w/v) LMNG, $0.006\%$ (w/v) CHS, 50 mM imidazole and 20 μM agonist. After that, the protein was eluted using 3 CVs of elution buffer containing 25 mM HEPES (pH 7.5), 100 mM NaCl, $10\%$ (v/v) glycerol, $0.01\%$ (w/v) LMNG, $0.002\%$ (w/v) CHS, 250 mM imidazole and 25 μM agonist. Finally, the complex was concentrated using the centrifugal filter with 100 kDa molecular weight cutoff and loaded onto a Superdex200 $\frac{10}{300}$ GL column (GE Healthcare) with buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, $0.00075\%$ (w/v) LMNG, $0.00025\%$ GDN (Anatrace, #GDN101), $0.0001\%$ (w/v) CHS, 100 μM TCEP. The fractions consisting of purified CB2R-Gi complex were collected and concentrated to 0.8–1.0 mg/mL for electron microscopy experiments.
## Cryo-EM grid preparation and data collection
For cryo-EM grids preparation of the CB2R-Gi complexes, 3 μL of the concentrated protein was loaded to a glow-discharged holey carbon grid (CryoMatrix Amorphous alloy film R$\frac{1.2}{1.3}$, 300 mesh), and subsequently were plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). The chamber of Vitrobot was set to $100\%$ humidity at 4 °C. The sample was blotted for 2.5 s with blot force 2. Cryo-EM images were collected on a Titan Krios microscope operated at 300 kV equipped with a Gatan Quantum energy filter, with a slit width of 20 eV, a Gatan K2 summit direct electron camera (Gatan). Images were taken at a dose rate of 8e−/Å2/s with a defocus range of −0.8 to −2.0 μm using SerialEM software64 in EFTEM nanoprobe mode, with 50 μm C2 aperture, at a calibrated magnification of 130,000 corresponding to a magnified pixel size of 1.04 Å. The total exposure time was 8.1 s and 45 frames were recorded per micrograph.
## Cryo-EM image processing
The cryo-EM data processing was performed with CryoSPARC65. For CB2R-Gαi-scFv16-APD371/LEI-102/HU308/CP55,940 dataset, a total of 7443, 5282, 7530, and 6473 movies were collected, respectively. For all datasets, patch motion correction was used for beam-induced motion correction. Contrast transfer function (CTF) parameters for each micrograph were determined by patch CTF estimation. Using Blob Picker in CryoSPARC to auto pick particles in the first 500 micrographs of CB2R-Gαi-scFv16-APD371 complex dataset and then 258,347 particles were extracted to conduct 2D classification. 9277 particles in good 2D patterns were selected as templates to pick better particles. 5,239,870, 3,398,611, 4,653,294, and 3,595,875 particles extracted, respectively, in a 256 Å box were divided into three hundred two-dimensional (2D) class averages with a maximum alignment resolution of 6 Å. Then, 1,152,146, 762,471, 355,832, and 440,292 particles were selected from good 2D classification after two round 2D classification, individually. Following 2D classification, these particles were subjected for ab initio reconstruction into four classes. After heterogeneous refinement, homogeneous refinement, non-uniform refinement and local refinement of the best-looking dataset in CryoSPARC, the final map has an indicated global resolution of 3.08, 2.98, 2.97, and 2.84 Å at a Fourier shell correlation (FSC) of 0.143, respectively. Local resolution was determined using the Bsoft package with half maps as input maps66.
## Model building and refinement
For CB2R-Gi-scFv16 complex, the CB2R-AM12033 cryo-EM structure and Gi protein in CB2R were used as the starting model. The model was docked into the EM density map using Chimera67, followed by iterative manual adjustment and rebuilding in COOT68 and phenix.real_space_refine in Phenix69. The model statistics were validated using MolProbity70. Structural figures were prepared in Chimera and PyMOL (http://www.pymol.org). The final refinement statistics were provided in Supplementary Table 4. The extent of any model overfitting during refinement was measured by refining the final model against one of the half-maps and by comparing the resulting map versus model FSC curves with the two half-maps and full model.
## Generation of mutants
The WT CB1R and CB2R genes were subcloned into vector pcDNA3.1 with an N-terminal HA signal peptide and FLAG-tag. Mutations were introduced by QuikChange PCR (as described by supplier).
## Transfection
24 h prior to transfection, HEK293T cells were seeded on 10 cm ø plates to reach approximately $50\%$ confluence at the start of transfection. The cells were transfected with 10 µg plasmid DNA of WT hCB2R or hCB1R receptor, or mutant receptor using the calcium phosphate precipitation method71. In short, a DNA-calcium mix was made containing 270 mM CaCl2 and 10 µg plasmid DNA to which Hank’s Balanced Salt Solution (HBSS; 280 mM NaCl, 10 mM KCl, 1.5 mM Na2HPO4, and 50 mM HEPES) was added in a 1:1 (v/v) ratio and mixed by aeration to create consistent calcium phosphate precipitates. For transfection, 1 mL DNA-calcium mix was added per 10 cm ø plate, followed by a 48 h incubation under a humidified atmosphere at 37 °C with $5\%$ CO2.
## Enzyme-linked immunosorbent assay (ELISA)
Receptor expression after transfection was measured in an enzyme-linked immunosorbent assay (ELISA). After 24 h of transfection, HEK293T cells were detached with phosphate-buffered saline (PBS)/EDTA and seeded into a sterile 96-well poly-D-lysine coated plate at a density of 100,000 cells per well and kept under a humidified atmosphere at 37 °C with $5\%$ CO2. After an additional 24 h, cells were washed with PBS and fixed with $4\%$ formaldehyde for 10 min at room temperature (rt). Cells were washed twice with Tris-buffered saline (TBS) and were blocked with TBS supplemented with $0.1\%$ TWEEN 20 (TBST) and $2\%$ BSA (w/v)) for 30 min at rt while shaking. Subsequently, the cells were incubated with monoclonal M2 mouse anti-FLAG primary antibody (1:4000) for 2 h at rt while shaking. After removal of the antibody, the cells were washed three times with TBST and incubated with the secondary goat anti-mouse HRP-conjugated antibody (1:10,000) for 1 h at rt while shaking. After a final wash with TBS, the cells were treated with 3,3’,5,5’-Tetramethylbenzidine (TMB, Sigma-Aldrich #T0440) in the dark for maximally 10 min at rt to visualize immunoreactivity. The reaction was quenched with 1 M H3PO4, and absorbance was read at 450 nm with a Wallac EnVision 2104 Multilabel reader (PerkinElmer).
## Membrane preparation
For membrane preparation, HEK293T cells were harvested 48 h after transfection. Cells were detached by scraping into 3 mL of PBS and subsequently centrifuged at 2000 × g for 5 min. Pellets were resuspended in ice-cold Tris buffer (50 mM Tris-HCl, pH 7.4) and homogenized with an Ultra Turrax homogenizer (IKA-Werke GmbH & Co. KG, Staufen, Germany). Cytosolic and membrane fractions were separated using a high-speed centrifugation step of 31,000 rpm in a Beckman Optima LE-80K ultracentrifuge with Ti70 Rotor for 20 min at 4 °C. After a second cycle of homogenization and centrifugation, the final pellets were resuspended in 50 mM Tris-HCl pH 7.4, 5 mM MgCl2 and stored in 100 µL aliquots at −80 °C until use. CHOK1_hCB2bgal cells were harvested when reaching $90\%$ confluence in 15 cm ø plates after one week subculture at a 1:6 ratio. Membrane preparation followed a similar procedure as described above. Final membrane pellets were resuspended in 50 mM Tris-HCl pH 7.4 and stored in 100 µL aliquots at −80 °C until use. Membrane protein concentrations were determined using a BCA protein determination assay as described by the manufacturer72.
## [3H]RO6957022 competition association assays
For assessment of kinetic agonist binding at hCB2R, [3H]RO6957022 competition association assays were executed. These assays were previously described with the main difference of incubation at 25 °C compared to 10 °C for identification of more distinct kinetic differences73. In short, prior to kinetic assessment of agonist binding, the affinity (IC50) of the agonists at the hCB2R was determined in [3H]RO6957022 displacement assays. CHOK1_hCB2bgal were thawed, homogenized, and subsequently diluted to 1 µg protein per well. When studying endocannabinoids, membranes were preincubated with 50 µM PMSF for 30 min. Membranes were incubated with ~1.5 nM [3H]RO6957022 and six increasing concentrations of competing agonists in a total volume of 100 µL assay buffer (50 mM Tris-HCl (pH 7.4), $0.1\%$ (w/v) BSA). Incubations were done for 2 h at 10 °C to reach equilibrium. Subsequently, in competition association assays, agonists were incubated at their IC50 concentration in the presence of ~1.5 nM [3H]RO6957022 in a total volume of 100 µL assay buffer at 10 °C. Competition was initiated by addition of membrane homogenates at different time points for 2 h. Nonspecific binding (NSB) was determined with 10 µM AM630 and organic solvent (DMSO or acetonitrile) concentrations were <$1\%$ in all samples. Total radioligand binding (TB) did not exceed $10\%$ of the amount added to prevent ligand depletion. Incubations were terminated by rapid vacuum filtration with ice-cold 50 mM Tris-HCl (pH 7.4), $0.1\%$ (w/v) BSA buffer through Whatman GF/C filters using a Filtermate 96-well harvester (PerkinElmer). Filters were dried for at least 30 min at 55 °C, and subsequently 25 µL MicroScint scintillation cocktail (PerkinElmer #6013621) was added per well. Filter-bound radioactivity was measured by scintillation spectrometry using a Microbeta2 2450 counter (PerkinElmer).
## [35S]GTPγS binding assays
G protein activation by agonists LEI-102, APD371, HU308, CP55,940 AEA, and 2-AG was measured by binding of radiolabeled [35S]GTPγS to the cannabinoid receptors as previously described25. In short, transient HEK293T membrane homogenates (10 µg/well) were diluted in assay buffer (50 mM Tris-HCl (pH 7.4), 5 mM MgCl2, 150 mM NaCl, 1 mM EDTA, $0.05\%$ BSA (w/v) and 1 mM DTT, freshly prepared every day) and were pretreated with 10 µg saponin and 1 µM GDP. For endocannabinoid samples, the membranes were additionally pretreated for 30 min with 50 µM PMSF before agonist addition. To determine the G protein activation, the membranes were incubated with 10 µM or six increasing concentrations of agonist (ranging from 0.01 nM to 10 µM) for 30 min at rt. Basal receptor activity was determined in the presence of vehicle only ($0.2\%$ DMSO/acetonitrile). [ 35S]GTPγS (0.3 nM) was added and the mixture was co-incubated for an additional 90 min at 25 °C while shaking at 400 rpm. Filtration was performed, and filter-bound radioactivity was determined as described under [3H]RO6957022 Competition Association Assays except for using ice-cold 50 mM Tris-HCl (pH 7.4), 5 mM MgCl2 buffer.
## [3H]CP55,940 homologous and heterologous displacement assays
Agonist affinity (Ki) on WT and mutant receptors was determined in [3H]CP55,940 displacement assays. The amount of transient HEK293T membrane, ranging from 0.75 µg to 10 µg protein per well, was chosen to obtain a specific [3H]CP55,940 binding window of 1200-1500 disintegrations per minute (dpm) except for the CB2R-QuadrupleTM1,7 mutant, for which a window of ~500 dpm could be obtained using 20 µg protein per well. Membranes were thawed and subsequently homogenized using the Ultra Turrax homogenizer. For the endocannabinoid assays, the membranes were preincubated for 30 min with 50 µM PMSF. Homologous displacement assays were performed with 1.5 nM final concentration [3H]CP55,940 and when necessary supplemented with an additional concentration of 0.55 nM [3H]CP55,940 in the presence of competing CP55,940 (ranging from 0.01 nM to 1 µM) in assay buffer (50 mM Tris-HCl (pH 7.4), 5 mM MgCl2, $0.1\%$ (w/v) BSA). Heterologous displacement assays were executed for LEI-102, APD371, HU308, AEA, and 2-AG using 1.5 nM final concentration [3H]CP55,940 with one concentration (10 µM) or six increasing concentrations (ranging from 0.1 nM to 10 µM) in assay buffer. For both assays, binding was initiated by addition of membrane homogenates to reach a final volume of 100 µL. NSB was determined using 10 µM CP55,940 and organic solvent (DMSO or acetonitrile) concentrations were <$1\%$ in all samples. TB did not exceed $10\%$ of the amount added to prevent ligand depletion. Incubation was done for 2 h at 25 °C to reach equilibrium. Filtration was performed, and filter-bound radioactivity was determined as described under [3H]RO6957022 Competition Association Assays except for using ice-cold 50 mM Tris-HCl (pH 7.4), 5 mM MgCl2, $0.1\%$ (w/v) BSA buffer.
## Cisplatin-induced nephropathy
Ten to twelve-week-old male/female C57BL/6J mice were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). CB2R knockout mice (CB2R−/−) and their wild-type littermates (CB2R+/+) were developed as described previously and had been backcrossed to a C57BL/6J background74. All animals were kept in a temperature-controlled environment (20–22 °C) with a 12 h light–dark cycle and were always allowed free access to food and water. All animal experiments reported in this manuscript complied with the National Institutes of Health “Guide for the Care and Use of Laboratory Animals” (NIH publication 86–23 revised 1985) and were approved by the Institutional Animal Care and Use Committee of the National Institute on Alcohol Abuse and Alcoholism (Bethesda, MD).
The well-established model of cisplatin-induced nephropathy was used63. Mice (CB2R−/− and CB2R+/+) were sacrificed 72 h after a single injection of cisplatin (cis-diamine platinum (II) dichloride (Sigma#P4394) 25 mg/kg i.p.; freshly dissolved in physiological saline) by cervical dislocation under deep anesthesia with $5\%$ isoflurane, for collection of blood and tissue samples. LEI-102 was given i.p. or by oral gavage (p.o.) at 0.3, 3.0, and 10 mg/kg every day, starting 1.5 h before the cisplatin exposure. The drug was dissolved in a vehicle of DMSO:Tween 80:saline, 1:1:18. After administration of LEI‐102, mice were killed by cervical dislocation under deep anesthesia with $5\%$ isoflurane, for collection of blood and tissue samples at the time described in the figure. The tetrad assay in mice has previously been described in detail29.
## Biochemistry, histopathology, immunostaining, real-time PCR
Markers of kidney dysfunction (BUN and CREA), histopathology (PAS staining), immunostaining or ELISA for 3-nitrotyrosine (3-NT; Cell Biolabs #STA-305) and 4-hydroxynonenal (4-HNE; Cell Biolabs#STA-838), and real-time PCR (Primers from Qiagen, SYBER Green Vita Scientific#MEIF01301, High-Capacity cDNA Reverse Transcription Kit, Thermo Fisher Scientific#4368813) for inflammatory cytokines were performed as previously described63. Tubular damage scores were determined based on the percentage of tubules showing epithelial necrosis where 0= normal; 1, <$10\%$; 2, 10–$25\%$; 3, 26–$75\%$; 4, >$75\%$. Tubular necrosis was defined as the loss of the proximal tubular brush border, blebbing of apical membranes, tubular epithelial cell detachment from the basement membrane, or intraluminal aggregation of cells and proteins. The morphometric examination was performed in a blinded manner. Ten fields were scored from each mouse kidneys at 200× magnification, and average scores were determined for each mouse. For final quantification graph, average tubular damage scores of six mice/group were plotted.
## Quantification and statistical analysis
All experimental data were analyzed using GraphPad Prism 9.0 (GraphPad Software Inc., San Diego, CA). All values obtained are means ± standard error of the mean (SEM) of at least three independent experiments performed in duplicate, unless stated otherwise.
From [3H]RO6957022 competition association assays, the kon and koff were determined by non-linear regression analysis, using the “kinetics of competitive binding” model as described by Motulsky and Mahan (Motulsky and Mahan, 1984):\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${K}_{a}={k}_{1}\cdot \left[L\right]\cdot {10}^{-9}+{k}_{2}$$\end{document}Ka=k1⋅L⋅10−9+k2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${K}_{b}={k}_{3}\cdot \left[I\right]\cdot {10}^{-9}+{k}_{4}$$\end{document}Kb=k3⋅I⋅10−9+k4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S=\sqrt{{({K}_{a}-{K}_{b})}^{2}+4\cdot {k}_{1}\cdot {k}_{3}\cdot \left[L\right]\cdot \left[I\right]\cdot {10}^{-18}}$$\end{document}S=(Ka−Kb)2+4⋅k1⋅k3⋅L⋅I⋅10−18\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${K}_{f}=0.5\cdot \left({K}_{a}+{K}_{b}+S\right)$$\end{document}Kf=0.5⋅Ka+Kb+S\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${K}_{s}=0.5\cdot \left({K}_{a}+{K}_{b}-S\right)$$\end{document}Ks=0.5⋅Ka+Kb−S\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q=\frac{B\max \cdot {k}_{1}\cdot \left[L\right]\cdot {10}^{-9}}{{K}_{f}-{K}_{s}}$$\end{document}Q=Bmax⋅k1⋅L⋅10−9Kf−Ks\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left[Y\right]=Q\cdot \left(\frac{{k}_{4}\cdot ({K}_{f}-{K}_{s})}{{K}_{f}\cdot {K}_{s}}+\frac{{k}_{4}-{K}_{f}}{{K}_{f}}\cdot {{{{{{\rm{e}}}}}}}^{\left(-{K}_{f}\cdot X\right)}-\frac{{k}_{4}-{K}_{s}}{{K}_{s}}\cdot {{{{{{\rm{e}}}}}}}^{\left(-{{K}_{S}} \cdot {X}\right)}\right)$$\end{document}Y=Q⋅k4⋅(Kf−Ks)Kf⋅Ks+k4−KfKf⋅e−Kf⋅X−k4−KsKs⋅e−KS⋅XWhere [L] is the radioligand concentration per experiment (~1.5 nM), I is the IC50 concentration of agonist (nM), X is the time (s), and Y is the specific binding of the radioligand (dpm). Ka and Kb are the observed association rate constants (kobs) of the radioligand and the agonist of interest, respectively. k1 and k3 are the association rate constants (kon in M−1s−1) of [3H]RO6957022 (determined per experiment) and the agonist of interest, respectively. Similarly, k2 and k4 are the dissociation rate constants (koff in s−1) of [3H]RO6957022 (experimentally determined at 4.3 × 10−4 s−1, data not shown) and the agonist of interest, respectively. The engagement time (ET in seconds) of the agonists of interest was determined at 1 µM of agonist using the equation ET = 1/(kon · 1 × 10−6). The residence time (RT in min) was calculated using the equation RT = 1/(60 · koff)75. The association and dissociation rate constants were used to calculate the kinetic KD using: KD = koff/kon.
[35S]GTPγS agonist responses on hCB2R constructs were baseline-corrected for the individual mutant’s basal activity. The responses were normalized to the basal activity of the construct ($0\%$) and top of the CP55,940 (for WT responses only) or WT curve (for mutants, $100\%$). The potency (pEC50) and efficacy (Emax) values were obtained by non-linear regression to a sigmoidal concentration-effect curve with a Hill slope of 1 by using the “log(agonist) vs response (three parameters)” model. [ 35S]GTPγS data from hCB1R constructs were expressed as fold over the mutant’s basal activity to also quantify the effects of CB2R selective agonists.
Displacement assays were baseline-corrected with NSB and normalized to this value ($0\%$) and TB ($100\%$). The equilibrium dissociation constants (KD) of [3H]CP55,940 on different mutants were calculated from homologous displacements by non-linear regression analysis, using the “one-site homologous” model. The half-maximal inhibitory concentrations (pIC50) of the agonists in [3H]CP55,940 and [3H]RO6957022 assays were obtained by non-linear regression analysis of the homologous and heterologous displacement curves and further converted into inhibitory constant pKi using the Cheng-Prusoff equation76. In which the experimentally determined KD for each construct was used for [3H]CP55,940 assays or 0.78 nM for [3H]RO6957022 assays (data not shown).
Differences in pEC50, Emax, pKD and pKi values for each mutant compared to WT were analyzed using a one-way Welch’s ANOVA with Dunnett’s T3 multiple comparisons test or an unpaired Student’s t-test with Welch’s correction. Significant differences are displayed as *$p \leq 0.05$; **$p \leq 0.01$, ***$p \leq 0.001$ and ****$p \leq 0.0001.$ For the animal experiments all the values are represented as mean ± SEM. Statistical analysis of the data was performed by analysis of variance (ANOVA) followed by Tukey’s post hoc test for multiple comparisons or t-test if appropriate. The analysis was conducted using GraphPad Prism 9 software. $p \leq 0.05$ was considered statistically significant.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary Information Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-37112-9.
## Source data
Source Data
## Peer review information
Nature Communications thanks Matthew Belousoff, Frank Park and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
## References
1. Mechoulam R, Hanus LO, Pertwee R, Howlett AC. **Early phytocannabinoid chemistry to endocannabinoids and beyond**. *Nat. Rev. Neurosci.* (2014.0) **15** 757-764. DOI: 10.1038/nrn3811
2. Pacher P, Batkai S, Kunos G. **The endocannabinoid system as an emerging target of pharmacotherapy**. *Pharmacol. Rev.* (2006.0) **58** 389-462. DOI: 10.1124/pr.58.3.2
3. Badowski ME, Yanful PK. **Dronabinol oral solution in the management of anorexia and weight loss in AIDS and cancer**. *Ther. Clin. Risk Manag.* (2018.0) **14** 643-651. DOI: 10.2147/TCRM.S126849
4. Grimison P. **Oral THC:CBD cannabis extract for refractory chemotherapy-induced nausea and vomiting: a randomised, placebo-controlled, phase II crossover trial**. *Ann. Oncol.* (2020.0) **31** 1553-1560. DOI: 10.1016/j.annonc.2020.07.020
5. Inglet S. **Clinical data for the use of cannabis-based treatments: a comprehensive review of the literature**. *Ann. Pharmacother.* (2020.0) **54** 1109-1143. DOI: 10.1177/1060028020930189
6. Jones E, Vlachou S. **A critical review of the role of the cannabinoid compounds Δ**. *Molecules* (2020.0) **25** 4930. DOI: 10.3390/molecules25214930
7. Adams IB, Martin BR. **Cannabis: pharmacology and toxicology in animals and humans**. *Addiction* (1996.0) **91** 1585-1614. DOI: 10.1111/j.1360-0443.1996.tb02264.x
8. Lucas CJ, Galettis P, Schneider J. **The pharmacokinetics and the pharmacodynamics of cannabinoids**. *Br. J. Clin. Pharmacol.* (2018.0) **84** 2477-2482. DOI: 10.1111/bcp.13710
9. Pacher P, Steffens S, Hasko G, Schindler TH, Kunos G. **Cardiovascular effects of marijuana and synthetic cannabinoids: the good, the bad, and the ugly**. *Nat. Rev. Cardiol.* (2018.0) **15** 151-166. DOI: 10.1038/nrcardio.2017.130
10. Riera R, Pacheco RL, Bagattini AM, Martimbianco ALC. **Efficacy and safety of therapeutic use of cannabis derivatives and their synthetic analogs: Overview of systematic reviews**. *Phytother. Res.* (2022.0) **36** 5-21. DOI: 10.1002/ptr.7263
11. Munro S, Thomas KL, Abu-Shaar M. **Molecular characterization of a peripheral receptor for cannabinoids**. *Nature* (1993.0) **365** 61-65. DOI: 10.1038/365061a0
12. Howlett AC. **International Union of Pharmacology. XXVII. Classification of cannabinoid receptors**. *Pharmacol. Rev.* (2002.0) **54** 161-202. DOI: 10.1124/pr.54.2.161
13. Pertwee RG. **International Union of Basic and Clinical Pharmacology. LXXIX. Cannabinoid receptors and their ligands: beyond CB**. *Pharmacol. Rev.* (2010.0) **62** 588-631. DOI: 10.1124/pr.110.003004
14. Howlett AC, Abood ME. **CB1 and CB2 receptor pharmacology**. *Adv. Pharmacol.* (2017.0) **80** 169-206. DOI: 10.1016/bs.apha.2017.03.007
15. Howlett AC, Blume LC, Dalton GD. **CB(1) cannabinoid receptors and their associated proteins**. *Curr. Med. Chem.* (2010.0) **17** 1382-1393. DOI: 10.2174/092986710790980023
16. Galiegue S. **Expression of central and peripheral cannabinoid receptors in human immune tissues and leukocyte subpopulations**. *Eur. J. Biochem.* (1995.0) **232** 54-61. DOI: 10.1111/j.1432-1033.1995.tb20780.x
17. Pacher P, Mechoulam R. **Is lipid signaling through cannabinoid 2 receptors part of a protective system**. *Prog. Lipid Res.* (2011.0) **50** 193-211. DOI: 10.1016/j.plipres.2011.01.001
18. Hanus L. **HU-308: a specific agonist for CB**. *Proc. Natl Acad. Sci. USA* (1999.0) **96** 14228-14233. DOI: 10.1073/pnas.96.25.14228
19. Guindon J, Hohmann AG. **Cannabinoid CB2 receptors: a therapeutic target for the treatment of inflammatory and neuropathic pain**. *Br. J. Pharmacol.* (2008.0) **153** 319-334. DOI: 10.1038/sj.bjp.0707531
20. Hussain MT, Greaves DR, Iqbal AJ. **The impact of cannabinoid receptor 2 deficiency on neutrophil recruitment and inflammation**. *DNA Cell Biol.* (2019.0) **38** 1025-1029. DOI: 10.1089/dna.2019.5024
21. Rizzo MD. **Targeting cannabinoid receptor 2 on peripheral leukocytes to attenuate inflammatory mechanisms implicated in HIV-associated neurocognitive disorder**. *J. Neuroimmune Pharmacol.* (2020.0) **15** 780-793. DOI: 10.1007/s11481-020-09918-7
22. Liu QR. **Anti-inflammatory and pro-autophagy effects of the cannabinoid receptor CB2R: possibility of modulation in type 1 diabetes**. *Front. Pharmacol.* (2021.0) **12** 809965. DOI: 10.3389/fphar.2021.809965
23. Brennecke B. **Cannabinoid receptor type 2 ligands: an analysis of granted patents since 2010**. *Pharm. Pat Anal.* (2021.0) **10** 111-163. DOI: 10.4155/ppa-2021-0002
24. Whiting ZM, Yin J, de la Harpe SM, Vernall AJ, Grimsey NL. **Developing the cannabinoid receptor 2 (CB2) pharmacopoeia: past, present, and future**. *Trends Pharmacol. Sci.* (2022.0) **43** 754-771. DOI: 10.1016/j.tips.2022.06.010
25. Soethoudt M. **Cannabinoid CB2 receptor ligand profiling reveals biased signalling and off-target activity**. *Nat. Commun.* (2017.0) **8** 13958. DOI: 10.1038/ncomms13958
26. Han S. **Discovery of APD371: identification of a highly potent and selective CB2 agonist for the treatment of chronic pain**. *ACS Med. Chem. Lett.* (2017.0) **8** 1309-1313. DOI: 10.1021/acsmedchemlett.7b00396
27. Yacyshyn BR. **Safety, pharmacokinetics, and efficacy of Olorinab, a peripherally acting, highly selective, full agonist of the cannabinoid receptor 2, in a phase 2a study of patients with chronic abdominal pain associated with Crohn’s disease**. *Crohn’s Colitis 360* (2021.0) **3** otaa089. DOI: 10.1093/crocol/otaa089
28. van der Stelt M. **Discovery and optimization of 1-(4-(pyridin-2-yl)benzyl)imidazolidine-2,4-dione derivatives as a novel class of selective cannabinoid CB2 receptor agonists**. *J. Med. Chem.* (2011.0) **54** 7350-7362. DOI: 10.1021/jm200916p
29. Mukhopadhyay P. **The novel, orally available and peripherally restricted selective cannabinoid CB2 receptor agonist LEI-101 prevents cisplatin-induced nephrotoxicity**. *Br. J. Pharmacol.* (2016.0) **173** 446-458. DOI: 10.1111/bph.13338
30. Soethoudt M. **Structure-kinetic relationship studies of cannabinoid CB2 receptor agonists reveal substituent-specific lipophilic effects on residence time**. *Biochem. Pharmacol.* (2018.0) **152** 129-142. DOI: 10.1016/j.bcp.2018.03.018
31. Hua T. **Crystal structure of the human cannabinoid receptor CB1**. *Cell* (2016.0) **167** 750-762.e714. DOI: 10.1016/j.cell.2016.10.004
32. Shao Z. **High-resolution crystal structure of the human CB1 cannabinoid receptor**. *Nature* (2016.0) **540** 602-606. DOI: 10.1038/nature20613
33. Li X. **Crystal structure of the human cannabinoid receptor CB2**. *Cell* (2019.0) **176** 459-467 e413. DOI: 10.1016/j.cell.2018.12.011
34. Hua T. **Activation and signaling mechanism revealed by cannabinoid receptor-Gi complex structures**. *Cell* (2020.0) **180** 655-665.e18. DOI: 10.1016/j.cell.2020.01.008
35. Xing C. **Cryo-EM structure of the human cannabinoid receptor CB2-Gi signaling complex**. *Cell* (2020.0) **180** 645-654.e13. DOI: 10.1016/j.cell.2020.01.007
36. 36.Hua, T. et al. Crystal structures of agonist-bound human cannabinoid receptor CB1. Nature10.1038/nature23272 (2017).
37. Shao Z. **Structure of an allosteric modulator bound to the CB1 cannabinoid receptor**. *Nat. Chem. Biol.* (2019.0) **15** 1199-1205. DOI: 10.1038/s41589-019-0387-2
38. Yang X. **Molecular mechanism of allosteric modulation for the cannabinoid receptor CB1**. *Nat. Chem. Biol.* (2022.0) **18** 831-840. DOI: 10.1038/s41589-022-01038-y
39. McAllister SD. **Structural mimicry in class A G protein-coupled receptor rotamer toggle switches: the importance of the F3.36(201)/W6.48(357) interaction in cannabinoid CB1 receptor activation**. *J. Biol. Chem.* (2004.0) **279** 48024-48037. DOI: 10.1074/jbc.M406648200
40. Diaz O, Dalton JAR, Giraldo J. **Revealing the mechanism of agonist-mediated cannabinoid receptor 1 (CB1) activation and phospholipid-mediated allosteric modulation**. *J. Med. Chem.* (2019.0) **62** 5638-5654. DOI: 10.1021/acs.jmedchem.9b00612
41. Krishna Kumar K. **Structure of a signaling cannabinoid receptor 1-G protein complex**. *Cell* (2019.0) **176** 448-458 e412. DOI: 10.1016/j.cell.2018.11.040
42. Bow EW, Rimoldi JM. **The structure-function relationships of classical cannabinoids: CB1/CB2 modulation**. *Perspect. Med. Chem.* (2016.0) **8** 17-39
43. Kapur A. **Mutation studies of Ser7.39 and Ser2.60 in the human CB1 cannabinoid receptor: evidence for a serine-induced bend in CB1 transmembrane helix 7**. *Mol. Pharmacol.* (2007.0) **71** 1512-1524. DOI: 10.1124/mol.107.034645
44. Javitch JA, Ballesteros JA, Weinstein H, Chen J. **A cluster of aromatic residues in the sixth membrane-spanning segment of the dopamine D2 receptor is accessible in the binding-site crevice**. *Biochemistry* (1998.0) **37** 998-1006. DOI: 10.1021/bi972241y
45. Xu W. **Comparison of the amino acid residues in the sixth transmembrane domains accessible in the binding-site crevices of mu, delta, and kappa opioid receptors**. *Biochemistry* (2001.0) **40** 8018-8029. DOI: 10.1021/bi002490d
46. Wang X. **Identification of V6.51L as a selectivity hotspot in stereoselective A2B adenosine receptor antagonist recognition**. *Sci. Rep.* (2021.0) **11** 14171. DOI: 10.1038/s41598-021-93419-x
47. Lin X. **Structural basis of ligand recognition and self-activation of orphan GPR52**. *Nature* (2020.0) **579** 152-157. DOI: 10.1038/s41586-020-2019-0
48. Hanson MA. **Crystal structure of a lipid G protein-coupled receptor**. *Science* (2012.0) **335** 851-855. DOI: 10.1126/science.1215904
49. Jakowiecki J, Filipek S. **Hydrophobic ligand entry and exit pathways of the CB1 cannabinoid receptor**. *J. Chem. Inf. Model* (2016.0) **56** 2457-2466. DOI: 10.1021/acs.jcim.6b00499
50. Stanley N, Pardo L, Fabritiis GD. **The pathway of ligand entry from the membrane bilayer to a lipid G protein-coupled receptor**. *Sci. Rep.* (2016.0) **6** 22639. DOI: 10.1038/srep22639
51. Szlenk CT, Gc JB, Natesan S. **Does the lipid bilayer orchestrate access and binding of ligands to transmembrane orthosteric/allosteric sites of G protein-coupled receptors**. *Mol. Pharmacol.* (2019.0) **96** 527-541. DOI: 10.1124/mol.118.115113
52. Bokoch MP. **Entry from the lipid bilayer: a possible pathway for inhibition of a peptide G protein-coupled receptor by a lipophilic small molecule**. *Biochemistry* (2018.0) **57** 5748-5758. DOI: 10.1021/acs.biochem.8b00577
53. Sykes DA. **Observed drug-receptor association rates are governed by membrane affinity: the importance of establishing “micro-pharmacokinetic/pharmacodynamic relationships” at the beta2-adrenoceptor**. *Mol. Pharmacol.* (2014.0) **85** 608-617. DOI: 10.1124/mol.113.090209
54. van der Velden WJC, Heitman LH, Rosenkilde MM. **Perspective: implications of ligand-receptor binding kinetics for therapeutic targeting of G protein-coupled receptors**. *ACS Pharmacol. Transl. Sci.* (2020.0) **3** 179-189. DOI: 10.1021/acsptsci.0c00012
55. Swanson ML, Regner KR, Moore BM, Park F. **Cannabinoid type 2 receptor activation reduces the progression of kidney fibrosis using a Mouse Model of unilateral ureteral obstruction**. *Cannabis Cannabinoid Res.* (2022.0) **7** 790-803. DOI: 10.1089/can.2021.0127
56. Chafik SG, Michel HE, El-Demerdash E. **The Cannabinoid-2 receptor agonist, 1-phenylisatin, protects against cisplatin-induced nephrotoxicity in mice**. *Life Sci.* (2022.0) **308** 120928. DOI: 10.1016/j.lfs.2022.120928
57. Trojnar E. **Cannabinoid-2 receptor activation ameliorates hepatorenal syndrome**. *Free Radic. Biol. Med.* (2020.0) **152** 540-550. DOI: 10.1016/j.freeradbiomed.2019.11.027
58. Cakir M, Tekin S, Doganyigit Z, Cakan P, Kaymak E. **The protective effect of cannabinoid type 2 receptor activation on renal ischemia-reperfusion injury**. *Mol. Cell Biochem.* (2019.0) **462** 123-132. DOI: 10.1007/s11010-019-03616-6
59. Pressly JD. **Selective cannabinoid 2 receptor stimulation reduces tubular epithelial cell damage after renal ischemia-reperfusion injury**. *J. Pharmacol. Exp. Ther.* (2018.0) **364** 287-299. DOI: 10.1124/jpet.117.245522
60. Nettekoven M. **Novel triazolopyrimidine-derived cannabinoid receptor 2 agonists as potential treatment for inflammatory kidney diseases**. *ChemMedChem* (2016.0) **11** 179-189. DOI: 10.1002/cmdc.201500218
61. Jenkin KA. **Renal effects of chronic pharmacological manipulation of CB2 receptors in rats with diet-induced obesity**. *Br. J. Pharmacol.* (2016.0) **173** 1128-1142. DOI: 10.1111/bph.13056
62. Barutta F. **Deficiency of cannabinoid receptor of type 2 worsens renal functional and structural abnormalities in streptozotocin-induced diabetic mice**. *Kidney Int.* (2014.0) **86** 979-990. DOI: 10.1038/ki.2014.165
63. Mukhopadhyay P. **Cannabinoid-2 receptor limits inflammation, oxidative/nitrosative stress, and cell death in nephropathy**. *Free Radic. Biol. Med.* (2010.0) **48** 457-467. DOI: 10.1016/j.freeradbiomed.2009.11.022
64. Mastronarde DN. **Automated electron microscope tomography using robust prediction of specimen movements**. *J. Struct. Biol.* (2005.0) **152** 36-51. DOI: 10.1016/j.jsb.2005.07.007
65. Punjani A, Rubinstein JL, Fleet DJ, Brubaker MA. **cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination**. *Nat. Methods* (2017.0) **14** 290-296. DOI: 10.1038/nmeth.4169
66. Heymann JB. **Single particle reconstruction and validation using Bsoft for the map challenge**. *J. Struct. Biol.* (2018.0) **204** 90-95. DOI: 10.1016/j.jsb.2018.07.003
67. Pettersen EF. **UCSF Chimera–a visualization system for exploratory research and analysis**. *J. Comput. Chem.* (2004.0) **25** 1605-1612. DOI: 10.1002/jcc.20084
68. Emsley P, Lohkamp B, Scott WG, Cowtan K. **Features and development of Coot**. *Acta Crystallogr. D Biol. Crystallogr.* (2010.0) **66** 486-501. DOI: 10.1107/S0907444910007493
69. Adams PD. **PHENIX: a comprehensive Python-based system for macromolecular structure solution**. *Acta Crystallogr. D Biol. Crystallogr.* (2010.0) **66** 213-221. DOI: 10.1107/S0907444909052925
70. Chen VB. **MolProbity: all-atom structure validation for macromolecular crystallography**. *Acta Crystallogr. D Biol. Crystallogr.* (2010.0) **66** 12-21. DOI: 10.1107/S0907444909042073
71. Chen C, Okayama H. **High-efficiency transformation of mammalian cells by plasmid DNA**. *Mol Cell Biol.* (1987.0) **7** 2745-2752. PMID: 3670292
72. Smith PK. **Measurement of protein using bicinchoninic acid**. *Anal. Biochem.* (1985.0) **150** 76-85. DOI: 10.1016/0003-2697(85)90442-7
73. Martella A. **A novel selective inverse agonist of the CB2 receptor as a radiolabeled tool compound for kinetic binding studies**. *Mol. Pharmacol.* (2017.0) **92** 389-400. DOI: 10.1124/mol.117.108605
74. Batkai S. **Cannabinoid-2 receptor mediates protection against hepatic ischemia/reperfusion injury**. *FASEB J.* (2007.0) **21** 1788-1800. DOI: 10.1096/fj.06-7451com
75. Copeland RA, Pompliano DL, Meek TD. **Drug-target residence time and its implications for lead optimization**. *Nat. Rev. Drug Discov.* (2006.0) **5** 730-739. DOI: 10.1038/nrd2082
76. Cheng Y, Prusoff WH. **Relationship between the inhibition constant (K1) and the concentration of inhibitor which causes 50 per cent inhibition (I50) of an enzymatic reaction**. *Biochem. Pharmacol.* (1973.0) **22** 3099-3108. DOI: 10.1016/0006-2952(73)90196-2
|
---
title: On the determinants and the role of the payers in the uptake of genetic testing
and data sharing in personalized health
authors:
- Veronika Kalouguina
- Joël Wagner
journal: Frontiers in Public Health
year: 2023
pmcid: PMC10017738
doi: 10.3389/fpubh.2023.920286
license: CC BY 4.0
---
# On the determinants and the role of the payers in the uptake of genetic testing and data sharing in personalized health
## Abstract
### Background
New health technologies and data offer tailored prevention and spot-on treatments, which can considerably reduce healthcare costs. In healthy individuals, insurers can participate in the creation of health capital through data and preventing the occurrence of a disease. In the onset of a disease, sequencing an individual's genome can provide information leading to the use of more efficient treatments. Both improvements are at the core of the “personalized health” paradigm. As a positive side effect, a reduction in healthcare costs is expected. However, the integration of personalized health in insurance schemes starts with a closer understanding of the demand drivers.
### Methods
Using novel data from a survey carried out in Switzerland, we determine the factors influencing the uptake and sharing of data from genetic tests. In our regression analyses, we use five sets of socioeconomic, lifestyle, health insurance, sentiment, and political beliefs variables. Furthermore, two framings assess the willingness to undertake a test and the readiness to share results with an insurer when the costs of the test are borne by the insurer or the individual.
### Results
We find that socioeconomic, lifestyle, or political belief variables have very little influence on the uptake of tests and the sharing of data. On the contrary, our results indicate that sentiment and insurance factors play a strong role. More precisely, if genetic tests are perceived as a mean to perform health prevention, this pushes individuals to take them. Furthermore, using the insurer's smartphone app leads to an increase of the likelihood to undergo a test and doubles the probability to share related data. Regarding insurance plans and deductible levels, there is no strong correlation neither with the willingness to take a test nor to share the data. Finally, individuals with complementary health insurance plans are less likely to share results. From the framings for the payment of genetic tests, our results indicate a positive effect of the insurer as a payer on the willingness to undertake tests as well as on data sharing.
### Conclusion
Our results lay the ground for a deeper understanding of the role of payers on health decisions and sharing of health-related data. In particular, we find that it is relevant for health insurers to engage with their clients.
## 1. Introduction
Genetic tests (GTs) have several purposes: in the case of a healthy individual, sequencing parts of the genome helps to evaluate the risk of developing a certain disease as well as to pass it to the next generation [1]. Newborn screening can reveal disorders that need early medication. Diagnose testing, which happens in the case of a sick individual, allows the medical team to understand the genetic root of the condition and to select the treatment minimizing adverse drug events [2, 3]. Finally, direct-to-consumer GTs allow individuals to obtain a genetic screening without healthcare intermediaries. The tests can provide genetic-based food intolerances, exercise plans, and, in certain cases, a risk profile for specific diseases such as breast cancer [see, e.g., Su [4]].1 GTs rely on very strong power of data. *Translating* genetic information gives the individual knowledge about the own risk level of a disease and hence the leverage to act on it. For example, by changing the lifestyle and health behaviors [9], one can reduce the probability of the occurrence of a disease. Furthermore, the results of the GT enable enlightened decisions to schedule a personalized check-up plan for the individual and to monitor those specific risks [10]. Finally, researchers as actors of the health ecosystem can run analyses with the anonymized data to understand which types of prevention work best for which predisposition to a disease. From a social sciences perspective, among the first steps to unlock the benefits of GTs, is to understand what drives individuals to take them. To grasp the general public attitude toward GTs and its willingness to undergo a GT (“genetic testing willingness,” GTW) provides policymakers and stakeholders (e.g., insurers) with insights useful to promote their uptake. However, so far, several authors in the literature solely focus on a particular condition to assess GTW. For instance, cancer susceptibility risk assessment is a recurrent subject under study [see, e.g., Fogel et al. [ 11]]. Often, the surveyed population and the criteria for an admittance in a particular study usually include family history and being at risk for a given disease [12].
In this article, to fill the gap, we provide a general study of the GTW and the willingness to share the related data (“data sharing willingness,” DSW), using novel data from an ad hoc survey carried out in Switzerland. We determine the factors influencing the uptake and sharing of data from GTs. Through regression analyses followed by a random forest robustness check, we use five sets of variables, socioeconomic, lifestyle, health insurance, sentiment, and political beliefs, and two framings. The two framings assess the GTW and the DSW of anonymized test results with the health insurer when the costs of the tests are either borne by the insurer or by the individual. Moreover, our survey design adds the effect of the payer dimension to our analysis through the framings. Including the health insurer as an actor has seldom been done and brings new results to this pane of the literature.
Our article hence focuses on two research questions: We find that socioeconomic, lifestyle, or political belief variables have very little or no influence on the uptake of GTs and the sharing of the results with an insurer, which is in line with the literature [13]. On the contrary, our results indicate that insurance and sentiment factors play a strong role. More precisely, if GTs are perceived as a mean to perform health prevention, this pushes individuals to take them by an increase in propensity of 10.9 pp. Furthermore, using the health insurer's smartphone app leads to an increase of $16.5\%$ in the GTW and of $27.6\%$ to anonymously share the related data with the insurer. Regarding insurance plans and deductible levels, there is no strong correlation neither with the GTW nor with the DSW. Finally, individuals with complementary health insurance plans are less likely to share anonymized test results with their insurer. Using framings for the payment of GTs, we seize the effect of the insurer as a payer on both GTW and DSW. Our results indicate a positive effect of the insurer as a payer on the GTW (+$24.8\%$) as well as the DSW with the health insurer (+$9.4\%$).
The remaining of the paper is organized as follows: Section 2 offers a literature review along the research questions and the methodology, as well as a description of the variables with descriptive statistics. Section 3 present both regression and random forest results. Finally, in Section 4, we conclude and provide a discussion for further research.
## 2.1. Literature review
The state of the existent literature is described by Sweeny et al. [ 13] as being “[...] rife with conflicting findings, inconsistent methodology, and uneven attention across test types and across predictors of genetic testing decisions.” One can find several clusters of studies in the academic research. They differ either by the nature of the GTs submitted for questioning or by the population under study. First, extant research mostly focuses on the GTW or the willingness to pay for a particular GT related to a certain disease, such as breast cancer [e.g., Armstrong et al. [ 14]], Alzheimer's disease [e.g., Kopits et al. [ 15]], or colon cancer [e.g., Lerman et al. [ 16]]. Few articles query on the willingness to do or to pay for GTs in general. Second, in many studies, the subject population is targeted and not randomly selected. The selection of the sample is usually based on criteria such as being at-risk for a certain condition. For instance, in the case of Dalpe et al. [ 12], women between 35 and 55 years were inquired about their interest to undergo GTs in search of a mutation which may lead to breast cancer. Following this restricted selection, the sample size usually ends up in < 1,000 individuals. Finally, a lot of research is conducted under a social sciences perspective rather than a economical view point; hence, we found very seldom health insurance as being an examined factor for the GTW. When the health insurer was mentioned, it was mostly presented in the perceived barriers section as a possible discriminator following a GT. As an instance, the fear of denial for coverage is discussed in multiple articles [e.g., Hall et al. [ 17], Allain et al. [ 18], Haga et al. [ 19], and Clayton et al. [ 20]].
Regarding drivers of the decision to take a GT, socioeconomic factors are often assessed. They include age, gender, education, employment status, marital status, and income. Throughout our literature review, we did not find consistent results for any of these factors. For instance, in Armstrong et al. [ 14] and Miron-Shatz et al. [ 21], older women are more likely to undergo genetic testing for breast cancer than younger ones. In Tubeuf et al. [ 22] or Wessel et al. [ 23], however, age does not play a role in the interest for genetic testing for retinal disease or diabetes type 2, respectively. These conflicting findings are backed up by Sweeny et al. [ 13], in their literature review. The authors find likewise that age has an unclear outcome on decision-making for genetic testing. They also have assessed the effects of the aforementioned socioeconomic factors and the results are the same to what is observed more recently by Wessel et al. [ 23]. Regarding socioeconomic factors, the results found in the literature do not reach a consensus either. Predictors such as gender, education, income, or marital status present different effects on the decisions-taking. Throughout the articles, results are ranging from a positive to negative effect with most studies not giving conclusive results.
Another interesting factor is the family health history, i.e., the existence or not in the close family of an individual who is suffering or suffered from a given health condition. Expectedly, in a majority of articles, the existence of a family member bearing a particular condition leads to an increase in the likelihood of the individual to undergo genetic testing. Blouin-Bougie et al. [ 24], Abdul Rahim et al. [ 25], and Sun et al. [ 26], to cite a few, document such results. Interestingly, a research on a sample of 1,960 British individuals by Sanderson et al. [ 27] presented opposing findings for the GTW for heart disease or cancer predisposition. In their results, individuals with a family history of heart disease are more likely to do a GT for heart disease, whereas individuals with cancer running in their family are less likely to undergo a GT for cancer. Again, in their systematic review of the literature, Sweeny et al. [ 13] confirmed that family health history displays either a positive relationship with GTs or no statistical relevance.
Despite the heterogeneity in socioeconomic factors, the literature nevertheless presents several consistent drivers displaying a clear effect on the willing to do or to pay. These drivers are psychological and they reflect the individual's view of the gains or losses a GT may result in. They are usually part of the health belief model [28], more precisely the perceived benefits and barriers of the tests, health motivations, and perceived susceptibility or severity. The most extensive literature is found on the effect of perceived benefits of genetic testing. These benefits can take several forms like the knowledge about the risks of getting a particular condition [e.g., Gollust et al. [ 29], Wessel et al. [ 23], Fogel et al. [ 11], Kauffman et al. [ 30], and Abdul Rahim et al. [ 25]], have adequate prevention [e.g., Lerman et al. [ 16] and Alanazy et al. [ 31]], or inform relatives of a possible risk [e.g., Smith and Croyle [32], Armstrong et al. [ 14], Hall et al. [ 17], Fogel et al. [ 11], and Sun et al. [ 26]]. These benefits are incentives for individuals to undergo testing and hence have a positive impact on the GTW. This is consistent and statistically significant throughout the literature [13]. The perceived barriers also play a role in the GT uptake decision. The most common fears are the financial consequences of the testing [e.g., Bosompra et al. [ 33], Alanazy et al. [ 31], and Sun et al. [ 26]] and the possible discrimination by employers and insurers [e.g., Lerman et al. [ 16], Armstrong et al. [ 14], Cameron et al. [ 34], and Dalpe et al. [ 12]].
In regard of the literature gaps and research avenues presented earlier, the aim of our research is two-fold. Our study builds on an original survey to address several gaps in the literature. By randomly selecting a representative sample of participants, we ensure the understanding of the GTW and of the DSW in a broad, lay population. In addition, the size of the sample gives us the opportunity to add a dimension using the payer of the GT as a framing.
## 2.2. Survey setup
To conduct our study, we created an original survey for which the collection of data was supported by a polling agency. The sample comprises 1,000 respondents from Switzerland evenly distributed by gender, by four age categories between 25 and 65 years, and by language regions with two-thirds from the German-speaking part and one-third from the French-speaking part.
After briefly explaining the purpose of GTs in the context of personalized health, we inquire individuals whether they are using or would be willing to use such type of technology. Subsequently, we focus on GTs and question individuals about factors which could incentivize or refrain them from performing such a test. We take advantage of this focused section to also analyze the effect of the price and the payer on individual's enthusiasm to do the GT through framing with different scenarios. Finally, we ask socioeconomic, sentiment, and political questions.
## 2.2.1. Response variables: Would you carry out such a genetic test?
The core of our questionnaire starts with an introductory paragraph providing the basic knowledge for the surveyed individuals and to set boundaries for a common understanding of genetic testing in the present research.
In Figure 1, one can observe that we first question the GTW without price information (questions A and B). Subsequently, the whole sample is then divided into two subsamples of randomly selected 500 individuals. The framing targets the payer of the GT. In Framing 1, the payer is the health insurer (question C1), whereas in Framing 2, it is the individual (question C2). Once the question about GTW following the framing is asked, both subsamples are inquired about the DSW of anonymized data with the health insurer (question D). We report relevant excerpts of the questionnaire in the Supplementary material. The questions A, B, C1, C2, and D correspond to the questions C3, C4, C5c, C5d, and C6, respectively.
**Figure 1:** *Survey setup, core questions, and framings.*
## 2.2.2. Explanatory variables
The first questions of our survey selected the participants. These questions inquired about age, gender, and postal code to select the respondents, and balanced the panel according to the criteria. The majority of the other questions leading to our explanatory variables were asked after the core questions. The first set of questions relates to socioeconomic factors and is composed of 10 variables. The second and third sets are insurance and lifestyle factors, containing four and six variables, respectively. The fourth set is made of political factors with three variables. The last set is the largest, assembling 24 variables regarding sentiment factors. Several variables present binary categories. Indeed, for some of them, the original categories were merged to create a binary outcome as to decrease the length of the model and avoid a potential overfit. Table 1 provides the list of all the used variables and a brief description of the variable itself, accompanied by the available categories, along the five sets of variables.
**Table 1**
| Variable | Description | Categories |
| --- | --- | --- |
| Socioeconomic factors | Socioeconomic factors | Socioeconomic factors |
| Gender | Gender of the respondent | Male, female |
| Age | Age class in years | 25–34, 35–44, 45–54, 55–65 |
| Region of residency | Canton defined by the spoken language | French-speaking, German-speaking |
| Nationality | Nationality of the respondent | Other, Swiss |
| Education | Higher education (above high school level) | No, yes |
| Professional situation | Current employment situation | Full-time employed, part-time employed, other |
| Subjective wealth | Subjective household wealth | Below average, above average |
| Marital status | Marital status | Married/registered partnership, other |
| Health | Self-rated health | Bad, average, good |
| Cancer history | History of cancer, cardiac or hereditary disease in close family | No, yes |
| Lifestyle factors | Lifestyle factors | Lifestyle factors |
| Alcohol consumption | Alcohol consumption | Everyday, sometimes, never |
| Cigarette consumption | Smoking habit | Everyday, sometimes, never |
| Greens consumption | Fruits and vegetables consumption | Everyday, sometimes, never |
| Sport | Exercising habit | At least once a week, less |
| Future planning | Interest of planing for the future | 0 to 1 by increments of 0.1 |
| Risk-loving | Readiness to take risks | 0 to 1 by increments of 0.1 |
| Insurance factors | Insurance factors | Insurance factors |
| Insurance plan | Mandatory health insurance plan | Basic, Health Maintenance Organization, family doctor, CallMed |
| Deductible | Mandatory health insurance level of deductible | CHF 300, 500–2,000, 2,500 |
| Complementary insurance | Complementary health insurance | No, yes |
| Insurer's app | Insurer's app for step or exercise count | No, yes |
| Political factors | Political factors | Political factors |
| Interest in politics | Interest in politics | No, yes |
| Political orientation | Political orientation assigned on the left | 0 to 1 by increments of 0.1 |
| Feeling close to a political party | Feeling close to a political party | No, yes |
| Sentiment factors | Sentiment factors | Sentiment factors |
| Incentive: curiosity | Curiosity is an incentive to undergo genetic testing | No, yes |
| Incentive: better health prevention | Take better care of health is an incentive to undergo genetic testing | No, yes |
| Incentive: help relatives prevention | Help relatives to take better care of their health is an incentive to undergo genetic testing | No, yes |
| Incentive: incentivize relatives | Incentivize relatives is an incentive to undergo genetic testing | No, yes |
| Incentive: disease risk information | Disease risk information is an incentive to undergo genetic testing | No, yes |
| Barrier: fear of discrimination | Fear of discrimination is a barrier to undergo genetic testing | No, yes |
| Barrier: test too costly | Fear of cost of test is a barrier to undergo genetic testing | No, yes |
| Barrier: family disapproves | Fear of family disapproving is a barrier to undergo genetic testing | No, yes |
| Barrier: induced lifestyle changes | Induced lifestyle changes is a barrier to undergo genetic testing | No, yes |
| Barrier: not want info | Not wanting to know the risks is a barrier to undergo genetic testing | No, yes |
| Barrier: family finances | Impact on family finances is a barrier to undergo genetic testing | No, yes |
| Impact: more difficult family insured | It will be more difficult for my family to get insured | No, yes |
| Impact: longer and better life | Genetic testing will promote a healthier and longer life | No, yes |
| Impact: testing will be common | Genetic testing will be common | No, yes |
| Impact: testing mandatory to be hired | Genetic testing will be necessary to get hired | No, yes |
| Impact: testing for insurance premiums | Sequencing asked prior premium establishment | No, yes |
| Impact: genetic passport | Everyone will have a genetic passport | No, yes |
| Impact: segregation good/bad | There will be a segregation between “good" and “bad" genomes | No, yes |
| Impact: discrimination of disabled | Disabled individuals will be discriminated | No, yes |
| Impact: government not able to protect | Government will not be able to protect individuals | No, yes |
| Impact: genetic testing for infants | all infants will have their genome sequenced | No, yes |
| Impact: genetic testing for fœtuses | All fœtuses will undergo genetic testing | No, yes |
| Usage of health-related apps | Usage of health-related apps | No, yes |
| Usage of health-related apps for prevention | Usage of health-related apps for prevention | No, yes |
## 2.2.3. Socioeconomic factors
This set starts with a question asking the survey respondent to indicate the gender with two choices of response, male or female. For the age, we collected integers which were gathered in four classes according to our selection criteria, each class containing $25\%$ of the sample. The classes are 25–34, 35–44, 45–54, and 55–65 years. The last selective variable is the region of residency, which can be German- or French-speaking depending on the postal code indicated by the individual. We also collect information about the respondent's nationality (Swiss/other) and education (below or above high school level). Another question concerned the professional situation, to which the responses were merged into “full-time employed,” “part-time employed,” or “other” categories. We subsequently asked about the subjective wealth of the individual which could be answered by below or above average and the marital status, which can be either “married/in a registered partnership” or “other”. Finally, the last two questions of this set dealt with health. In one question, the respondents had to rate their health from “very bad,” “bad,” “fairly good,” “good,” and “very good,” which response we classified into “bad” for the two worst levels, “average” for the middle level, and “good” for the two best levels chosen. The last question is whether the participant has a history of cancer, cardiac, or hereditary disease in the immediate family.
## 2.2.4. Insurance factors
This set relates to the health insurance subscribed by the individual. In Switzerland, the mandatory health insurance policy has two features: the plan and the annual deductible. Hence, the first question inquires about the insurance plan, which can be of several nature: basic, Health Maintenance Organization, family doctor, or CallMed. The second question regards the deductible which can be CHF 300, CHF 500, CHF 1,000, CHF 1,500, CHF 2,000, or CHF 2,500. Usually, it is the two extremes that are favored, hence we merged the levels in the middle (CHF 500–2,000) to obtain a three-level scale. Alongside the mandatory health insurance, the individual can take out an optional complementary insurance; we, therefore, ask if he/she holds such a policy. Finally, we have a variable (insurer's app) indicating whether the person has an app from his/her health insurance for recording activity or counting steps.
## 2.2.5. Lifestyle factors
Three questions start by inquiring the individual about his/her habits. These questions concern alcohol, cigarettes, and greens (vegetables and fruits) consumption, to which the possible responses were daily, several times a week, once a week, once every 2 weeks, once a month, less regularly, or never. We subsequently merged the responses to obtain a three-level categorical variable with “everyday,” “sometimes,” and “never” as outcomes. Physical exercise (sport) was also taken into account by a question asking the frequency at which the individual exercises. The possible answers being several times a week, once a week, less regularly, and never were pooled together to create a binary variable: at least once a week or less. To conclude this set, we dig deeper into the person's behavior by asking for his/her interest in planning for the future, together with readiness to take risks. The answers were based on an 11-point Likert scale ranging from “not interested at all” to “very interested” (future planning risk-loving).
## 2.2.6. Political beliefs factors
Our fourth and shortest set includes three questions about political beliefs. In the first question, individuals had to express their interest in politics from the possible “not at all interested,” “slightly interested,” “fairly interested,” or “very interested” answers. The second question asked the individual to rate his/her political orientation on an 11-point Likert scale going from “left” to “right.” For the last question, we presented several political parties (with the “another several parties,” “I do not want to disclose,” and “I do not relate to any” options) and asked the person to select which party they feel the closest to. We then extracted a binary outcome indicating if the participant felt close to a political party or not.
## 2.2.7. Sentiment factors
This last set is the largest one with 22 variables stemming from three questions and two additional health-related apps questions. In the three first questions, several statements are given to which the respondent had to chose a level of agreement on a 5-point Likert scale ranging from “strongly disagree” to “completely agree.” Subsequently, we code the answers as “is an incentive” for individuals ticking the “completely agree” and the following level and “not an incentive” for the other responses.
The first question suggests incentives to undergo genetic testing: I am curious about my genetic makeup; my results could help me take better care of my health; my results could help my relatives to take better care of their health; it could incentivize my relatives to undergo genetic testing for themselves and my results could provide useful information about my hereditary diseases or my risk cancer.
Similarly, the second question cites potential barriers to genetic testing. These hurdles being: I fear a possible discrimination; I fear the test would be too expensive; some members of my family could disapprove me taking a test; knowing my cancer risk may force me to lead a different lifestyle; I don't want to know what potential illness I might have in future; and I think my results could have a strong impact on my family's finances.
Ultimately, to capture the outlook of the individual on GT and his/her beliefs regarding GT developments, we have a series of 11 questions. The following sentences were displayed to which the respondent had to chose a level of agreement. It will be more difficult for my family members to get an insurance policy. Knowledge related to genetics will lead to fewer illnesses and longer life expectancy. It will be very common to perform GTs. Future employees will have to undergo genetic testing before being hired. Insurance companies will request a sequencing of our genome to establish premium levels. In the future, we will all have a genetic passport. There will be a segregation in our society between “good” and “bad” genomes. People with disabilities will be less accepted in society. The government will not be able to protect citizens from the negative aspects of GTs. The genome of all infants will be sequenced to establish their genetic profile and prevent development of certain diseases. Finally, all pregnant women will undergo genetic testing to determine if the fetus carries a disease.
## 2.3.1. Response variables: would you carry out such a genetic test?
In this section, we perform a statistical analysis on the responses derived from the core questions presented in Section 2.2.1. In Figure 2, we display the mean values and confidence intervals for the answers to each question. The figure is divided in to three sections. The left section represents the means of the whole sample of 1,000 individuals, the middle section represents the means of the subsample presented with the first framing, insurer as a payer, and the right section, the subsample from the second framing, individual self-payer. On the left of each section of the figure, the black dot illustrates the mean level of agreement in question A in Figure 1 for the whole sample, for the insurer payer framing in the middle, and for the self payer framing on the right. The same logic applies to the red and yellow dots, which represent the means for questions B and D. In the second section, the green dot concerns the answers for those who had the insurer framing and the blue dot represents the answers for the self-payer framing. In addition, the red line represents a $99\%$ confidence interval and the black line a $95\%$. The numbers corresponding to the $95\%$ confidence interval can be found in Table 2.
**Figure 2:** *Average level of agreement with 95 and $99\%$ confidence intervals. A1 and A2 correspond to the level of agreement for question A in each subsample of $$n = 500$$, respectively. B1 and B2 correspond to the level of agreement for question B in each subsample of $$n = 500$.$ C1 and C2 correspond to the level of agreement for question C in each subsample of $$n = 500$.$* TABLE_PLACEHOLDER:Table 2 As one can first see, for the result of the baseline GTW, half of individuals ($50.9\%$) agreed that they would be willing to undergo a GT. The distribution of this answer is not statistically different in the framed groups. Comparing with similar studies, in a randomly selected sample of 383 individuals by Smith and Croyle [32], $47.3\%$ of the interviewees stated that they are very interested in taking a GT for colon cancer and $16.1\%$ stated that they are not interested. More recently, in a study conducted in Saudi Arabia, authors assessed willingness to undergo presymptomatic genetic testing for Alzheimer's disease and obtained a level of agreement of $59.9\%$ for either one of the two presented GTs and $45.1\%$ for both tests [31]. Our results hence corroborate findings for similar surveys in the literature.
Once price information is displayed, we observe the number of agreeing respondents drop from 509 individuals to 381. We note for this question that, aside from the price range of CHF 100 to CHF 400, no payer was specified. This drop can be explained by the fact that the price may be higher than expected or renders the test more tangible as knowing the price brings the individual closer to the concept of buying the product. Another explanation could be the price itself, which can be a burden for some individuals. It will be interesting to test this hypothesis in the regressions with the income variable. In addition, the two samples used in the framings do not have statistically different means with $99\%$ confidence.
Subsequently, when the framings are applied, a clear cut appears. For the group framed with the insurer as a payer, the share of surveyed individuals agreeing to undergo the testing increases to attain $60.0\%$ whereas those framed to be the sole payer of the test decreases as low as $35.2\%$. These means are statistically significantly different at $95\%$ as their confidence intervals do not overlap. One can hypothesize that the insurer as a payer triggers more individuals to undergo the test because of the cost relief. We later test this hypothesis with regressions to understand the difference between potential drivers of the GTW.
Finally, interesting results can already be seen for the DSW of anonymized data from GTs with the health insurer. When merged together, the whole sample exhibits a DSW of $33.5\%$. However, in the subsamples, we observe a clear cleavage. Indeed, the two groups display statistically significantly different means at a confidence level of $95\%$, hinting that the role of the payer is essential in this regard. Regression analysis allows us to study this relationship and suggest possible correlation between the payer and the readiness of an individual to share health-related data with the insurer.
## 2.3.2. Explanatory variables
In a preliminary analysis, we have a look at the descriptive statistics for the socioeconomic, insurance, lifestyle, and political beliefs in Table 3 and for the sentiment factors in Table 4. For each variable, the sample column displays the frequency of the variable in the whole sample. In the four following columns, we display the level of agreement to the questions introduced in Section 2.2.1, i.e., A–genetic testing, B–genetic testing after price display, C1–genetic testing with insurer as a payer, C2–genetic testing with the individual as a payer, and D–data sharing with the health insurer, by variable.
Considering the descriptive statistics, we get a hint on possible correlations between the explanatory variables and the core questions. In the first set of variables, the socioeconomic factors, two variables stand out—age and nationality. Noticeable changes in the share of individuals who are willing to either undergo the test or share the data take place for the older group in the sample. For instance, the GTW in the group 55–65 years drops by as much as 18 percentage points (pp) when compared to the 35–44 in question A. This gap increases to 25 pp for the GTW when the insurer is the payer (C1). This difference is also true for the DSW with a disparity of 11 pp between the two groups. The older individuals in our sample seem to be reluctant to taking a GT as well as sharing the related anonymized data with their health insurer. We hence expect this effect to emerge in the regressions. The same conclusion can be drawn for the Swiss nationals in our sample. As a matter of fact, disregarding the price display or the payer of the test, they present a lower level of GTW and DSW, suggesting that Swiss are less open to these ideas. Moving on to insurance factors, some sparse but clear effects can be seen from having a complementary health insurance. The strongest positive effect for those who declared holding a complementary insurance policy intervene when cost comes into play, i.e., when the price is displayed or when the individual is the payer of the GT. On these GTW, the increase is by roughly 10 pp. In addition, the correlation between having an insurer's app for step or exercise count and GTW as well as DSW is quite strong and positive. Going to the next set, the lifestyle factors, only one variable has a clear and consistent pattern along its categories. Indeed, as the level of interest of planning for future increases, so does the share of individuals who present a positive GTW and DSW. As an example, for question A, the proportion of individuals who would be willing to get a GT rises from $29\%$ among individuals who indicated having the lowest level of interest in future planning to $71\%$ for those who have the highest.
For the last set on Table 3, the political factors, it is difficult to establish any hypothesis on the impact of these variables. The GTW proportions do not seem to follow a clear pattern and to display any correlation.
Table 4 contains the sentiment factors. The related variables come from three categories of questions: potential incentives for GTs, potential barriers to genetic testing, and impact of genetic testing on society. We first focus on the potential incentives to undergo genetic testing. According to our statistics, for each variable, there is a strong discrepancy between individuals who agreed with the statement and those who did not. As an example, individuals who agreed being curious about their genetic makeup is a good incentive for them to get GT are almost three times more likely to undergo genetic testing as well as share the related anonymized data with their health insurer than those who were not curious. This observation holds for all the variables in the set for a minimum difference of two-fold. Regarding the barriers, the divergences in the answers is less striking. Only not wanting to know the risks and the fear of possibly induced life changes are potential factors diminishing GTW. Finally, for variables indicating the general outlook of individuals on the GT in society, several factors show relevance. One can spot four variables: testing will be common, all fetuses as well as infants will undergo GTs, everybody will have a genetic passport, and knowledge based on genetics will increase life expectancy and promote better health. The individuals who agreed with these statements are more likely to undergo GTs. Interestingly, those who agreed that GTs will be mandatory to be hired are $63\%$ more likely to share anonymized data with their health insurer.
## 2.4. Methodology
We perform all regressions in Section 3 using the R software. Equation [1] describes the regression of each of the interest variables, A, B, C1, C2, and D, that we denote Wi. Each *Wi is* regressed on the five groups of factors, i.e., socioeconomic, lifestyle, insurance, political, and sentiment factors variables that form the set of variables X. For Wi, we merged the possible responses into a binary variable taking the value 1 if “likely” or “very likely” was selected, and 0 otherwise. Using Akaike's information criterion (AIC), we selected the logit link function for the regression as it displayed a lower AIC. The following Equation 1 is used for all sets of explanatory variables defined by the vector X Where j represents each group of explanatory factors in the set X and k, each variable within this group. The β0 and βXjk coefficients correspond to the baseline, respectively, and the regression coefficients are linked to the variables Xjk.
Furthermore, to facilitate interpretation and comparison between effects, for binary variables, we translated the βXjk coefficients into their probabilities (expressed in %) of obtaining 1 for Wi. The formula for the effect of a coefficient jk for a particular WILi is the following: After regressing the four Wi variables separately on the five groups of factors, we perform an overall regression combining all factors in a single regression model. Subsequently, we select the most relevant variables using a forward and backward variable selection with the stepAIC2 function in R [35]. This procedure allows to check for coefficients robustness and capture the most relevant explanatory variables. Finally, as an additional information, we use the randomForest package in R3 [36] to obtain an importance ranking of the effect of the variables on the GTW and DSW.
## 3.1. Results from regression analysis
In this subsection, we present regression results separately for each of the five sets of socioeconomic, insurance, lifestyle, political beliefs, and sentiment factors. For each of the four variables, we display the β coefficients, their equivalent p in terms of probabilities, and the significance. For categorical variables, the baseline is defined by the most frequent category in the sample. Following these regression results, we will present a confusion matrix and perform several robustness checks in Section 3.2.
## 3.1.1. Socioeconomic factors
From Table 5, we observe that only few factors are significant drivers for either GTW or DSW. As expected from the literature review and the statistical analysis, except for nationality and age, the gender, education, professional status, marital status, wealth, and region of residency do not explain responses from individuals. Two regression results however confirm findings from the statistical outlook. The age and Swiss nationality do influence the GTW. As noted by the 18 pp decrease in the individuals aged 55–65 years, their GTW is distinctively lower than for other categories. They display a reduction by $17\%$ in willingness compared to the baseline categories of 35–44 years. This decrease is however solely significant in the questions with the baseline willingness (A) and when the insurer is the payer (C1), where we observe a decrease (of −$18.5\%$). The same observation holds for the DSW. According to our results, respondents between the ages of 55 and 65 years are $11.9\%$ less likely to share their anonymized GTs result with the health insurer, everything else kept constant. The second variable with significant impact on questions A, B, C1, and C2 is nationality. Individuals with Swiss nationality seem less open to the idea of genetic testing, disregarding the price display or the payer, with strong significance. To conclude with this set of variables, cancer history and health present rather intriguing results. One would hypothesize that an individual who has a case of cancer in his/her close family is more enthusiastic regarding genetic testing but this hypothesis is only statistically verified for the baseline GTW, before any price is given. This inconclusive result can also be found in literature where authors either find a positive [25, 26] or a mitigated effect [27]. Similarly, another belief could be that the health of the respondent comes into the decision process to undergo a GT. Our results seem to annihilate such a relationship as the variable does not present significant coefficients. Nevertheless, it is interesting to notice that when the level of agreement for genetic testing drops from question A to question B when the price range is displayed, wealthier individuals do not seem to be less affected as wealth is not show significant.
**Table 5**
| Model | A– Baseline GTW | A– Baseline GTW.1 | A– Baseline GTW.2 | B–Price display | B–Price display.1 | B–Price display.2 | C1–Insurer payer | C1–Insurer payer.1 | C1–Insurer payer.2 | C2–Self payer | C2–Self payer.1 | C2–Self payer.2 | D–Data sharing | D–Data sharing.1 | D–Data sharing.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Model | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. |
| Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) |
| Female | 0.099 | +2.41% | | 0.130 | +3.22% | | 0.172 | +2.219% | | 0.000 | +0.00% | | −0.310 | −7.271% | * |
| Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) |
| 25 – 34 years | −0.113 | −2.81% | | −0.041 | −1.00% | | −0.164 | −2.44% | | 0.278 | +6.73% | | −0.047 | −1.13% | |
| 45 – 54 years | −0.337 | −8.39% | | −0.208 | −5.06% | | −0.790 | −13.95% | ** | −0.129 | −2.99% | | −0.047 | −1.124% | |
| 55 – 64 years | −0.696 | −17.19% | *** | −0.313 | −7.54% | | −0.998 | −18.52% | *** | −0.196 | −4.51% | | −0.522 | −11.87% | ** |
| Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) |
| Yes | −0.630 | −15.61% | *** | −0.635 | −14.66% | *** | −0.925 | −16.88% | *** | −0.589 | −12.63% | ** | −0.292 | −6.87% | |
| Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) |
| Yes | 0.249 | +6.00% | | 0.164 | +4.07% | | −0.007 | −0.13% | | −0.023 | −0.54% | | −0.208 | −4.94% | |
| Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) |
| Part-time | −0.245 | −6.08% | | −0.336 | −8.07% | | −0.609 | −10.24% | * | −0.219 | −5.02% | | −0.101 | −2.42% | |
| Other | 0.322 | +7.70% | | −0.110 | −2.69% | | 0.141 | +1.83% | | −0.206 | −4.74% | | 0.017 | +0.40% | |
| Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) |
| Above average | 0.155 | +3.78% | | 0.202 | +5.03% | | 0.052 | +0.69% | | 0.602 | +14.81% | ** | −0.229 | −5.42% | |
| Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) |
| Yes | 0.182 | +4.40% | | 0.139 | +3.46% | | −0.089 | −1.30% | | 0.233 | +5.63% | | 0.229 | +5.64% | |
| Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) |
| Yes | 0.382 | +9.05% | ** | 0.178 | +4.43% | | 0.002 | −0.00% | | 0.061 | +1.44% | | −0.045 | −1.08% | |
| Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) |
| Average | 0.069 | +1.70% | | 0.031 | +0.76% | | 0.662 | +7.28% | | −0.021 | −0.50% | | 0.265 | +6.54% | |
| Good | 0.120 | +2.93% | | 0.237 | +5.89% | | 0.346 | +4.24% | | 0.046 | +1.09% | | −0.005 | −0.12% | |
| Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) | Region (baseline: French) |
| German | −0.194 | −4.81% | | 0.008 | +0.19% | | −0.351 | −5.49% | | 0.006 | +0.14% | | 0.313 | +7.75% | * |
| Constant | 0.257 | | | −0.227 | | | 1.612 | | *** | −0.488 | | | −0.342 | | |
| N | 1000 | | | 1000 | | | 500 | | | 500 | | | 1000 | | |
## 3.1.2. Lifestyle factors
Among the lifestyle factors displayed in Table 6, only one variable displays a significant and consistent effect throughout all regressions: being keen on planning for the future. This variable is considered on a scale from 0 to 1, on which the individuals had to place their tendency of planning for the future. According to our results, the higher the level, the more likely is the respondent to undergo a GT, disregarding added information about the price or the payer. The same result is valid for propensity to share anonymized data. This correlation is coherent considering that in our survey, we deal with genetic testing for preventive purposes, hence for planning future medical examinations and potential diseases. Other health-related covariates do not affect individuals' decision-making, suggesting that this decision does not necessarily stem from health considerations, as already outlined by the absent correlation with the health variable in Table 5.
**Table 6**
| Model | A– Baseline GTW | A– Baseline GTW.1 | A– Baseline GTW.2 | B–Price display | B–Price display.1 | B–Price display.2 | C1–Insurer payer | C1–Insurer payer.1 | C1–Insurer payer.2 | C2–Self payer | C2–Self payer.1 | C2–Self payer.2 | D–Data sharing | D–Data sharing.1 | D–Data sharing.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Model | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. |
| Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) |
| Sometimes | −0.185 | −0.67% | | 0.056 | +0.00% | | −0.200 | +0.00% | | 0.150 | +0.75% | | 0.015 | +0.11% | |
| Never | −0.299 | −1.02% | | 0.179 | +0.00% | | 0.020 | +0.00% | | −1.123 | −3.38% | | 0.153 | +1.79% | |
| Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) |
| Sometimes | 0.591 | +3.01% | ** | 0.298 | +0.00% | | 0.482 | +0.00% | | 0.188 | +0.96% | | 0.502 | +6.79% | ** |
| Never | 0.349 | +1.58% | * | 0.223 | +0.00% | | −0.024 | +0.00% | | 0.152 | +0.76% | | 0.069 | +0.75% | |
| Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) |
| Sometimes | 1.097 | +7.14% | | 14.932 | +5.91% | | 16.322 | +21.79% | | 0.298 | +1.62% | | −0.328 | −3.36% | |
| Never | 1.063 | +6.81% | | 15.109 | +6.97% | | 16.376 | +22.73% | | 0.482 | +2.87% | | −0.339 | −3.46% | |
| Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) |
| Yes | 0.253 | +1.09% | | 0.115 | +0.00% | | −0.188 | +0.00% | | 0.153 | +0.77% | | 0.240 | +2.94% | |
| Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future |
| | 0.225 | | *** | 0.241 | | *** | 0.222 | | *** | 0.197 | | *** | 0.129 | | *** |
| Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks |
| | 0.051 | | | 0.053 | | | 0.068 | | | 0.054 | | | 0.059 | | |
| Constant | −3.167 | | *** | −17.736 | | | −17.640 | | | −2.967 | | ** | −1.915 | | ** |
| N | 1000 | | | 1000 | | | 500 | | | 500 | | | 1000 | | |
## 3.1.3. Political belief factors
Regarding political factors, the results from Table 7 are clear, there is no correlation between political belongings and GTs decisions. A plausible explanation could be that the subject is too new to be politicized. No party in Switzerland yet has formulated a clear opinion on the subject, neither on the related data. Hence, the belonging to a party or a movement of thought does not translate in a clear differentiation between individuals' responses.
**Table 7**
| Model | A– Baseline GTW | A– Baseline GTW.1 | A– Baseline GTW.2 | B–Price display | B–Price display.1 | B–Price display.2 | C1–Insurer payer | C1–Insurer payer.1 | C1–Insurer payer.2 | C2–Self payer | C2–Self payer.1 | C2–Self payer.2 | D–Data sharing | D–Data sharing.1 | D–Data sharing.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Model | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. |
| Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) |
| Yes | 0.151 | +3.77% | | 0.188 | +4.47% | | 0.018 | +0.46% | | 0.184 | +4.44% | | −0.059 | −1.22% | |
| Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) | Feeling close to a political party (baseline: No) |
| Yes | 0.000 | +0.02% | | 0.107 | +2.50% | | −0.099 | −2.45% | | −0.075 | −1.76% | | 0.167 | +3.58% | |
| Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) |
| | 0.191 | +4.78% | | −0.183 | −4.15% | | 0.390 | +9.26% | | −0.338 | −7.59% | | 0.259 | +5.63% | |
| Constant | −0.131 | | | −0.542 | | ** | 0.259 | | | −0.486 | | * | −0.884 | | *** |
| N | 1000 | | | 1000 | | | 500 | | | 500 | | | 1000 | | |
## 3.1.4. Insurance factors
In the set of insurance, we find that the chosen features of the mandatory health insurance (insurance plan and deductible) do not allow to consistently distinguish individuals who are more willing to take a GT. Three other variables, however, allow to do so. In our sample, we document that individuals who hold a complementary health insurance policy display a different behavior. More precisely, the factor comes into play when the decision to undergo a GT is faced with the cost, that is, in questions B and C2. For both cases, individuals who do own such a policy are more willing to undergo a GT by 9.33 and $14.70\%$, respectively, as reported in Table 8. A potential explanation could be that individuals with a complementary health insurance are less cost-conscious as the healthcare costs are alleviated. This usually leads to an increase in healthcare consumption as highlighted by Schmitz [37], thus encompassing genetic testing. Another interesting effect induced by this variable is the decrease in the DSW with the health insurer (question D): having a complementary health insurance renders individuals less likely by $7.92\%$ to share the data. This may be correlated with the fact that the calculation of premiums for complementary health insurance in Switzerland, contrary to basic health insurance, is based, among other characteristics, on the health condition and family history. The next variable is binary, indicating whether the individual has an app from the insurer for step counting or recording exercise, participants who ticked “yes” have an increased GTW, except when the insurer is the payer, in which case, the coefficient is not significant. This outcome is rather intriguing and an underlying rationale could be that individuals who are interested in their health in the first place are more likely to download the health app. This interest then makes them more likely to be interested in performing a GT, unless when it is the insurer who is the payer, where more respondents are more interested in general, thus annihilating the significance of the difference. When it comes to sharing the anonymized data from the GT with the health insurer, the same rationale can be applied. In fact, these individuals that already share data from the app with the health insurer are $27.9\%$ ($p \leq 0.001$) more willing to share GT data.
**Table 8**
| Model | A– Baseline GTW | A– Baseline GTW.1 | A– Baseline GTW.2 | B–Price display | B–Price display.1 | B–Price display.2 | C1–Insurer payer | C1–Insurer payer.1 | C1–Insurer payer.2 | C2–Self payer | C2–Self payer.1 | C2–Self payer.2 | D–Data sharing | D–Data sharing.1 | D–Data sharing.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Model | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. |
| Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) |
| Standard | 0.164 | +0.54% | | 0.457 | +0.61% | ** | 0.016 | +0.50% | | 0.412 | +0.60% | | −0.229 | −5.11% | |
| HMO | −0.171 | −4.22% | | −0.113 | −2.18% | | −0.060 | −1.45% | | −0.413 | +6.79% | | 0.040 | +0.90% | |
| CallMed | −0.044 | −1.10% | | −0.034 | −0.72% | | 0.088 | +2.10% | | −0.282 | +7.82% | | −0.584 | −12.12% | * |
| Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) |
| CHF 500 – 2000 | −0.134 | −3.32% | | −0.034 | −0.72% | | 0.033 | +0.78% | | 0.279 | +12.53% | | 0.196 | +4.59% | |
| CHF 2500 | −0.024 | −0.61% | | 0.239 | +4.79% | | 0.085 | +2.01% | | 0.461 | +14.07% | * | 0.053 | +1.19% | |
| Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) |
| Yes | 0.008 | +0.18% | | 0.445 | +9.33% | ** | −0.179 | −4.36% | | 0.535 | +14.69% | * | −0.391 | −8.46% | ** |
| Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) |
| Yes | 0.671 | +16.47% | *** | 0.793 | +17.63% | *** | 0.418 | +9.46% | | 0.530 | +14.65% | * | 1.147 | +27.88% | *** |
| Constant | −0.074 | | | −1.126 | | *** | 0.396 | | * | −1.334 | | *** | −0.581 | | *** |
| N | 1000 | | | 1000 | | | 500 | | | 500 | | | 1000 | | |
## 3.1.5. Sentiment factors
Finally, our last set of variables included in the regression model in Table 9 exhibits the most significant correlations. Thereby, several variables are worth particular attention. The two first are curiosity and disease risk information. Whereas, the logic behind genetic testing take up, i.e., curiosity driving the GTW, is sound, the result generated by the second (not significant) variable is intriguing. Indeed, our model suggests that it is the simple curiosity rather than any health-related considerations, captured by the disease risk information variable, that drive the GT decision. This observation has already been made several times through our analysis with the health and lifestyle variables, thus giving further confirmation. Moreover, the curiosity is self-based as it is only enough for GT itself and does not extend to the DSW with the health insurer. Another pair of factors, however, present a pattern and they both display altruistic features. For the individuals stating that helping relatives or incentivize them to do a GT is a rather strong incentive for them to undergo one, they present different behaviors in certain cases. When the price is not yet displayed, in question A, or when it is the insurer who is the payer, in question C1, these incentives seem to differentiate respondents' choices. The effects range from $5.5\%$ of increase in the GTW in question A for helping relatives take better care of their health to $22.4\%$ for incentivizing a relative to undergo a test when the insurer pays for it. However, this altruism stops when individuals have to pay themselves. Ultimately, for those who could undergo a GT to incentivize relatives to do so, they are more likely to be willing to share these results with the health insurer.
**Table 9**
| Model | A– Baseline GTW | A– Baseline GTW.1 | A– Baseline GTW.2 | B–Price display | B–Price display.1 | B–Price display.2 | C1–Insurer payer | C1–Insurer payer.1 | C1–Insurer payer.2 | C2–Self payer | C2–Self payer.1 | C2–Self payer.2 | D–Data sharing | D–Data sharing.1 | D–Data sharing.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Model | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. | βk | p k | sig. |
| Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) |
| Curiosity | 1.130 | +18.92% | *** | 1.121 | +15.77% | *** | 0.923 | +19.38% | *** | 1.136 | +19.14% | *** | 0.197 | +3.15% | |
| Better health prevention | 0.732 | +10.88% | *** | 0.441 | +4.83% | * | 0.503 | +9.72% | | 0.465 | +6.35% | | 0.408 | +6.90% | |
| Disease risk information | 0.356 | +4.66% | | 0.229 | +2.30% | | 0.072 | +1.28% | | −0.097 | −1.12% | | −0.438 | −5.63% | |
| Help relatives health prevention | 0.413 | +5.51% | * | 0.340 | +3.58% | | 0.672 | +13.47% | * | 0.288 | +3.70% | | 0.228 | +3.67% | |
| Incentivize relatives to undergo test | 0.547 | +7.65% | ** | 0.201 | +1.99% | | 1.046 | +22.37% | ** | 0.316 | +4.09% | | 0.405 | +6.85% | * |
| Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) |
| Impact on family finances | −0.027 | −0.34% | | −0.210 | −1.82% | | 0.629 | +12.49% | * | −0.221 | −2.40% | | 0.262 | +4.25% | |
| Family would disapprove | 0.129 | +1.54% | | 0.239 | +2.41% | | −0.386 | −5.79% | | −0.268 | −2.86% | | 0.273 | +4.44% | |
| Fear test too costly | −0.059 | −0.69% | | −0.607 | −4.46% | *** | 0.580 | +11.41% | * | −0.593 | −5.58% | * | −0.042 | −0.60% | |
| Do not want to know | −0.837 | −7.13% | *** | −0.585 | −4.33% | ** | −1.343 | −14.85% | *** | −0.175 | −1.94% | | −0.073 | −1.04% | |
| Induced lifestyle changes | 0.236 | +2.95% | | 0.234 | +2.34% | | 0.560 | +10.97% | * | −0.305 | −3.20% | | 0.183 | +2.91% | |
| Fear of discrimination | −0.365 | −3.72% | | 0.198 | +1.95% | | 0.005 | +0.12% | | 0.116 | +1.38% | | −0.127 | −1.79% | |
| Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) |
| Family discriminated for insurance | 0.034 | +0.38% | | −0.028 | −0.29% | | −0.470 | −6.87% | | 0.225 | +2.82% | | −0.259 | −3.52% | |
| Fewer illnesses, longer life | 0.110 | +1.30% | | 0.383 | +4.10% | * | 0.567 | +11.11% | * | 0.266 | +3.39% | | 0.331 | +5.49% | * |
| Testing will be common | 0.712 | +10.52% | *** | 0.675 | +8.09% | *** | 0.179 | +3.23% | | 1.052 | +17.34% | *** | 0.315 | +5.20% | |
| Testing mandatory to be hired | 0.229 | +2.85% | | 0.339 | +3.56% | | 0.105 | +1.87% | | 0.537 | +7.54% | | 0.932 | +17.96% | *** |
| Testing for insurance premiums | −0.309 | −3.22% | | −0.655 | −4.72% | ** | −0.026 | −0.41% | | −0.328 | −3.42% | | −0.357 | −4.72% | * |
| Genetic passport for all | 0.098 | +1.14% | | 0.657 | +7.82% | *** | −0.223 | −3.51% | | 0.592 | +8.46% | * | 0.257 | +4.17% | |
| Segregation bad/good genomes | 0.172 | +2.09% | | 0.174 | +1.70% | | −0.205 | −3.23% | | −0.129 | −1.46% | | −0.342 | −4.54% | |
| Discrimination of handicaped | −0.257 | −2.73% | | 0.094 | +0.87% | | −0.270 | −4.18% | | −0.014 | −0.20% | | −0.246 | −3.36% | |
| Government not able to protect | −0.702 | −6.28% | *** | −0.238 | −2.40% | | −0.600 | −8.43% | * | −0.255 | −2.74% | | −0.070 | −1.00% | |
| Sequencing of infants genome | 0.247 | +3.11% | | −0.087 | −0.81% | | −0.100 | −1.61% | | 0.152 | +1.84% | | 0.220 | + 3.54% | |
| Sequencing of fœtuses genome | −0.275 | −2.90% | | −0.172 | −1.52% | | 0.786 | +16.10% | ** | −0.281 | −2.98% | | −0.148 | −2.08% | |
| Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) |
| Yes | 0.784 | +11.86% | *** | 0.509 | + 5.73% | * | 0.644 | +12.84% | * | 0.021 | +0.22% | | 0.620 | +11.09% | ** |
| Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) |
| Yes | 0.166 | +2.02% | | 0.322 | +3.36% | | 0.087 | +1.54% | | 0.318 | +4.13% | | −0.348 | −4.60% | * |
| Constant | −1.868 | | *** | −2.157 | | *** | −1.292 | | *** | −1.858 | | *** | −1.501 | | *** |
| N | 1000 | | | 1000 | | | 500 | | | 500 | | | 1000 | | |
Regarding deterrents, cost is an issue. The fear that the test is too costly especially arises when the price is displayed in question B. With strong significance, individuals for whom the cost may be a hurdle are $4.5\%$ less likely to undergo the test in general and $5.6\%$ when they are the sole payer. On the contrary, when it is the insurer who is supposed to pay for the test, respondents who had an issue with the expenditure are now $11.4\%$ more likely to undergo the test once that burden is taken away. The last significant variable in this group of barriers is the lack of desire to know what potential disease one could have in the future. Not wanting to know correlates with a decrease in $7\%$ in the overall willingness (question A) and of $14\%$ when the insurer bears the cost. This correlation, nevertheless disappears in the last two regressions, question C1 and DSW. Interestingly, fear of discrimination is not significant in our model, despite being fairly present in the literature [[12, 34] to name a few].
Subsequently, we capture the outlook of the respondents on genetic testing and its future. We first note that those who agree or completely agree that GT will be common are distinguishable when price comes in question from those who did not. Their belief pushes them to perform the test when the price is displayed, giving an edge compared to those who do not believe so. Another belief—that the government will not be able to protect its citizens against negative aspects of genetic testing—has a significant impact. It translates into a decrease in the GTW in models A and C1. Especially when the cost of the test is taken care of by the health insurer, the individuals who share this opinion are $8.4\%$ less likely to undergo the test. Compelling enough, this attitude does not give a significant difference when it comes to deciding whether to share the data with the health insurer. Regarding that last question, the DSW with the health insurer, respondents who agree that testing will be mandatory before being hired are $17.9\%$ more likely to do so. Curiously, this perspective, though, does not make them more likely to perform the test. Finally, we study the usage of health-related applications. First, the respondents who use health-related apps for step counting, sleep cycle, or women's health, for instance, have a higher propensity of accepting to undergo the test, except when they are the payers. This could be easily explained by the fact that these individuals are already familiar with health technologies and are willing to use them to monitor their health. However, these results hint again that this behavior is not driven by health considerations but rather by curiosity. This observation being backed up several times in our study is once again confirmed by the non-significance of the last usage of health-related apps for prevention factor. Regarding DSW, these last two variables present conflicting results, suggesting that those who use health apps are more likely to share the anonymized data but using this app for prevention renders them less likely to do so.
## 3.1.6. Effect of the payer framing
In this section, we document the effect of the health insurer as a payer framing on GTW as well as DSW, as outlined in Section 2.2 and Figure 1. We capture this effect by introducing an “insurer framing” dummy variable in the GTW regressions of question C and the DSW of question D. To this aim, we first aggregate the data of questions C1 and C2, and we subsequently control for the framing by regressing the outcomes on the health insurer as a payer binary variable. By doing so, we witness a difference in outcome between the two groups, as suggested by the statistical analysis. The results of our regression in Table 10 corroborate with the observation made in the descriptive statistics analysis—the two framings present significantly different outcomes on the GTW. As the coefficient suggests, individuals who were told that it is the health insurer who should finance these GTs are $24.8\%$ more likely to undergo the test, compared to individuals who would bear the cost of the test themselves. One can hypothesize that the insurer as a payer triggers more individuals to undergo the test because of the cost relief. To verify this conjecture, we run a subsidiary regression with the interaction term Health insurer framing × Fear that test would be too costly. When crossed with the health insurer as a payer variable, the fear of the test to be too costly is statistically significant at a $90\%$ confidence level and has a coefficient of 0.724, thus validating the hypothesis that the health insurer as a payer alleviates the fear that the test may be too costly.
**Table 10**
| Model | C–GTW | C–GTW.1 | C–GTW.2 | D–Data sharing | D–Data sharing.1 | D–Data sharing.2 |
| --- | --- | --- | --- | --- | --- | --- |
| Model | βk | p k | sig. | βk | p k | sig. |
| Insurer framing (baseline: No) | Insurer framing (baseline: No) | Insurer framing (baseline: No) | Insurer framing (baseline: No) | Insurer framing (baseline: No) | Insurer framing (baseline: No) | Insurer framing (baseline: No) |
| Yes | 1.016 | +24.81% | *** | 0.406 | +9.18% | ** |
| Constant | −0.585 | | *** | −0.846 | | *** |
| Observations | 1000 | | | | | |
When moving to the DSW, we as well witness a difference in outcome between both groups, as suggested by the statistical analysis. Our regression coefficient provides empirical evidence that the individuals for whom the health insurer is the payer of the GT would be $9.18\%$ more likely to share the test's anonymized data with the health insurer, when compared to individuals who are the sole payers of the GTs with $99\%$ confidence. Making use of framings to carry another dimension into the analysis of GTW and DSW, we highlight the critical importance of the payer of these tests. From the findings in this framework emerge a new perspective in which the health insurer and the insured establish a collaboration relationship. When the health insurer pays for the GT to be undergone by the insured, which can lead to actionable information, this can be viewed as an investment into the individual's health capital. In return, the insured shares the anonymized data. A possible explanation of this significantly different behavior could be that it stems from a latent feeling of indebtedness toward the health insurer, rather than collaboration. However, it is not possible to determine the extent to which this might play a role in practice.
## 3.2. Robustness checks and additional analysis
We now assess the robustness of our results by performing several checks. First, in Table 11, we produce confusion matrices on 10,000 bootstrapped samples, providing the mean accuracy for the five models in regard to each variable of interest. The mean accuracy spreads between 51 and $79\%$. The best performing model at explaining GTW is sentiment-related. Its accuracy ranges between 71 and $79\%$. Unsurprisingly, the model that performs the worst concerns political belief factors: there are no significant variables for this model. Finally, it is usually the DSW that is best explained (model D).
**Table 11**
| Model | A–Baseline GTW | B–Price display | C1–Insurer payer | C2–Self payer | D–Data sharing |
| --- | --- | --- | --- | --- | --- |
| Socioeconomic | 0.61 | 0.63 | 0.65 | 0.66 | 0.64 |
| Insurance | 0.55 | 0.63 | 0.6 | 0.65 | 0.69 |
| Lifestyle | 0.62 | 0.65 | 0.64 | 0.66 | 0.67 |
| Political belief | 0.51 | 0.62 | 0.6 | 0.65 | 0.67 |
| Sentiment | 0.79 | 0.76 | 0.79 | 0.76 | 0.71 |
## 3.2.1. Total regression, StepAIC, and reduced form regression results
For the second robustness check, we calibrate several regression models to test the sensitivity of our coefficients. In a robust model, coefficients should almost not vary when new variables are introduced, a case that we simulate by running a regression making use of all our variables. Another aim of conducting a regression comprising all the variables is to subsequently reduce the model with a selection based on the AIC. This procedure keeps the variables that improve the explanatory power of the model and hence provide another mapping of variable importance in GT decision. In Table 12, we display both the total regression model and the reduced model. A first observation, we can make regards for the robustness of the coefficients. Expectedly, we can notice that coefficients of significant variables vary much less than those that are not significant. For instance, health, a variable that is not significant, has a coefficient that changes from 0.120 to −0.153 in the case of good health reported by the surveyed person. The factor curiosity, on the contrary, has a stable coefficient with only a minor change from 1.130 to 1.150 from the reduced to the total regression. Another interesting perspective is the change in the significance of the coefficients. Merging all the variables together has confirmed previous findings pointing at the importance of sentiment-related factors. Indeed, in the total regression, other sets of variables which displayed a few significant drivers in the separate models lose their importance when merged together, leaving almost solely significance to the sentiment factors. Finally, if we take a closer look at the analysis of the DSW, we notice that two variables remain highly significant and bear a strong coefficient: having a health insurer's application and the insurer's framing (framing 2). The latter even displays a stronger coefficient, increasing from 0.409 to 0.575 while remaining significant at a $99\%$ level of confidence. These findings confirm the importance of the relationship between the insurer and the respondent in the DSW.
**Table 12**
| Model | A–Baseline GTW | A–Baseline GTW.1 | A–Baseline GTW.2 | A–Baseline GTW.3 | B–Price display | B–Price display.1 | B–Price display.2 | B–Price display.3 | C1–Insurer payer | C1–Insurer payer.1 | C1–Insurer payer.2 | C1–Insurer payer.3 | C2–Self payer | C2–Self payer.1 | C2–Self payer.2 | C2–Self payer.3 | D–Data sharing | D–Data sharing.1 | D–Data sharing.2 | D–Data sharing.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Model | Total | Total | Reduced | Reduced | Total | Total | Reduced | Reduced | Total | Total | Reduced | Reduced | Total | Total | Reduced | Reduced | Total | Total | Reduced | Reduced |
| Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) |
| Female | 0.088 | | | | 0.193 | | | | 0.407 | | | | −0.179 | | | | −0.264 | | −0.324 | * |
| Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) |
| 25 – 34 years | 0.021 | | 0.041 | | 0.081 | | | | 0.369 | | 0.297 | | 0.496 | | | | 0.032 | | | |
| 45 – 54 years | −0.224 | | −0.154 | | −0.178 | | | | −0.880 | * | −0.877 | * | 0.216 | | | | 0.010 | | | |
| 55 – 64 years | −0.626 | * | −0.567 | * | −0.194 | | | | −0.510 | | | | | | | | | | | |
| Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) |
| Yes | −0.458 | * | −0.487 | * | −0.529 | ** | −0.555 | ** | −0.836 | * | −0.782 | * | −0.541 | | −0.509 | * | −0.085 | | | |
| Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) |
| Yes | 0.231 | | | | 0.040 | | | | 0.028 | | | | −0.317 | | | | −0.254 | | −0.229 | |
| Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) |
| Part-time | 0.074 | | 0.074 | | −0.127 | | | | −0.421 | | | | 0.097 | | | | 0.032 | | | |
| Other | 0.581 | * | 0.561 | * | 0.091 | | | | 0.174 | | | | 0.126 | | | | 0.075 | | | |
| Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) |
| Above average | 0.251 | | | | 0.209 | | | | 0.040 | | | | 0.885 | ** | 0.657 | ** | −0.303 | | −0.250 | |
| Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) |
| Yes | 0.266 | | 0.301 | | −0.077 | | | | −0.131 | | | | 0.067 | | | | 0.133 | | | |
| Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) |
| Yes | 0.266 | | | | −0.039 | | | | −0.277 | | | | −0.234 | | | | −0.123 | | | |
| Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) |
| Average | −0.038 | | | | 0.214 | | | | 0.755 | | | | 0.332 | | | | 0.294 | | 0.458 | |
| Good | −0.153 | | | | 0.246 | | | | 0.023 | | | | 0.100 | | | | −0.086 | | 0.131 | |
| Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) |
| German-speaking | 0.047 | | | | 0.495 | * | 0.507 | ** | −0.202 | | | | 0.339 | | | | 0.558 | ** | 0.580 | *** |
| Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) |
| Sometimes | 0.055 | | | | 0.306 | | 0.327 | | 0.054 | | | | −0.018 | | | | 0.156 | | | |
| Never | 0.041 | | | | 1.026 | * | 0.934 | * | 0.563 | | | | −0.199 | | | | 0.461 | | | |
| Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) |
| Sometimes | 0.655 | ** | 0.627 | ** | 0.063 | | | | 0.169 | | | | −0.143 | | | | 0.284 | | | |
| Never | 0.448 | * | 0.451 | * | 0.198 | | | | −0.660 | | | | 0.301 | | | | −0.016 | | | |
| Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) |
| Sometimes | 0.338 | | | | 15.473 | | 14.815 | | 15.171 | | | | 0.322 | | | | −0.740 | | | |
| Never | 0.247 | | | | 15.603 | | 14.971 | | 15.142 | | | | 0.622 | | | | −0.651 | | | |
| Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) |
| Yes | 0.215 | | | | −0.207 | | | | −0.400 | | | | −0.071 | | | | 0.107 | | | |
| Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future |
| | 0.023 | | | | 0.060 | | | | 0.090 | | | | −0.011 | | | | 0.041 | | | |
| Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks |
| | 0.020 | | | | −0.005 | | | | 0.121 | | 0.095 | | −0.031 | | | | 0.026 | | | |
| Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) |
| Standard | 0.160 | | | | 0.495 | * | 0.407 | * | 0.443 | | | | 0.605 | | | | −0.310 | | −0.278 | |
| HMO | −0.350 | | | | −0.273 | | −0.261 | | −0.149 | | | | −0.099 | | | | 0.095 | | 0.128 | |
| CallMed | −0.254 | | | | −0.327 | | −0.299 | | −0.381 | | | | −0.367 | | | | −0.735 | ** | −0.760 | ** |
| Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) |
| CHF 500 – 2000 | −0.203 | | | | −0.047 | | | | 0.072 | | | | 0.469 | | 0.519 | | 0.190 | | | |
| CHF 2500 | −0.084 | | | | 0.274 | | | | 0.022 | | | | 0.780 | * | 0.566 | * | 0.134 | | | |
| Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) |
| Yes | 0.002 | | | | 0.681 | *** | 0.680 | *** | −0.131 | | | | 0.613 | * | 0.419 | | −0.423 | * | −0.447 | ** |
| Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) |
| Yes | 0.402 | | 0.340 | | 0.674 | ** | 0.565 | ** | 0.127 | | | | 0.200 | | | | 0.930 | *** | 1.043 | *** |
| Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) |
| Yes | | | | | | | | | | | | | | | | | 0.575 | *** | 0.549 | *** |
| Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) |
| Curiosity | 1.115 | *** | 1.126 | *** | 1.104 | *** | 1.200 | *** | 0.951 | ** | 0.843 | ** | 1.084 | *** | 1.211 | *** | 0.154 | | | |
| Better health prevention | 0.733 | ** | 0.848 | *** | 0.392 | | 0.553 | ** | 0.418 | | 0.575 | | 0.680 | | 0.501 | | 0.505 | * | 0.419 | * |
| Disease risk information | 0.407 | | 0.473 | * | 0.264 | | | | 0.316 | | | | −0.022 | | | | −0.365 | | | |
| Help relatives health prevention | 0.410 | | 0.418 | * | 0.350 | | 0.496 | ** | 0.556 | | 0.686 | * | 0.438 | | 0.423 | | 0.096 | | | |
| Incentivize relatives to undergo test | 0.537 | * | 0.516 | * | 0.263 | | | | 1.312 | *** | 1.175 | *** | 0.373 | | | | 0.486 | * | 0.561 | ** |
| Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) | Is a reason not to undergo genetic testing (baseline: No) |
| Impact on family finances | −0.031 | | | | −0.203 | | | | 0.779 | * | 0.691 | * | −0.340 | | | | 0.297 | | 0.335 | * |
| Family would disapprove | 0.071 | | | | 0.289 | | | | −0.553 | | −0.515 | | −0.280 | | | | 0.200 | | | |
| Test too costly | 0.015 | | | | −0.594 | ** | −0.589 | *** | 0.615 | * | 0.583 | * | −0.631 | * | −0.740 | ** | −0.030 | | | |
| Do not want to | −0.854 | *** | −0.754 | *** | −0.685 | *** | −0.614 | ** | −1.364 | *** | −1.429 | *** | −0.250 | | | | −0.138 | | | |
| Induced lifestyle changes | 0.183 | | | | 0.149 | | | | 0.356 | | 0.502 | | −0.432 | | | | 0.142 | | | |
| Fear of discrimination | −0.400 | | −0.429 | * | 0.309 | | 0.350 | | 0.136 | | | | 0.303 | | | | −0.039 | | | |
| Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) | Impact of genetic tests on society (baseline: No) |
| Family discriminated for insurance | 0.027 | | | | 0.110 | | | | −0.749 | * | −0.626 | * | 0.373 | | | | −0.224 | | | |
| Fewer illnesses, longer life | 0.058 | | | | 0.311 | | 0.348 | * | 0.715 | * | 0.707 | ** | 0.055 | | | | 0.401 | * | 0.439 | ** |
| Testing will be common | 0.719 | *** | 0.807 | *** | 0.752 | *** | 0.800 | *** | 0.264 | | | | 1.142 | *** | 0.918 | *** | 0.422 | * | 0.514 | ** |
| Testing mandatory to be hired | 0.138 | | | | 0.344 | | | | 0.083 | | | | 0.707 | | 0.474 | | 0.830 | *** | 0.861 | *** |
| Testing for insurance premiums | −0.367 | | −0.312 | | −0.838 | *** | −0.699 | *** | −0.098 | | | | −0.435 | | | | −0.437 | * | −0.460 | * |
| Genetic passport for all | 0.107 | | | | 0.744 | *** | 0.680 | *** | −0.353 | | | | 0.837 | ** | 0.656 | * | 0.183 | | | |
| Segregation bad/good genomes | 0.195 | | | | 0.268 | | | | 0.027 | | | | −0.106 | | | | −0.367 | | −0.358 | * |
| Discrimination of handicaped | −0.255 | | | | 0.010 | | | | −0.231 | | | | −0.137 | | | | −0.323 | | −0.347 | |
| Government not able to protect | −0.723 | *** | −0.661 | *** | −0.266 | | | | −0.393 | | −0.611 | * | −0.382 | | −0.468 | | 0.009 | | | |
| Sequencing of infants genome | 0.262 | | 0.307 | | −0.102 | | | | −0.130 | | | | 0.262 | | | | 0.214 | | | |
| Sequencing of fœtuses genome | −0.427 | * | −0.377 | | −0.280 | | | | 0.729 | * | 0.658 | * | −0.500 | | | | −0.162 | | | |
| Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) |
| Yes | 0.737 | ** | 0.929 | *** | 0.348 | | 0.403 | | 0.624 | | 0.796 | ** | −0.092 | | | | 0.510 | * | 0.431 | * |
| Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) |
| Yes | 0.273 | | | | 0.409 | * | 0.385 | * | 0.316 | | | | 0.407 | | 0.368 | | −0.211 | | | |
| Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) |
| Yes | 0.044 | | | | −0.026 | | | | 0.132 | | | | −0.085 | | | | −0.039 | | | |
| Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) |
| Yes | −0.249 | | | | 0.007 | | | | −0.791 | * | −0.662 | * | −0.297 | | | | −0.062 | | | |
| Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) | Political orientation (baseline: left) |
| | 0.176 | | | | −0.431 | | | | 1.740 | ** | 1.413 | * | −0.820 | | −0.827 | | 0.223 | | | |
| Constant | −2.784 | * | −1.981 | *** | −18.859 | | −17.768 | | −17.287 | | −1.449 | * | −2.725 | | −2.206 | *** | −1.759 | | −2.040 | *** |
| N | 1000 | | | | 1000 | | | | 500 | | | | 500 | | | | 1000 | | | |
Using the same Table 12 but now looking at the coefficients of the reduced model, we have another evidence of the importance of sentiment variables as well as insurance-related ones. The variables remaining in the reduced models mostly come from the sentiment factors and for the DSW decision variables from the insurance factors.
## 3.2.2. Random forest
As a last analysis and robustness check for the importance of factors in the decision process of GTW or data sharing, we report results obtained from the random forest application.4 In our case, we classify each respondent whether he/she is likely to undergo a GT, respectively, to share the data or not. The algorithm performs the best classification and we extract the ranking of each variable, which is presented in Table 13 (see column “RF”). The variables considered as the most important are the ones that allow as soon as possible to classify the highest number of individuals into either group with the highest accuracy. We find that the first ranks stem from the incentive sentiment factors for GTW, for insurer's app usage and genetic testing impact for DSW.
**Table 13**
| Model | A–Baseline GTW | A–Baseline GTW.1 | B–Price display | B–Price display.1 | C1–Insurer payer | C1–Insurer payer.1 | C2–Self payer | C2–Self payer.1 | D–Data sharing | D–Data sharing.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Model | Reduced | RF | Reduced | RF | Reduced | RF | Reduced | RF | Reduced | RF |
| Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) | Gender (baseline: Male) |
| Female | | | | | | | | | ✓ | |
| Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) | Age (baseline: 35 – 44 years) |
| 25 – 34 years | ✓ | | | | ✓* | (10) | | | | |
| 45 – 54 years | ✓ | | | | ✓* | (10) | | | | |
| 55 – 64 years | ✓* | | | | ✓ | (10) | | | | |
| Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) | Swiss nationality (baseline: No) |
| Yes | ✓* | | ✓** | | ✓* | | ✓* | | | |
| Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) | Higher education (baseline: No) |
| Yes | | | | | | | | | ✓ | |
| Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) | Professional status (baseline: Full-time employed) |
| Part-time | ✓ | | | | | | | | | |
| Other | ✓* | | | | | | | | | |
| Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) | Subjective wealth (baseline: Below average) |
| Above average | | | | | | | ✓** | (10) | ✓ | |
| Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) | Married (baseline: No) |
| Yes | ✓ | | | | | | | | | |
| Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) | Cancer history (baseline: No) |
| Yes | | | | | | | | | | |
| Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) | Health (baseline: Bad) |
| Average | | | | | | | | | ✓ | |
| Good | | | | | | | | | ✓ | |
| Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) | Region (baseline: French-speaking) |
| German-speaking | | | ✓** | | | | | | ✓*** | |
| Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) | Alcohol consumption (baseline: Everyday) |
| Sometimes | | | ✓ | | | | | | | |
| Never | | | ✓** | | | | | | | |
| Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) | Cigarettes consumption (baseline: Everyday) |
| Sometimes | ✓** | | | | | | | | | |
| Never | ✓* | | | | | | | | | |
| Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) | Fruits and vegetables consumption (baseline: Everyday) |
| Sometimes | | | ✓ | | | | | | | |
| Never | | | ✓ | | | | | | | |
| Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) | Sport at least once a week (baseline: No) |
| Yes | | | | | | | | | | |
| Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future | Level of planning for the future |
| | | | | (10) | | | | | | |
| Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks | Level of loving taking risks |
| | | | | | ✓ | | | | | |
| Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) | Insurance plan (baseline: Family doctor) |
| Standard | | | ✓* | | | | | | ✓ | |
| HMO | | | ✓ | | | | | | ✓ | |
| CallMed | | | ✓ | | | | | | ✓** | |
| Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) | Insurance deductible (baseline: CHF 300) |
| CHF 500 – 2000 | | | | | | | ✓ | | | |
| CHF 2500 | | | | | | | ✓* | | | |
| Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) | Complementary insurance (baseline: No) |
| Yes | | | ✓*** | | | | ✓ | | ✓** | |
| Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) | Insurer's app (baseline: No) |
| Yes | ✓ | | ✓** | | | | ✓ | | ✓*** | (1) |
| Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) | Insurer's framing (baseline: No) |
| Yes | | | | | | | | | ✓*** | |
| Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) | Is an incentive to undergo genetic testing (baseline: No) |
| Curiosity | ✓*** | (1) | ✓*** | (1) | ✓** | (2) | ✓*** | (1) | | |
| Better health prevention | ✓*** | (2) | ✓** | (2) | ✓ | (1) | ✓ | (2) | ✓* | |
| Disease risk information | ✓* | (3) | | (3) | | (3) | | (5) | | |
| Help relatives health prevention | ✓* | (5) | ✓** | (4) | ✓* | (5) | ✓ | (4) | | (7) |
| Incentivize relatives to undergo test | ✓* | (4) | | (6) | ✓*** | (4) | | | ✓** | (6) |
| Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) | Is not an incentive to undergo genetic testing (baseline: No) |
| Impact on family finances | | | | | ✓* | | | | ✓* | |
| Family would disapprove | | | | | ✓ | | | | | |
| Test too costly | | | ✓*** | | ✓* | | ✓** | (9) | | |
| Do not want to | ✓*** | (8) | ✓** | (8) | ✓*** | | | | | |
| Induced lifestyle changes | | | | | | | | | | |
| Fear of discrimination | ✓* | | ✓ | | | | | | | |
| Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) | Impact of genetic tests (baseline: No) |
| Family discriminated for insurance | | | | | ✓* | | | | | |
| Fewer illnesses, longer life | | (9) | | (8) | ✓* | (7) | ✓** | (7) | ✓** | (3) |
| Testing will be common | ✓*** | (7) | ✓*** | (5) | | | ✓*** | (3) | ✓** | (4) |
| Testing mandatory to be hired | | | | | | | ✓ | | ✓*** | (2) |
| Testing for insurance premiums | ✓ | | ✓*** | | | | | | ✓* | |
| Genetic passport for all | | | ✓*** | (7) | | | ✓* | (8) | | (9) |
| Segregation bad/good genomes | | | | | | | | | ✓* | |
| Discrimination of handicaped | | | | | | | | | ✓ | |
| Government not able to protect | ✓*** | | | | ✓* | | ✓ | | | |
| Sequencing of infants genome | ✓ | | | | | | | | | (5) |
| Sequencing of fœtuses genome | ✓ | | | (9) | ✓* | | | | | |
| Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) | Usage of health-related apps (baseline: No) |
| Yes | ✓*** | (6) | ✓ | | ✓** | | | | ✓* | |
| Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) | Usage of health-related apps for prevention (baseline: No) |
| Yes | | (10) | ✓* | (9) | | (6) | ✓ | | | (10) |
| Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) | Political interest (baseline: No) |
| Yes | | | | | | | | | | |
| Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) | Belong to a political party (baseline: CHF 300) |
| Yes | | | | | ✓* | | | | | |
| Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) | Political orientation (baseline: Left) |
| | | | | | ✓ | | ✓ | | | |
| N | 1000 | | 1000 | | 500 | | 500 | | 1000 | |
## 4. Discussion and conclusion
GTs by essence give access to personalized health. Understanding what drives the decision to undergo these tests and the associated fears is crucial for personalized health-oriented policies. The two-fold aim of this article is reflected in the design of the ad hoc survey. To fill the gap in the literature and better understand health-related decisions, we first analyzed the factors influencing genetic testing uptake as well as the sharing of the anonymized data from the GT with the health insurer. To do so, we ran regressions on five sets of variables susceptible to influence individuals' behavior regarding their GTW decision, including socioeconomic, insurance, lifestyle, political beliefs, and sentiment factors. We find that socioeconomic (age and nationality) and lifestyle factors (smoking habits and planning for the future) have little or inconsistent significant influence, while the insurance (complementary health insurance and insurer's app usage) and sentiment factors (e.g., curiosity and health prevention) present strong and significant results regarding GTW. These findings are corroborated by random forest modeling robustness checks. For instance, the perception of GT as a mean for health prevention pushes the individual's propensity for testing by more then 10 pp. Furthermore, following a GT, an individual is $27.6\%$ more likely to share the anonymized results with the health insurer if the individual already has an app from the insurer. Curiosity about one's genetic making is, overall, the strongest explanatory variable throughout all our models. Respondents who stated that curiosity for them would be an incentive to undergo genetic testing are on average $18\%$ more likely to undergo the test, disregarding the display of the price or the payer.
Subsequently, making use of framings in the design of our survey, we are able to shed light on the relationship between GTW along with the related data sharing and the nature of the payer of this GT, namely the individual itself or the health insurer. Our model is able to capture the critical importance of the payer in the decision process of undergoing the test and sharing anonymized genetic data. We provide empirical evidence of the impact of the health insurer as a payer on to GTW and DSW. Precisely, when the health insurer should be the payer, GTW and DSW increase by 24.8 and $9.2\%$, respectively.
The empirical results that this article provides are relevant for several streams of research. On the academic side, we lay the ground for a deeper understanding of the presence of a payer on health decisions as well as sharing of health-related data. We confirm findings from the extant body of literature on the relevance of number of factors influencing the GTW (cf. Section 2). As a novel result, for insurance, practitioners, we present the relevance of collaboration between clients and their insurance. With that in mind, an interesting avenue for further research may be, for example, how the amount of the insurance coverage of genetic testing influences preferences. However, while we believe that our set of variables is quite extensive, further uncaptured idiosyncratic characteristics may play a role in the decision process. For example, our study disregards ethical aspects, the mean of delivery of GT information, clinical counseling, and the limitations of GTs. Furthermore, privacy concerns are important in the context of personalized health [see Deruelle et al. [ 38]]. We conducted our research on survey-collected data, which intrinsically carries several biases. Self-reported data include flaws such as social desirability [see Gittelman et al. [ 39]] or health specific biases [documented in Bound et al. [ 40]]. Hence, results are to be taken with hindsight and a robustness test on another type of data (such as panel data, to get rid of confounding variables) could improve the results. In addition, the results we obtain are valid for Switzerland or countries under the same healthcare system. Extending this research to other models of healthcare would further increase the knowledge on health decisions.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements.
## Author contributions
VK performed the statistical analysis and wrote the first draft of the manuscript. All authors contributed to conception and design of the study, contributed to manuscript revision, read, and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpubh.2023.920286/full#supplementary-material
## References
1. Perkins BA, Caskey CT, Brar P, Dec E, Karow DS, Kahn AM. **Precision medicine screening using whole-genome sequencing and advanced imaging to identify disease risk in adults**. *Proc Natl Acad Sci USA* (2018) **115** 3686. DOI: 10.1073/pnas.1706096114
2. Jin J, Wu X, Yin J, Li M, Shen J, Li J. **Identification of genetic mutations in cancer: challenge and opportunity in the new era of targeted therapy**. *Front Oncol* (2019) **9** 263. DOI: 10.3389/fonc.2019.00263
3. Lima ZS, Ghadamzadeh M, Arashloo FT, Amjad G, Ebadi MR, Younesi L. **Recent advances of therapeutic targets based on the molecular signature in breast cancer: genetic mutations and implications for current treatment paradigms**. *J Hematol Oncol* (2019) **12** 38. DOI: 10.1186/s13045-019-0725-6
4. Su P. **Direct-to-consumer genetic testing: a comprehensive view**. *Yale J Biol Med* (2013) **86** 359-365. PMID: 24058310
5. Horton R, Crawford G, Freeman L, Fenwick A, Wright CF, Lucassen A. **Direct-to-consumer genetic testing**. *BMJ* (2019) **367** 15688. DOI: 10.1136/bmj.l5688
6. Ensenauer RE, Michels VV, Reinke SS. **Genetic testing: practical, ethical, and counseling considerations**. *Mayo Clinic Proc* (2005) **80** 63-73. DOI: 10.1016/S0025-6196(11)62960-1
7. Ordovas JM, Ferguson LR, Tai ES, Mathers JC. **Personalised nutrition and health**. *BMJ* (2018) **361** 2173. DOI: 10.1136/bmj.k2173
8. Verma M, Hontecillas R, Tubau-Juni N, Abedi V, Bassaganya-Riera J. **Challenges in personalized nutrition and health**. *Front Nutr* (2018) **5** 117. DOI: 10.3389/fnut.2018.00117
9. Horne J, Madill J, O'Connor C, Shelley J, Gilliland J. **A systematic review of genetic testing and lifestyle behaviour change: are we using high-quality genetic interventions and considering behaviour change theory?**. *Lifestyle Genomics* (2018) **11** 49-63. DOI: 10.1159/000488086
10. McGeoch L, Saunders CL, Griffin SJ, Emery JD, Walter FM, Thompson DJ. **Risk prediction models for colorectal cancer incorporating common genetic variants: a systematic review**. *Cancer Epidemiol Biomarkers Prev* (2019) **28** 1580. DOI: 10.1158/1055-9965.EPI-19-0059
11. Fogel AL, Jaju PD, Li S, Halpern-Felsher B, Tang JY, Sarin KY. **Factors influencing and modifying the decision to pursue genetic testing for skin cancer risk**. *J Am Acad Dermatol* (2017) **76** 829-835.e1. DOI: 10.1016/j.jaad.2016.11.050
12. Dalpe G, Feze IN, Salman S, Joly Y, Hagan J, Levesque E. **Breast cancer risk estimation and personal insurance: a qualitative study presenting perspectives from canadian patients and decision makers**. *Front Genet* (2017) **8** 128. DOI: 10.3389/fgene.2017.00128
13. Sweeny K, Ghane A, Legg AM, Huynh HP, Andrews SE. **Predictors of genetic testing decisions: a systematic review and critique of the literature**. *J Genet Counsel* (2014) **23** 263-88. DOI: 10.1007/s10897-014-9712-9
14. Armstrong K, Calzone K, Stopfer J, Fitzgerald G, Coyne J, Weber B. **Factors associated with decisions about clinical BRCA1/2 testing**. *Cancer Epidemiol Biomarkers Prev* (2000) **9** 1251-4. PMID: 11097234
15. Kopits IM, Chen C, Roberts JS, Uhlmann W, Green RC. **Willingness to pay for genetic testing for alzheimer's disease: a measure of personal utility**. *Genetic Testing Mol Biomarkers* (2011) **15** 871-5. DOI: 10.1089/gtmb.2011.0028
16. Lerman C, Marshall J, Audrain J, GomezCaminero A. **Genetic testing for colon cancer susceptibility: anticipated reactions of patients and challenges to providers**. *Int J Cancer* (1996) **69** 58-61. DOI: 10.1002/(SICI)1097-0215(19960220)69:1<58::AID-IJC15>3.0.CO;2-G
17. Hall MA, McEwen JE, Barton JC, Walker AP, Howe EG, Reiss JA. **Concerns in a primary care population about genetic discrimination by insurers**. *Genet Med* (2005) **7** 311-6. DOI: 10.1097/01.GIM.0000162874.58370.C0
18. Allain DC, Friedman S, Senter L. **Consumer awareness and attitudes about insurance discrimination post enactment of the genetic information nondiscrimination act**. *Familial Cancer* (2012) **11** 637-44. DOI: 10.1007/s10689-012-9564-0
19. Haga SB, Barry WT, Mills R, Ginsburg GS, Svetkey L, Sullivan J. **Public knowledge of and attitudes toward genetics and genetic testing**. *Genet Testing Mol Biomarkers* (2013) **17** 327-35. DOI: 10.1089/gtmb.2012.0350
20. Clayton EW, Halverson CM, Sathe NA, Malin BA. **A systematic literature review of individuals' perspectives on privacy and genetic information in the United States**. *PLoS ONE* (2018) **13** e0204417. DOI: 10.1371/journal.pone.0204417
21. Miron-Shatz T, Hanoch Y, Katz BA, Doniger GM, Ozanne EM. **Willingness to test for BRCA1/2 in high risk women: influenced by risk perception and family experience, rather than by objective or subjective numeracy?**. *Judgment Decis Making* (2015) **10** 15. DOI: 10.1017/S1930297500005180
22. Tubeuf S, Willis T, Potrata B, Grant H, Allsop M, Ahmed M. **Willingness to pay for genetic testing for inherited retinal disease**. *Eur J Human Genet* (2015) **23** 285-91. DOI: 10.1038/ejhg.2014.111
23. Wessel J, Gupta J, de Groot M. **Factors motivating individuals to consider genetic testing for type 2 diabetes risk prediction**. *PLoS ONE* (2016) **11** e0147071. DOI: 10.1371/journal.pone.0147071
24. Blouin-Bougie J, Amara N, Bouchard K, Simard J, Dorval M. **Disentangling the determinants of interest and willingness-to-pay for breast cancer susceptibility testing in the general population: a cross-sectional Web-based survey among women of Québec (Canada)**. *BMJ Open* (2018) **8** e016662. DOI: 10.1136/bmjopen-2017-016662
25. Abdul Rahim HF, Ismail SI, Hassan A, Fadl T, Khaled SM, Shockley B. **Willingness to participate in genome testing: a survey of public attitudes from Qatar**. *J Hum Genet* (2020) **65** 1067-73. DOI: 10.1038/s10038-020-0806-y
26. Sun S, Li S, Ngeow J. **Factors shaping at-risk individuals' decisions to undergo genetic testing for cancer in Asia**. *Health Soc Care Commun* (2020) **28** 1569-77. DOI: 10.1111/hsc.12981
27. Sanderson S, Wardle J, Jarvis M, Humphries S. **Public interest in genetic testing for susceptibility to heart disease and cancer: a population-based survey in the UK**. *Prev Med* (2004) **39** 458-64. DOI: 10.1016/j.ypmed.2004.04.051
28. Rosenstock IM. **Historical origins of the health belief model**. *Health Educ Monogr* (1974) **2** 328-35. DOI: 10.1177/109019817400200403
29. Gollust SE, Gordon ES, Zayac C, Griffin G, Christman MF, Pyeritz RE. **Motivations and perceptions of early adopters of personalized genomics: perspectives from research participants**. *Public Health Genomics* (2012) **15** 22-30. DOI: 10.1159/000327296
30. Kauffman TL, Irving SA, Leo MC, Gilmore MJ, Himes P, McMullen CK. **The NextGen study: patient motivation for participation in genome sequencing for carrier status**. *Mol Genet Genomic Med* (2017) **5** 508-15. DOI: 10.1002/mgg3.306
31. Alanazy MH, Alghsoon KA, Alkhodairi AF, Binkhonain FK, Alsehli TN, Altukhaim FF. **Public willingness to undergo presymptomatic genetic testing for Alzheimer's disease**. *Neurol Res Int* (2019) **2019** 2570513. DOI: 10.1155/2019/2570513
32. Smith K, Croyle R. **Attitudes toward genetic testing for colon-cancer risk**. *Am J Public Health* (1995) **85** 1435-8. DOI: 10.2105/AJPH.85.10.1435
33. Bosompra K, Flynn BS, Ashikaga T, Rairikar CJ, Worden JK, Solomon LJ. **Likelihood of undergoing genetic testing for cancer risk: a population-based study**. *Prev Med* (2000) **30** 155-66. DOI: 10.1006/pmed.1999.0610
34. Cameron LD, Sherman KA, Marteau TM, Brown PM. **Impact of genetic risk information and type of disease on perceived risk, anticipated affect, and expected consequences of genetic tests**. *Health Psychol* (2009) **28** 307-16. DOI: 10.1037/a0013947
35. Venables WN, Ripley BD. **Random and mixed effects**. *Modern Applied Statistics with S. Statistics and Computing* (2002) 271-300
36. Breiman L. **Random forests**. *Mach Learn* (2001) **45** 5-32. DOI: 10.1023/A:1010933404324
37. Schmitz H. **More health care utilization with more insurance coverage? Evidence from a latent class model with German data**. *Appl Econ* (2012) **44** 4455-68. DOI: 10.1080/00036846.2011.591733
38. Deruelle T, Kalouguina V, Trein P, Wagner J. **Designing privacy in personalized health: an empirical analysis**. *Bid Data Soc* (2023)
39. Gittelman S, Lange V, Cook WA, Frede SM, Lavrakas PJ, Pierce C. **Accounting for social-desirability bias in survey sampling**. *J Advert Res* (2015) **55** 242-54. DOI: 10.2501/JAR-2015-006
40. Bound J, Brown C, Mathiowetz N. **Chapter 59-measurement error in survey data**. *Handbook Econometr* (2001) **5** 3705-43. DOI: 10.1016/S1573-4412(01)05012-7
|
---
title: Angiotensin II type 2 receptor activation preserves megalin in the kidney and
prevents proteinuria in high salt diet fed rats
authors:
- Kalyani Kulkarni
- Sanket Patel
- Riyasat Ali
- Tahir Hussain
journal: Scientific Reports
year: 2023
pmcid: PMC10017765
doi: 10.1038/s41598-023-31454-6
license: CC BY 4.0
---
# Angiotensin II type 2 receptor activation preserves megalin in the kidney and prevents proteinuria in high salt diet fed rats
## Abstract
Proteinuria is a risk factor for and consequence of kidney injury. Angiotensin II type 2 receptor (AT2R) is an emerging reno-protective target and is anti-proteinuric under pathological conditions, including high salt-fed obese animals. However, the mechanisms remain unknown, particularly whether the anti-proteinuric activity of AT2R is independent of its anti-hypertensive and anti-inflammatory effects. In the present study, obese Zucker rats were fed high sodium ($4\%$) diet (HSD) for 48 h, a time in which blood pressure does not change. HSD caused proteinuria without affecting glomerular slit diaphragm proteins (nephrin and podocin), glomerular filtration rate, inflammatory and fibrotic markers (TNFα, IL-6, and TGF-β), ruling out glomerular injury, inflammation and fibrosis but indicating tubular mechanisms of proteinuria. At cellular and molecular levels, we observed a glycogen synthase kinase (GSK)-3β-mediated megalin phosphorylation, and its subsequent endocytosis and lysosomal degradation in HSD-fed rat kidneys. Megalin is a major proximal tubular endocytic protein transporter. The AT2R agonist C21 (0.3 mg/kg/day, i.p.) administration prevented proteinuria and rescued megalin surface expression potentially by activating Akt-mediated phosphorylation and inactivation of GSK-3β in HSD-fed rat kidneys. Overall, AT2R has a direct anti-proteinuric activity, potentially via megalin regulation, and is suggested as a novel target to limit kidney injury.
## Introduction
Proteinuria/albuminuria is a consequence and a risk factor for chronic kidney disease (CKD) under various pathological conditions, mainly hypertension and diabetes. Recently, angiotensin II type 2 receptor (AT2R), a component of the protective arm of the renin angiotensin system, has been reported as reno-protective, anti-proteinuric, anti-inflammatory, anti-fibrotic and anti-hypertensive, including in high salt diet (HSD)-fed animals1–11. Generally, these studies were chronic and the anti-proteinuric effects of AT2R were associated with a reduction in blood pressure. Specifically, our laboratory has reported that obese Zucker rats (OZR) when fed with HSD for 14 days exhibited proteinuria and tubulointerstitial injury, which were attenuated by AT2R activation1. Since high salt-intake induces inflammation, fibrosis and causes an increase in blood pressure, particularly in obesity112, these processes could be responsible for proteinuria and kidney injury. Considering the opposing effects of AT2R and high salt-intake, it is unknown whether anti-proteinuric effects were due to anti-hypertensive and anti-inflammatory effects of AT2R or a direct action related to AT2R activation in HSD fed animals. Therefore, present study is designed to elucidate the anti-proteinuric novel mechanism of AT2R activation independent of changes in blood pressure, inflammation, and fibrosis.
Tubular proteinuria is characterized as an impaired or reduced uptake of glomerular filtrate proteins by proximal tubular epithelial cells13–16. Megalin mediated endocytosis is an important function of proximal tubule epithelial cells to reabsorb proteins filtered through glomerulus. Megalin, a low-density lipoprotein receptor, is expressed abundantly on the apical surface of the proximal tubule epithelial cells and is a fast-recycling endocytic receptor. There is evidence that megalin is phosphorylated by GSK3β 17,18 and GSK3β -induced phosphorylation negatively regulates receptor recycling and reduces cell surface expression of the receptor18. Megalin is endocytosed after binding to filtered proteins and is recycled back to the plasma membrane after directing the bound proteins to lysosomal degradation17,19–23. Cubilin is another endocytic protein expressed on the plasma membrane24–26. Under normal physiological conditions, these receptors are responsible for tubular clearance of low and high molecular weight proteins and are recycled back to the plasma membrane25,26. Reduction in the functional activity of megalin/cubilin leads to proteinuria, which, if not resolved, can lead to inflammation, fibrosis, kidney injury culminating into cardiovascular diseases, in the long-term20,21. The AT2R is linked to Akt phosphorylation (activation)27 and GSK3β is inactivated upon phosphorylation by Akt28. Considering that GSK3β induces phosphorylation of megalin and negatively regulates its recycling reducing cell surface expression29,30, we hypothesize that AT2R-mediated activation of Akt reduces GSK3β activity and megalin phosphorylation, which in turn protects cell surface megalin expression and reduces proteinuria induced by high salt-intake. This hypothesis was tested in obese rats fed high sodium diet and administered with the AT2R novel agonist compound-21 (C21), which is a novel agonist and well-studied for its specificity and efficacy for various AT2R-mediated function, including human clinical studies27.
## Methods
The Institutional Animal Care and Use Committee at University of Houston approved these protocols.
## Ethical approval
All methods in this study are reported in accordance with the ARRIVE guidelines (https://arriveguidelines.org/) that maximize the quality and reliability of published research, and enabling others to better scrutinize, evaluate and reproduce it. Humans are not involved in this study. All the methods were carried out in accordance with relevant guidelines and regulations.
## Animals
Male OZR, 11–14 weeks old were purchased from Envigo, Indianapolis. The animals were acclimatized for a week at the University of Houston animal care facility upon arrival. In vivo experimental protocols used in this study were approved by the Institutional Animal Care and Use Committee at the University of Houston. The animals were fed with NSD ($0.4\%$; TekLad TD.99215, Harlan laboratories) or HSD ($4\%$, TekLad TD.92034, Harlan laboratories) and treated with AT2R agonist C21 (0.3 mg/kg/day i.p.) for 2 days. The specificity of C21 in vivo as well as in vitro studies in our laboratory has been tested by blocking its effects with the AT2R antagonist PD1233194,7. Rats were placed in metabolic cages during the study for urine collection. Body weight, food intake, water consumption and urine output were measured at 24 and 48 h. At the end of the study, blood was collected through cardiac puncture under isoflurane (2–$3\%$) anesthesia, processed for plasma, and stored at − 80 °C. Kidney cortices were collected and a part of it was embedded in OCT and stored at − 80 °C.
## Proteinuria and albuminuria
Urinary protein was measured by pyrogallol red (PR)-molybdate method. Briefly, to 5 μL of centrifuged urine sample, 200 μL PR-molybdate reagent was added and allowed to react for 10 min at 37 °C. Absorbance was read at 600 nm to measure total protein (mg/mL). Urinary albumin was determined by Nephrat II competitive ELISA kit (catalog# NR002 Ethos biosciences) according to manufacturer’s protocol. Urinary protein and albumin were normalized with urine volume (mL/hr) and reported as excretion rate in mg/hr.
## Immunoblotting
The expression of podocin, nephrin, pAkt and pGSK3β in the kidney cortices was determined by western blot analysis. Equal amount of protein (20 μg for podocin and nephrin, and 100 μg for pAkt and pGSK3β) was loaded at 4–$20\%$ SDS-PAGE, transferred to activated PVDF membrane, and immunoblotted with anti-podocin, anti-nephrin, anti-phospho-Akt (Ser-473) and anti-phospho-GSK3β (Ser9) respectively. β-Actin was used as loading control for podocin and nephrin and total Akt and total GSK3β were used to normalize pAkt and pGSK3β. For dot blot analysis, an equal amount of protein (10 μg) was directly spotted onto the activated PVDF membrane. The membrane was then incubated with specific anti-megalin or anti-cubilin antibody in $5\%$ BSA-PBST (phosphate buffered saline containing $0.05\%$ tween-20) overnight at 4 °C. The membrane was washed with PBST (5 mL, 10 min × 3), immunoprobed with relevant secondary HRP-conjugated antibodies, namely goat-anti-mouse IgG secondary antibody, goat-anti-rabbit IgG secondary antibody, for 1 h at room temperature, washed with PBST and the electrochemiluminescence signal was recorded and the bands density was analyzed (BioRad ChemiDoc MP Imaging System or Li-Cor Odyssey Fc Imager). The original blots are provided in SI Fig. S1-4.
## Separation of phospho-megalin, phospho-Akt and phospho-GSK3β
To determine the phosphorylated proteins, SuperSep Phos-tag (50 μmol/l), $7.5\%$, 17 well, 83 × 100 × 3.9 mm (FUJIFILM Wako Pure Chemical Corporation catalog# 198–17,981) was used. Phos-tag gel is a novel method which separates phospho-proteins based on migration and band shift which relies on complex formation ability. Phospho-proteins separate at a higher level compared with the non-phosphorylated form. Phos-tag allows to study phospho-proteins independent of phospho-specific antibody. Moreover, the stripping procedure is not required, hence, this is the method of choice to study phospho-proteins independent of loading control (e.g., GAPDH, beta-actin, etc.). The samples were prepared using RIPA lysis buffer without EDTA (150 mM NaCl, $1\%$ NP-40, $1\%$ sodium deoxycholate, Tris–HCl and $1\%$ SDS with protein phosphatase inhibitor) and samples were separated for 9 h. at 10 mA. Before transferring the gel on the PVDF membrane, the gel was washed with transfer buffer containing 10 mmol/L EDTA and 3 times for 10 min each. The gel was then immersed in transfer buffer without EDTA for 10 min. Transfer buffer without EDTA was used to transfer the proteins on the PVDF membrane. Immunoblotting with megalin antibody, Akt or GSK3β antibody was performed traditionally as explained earlier. The chemiluminescence signal was recorded and analyzed densitometrically by BioRad ChemiDoc MP Imaging System or Li-Cor Odyssey Fc Imager. The data is represented as the density ratio of the phospho- to non-phospho bands. The original blots are provided in SI Fig. S5. The validation of antibody by comparing the binding of boiled (vs. unboiled) antibody along with the specificity of secondary antibody is provided (SI Fig. S6).
## Immunofluorescence and colocalization
Approximately 20 μm thick sections were used for immunofluorescence experiment. The sections were incubated and permeabilized/blocked with $0.4\%$ BSA, $0.2\%$ saponin and $1\%$ of the animal serum (donkey) in which the secondary antibody is raised in 1X PBS for 1 h. at room temperature. This blocking buffer was discarded and 1X PBS containing the anti-megalin, anti-cubilin antibody and/or anti-LAMP1 in $0.2\%$ BSA and $0.1\%$ saponin was added to the sections and incubated overnight at 4 °C. The sections were washed with 1X PBS (10 min × 4) and secondary antibodies for megalin, cubilin, and/or LAMP1 in 1X PBS containing $0.2\%$ BSA and $0.1\%$ saponin was added and incubated for 2 h. at room temperature. The sections were washed (10 min × 4) with 1X PBS and incubated with DAPI (catalog# D1306, Thermo Fisher Scientific, 1:3000, 5 mg/mL stock) for 10 min followed by washing 3-times with PBS. The sections and coverslip were mounted on slides with glycerol and were imaged using Leica confocal microscope (DMi8). We have acquired images at the depth of 1 micron Z stacks (pinhole 1 AU, 63 × objective lens, HyD or PMT detector, pixel format/dimension 1024 × 1024; sequential scanning mode). Moreover, we have confirmed the membrane vs cytosolic fluorescence signal of megalin by XZ or YZ image planes (Fig. S7).
## Creatinine and GFR measurements
Urinary creatinine was measured using BioAssay systems kit (catalog# DICT500) according to the manufacturer’s protocol. The data was reported as mg/day. Plasma creatinine was measured by Arbor Assays kit (catalog# KB02-H1) according to the manufacturer’s protocol. The values were reported as mg/dL. The GFR was calculated creatinine clearance method.
## Atomic absorption spectroscopy
This method was used to measure urinary and plasma sodium. The standards and samples were prepared according to the company’s protocol and the data was calculated using Beer’s law.
## Quantitative RT-PCR analysis for mRNA expression
Total RNA from frozen kidneys was extracted using the RNAEasy kit (Qiagen) according to the manufacturer’s protocol. A total of 500 ng of RNA were reversed transcribed into cDNA using ReverTra Ace qPCR RT Master Mix with gDNA remover (Diagnocine). This cDNA was used to semi-quantitate cytokines (TNF-α, IL-10, TGF-β, megalin and cubilin) using Thunderbird SYBR qPCR master mix (Diagnocine) in CFX Connect RT-PCR (Bio-Rad). Specific quantitative PCR primers for TNF-α (catalog# RP300044), TGF-β (catalog# RP300111), and IL-6 (catalog# RP300072) were purchased from Sino Biological, and megalin (catalog# 316,614,765 F [Sequence: TGG AAT CTC CCT TGA TCC TG], catalog# 316,614,766 R [Sequence: TGT TGC TGC CAT CAG TCT TC]) and cubilin (catalog# 316,614,763 F [Sequence: GCA CTG GCA ATG AAC TAG CA], catalog# 316,614,764 R [Sequence: TGA TCC AGG AGC ACT CTG TG]) from Integrated DNA Technologies. Expression of each gene was normalized to β -actin (catalog# VRPS-97, Real Time Primers, LLC), and relative fold expression values were calculated using a DD threshold cycle method.
## Chemicals
Anti-podocin (Santa Cruz, catalog# sc-518088), anti-nephrin (Santa Cruz, catalog# sc-377246), anti-phospho-Akt (Ser-473) (Cell Signaling, catalog# 9271), anti-phospho-GSK3β (Ser9) (Cell Signaling, catalog# 9336), β-actin (Santa Cruz, catalog# sc-47778), total Akt (Cell Signaling, catalog# 9272), total GSK3β (Cell Signaling, catalog# 9315), anti-megalin (Santa Cruz, catalog# sc-515750), anti-cubilin antibody (Santa Cruz, catalog# sc-518059), anti-LAMP1 (Development Studies Hybridoma Bank, catalog# 1D4B), Alexa fluor 488 anti-mouse for megalin, Alexa fluor 488 anti-rabbit for cubilin, Alexa fluor 568 anti-rat for LAMP1.
## Statistical analysis
The data were analyzed using GraphPad Prism Version 9.1.2 [225]. Data are represented as mean ± sem. Statistical analysis was performed using one-way or two-way ANOVA with Fisher’s LSD for multiple comparisons and *$p \leq 0.05$ versus NSD and #$p \leq 0.05$ versus HSD considered statistically significant.
## Body weight
The body weight of animals among the study groups remain unchanged (NSD: 576 ± 23 g; HSD: 586 ± 22 g; HSD + C21: 596 ± 17 g).
## Renal function parameters
Urinary protein excretion was found to be increased by HSD feeding for 24 h. (HSD: 5.0 ± 0.6 mg/hr. vs. NSD: 2.2 ± 0.4 mg/hr.) and 48 h. (HSD: 6.1 ± 0.8 mg/hr. vs. NSD: 1.1 ± 0.8 mg/hr.). C21 treatment significantly reduced urinary protein excretion for both 24 and 48 h. respectively as compared to HSD feeding (HSD + C21: 3.2 ± 0.4 mg/hr. vs. HSD: 5 ± 0.6 mg/hr. at 24 h. and HSD + C21: 2.8 ± 0.4 mg/hr. vs. HSD: 6.1 ± 0.8 mg/hr. at 48 h.; Fig. 1a). However, urinary albumin excretion remained unchanged in HSD fed rats when compared to NSD group and C21 treatment also did not show any effect at both the time points (Fig. 1b). HSD feeding and C21 treatment did not cause any change in the plasma creatinine (Fig. 1c) and GFR in obese rats. ( Fig. 1d).Figure 1Effect of C21 treatment on renal function parameters of OZR fed with HSD or HSD + C21 for 48 h. Total urinary protein (proteinuria; a), total urinary albumin (albuminuria; b), plasma creatinine (c) and GFR (d). HSD and HSD + C21 treatment had no effect on plasma sodium concentration (e) however, HSD and HSD + C21 treatment groups exhibited a significant amount of sodium excretion in the urine as compared to NSD group (f). The values are represented as mean + sem; two-way ANOVA and one-way ANOVA followed by Fisher’s LSD test. * $p \leq 0.05$ versus NSD and #$p \leq 0.05$ versus HSD.
## Plasma and urinary Na+ concentration
Compared with that in the NSD group, the plasma Na+ concentration in the HSD and HSD + C21 group did not change, as expected (Fig. 1e) but HSD and HSD + C21 fed rats excreted significantly higher amount of urinary Na+ (Fig. 1f; HSD: 10.9 ± 1.7 mMol/day vs. NSD: 0.7 ± 0.2 mMol/day; HSD + C21: 9.29 ± 2.1 mMol/day vs. NSD: 0.7 ± 0.2 mMol/day), compared with NSD rats.
## Expression of endocytic receptors
Densitometric analysis of dot blots revealed that HSD feeding caused a decrease in megalin expression and this decrease was prevented by C21 treatment (Fig. 2a). Megalin mRNA expression as measured by qPCR did not change in either HSD or HSD + C21 rats as compared to NSD rats (Fig. 2b). Densitometric analysis of phostag gel (western blot) revealed that HSD group exhibited significant increase in megalin phosphorylation as compared to NSD treatment group. Moreover, megalin phosphorylation was significantly reduced by C21 treatment compared with HSD alone (Fig. 2c). However, protein expression as measured by dot blot or mRNA of cubilin remained similar in all the three groups (Fig. 3a, b, respectively).Figure 2Expression of megalin through dot blot (a), mRNA expression of megalin (b) and western blot of phosphorylated and unphosphorylated-megalin (c). Densitometry was normalized by ponceau (a) and the ratio of phospho to non-phospho megalin expression on phostag gel is shown in 2c. Lanes 10 and 11 from Fig. 2c has been removed. Values are represented as mean ± sem; and one-way ANOVA followed by Fisher’s LSD test respectively (b, c). * $p \leq 0.05$ versus NSD and #$p \leq 0.05$ versus HSD.Figure 3Expression of cubilin through dot blot (a) and mRNA expression of cubilin (b). Densitometry was normalized by ponceau (a). Values are represented as mean + sem; one-way ANOVA followed by Fisher’s LSD test respectively.
## Surface expression of megalin and cubilin through immunofluorescence
Confocal immunofluorescence microcopy revealed that in the NSD treated group, megalin was mostly present on the tubular cell surface (Fig. 4a). However, in the HSD-fed animals, megalin was found to be present mostly in the cytosolic compartment with negligible expression on the cell surface (Fig. 4b). Whereas treatment with C21 restored megalin localization back to the cell surface (Fig. 4c). Cubilin was found to be present on the tubular cell surface as well as in the cytosol of HSD treated rats whereas, in HSD + C21 treated and NSD fed rats, cubilin mostly was present on the cell surface of the tubules (Fig. 4d–f). Tissue permeabilization was validated using propidium iodide which is an impermeable nuclear dye in permeabilized kidney tissue (Fig. 4j) versus non-permeabilized kidney tissue (Fig. 4k). The tubular surface location was validated by phalloidin in permeabilized kidney tissue (Fig. 4l, n, p) versus non-permeabilized kidney tissue (Fig. 4m, o, q).Figure 4Representative images of cellular localization of megalin (a), cubilin (b) and megalin + LAMP1 (lysosomal marker) colocalization as determined by immunofluorescence in the kidney cortex. NSD- OZR fed with normal salt ($0.4\%$, a, d, g), HSD-OZR fed with high salt ($4\%$, b, e, h), and HSD + C21- OZR fed with high salt and treated with C21 (c, f, i). White arrows in (a, b, c) indicate megalin; (d, e, f) indicate cubilin and (g, h, i) indicate megalin + LAMP1. Kidney tissues (j, l, n, p) were permeabilized with saponin and tween 20 and (k, m, o, q) were not permeabilized. Kidney tissues (j and k) were stained with propidium iodide to demonstrate appropriate permeabilization. White arrows in panel “j” indicate nucleus. Images (l, n, p) indicate lumen (designated as “L” with white arrows) or apical membrane (red; we used phalloidin as the tubular plasma membrane marker) and megalin (green). All the images are of single plane as shown in figure S7. Scale 50 µm.
## Colocalization of megalin and lysosomal marker LAMP-1
LAMP-1 is a protein marker for lysosomes. Co-labeling of LAMP-1 and megalin revealed that megalin was co-localized with LAMP-1 in tubular cells of HSD-fed rats (Fig. 4h) and not in the NSD fed (Fig. 4g) or in HSD + C21 rats (Fig. 4i).
## Expression of glomerular injury markers
The expression of nephrin and podocin (Fig. 5a and b respectively) as measured by western blot revealed no significant difference in HSD or HSD + C21 groups as compared to NSD group. Figure 5Representative western blots of expression of glomerular injury markers nephrin (a) and podocin (b) and the blots are used from the respective gels. Densitometry of the bands was normalized with β-actin expression. Quantitative mRNA analysis of TNF-α, IL-6 and TGF-β (c–e). Values are represented as mean ± sem; one-way ANOVA followed by Fisher’s LSD test.
## Inflammatory and fibrotic markers
The inflammatory cytokines TNF-α (Fig. 5c), IL-6 (Fig. 5d) and the fibrotic marker TGF- β (Fig. 5e) were quantitated by measuring their mRNA and there was no difference in their levels between the NSD, HSD and HSD + C21 treatment groups.
## Akt/GSK3β activity
Akt phosphorylation as measured by p-S473-Akt antibody and Phostag western blotting with Akt antibody, was not affected in rats treated with HSD as compared to NSD group. Whereas Akt phosphorylation was significantly increased in HSD + C21 treatment group when compared to NSD and HSD groups (Fig. 6a and b). Similarly, GSK3β phosphorylation was measured by p-S9 GSK3β antibody and Phostag western blotting with GSK3β antibody. Both the methods revealed that HSD caused a modest decrease in the GSK3β phosphorylation, which was reversed in C21 + HSD group (Fig. 6c and d). However, the changes in GSK3β phosphorylation observed by p-S9 GSK3β antibody did not achieve statistical significance, although there was $44\%$ decrease by HSD compared with NSD and $166\%$ increase by C21 treatment. Figure 6Representative western blot and phos-tag gel images of pAkt (6a and 6b) respectively and pGSK3 β (6c and 6d) respectively. Densitometry of phospho-serine bands was normalized total Akt and total GSK3 β (6a and 6c) respectively, the ratio of phospho- to non-phospho-Akt and GSK3 β respectively (6b and 6d). Values are represented as mean ± sem; one-way ANOVA followed by Fisher’s LSD test. * $p \leq 0.05$ versus NSD and #$p \leq 0.05$ versus HSD.
## Discussion
This study investigates the early cellular and molecular mechanism of HSD-induced proteinuria and its protection by AT2R activation. Our data reveals that HSD intake causes proteinuria, which is associated with increased megalin phosphorylation and reduced cell surface expression in the kidney. Sub-cellular localization reveals the presence of megalin in the lysosomes suggesting that the reduced expression of megalin could be due to its degradation in the lysosomes. Also, non-phosphorylated (active) form of GSK3β which is a megalin phosphorylating enzyme is modestly decreased in HSD group. AT2R activation led to an increase in phosphorylated inactive form of GSK3β, reduction in megalin phosphorylation, prevention of lysosomal megalin localization, restoration of the surface megalin expression, and prevention of the onset of proteinuria. The AT2R activation enhances the Akt activity, which potentially might be responsible for reduction in GSK3β activity via its phosphorylation. The schematic model depicting the hypothesis and proposed mechanisms are provided in the Fig. 7.Figure 7The schematic model supporting hypothesis.
Obesity is generally believed to be salt-sensitive in terms of kidney dysfunction, cardiovascular diseases, and hypertension31,32. High salt intake is known to be pro-fibrotic, pro-inflammatory and pro-oxidative stress12,33,34. Contrary to the high salt intake, the AT2R activation is anti-fibrotic and anti-inflammatory2,5,35. So, it can be argued that HSD may have caused damage to the tubules affecting megalin function through these processes, which are counteracted by AT2R activation thus restoring megalin recycling and function and improving proteinuria. However, our data reveals that 2 days of HSD feeding does not affect inflammatory markers (TNF-α and IL-6) and fibrotic marker (TGF-β) in the kidney. Nephrin and podocin are glomerular slit diaphragm proteins and their loss indicates glomerular injury. In this study, nephrin and podocin expression is not altered, so this rules out the possibility of a potential glomerular damage by HSD at 48 h. This data is further supported by the observation that GFR remained the same. Overall, this data suggests that renal damage or injurious processes i.e., inflammation, and fibrosis are unlikely to be the mechanisms responsible for proteinuria and disruption in megalin recycling to the cell surface in this early period of HSD feeding and that the protective effect of the AT2R agonist treatment in restoring megalin recycling and preventing proteinuria is likely the result of direct molecular and cellular effects. However, inflammation, fibrosis and structural and functional injury will increase by HSD intake if continued for a longer period as reported in obese rats1 and in normal mice placed on HSD12 Specifically, in obese rats, HSD feeding over 2-weeks period caused decrease in GFR, glomerular and tubular injury, infiltration of immune cells and fibrosis12. Also in normal mice, HSD feeding over 7-days period led to glomerular injury associated with inflammation and fibrosis12.
Megalin is clustered in clathrin coated pits and is delivered to early endosomes to recycle back to the plasma membrane36,37. Alteration in this cycle causes megalin to fuse with lysosome for degradation29 thus, likely reduces megalin surface expression and protein transport leading to proteinuria. GSK3β is one of the kinases which at unphosphorylated state is active and has been suggested to phosphorylate megalin and impair its recycling and reducing its cell surface expression29. In the present study, HSD causes a modest increase in activity of GSK3β (reduced phosphorylation) which may be responsible for increased megalin phosphorylation. However, it’s not known as to what have caused a reduction in GSK3β phosphorylation upon HSD intake. It is likely that an increase in megalin phosphorylation via GSK3β impaired megalin recycling and its lysosomal degradation, as we observed in our study that megalin is localized with the lysosomal marker LAMP1. Since megalin mRNA remains unchanged in HSD group, this further supports the notion that it is the degradation, not the reduced synthesis, that leads to the overall decrease in megalin expression. The remaining megalin seems to be present mainly in the cytosol, not on the cell surface, which is necessary for its protein uptake/transport function. AT2R is known to activate Akt pathway38 and that Akt has a diverse function including phosphorylation of GSK3β leading to its inactivity30,39. In our study AT2R activation causes an increase in Akt phosphorylation. It is likely that AT2R activation reduced megalin phosphorylation via Akt/GSK3β pathway. This reduced phosphorylation may have prevented megalin trafficking toward lysosomes, and restored recycling process and megalin localization on the plasma membrane for its endocytic function. However, additional mechanisms, which are yet to be explored, may be involved in the impairment and restoration of megalin function in response to HSD and AT2R activation, respectively.
The reduced surface megalin expression can be a major mechanism of early increase in LMW proteinuria, while albuminuria, which makes < $10\%$ of the total proteinuria in our study, remained unchanged. Since albumin and other high molecular proteins (> 68 kDa) makes a fraction of the filtered proteins in the absence of glomerular damage, it is likely that large proteins are effectively reabsorbed by cubilin with the remaining megalin. However, escaping of proteins from proximal tubule reabsorption and the subsequent passage through nephron can cause inflammation and fibrosis in the long-term causing tubulointerstitial and cardiovascular diseases. Our study reports that AT2R prevents megalin disruption presenting this as an early mechanism of HSD-induced proteinuria in HSD-fed obese rats and provides a basis for long-term beneficial effects as reported in several pre-clinical models of kidney diseases1,40,41.
## Summary
Overall this study provided molecular mechanisms associated with high salt-induced proteinuria and its reversal by AT2R activation, particularly independent of hypertension, inflammation and fibrosis which themselves are risk factors of proteinuria and kidney injury. Specifically, this study suggest that increased activity of GSK3β in response to HSD as a mechanism responsible for megalin recycling disruption and reduced expression. The reversal of this molecular process by AT2R activation via Akt pathway presents a potential mechanims of reducing HSD-induced protenuria with potentially protecting kidney injury in the long-term as has been reported earlier.
## Supplementary Information
Supplementary Figures. The online version contains supplementary material available at 10.1038/s41598-023-31454-6.
## References
1. Patel SN, Ali Q, Hussain T. **Angiotensin II type 2-receptor agonist C21 reduces proteinuria and oxidative stress in kidney of high-salt-fed obese zucker rats**. *Hypertension* (2016.0) **67** 906-915. DOI: 10.1161/HYPERTENSIONAHA.115.06881
2. Ali Q, Patel S, Hussain T. **Angiotensin AT**. *Am. J. Physiol. Renal. Physiol.* (2015.0) **308** F1379-1385. DOI: 10.1152/ajprenal.00002.2015
3. Sabuhi R, Ali Q, Asghar M, Al-Zamily NR, Hussain T. **Role of the angiotensin II AT**. *Am. J. Physiol. Renal. Physiol.* (2011.0) **300** F700-706. DOI: 10.1152/ajprenal.00616.2010
4. Dhande I, Ali Q, Hussain T. **Proximal tubule angiotensin AT**. *Hypertension* (2013.0) **61** 1218-1226. DOI: 10.1161/HYPERTENSIONAHA.111.00422
5. Dhande I, Ma W, Hussain T. **Angiotensin AT**. *Hypertens. Res.* (2015.0) **38** 21-29. DOI: 10.1038/hr.2014.132
6. Ali Q, Dhande I, Samuel P, Hussain T. **Angiotensin type 2 receptor null mice express reduced levels of renal angiotensin II type 2 receptor/angiotensin (1–7)/Mas receptor and exhibit greater high-fat diet-induced kidney injury**. *J. Renin Angiotensin Aldosterone Syst.* (2016.0) **17** 1470320316661871. DOI: 10.1177/1470320316661871
7. 7.Patel, S., Dhande, I., Gray, E. A., Ali, Q. & Hussain, T. Prevention of lipopolysaccharide-induced CD11b(+) immune cell infiltration in the kidney: role of AT(2) receptors. Biosci. Rep.39 (2019).
8. Fatima N, Patel S, Hussain T. **Angiotensin AT**. *Front. Pharmacol.* (2021.0) **12** 600163. DOI: 10.3389/fphar.2021.600163
9. Abadir PM, Walston JD, Carey RM, Siragy HM. **Angiotensin II type-2 receptors modulate inflammation through signal transducer and activator of transcription proteins 3 phosphorylation and TNFα production**. *J. Interferon Cytokine Res.* (2011.0) **31** 471-474. DOI: 10.1089/jir.2010.0043
10. Kemp BA. **AT**. *Circ. Res.* (2014.0) **115** 388-399. DOI: 10.1161/circresaha.115.304110
11. Matavelli LC, Huang J, Siragy HM. **Angiotensin AT**. *Hypertension* (2011.0) **57** 308-313. DOI: 10.1161/hypertensionaha.110.164202
12. Teixeira DE. **A high salt diet induces tubular damage associated with a pro-inflammatory and pro-fibrotic response in a hypertension-independent manner**. *Biochim. Biophys. Acta Mol. Basis Dis.* (2020.0) **1866** 165907. DOI: 10.1016/j.bbadis.2020.165907
13. Ballermann BJ, Nystrom J, Haraldsson B. **The glomerular endothelium restricts albumin filtration**. *Front. Med. (Lausanne)* (2021.0) **8** 766689. DOI: 10.3389/fmed.2021.766689
14. D'Amico G, Bazzi C. **Pathophysiology of proteinuria**. *Kidney Int.* (2003.0) **63** 809-825. DOI: 10.1046/j.1523-1755.2003.00840.x
15. Miner JH. **Renal basement membrane components**. *Kidney Int.* (1999.0) **56** 2016-2024. DOI: 10.1046/j.1523-1755.1999.00785.x
16. Tojo A, Kinugasa S. **Mechanisms of glomerular albumin filtration and tubular reabsorption**. *Int. J. Nephrol.* (2012.0) **2012** 481520. DOI: 10.1155/2012/481520
17. Cabezas F. **Megalin/LRP2 expression is induced by peroxisome proliferator-activated receptor -alpha and -gamma: implications for PPARs' roles in renal function**. *PLoS One* (2011.0) **6** e16794. DOI: 10.1371/journal.pone.0016794
18. Surendran K, Vitiello SP, Pearce DA. **Lysosome dysfunction in the pathogenesis of kidney diseases**. *Pediatr. Nephrol.* (2014.0) **29** 2253-2261. DOI: 10.1007/s00467-013-2652-z
19. Alves SAS. **Surface megalin expression is a target to the inhibitory effect of bradykinin on the renal albumin endocytosis**. *Peptides* (2021.0) **146** 170646. DOI: 10.1016/j.peptides.2021.170646
20. Anand IS. **Proteinuria, chronic kidney disease, and the effect of an angiotensin receptor blocker in addition to an angiotensin-converting enzyme inhibitor in patients with moderate to severe heart failure**. *Circulation* (2009.0) **120** 1577-1584. DOI: 10.1161/circulationaha.109.853648
21. Currie G, Delles C. **Proteinuria and its relation to cardiovascular disease**. *Int. J. Nephrol. Renovasc. Dis.* (2013.0) **7** 13-24. DOI: 10.2147/ijnrd.S40522
22. Liu D. **Megalin/cubulin-lysosome-mediated albumin reabsorption is involved in the tubular cell activation of NLRP3 inflammasome and tubulointerstitial inflammation**. *J. Biol. Chem.* (2015.0) **290** 18018-18028. DOI: 10.1074/jbc.M115.662064
23. Nielsen R, Christensen EI, Birn H. **Megalin and cubilin in proximal tubule protein reabsorption: from experimental models to human disease**. *Kidney Int.* (2016.0) **89** 58-67. DOI: 10.1016/j.kint.2015.11.007
24. Coudroy G. **Contribution of cubilin and amnionless to processing and membrane targeting of cubilin-amnionless complex**. *J. Am. Soc. Nephrol.* (2005.0) **16** 2330-2337. DOI: 10.1681/asn.2004110925
25. Christensen EI, Birn H. **Megalin and cubilin: synergistic endocytic receptors in renal proximal tubule**. *Am. J. Physiol. Renal. Physiol.* (2001.0) **280** F562-573. DOI: 10.1152/ajprenal.2001.280.4.F562
26. Yammani RR, Seetharam S, Seetharam B. **Cubilin and megalin expression and their interaction in the rat intestine: effect of thyroidectomy**. *Am. J. Physiol. Endocrinol. Metabil.* (2001.0) **281** E900-907. DOI: 10.1152/ajpendo.2001.281.5.E900
27. Steckelings UM. **The angiotensin AT(2) receptor: from a binding site to a novel therapeutic target**. *Pharmacol. Rev.* (2022.0) **74** 1051-1135. DOI: 10.1124/pharmrev.120.000281
28. Ruvolo PP. **Phosphorylation of GSK3α/β correlates with activation of AKT and is prognostic for poor overall survival in acute myeloid leukemia patients**. *BBA Clin.* (2015.0) **4** 59-68. DOI: 10.1016/j.bbacli.2015.07.001
29. Yuseff MI, Farfan P, Bu G, Marzolo MP. **A cytoplasmic PPPSP motif determines megalin's phosphorylation and regulates receptor's recycling and surface expression**. *Traffic* (2007.0) **8** 1215-1230. DOI: 10.1111/j.1600-0854.2007.00601.x
30. Zhou X, Wang H, Burg MB, Ferraris JD. **Inhibitory phosphorylation of GSK-3β by AKT, PKA, and PI3K contributes to high NaCl-induced activation of the transcription factor NFAT5 (TonEBP/OREBP)**. *Am. J. Physiol. Renal. Physiol.* (2013.0) **304** F908-917. DOI: 10.1152/ajprenal.00591.2012
31. Aparicio A. **Estimation of salt intake assessed by urinary excretion of sodium over 24 h in Spanish subjects aged 7–11 years**. *Eur. J. Nutr.* (2017.0) **56** 171-178. DOI: 10.1007/s00394-015-1067-y
32. Wójcik M, Kozioł-Kozakowska A. **Obesity, sodium homeostasis, and arterial hypertension in children and adolescents**. *Nutrients* (2021.0) **13** 4032. DOI: 10.3390/nu13114032
33. Ferreira DN. **Salt-induced cardiac hypertrophy and interstitial fibrosis are due to a blood pressure-independent mechanism in Wistar rats**. *J. Nutr.* (2010.0) **140** 1742-1751. DOI: 10.3945/jn.109.117473
34. Hijmans RS. **High sodium diet converts renal proteoglycans into pro-inflammatory mediators in rats**. *PLoS One* (2017.0) **12** e0178940. DOI: 10.1371/journal.pone.0178940
35. Bhat SA, Sood A, Shukla R, Hanif K. **AT**. *Mol. Neurobiol.* (2019.0) **56** 3005-3023. DOI: 10.1007/s12035-018-1272-9
36. De S, Kuwahara S, Saito A. **The endocytic receptor megalin and its associated proteins in proximal tubule epithelial cells**. *Membranes (Basel)* (2014.0) **4** 333-355. DOI: 10.3390/membranes4030333
37. Ren Q. **Distinct functions of megalin and cubilin receptors in recovery of normal and nephrotic levels of filtered albumin**. *Am. J. Physiol. Renal. Physiol.* (2020.0) **318** F1284-F1294. DOI: 10.1152/ajprenal.00030.2020
38. Carrillo-Sepulveda MA. **Emerging role of angiotensin type 2 receptor (AT**. *PLoS One* (2013.0) **8** e61982. DOI: 10.1371/journal.pone.0061982
39. Fang X. **Phosphorylation and inactivation of glycogen synthase kinase 3 by protein kinase A**. *Proc. Natl. Acad. Sci. U. S. A.* (2000.0) **97** 11960-11965. DOI: 10.1073/pnas.220413597
40. Koulis C. **AT**. *Hypertension* (2015.0) **65** 1073-1081. DOI: 10.1161/HYPERTENSIONAHA.115.05204
41. Pandey A, Gaikwad AB. **AT(2) receptor agonist compound 21: a silver lining for diabetic nephropathy**. *Eur. J. Pharmacol.* (2017.0) **815** 251-257. DOI: 10.1016/j.ejphar.2017.09.036
|
---
title: Zinc and iron dynamics in human islet amyloid polypeptide-induced diabetes
mouse model
authors:
- Ayako Fukunaka
- Mari Shimura
- Takayuki Ichinose
- Ofejiro B. Pereye
- Yuko Nakagawa
- Yasuko Tamura
- Wakana Mizutani
- Ryota Inoue
- Takato Inoue
- Yuto Tanaka
- Takashi Sato
- Tatsuya Saitoh
- Toshiyuki Fukada
- Yuya Nishida
- Takeshi Miyatsuka
- Jun Shirakawa
- Hirotaka Watada
- Satoshi Matsuyama
- Yoshio Fujitani
journal: Scientific Reports
year: 2023
pmcid: PMC10017767
doi: 10.1038/s41598-023-30498-y
license: CC BY 4.0
---
# Zinc and iron dynamics in human islet amyloid polypeptide-induced diabetes mouse model
## Abstract
Metal homeostasis is tightly regulated in cells and organisms, and its disturbance is frequently observed in some diseases such as neurodegenerative diseases and metabolic disorders. Previous studies suggest that zinc and iron are necessary for the normal functions of pancreatic β cells. However, the distribution of elements in normal conditions and the pathophysiological significance of dysregulated elements in the islet in diabetic conditions have remained unclear. In this study, to investigate the dynamics of elements in the pancreatic islets of a diabetic mouse model expressing human islet amyloid polypeptide (hIAPP): hIAPP transgenic (hIAPP-Tg) mice, we performed imaging analysis of elements using synchrotron scanning X-ray fluorescence microscopy and quantitative analysis of elements using inductively coupled plasma mass spectrometry. We found that in the islets, zinc significantly decreased in the early stage of diabetes, while iron gradually decreased concurrently with the increase in blood glucose levels of hIAPP-Tg mice. Notably, when zinc and/or iron were decreased in the islets of hIAPP-Tg mice, dysregulation of glucose-stimulated mitochondrial respiration was observed. Our findings may contribute to clarifying the roles of zinc and iron in islet functions under pathophysiological diabetic conditions.
## Introduction
Approximately a third of the human proteome contains metal cations, either in the form of cofactors with catalytic functions or as structural support. To guarantee the proper maintenance of the homeostasis of these metals, cells and organisms have evolved highly sophisticated machinery involved in the transport, storage, and distribution of metals1. Studies have also shown that the imbalance of these metal ions in some tissues is closely associated with the onset and/or progression of various diseases. For example, patients with chronic hepatitis C frequently demonstrate iron overload in the serum and liver. In Alzheimer’s disease patients, imbalance of metal ions in the brain are frequently observed and are thought to be associated with Aβ deposition and tau hyperphosphorylation, a hallmark of Alzheimer’s disease2,3.
Pancreatic islets consist of 5 types of endocrine cells, including insulin producing β cells. Pancreatic β cells comprise approximately $80\%$ of the cells in the islets and are known to contain very high concentrations of zinc compared with other islet cells4. In particular, insulin secretory granules have been shown to have a high zinc content and are packed along with islet amyloid polypeptide (IAPP) within β cells5. Iron is also indispensable for β cells, considering the importance of mitochondrial function in β cells. It has been shown that iron is important for normal glucose-stimulated insulin secretion; however, excess iron causes oxidative stress and increases apoptosis in β cells6,7. Thus, zinc and iron play essential roles in pancreatic β-cell biology, and the dysfunction of their homeostasis has been reported to be implicated in the pathogenesis of type 2 diabetes4,6.
Whereas most previous studies on the basic pathogenic mechanism of diabetes have used mouse models, there are many differences between human and mouse diabetes. For instance, the structure and function of pancreatic islets are different between mouse models and humans8. One of the most intriguing differences between mouse diabetes models and human patients is that the deposition of islet amyloid is found in more than $90\%$ of human diabetes patients, but not in mouse models. This is owing to differences in their amino acid residues of IAPP (also known as amylin) between them9. Human and mouse IAPP share similar amino acid sequences in their N- and C-terminal regions, whereas their amino acids in the middle region substantially differ. Rodent IAPP has a proline substitution in the middle region, and hence does not form a β-sheet structure, and thus lacks the ability to self-oligomerize IAPP and form amyloid. To investigate the role of human IAPP (hIAPP), several mouse models have been developed, such as transgenic hIAPP overexpression models, and hIAPP knock-in mice10–13. These models clearly demonstrated that the expression of hIAPP induces toxic effects on β cells, likely owing to apoptosis and amyloidogenesis, although these mice models do not exhibit obesity. However, the molecular mechanisms by which hIAPP exerts cytotoxic effects have not yet been clarified.
In this study, we have investigated the dynamics of elements of islets derived from hIAPP transgenic (hIAPP-Tg) mice. We applied synchrotron X-ray scanning fluorescence microscopy (SXFM) and inductively coupled plasma mass spectrometry (ICP-MS) for imaging and quantitative analysis of elements, respectively. SXFM enables the mapping of multiple intracellular elements at the sub-organelle level by the combination of a synchrotron radiation source and a sub-100-nm X-ray beam focusing system14–20, and ICP-MS can quantify the concentrations of multiple elements with high sensitivity and precision (Fig. S1). Notably, zinc in the islets of hIAPP-Tg mice was decreased in the early stage of diabetes, whereas iron was reduced with the progression of hyperglycemia. We discuss the possible association between the decrease in these 2 essential elements and the progression of diabetes based on the hIAPP expression.
## hIAPP-Tg mice demonstrate aging-dependent hyperglycemia and amyloid deposition in their islets
To analyze the phenotype of hIAPP-Tg mice, blood glucose measurements and hematoxylin & eosin (HE) staining were performed using 5, 8, 12, 16, and 32-week-old in wild type (WT) and hIAPP-Tg mice. Non-fasting blood glucose levels were slightly increased in 5-week-old hIAPP-Tg mice compared with littermate WT mice, although the difference was not significant. hIAPP-Tg mice showed a significant increase in blood glucose level after 8-week-old (Fig. 1A). The islets of hIAPP-Tg mice appeared to have almost the same morphology as the islets of WT mice until 12 weeks of age. However, after 12 weeks, some islets of hIAPP-Tg mice became smaller and irregular in shape (Fig. 1B). Furthermore, insulin signals gradually decreased in hIAPP-Tg mice with aging (Fig. S2A), which is consistent with previous report10,21–23. Notably, a slightly eosinophilic appearance, which is the typical staining pattern of amyloid, was observed in 32-week-old hIAPP-Tg mice (Fig. S2B, arrows). The accumulation of amyloid was also detected in 32-week-old hIAPP-Tg mice by Thioflavin-T staining, which is a classical procedure to detect amyloid (Fig. S2B and C, arrows). These data confirm age-dependent hyperglycemia and deposition of islet amyloid in hIAPP-Tg mice, as previously reported10,11,13,21–23.Figure 1Phenotype of hIAPP-Tg mice. ( A) Non-fasting blood glucose levels of WT mice ($$n = 3$$–11) and hIAPP-Tg mice ($$n = 4$$–9). ( B) HE staining of pancreas sections from 5-, 8-, 12-, 16-, and 32-week-old hIAPP-Tg mice and WT mice. Data are shown as the mean ± SEM. ** $p \leq 0.01$ (WT vs. hIAPP). Scale bar, 100 μm.
## Iron and zinc dynamically decreased in the islets of aged hIAPP-Tg mice, a phenotype associated with diabetes
We performed SXFM imaging to observe the distribution of elements in the pancreatic islets. X-ray fluorescence energy spectrum showed that iron (X-ray emission lines: FeKα or FeKβ), zinc (ZnKα or ZnKβ), and bromine were substantially decreased in the islets of 32-week-old hIAPP-Tg mice compared with WT mice (black-colored lines, Fig. 2A), while all Compton and Elastic scattering X-rays, which influenced by the thickness of a sample basement, were quite repeatable (Fig. 2A). Bromine is abundantly contained in animal foods and bedding; however, even WT mice in our study showed variable bromine concentrations in their pancreatic tissue, resulting that bromine decrease shown in Fig. 2A was not repeated (Fig. S3A). We hence concluded that the differences in bromine were caused by individual differences. We focused on zinc and iron, as they are essential elements. Their mapping data showed a significant reduction at FeKα and ZnKα in the entire region of islets, whereas phosphorus and calcium mapping were comparable to those of WT mice (Fig. 2B). Another independent experiment using islets of 32-week-old hIAPP-Tg mice showed similar results, demonstrating the reproducibility of the results (Fig. S3). Similar results were also obtained from the islets of 16-week-old hIAPP-Tg mice (Fig. 2C), suggesting that the decrease in zinc and iron probably starts before 16 weeks of age in hIAPP-Tg mice (Fig. 2). Quantitative analysis of islets by ICP-MS was not possible because it is difficult to isolate islets from the pancreatic tissues hIAPP-Tg mice more than 16 week of age, for an unknown reason. Figure 2X-ray fluorescence images of islets from 16- and 32-week-old WT and hIAPP-Tg mice. ( A) X-ray fluorescence spectra of islet sections from 32-week-old mice. Arrows indicate peak signals of FeKα, FeKβ, ZnKα, and ZnKβ x-ray emission lines. Measurement was performed three times for each section. Gray line: a spectrum for a section from control mouse; black line: a spectrum for a section from hIAPP. X-ray energy, 15 keV; beam size, 500 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document}× 500 nm. ( B) Representative mapping images of (A). ( C) Mapping images of islets from indicated 16-week-old mice. WT: WT mice, hIAPP: hIAPP-Tg mice, Color bar, femtograms per square micrometer; DIC: differential interference contrast image; Scale bar, 20 μm.
## Zinc in hIAPP-Tg mouse islets decreased in the early stage of diabetes, while iron gradually decreased with the progression of hyperglycemia
To clarify whether zinc or iron dynamics were affected by the development of diabetes, we first analyzed 5-week-old hIAPP-Tg mice and found that their blood glucose levels were comparable to those of WT mice (Fig. 1A). We performed SXFM imaging on pancreatic islet tissues (Fig. 3A) and isolated islets from pancreatic tissues (Fig. 3B). Both mapping data suggested that zinc signals in the islets tended to be decreased in hIAPP-Tg mice, whereas iron signals was comparable to that of WT mice. The signal intensities of the mapping from isolated islets also suggested that zinc was significantly decreased in hIAPP-Tg mice ($$p \leq 0.0013$$), whereas iron was not ($$p \leq 0.324$$, Fig. 3C). To confirm these results quantitatively, we performed ICP-MS using isolated islets from pancreatic tissues. The results from the ICP-MS analysis supported the mapping data (Fig. 3D), suggesting that iron in the islets of hIAPP-Tg mice did not change, although zinc significantly decreased in the early stage of diabetes. Figure 3Decreased zinc in islets of 5-week-old hIAPP-Tg mice. ( A) Mapping images of pancreatic islets from 5-week-old mice. ( B) Mapping images of isolated islets from 5-week-old mice. Color bar, femtograms per square micrometer; DIC: differential interference contrast image; Scale bar, 20 μm. ( C) Quantification of signal intensities in (B). ( D) Relative amounts of each element in isolated islets were measured using ICP-MS ($$n = 4$$ for each genotype). WT: WT mice, hIAPP: hIAPP-Tg mice, Data are shown as means ± SEM. ** $p \leq 0.01$ (WT vs. hIAPP).
We next analyzed using pancreatic tissue and isolated islets of hIAPP-Tg mice and WT mice at 8 weeks of age, when blood glucose levels had started to be significantly increased without any obvious weight changes (Figs. 1A and S4). Mapping data from pancreatic tissue (Fig. 4A) and isolated islets from pancreatic tissues (Fig. 4B) of hIAPP-Tg mice showed that zinc decreased significantly ($$p \leq 0.0091$$), and iron showed a decreasing tendency ($$p \leq 0.058$$) in the islets of hIAPP-Tg mice (Fig. 4A–C). ICP-MS showed that zinc decreased significantly ($$p \leq 0.0004$$), whereas iron did not ($$p \leq 0.209$$) in the islets of hIAPP-Tg mice (Fig. 4D). These data suggest that zinc remains obviously decreased, while iron did not much change in the islets of 8-week-old hIAPP-Tg mice. Figure 4Iron level in islets is partially decreased in 8-week-old hIAPP-Tg mice. ( A) Mapping images of pancreatic islets from 8-week-old mice. ( B) Mapping images of isolated islets from 8-week-old mice. Color bar, femtograms per square micrometer; DIC: differential interference contrast image; Scale bar, 20 µm. ( C) Quantification of signal intensities in (B). ( D) Relative amounts of each element in isolated islets were measured using ICP-MS ($$n = 5$$,4 for each genotype). WT: WT mice, hIAPP: hIAPP-Tg mice, Data are shown as means ± SEM. ** $p \leq 0.01$ (WT vs. hIAPP).
We next analyzed 12-week-old hIAPP-Tg mice, in which significant blood glucose level changes were observed without any obvious weight changes (Figs. 1A and S4). We found a slight decrease in iron in the core region of 12-week-old hIAPP-Tg mouse islets, where β cells are localized8. On the other hand, zinc remained low in the entire islets (Fig. 5A). Data from ICP-MS showed that iron ($$p \leq 0.0087$$) and zinc ($$p \leq 0.0013$$) decreased significantly in the islets of hIAPP-Tg mice in diabetic conditions (Fig. 5B). Taken together, these data suggest that the decrease in zinc reflected a prediabetic or the early stages of diabetes. In contrast, iron in the islets of hIAPP-Tg mice gradually decreased concurrently with the development of diabetic phenotypes after 8–12 weeks of age. Figure 5Decreased iron in the islets of 12-week-old hIAPP-Tg mice. ( A) Mapping images of pancreatic islets from 12-week-old mice. Color bar, femtograms per square micrometer; DIC: differential interference contrast image; Scale bar, 20 µm. ( B) Relative amounts of each element in isolated islets were measured using ICP-MS ($$n = 4$$ for genotype). WT: WT mice, hIAPP: hIAPP-Tg mice, Data are given as means ± SEM. ** $p \leq 0.01$ (WT vs. hIAPP).
## The possible association between metal contents and cellular functions
To investigate the possible association between blood glucose levels and zinc and iron, we investigated the glucose-stimulated insulin secretion (GSIS) in pancreatic islet cells. GSIS from isolated islets was significantly reduced in 12-week-old hIAPP-Tg mice (Fig. 6A and B, right panel), while GSIS was not significantly changed in 5-week-old hIAPP-Tg mice compared to WT mice (Fig. 6A and B, left panel). Given that there was no difference in KCl-induced insulin secretion between the two groups, the deficiency of signaling events upstream of KATP channel closure appears to be responsible for the defects in GSIS. Considering the fact that the mitochondrion is a key machinery involved in the regulation of GSIS in β cells and studies have shown that disruption in zinc and iron homeostasis may seriously affect mitochondrial function, leading to an impaired energy state and susceptibility to disease development24,25, we investigated mitochondrial function in pancreatic islets cells. The analysis of mitochondrial respiration using a flux analyzer demonstrated that glucose-stimulated mitochondrial respiration (Acute Res) was significantly suppressed in the islets of 12-week-old hIAPP-Tg mice compared with WT mice (Fig. 7A and B, right panels). In contrast, there was no change in Acute Res of 5-week-old hIAPP-Tg mice when compared to WT mice (Fig. 7A and B, left panels). Maximal respiration (Max Res) induced by the addition of the uncoupler FCCP was upregulated in 5-week-old hIAPP-Tg mice, which might be due to the upregulation of compensation machinery for hIAPP toxicity (Fig. 7B, left panel). These results suggest that changes in both iron and/or zinc dynamics in pancreatic islets may affect Acute Res in the islets. Figure 6Glucose-stimulated insulin secretion (GSIS) is decreased in 12-week-old hIAPP-Tg mice. ( A) GSIS in 5 and 12-week-old indicated islets. Islets were incubated in KRB buffer containing 2.8 or 16.7 mM glucose, or 40 mM KCl for 60 min ($$n = 4$$ per group). ( B) Insulin content in 5 and 12-week-old mice islets ($$n = 4$$ per group). WT: WT mice, hIAPP: hIAPP-Tg mice, Data are shown as means ± SEM. * $p \leq 0.05$ (WT vs. hIAPP).Figure 7Mitochondrial function of 5- and 12-week-old hIAPP-Tg mice. ( A) Fold-increase in the oxygen consumption rate of mice ($$n = 4$$ for 5-week-old mice, $$n = 5$$ for 12-week-old mice). ( B) *Analyzed data* are shown as a graph. Non-mito: nonmitochondrial oxygen consumption; Basal Res: basal respiration, Max Res: maximal respiration, Proton Leak, ATP Pro: ATP production, Spare Res: spare respiratory capacity, Acute Res: acute response. Max Res and Acute Res are calculated by the equation as (maximum rate measurement after FCCP injection)—(non-mitochondrial respiration) and (late rate measurement before oligomycin injection)—(last rate measurement before acute injection), respectively. WT: WT mice, hIAPP: hIAPP-Tg mice, Data are shown as means ± SEM. * $p \leq 0.05$ (WT vs. hIAPP).
## Discussion
Previous studies have suggested that relatively high concentrations of zinc and iron were necessary for the normal function of pancreatic β cells4,7,26,27. Our data showed that a decrease in zinc and iron is associated with hyperglycemia in the islets of hIAPP-Tg mice, implicating an association between decreased zinc and/or iron in islets and the onset of human diabetes.
## Decrease in zinc in islets
Notably, imaging and quantitative analysis of metals in islets showed a decrease in zinc in the pancreatic islets of hIAPP-Tg mice in the early stage of diabetes. The question then arises as to whether the decrease in zinc is due to the expression of hIAPP. To clarify that, we transiently expressed hIAPP in the rat insulinoma INS-1 cell line and analyzed zinc and iron contents in the cells (Fig. S5). Zinc content in INS-1 cells expressing hIAPP was comparable to that in control cells, suggesting that the decreased zinc content was not due to only hIAPP expression. A decrease in zinc may require a long-term expression of hIAPP or/and its associated environmental conditions such as inflammation in vivo28. Notably, a significant decrease in zinc content in pancreatic tissues from various genetic mouse models of type 2 diabetes in the early stage of the disease, such as db/db mice (having a mutation in the leptin receptor) and ob/ob mice (having a mutation in the leptin gene) have been reported29,30, suggesting that the decrease in zinc is not dependent only on hIAPP expression. A recent study reported that the treatment of β cells from the islets of WT mice with inflammatory cytokines caused a significant reduction in zinc31. Therefore, a decrease in zinc may be associated with chronic inflammation prior to the onset of diabetes.
## Decrease in iron in islets
It has been reported that iron is increased in patients with type 2 diabetes, and therapeutic phlebotomy, a procedure that is used to reduce iron levels, was found to improve β-cell function in patients with pathological iron overload32,33. These findings suggest that there is an imbalance of iron metabolism in diabetic patients. However, the actual iron concentration within the islets was not reported. We observed that iron content in islets gradually decreased concurrently with the development of diabetic phenotype in hIAPP-Tg mice. Recently, mice lacking iron regulatory protein 2; a regulator of cellular iron homeostasis were reported to have a decreased iron content in β cells, and developed diabetes6, which is consistent with our result that a decrease in iron occurs concurrently with the onset of diabetes, although the mechanism of the decrease islet iron content remains unclear to date.
## A decrease in zinc and/or iron is associated with the dysfunction of mitochondria in β cells
It is important to understand how the decrease in zinc and/or iron alters the function of islets. Cellular iron deficiency results in reduced activity of Fe-S cluster-based complexes in the mitochondria, which is associated with impaired mitochondrial respiration34, and the restoration of intracellular iron levels can reverse these effects34,35. Zinc has been reported to restore impaired mitochondrial pyruvate transport, oxidative phosphorylation, and ultimate energy metabolism36. Our data showed an alteration in glucose-stimulated mitochondrial respiration following zinc and/or iron was reduced in hIAPP-Tg mice older than 12 weeks of age. This is consistent with a previous study which reported similar alterations in glucose-stimulated mitochondrial respiration in rats transgenic for hIAPP (HIP rats); an alteration associated with activation of the HIF1α/PFKFB3 signaling pathway leading to the disengagement of glycolysis from the mitochondrial TCA cycle24,37. The authors of this study suggested that this adaptive metabolic response induces β-cell dysfunction accompanied by a deficient response to glucose stimulation24. Notably, our data revealed that 12-week-old of hIAPP-Tg mice showed iron deficiency in islets, which might trigger the activation of the HIF1α signaling pathway and lead to increased glycolysis38.
## Study limitations and future perspectives
It remains unclear whether a decrease in zinc and/or iron in islets is the result of or the cause of diabetes. However, it is possible that a decrease in zinc and/or iron worsens the situation. In our study, hIAPP-Tg mice show a decrease in iron and zinc at 12 weeks old, and dysregulation in glucose-stimulated mitochondrial respiration that might lead to the dysfunction of mitochondrial function in β cells. As this study presents, we only identified the association between metal dynamics and the progressive dysregulation of islets, further analysis is needed to resolve the link between metal levels and islet functions using hIAPP-Tg mice or islets from diabetes patients. A combination of metal analysis with other omics studies will provide clarifications on the pathophysiological relevance of iron and zinc dynamics in diabetes39.
The interaction of metal ions with hIAPP may also affect its structure, causing the formation of misfolded IAPP, which can undergo oligomer and amyloid formation40,41. There are several in vitro studies reporting on the effects of metal ions on the formation of soluble hIAPP oligomers and mature amyloid42–47; however, there are limited information on this issue regarding the physiological conditions in studies using mouse models or human samples. SXFM will become a useful tool to visualize the interaction between metals and hIAPP-derived amyloid formation. Further analysis is needed in the future to clarify the association between metals and hIAPP amyloid formation.
## Mouse experiments
All mice were housed in specific pathogen-free barrier facilities, maintained under a 12-h light/dark cycle, given water ad libitum, and fed with standard rodent chow (Oriental Yeast, Tokyo, Japan). Blood glucose levels were measured using a glucose analyzer (Glutest Mint, Sanwa Chemical Co., Nagoya, Japan). hIAPP-Tg mice were obtained from Jackson Laboratories (Strain No. 008232) and mice were backcrossed with C57BL/6 J mice for more than seven generations. Mice were euthanized by anesthesia with isoflurane.
## Pancreatic islet culture
Mouse islet isolation was performed as described previously4,12,48. Isolated islets were cultured in RPMI 1640 medium supplemented with $10\%$ fetal bovine serum (FBS) and $1\%$ penicillin/streptomycin.
## GSIS
GSIS from mouse islets was investigated as described previously48. Briefly, eight size-matched islets were incubated in HRKB buffer containing glucose or KCl for 60 min. Insulin concentration of supernatants of isolated islets was analyzed by mouse insulin enzyme-linked immunosorbent assay kit (Morinaga Institute of Biological Science, Yokohama, Japan). Insulin secretion data were corrected by the insulin content of the islets.
## SXFM
Differential interference contrast images of each tissue section were obtained prior to SXFM measurements using an Olympus DP22 microscope. SXFM was performed using the undulator beamline BL29XU at the SPring-8 synchrotron radiation facility by combining a Kirkpatrick-Baez type X-ray focusing system, xy scanning stage for scanning sample mounting, and an energy dispersive X-ray detector (Vortex-90EX; Hitachi High-Technologies Science America, Inc., Northridge, CA, USA)15. Monochromatic X-rays at 15 keV were focused down to 500 × 500 nm2 for a large-area scan. A typical photon flux for the 500-nm beam is approximately 2 × 1011 photons/s. X-ray fluorescence spectra were recorded using a 1-s exposure for each pixel. When the region of interest was relatively wide, it was separated into several parts for the measurement. The fluorescence signals of each element of interest were extracted and normalized based on the incident beam intensity. After scanning the whole area, the distributions of various elements were visualized digitally. SXFM produced superimposed signals from the samples in the vertical direction. In addition to the mapping images, the element concentration per area (µm2) was analyzed quantitatively using thin nickel and platinum films, the thickness and density of which were decided in advance. Signal intensities per interested region in the TIFF images were acquired using ImageJ software (National Institutes of Health, Bethesda, MD, USA).
## ICP-MS
For isolated islet samples of small amounts (from 5, 8, and 12-week-old-mice), 0.5 mL of nitric acid solution in a PFA jar (ARAM Co., Osaka, Japan) was used. The jar was heated at 100 °C for 20 min using an ETHOS 1 microwave oven (Milestone Srl, Sorisole, Italy). Concentrations of Ca, Fe, P, and Zn were determined using ICP-MS (ELEMENT XR, Thermo Fisher Scientific Inc, Bremen, Germany) with resolution R-10,000 for 44Ca, and R-4,000 for 31P, 56Fe, and 66Zn. The INS-1 cells were digested with 0.5 mL of nitric acid (Tamapure-AA-100, Tama Chemical Co. Ltd., Kanagawa, Japan) at 180 °C for 20 min in an ETHOS 1 microwave oven (Milestone Srl, Sorisole, Italy) and then diluted with ultrapure water (manufactured by PURELAB Option-R 7 and PURELAB flex UV, Veolia Water Solutions and Technologies, Paris, France) to a 5-mL volume. Concentrations of P, Ca, Fe, and Zn were determined using ICP-MS (ELEMENT XR, Thermo Fisher Scientific Inc., Bremen, Germany) with resolution R-10,000 for 44Ca, and R-4,000 for 31P, 56Fe, and 66Zn.
## Mitochondrial respiration in islets
Mitochondrial respiration in mouse islets was measured as previously described49. Briefly, isolated islets (20 islets/well) from hIAPP-Tg mice and WT mice were incubated in RPMI 1640 medium containing 5.6 mM glucose, 1 mM pyruvate, and $10\%$ FBS. The islets were washed with PBS and seeded into poly-L-lysine coated XF96 cell culture microplates (Agilent Technologies, Palo Alto, CA). The culture microplates were then centrifuged at 500 rpm for 7 min at room temperature and incubated for 1 to 2 h at 37 °C in a non-CO2 incubator. XF RPMI Medium (Agilent Technologies) containing 5.6 mM glucose, 1 mmol/L pyruvate, and 2 mM L-glutamine was used as the assay medium. The oxygen consumption rate and extracellular acidification rate were measured using a Seahorse SX96 Analyzer (Agilent Technologies). Islets were sequentially exposed to glucose (11.1 mM), oligomycin (4 µM), carbonyl cyanide 4-phenylhydrazone (FCCP) (10 µM), and rotenone/antimycin A (2.5 µM). Wave 2.6.0 software (Agilent Technologies) was used to analyze non-mitochondrial oxygen consumption, basal respiration, maximal respiration, proton leak, ATP production, spare respiratory capacity, and acute response. Non-mitochondrial oxygen consumption is calculated by the equation as minimum rate measurement after rotenone/antimycin A injection. Basal respiration is calculated by the equation as (last rate measurement before the first injection)—(non-mitochondrial respiration rate). Maximal respiration is calculated by the equation as (maximum rate measurement after FCCP injection)—(non-mitochondrial respiration). Proton leak is calculated by the equation as (minimum rate measurement after oligomycin injection)—(non-mitochondrial respiration). ATP production is calculated by the equation as (last rate measurement before oligomycin injection)—(minimum rate measurement after oligomycin injection). Spare respiratory capacity is calculated by the equation as (maximal respiration)—(basal respiration). Acute response is calculated by the equation as (last rate measurement before oligomycin injection)—(last rate measurement before acute injection).
## Statistical analysis
All quantitative data were reported as the mean ± SEM. Student t-test was performed for comparison between groups. Welch’s t-test was performed in the experiments of Figs. 3C, 4C. $p \leq 0.05$ was considered a significant difference between control an experimental group.
## Ethical approval
All experimental animal care was performed in accordance with institutional and national guidelines and regulations. The study protocol was approved by the Institutional Animal Care and Use Committee of Gunma University (Permit #19–025). The study is reported in accordance with ARRIVE guidelines.
## Supplementary Information
Supplementary Information 1.Supplementary Information 2.Supplementary Information 3. The online version contains supplementary material available at 10.1038/s41598-023-30498-y.
## References
1. McRae R, Bagchi P, Sumalekshmy S, Fahrni CJ. **In situ imaging of metals in cells and tissues**. *Chem. Rev.* (2009) **109** 4780-4827. DOI: 10.1021/cr900223a
2. Lei P, Ayton S, Bush AI. **The essential elements of Alzheimer's disease**. *J. Biol. Chem.* (2021) **296** 100105. DOI: 10.1074/jbc.REV120.008207
3. Feld JJ, Liang TJ. **Hepatitis C – identifying patients with progressive liver injury**. *Hepatology* (2006) **43** S194-206. DOI: 10.1002/hep.21065
4. Tamaki M. **The diabetes-susceptible gene SLC30A8/ZnT8 regulates hepatic insulin clearance**. *J. Clin. Invest.* (2013) **123** 4513-4524. DOI: 10.1172/jci68807
5. Fukunaka A, Fujitani Y. **Role of zinc homeostasis in the pathogenesis of diabetes and obesity**. *Int. J. Mol. Sci.* (2018). DOI: 10.3390/ijms19020476
6. Santos M. **Irp2 regulates insulin production through iron-mediated Cdkal1-catalyzed tRNA modification**. *Nat. Commun.* (2020) **11** 296. DOI: 10.1038/s41467-019-14004-5
7. Berthault C, Staels W, Scharfmann R. **Purification of pancreatic endocrine subsets reveals increased iron metabolism in beta cells**. *Mol. Metab.* (2020) **42** 101060. DOI: 10.1016/j.molmet.2020.101060
8. Campbell JE, Newgard CB. **Mechanisms controlling pancreatic islet cell function in insulin secretion**. *Nat. Rev. Mol. Cell. Biol.* (2021) **22** 142-158. DOI: 10.1038/s41580-020-00317-7
9. Haataja L, Gurlo T, Huang CJ, Butler PC. **Islet amyloid in type 2 diabetes, and the toxic oligomer hypothesis**. *Endocr. Rev.* (2008) **29** 303-316. DOI: 10.1210/er.2007-0037
10. Janson J. **Spontaneous diabetes mellitus in transgenic mice expressing human islet amyloid polypeptide**. *Proc. Natl. Acad. Sci. U. S. A.* (1996) **93** 7283-7288. DOI: 10.1073/pnas.93.14.7283
11. Kim J. **Amyloidogenic peptide oligomer accumulation in autophagy-deficient β cells induces diabetes**. *J. Clin. Invest.* (2014) **124** 3311-3324. DOI: 10.1172/jci69625
12. Shigihara N. **Human IAPP-induced pancreatic β cell toxicity and its regulation by autophagy**. *J. Clin. Invest.* (2014) **124** 3634-3644. DOI: 10.1172/jci69866
13. Butler AE. **Diabetes due to a progressive defect in beta-cell mass in rats transgenic for human islet amyloid polypeptide (HIP Rat): A new model for type 2 diabetes**. *Diabetes* (2004) **53** 1509-1516. DOI: 10.2337/diabetes.53.6.1509
14. Shimura M. **Element array by scanning X-ray fluorescence microscopy after cis-diamminedichloro-platinum(II) treatment**. *Cancer Res.* (2005) **65** 4998-5002. DOI: 10.1158/0008-5472.Can-05-0373
15. Shimura M. **Imaging of intracellular fatty acids by scanning X-ray fluorescence microscopy**. *Faseb J.* (2016) **30** 4149-4158. DOI: 10.1096/fj.201600569R
16. Matsuyama S, Maeshima K, Shimura M. **Development of X-ray imaging of intracellular elements and structure**. *J. Anal. Atom. Spectrom.* (2020) **35** 1279-1294. DOI: 10.1039/d0ja00128g
17. Szyrwiel Ł. **A novel branched TAT(47–57) peptide for selective Ni(2+) introduction into the human fibrosarcoma cell nucleus**. *Metallomics* (2015) **7** 1155-1162. DOI: 10.1039/c5mt00021a
18. Imai R. **Chromatin folding and DNA replication inhibition mediated by a highly antitumor-active tetrazolato-bridged dinuclear platinum(II) complex**. *Sci. Rep.* (2016) **6** 24712. DOI: 10.1038/srep24712
19. Tanaka YK. **Formation mechanism and toxicological significance of biogenic mercury selenide nanoparticles in human hepatoma HepG2 cells**. *Chem. Res. Toxicol.* (2021) **34** 2471-2484. DOI: 10.1021/acs.chemrestox.1c00231
20. Matsuyama S. **Scanning protein analysis of electrofocusing gels using X-ray fluorescence**. *Metallomics* (2013) **5** 492-500. DOI: 10.1039/c3mt20266f
21. Hull RL. **Genetic background determines the extent of islet amyloid formation in human islet amyloid polypeptide transgenic mice**. *Am. J. Physiol. Endocrinol. Metab.* (2005) **289** E703-709. DOI: 10.1152/ajpendo.00471.2004
22. Mukherjee A. **Induction of IAPP amyloid deposition and associated diabetic abnormalities by a prion-like mechanism**. *J. Exp. Med.* (2017) **214** 2591-2610. DOI: 10.1084/jem.20161134
23. Kim J. **An autophagy enhancer ameliorates diabetes of human IAPP-transgenic mice through clearance of amyloidogenic oligomer**. *Nat. Commun.* (2021) **12** 183. DOI: 10.1038/s41467-020-20454-z
24. Montemurro C. **IAPP toxicity activates HIF1α/PFKFB3 signaling delaying β-cell loss at the expense of β-cell function**. *Nat. Commun.* (2019) **10** 2679. DOI: 10.1038/s41467-019-10444-1
25. Weiss A. **Zn-regulated GTPase metalloprotein activator 1 modulates vertebrate zinc homeostasis**. *Cell* (2022) **185** 2148-2163.e2127. DOI: 10.1016/j.cell.2022.04.011
26. Rutter GA, Chimienti F. **SLC30A8 mutations in type 2 diabetes**. *Diabetologia* (2015) **58** 31-36. DOI: 10.1007/s00125-014-3405-7
27. Dwivedi OP. **Loss of ZnT8 function protects against diabetes by enhanced insulin secretion**. *Nat. Genet.* (2019) **51** 1596-1606. DOI: 10.1038/s41588-019-0513-9
28. Westwell-Roper CY. **IL-1 mediates amyloid-associated islet dysfunction and inflammation in human islet amyloid polypeptide transgenic mice**. *Diabetologia* (2015) **58** 575-585. DOI: 10.1007/s00125-014-3447-x
29. Simon SF, Taylor CG. **Dietary zinc supplementation attenuates hyperglycemia in db/db mice**. *Exp. Biol. Med.* (2001) **226** 43-51. DOI: 10.1177/153537020122600107
30. Begin-Heick N, Dalpe-Scott M, Rowe J, Heick HM. **Zinc supplementation attenuates insulin secretory activity in pancreatic islets of the ob/ob mouse**. *Diabetes* (1985) **34** 179-184. DOI: 10.2337/diab.34.2.179
31. Slepchenko KG. **Synchrotron fluorescence imaging of individual mouse beta-cells reveals changes in zinc, calcium, and iron in a model of low-grade inflammation**. *Metallomics* (2021). DOI: 10.1093/mtomcs/mfab051
32. Creighton Mitchell T, McClain DA. **Diabetes and hemochromatosis**. *Curr. Diab. Rep.* (2014) **14** 488. DOI: 10.1007/s11892-014-0488-y
33. Harrison AV, Lorenzo FR, McClain DA. **Iron and the pathophysiology of diabetes**. *Annu. Rev. Physiol.* (2022). DOI: 10.1146/annurev-physiol-022522-102832
34. Rensvold JW. **Complementary RNA and protein profiling identifies iron as a key regulator of mitochondrial biogenesis**. *Cell Rep.* (2013) **3** 237-245. DOI: 10.1016/j.celrep.2012.11.029
35. Hoes MF. **Iron deficiency impairs contractility of human cardiomyocytes through decreased mitochondrial function**. *Eur. J. Heart Fail.* (2018) **20** 910-919. DOI: 10.1002/ejhf.1154
36. Yang X. **Zinc enhances the cellular energy supply to improve cell motility and restore impaired energetic metabolism in a toxic environment induced by OTA**. *Sci. Rep.* (2017) **7** 14669. DOI: 10.1038/s41598-017-14868-x
37. Nomoto H. **Activation of the HIF1α/PFKFB3 stress response pathway in beta cells in type 1 diabetes**. *Diabetologia* (2020) **63** 149-161. DOI: 10.1007/s00125-019-05030-5
38. Bianchi L, Tacchini L, Cairo G. **HIF-1-mediated activation of transferrin receptor gene transcription by iron chelation**. *Nucleic Acids Res.* (1999) **27** 4223-4227. DOI: 10.1093/nar/27.21.4223
39. Morel JD. **The mouse metallomic landscape of aging and metabolism**. *Nat. Commun.* (2022) **13** 607. DOI: 10.1038/s41467-022-28060-x
40. Milardi D. **Proteostasis of Islet amyloid polypeptide: A molecular perspective of risk factors and protective strategies for type II diabetes**. *Chem. Rev.* (2021) **121** 1845-1893. DOI: 10.1021/acs.chemrev.0c00981
41. Bhowmick DC, Kudaibergenova Z, Burnett L, Jeremic AM. **Molecular mechanisms of amylin turnover, misfolding and toxicity in the pancreas**. *Molecules* (2022). DOI: 10.3390/molecules27031021
42. Brender JR. **Role of zinc in human islet amyloid polypeptide aggregation**. *J. Am. Chem. Soc.* (2010) **132** 8973-8983. DOI: 10.1021/ja1007867
43. Sciacca MF. **Cations as switches of amyloid-mediated membrane disruption mechanisms: Calcium and IAPP**. *Biophys. J.* (2013) **104** 173-184. DOI: 10.1016/j.bpj.2012.11.3811
44. Salamekh S. **A two-site mechanism for the inhibition of IAPP amyloidogenesis by zinc**. *J. Mol. Biol.* (2011) **410** 294-306. DOI: 10.1016/j.jmb.2011.05.015
45. Brender JR, Salamekh S, Ramamoorthy A. **Membrane disruption and early events in the aggregation of the diabetes related peptide IAPP from a molecular perspective**. *Acc. Chem. Res.* (2012) **45** 454-462. DOI: 10.1021/ar200189b
46. Brender JR. **Zinc stabilization of prefibrillar oligomers of human islet amyloid polypeptide**. *Chem. Commun.* (2013) **49** 3339-3341. DOI: 10.1039/c3cc40383a
47. Khemtemourian L. **Investigation of the effects of two major secretory granules components, insulin and zinc, on human-IAPP amyloid aggregation and membrane damage**. *Chem. Phys. Lipids* (2021) **237** 105083. DOI: 10.1016/j.chemphyslip.2021.105083
48. Ebato C. **Autophagy is important in islet homeostasis and compensatory increase of beta cell mass in response to high-fat diet**. *Cell Metab.* (2008) **8** 325-332. DOI: 10.1016/j.cmet.2008.08.009
49. Inoue R. **Uncoupling protein 2 and aldolase B impact insulin release by modulating mitochondrial function and Ca(2+) release from the ER**. *iScience* (2022) **25** 104603. DOI: 10.1016/j.isci.2022.104603
|
---
title: Dietary protection against the visual and motor deficits induced by experimental
autoimmune encephalomyelitis
authors:
- Katarzyna Zyla-Jackson
- Dorothy A. Walton
- Kendra S. Plafker
- Susan Kovats
- Constantin Georgescu
- Richard S. Brush
- Madison Tytanic
- Martin-Paul Agbaga
- Scott M. Plafker
journal: Frontiers in Neurology
year: 2023
pmcid: PMC10017782
doi: 10.3389/fneur.2023.1113954
license: CC BY 4.0
---
# Dietary protection against the visual and motor deficits induced by experimental autoimmune encephalomyelitis
## Abstract
### Introduction
Five to eight percent of the world population currently suffers from at least one autoimmune disorder. Despite multiple immune modulatory therapies for autoimmune demyelinating diseases of the central nervous system, these treatments can be limiting for subsets of patients due to adverse effects and expense. To circumvent these barriers, we investigated a nutritional intervention in mice undergoing experimental autoimmune encephalomyelitis (EAE), a model of autoimmune-mediated demyelination that induces visual and motor pathologies similar to those experienced by people with multiple sclerosis (MS).
### Methods
EAE was induced in female and male mice and the impact of limiting dietary carbohydrates by feeding a ketogenic diet (KD) enriched in medium chain triglycerides (MCTs), alpha-linolenic acid (an omega-3 fatty acid), and fiber was evaluated in both a preventive regimen (prior to immunization with MOG antigen) and an interventional regimen (following the onset of symptoms). Motor scores were assigned daily and visual acuity was measured using optokinetic tracking. Immunohistochemical analyses of optic nerves were done to assess inflammatory infiltrates and myelination status. Fatty acid and cytokine profiling from blood were performed to evaluate systemic inflammatory status.
### Results
The KD was efficacious when fed as a preventive regimen as well as when initiated as an interventional regimen following symptom onset. The KD minimally impacted body weight during the experimental time course, increased circulating ketones, prevented motor and ocular deficits, preserved myelination of the optic nerve, and reduced infiltration of immune cells to optic nerves. The KD also increased anti-inflammatory-associated omega-3 fatty acids in the plasma and reduced select cytokines in the circulation associated with EAE-mediated pathological inflammation.
### Discussion
In light of ongoing clinical trials using dietary strategies to treat people with MS, these findings support that a KD enriched in MCTs, omega-3 fatty acids, and fiber promotes a systemic anti-inflammatory milieu and ameliorates autoimmune-induced demyelinating visual and motor deficits.
## Introduction
Multiple sclerosis (MS) is an autoimmune demyelinating disease of the central nervous system (CNS) that causes severe disabilities. Deficits include losses of mobility, balance and coordination, blindness, depression, fatigue, memory loss, and decreased quality of life. MS is more common in females, with onset typically occurring in the third and fourth decades of life [reviewed in (1–3)].
The increased incidence of MS and other autoimmune disorders in recent decades coincides with global increases in obesity, hyperglycemia, hyperinsulinemia, insulin resistance, dyslipidemia, and type 2 diabetes [4]. Mounting evidence points to the excessive consumption of ultra-processed foods containing highly-processed carbohydrates and pro-inflammatory fats as drivers of metabolic syndrome (5–7), poor health outcomes [8, 9], and all-cause mortality [10]. These ultra-processed, hyperpalatable, calorically-dense foodstuffs promote hyperphagia [11], and can exacerbate autoimmunity by disrupting microbiome-host symbiosis and promoting systemic inflammation [12]. Evidence from clinical trials and animal studies support that diet profoundly impacts MS severity and disease trajectory (13–15), consistent with the observation that insulin resistance and adiposity correlate with more severe disability scores for people with MS [16, 17].
To determine the impact of reducing carbohydrate-laden foods in the diet, we investigated the therapeutic efficacy of a ketogenic diet (KD) in the mouse MOG35−55-experimental autoimmune encephalomyelitis (MOG35−55-EAE) model, hereafter referred to as “EAE.” This model of autoimmune-mediated demyelination induces visual and motor pathologies similar to those experienced by people with MS. Despite the caveats and limitations of pre-clinical models (e.g., the lack of genetic diversity inherent to humans), EAE rodent studies have proven valuable with respect to the development of FDA-approved therapeutics for MS [18].
The studies presented here were done to determine whether the KD developed by D'Agostino and colleagues [19] that is enriched in fiber and contains medium chain triglycerides [MCTs; caprylic acid (C8) and capric acid (C10)] along with flaxseed oil and canola oil as the sources of fat can preserve motor and visual function in both male and female C57BL/6J mice undergoing EAE. Previous studies of nutritional interventions using this EAE model have demonstrated that caloric restriction attenuates EAE motor deficits and correlates with reduced levels of IL-6 and leptin [20]. Likewise, fast-mimicking diets [21, 22], intermittent fasting [23, 24], and KDs [22, 25, 26] reduce EAE motor disabilities, and a KD improved long-term potentiation, spatial learning, and memory [26].
Nutritional ketosis has been used since the 1920s to treat children with drug-refractory seizures [27, 28], and KDs have gained popularity for weight loss and as an anti-diabetic strategy because the diet suppresses hunger and reduces energy intake (29–31). These and other findings [reviewed in [32, 33]] have led to multiple clinical trials (e.g., NCT01538355, NCT03718247, NCT01915433, and NCT05007483) including “Nutritional Approaches in Multiple Sclerosis” (NAMS; NCT03508414), a randomized controlled clinical trial in Germany for people with active MS to compare a KD vs. a fasting protocol vs. a fat-modified standard diet over an 18-month period [34].
KDs can differ in composition but share at least two properties. The diets are comprised primarily of fats with moderate protein content and low amounts of carbohydrates, typically in the range of 0–70 g daily for humans, excluding indigestible fiber. Secondly, KDs induce the liver to produce the ketone bodies β-hydroxybutyrate, acetoacetate, and acetone. People in nutritional ketosis have circulating ketones of 0.5–4 mM with blood glucose <150 mg/dl, irrespective of fasting or fed state. This nutritional ketosis is distinguished from the pathological ketoacidosis associated with uncontrolled diabetes that yields ketones exceeding 25 mM and blood glucose >240 mg/dl.
Here we report that a KD can robustly protect against EAE-mediated motor and vision loss concomitant with reducing immune cell infiltration and preserving myelination of the optic nerve in both female and male mice. Functional preservation and protection against neurodegeneration were robust when the KD was fed as a preventive regimen prior to immunization to induce EAE and likewise when implemented as an intervention after symptom onset, demonstrating the translational feasibility of this nutritional approach. The KD increased circulating levels of multiple omega 3 (ω3) fatty acids associated with endogenous resolution pathways of acute inflammation and reduced circulating factors associated with neutrophil-mediated inflammation and MS pathogenicity. Together, these findings show that an MCT-based KD enriched in fiber confers neuroprotection and can reverse the loss of motor and visual function caused by autoimmune-mediated demyelination.
## Mice
Male and female C57BL/6J mice were housed in microisolator cages ($$n = 4$$–5 per cage) under a 12-h light/dark cycle and fed ad libitum [Picolab® Rodent Diet (cat # 5053)]. All animal care and experimental procedures were performed in compliance with ARRIVE guidelines, an Oklahoma Medical Research Foundation Institutional Animal Care and Use Committee (IACUC)-approved protocol, and complied with standards delineated by the Institute for Laboratory Animal Research. These studies adhered to The Association for Research in Vision and Ophthalmology (ARVO) statement for the Use of Animals in Research. All studies used C57BL/6J female and male mice. Breeders were purchased from Jackson Laboratories (stock # 000664). At the termination of experiments, mice were anesthetized to collect blood by cardiac puncture and subsequently sacrificed by CO2 asphyxiation followed by cervical dislocation.
## Diet compositions
Teklad control (TD.170645; CD) and ketogenic (TD.10911; KD) diets were obtained from Envigo, Inc., and custom formulated with the assistance of a company nutritionist based on [19]. At the macronutrient level, the KD provides 4.7 Kcal/g with $22.4\%$ Kcal from protein, $0.5\%$ Kcal from carbohydrate, and $77.1\%$ Kcal from fat. The CD provides 3.6 Kcal/g with $20.4\%$ Kcal from protein, $69.3\%$ Kcal from carbohydrate, and $10.4\%$ Kcal from fat. Macromolecular compositions and ingredients for both diets are provided in Figure 1A and Supplementary Figure 1A. Food and water were provided ad libitum and additional hydration/electrolytes were provided to animals showing signs of dehydration using intraperitoneal saline injections.
**Figure 1:** *The KD stabilizes body weight and blood glucose but increases circulating ketones. (A) Macromolecular compositions of diets. The sources of carbohydrate in the CD are corn starch, maltodextrin, sucrose, and fiber in the form of cellulose. The only carbohydrate in the KD is cellulose. The fat content of the KD is derived from MCTs (C8 and C10), flaxseed oil, and canola oil. (B) Outline of experimental approach for the preventive regimen. Mice were fed a KD or CD for 2 weeks prior to MOG35−55 immunization to acclimate and were maintained on their respective diets for the duration of the experiment. Motor scores and visual acuity were tracked daily, and blood draws (red drop symbol) for glucose and ketone levels were taken at days −14, 0, and 21 dpi. Tissues and plasma were harvested 3 weeks post immunization at study termination. (C–E) Graphs of body weights (C), blood glucose levels (D), and blood ketones (E) as a function of days before and after MOG35−55 immunization. For graphs, blue and green traces show female and male mice on KD, respectively. Red and purple traces show female and male mice on CD, respectively. N = 10–17 mice/sex. Whisker-bar standard deviations and p-values for differences between curves in C were computed with mixed linear model implementation of lme function, nlme R package. Asterisks denote statistical significance: *p < 0.05, **p < 0.01, ****p < 0.0001; ns, not statistically significant. Data compiled from 4 to 6 independent experiments.*
## MOG-EAE
EAE was induced in 10–12 week old mice by subcutaneous flank injection of 150 μg of myelin oligodendrocyte glycoprotein peptide (residues 35–55; MOG35−55; Genemed Synthesis, Inc., San Francisco, USA) emulsified in incomplete Freund's adjuvant (Thermo Fisher Scientific; DF0639606) supplemented with 5 mg/ml heat-inactivated *Mycobacterium tuberculosis* (Thermo Fisher Scientific; DF3114338). Mice were injected intraperitoneally with 250 ng *Bordetella pertussis* toxin (List Biological Laboratories, Inc. #181) the day of and 2 days following MOG35−55 immunization. For the prevention experiments, mice were euthanized for post-mortem histology 21–22 days post-immunization (dpi). For the intervention experiments (Figure 7), all mice were maintained on standard chow until symptom onset at which time animals were switched to either the KD or CD and followed until 28 dpi. For the study presented in Supplementary Figures 7B, C, KD-fed mice were followed until 35 dpi. Manifestations of progressive ascending paralysis of classical EAE were assessed daily, using a more granular scoring system than we previously described (35–37): 0—no disease, 0.5—reduced tail tone, 1—loss of tail tone, 1.5—limp tail and ataxia, 2—hind limb paresis, 2.5—one hind limb paralyzed, 3—complete hind limb paralysis, 3.5—complete hind limb paralysis and forelimb weakness, 4—hind limb paralysis and forelimb paresis, 5—moribund or dead. Mice were weighed daily to ensure weight loss did not exceed $25\%$ of starting weight at the time of immunization.
## Visual acuity assessment
Visual acuity threshold was measured daily by OKT response using Optometry software and apparatus (Cerebral Mechanics Inc., Alberta, Canada) as previously described (35–37). Briefly, mice are placed on a pedestal inside a box with a virtual cylinder consisting of vertical lines projected on four computer screens of the box surrounding the animal. The vertical lines rotate at varying frequencies, and tracking behavior is assessed in a stepwise manner as the thickness of the lines are reduced. Visual acuity is represented as the highest spatial frequency at which mice track the rotating cylinder. Optokinetic tracking is a temporal-to-nasal reflex, and therefore counter-clockwise and clockwise rotations exclusively test the right and left eye, respectively. When EAE-induced motor deficits rendered mice without the balance and stability to adequately perform OKT testing, a measurement was not recorded for the affected mouse on that day.
## Retinal flatmount analysis
Retinal flatmounts were prepared and RGCs labeled and quantified as described (35–37) using anti-Brn3a (goat; Santa Cruz, Santa Cruz, CA; sc31984; 1:500), Alexa546nm Fluor-conjugated donkey anti-goat IgG (Molecular Probes; 1:1,000), and Hoechst 33342 (2 μg/ml), all diluted in $3\%$ BSA/$10\%$ donkey serum/PBS. Retinas were washed in PBS before mounting with Prolong Gold mounting medium (Life Technologies, Grand Island, NY) and examined with a Nikon 80i microscope with a 60X objective. Images were captured with a DXM1200C camera using NIS-Elements software (Nikon, Inc. Tokyo, Japan). Photomicrographs were captured from the four leaflets comprising the flatmount with representative images captured from the peripheral, medial, and central retina within each quadrant, yielding 12 pictures per retina. Images were identically contrast-enhanced. Brn3a-positive RGCs were counted manually using the FIJI Cell Counter Plugin.
## Fatty acid analysis
Plasma fatty acid profiles were determined by extracting total lipids from 40 to 50 μl of plasma following the method of Bligh and Dyer [38] with slight modifications [39]. The purified total lipid extracts were stored under nitrogen until use. To these lipid extracts, 50 nmol each of 15:0 and 17:0 internal standards were added and the total extracts subjected to acid hydrolysis/methanolysis to generate fatty acid methyl esters (FAMEs) [40]. All reagents for fatty acid extraction and derivatization were of the highest quality available from Sigma-Aldrich. FAMEs were identified using an Agilent Technologies 7890A gas chromatograph with a 5975C inert XL mass spectrometer detector (Agilent Technologies, Lexington, MA) as described [40]. The gas chromatograph-mass spectrometer was operated in the electron impact total ion monitoring mode. The injection volume was 1 μl and the inlet, held at 325°C, was set to pulsed splitless mode. An Agilent Technologies HP-5MS column (30 m × 0.25 mm × 0.25 μm) was used with a helium carrier gas flow rate of 1.2 ml/min. The oven temperature began at 130°C for 1.0 min, was ramped to 170°C at 6.8°C/min, and was then ramped to 215°C at 2.9°C/min. After holding at 215°C for 15.0 min, the oven was ramped to 260°C at 20°C/min and held for 5.0 min. The oven was then ramped to 325°C at 15°C/min and held for 18.0 min. The mass spectrometer transfer line, ion source, and quadrupole temperatures were 325, 230, and 150°C, respectively.
FAMEs were quantified using an Agilent Technologies 6890N gas chromatograph with flame ionization detector (GC-FID) [41]. Sample concentrations were determined by comparison to internal standards 15:0 and 17:0. The injection volume was 1 μl and the inlet, held at 290°C, was set to pulsed split mode (10:1 ratio). An Agilent Technologies DB-23 column (60 m × 0.32 mm × 0.25 μm) was used with a hydrogen carrier gas constant pressure of 13.1 psi. The oven temperature began at 130°C for 0.8 min, was ramped to 170°C at 8.2°C/min, and was then ramped to 215°C at 3.5°C/min. After holding at 215°C for 9.5 min, the oven was ramped to 230°C at 50°C/min, and was then held for 8 min. The oven was then ramped to 290°C at 12.0°C/min and was held for 12 min. The detector was held at 290°C. Data for the diets is represented as μg of each fatty acid per mg of diet, and data from plasma is represented as nmol of fatty acid per mg of plasma protein.
## Blood glucose and ketone measurements
Blood ketone levels (non-fasting) were measured with a Precision Xtra Blood Ketone Monitoring System and fasting blood glucose was measured after 6 h of food withdrawal using a True Metrix Blood Glucose Meter. Blood was drawn from the tail vein.
## Cytokine analyses
Twenty-six cytokines and chemokines in the plasma of EAE mice 21 dpi were assayed using custom xMAP multiplex cytokine panels from Biotechne, Inc. according to the manufacturer's recommendations and processed in the OMRF Arthritis and Clinical Immunology Human Phenotyping Core. Six factors were below the detection limit of the assay in all samples (IL-1β, IL-2, IL-4, IL-13, IL-17A, and IL-27) and excluded from the table in Figure 6.
## Immunohistochemistry of optic nerves
To assess oligodendrocytes (as a marker of myelination) as well as lymphocyte and macrophage infiltration, we incubated sequential de-paraffinized sections with anti- 2′,3′-cyclic-nucleotide 3′-phosphodiesterase (CNPase; mouse CNPase: 1:200 dilution; Biolegend, Inc.), anti-CD3 (rabbit; 1:250; Abcam) or anti-Iba1 (mouse; 1:200; Millipore, Inc.) antibodies after antigen retrieval in R-buffer B. Images were captured using a Nikon TE2000 fluorescent microscope. Quantification of staining was performed using FIJI Software. CNPase, Iba1, and CD3 staining on sections spanning the entire length of the optic nerve were quantified by a masked experimentalist using representative images from both ends and the middle of the optic nerve. The number of positive cells within a defined area were counted using the Cell Counter Plugin, and the number of positive-stained cells per 10,000 pixels was calculated for each optic nerve.
## Ex vivo stimulation of splenocytes
Twenty-one dpi, single cell suspensions from spleens were prepared. 2 × 106 cells were plated in triplicate in complete RPMI (supplemented with $10\%$ FBS, 2 mM L-glutamine, $1\%$ pen/strep, 2 mM β-mercaptoethanol, non-essential amino acids, 1 mM sodium pyruvate). 10 μg/ml murine MOG35−55 peptide was added to the “stimulated” wells and incubated at 37°C for 24 h. For the final 5 h of the 24 h incubation, samples were spiked with 50 ng/ml PMA (Sigma, Inc. cat # P1585), 500 ng/ml ionomycin (Sigma, Inc. cat # IO634), and BD Golgi StopTM containing monensin (0.66 μl/ml; BD Biosciences). Cells from triplicate wells were combined, stained with a fixable viability dye (Biolegend) and treated with Fc block (TruStain FcX PLUS; BioLegend, Inc. cat # 156604) subsequent to staining with the following anti-mouse Abs against cell surface markers (all Abs from eBioscience, Inc.): anti-CD3 (145-2C11), anti-CD4 (GK1.5), anti-CD8 (53-6.7), and anti-CD44 (IM7). Cells were fixed and permeabilized with BD Cytofix/CytopermTM Plus (BD Biosciences, Inc.) and stained with anti-IL17 (TC11-18H10.1) and anti-IFN-γ (XMG1.2) to detect the respective cytokines intracellularly. Samples were processed on a LSRII flow cytometer (BD Biosciences) and data analyzed with FlowJo version 10.7.1 software (BD Biosciences, Inc.).
## Statistical analysis
Confidence intervals and p-values for the statistical significance of each studied effect in the longitudinal or binocular data were determined by fitting the data to a linear mixed-effects model, using the lme function implemented in the nlme R package, as detailed in Larabee et al. [ 35], Axtell et al. [ 42], Laird and Ware [43], and Lindstrom and Bates [44]. This function is an extended version of regular linear regression but can accommodate complex data collection design features, such as longitudinal measurements, nested layers, and within-group correlation. Standard testing methods, such as the Student's paired or unpaired t-test and the Mann–Whitney exact test, with Bonferroni correction for multiple testing, were employed when no embedding was involved. Graphpad Prism 9 was used to display the results. Testing for differences in marker levels among the five mice groups in Figure 4 was performed following log transformations for improving normality by unequal group variance ANOVA and Tukey's post hoc adjustment. Differences in cytokine and chemokine levels in Figure 6 were assessed with robust linear regression using the rlm function in MASS R package, and p-value computations relied on robust F-Testing (Wald) performed with f.robftest function from sfsmisc package. Principal component exploratory analysis (PCA) and biplot generation were carried out with specific functions from stats package in R. Multidimensional $95\%$ confidence regions were added to the biplot using the draw.ellipse function from the plotrix package. Throughout the manuscript, the asterisks code denoting significance is: *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$, ****$p \leq 0.0001$; ns, not statistically significant.
## Results
The current study tested the efficacy of a KD in the EAE mouse model of autoimmune-mediated demyelination. Many pre-clinical studies that have demonstrated therapeutic efficacy with a KD have used formulations containing lard, soybean oil, and hydrogenated fats [e.g., [22, 45, 46]]. In contrast, the KD used here is enriched in fiber and the sources of fat are MCTs, flaxseed oil, and canola oil (Figure 1A). This diet was designed to be anti-inflammatory by minimizing ω6 fatty acids and hydrogenated fats [19]. A complete list of the diet ingredients is provided in Supplementary Figure 1A.
The experimental paradigm used to test the KD in a preventive regimen is shown in Figure 1B. Animals were fed either a KD or an ingredient-matched control diet (CD) for 2 weeks prior to MOG35−55 peptide immunization to acclimate mice to the diet based on previous strategies testing dietary impacts on CNS autoimmunity in the EAE model [e.g., [47]]. Mice were maintained on their respective diets for the duration of the experiment during which motor scores, vision, and body weights were longitudinally tracked and blood was collected. Three weeks post-immunization, mice were euthanized for post-mortem analyses. Notably, consumption of the KD for 5 weeks in the absence of EAE led to modest weight reduction whereas 5 weeks of the CD did not (Supplementary Figure 1B). Animals were randomized to KD or CD groups and had comparable weights within each sex at study initiation (Figure 1C). From immunization to completion of experiments at 21dpi, among EAE mice fed the KD, females slightly gained weight whereas males were weight stabilized. In contrast, both sexes on the CD undergoing EAE lost body weight (Supplementary Figure 1C), consistent with sickness behavior (e.g., inflammatory anorexia) [48]. Compared to the CD, the KD reduced fasting blood glucose levels after 2 weeks of feeding despite no differences at study onset or termination (Figure 1D). Elevated non-fasting ketones are a defining molecular signature of a KD [49] and circulating, non-fasting ketones (β-hydroxybutyrate specifically) were increased in mice fed the KD for 2 weeks and persisted through the duration of the study (Figure 1E). Mice on the CD also had moderately elevated circulating ketones 3 weeks after immunization, likely related to sickness behavior (e.g., inflammatory anorexia) [48]. These data show that the KD lowered fasting blood glucose and induced nutritional ketosis but did not promote obesity during 5 week-long experiments.
## KD prevents the onset of motor and visual deficits in EAE mice
Remarkably, mice fed the KD were spared the ascending paralysis and motor deficits induced by EAE (Figure 2A), with comparable efficacy observed for females and males (Figures 2B, C, respectively). Because optic neuritis and visual disturbances are major sequelae of autoimmune demyelinating disease in humans (50–52), we complemented the motor score measurements with visual acuity measurements using optokinetic tracking (OKT). OKTs were recorded at the initiation of the diets, prior to MOG35−55 immunization, and then daily beginning ~10 dpi, coincident with symptom onset in CD-fed animals (Figure 1B). As we reported previously [36], mice subjected to MOG-EAE typically undergo episodic monocular vision loss, consistent with what is experienced by people with relapsing-remitting MS (RRMS) [53, 54]. These deficits are readily detectable as diminished visual acuity in one eye [35, 36]. The OKT measures each eye separately, so every mouse has a designated more affected (MA) and less affected (LA) eye. Coincident with the mitigation of motor deficits, the KD preserved vision in the MA eyes of EAE mice (Figure 3A). Although visual deficits were minimal in the LA eyes of mice on either diet, the LA eyes of KD-fed mice retained better vision than their CD-fed counterparts (Figure 3B, inset). In alignment with the efficacy of the KD on motor function, this preservation of vision was independent of sex (Supplementary Figures 2A, B).
**Figure 2:** *The KD prevents EAE-induced motor deficits. (A) Daily motor scores of EAE mice on the KD (green) vs. the CD (orange). (B) Daily motor scores as a function of diet for female mice. Blue trace shows data for KD-fed mice and red trace shows data for CD-fed mice. (C) Daily motor scores as a function of diet for male mice. Green trace shows data for KD-fed mice and purple trace shows data for CD-fed mice. N = 10–17 mice/sex/diet. Whisker-bar standard deviations and p-values for differences between curves were computed with mixed linear model implementation of lme function, nlme R package. Asterisks denote statistical significance: **p < 0.01, ****p < 0.0001. All data compiled ≥4 independent experiments.* **Figure 3:** *The KD prevents visual acuity deficits and spares retinal ganglion cells in the central retinas of EAE mice. (A) OKT measurements of visual acuity of the more affected (MA) eyes of EAE mice fed the KD (green) or CD (orange); n = 14–17 mice/diet. (B) Same as (A) for less affected (LA) eyes; n = 10–12 mice/diet. Inset with adjusted scales of x-axis and y-axis highlights significant differences in visual acuity between diets. Whisker-bar standard deviations and p-values for differences between curves in (A, B) were computed with mixed linear model implementation of lme function, nlme R package. (C) Representative photomicrographs of Brn3A+ RGCs in the central, medial, and peripheral retinas of mice as a function of diet. Examples are shown for both the MA and the LA eyes. (D) Graph of Brn3A+ RGC cell counts in central, medial, and peripheral retinas of MA eyes as a function of diet. Data are color coded based on diet as in (A, B). Data compiled ≥4 independent experiments. **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant.*
The axons of retinal ganglion cells (RGCs) are bundled together into the optic nerve and episodes of optic neuritis induce RGC apoptosis (55–59). We therefore quantified RGC counts and found that the functional preservation of vision mediated by the KD was accompanied by a sparing of Brn3A+ RGCs in the central and medial retina of MA eyes from EAE mice (Figures 3C, D). As we reported previously [36], RGCs in the peripheral retina are not typically lost during MOG-EAE (Figure 3D), fitting with RGC loss in longstanding MS cases being most prominent in the central retina, nearest the optic nerve head [60]. When analyzed as a function of diet and sex, the KD-mediated sparing of RGCs in the MA eyes was significant in the central retina for both females and males (Supplementary Figure 2C).
## The KD preserves oligodendrocytes and restricts inflammatory infiltration of the optic nerve in EAE mice
To complement the OKT and RGC analyses, immunohistochemistry (IHC) was performed on paraffin-embedded optic nerves from the MA eyes of the mice analyzed for visual function in Figure 3. For each marker, labeling was quantified from sections captured along the entire length of the optic nerve, and optic nerves from healthy (i.e., no EAE) mice were included for comparison. CNPase, an oligodendrocyte marker and proxy of myelination, was largely preserved on the optic nerves of KD-fed mice (Figure 4A). CNPase labeling was quantified by comparing perinuclear staining, with the specificity of this staining corroborated by labeling optic nerves from healthy mice (Figure 4A, left panel inset). Labeling for Iba1 and CD3 to mark macrophages/microglia and T cells, respectively, revealed that mice fed the KD had reduced Iba1 and CD3 labeling of the optic nerve compared to mice fed the CD (Figures 4B, C, respectively).
**Figure 4:** *The KD preserves oligodendrocytes on the optic nerve and reduces immune cell infiltrates. (A–C) Representative photomicrographs of paraffin-embedded optic nerve sections labeled for the indicated markers (CNPase, Iba1, and CD3, A–C respectively). Optic nerves from the MA eyes of EAE mice consuming the indicated diets and from healthy (no-EAE) mice on standard chow were compared. EAE MA optic nerves were taken from the mice analyzed functionally by OKT in Figure 3. Sections spanning the entire length of the optic nerve were labeled with each marker. Insets in (A) highlight the perinuclear CNPase labeling that was quantified. Mean values for CNPase, Iba1, and CD3 labeling of optic nerve sections are shown in the graphs to the right of the photomicrographs. Whisker bars depict 95% confidence intervals for each variable and horizontal bars with accompanying p-values above the bars indicate statistical significance as calculated by unequal group variance ANOVA with Tukey's post hoc adjustment. Size bar in (C) corresponds to 100 μm. (D) Principal components (PC) biplot to summarize, in 2d projection, the similarities among mice groups and their relationships to the five measured scores (i.e., motor scores, changes in OKT from baseline, CNPase, Iba1, CD3). The dots on the graph are samples, colored according to their phenotype group, with the measured variables shown as arrows. Arrow coordinates on the two axes show each variable contributing loading on the first two principal components. Multidimensional 95% confidence regions for each mouse group, projected in 2d, are shown as ellipses. All data acquired from n ≥ 4 mice/sex/diet. The graph to the right of the PC biplot makes explicit, for the five markers, the difference between the two clusters of mice (Healthy + KD vs. CD) separated by the first principal component. Whisker bars depicting 95% confidence intervals are shown for each variable both in Healthy + KD mice and in CD mice with the p-values assessing the significance of their differences marked on the graph.*
When the data were further examined to determine the combined impacts of diet and sex, both females and males on the KD had significantly preserved CNPase labeling compared to their CD-fed counterparts (Supplementary Figure 3A). Likewise, KD-fed male optic nerves had decreased Iba1 and CD3 labeling compared to CD-fed males (Supplementary Figures 3B, C, respectively). Due to higher variability between samples, differences in Iba1 and CD3 labeling did not reach significance between the KD and CD female mice (Supplementary Figures 3B, C, respectively), despite the preservation of visual acuity for female mice on the KD (Supplementary Figure 2A). Notably, all groups had statistically significant increases in Iba1 labeling compared to optic nerves from healthy mice (Supplementary Figure 3B), despite no loss of visual acuity for the KD group (Figure 3A). For CD3 labeling, females and males on the KD were not statistically different than healthy mice whereas EAE males and females on the CD showed significant increases in CD3+ infiltrates (Supplementary Figure 3C).
Principal components exploratory analysis (Figure 4D) confirmed that the relatively higher CNPase labeling and lower Iba1 and CD3 labeling for mice fed the KD correlated with lower motor and visual acuity deficits. The first principal component reveals that the KD-fed mice are statistically comparable to healthy mice but significantly different from the CD-fed group, presenting significantly improved motor and visual acuity scores and higher CNPase levels, along with lower Iba1 and CD3 levels. The second component suggests marginal separation of the healthy mice group from KD-fed mice, in terms of Iba1 and CD3 levels.
## A KD increases circulating ω3 fatty acids in EAE mice
To test the hypothesis that the therapeutic efficacy of the KD derives from inducing a systemic anti-inflammatory milieu, we analyzed a panel of circulating fatty acids (FAs) in the plasma of EAE mice 21 dpi (Figure 5A). With a focus on those FAs showing statistically significant differences between diets, we observed that, compared to the CD, the KD elevated circulating levels of 18:3n3, 20:4n3, 20:5n3, and 22:5n3 whereas the saturated fats, 14:0 and 16:0, the mono-saturated fat 16:1, and 20:4n6 (arachidonic acid; AA) were all decreased (Figure 5B). Likewise, the ratio of 20:4n6 to (20:5n3 + 22:6n3) [i.e., AA/(EPA + DHA)], a proxy of inflammatory status, was decreased by the KD (Figure 5B). Notably, 20:5n3, 22:5n3, and 22:6n3 are biosynthetic precursors of E-series, T-series, and D-series resolvins, respectively, specialized pro-resolving lipid mediators (SPMs) that restrict the development of chronic inflammation by dampening acute inflammation [reviewed in [61]]. Consistent with the KD limiting systemic inflammation, 20:3n9 (eicosatrienoic acid; ETA) levels were decreased (Figure 5B). This ω9 fatty acid can be converted to the proinflammatory C3 and D3 leukotrienes [62].
**Figure 5:** *The KD increases plasma ω3 fatty acid content and suppresses levels of pro-inflammatory ω6 fatty acids. (A) List of fatty acids analyzed with the accompanying omega designations (aka. lipid), common names, and systematic names. (B) Graph depicting nmols of individual fatty acids per mg of protein in the plasma that are statistically significantly different (*p ≤ 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by Mann-Whitney test with Bonferroni correction) between EAE mice fed the KD (green) vs. the CD (orange). The ratio of arachidonic acid (AA; 20:4n6) to (EPA +DHA) (20:5n3 + 22:6n3) is included as a proxy of systemic inflammation. Graph uses log scale on the y-axis with data derived from n ≥ 6 mice per diet from ≥2 independent experiments.*
We also observed several circulating FAs changed as a combined function of sex and diet. Females fed the KD had increased serum levels of 18:2n6, 20:1, and 22:1 along with decreases in 16:1, 18:1, 20:2n6, and 20:4n6 (Supplementary Figure 4A). EAE male mice fed the KD had elevated levels of the ω3 fatty acid 20:5n3 (Supplementary Figure 4B), the intermediate in the conversion of 20:4n3 to 22:5n3 [63]. 20:5n3 was also increased in KD-fed females but did not reach significance because of high variability among 20:5n3 levels within female CD-fed mice (data not shown). Together, these data are consistent with the KD conferring protection from EAE pathologies at least in part by enriching the systemic milieu with ω3 fatty acids associated with SPM-mediated neuroprotection and decreasing fatty acids, such as 20:3n9, that are precursors for the biosynthesis of pro-inflammatory leukotrienes.
To determine which of the enriched circulating ω3 fatty acids in EAE mice consuming the KD were contributed directly by the diet, we analyzed the fatty acid content of each diet (Supplementary Figure 4C). Notably, compared to the CD, the additional flaxseed oil in the KD led to an enrichment by 4.3 μg fatty acid/mg sample (or 1.6-fold) in 18:3n3 (α-linolenic acid; ALA). 18:3n3 can be biosynthetically converted in vivo to the corresponding 20:4n3, 20:5n3, and 22:5n3 [63, 64]. As 20:4n3, 20:5n3, and 22:5n3 were absent from the KD and CD diets, the elevated levels of these ω3 fatty acids in the plasma of EAE mice consuming the KD (Figure 5B) appear to be derived from the endogenous conversion of 18:3n3.
## Cytokines impacted by the KD in EAE mice
Complementary evidence supporting the anti-inflammatory milieu resulting from the KD came from an analysis of 26 different cytokines and chemokines in the plasma 21 dpi, of which 20 were detectable in some or all samples within the sensitivity of the multiplex assay (Figure 6A). Of the cytokines detected, granulocyte-colony stimulating factor (G-CSF), C-X-C Motif Chemokine Ligand 2 (CXCL2), C-C motif chemokine ligand 11 (CCL11), and IL-6 were all significantly decreased 21 dpi in the plasma of EAE mice consuming the KD (Figure 6B). Two sex-specific changes were also detected (Supplementary Figure 5). EAE females on the KD had elevated monocyte-colony stimulating factor (M-CSF) whereas males had reduced C-reactive protein (CRP). Curiously, CRP levels in KD female mice were comparable to the levels in KD-fed males but CRP was not elevated in the female CD-fed group (Supplementary Figure 5).
**Figure 6:** *The KD suppresses levels of G-CSF, CXCL2, and other markers of inflammation. (A) Table of cytokine and chemokine measurements from the plasma of EAE mice. Asterisks and bolding highlight factors that are statistically significantly different as a function of diet. “AVG” denotes average and “SD” denotes standard deviation. “ND” indicates that all samples analzyed were below the detection limit of the assay and “N/A” indicates that insufficient numbers of samples were within the detectable range of the assay to perform statistical analyses. (B) Graphical presentation of the four factors that differ in the plasma 21 dpi as a function of diet. EAE mice consumed either the KD (green) or a CD (orange), 4–8 mice per diet per sex from ≥2 independent experiments. Statistical significance for cytokines and chemokines was determined with robust linear regression using the rlm function in MASS R package. p-values relied on robust F-Testing (Wald) performed with f.robftest function from sfsmisc package. For graphs in (B), *p < 0.05, ***p < 0.001.*
## EAE mice fed a KD mount a T-cell response to MOG35−55 immunization
To rule out that the protection conferred by the KD in the preventive regimen was not a consequence of the diet blunting an immune response to the MOG antigen, splenoctyes were isolated 21 dpi from EAE mice on each diet, stimulated ex vivo in culture with MOG35−55 peptide, and subsequently with PMA, iononmycin, and monensin prior to antibody labeling to identify activated T cells (i.e., CD3+, CD4+, CD44++) and their respective intracellular levels of IL-17 and IFN-γ by flow cytometry (Supplementary Figure 6A). IL-17 and IFN-γ expression by MOG35−55-activated splenic T cells is a signature EAE response following immunization with MOG [e.g., [65, 66]]. For comparison purposes, unstimulated sets of splenocytes from mice on either the KD or CD were similarly stained for intracellular IL-17 and IFN-γ. These analyses showed that the percent of live CD3+CD4+ CD44++ T cells expressing IL-17 in the spleen was comparable between diets whereas a greater faction of stimulated splenocytes from KD-fed animals expressed IFN-γ vs. their CD-fed counterparts (Supplementary Figures 6A, B). Notably, the level of activated CD3+CD4+ CD44++ T cells was similar between the KD and CD groups (Supplementary Figure 6C).
Of note, we observed a small of percentage of “breakthrough” mice consuming the KD that began to exhibit EAE symptoms ~19–21 dpi (Supplementary Figure 7A). To determine whether all or most KD-fed mice would manifest EAE symptoms if the experimental time course was extended, we tracked EAE mice on the KD for an additional 2 weeks (i.e., 2 weeks pre-immunization plus 5 weeks post-immunization). Compared to the $90\%$ of CD-fed mice ($100\%$ males and $80\%$ females) that showed symptoms within 10–16 dpi (11–12 dpi for males and 10–16 dpi for females), extending the study out to 7 weeks total revealed that disease incidence among KD-fed mice was ~ $33\%$ ($20\%$ for females and $50\%$ for males) with symptoms not detected until 22–33 dpi (22–33 dpi for males and 31–32 dpi for females; Supplementary Figure 7B). Furthermore, the severity of motor deficits among the subset of KD-fed animals that eventually developed symptoms was significantly reduced compared to their CD-fed counterparts (Supplementary Figure 7C). Together, these lines of evidence support that although most mice on the KD are protected from the overt pathologies induced by the EAE model, the underlying mechanism is not a failure to mount a T cell response to immunization with the MOG antigen. Collectively, the KD effectively prevented disease onset in most mice and mitigated disease severity in the subset of mice that manifested functional deficits.
**Figure 7:** *The KD restores motor and visual function to EAE mice when fed following symptom onset. (A) Diagram of the interventional regimen. Mice were immunized with MOG35−55 (day 0), switched to either KD or CD diets following symptom onset (10–14 dpi), and maintained on the respective diets for the duration of the experiment (28 dpi). Motor scores and visual acuity were tracked as indicated. Daily motor scores of female (B) and male mice (C) switched to the indicated diets following the onset of motor and visual deficits. Blue and green traces show female and male mice on KD, respectively, with red and purple traces showing female and male mice on CD, respectively. n = 13 mice/sex/diet. (D) OKT measurements of visual acuity of the more affected (MA; top graph) and less affected (LA; bottom graph) eyes in female mice on the indicated diets. n = 13 mice/diet. (E) Same as (D) for male mice. n = 13 mice/diet. All data graphed with the x-axis representing “days post-symptom onset and diet switch”. N.B.: DAY 0 is the day before symptom onset and is denoted by a motor score of zero. Asterisks in all graphs denote statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Data compiled ≥4 independent experiments.*
## The KD as an interventional regimen restores motor and visual function
Complementary studies in female and male mice were performed to determine the efficacy of the KD as an interventional regimen (Figure 7A). All mice were maintained on standard chow until EAE symptom onset. At the first observable sign of motor and visual deficits, mice were switched to either the KD or the CD for the remainder of the study. Within 4 days of consuming the KD, motor (Figures 7B, C) and visual (Figures 7D, E) deficits were significantly mitigated in both sexes and continued to improve such that motor and visual function were restored to near baseline levels by the termination of the study. Notably, these studies spanned 26–30 days from the day of MOG immunization until termination but the data are graphed with the x-axis representing “days post-symptom onset and diet switch” to normalize for all mice not manifesting their respective initial functional deficits on the same day. These results complement the above prevention studies and importantly, confer translational relevance and feasibility to this dietary strategy.
## Discussion
Previous studies established that dietary interventions including caloric restriction [20], a fast-mimicking diet [21, 22], intermittent fasting [23, 24], and a KD [22, 25, 26] can ameliorate EAE motor and cognitive deficits in EAE mice. Our data expand on these findings by demonstrating that a KD preserves visual acuity (Figure 3) and myelination of the optic nerve, reduces Iba1+ and CD3+ infiltrates (Figure 4), and spares RGCs (Figure 3). We further show that the benefits extend comparably to female and male mice and that the KD increased circulating levels of the anti-inflammatory ω3 fatty acids 18:3n3, 20:4n3, 20:5n3, and 22:5n3 while decreasing the pro-inflammatory fatty acids 20:3n9 and 20:4n6 (Figure 5). We also detected a KD-mediated decrease in circulating cytokines implicated in EAE and MS, namely G-CSF, CXCL2, CCL11, and IL-6 (Figure 6B). When tested as an intervention, the KD promoted rapid and nearly complete recovery of motor and visual function (Figure 7). This finding is particularly promising for establishing the clinical feasibility of acutely implementing the KD in response to symptom relapse.
The diet composition of the KD used in the present study [19] differs with previous EAE studies that used KDs containing butter, corn oil, and lard [e.g., [22, 26]], sources of fat associated with hepatic dysfunction and inflammation [67, 68]. The different KDs were efficacious in mitigating EAE sequelae, but a consideration of the cardiometabolic health risks of chronically consuming high fat diets is of relevance to people with MS as these individuals will likely implement long-term nutritional interventions. In this regard, our studies support that the use of a KD enriched in ω3 fatty acids including α-linolenic acid (ALA, 18:3n3) can elicit therapeutic benefits without promoting weight gain or increasing inflammation (Supplementary Figure 1 and Figure 6).
Because different KD formulations mitigate EAE pathologies despite containing various pro- and anti-inflammatory fats [22, 26], the therapeutic efficacy of KDs may derive from the production of ketones and/or the dramatic limitation of simple sugars and starch consumption common to these regimens. Consistent with these hypotheses, high dietary glucose increases disease severity in the EAE model by promoting TH17 cell differentiation via ROS-mediated TGF-β activation [69]. In clinical work, a KD improved glycemic control and reduced the medications of people with obesity and diabetes more effectively than a $55\%$ carbohydrate low glycemic index diet [70]. For people with (pre)diabetes, a KD reduced glycosylated hemoglobin (HbA1C), fasting glucose, fasting insulin, weight, blood pressure, triglycerides, alanine aminotransferase, and high-density lipoprotein, consistent with safety and tolerability for long-term adherence. Remarkably, $53\%$ of enrolled participants achieved disease resolution [30]. Individuals that are overweight and implemented a KD also had decreased body weight, insulin resistance, and serum markers of inflammation (e.g., TNF-α, IL-6, IL-8, MCP-1) (71–73). These normalizing effects on insulin sensitivity and insulin resistance benefit people with metabolic syndrome and diabetes [reviewed in [74]] and likely extend to people with MS, as insulin resistance is associated with elevated disability scores [16, 17]. Clinical trials with small cohorts of people with MS have also shown promising responses using (modified) paleolithic diets that eliminate or dramatically reduce the consumption of simple carbohydrates, ultra-processed foods, and other putative disease aggravators (e.g., gluten, dairy, legumes) (13, 75–81).
A potential mechanistic contribution mediating the efficacy of the KD in the present study comes from the anti-inflammatory actions of resolvins, a family of SPMs derived from ALA. In alignment with this idea, elevated dietary consumption of 18:3n3 has been inversely linked with the risk of developing MS [82]. ALA was enriched in the KD (Supplementary Figure 4C) as well as in the plasma of EAE mice fed the KD (Figure 5B). Moreover, the ω3 fatty acids in the plasma of mice fed the KD may directly contribute to lower levels of inflammatory infiltrates in the MA optic nerves (Figure 4) as ω3 fatty acids enhanced lesion recovery by decreasing phagocytic infiltration to the corpus callosum following demyelination [83].
The plasma of KD EAE females and males was enriched in ETA (20:4n3) and DPA (22:5n3; Figure 5 and Supplementary Figures 4A, B) and males were also enriched with EPA (20:5n3; Supplementary Figure 4B). 20:5n3 and 22:5n3 are intermediate biosynthetic precursors of E-series and T-series resolvins, respectively. DHA (22:6n3) and arachidonic acid (20:4n6), the intermediates from which the Resolvin D-series and the lipoxins are synthesized, respectively, were not increased and arachidonic acid levels were reduced. However, the lack of elevated DHA and arachidonic acid does not necessarily preclude contributions from their respective SPMs as efficient synthesis reactions could limit accumulation of these intermediates. A study using the MOG-EAE model and standard chow over a 45-day course showed PUFA metabolism is compromised during disease progression and that daily administration of exogenous Resolvin D1 decreased EAE pathologies [84]. Clinical work has identified imbalances in pro-inflammatory eicosanoids and SPM levels in the plasma of people with MS as a function of disease progression and severity. The expression of multiple biosynthetic enzymes and receptors for SPMs were impaired in peripheral blood mononuclear cells from people with MS and their monocytes were less responsive to SPMs in culture [85]. Such deficiencies in the SPM synthetic machinery may underlie the observation that elevated EPA in the serum of people with MS tracks with increased severity on the expanded disability status scale (EDSS). This study additionally reported that circulating arachidonic acid levels associate with relapse-free status [86]. Notably, study participants were taking IFN-β or other disease-modifying therapies, which confounds comparing these data directly to our study or other stand-alone dietary intervention studies.
Fitting with the KD promoting a systemic anti-inflammatory milieu, we detected reductions in four pro-inflammatory cytokines and chemokines in the circulation as a function of diet: G-CSF, CXCL2, CCL11, and IL-6 (Figure 6). G-CSF and CXCL2 were the two pro-inflammatory cytokines reduced in both sexes fed the KD (Supplementary Figure 5). G-CSF levels are typically kept low but rapidly increase in response to stress and inflammation to stimulate the production and maturation of granulocytes and neutrophils. G-CSF governs early signaling necessary for EAE disease induction [87] with the neutrophils produced driving multiple steps of EAE and MS progression [e.g., (88–92)], fitting with the KD reducing G-CSF in the circulation and blocking symptom onset.
CXCL2 levels were also reduced by the KD (Figure 6B). This factor is also known as macrophage inflammatory protein 2-α (MIP2-α), is produced by macrophages and neutrophils at sites of inflammation, and functions to recruit neutrophils during inflammation [93]. Transient receptor potential melastatin 2 knockout mice also have reduced CXCL2 and a suppression of neutrophil infiltration into the CNS during EAE [94]. Thus, reduced levels of circulating CXCL2 are consistent with disease mitigation (95–97) and fit logically with lower G-CSF levels limiting neutrophil maturation.
KD fed mice had reduced CCL11, also known as eosinophil chemotactic protein (eotaxin-1; Figure 6B). This chemokine is a putative biomarker of disease duration in people with secondary progressive MS [98]. Curiously, in a rat EAE model, increased CCL11 expression was associated with a TH2 anti-inflammatory response but did not correlate with eosinophil recruitment. Elevated CCL11 was detected in the CSF and lymph nodes (but not in the serum) and correlated with decreased ED1+/Iba1+ macrophages in the spinal cord and with protecting the integrity of the blood-brain barrier [99]. In contrast, CCL11 was found to be elevated in the spinal cords of mice undergoing MOG35−55 EAE [100]. As CCL11 has been implicated in systemic inflammation and as a pathogenic factor in a range of neurodegenerative and neuroinflammatory diseases [reviewed in [101]], reduced CCL11 in the serum by the KD is consistent with an anti-inflammatory milieu.
The KD reduced circulating levels of IL-6 (Figure 6B), a pro-inflammatory cytokine secreted by astrocytes, macrophages, and microglia as well as other cell types in the CNS (e.g., neurons and endothelial cells) [102]. IL-6 is detectable in brain lesions [103] and in the CSF from people with MS [104] and plays critical roles in MS and EAE pathophysiology that include compromising blood-brain barrier integrity in combination with IL-17A [105], cooperating with transforming growth factor-β to drive the differentiation and expansion of auto-reactive TH17 cells (106–109), and damaging myelin [102]. Reduced circulating IL-6 by a KD has been reported [e.g., (110–116)] although studies in select populations [e.g., [117, 118]] and some mouse models have also reported that a KD may not change or may even increase levels of this cytokine [e.g., [119]]. Reduced levels of IL-6 in the serum of KD-fed EAE mice are consistent with this dietary approach blunting CNS disease burden and maintaining motor and visual functions.
In conclusion, our work demonstrates the efficacy of a KD to preserve motor and visual function in mice undergoing autoimmune demyelinating disease. This dietary strategy limits systemic inflammation by reducing key cytokines involved in mediating the infiltration, activation, and differentiation of auto-reactive T cells and neutrophils into the CNS. We further posit that this KD provides an abundance of ω3 fatty acids for SPM biosynthesis, the products of which restrict acute inflammatory responses to self-antigens from transitioning to chronic inflammation and tissue damage. The observation that initiating the KD beginning at the time of symptom onset can resolve both motor and visual deficits (Figure 7) supports the potential of this diet for direct translational application and improved patience compliance, a current barrier for nutrition-based therapeutic strategies (120–122).
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The animal study was reviewed and approved by Oklahoma Medical Research Foundation Institutional Animal Care and Use Committee.
## Author contributions
KZ-J: experimental design, acquisition and analysis of data, figure preparation, and writing and editing. DW, SK, RB, MT, and M-PA: acquisition and analysis of data and editing. KP: acquisition and analysis of data, figure preparation, and writing and editing. CG: statistical analyses and editing. SP: experimental design, figure preparation, and writing and editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Author disclaimer
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur.2023.1113954/full#supplementary-material
## References
1. Leray E, Moreau T, Fromont A, Edan G. **Epidemiology of multiple sclerosis**. *Rev Neurol.* (2016) **172** 3-13. DOI: 10.1016/j.neurol.2015.10.006
2. Tullman MJ. **Overview of the epidemiology, diagnosis, and disease progression associated with multiple sclerosis**. *Am J Manag Care.* (2013) **19** S15-20. PMID: 23544716
3. Walton C, King R, Rechtman L, Kaye W, Leray E, Marrie RA. (2020) **26** 1816-21. DOI: 10.1177/1352458520970841
4. Versini M, Jeandel PY, Rosenthal E, Shoenfeld Y. **Obesity in autoimmune diseases: not a passive bystander**. *Autoimmun Rev.* (2014) **13** 981-1000. DOI: 10.1016/j.autrev.2014.07.001
5. Ivancovsky-Wajcman D, Fliss-Isakov N, Webb M, Bentov I, Shibolet O, Kariv R. **Ultra-processed food is associated with features of metabolic syndrome and non-alcoholic fatty liver disease**. *Liver Int.* (2021) **41** 2635-45. DOI: 10.1111/liv.14996
6. Martinez Steele E, Juul F, Neri D, Rauber F, Monteiro CA. **Dietary share of ultra-processed foods and metabolic syndrome in the US adult population**. *Prev Med.* (2019) **125** 40-8. DOI: 10.1016/j.ypmed.2019.05.004
7. Tavares LF, Fonseca SC, Garcia Rosa ML, Yokoo EM. **Relationship between ultra-processed foods and metabolic syndrome in adolescents from a Brazilian Family Doctor Program**. *Public Health Nutr.* (2012) **15** 82-7. DOI: 10.1017/S1368980011001571
8. Fiolet T, Srour B, Sellem L, Kesse-Guyot E, Alles B, Mejean C. **Consumption of ultra-processed foods and cancer risk: results from NutriNet-Sante prospective cohort**. *BMJ.* (2018) **360** k322. DOI: 10.1136/bmj.k322
9. Mendonca RD, Lopes AC, Pimenta AM, Gea A, Martinez-Gonzalez MA, Bes-Rastrollo M. **Ultra-processed food consumption and the incidence of hypertension in a mediterranean cohort: the seguimiento universidad de navarra project**. *Am J Hypertens.* (2017) **30** 358-66. DOI: 10.1093/ajh/hpw137
10. Schnabel L, Kesse-Guyot E, Alles B, Touvier M, Srour B, Hercberg S. **Association between ultraprocessed food consumption and risk of mortality among middle-aged adults in France**. *JAMA Intern Med.* (2019) **179** 490-8. DOI: 10.1001/jamainternmed.2018.7289
11. Hall KD, Ayuketah A, Brychta R, Cai H, Cassimatis T, Chen KY. **Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of**. *Cell Metab* (2019) **30** 67-77. DOI: 10.1016/j.cmet.2019.05.008
12. Thorburn AN, Macia L, Mackay CR. **Diet, metabolites, and “western-lifestyle” inflammatory diseases**. *Immunity.* (2014) **40** 833-42. DOI: 10.1016/j.immuni.2014.05.014
13. Wahls TL, Chenard CA, Snetselaar LG. **Review of two popular eating plans within the multiple sclerosis community: low saturated fat and modified paleolithic**. *Nutrients* (2019) **11** 352. DOI: 10.3390/nu11020352
14. Stoiloudis P, Kesidou E, Bakirtzis C, Sintila SA, Konstantinidou N, Boziki M. **The role of diet and interventions on multiple sclerosis: a review**. *Nutrients* (2022) **14** 1150. DOI: 10.3390/nu14061150
15. Sato W, Yamamura T. **Multiple sclerosis: possibility of a gut environment-induced disease**. *Neurochem Int.* (2019) **130** 104475. DOI: 10.1016/j.neuint.2019.104475
16. Oliveira SR, Kallaur AP, Lopes J, Colado Simao AN, Reiche EM, de Almeida ERD. **Insulin resistance, atherogenicity, and iron metabolism in multiple sclerosis with and without depression: associations with inflammatory and oxidative stress biomarkers and uric acid**. *Psychiatry Res.* (2017) **250** 113-20. DOI: 10.1016/j.psychres.2016.12.039
17. Oliveira SR, Simao AN, Kallaur AP, de Almeida ER, Morimoto HK, Lopes J. **Disability in patients with multiple sclerosis: influence of insulin resistance, adiposity, and oxidative stress**. *Nutrition.* (2014) **30** 268-73. DOI: 10.1016/j.nut.2013.08.001
18. Glatigny S, Bettelli E. **Experimental autoimmune encephalomyelitis (EAE) as animal models of multiple sclerosis (MS)**. *Cold Spring Harb Perspect Med* (2018) **8** a028977. DOI: 10.1101/cshperspect.a028977
19. Brownlow ML, Benner L, D'Agostino D, Gordon MN, Morgan D. **Ketogenic diet improves motor performance but not cognition in two mouse models of Alzheimer's pathology**. *PLoS ONE.* (2013) **8** e75713. DOI: 10.1371/journal.pone.0075713
20. Piccio L, Stark JL, Cross AH. **Chronic calorie restriction attenuates experimental autoimmune encephalomyelitis**. *J Leukoc Biol.* (2008) **84** 940-8. DOI: 10.1189/jlb.0208133
21. Bai M, Wang Y, Han R, Xu L, Huang M, Zhao J. **Intermittent caloric restriction with a modified fasting-mimicking diet ameliorates autoimmunity and promotes recovery in a mouse model of multiple sclerosis**. *J Nutr Biochem.* (2021) **87** 108493. DOI: 10.1016/j.jnutbio.2020.108493
22. Choi IY, Piccio L, Childress P, Bollman B, Ghosh A, Brandhorst S. **A diet mimicking fasting promotes regeneration and reduces autoimmunity and multiple sclerosis symptoms**. *Cell Rep.* (2016) **15** 2136-46. DOI: 10.1016/j.celrep.2016.05.009
23. Kafami L, Raza M, Razavi A, Mirshafiey A, Movahedian M, Khorramizadeh MR. **Intermittent feeding attenuates clinical course of experimental autoimmune encephalomyelitis in C57BL/6 mice**. *Avicenna J Med Biotechnol.* (2010) **2** 47-52. PMID: 23407146
24. Razeghi Jahromi S, Ghaemi A, Alizadeh A, Sabetghadam F, Moradi Tabriz H, Togha M. **Effects of intermittent fasting on experimental autoimune encephalomyelitis in C57BL/6 mice**. *Iran J Allergy Asthma Immunol.* (2016) **15** 212-9. PMID: 27424136
25. Duking T, Spieth L, Berghoff SA, Piepkorn L, Schmidke AM, Mitkovski M. **Ketogenic diet uncovers differential metabolic plasticity of brain cells**. *Sci Adv* (2022) **8** eabo7639. DOI: 10.1126/sciadv.abo7639
26. Kim DY, Hao J, Liu R, Turner G, Shi FD, Rho JM. **Inflammation-mediated memory dysfunction and effects of a ketogenic diet in a murine model of multiple sclerosis**. *PLoS ONE.* (2012) **7** e35476. DOI: 10.1371/journal.pone.0035476
27. Martin-McGill KJ, Jackson CF, Bresnahan R, Levy RG, Cooper PN. **Ketogenic diets for drug-resistant epilepsy**. *Cochrane Database Syst Rev.* (2018) **11** CD001903. DOI: 10.1002/14651858.CD001903.pub4
28. Wheless JW. **History of the ketogenic diet**. *Epilepsia.* (2008) **49** 3-5. DOI: 10.1111/j.1528-1167.2008.01821.x
29. Abbasi J. **Interest in the ketogenic diet grows for weight loss and type 2 diabetes**. *JAMA.* (2018) **319** 215-7. DOI: 10.1001/jama.2017.20639
30. Athinarayanan SJ, Adams RN, Hallberg SJ, McKenzie AL, Bhanpuri NH, Campbell WW. **Long-term effects of a novel continuous remote care intervention including nutritional ketosis for the management of type 2 diabetes: a 2-year non-randomized clinical trial**. *Front Endocrinol.* (2019) **10** 348. DOI: 10.3389/fendo.2019.00348
31. Saslow LR, Daubenmier JJ, Moskowitz JT, Kim S, Murphy EJ, Phinney SD. **Twelve-month outcomes of a randomized trial of a moderate-carbohydrate versus very low-carbohydrate diet in overweight adults with type 2 diabetes mellitus or prediabetes**. *Nutr Diabetes.* (2017) **7** 304. DOI: 10.1038/s41387-017-0006-9
32. Storoni M, Plant GT. **The therapeutic potential of the ketogenic diet in treating progressive multiple sclerosis**. *Mult Scler Int.* (2015) **2015** 681289. DOI: 10.1155/2015/681289
33. Wilhelm C, Surendar J, Karagiannis F. **Enemy or ally? Fasting as an essential regulator of immune responses**. *Trends Immunol.* (2021) **42** 389-400. DOI: 10.1016/j.it.2021.03.007
34. Bahr LS, Bock M, Liebscher D, Bellmann-Strobl J, Franz L, Pruss A. **Ketogenic diet and fasting diet as Nutritional Approaches in Multiple Sclerosis (NAMS): protocol of a randomized controlled study**. *Trials.* (2020) **21** 3. DOI: 10.1186/s13063-019-3928-9
35. Larabee CM, Desai S, Agasing A, Georgescu C, Wren JD, Axtell RC. **Loss of Nrf2 exacerbates the visual deficits and optic neuritis elicited by experimental autoimmune encephalomyelitis**. *Mol Vis.* (2016) **22** 1503-13. PMID: 28050123
36. Larabee CM, Hu Y, Desai S, Georgescu C, Wren JD, Axtell RC. **Myelin-specific Th17 cells induce severe relapsing optic neuritis with irreversible loss of retinal ganglion cells in C57BL/6 mice**. *Mol Vis.* (2016) **22** 332-41. PMID: 27122964
37. Zyla K, Larabee CM, Georgescu C, Berkley C, Reyna T, Plafker SM. **Dimethyl fumarate mitigates optic neuritis**. *Mol Vis.* (2019) **25** 446-61. PMID: 31523122
38. Bligh EG, Dyer WJ. **A rapid method of total lipid extraction and purification**. *Can J Biochem Physiol.* (1959) **37** 911-7. DOI: 10.1139/o59-099
39. Li F, Marchette LD, Brush RS, Elliott MH, Le YZ, Henry KA. **DHA does not protect ELOVL4 transgenic mice from retinal degeneration**. *Mol Vis.* (2009) **15** 1185-93. PMID: 19536303
40. Agbaga MP, Merriman DK, Brush RS, Lydic TA, Conley SM, Naash MI. **Differential composition of DHA and very-long-chain PUFAs in rod and cone photoreceptors**. *J Lipid Res.* (2018) **59** 1586-96. DOI: 10.1194/jlr.M082495
41. Yu M, Benham A, Logan S, Brush RS, Mandal MNA, Anderson RE. **ELOVL4 protein preferentially elongates 20:5n3 to very long chain PUFAs over 20:4n6 and 22:6n3**. *J Lipid Res.* (2012) **53** 494-504. DOI: 10.1194/jlr.M021386
42. Axtell RC, de Jong BA, Boniface K, van der Voort LF, Bhat R, De Sarno P. **T helper type 1 and 17 cells determine efficacy of interferon-beta in multiple sclerosis and experimental encephalomyelitis**. *Nat Med.* (2010) **16** 406-12. DOI: 10.1038/nm.2110
43. Laird NM, Ware JH. **Random-effects models for longitudinal data**. *Biometrics.* (1982) **38** 963-74. DOI: 10.2307/2529876
44. Lindstrom ML, Bates DM. **Nonlinear mixed effects models for repeated measures data**. *Biometrics.* (1990) **46** 673-87. DOI: 10.2307/2532087
45. Goldberg EL, Asher JL, Molony RD, Shaw AC, Zeiss CJ, Wang C. **beta-hydroxybutyrate deactivates neutrophil NLRP3 inflammasome to relieve gout flares**. *Cell Rep.* (2017) **18** 2077-87. DOI: 10.1016/j.celrep.2017.02.004
46. Salberg S, Weerwardhena H, Collins R, Reimer RA, Mychasiuk R. **The behavioural and pathophysiological effects of the ketogenic diet on mild traumatic brain injury in adolescent rats**. *Behav Brain Res.* (2019) **376** 112225. DOI: 10.1016/j.bbr.2019.112225
47. Haghikia A, Jorg S, Duscha A, Berg J, Manzel A, Waschbisch A. **Dietary fatty acids directly impact central nervous system autoimmunity**. *Immunity.* (2015) **43** 817-29. DOI: 10.1016/j.immuni.2015.09.007
48. Sanna V, Di Giacomo A, La Cava A, Lechler RI, Fontana S, Zappacosta S. **Leptin surge precedes onset of autoimmune encephalomyelitis and correlates with development of pathogenic T cell responses**. *J Clin Invest.* (2003) **111** 241-50. DOI: 10.1172/JCI200316721
49. Miller VJ, Villamena FA, Volek JS. **Nutritional ketosis and mitohormesis: potential implications for mitochondrial function and human health**. *J Nutr Metab.* (2018) **2018** 5157645. DOI: 10.1155/2018/5157645
50. Kale N. **Optic neuritis as an early sign of multiple sclerosis**. *Eye Brain.* (2016) **8** 195-202. DOI: 10.2147/EB.S54131
51. Papp V, Magyari M, Aktas O, Berger T, Broadley SA, Cabre P. **Worldwide incidence and prevalence of neuromyelitis optica: a systematic review**. *Neurology.* (2021) **96** 59-77. DOI: 10.1212/WNL.0000000000011153
52. Reindl M, Rostasy K. **MOG antibody-associated diseases**. *Neurol Neuroimmunol Neuroinflamm.* (2015) **2** e60. DOI: 10.1212/NXI.0000000000000060
53. **Experience of the optic neuritis treatment trial. Optic Neuritis Study Group**. *Arch Ophthalmol* (1991) **109** 1673-8. DOI: 10.1001/archopht.1991.01080120057025
54. Atkins EJ, Biousse V, Newman NJ. **The natural history of optic neuritis**. *Rev Neurol Dis.* (2006) **3** 45-56. DOI: 10.1055/s-2007-979683
55. Guo J, Li B, Wang J, Guo R, Tian Y, Song S. **Protective effect and mechanism of nicotinamide adenine dinucleotide against optic neuritis in mice with experimental autoimmune encephalomyelitis**. *Int Immunopharmacol.* (2021) **98** 107846. DOI: 10.1016/j.intimp.2021.107846
56. Hobom M, Storch MK, Weissert R, Maier K, Radhakrishnan A, Kramer B. **Mechanisms and time course of neuronal degeneration in experimental autoimmune encephalomyelitis**. *Brain Pathol.* (2004) **14** 148-57. DOI: 10.1111/j.1750-3639.2004.tb00047.x
57. Sattler MB, Merkler D, Maier K, Stadelmann C, Ehrenreich H, Bahr M. **Neuroprotective effects and intracellular signaling pathways of erythropoietin in a rat model of multiple sclerosis**. *Cell Death Differ.* (2004) **11** S181-92. DOI: 10.1038/sj.cdd.4401504
58. Shindler KS, Guan Y, Ventura E, Bennett J, Rostami A. **Retinal ganglion cell loss induced by acute optic neuritis in a relapsing model of multiple sclerosis**. *Mult Scler.* (2006) **12** 526-32. DOI: 10.1177/1352458506070629
59. Shindler KS, Ventura E, Dutt M, Rostami A. **Inflammatory demyelination induces axonal injury and retinal ganglion cell apoptosis in experimental optic neuritis**. *Exp Eye Res.* (2008) **87** 208-13. DOI: 10.1016/j.exer.2008.05.017
60. Green AJ, McQuaid S, Hauser SL, Allen IV, Lyness R. **Ocular pathology in multiple sclerosis: retinal atrophy and inflammation irrespective of disease duration**. *Brain* (2010) **133** 1591-601. DOI: 10.1093/brain/awq080
61. Zahoor I, Giri S. **Specialized pro-resolving lipid mediators: emerging therapeutic candidates for multiple sclerosis**. *Clin Rev Allergy Immunol.* (2021) **60** 147-63. DOI: 10.1007/s12016-020-08796-4
62. Hammarstrom S. **Conversion of 5,8,11-eicosatrienoic acid to leukotrienes C3 and D3**. *J Biol Chem.* (1981) **256** 2275-9. DOI: 10.1016/S0021-9258(19)69773-5
63. Hulbert AJ, Kelly MA, Abbott SK. **Polyunsaturated fats, membrane lipids and animal longevity**. *J Comp Physiol B.* (2014) **184** 149-66. DOI: 10.1007/s00360-013-0786-8
64. Burdge GC, Wootton SA. **Conversion of alpha-linolenic acid to eicosapentaenoic, docosapentaenoic and docosahexaenoic acids in young women**. *Br J Nutr.* (2002) **88** 411-20. DOI: 10.1079/BJN2002689
65. El-behi M, Rostami A, Ciric B. **Current views on the roles of Th1 and Th17 cells in experimental autoimmune encephalomyelitis**. *J Neuroimmune Pharmacol* (2010) **5** 189-97. DOI: 10.1007/s11481-009-9188-9
66. Imler TJ, Petro TM. **Decreased severity of experimental autoimmune encephalomyelitis during resveratrol administration is associated with increased IL-17+IL-10+ T cells, CD4(-) IFN-gamma+ cells, and decreased macrophage IL-6 expression**. *Int Immunopharmacol.* (2009) **9** 134-43. DOI: 10.1016/j.intimp.2008.10.015
67. Garbow JR, Doherty JM, Schugar RC, Travers S, Weber ML, Wentz AE. **Hepatic steatosis, inflammation, and ER stress in mice maintained long term on a very low-carbohydrate ketogenic diet**. *Am J Physiol Gastrointest Liver Physiol.* (2011) **300** G956-67. DOI: 10.1152/ajpgi.00539.2010
68. Jornayvaz FR, Jurczak MJ, Lee HY, Birkenfeld AL, Frederick DW, Zhang D. **A high-fat, ketogenic diet causes hepatic insulin resistance in mice, despite increasing energy expenditure and preventing weight gain**. *Am J Physiol Endocrinol Metab.* (2010) **299** E808-15. DOI: 10.1152/ajpendo.00361.2010
69. Zhang D, Jin W, Wu R, Li J, Park SA, Tu E. **High glucose intake exacerbates autoimmunity through reactive-oxygen-species-mediated TGF-beta cytokine activation**. *Immunity.* (2019) **51** 671-81. DOI: 10.1016/j.immuni.2019.08.001
70. Westman EC, Yancy WS, Mavropoulos JC, Marquart M, McDuffie JR. **The effect of a low-carbohydrate, ketogenic diet versus a low-glycemic index diet on glycemic control in type 2 diabetes mellitus**. *Nutr Metab.* (2008) **5** 36. DOI: 10.1186/1743-7075-5-36
71. Forsythe CE, Phinney SD, Fernandez ML, Quann EE, Wood RJ, Bibus DM. **Comparison of low fat and low carbohydrate diets on circulating fatty acid composition and markers of inflammation**. *Lipids.* (2008) **43** 65-77. DOI: 10.1007/s11745-007-3132-7
72. Hallberg SJ, McKenzie AL, Williams PT, Bhanpuri NH, Peters AL, Campbell WW. **Effectiveness and safety of a novel care model for the management of type 2 diabetes at 1 year: an open-label, non-randomized, controlled study**. *Diabetes Ther.* (2018) **9** 583-612. DOI: 10.1007/s13300-018-0373-9
73. Hyde PN, Sapper TN, Crabtree CD, LaFountain RA, Bowling ML, Buga A. **Dietary carbohydrate restriction improves metabolic syndrome independent of weight loss**. *JCI Insight* (2019) **4** e128308. DOI: 10.1172/jci.insight.128308
74. Volek JS, Phinney SD, Krauss RM, Johnson RJ, Saslow LR, Gower B. **Alternative dietary patterns for americans: low-carbohydrate diets**. *Nutrients* (2021) **13** 3299. DOI: 10.3390/nu13103299
75. Bisht B, Darling WG, Grossmann RE, Shivapour ET, Lutgendorf SK, Snetselaar LG. **A multimodal intervention for patients with secondary progressive multiple sclerosis: feasibility and effect on fatigue**. *J Altern Complement Med.* (2014) **20** 347-55. DOI: 10.1089/acm.2013.0188
76. Bisht B, Darling WG, Shivapour ET, Lutgendorf SK, Snetselaar LG, Chenard CA. **Multimodal intervention improves fatigue and quality of life in subjects with progressive multiple sclerosis: a pilot study**. *Degener Neurol Neuromuscul Dis.* (2015) **5** 19-35. DOI: 10.2147/DNND.S76523
77. Bisht B, Darling WG, White EC, White KA, Shivapour ET, Zimmerman MB. **Effects of a multimodal intervention on gait and balance of subjects with progressive multiple sclerosis: a prospective longitudinal pilot study**. *Degener Neurol Neuromuscul Dis.* (2017) **7** 79-93. DOI: 10.2147/DNND.S128872
78. Fellows Maxwell K, Wahls T, Browne RW, Rubenstein L, Bisht B, Chenard CA. **Lipid profile is associated with decreased fatigue in individuals with progressive multiple sclerosis following a diet-based intervention: results from a pilot study**. *PLoS ONE.* (2019) **14** e0218075. DOI: 10.1371/journal.pone.0218075
79. Irish AK, Erickson CM, Wahls TL, Snetselaar LG, Darling WG. **Randomized control trial evaluation of a modified Paleolithic dietary intervention in the treatment of relapsing-remitting multiple sclerosis: a pilot study**. *Degener Neurol Neuromuscul Dis.* (2017) **7** 1-18. DOI: 10.2147/DNND.S116949
80. Lee JE, Bisht B, Hall MJ, Rubenstein LM, Louison R, Klein DT. **A multimodal, nonpharmacologic intervention improves mood and cognitive function in people with multiple sclerosis**. *J Am Coll Nutr.* (2017) **36** 150-68. DOI: 10.1080/07315724.2016.1255160
81. Wahls TL, Titcomb TJ, Bisht B, Eyck PT, Rubenstein LM, Carr LJ. **Impact of the Swank and Wahls elimination dietary interventions on fatigue and quality of life in relapsing-remitting multiple sclerosis: the WAVES randomized parallel-arm clinical trial**. *Mult Scler J Exp Transl Clin.* (2021) **7** 20552173211035399. DOI: 10.1177/20552173211035399
82. Bjornevik K, Chitnis T, Ascherio A, Munger KL. **Polyunsaturated fatty acids and the risk of multiple sclerosis**. *Mult Scler.* (2017) **23** 1830-8. DOI: 10.1177/1352458517691150
83. Penkert H, Bertrand A, Tiwari V, Breimann S, Muller SA, Jordan PM. **Proteomic and lipidomic profiling of demyelinating lesions identifies fatty acids as modulators in lesion recovery**. *Cell Rep.* (2021) **37** 109898. DOI: 10.1016/j.celrep.2021.109898
84. Poisson LM, Suhail H, Singh J, Datta I, Denic A, Labuzek K. **Untargeted plasma metabolomics identifies endogenous metabolite with drug-like properties in chronic animal model of multiple sclerosis**. *J Biol Chem.* (2015) **290** 30697-712. DOI: 10.1074/jbc.M115.679068
85. Kooij G, Troletti CD, Leuti A, Norris PC, Riley I, Albanese M. **Specialized pro-resolving lipid mediators are differentially altered in peripheral blood of patients with multiple sclerosis and attenuate monocyte and blood-brain barrier dysfunction**. *Haematologica.* (2020) **105** 2056-70. DOI: 10.3324/haematol.2019.219519
86. Villoslada P, Alonso C, Agirrezabal I, Kotelnikova E, Zubizarreta I, Pulido-Valdeolivas I. **Metabolomic signatures associated with disease severity in multiple sclerosis**. *Neurol Neuroimmunol Neuroinflamm.* (2017) **4** e321. DOI: 10.1212/NXI.0000000000000321
87. Rumble JM, Huber AK, Krishnamoorthy G, Srinivasan A, Giles DA, Zhang X. **Neutrophil-related factors as biomarkers in EAE and MS**. *J Exp Med.* (2015) **212** 23-35. DOI: 10.1084/jem.20141015
88. Aube B, Levesque SA, Pare A, Chamma E, Kebir H, Gorina R. **Neutrophils mediate blood-spinal cord barrier disruption in demyelinating neuroinflammatory diseases**. *J Immunol.* (2014) **193** 2438-54. DOI: 10.4049/jimmunol.1400401
89. Jiang W, St-Pierre S, Roy P, Morley BJ, Hao J, Simard AR. **Infiltration of CCR2+Ly6Chigh proinflammatory monocytes and neutrophils into the central nervous system is modulated by nicotinic acetylcholine receptors in a model of multiple sclerosis**. *J Immunol.* (2016) **196** 2095-108. DOI: 10.4049/jimmunol.1501613
90. Levesque SA, Pare A, Mailhot B, Bellver-Landete V, Kebir H, Lecuyer MA. **Myeloid cell transmigration across the CNS vasculature triggers IL-1beta-driven neuroinflammation during autoimmune encephalomyelitis in mice**. *J Exp Med.* (2016) **213** 929-49. DOI: 10.1084/jem.20151437
91. Steinbach K, Piedavent M, Bauer S, Neumann JT, Friese MA. **Neutrophils amplify autoimmune central nervous system infiltrates by maturing local APCs**. *J Immunol.* (2013) **191** 4531-9. DOI: 10.4049/jimmunol.1202613
92. Yan Z, Yang W, Parkitny L, Gibson SA, Lee KS, Collins F. **Deficiency of Socs3 leads to brain-targeted EAE**. *JCI Insight* (2019) **5** e126520. DOI: 10.1172/jci.insight.126520
93. Ghafouri-Fard S, Honarmand K, Taheri M. **A comprehensive review on the role of chemokines in the pathogenesis of multiple sclerosis**. *Metab Brain Dis.* (2021) **36** 375-406. DOI: 10.1007/s11011-020-00648-6
94. Tsutsui M, Hirase R, Miyamura S, Nagayasu K, Nakagawa T, Mori Y. **TRPM2 exacerbates central nervous system inflammation in experimental autoimmune encephalomyelitis by increasing production of CXCL2 chemokines**. *J Neurosci.* (2018) **38** 8484-95. DOI: 10.1523/JNEUROSCI.2203-17.2018
95. Carlson T, Kroenke M, Rao P, Lane TE, Segal B. **The Th17-ELR+ CXC chemokine pathway is essential for the development of central nervous system autoimmune disease**. *J Exp Med.* (2008) **205** 811-23. DOI: 10.1084/jem.20072404
96. Matejuk A, Dwyer J, Ito A, Bruender Z, Vandenbark AA, Offner H. **Effects of cytokine deficiency on chemokine expression in CNS of mice with EAE**. *J Neurosci Res.* (2002) **67** 680-8. DOI: 10.1002/jnr.10156
97. Stoolman JS, Duncker PC, Huber AK, Giles DA, Washnock-Schmid JM, Soulika AM. **An IFNgamma/CXCL2 regulatory pathway determines lesion localization during EAE**. *J Neuroinflammation.* (2018) **15** 208. DOI: 10.1186/s12974-018-1237-y
98. Huang J, Khademi M, Fugger L, Lindhe O, Novakova L, Axelsson M. **Inflammation-related plasma and CSF biomarkers for multiple sclerosis**. *Proc Natl Acad Sci U S A.* (2020) **117** 12952-60. DOI: 10.1073/pnas.1912839117
99. Adzemovic MZ, Ockinger J, Zeitelhofer M, Hochmeister S, Beyeen AD, Paulson A. **Expression of Ccl11 associates with immune response modulation and protection against neuroinflammation in rats**. *PLoS ONE.* (2012) **7** e39794. DOI: 10.1371/journal.pone.0039794
100. Ruppova K, Lim JH, Fodelianaki G, August A, Neuwirth A. **Eosinophils are dispensable for development of MOG(35-55)-induced experimental autoimmune encephalomyelitis in mice**. *Immunol Lett.* (2021) **239** 72-6. DOI: 10.1016/j.imlet.2021.09.001
101. Nazarinia D, Behzadifard M, Gholampour J, Karimi R, Gholampour M. **Eotaxin-1 (CCL11) in neuroinflammatory disorders and possible role in COVID-19 neurologic complications**. *Acta Neurol Belg.* (2022) **122** 865-9. DOI: 10.1007/s13760-022-01984-3
102. Erta M, Quintana A, Hidalgo J. **Interleukin-6, a major cytokine in the central nervous system**. *Int J Biol Sci.* (2012) **8** 1254-66. DOI: 10.7150/ijbs.4679
103. Maimone D, Guazzi GC, Annunziata P. **IL-6 detection in multiple sclerosis brain**. *J Neurol Sci.* (1997) **146** 59-65. DOI: 10.1016/S0022-510X(96)00283-3
104. Stampanoni Bassi M, Iezzi E, Drulovic J, Pekmezovic T, Gilio L, Furlan R. **IL-6 in the cerebrospinal fluid signals disease activity in multiple sclerosis**. *Front Cell Neurosci.* (2020) **14** 120. DOI: 10.3389/fncel.2020.00120
105. Setiadi AF, Abbas AR, Jeet S, Wong K, Bischof A, Peng I. **IL-17A is associated with the breakdown of the blood-brain barrier in relapsing-remitting multiple sclerosis**. *J Neuroimmunol.* (2019) **332** 147-54. DOI: 10.1016/j.jneuroim.2019.04.011
106. Bettelli E, Carrier Y, Gao W, Korn T, Strom TB, Oukka M. **Reciprocal developmental pathways for the generation of pathogenic effector TH17 and regulatory T cells**. *Nature.* (2006) **441** 235-8. DOI: 10.1038/nature04753
107. Mangan PR, Harrington LE, O'Quinn DB, Helms WS, Bullard DC, Elson CO. **Transforming growth factor-beta induces development of the T(H)17 lineage**. *Nature.* (2006) **441** 231-4. DOI: 10.1038/nature04754
108. Serada S, Fujimoto M, Mihara M, Koike N, Ohsugi Y, Nomura S. **IL-6 blockade inhibits the induction of myelin antigen-specific Th17 cells and Th1 cells in experimental autoimmune encephalomyelitis**. *Proc Natl Acad Sci U S A.* (2008) **105** 9041-6. DOI: 10.1073/pnas.0802218105
109. Veldhoen M, Hocking RJ, Atkins CJ, Locksley RM, Stockinger B. **TGFbeta in the context of an inflammatory cytokine milieu supports de novo differentiation of IL-17-producing T cells**. *Immunity.* (2006) **24** 179-89. DOI: 10.1016/j.immuni.2006.01.001
110. Ma S, Huang Q, Tominaga T, Liu C, Suzuki K. **An 8-week ketogenic diet alternated interleukin-6, ketolytic and lipolytic gene expression, and enhanced exercise capacity in mice**. *Nutrients* (2018) **10** 1696. DOI: 10.3390/nu10111696
111. Nakamura K, Tonouchi H, Sasayama A, Ashida K. **A ketogenic formula prevents tumor progression and cancer cachexia by attenuating systemic inflammation in colon 26 tumor-bearing mice**. *Nutrients* (2018) **10** 206. DOI: 10.3390/nu10020206
112. Nandivada P, Fell GL, Pan AH, Nose V, Ling PR, Bistrian BR. **Eucaloric ketogenic diet reduces hypoglycemia and inflammation in mice with endotoxemia**. *Lipids.* (2016) **51** 703-14. DOI: 10.1007/s11745-016-4156-7
113. Norwitz NG, Winwood R, Stubbs BJ, D'Agostino DP, Barnes PJ. **Case report: ketogenic diet is associated with improvements in chronic obstructive pulmonary disease**. *Front Med.* (2021) **8** 699427. DOI: 10.3389/fmed.2021.699427
114. Thambi M, Nathan J, Bailur S, Unnikrishnan MK, Ballal M, Radhakrishnan K. **Is the antiseizure effect of ketogenic diet in children with drug-resistant epilepsy mediated through proinflammatory cytokines?**. *Epilepsy Res.* (2021) **176** 106724. DOI: 10.1016/j.eplepsyres.2021.106724
115. Yang X, Cheng B. **Neuroprotective and anti-inflammatory activities of ketogenic diet on MPTP-induced neurotoxicity**. *J Mol Neurosci.* (2010) **42** 145-53. DOI: 10.1007/s12031-010-9336-y
116. Zhu Y, Tang X, Cheng Z, Dong Q, Ruan G. **The anti-inflammatory effect of preventive intervention with ketogenic diet mediated by the histone acetylation of mGluR5 promotor region in rat Parkinson's disease model: a dual-tracer PET study**. *Parkinsons Dis.* (2022) **2022** 3506213. DOI: 10.1155/2022/3506213
117. Bertoli S, Neri IG, Trentani C, Ferraris C, De Amicis R, Battezzati A. **Short-term effects of ketogenic diet on anthropometric parameters, body fat distribution, and inflammatory cytokine production in GLUT1 deficiency syndrome**. *Nutrition.* (2015) **31** 981-7. DOI: 10.1016/j.nut.2015.02.017
118. Fraser DA, Thoen J, Djoseland O, Forre O, Kjeldsen-Kragh J. **Serum levels of interleukin-6 and dehydroepiandrosterone sulphate in response to either fasting or a ketogenic diet in rheumatoid arthritis patients**. *Clin Exp Rheumatol.* (2000) **18** 357-62. PMID: 10895373
119. Vidali S, Aminzadeh-Gohari S, Feichtinger RG, Vatrinet R, Koller A, Locker F. **The ketogenic diet is not feasible as a therapy in a CD-1 nu/nu mouse model of renal cell carcinoma with features of Stauffer's syndrome**. *Oncotarget.* (2017) **8** 57201-15. DOI: 10.18632/oncotarget.19306
120. Longo VD, Panda S. **Fasting, circadian rhythms, and time-restricted feeding in healthy lifespan**. *Cell Metab.* (2016) **23** 1048-59. DOI: 10.1016/j.cmet.2016.06.001
121. Luis D, Zlatkis K, Comenge B, Garcia Z, Navarro JF, Lorenzo V. **Dietary quality and adherence to dietary recommendations in patients undergoing hemodialysis**. *J Ren Nutr.* (2016) **26** 190-5. DOI: 10.1053/j.jrn.2015.11.004
122. Mellor R, Saunders-Dow E, Mayr HL. **Scope of use and effectiveness of dietary interventions for improving health-related outcomes in veterans: a systematic review**. *Nutrients* (2022) **14** 2094. DOI: 10.3390/nu14102094
|
---
title: Coenzyme Q10 supplementation improves the motor function of middle-aged mice
by restoring the neuronal activity of the motor cortex
authors:
- Ritsuko Inoue
- Masami Miura
- Shuichi Yanai
- Hiroshi Nishimune
journal: Scientific Reports
year: 2023
pmcid: PMC10017826
doi: 10.1038/s41598-023-31510-1
license: CC BY 4.0
---
# Coenzyme Q10 supplementation improves the motor function of middle-aged mice by restoring the neuronal activity of the motor cortex
## Abstract
Physiological aging causes motor function decline and anatomical and biochemical changes in the motor cortex. We confirmed that middle-aged mice at 15–18 months old show motor function decline, which can be restored to the young adult level by supplementing with mitochondrial electron transporter coenzyme Q10 (CoQ10) as a water-soluble nanoformula by drinking water for 1 week. CoQ10 supplementation concurrently improved brain mitochondrial respiration but not muscle strength. Notably, we identified an age-related decline in field excitatory postsynaptic potential (fEPSP) amplitude in the pathway from layers II/III to V of the primary motor area of middle-aged mice, which was restored to the young adult level by supplementing with CoQ10 for 1 week but not by administering CoQ10 acutely to brain slices. Interestingly, CoQ10 with high-frequency stimulation induced NMDA receptor-dependent long-term potentiation (LTP) in layer V of the primary motor cortex of middle-aged mice. Importantly, the fEPSP amplitude showed a larger input‒output relationship after CoQ10-dependent LTP expression. These data suggest that CoQ10 restores the motor function of middle-aged mice by improving brain mitochondrial function and the basal fEPSP level of the motor cortex, potentially by enhancing synaptic plasticity efficacy. Thus, CoQ10 supplementation may ameliorate the age-related decline in motor function in humans.
## Introduction
The age-related decrease in motor function can be caused by a loss of muscle mass and strength (sarcopenia), denervation of neuromuscular junctions (NMJs), loss of motor neurons in the spinal cord, and reduced function of the brain motor cortex. In mice, a decline in motor function was observed in middle-aged mice (15 months old); earlier than the drop in survival rate that typically occurs at approximately 24 months of age1–3. In the human motor cortex, physiological aging causes cortical atrophy, altered excitability, and decreased neurotransmitter levels 4–6. Voluntary activation of skeletal muscles is impaired during aging, especially in elderly individuals who are weak or in poor physical condition5. Elderly individuals show a decrease in the firing rate of lower motor neurons, which may be related, at least in part, to decreased activity of the motor cortex. In rodents, middle-aged mice show motor function impairment and increased phosphorylation of α-synuclein and a decreased level of vesicular glutamate transporter 1 (VGluT1) in motor cortex compared to those of young adult mice1,7. These impairments in the motor cortex may be part of the underlying mechanism that leads to age-related motor function decline.
Brain aging also causes deficits in the mitochondrial oxidative phosphorylation system, producing ATP necessary to fulfill brain neuronal functions8. The basal ganglia putamen of old rhesus monkeys showed decreased mitochondrial functions of ATP synthesis and calcium buffering, which correlated with age-related motor deficits8. Interestingly, mitochondrial function measured by peroxide production was higher in synaptic mitochondria than in nonsynaptic mitochondria in rat brains. Furthermore, mitochondrial respiration function decreased significantly with age only in synaptic mitochondria but not in nonsynaptic mitochondria among 14- and 17-month-old mice compared to those of 3-month-old mice9,10. These data suggest an age-related functional decline of mitochondria located at synapses in the brain. Mechanistically, mitochondrial oxidative phosphorylation requires the electron transporter coenzyme Q. In mice, coenzyme Q9 and Q10 are used to transport electrons from complexes I and II to complex III for ATP synthesis11–13. The level of coenzyme Q10 (CoQ10) decreases during aging in rodents and humans1,14–16. Interestingly, exogenous administration of CoQ10 ameliorated motor impairment and the brain mitochondrial respiration rate in aged mice1,17.
These studies demonstrate the age-related decline in motor functions, motor cortex functions, and synaptic mitochondrial functions. However, further research is needed to reveal what kind of electrophysiological impairments underlie the age-related decline in motor functions and to develop intervention methods to rescue the motor function and motor cortex neuronal activity of elderly or aged animals. This study tested whether the age-related decline in motor functions could be reversed by supplementing the mitochondrial component that decreases during aging. CoQ10 was supplemented by drinking water in middle-aged mice, which led to motor function improvements. To investigate the mechanism activated by CoQ10, we measured muscle strength, brain mitochondrial function, and electrophysiological activity in the motor cortex of middle-aged mice supplemented with CoQ10.
## CoQ10 supplementation reversed age-related decline in motor function
We compared the motor function of young adult mice (6 months old) and middle-aged mice (15–16 months old) by performing the pole test following supplementation with water-soluble nanoformula-type CoQ10 by drinking water (150 μM, 40SP, Petroeuroasia) for 10–13 days or without supplementation (Fig. 1a). The pole test is used to evaluate motor coordination deficit18–20 by measuring the time required for mice to orient their body and feet completely downward at the top of a vertical pole (T-turn) and the total time to descend to the floor of the experimental cage (T-total). The 15-month-old middle-aged mice required a significantly longer time for the T-turn than the young adult mice (young adult control, 1.50 ± 0.07 s, middle-aged control, 2.24 ± 0.13 s). CoQ10 supplementation in middle-aged mice improved motor function (T-turn) by $25.76\%$ to a level similar to that of the young adult controls (Fig. 1a left). There was a significant interaction between CoQ10 supplementation and age, a significant main effect of supplementation, and a significant main effect of age (Fig. 1a left; the interaction between supplementation and age $$p \leq 0.0296$$, F [1,76] = 4.919; the main effect of supplementation $$p \leq 0.0009$$, F [1, 76] = 11.95; the main effect of age $p \leq 0.0001$, F [1, 76] = 25.77, two-way ANOVA; young adult control compared to middle-aged control $p \leq 0.0001$; young adult CoQ10 compared to middle-aged control, $p \leq 0.0001$; middle-aged control compared to middle-aged CoQ10, $$p \leq 0.0008$$, Bonferroni’s multiple comparison test). Measurements of the total time (T-total) revealed a significant main effect of supplementation but did not show any main effect of age or a supplementation-age interaction (Fig. 1a right; the interaction between supplementation and age $$p \leq 0.1386$$, F [1,76] = 2.240; the main effect of supplementation $$p \leq 0.0043$$, F [1, 76] = 8.686, two-way ANOVA). The age-dependent decline in motor function in the pole test and recovery by water-soluble nanoformula-type CoQ10 supplementation is consistent with a previous study1.Figure 1CoQ10 supplementation by drinking water restored motor function in middle-aged mice. Motor function was evaluated by the time to complete each aspect of the pole test. A T-turn represents the time required for a mouse to orient the body and feet downward at the top of a vertical pole. T-total represents the time required for the mouse to complete the T-turn and climb down to the experimental cage floor. ( a) Pole test latency of young adult and middle-aged mice treated with drinking water supplemented with or without CoQ10 for 10–13 days. Young adult mice (6 months old; control, $$n = 20$$; CoQ10, $$n = 20$$) and middle-aged mice (15 months old; control, $$n = 20$$; CoQ10, $$n = 20$$) were used in the test (****$p \leq 0.0001$; ***$p \leq 0.001$, two-way ANOVA with Bonferroni's multiple comparison test). ( b) Wire hanging latency of young adult and middle-aged mice supplemented with CoQ10 for 1 week compared to age-matched controls. The latency to fall did not show a difference with CoQ10 supplementation but was shortened by aging ($$n = 20$$ in each group of young adult control and CoQ10 and middle-aged control and CoQ10; no significant interaction, two-way ANOVA). ( c) Middle-aged mice supplemented with CoQ10 for 33–36 days (approximately 1 month) showed improvement of pole test latency compared to age-matched controls (16-month-old; control, $$n = 20$$, CoQ10, $$n = 19$$; ***$p \leq 0.001$, Welch's t test). ( d) Wire hanging latency of middle-aged mice supplemented with CoQ10 for 1 month compared to age-matched controls (16-month-old; control, $$n = 20$$; CoQ10, $$n = 19$$; no significant difference, Welch's t test). Values are expressed as the mean ± standard error of the mean (SEM) of independent experimental groups. For details of the data, see Supplementary Table 3.
In this study, we also tested the effect of CoQ10 supplementation in young adult mice to reveal whether the beneficial effect of CoQ10 is an age-specific effect or a general effect. Interestingly, CoQ10 supplementation did not change the T-turn of young adult mice (Fig. 1a left; young adult control compared to young adult CoQ10, $p \leq 0.9999$, Bonferroni's multiple comparison test) and revealed the significant main effect of supplementation on the T-total without the main effect of age or the supplementation-age interaction (Fig. 1a right). These results revealed that the beneficial effect of CoQ10 supplementation is stronger in middle-aged mice showing a decline in motor function.
We analyzed whether CoQ10 supplementation for 1 week affects muscle strength by measuring the wire hanging latency of young adult and middle-aged mice with or without CoQ10 supplementation. There was no interaction between CoQ10 supplementation and age or main effect of supplementation. However, there was a significant main effect of age, suggesting an age-related decline in muscle strength (Fig. 1b; the interaction between supplementation and age $$p \leq 0.9440$$, F [1, 76] = 0.0050; the main effect of age $p \leq 0.0001$, F [1,76] = 22.42, two-way ANOVA). CoQ10 supplementation did not affect the wire hanging latency of either young adult or middle-aged mice.
Furthermore, we tested the effect of extended CoQ10 supplementation by drinking water for approximately 1 month in middle-aged mice. This longer-term treatment also improved the pole test latency (Fig. 1c; T-turn by $28.62\%$, $$p \leq 0.0002$$, t [34] = 4.253; T-total by $25.94\%$, $$p \leq 0.0004$$, t (36.93) = 3.875, Welch's t test), suggesting that the beneficial effect of CoQ10 supplementation is maintained and there is no desensitization during the first month. In addition, there was no significant difference between the wire hanging latency of middle-aged mice with or without CoQ10 supplementation for 1 month (Fig. 1d; $$p \leq 0.6387$$, t (31.63) = 0.4741, Welch's t test). These results suggest no correlation between muscle strength and the improved motor function caused by CoQ10 supplementation.
## CoQ10 supplementation improved brain mitochondrial respiration in middle-aged mice
A previous study tested the effect of water-soluble nanoformula-type CoQ10 on motor and brain mitochondrial functions in different sets of animals and did not confirm them concurrently in the same animal1. In the current study, water-soluble nanoformula-type CoQ10 was administered to middle-aged mice for more than 1 month, and behavioral tests and measurements of brain mitochondrial function were performed in the same animals. A brain mitochondrial fraction was purified to measure the NADH-dependent (complex I-mediated) mitochondrial oxygen consumption rate (OCR). The brain mitochondrial OCR increased significantly in the CoQ10-supplemented middle-aged mice compared to that of the age-matched non-drug controls (Fig. 2; $$p \leq 0.0459$$, t (22.23) = 2.115, Welch's t test). These results suggested that brain mitochondrial function and motor function were concurrently restored in middle-aged mice supplemented with CoQ10 by drinking water. Figure 2CoQ10 supplementation by drinking water concurrently restored brain mitochondrial respiration in middle-aged mice. The oxygen consumption rate (OCR) of the brain mitochondrial fraction was measured using high-resolution respirometry (Oxygraph-2k). Middle-aged mice were used in this measurement following CoQ10 supplementation for 40–76 days (control, $$n = 17$$; CoQ10, $$n = 17$$; *$p \leq 0.05$, Welch's t test). Values are expressed as the mean ± SEM of independent experimental groups. For details of the data, see Supplementary Table 3.
## CoQ10 supplementation restored motor cortex neuronal activity
CoQ10 supplementation improved the motor function of middle-aged mice without enhancing muscle strength. Therefore, we investigated whether water-soluble nanoformula-type CoQ10 (Aqua Q10L10-NF) supplementation affects neuronal activity in the motor cortex. Synaptic mitochondria in the brain show age-related functional decline9,10. Brain mitochondrial activity is critical for maintaining normal synaptic function21.
The motor cortex is anatomically and functionally divided into the primary motor (M1) and the secondary motor (M2) cortices. In this study, we analyzed the two motor regions separately. We prepared cortical slices, stimulated layers II/III, and recorded the responses from layer V neurons to analyze the major intralaminar excitatory connection between layers II/III and V in the motor cortex22–28. fEPSP amplitudes were significantly reduced on average by 35.30 ± $1.85\%$ in the M1 region of middle-aged mice compared to young adult mice in the tested range of 20–80 μA current stimuli (Fig. 3a left). The field excitatory postsynaptic potential (fEPSP) amplitude of young adult mice and middle-aged mice, both without CoQ10 supplementation, showed no interaction but a significant effect of age by two-way repeated-measures ANOVA (Fig. 3a left; the interaction between stimulus intensity and age $$p \leq 0.2778$$, F [6, 246] = 1.257; the main effect of age $$p \leq 0.0231$$, F [1, 41] = 5.572). These results suggested that physiological aging altered neuronal activity in the connection between layers II/III and V in the motor cortex. However, there was no significant interaction or main effect of age by two-way repeated-measures ANOVA in the fEPSP amplitude in the M2 region between young adult and middle-aged mice (Fig. 3a right; the interaction between stimulus intensity and age $$p \leq 0.0941$$, F [6, 198] = 1.835; the main effect of age $$p \leq 0.1107$$, F [1, 33] = 2.687, two-way repeated-measures ANOVA). These results suggest an age-dependent decline in neuronal activity in a motor cortex region-specific manner. Figure 3An age-related decline in synaptic transmission in the motor cortex of middle-aged mice and an improvement with CoQ10 supplementation by drinking water. The field excitatory postsynaptic potentials (fEPSPs) in layer V of the primary motor (M1) and secondary motor (M2) cortices were recorded separately using a multi-electrode array and stimulating the pathway from layers II/III to layer V with a single glass electrode. ( a, left) fEPSP amplitudes in the layer V of M1 region showed a significant decrease with age (young adult, $$n = 14$$ slices from 5 mice; middle-aged, $$n = 29$$ slices from 10 mice; *$p \leq 0.05$). ( a, right) In the M2 region, there were no significant differences between the young adult and middle-aged groups (young adult, $$n = 9$$ slices from 4 mice; middle-aged, $$n = 26$$ slices from 10 mice). ( b, left) The fEPSP amplitudes in the layer V of M1 region of middle-aged mice supplemented with CoQ10 for 1 week were significantly higher than those of age-matched controls in the range of 50–80 μA current stimuli (middle-aged + CoQ10, $$n = 11$$ slices from 5 mice; middle-aged, $$n = 29$$ slices from 10 mice; **$p \leq 0.01$, *$p \leq 0.05$, Bonferroni's multiple comparison test). ( b, right) There were no significant differences between fEPSP amplitudes of CoQ10-treated middle-aged mice and age-matched controls in the layer V of M2 region (middle-aged + CoQ10, $$n = 16$$ slices from 5 mice; middle-aged, $$n = 26$$ slices from 10 mice). The middle-aged control data in (b) are identical to those in (a). Values are expressed as the mean ± SEM of independent experimental groups. Statistical analyses were performed using two-way repeated-measures ANOVA. For details of the data, see Supplementary Table 3.
Next, we tested whether CoQ10 supplementation affects the age-related decline in fEPSP amplitude in middle-aged mice. Cortical slices were prepared from the brains of middle-aged mice supplemented with CoQ10 for 1 week and compared to those of age-matched controls without CoQ10 supplementation. fEPSP amplitudes were significantly increased on average by 61.51 ± $1.12\%$ in the M1 region of CoQ10-supplemented middle-aged mice compared to middle-aged control mice in the range of 50–80 μA current stimuli (Fig. 3a left; Middle-aged compared to Middle-aged CoQ10, 50–80 μA: $$p \leq 0.0076$$–0.0464, Bonferroni's multiple comparison test). The fEPSP amplitude in the layer V of M1 region of the CoQ10-supplemented middle-aged mice and that of the age-matched controls showed a significant interaction and a significant difference between mice with and without CoQ10 supplementation (Fig. 3b left; the interaction between stimulus intensity and supplementation $p \leq 0.0001$, F [6, 228] = 5.064; the main effect of supplementation $$p \leq 0.0127$$, F [1, 38] = 6.843, two-way repeated-measures ANOVA). On the other hand, there was no significant interaction or difference in the fEPSP amplitude in the layer V of M2 region between mice with and without CoQ10 supplementation (Fig. 3b right; the interaction between stimulus intensity and supplementation $$p \leq 0.9032$$, F [6, 240] = 0.3607; the main effect of supplementation $$p \leq 0.9095$$, F [1, 40] = 0.0131, two-way repeated-measures ANOVA). In summary, the connection between layers II/III and layer V in the mouse motor cortex showed an age-related decline in fEPSP amplitude in the layer V of M1 region, which was restored by CoQ10 supplementation by drinking water for 1 week. There was no significant interaction or difference in the fEPSP amplitude in the layer V of M1 region between young adult mice without CoQ10 supplementation and middle-aged mice with CoQ10 supplementation (Data shown in Fig. 3a left and 3b left; the interaction between stimulus intensity and age $$p \leq 0.5174$$, F [6, 138] = 0.8719; the main effect of age $$p \leq 0.8316$$, F [1, 23] = 0.0463, two-way repeated-measures ANOVA).
## CoQ10 supplementation did not affect short-term plasticity in the motor cortex
We assessed short-term synaptic plasticity of the major intralaminar excitatory connection from layers II/III to layer V in the motor cortex22,28 using cortical slices and measuring paired-pulse ratios (PPRs) following paired stimulation at 25- to 500-ms intervals. There was no significant interaction or main effect of age group and stimulus interval on PPRs in the M1 region (Fig. 4a left; the interaction between interval and age $$p \leq 0.8607$$, F [4, 156] = 0.3252; the main effect of age $$p \leq 0.7295$$, F [1, 39] = 0.1213, two-way repeated-measures ANOVA). In the M2 region, there was a significant interaction between age group and stimulus interval but there was not a significant main effect of age, and Bonferroni's multiple comparison showed no significant differences in PPRs between young adult and middle-aged mice among all stimulus intervals (Fig. 4a right; the interaction between interval and age $$p \leq 0.0293$$, F [4, 124] = 2.790; the main effect of age $$p \leq 0.4784$$, F [1, 31] = 0.5150, two-way repeated-measures ANOVA). In addition, there were no significant interactions or differences in PPRs in the M1 and M2 regions of middle-aged mice with and without CoQ10 supplementation for 1 week (Fig. 4b; M1, the interaction between interval and supplementation $$p \leq 0.6793$$, F [4, 132] = 0.5777; the main effect of supplementation $$p \leq 0.7833$$, F [1, 33] = 0.0769; M2, the interaction between interval and supplementation $$p \leq 0.1694$$, F [4, 136] = 1.633; the main effect of supplementation $$p \leq 0.1640$$, F [1, 34] = 2.023, two-way repeated-measures ANOVA). These results suggested that physiological aging and CoQ10 supplementation did not affect the short-term plasticity of the layers II/III to V connection in the motor cortex in these preparations. Figure 4Short-term plasticity in the motor cortex was not affected by age or CoQ10 supplementation. ( a) Paired pulse ratios (PPRs) at various stimulus intervals (25, 50, 100, 200, and 500 ms) were comparable in the M1 region (young adult, $$n = 14$$ slices from 5 mice; middle-aged, $$n = 27$$ slices from 9 mice) and the M2 region (young adult, $$n = 9$$ slices from 4 mice; middle-aged, $$n = 24$$ slices from 10 mice; no significant difference by age). ( b) CoQ10 supplementation did not alter the PPRs in the M1 or M2 regions of middle-aged mice compared to those of the age-matched controls (M1: middle-aged + CoQ10, $$n = 8$$ slices from 4 mice; middle-aged, $$n = 27$$ slices from 9 mice; M2: middle-aged + CoQ10, $$n = 12$$ slices from 5 mice; middle-aged, $$n = 24$$ slices from 10 mice; no significant differences with supplementation). The middle-aged control data in (b) are identical to those in (a). Values are expressed as the mean ± SEM of independent experimental groups. Statistical analyses were performed using two-way repeated-measures ANOVA. For details of the data, see Supplementary Table 3.
## Acute CoQ10 treatment induced NMDA receptor-dependent LTP
Mitochondria have effects on age-related synaptic plasticity29. Therefore, we studied the involvement of plasticity enhancement as a mechanism that augmented the fEPSP amplitude in the motor cortex of CoQ10-supplemented middle-aged mice. A larger fEPSP amplitude is recorded after long-term potentiation (LTP) induction30,31, and LTP can be induced in the motor cortex by motor-skill learning or several induction methods30,32. Furthermore, a larger fEPSP amplitude can be retained for months in the motor cortex after motor-skill learning33.
We tested whether acute CoQ10 administration (50 μM) to brain slices could enhance fEPSP amplitude in the connection between layers II/III and layer V in the motor cortex. Figure 5a shows normalized fEPSP amplitudes in the presence of CoQ10 and during CoQ10 washout. Acute CoQ10 administration alone did not augment fEPSP amplitude in the layer V of M1 region after the treatment (Fig. 5a; middle-aged without stimulation, averaged fEPSP, between − 2 and 0 min and 25–27 min $$p \leq 0.5931$$, t [4] = 0.5799, paired t test). The combination of CoQ10 administration (50 μM, 20–25 min before and during LTP induction) and high-frequency stimulation induced an LTP of 122.39 ± $7.15\%$ of baseline in middle-aged mice (Fig. 5a; middle-aged with stimulation, averaged fEPSP, between − 2 and 0 min and 25–27 min $$p \leq 0.0061$$, t [17] = 3.133, paired t test). The mean normalized fEPSP amplitude at 25–27 min after CoQ10 administration was significantly higher with high-frequency stimulation than without stimulation (Fig. 5b, without stimulation, $$n = 5$$; with stimulation, $$n = 18$$; $$p \leq 0.0107$$, t (19.67) = 2.819, Welch's t test).Figure 5Acute CoQ10 administration enhanced LTP in the motor cortex of middle-aged mice. The fEPSP amplitude in the layer V of M1 region was recorded with a single glass electrode. The stimulation electrode was placed in layers II/III as described in the methods section. ( a) The fEPSP amplitude increased compared to baseline amplitude in slices treated acutely with CoQ10 and high-frequency stimulation but not without high-frequency stimulation. The graph shows plots of the mean normalized fEPSP amplitude recorded in slices prepared from middle-aged mice with CoQ10 (without stim, $$n = 5$$ slices from 4 middle-aged mice; with stim, $$n = 18$$ slices from 10 mice). fEPSP amplitudes were normalized to baseline amplitudes before stimulation. The horizontal lines indicate the time of CoQ10 administration (50 μM, 20–25 min), and the arrows indicate the timing of the high-frequency stimulation (three trains of 100 pulses at 100 Hz applied at 15-s intervals). ( b) The mean normalized fEPSP amplitude at 25–27 min after CoQ10 administration with and without high-frequency stimulation (*$p \leq 0.05$, Welch's t test). ( c) Different magnitudes of LTP were induced in slices prepared from young adult and middle-aged mice treated acutely with CoQ10 and high-frequency stimulation. The plots of normalized fEPSP amplitudes are shown as in (a). ( d) The mean normalized fEPSP amplitudes at 25–27 min after high-frequency stimulation with and without CoQ10 administration (control: young adult, $$n = 20$$ slices from 10 mice; middle-aged, $$n = 16$$ slices from 9 mice; CoQ10: young adult, $$n = 19$$ slices from 4 mice; middle-aged, $$n = 18$$ slices from 10 mice; young adult with CoQ10 compared to middle-aged with CoQ10; *$p \leq 0.05$, two-way ANOVA with Bonferroni's multiple comparison test). The high-frequency stimulation induced slight LTP (105–$108\%$) in slices of young adult mice with and without CoQ10 administration and middle-aged mice without CoQ10 administration. Acute CoQ10 administration enhanced the magnitude of LTP on average by $22.39\%$. The middle-aged with CoQ10 data in (d) are identical to those in (b). The control and experimental groups had different numbers of mice because few experiments were discarded due to the baseline variation being greater than $10\%$ in the first 20 min of recording. Values are expressed as the mean ± SEM of independent experimental groups. For details of the data, see Supplementary Table 3.
Next, we evaluated the magnitude of LTP in slices taken from young adult and middle-aged mice by measuring changes in normalized fEPSP amplitude before and after LTP induction with high-frequency stimulation with and without acute CoQ10 administration. High-frequency stimulation with acute CoQ10 administration induced LTP in slices of young adult and middle-aged mice (Fig. 5c; CoQ10: young adult, $$p \leq 0.0051$$, t [18] = 3.187; middle-aged, $$p \leq 0.0061$$, t [17] = 3.133, paired t test). The high-frequency stimulation without CoQ10 administration induced LTP in slices of young adult and middle-aged mice similarly (Fig. 5d; Control, young adult: 108.1 ± $1.98\%$ of baseline, $$p \leq 0.0006$$, t [19] = 4.092; Middle-aged: 108.03 ± $1.35\%$ of baseline, $p \leq 0.0001$, t [15] = 5.962, paired t test). The fEPSP amplitude in slices of young adult and middle-aged mice after LTP induction with or without acute CoQ10 administration showed a significant interaction and a significant difference of age by two-way ANOVA. The magnitude of LTP was significantly greater in CoQ10-treated slices of middle-aged mice than CoQ10-treated slices of young adult mice (Fig. 5d; the interaction between treatment and age $$p \leq 0.0355$$, F [1, 69] = 4.598; the main effect of treatment $$p \leq 0.0371$$, F [1, 69] = 4.519, two-way ANOVA; young adult CoQ10 compared to middle-aged CoQ10, $$p \leq 0.0195$$, Bonferroni’s multiple comparison test). These results suggested that exogenous CoQ10 and increased neuronal activity enhanced the synaptic plasticity efficacy of middle-aged mice.
An age-related increase in NMDA receptor-dependent LTP has been observed in rat hippocampal slices34. Therefore, the NMDA receptor selective antagonist 2-amino-5-phosphonovaleric acid (APV) was applied with CoQ10 during LTP induction in the motor cortex to examine the role of NMDA receptors in the age-related LTP induction described in Fig. 5. The high-frequency stimulation in the presence of CoQ10 (50 μM, 23–25 min before and during LTP induction) induced an LTP of on average 110.88 ± $1.77\%$ of baseline in the M1 region of middle-aged mice (Fig. 6a, CoQ10; averaged fEPSP, between − 2 to 0 min and 58 to 60 min $$p \leq 0.0001$$, t [10] = 6.139, paired t test). However, the high-frequency stimulation in the presence of CoQ10 and APV (each 50 μM, 23–25 min before and during LTP induction) failed to induce LTP in slices taken from the same mice (Fig. 6a, CoQ10 + APV: 102.94 ± $2.65\%$ of baseline; averaged fEPSP, between − 2 to 0 min and 58 to 60 min $$p \leq 0.2924$$, t [10] = 1.111, paired t test). APV significantly blocked CoQ10-dependent LTP induction to a level similar to that of the control (Fig. 6b; $$p \leq 0.0167$$, F [2, 30] = 4.703, one-way ANOVA; control compared to CoQ10, $$p \leq 0.0343$$; CoQ10 compared to CoQ10 + APV, $$p \leq 0.0414$$, Bonferroni's multiple comparison test). These results suggested that CoQ10-dependent LTP of the M1 region in middle-aged mice was dependent on NMDA receptors. Figure 6CoQ10-dependent LTP was blocked by an NMDA receptor antagonist and increased the basal fEPSP amplitudes. ( a) The graph shows the averaged time course of the normalized fEPSP amplitude recorded in layer V in slices prepared from middle-aged mice with high-frequency stimulation alone (control), in the presence of CoQ10, and in the presence of CoQ10 with APV (each $$n = 11$$ slices from 11 mice). The ordinates represent normalized fEPSP amplitude, where $100\%$ corresponds to the averaged amplitude recorded before high-frequency stimulation, and the abscissa represents the time of recording. The horizontal line above the plots indicates the time of drug application. The arrows indicate the timing of the high-frequency stimulation (three trains of 100 pulses at 100 Hz applied at 15-s intervals). The inserts on the right show traces from representative recordings. Each trace is the average of 2 min immediately before the high-frequency stimulation (1, CoQ10; 3, CoQ10 + APV) and 2 min at the 58- to 60-min time point (2, CoQ10; 4, CoQ10 + APV). ( b) The blockage of NMDA receptors with APV (50 μM, 23–25 min before and during LTP induction) in the presence of CoQ10 occluded the LTP induction in the layer V of M1 region of the middle-aged mice (control, CoQ10, CoQ10 + APV, each $$n = 11$$, *$p \leq 0.05$, one-way ANOVA with Bonferroni's multiple comparison test). ( c) Average fEPSP amplitudes before (before) and 1 h after (after) the high-frequency stimulation recorded in ACSF in a range from 10 to 90 μA current stimuli (each $$n = 11$$ slices from 11 mice). Statistical analyses were performed using two-way repeated-measures ANOVA with Bonferroni's multiple comparisons for the control group, the CoQ10 group, and the CoQ10 + APV group (****$p \leq 0.0001$; ***$p \leq 0.001$; **$p \leq 0.01$; *$p \leq 0.05$). Values are the mean ± SEM of independent experimental groups. For details of the data, see Supplementary Table 3.
## Acute CoQ10 treatment augmented basal fEPSP amplitude
We hypothesized that CoQ10-dependent LTP might be part of the mechanism augmenting the basal fEPSP amplitude in middle-aged mice supplemented with CoQ10 by drinking water. Figure 6c shows the average amplitude of fEPSPs recorded from 5 trials of each current stimulus in 1 brain slice before the LTP experiment and 1 h after the high-frequency stimulation shown in Fig. 6a, b. In the control condition, there was no significant main effect between the fEPSP amplitudes before/after the high-frequency stimulation, but there was a significant interaction between the fEPSP amplitudes before/after stimulation and the stimulus intensity (the interaction between the fEPSP amplitudes before/after stimulation and the stimulus intensity $p \leq 0.0001$, F [8, 80] = 5.145; the main effect of the fEPSP amplitudes before/after stimulation $$p \leq 0.4513$$, F [1, 10] = 0.6145, the main effect of the stimulus intensity $p \leq 0.0001$, F [8, 80] = 102.2; two-way repeated-measures ANOVA). Bonferroni's multiple comparison showed significant differences between the fEPSP amplitudes before/after the high-frequency stimulation among stimulus intensities between 60 and 90 μA (Fig. 6c left; Before compared to After, 60–90 μA: $$p \leq 0.001$$ to 0.0279). However, when CoQ10-dependent LTP expression was observed, there was a significant interaction between the fEPSP amplitudes before/after stimulation and the stimulus intensity and a significant difference between the fEPSP amplitudes before/after stimulation. The fEPSP amplitudes increased significantly at an average of 115.95 ± $1.61\%$ between the two recording time points (Fig. 6c center; the interaction between the fEPSP amplitudes before/after stimulation and the stimulus intensity $$p \leq 0.0067$$, F [8, 80] = 2.912; the main effect of the fEPSP amplitudes before/after stimulation $$p \leq 0.0031$$, F [1, 10] = 14.92, two-way repeated-measures ANOVA; Before compared to After, 10–90 μA: $p \leq 0.0001$ to $$p \leq 0.0031$$, Bonferroni's multiple comparison test). In contrast, when coadministration of CoQ10 and APV occluded LTP expression, there was a significant interaction between the fEPSP amplitudes before/after stimulation and the stimulus intensity by two-way repeated-measures ANOVA and a significant difference between the fEPSP amplitudes before/after stimulation. The fEPSP amplitudes were significantly smaller among 20–50 μA stimuli (Fig. 6c right; the interaction between the fEPSP amplitudes before/after stimulation and the stimulus intensity $$p \leq 0.0057$$, F [8, 80] = 2.980; the main effect of fEPSP amplitudes before/after stimulation $$p \leq 0.0076$$, F [1, 10] = 11.09, the main effect of the stimulus intensity $p \leq 0.0001$, F [8, 80] = 88.42; two-way repeated-measures ANOVA; Before compared to After, 20–50 μA: $$p \leq 0.0001$$ to 0.0359, Bonferroni's multiple comparison test). These results suggested that basal fEPSP amplitudes were augmented when LTP expression was observed; therefore, CoQ10-dependent LTP may have improved the fEPSP amplitudes of the M1 region motor cortex of the CoQ10-supplemented middle-aged mice.
## Discussion
The middle-aged mice showed an age-related decline in motor function (Fig. 1a, c). Concomitantly, the M1 region of the middle-aged mice showed an age-related decline in fEPSP amplitude in the pathway from layers II/III to layer V neurons (Fig. 3a). The decreased motor function and fEPSP amplitude were reverted to the young adult level by supplementing CoQ10 by drinking water for 1 week (Figs. 1a, 3b). Furthermore, acute CoQ10 treatment of brain slices induced LTP in the layer V of M1 region of middle-aged mice (Fig. 5a, c). This LTP induction depended on exogenous CoQ10, high-frequency stimulation, and NMDA receptors; however, acute CoQ10 administration alone did not alter the fEPSP amplitude (Fig. 5a). Coadministration of CoQ10 and APV reduced basal synaptic transmission (Fig. 6c, right), which also indicates the contribution of NMDA receptors in the pathway from layers II/III to layer V neurons at LTP induction. These results suggested that a change in the efficacy of plasticity may be the underlying mechanism for the fEPSP amplitude recovery by CoQ10 treatment. Indeed, we demonstrated that CoQ10-dependent LTP in the layer V of M1 region translates to enhanced fEPSP amplitude (Fig. 6c). To our knowledge, this report is the first to demonstrate that the pathway from layers II/III to V of the M1 region shows (a) an age-related decrease in fEPSP amplitude and (b) LTP in middle-aged mice. We identified an age-related alteration and CoQ10 and NMDA receptor dependency of LTP induction in the M1 region.
The efficacy of CoQ10 depends on its formulation35. Nanoformulations of CoQ10 have higher bioavailability than regular CoQ10 and have been reported to increase brain CoQ10 content and protect neurons by oral administration7,36. A previous study by Takahashi et al.1 used the water-soluble nanoformula product of Nisshin Pharma (Aqua Q10L10). To test whether the beneficial effect of CoQ10 supplementation could be generalized, we used a water-soluble nanoformula-type CoQ10 from Petroeuroasia (40SP) in the behavioral and OCR analyses of this study. CoQ10 (40SP) showed a beneficial effect on motor function and the oxygen consumption rate of brain mitochondria in middle-aged mice, similar to Aqua Q10L10. These results demonstrated that the beneficial effect of CoQ10 supplementation could be confirmed in water-soluble nanoformula-type CoQ10 from at least two different sources and suggested that the beneficial effect of CoQ10 could be generalized.
Elderly individuals suffer a progressive loss of muscle mass and strength (sarcopenia) and motor function37–41. However, the motor deficit in middle-aged mice is less likely to be due to motor neuron loss, NMJ denervation, or muscle atrophy. NMJ denervation is not detected significantly at or earlier than 18 months of age in mice42,43. Similarly, the maintenance of NMJ number suggests that spinal motor neurons are preserved in middle-aged mice44. A decline in muscle contractility is less prominent earlier than 20 months of age in mice44, and we also confirmed that muscle strength did not change significantly with CoQ10 supplementation (Fig. 1b, d).
Age-related changes in electrophysiological activity in layer V have been linked to motor function deficits in humans and animals. Middle-aged humans (between the late 50 s and early 60 s) showed more intracortical inhibition and less intracortical facilitation in the motor cortex than young adults when examined using transcranial magnetic stimulation45. Elderly individuals in their 70 s exhibited similar but more profound intracortical inhibition and less intracortical facilitation46. These data suggested an age-related decline in neuronal activity in the motor cortex of humans due to an altered balance of excitatory and inhibitory circuits. The correlation of hypoexcitability in the motor cortex and behavioral defects has also been implicated in chronic obstructive pulmonary disease (COPD) and amyotrophic lateral sclerosis (ALS) patients 47,48. In contrast, motor function improved when the activity of motor cortex layer V neurons was increased using optogenetic stimulation in Parkinson's disease model mice49. Layer V pyramidal neurons directly evoke or control the rhythm of whisker movements in rodents50. These observations suggest that the excitability level of layer V neurons of the motor cortex is important to maintain motor functions.
Neuronal plasticity enhancers augment motor-skill learning or accelerate rehabilitation after brain damage51,52. In the rat motor cortex, the LTP-like plasticity of M1 region augments motor-skill learning and rehabilitation effects53. LTP can be induced in the motor cortex by motor-skill learning32,33, and a larger fEPSP amplitude can be stabilized for months in the motor cortex after motor-skill learning33. LTP is also naturally induced by the environment or sensory stimuli. Enriched environmental exposure changed cellular excitability and synaptic transmission, induced NMDA-dependent LTP54,55 and enhanced learning56. Sensory stimulation, such as rhythmic stimulation of whiskers, also induced NMDA-dependent LTP57,58. An age-related increase in NMDA receptor-dependent LTP has been observed in rat hippocampal slices34. These types of LTP may have been induced in the middle-aged mice supplemented with CoQ10 by drinking water and contributed to recovering fEPSP amplitude and motor function.
CoQ10 supplementation by drinking water in middle-aged mice enhanced complex I activity in the brain mitochondria (Fig. 2). Considering that oxidative phosphorylation is associated with oxidative stress59,60, CoQ10 supplementation might induce higher oxidative stress, and excessive oxidative stress impairs cognitive function61,62. However, CoQ10 supplementation by drinking water reduces oxidative stress and improves cognitive function63. Furthermore, exogenous CoQ10 administration is protective against age-related and pathological oxidative stress64,65. Therefore, the antioxidant status in the brain of the CoQ10-supplemented middle-aged mice may be beneficial overall for the behavioral outcome of these mice in the current study.
CoQ10 supplementation had a beneficial effect on the motor function of the middle-aged mice and rescued their behavior to the young adult level (Fig. 1a). The beneficial effect of CoQ10 was partially achieved by enhancing the excitability level of layer V neurons in the M1 region. We hypothesized that the basal fEPSP amplitude level was enhanced in the middle-aged mice during CoQ10 supplementation (Fig. 3) by the continuation and retention of LTP-like plasticity33,66, such as CoQ10-dependent LTP (Figs. 5, 6). The enhanced LTP induction efficacy augments the rehabilitation-like effect to improve the pole test latency (Fig. 1a, b). These effects might be similar to the recovery of motor function after stroke and nervous system damage by rehabilitation training, which makes use of the plasticity and recovery function of the central nervous system. Therefore, these results suggest the possibility of translational application of CoQ10 supplementation in the following circumstances: [1] oral CoQ10 administration as preventive care for age-related motor decline and [2] the enhancement of plasticity of the primary motor cortex to improve motor function in elderly individuals.
## Animals
All experimental procedures were approved by the Animal Care and Use Committee of the Tokyo Metropolitan Institute for Geriatrics and Gerontology. All experiments were carried out in accordance with the approved animal care and use protocol and the Guidelines for Care and Use of Laboratory Animals. The authors complied with the ARRIVE 2.0 guidelines67,68. C57BL/6NCr male mice were purchased from Japan SLC Inc. (Shizuoka, Japan) at 4 weeks old. The mice were housed in groups of two to five per cage and maintained in a temperature- and humidity-controlled environment with a 12-h light/dark cycle. We used a total of 56 young adult mice (6–8 months old) and 86 middle-aged mice (15–18 months old) given ad libitum food and water. For details of animal numbers, see Supplementary Table 1.
## CoQ10-supplemented mouse experiments
CoQ10-supplemented mice were analyzed by behavioral experiments, measurement of brain mitochondrial respiration, and electrophysiological recording with a multi-electrode array. Drinking water containing 150 μM water-soluble nanoformula-type CoQ10 (Figs. 1, 2, Coenzyme Q10 $40\%$ Water-dispersive Powder 40SP, Petroeuroasia Co. Ltd., Shizuoka, Japan; Figs. 3, 4, Aqua Q10L10-NF, Nisshin Pharma Inc., Tokyo, Japan) were prepared in light-protected bottles twice weekly and given ad libitum to mice until sacrifice based on the preceding studies1,17. An overview of the mouse experiments is shown in Fig. 7.Figure 7Schematic diagram of the experimental design for CoQ10-supplemented mice. Mice were supplemented with CoQ10 by drinking water for at least 1 week before starting the experiment. Age-matched control mice were given normal drinking water. The mice were evaluated using the wire hanging test at 7 days after starting CoQ10 supplementation for the young adult and middle-aged groups and again at 30 days for the middle-aged group. The pole test was performed 10–13 days after starting the supplementation for the young adult and middle-aged groups and again at 33–36 days for the middle-aged group. The brain mitochondrial respiration rates were measured between 40 and 76 days after starting CoQ10 supplementation. Therefore, the mice were 15 months old at the beginning and approximately 17 months old at the completion of the experiment. Electrophysiological recording with a multi-electrode array was performed using 15-month-old mice supplemented with CoQ10 by drinking water for 7 days. Figure numbers indicate the relative timing of corresponding experiments.
## Behavioral experiments
We performed a priori power analysis using G-Power software69 to estimate the required sample size for behavioral experiments. The experiments aimed to analyze the effects of aging and CoQ10 supplementation with $95\%$ actual power using a two-way analysis of variance (ANOVA) with four groups, a significance level of $p \leq 0.05$ and an effect size of $f = 0.4.$ The required total sample size was 84 mice, 21 mice per group. We decided to use 20 mice in each group and house five mice per cage for social housing. The middle-aged mouse group had one less data point due to death (Fig. 1c, d). Mice were handled by the experimenter for three consecutive days for habituation and then sequentially tested on the wire hanging and pole tests. The behavior tests were performed by personnel blinded to the treatment group and the animals were randomized.
## Wire hanging test
The four-limb wire hanging test (O'Hara & Co., Ltd., Tokyo, Japan) was performed as described previously70. The latency to fall from the grid was recorded from two trials with a 30 min intertrial interval. The longer latency was considered the representative value for the mice.
## Pole test
The pole test was first designed to evaluate bradykinesia in a Parkinson's disease murine model and has been used to measure motor coordination deficits18–20. Initially, mice were habituated to an experimental cage and the pole (length 45 cm, diameter 1 cm). The day before the test, four training trials were conducted. During the test, the time required for mice to turn their body and feet completely downward (T-turn) and the total time to descend to the floor of the experimental cage (T-total) were measured with 15 min intertrial intervals. The average of five test trials was used as the representative value.
## Measurement of mitochondrial respiration
After the behavioral experiments (16 months old; control, $$n = 20$$; CoQ10, $$n = 19$$), mitochondrial fractions from one brain hemisphere were isolated as previously described17. We measured NADH-linked (Complex I) respiration of brain mitochondrial fractions, which declined with age1. The OCRs of mitochondrial fractions in mitochondrial respiration medium (MiR05; 0.5 mM EGTA, 3 mM MgCl2, 60 mM lactobionic acid, 20 mM taurine, 10 mM KH2PO4, 20 mM HEPES, 110 mM D-sucrose, 1 g/l bovine serum albumin; pH 7.1) containing 10 mM cytochrome c were determined at 37 °C using high-resolution respirometry (Oxygraph-2k; Oroboros Instruments, Innsbruck, Austria). NADH-linked respiration was assessed by adding 2.5 mM ADP in the presence of 5 mM malate and 10 mM glutamate, as previously described17,71. The data were normalized to total protein of the mitochondrial fraction in the high-resolution respirometry chamber (1.23–2.00 mg of protein). The respiration rates were analyzed by pairing the control and CoQ10-supplemented groups in each measurement by personnel blinded to the treatment group. Three samples in the control group were not measurable due to insufficient sample volume caused by a human error. Two samples in the CoQ10 group were not measurable due to an equipment failure of the Oxygraph-2k.
## Brain slice preparation for electrophysiology
Mouse brains were cut at a 15°–20° angle inclined rostrally against the coronal plane of the cortex yielding slices with apical dendrites of layer V neurons parallel to the cut surface22. Brains were cut into 300 μm thick slices using a Pro 7 Linear Microslicer (Dosaka, Kyoto, Japan) in chilled artificial cerebrospinal fluid (ACSF, 124 mM NaCl, 3 mM KCl, 1 mM NaH2PO4, 1.2 mM MgCl2, 2.4 mM CaCl2, and 10 mM glucose) with 26 mM NaHCO3 bubbled with $95\%$ O2 and $5\%$ CO2 for oxygenation and pH adjustment to pH 7.4. The slices were incubated in 30 °C ACSF for 1 h for recovery and then maintained in room temperature (23–25 °C) ACSF until recordings. We selected a small but empirically adequate sample size for all electrophysiological experiments because this was the first evaluation.
## Multi-electrode array recording
We separately analyzed the primary motor (M1) and secondary motor (M2) cortices in the motor cortex (approximately + 0.8 to + 1.2 mm from bregma based on the mouse brain atlas by Paxinos and Franklin72). Evoked field excitatory postsynaptic potentials (fEPSPs) were recorded with a multi-electrode array (60pMEA$\frac{200}{30}$iR-Ti; Multi Channel Systems, Reutlingen, Germany) at room temperature by placing the multiple electrodes in the layer V of M1 or M2 regions. A stimulating glass electrode filled with 1 M NaCl (resistance < 1 MΩ) was placed in layers II/III of the motor cortex. Signals were sampled at room temperature at 50 kHz using a multi-electrode array 1060 amplifier with a band pass filter (3 kHz) (Multi Channel Systems), digitized with a Digidata 1440 series acquisition interface (Molecular Devices, San Jose, USA), and analyzed with pCLAMP10 software (Molecular Devices). Among the electrodes in layer V, we analyzed data from one electrode that recorded the largest fEPSP amplitude at 80 µA current stimuli. A paired-pulse ratio (PPR) was calculated as the ratio of fEPSPs (second amplitude/first amplitude) recorded during paired stimulation (60 μA) in 25- to 500-ms intervals. At the end of the recordings, an AMPA and kainate receptor antagonist, 6-cyano-7-nitroquinoxa-line-2,3-dione (CNQX, 10 μM), and an NMDA receptor antagonist, 2-amino-5-phosphonovaleric acid (APV, 25 μM), were bath-applied to block synaptic transmission and to confirm the disappearance of fEPSPs (data not shown). We used a total of 20 mice (young adult: $$n = 5$$; middle-aged control: $$n = 10$$; middle-aged supplemented with CoQ10: $$n = 5$$). The recordings were performed on brain slices in randomized order.
## Single glass electrode recording
Evoked fEPSPs were recorded at 30 °C with a borosilicate glass electrode (3.0–4.5 MΩ resistance) filled with ACSF and placed in the layer V of M1 region. A stimulating glass electrode filled with 1 M NaCl was placed in layers II/III of the M1 region in the radial direction from the recording electrode. Signals were sampled at 10 kHz and filtered at 1 kHz using an EPC 10 amplifier (HEKA Elektronik, Lambrecht/Pfalz, Germany) and analyzed offline with FITMASTER software (version 2 × 90.2, HEKA Elektronik). Baseline fEPSPs were evoked with short pulses (100 μs at 0.067 Hz) and recorded for at least 10 min preceding CoQ10 administration. The stimulus intensity was adjusted to a level where 65–$80\%$ of the maximal fEPSP amplitude was evoked. We used slices prepared from the same animals and recorded using alternating stimulus and pharmacological treatments on the same day. The experiment was discarded if the baseline variation was greater than $10\%$ in the first 20 min of recording. LTP was induced with three trains of high-frequency stimulation consisting of 100 pulses at 100 Hz applied at 15-s intervals. The magnitude of LTP was expressed as the % change in the average fEPSP amplitude obtained from 25 to 27 min (Fig. 5) or 58 to 60 min (Fig. 6) after LTP induction to the average amplitude of baseline fEPSP measured during the 2 min before the high-frequency stimulation. We used a total of 26 mice (young adult: $$n = 11$$; middle-aged: $$n = 15$$) in Fig. 5.
For the APV experiments, water-soluble nanoformula-type CoQ10 and APV (each 50 μM) dissolved in ACSF were bath-applied to the chamber from 23 to 25 min before LTP induction and then washed out after high-frequency stimulation. Immediately before and after the LTP experiments, the input‒output relationship was examined by varying the stimulus intensity. Three conditions (ACSF, CoQ10, CoQ10 + APV) were tested in each mouse, and data from mice that showed more than $5\%$ CoQ10-dependent LTP were analyzed (Fig. 6, 11 mice were analyzed among 16 mice tested at 15–18 months old with the same birthdate). The recordings were performed on brain slices prepared from the same animals and treated in randomized order with three experimental conditions.
## Drugs
APV was purchased from Tocris Bioscience (Bristol, UK). All other drugs were purchased from Sigma‒Aldrich (St. Louis, USA).
## Statistics
Statistical differences of three or more groups were assessed using one-way analysis of variance (ANOVA) or two-way repeated-measures ANOVA with a multiple comparison test with Bonferroni's correction (Prism version 8.4.3, GraphPad Software, La Jolla, USA). Statistical differences between the two groups or conditions were assessed using the two-tailed Welch's t test or paired t test. All values are expressed as the mean ± SEM. Statistical significance was set at $p \leq 0.05.$ For details of the statistical analyses, see Supplementary Table 2.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-31510-1.
## References
1. Takahashi K, Ohsawa I, Shirasawa T, Takahashi M. **Early-onset motor impairment and increased accumulation of phosphorylated alpha-synuclein in the motor cortex of normal aging mice are ameliorated by coenzyme Q**. *Exp. Gerontol.* (2016) **81** 65-75. DOI: 10.1016/j.exger.2016.04.023
2. Turturro A. **Growth curves and survival characteristics of the animals used in the Biomarkers of Aging Program**. *J. Gerontol. A Biol. Sci. Med. Sci.* (1999) **54** B492-501. DOI: 10.1093/gerona/54.11.b492
3. Yuan R. **Genetic coregulation of age of female sexual maturation and lifespan through circulating IGF1 among inbred mouse strains**. *Proc. Natl. Acad. Sci. U. S. A.* (2012) **109** 8224-8229. DOI: 10.1073/pnas.1121113109
4. Seidler RD. **Motor control and aging: Links to age-related brain structural, functional, and biochemical effects**. *Neurosci. Biobehav. Rev.* (2010) **34** 721-733. DOI: 10.1016/j.neubiorev.2009.10.005
5. Tieland M, Trouwborst I, Clark BC. **Skeletal muscle performance and ageing**. *J. Cachexia Sarcopenia Muscle* (2018) **9** 3-19. DOI: 10.1002/jcsm.12238
6. Cassady K. **Sensorimotor network segregation declines with age and is linked to GABA and to sensorimotor performance**. *Neuroimage* (2019) **186** 234-244. DOI: 10.1016/j.neuroimage.2018.11.008
7. Takahashi M, Takahashi K. **Water-soluble CoQ10 as a promising anti-aging agent for neurological dysfunction in brain mitochondria**. *Antioxidants (Basel)* (2019). DOI: 10.3390/antiox8030061
8. Grimm A, Eckert A. **Brain aging and neurodegeneration: From a mitochondrial point of view**. *J. Neurochem.* (2017) **143** 418-431. DOI: 10.1111/jnc.14037
9. Lores-Arnaiz S, Bustamante J. **Age-related alterations in mitochondrial physiological parameters and nitric oxide production in synaptic and non-synaptic brain cortex mitochondria**. *Neuroscience* (2011) **188** 117-124. DOI: 10.1016/j.neuroscience.2011.04.060
10. Lores-Arnaiz S. **Brain cortex mitochondrial bioenergetics in synaptosomes and non-synaptic mitochondria during aging**. *Neurochem. Res.* (2016) **41** 353-363. DOI: 10.1007/s11064-015-1817-5
11. Friedrich T. **Two binding sites of inhibitors in NADH: Ubiquinone oxidoreductase (complex I). Relationship of one site with the ubiquinone-binding site of bacterial glucose: Ubiquinone oxidoreductase**. *Eur. J. Biochem.* (1994) **219** 691-698. DOI: 10.1111/j.1432-1033.1994.tb19985.x
12. Brandt U. **Proton-translocation by membrane-bound NADH: Ubiquinone-oxidoreductase (complex I) through redox-gated ligand conduction**. *Biochim. Biophys. Acta* (1997) **1318** 79-91. DOI: 10.1016/s0005-2728(96)00141-7
13. Scheffler IE. **Molecular genetics of succinate: Quinone oxidoreductase in eukaryotes**. *Prog. Nucleic Acid Res. Mol. Biol.* (1998) **60** 267-315. DOI: 10.1016/s0079-6603(08)60895-8
14. Battino M. **Coenzyme Q content in synaptic and non-synaptic mitochondria from different brain regions in the ageing rat**. *Mech. Ageing Dev.* (1995) **78** 173-187. DOI: 10.1016/0047-6374(94)01535-t
15. Beyer RE. **Tissue coenzyme Q (ubiquinone) and protein concentrations over the life span of the laboratory rat**. *Mech. Ageing Dev.* (1985) **32** 267-281. DOI: 10.1016/0047-6374(85)90085-5
16. Kalen A, Appelkvist EL, Dallner G. **Age-related changes in the lipid compositions of rat and human tissues**. *Lipids* (1989) **24** 579-584. DOI: 10.1007/BF02535072
17. Takahashi K, Takahashi M. **Exogenous administration of coenzyme Q10 restores mitochondrial oxygen consumption in the aged mouse brain**. *Mech. Ageing Dev.* (2013) **134** 580-586. DOI: 10.1016/j.mad.2013.11.010
18. Ogawa N, Hirose Y, Ohara S, Ono T, Watanabe Y. **A simple quantitative bradykinesia test in MPTP-treated mice**. *Res. Commun. Chem. Pathol. Pharmacol.* (1985) **50** 435-441. PMID: 3878557
19. Matsuura K, Kabuto H, Makino H, Ogawa N. **Pole test is a useful method for evaluating the mouse movement disorder caused by striatal dopamine depletion**. *J. Neurosci. Methods* (1997) **73** 45-48. DOI: 10.1016/s0165-0270(96)02211-x
20. Leconte C. **Histological and behavioral evaluation after traumatic brain injury in mice: A ten months follow-up study**. *J. Neurotrauma* (2020) **37** 1342-1357. DOI: 10.1089/neu.2019.6679
21. Li Z, Okamoto K, Hayashi Y, Sheng M. **The importance of dendritic mitochondria in the morphogenesis and plasticity of spines and synapses**. *Cell* (2004) **119** 873-887. DOI: 10.1016/j.cell.2004.11.003
22. Weiler N, Wood L, Yu J, Solla SA, Shepherd GMG. **Top-down laminar organization of the excitatory network in motor cortex**. *Nat. Neurosci.* (2008) **11** 360-366. DOI: 10.1038/nn2049
23. Anderson CT, Sheets PL, Kiritani T, Shepherd GM. **Sublayer-specific microcircuits of corticospinal and corticostriatal neurons in motor cortex**. *Nat. Neurosci.* (2010) **13** 739-744. DOI: 10.1038/nn.2538
24. Yu J. **Local-circuit phenotypes of layer 5 neurons in motor-frontal cortex of YFP-H mice**. *Front. Neural Circuits* (2008) **2** 6. DOI: 10.3389/neuro.04.006.2008
25. Shepherd GM. **Corticostriatal connectivity and its role in disease**. *Nat. Rev. Neurosci.* (2013) **14** 278-291. DOI: 10.1038/nrn3469
26. Kiritani T, Wickersham IR, Seung HS, Shepherd GM. **Hierarchical connectivity and connection-specific dynamics in the corticospinal-corticostriatal microcircuit in mouse motor cortex**. *J. Neurosci.* (2012) **32** 4992-5001. DOI: 10.1523/JNEUROSCI.4759-11.2012
27. Alexander GE, DeLong MR, Strick PL. **Parallel organization of functionally segregated circuits linking basal ganglia and cortex**. *Annu. Rev. Neurosci.* (1986) **9** 357-381. DOI: 10.1146/annurev.ne.09.030186.002041
28. Hooks BM. **Laminar analysis of excitatory local circuits in vibrissal motor and sensory cortical areas**. *PLoS Biol.* (2011) **9** e1000572. DOI: 10.1371/journal.pbio.1000572
29. Todorova V, Blokland A. **Mitochondria and synaptic plasticity in the mature and aging nervous system**. *Curr. Neuropharmacol.* (2017) **15** 166-173. DOI: 10.2174/1570159x14666160414111821
30. Rioult-Pedotti MS, Friedman D, Donoghue JP. **Learning-induced LTP in neocortex**. *Science* (2000) **290** 533-536. DOI: 10.1126/science.290.5491.533
31. Barbati SA. **Enhancing plasticity mechanisms in the mouse motor cortex by anodal transcranial direct-current stimulation: The contribution of nitric oxide signaling**. *Cereb. Cortex* (2020) **30** 2972-2985. DOI: 10.1093/cercor/bhz288
32. Hess G. **Synaptic plasticity of local connections in rat motor cortex**. *Acta Neurobiol. Exp. (Wars)* (2004) **64** 271-276. PMID: 15366258
33. Rioult-Pedotti MS, Donoghue JP, Dunaevsky A. **Plasticity of the synaptic modification range**. *J. Neurophysiol.* (2007) **98** 3688-3695. DOI: 10.1152/jn.00164.2007
34. Pinho J. **Enhanced LTP in aged rats: Detrimental or compensatory?**. *Neuropharmacology* (2017) **114** 12-19. DOI: 10.1016/j.neuropharm.2016.11.017
35. Lopez-Lluch G, Del Pozo-Cruz J, Sanchez-Cuesta A, Cortes-Rodriguez AB, Navas P. **Bioavailability of coenzyme Q10 supplements depends on carrier lipids and solubilization**. *Nutrition* (2019) **57** 133-140. DOI: 10.1016/j.nut.2018.05.020
36. Sikorska M. **Nanomicellar formulation of coenzyme Q10 (Ubisol-Q10) effectively blocks ongoing neurodegeneration in the mouse 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine model: Potential use as an adjuvant treatment in Parkinson's disease**. *Neurobiol. Aging* (2014) **35** 2329-2346. DOI: 10.1016/j.neurobiolaging.2014.03.032
37. Clark BC. **Neuromuscular changes with aging and sarcopenia**. *J. Frailty Aging* (2019) **8** 7-9. DOI: 10.14283/jfa.2018.35
38. Gonzalez-Freire M, de Cabo R, Studenski SA, Ferrucci L. **The neuromuscular junction: Aging at the crossroad between nerves and muscle**. *Front. Aging Neurosci.* (2014) **6** 208. DOI: 10.3389/fnagi.2014.00208
39. Manini TM, Hong SL, Clark BC. **Aging and muscle: A neuron's perspective**. *Curr. Opin. Clin. Nutr. Metab. Care* (2013) **16** 21-26. DOI: 10.1097/MCO.0b013e32835b5880
40. Roubenoff R. **Sarcopenia and its implications for the elderly**. *Eur. J. Clin. Nutr.* (2000) **54** S40-47. DOI: 10.1038/sj.ejcn.1601024
41. Willadt S, Nash M, Slater C. **Age-related changes in the structure and function of mammalian neuromuscular junctions**. *Ann. N. Y. Acad. Sci.* (2018) **1412** 41-53. DOI: 10.1111/nyas.13521
42. Valdez G. **Attenuation of age-related changes in mouse neuromuscular synapses by caloric restriction and exercise**. *Proc. Natl. Acad. Sci. U. S. A.* (2010) **107** 14863-14868. DOI: 10.1073/pnas.1002220107
43. Sheth KA. **Muscle strength and size are associated with motor unit connectivity in aged mice**. *Neurobiol. Aging* (2018) **67** 128-136. DOI: 10.1016/j.neurobiolaging.2018.03.016
44. Chugh D. **Neuromuscular junction transmission failure is a late phenotype in aging mice**. *Neurobiol. Aging* (2020) **86** 182-190. DOI: 10.1016/j.neurobiolaging.2019.10.022
45. Kossev AR, Schrader C, Dauper J, Dengler R, Rollnik JD. **Increased intracortical inhibition in middle-aged humans; a study using paired-pulse transcranial magnetic stimulation**. *Neurosci. Lett.* (2002) **333** 83-86. DOI: 10.1016/s0304-3940(02)00986-2
46. McGinley M, Hoffman RL, Russ DW, Thomas JS, Clark BC. **Older adults exhibit more intracortical inhibition and less intracortical facilitation than young adults**. *Exp. Gerontol.* (2010) **45** 671-678. DOI: 10.1016/j.exger.2010.04.005
47. Alexandre F. **Specific motor cortex hypoexcitability and hypoactivation in COPD patients with peripheral muscle weakness**. *BMC Pulm. Med.* (2020) **20** 1. DOI: 10.1186/s12890-019-1042-0
48. Khedr EM, Ahmed MA, Hamdy A, Shawky OA. **Cortical excitability of amyotrophic lateral sclerosis: Transcranial magnetic stimulation study**. *Neurophysiol. Clin.* (2011) **41** 73-79. DOI: 10.1016/j.neucli.2011.03.001
49. Sanders TH, Jaeger D. **Optogenetic stimulation of cortico-subthalamic projections is sufficient to ameliorate bradykinesia in 6-ohda lesioned mice**. *Neurobiol. Dis.* (2016) **95** 225-237. DOI: 10.1016/j.nbd.2016.07.021
50. Brecht M, Schneider M, Sakmann B, Margrie TW. **Whisker movements evoked by stimulation of single pyramidal cells in rat motor cortex**. *Nature* (2004) **427** 704-710. DOI: 10.1038/nature02266
51. Zemmar A. **Neutralization of Nogo-A enhances synaptic plasticity in the rodent motor cortex and improves motor learning in vivo**. *J. Neurosci.* (2014) **34** 8685-8698. DOI: 10.1523/JNEUROSCI.3817-13.2014
52. Abe H. **CRMP2-binding compound, edonerpic maleate, accelerates motor function recovery from brain damage**. *Science* (2018) **360** 50-57. DOI: 10.1126/science.aao2300
53. Bundy DT, Guggenmos DJ, Murphy MD, Nudo RJ. **Chronic stability of single-channel neurophysiological correlates of gross and fine reaching movements in the rat**. *PLoS ONE* (2019) **14** e0219034. DOI: 10.1371/journal.pone.0219034
54. Irvine GI, Logan B, Eckert M, Abraham WC. **Enriched environment exposure regulates excitability, synaptic transmission, and LTP in the dentate gyrus of freely moving rats**. *Hippocampus* (2006) **16** 149-160. DOI: 10.1002/hipo.20142
55. Stein LR, O'Dell KA, Funatsu M, Zorumski CF, Izumi Y. **Short-term environmental enrichment enhances synaptic plasticity in hippocampal slices from aged rats**. *Neuroscience* (2016) **329** 294-305. DOI: 10.1016/j.neuroscience.2016.05.020
56. Bednarek E, Caroni P. **beta-Adducin is required for stable assembly of new synapses and improved memory upon environmental enrichment**. *Neuron* (2011) **69** 1132-1146. DOI: 10.1016/j.neuron.2011.02.034
57. Gambino F. **Sensory-evoked LTP driven by dendritic plateau potentials in vivo**. *Nature* (2014) **515** 116-119. DOI: 10.1038/nature13664
58. Cheyne JE, Montgomery JM. **The cellular and molecular basis of in vivo synaptic plasticity in rodents**. *Am. J. Physiol. Cell Physiol.* (2020) **318** C1264-C1283. DOI: 10.1152/ajpcell.00416.2019
59. Cui H, Kong Y, Zhang H. **Oxidative stress, mitochondrial dysfunction, and aging**. *J. Signal Transduct.* (2012) **2012** 646354. DOI: 10.1155/2012/646354
60. Shigenaga MK, Hagen TM, Ames BN. **Oxidative damage and mitochondrial decay in aging**. *Proc. Natl. Acad. Sci. U. S. A.* (1994) **91** 10771-10778. DOI: 10.1073/pnas.91.23.10771
61. Forster MJ. **Age-related losses of cognitive function and motor skills in mice are associated with oxidative protein damage in the brain**. *Proc. Natl. Acad. Sci. USA* (1996) **93** 4765-4769. DOI: 10.1073/pnas.93.10.4765
62. Massaad CA, Klann E. **Reactive oxygen species in the regulation of synaptic plasticity and memory**. *Antioxid. Redox Signal.* (2011) **14** 2013-2054. DOI: 10.1089/ars.2010.3208
63. Monsef A, Shahidi S, Komaki A. **Influence of chronic coenzyme Q10 supplementation on cognitive function, learning, and memory in healthy and diabetic middle-aged rats**. *Neuropsychobiology* (2019) **77** 92-100. DOI: 10.1159/000495520
64. Diaz-Casado ME. **The paradox of coenzyme Q(10) in aging**. *Nutrients* (2019). DOI: 10.3390/nu11092221
65. Vegh C. **Combined ubisol-Q(10) and ashwagandha root extract target multiple biochemical mechanisms and reduces neurodegeneration in a paraquat-induced rat model of Parkinson's disease**. *Antioxidants (Basel)* (2021). DOI: 10.3390/antiox10040563
66. Cantarero G, Lloyd A, Celnik P. **Reversal of long-term potentiation-like plasticity processes after motor learning disrupts skill retention**. *J. Neurosci.* (2013) **33** 12862-12869. DOI: 10.1523/JNEUROSCI.1399-13.2013
67. Kilkenny C, Browne WJ, Cuthill IC, Emerson M, Altman DG. **Improving bioscience research reporting: The ARRIVE guidelines for reporting animal research**. *PLoS Biol.* (2010) **8** e1000412. DOI: 10.1371/journal.pbio.1000412
68. Percie du Sert N. **Reporting animal research: Explanation and elaboration for the ARRIVE guidelines 20**. *PLoS Biol.* (2020) **18** e3000411. DOI: 10.1371/journal.pbio.3000411
69. Faul F, Erdfelder E, Lang AG, Buchner A. **G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences**. *Behav. Res. Methods* (2007) **39** 175-191. DOI: 10.3758/bf03193146
70. Yanai S, Endo S. **Functional aging in male C57BL/6J mice across the life-span: A systematic behavioral analysis of motor, emotional, and memory function to define an aging phenotype**. *Front. Aging Neurosci.* (2021) **13** 6621. DOI: 10.3389/fnagi.2021.697621
71. Takahashi K, Miura Y, Ohsawa I, Shirasawa T, Takahashi M. **In vitro rejuvenation of brain mitochondria by the inhibition of actin polymerization**. *Sci. Rep.* (2018) **8** 15585. DOI: 10.1038/s41598-018-34006-5
72. Paxinos G, Franklin KBJ. *Paxinos and Franklin's the Mouse Brain in Stereotaxic Coordinates* (2013)
|
---
title: 'Prenatal marijuana exposure and visual perception in toddlers: Evidence of
a sensory processing deficit'
authors:
- Beth A. Bailey
- Jahla B. Osborne
journal: Frontiers in Pediatrics
year: 2023
pmcid: PMC10017869
doi: 10.3389/fped.2023.1113047
license: CC BY 4.0
---
# Prenatal marijuana exposure and visual perception in toddlers: Evidence of a sensory processing deficit
## Abstract
### Background
Research has identified a link between prenatal marijuana exposure and multiple outcomes in children, including cognitive development. Several studies have found specific differences in sensory processing and attention, with visual perception especially impacted in school age children. The current study explored whether this effect is evident at an earlier age, and thus our goal was to investigate the relationship between in-utero marijuana exposure and sensory processing capabilities in toddlers. We hypothesized that in-utero marijuana exposure throughout pregnancy would specifically predict visual sensory hyperactivity in children as young as 15 months of age.
### Methods
Participants were 225 15-month-old children whose mothers were recruited during pregnancy. Substance exposure was prospectively collected and biochemically verified, with marijuana coded as no exposure, 1st trimester exposure only, or exposure throughout pregnancy. The Infant Toddler Sensory Profile evaluated 5 domains of sensory processing (visual, auditory, tactile, vestibular, oral).
### Results
Prenatal marijuana exposure throughout pregnancy, but not when limited to the first trimester, predicted a two-fold increased likelihood of scoring in a range indicating high levels of seeking out and potentially over-attending to visual stimulation after controlling for potentially confounding factors including other prenatal exposures. Marijuana exposure was not significantly related to other processing domains.
### Conclusion
Results indicate that links previously identified between prenatal marijuana exposure and visual function and attention may already be evident at 15 months of age, and also suggest an impact related to continuous/later pregnancy exposure. Our findings, as well as those from previous studies, all suggest visual processing differences for exposed children, differences that may predict emerging issues with visual attention and habituation. As legalization of marijuana continues to increase, further research is clearly needed to examine specific teratologic effects associated with use during pregnancy.
## Introduction
Marijuana is the most commonly used drug in the United States, with $15.9\%$ of the American population (43.5 million people) self-reporting marijuana use in 2018 [1]. Reasons reported for using marijuana include enjoyment, stress relief, and pain relief to name a few (2–6). While marijuana has been shown to reduce pain and improve sleep quality [6], the impact of marijuana use during pregnancy remains unclear. The American College of Obstetricians & Gynecologists (ACOG) advises against the use of marijuana during pregnancy [7]. However, it is estimated that at least $4.2\%$ of pregnant women still report using marijuana [8]. Reports indicate pregnant women may use marijuana for a variety of reasons, such as nausea and/or anxiety management [8]. Studies investigating the association between in-utero exposure to marijuana and birth outcomes provide mixed results. For example, Crume et al. only reported increased odds of delivering a low birth weight infant (<2,500 g), but not increased odds of neonatal intensive care (NICU) admission or preterm birth (delivery prior to 37 completed weeks gestation) in a sample from Colorado [9]. Conversely, Ko et al. found no association between in-utero marijuana exposure and low birth weight using PRAMS data [10], while our research team demonstrated increased risk of low birth weight, preterm delivery, and NICU admission following prenatal marijuana exposure in a multi-state sample [11]. Most recently, a meta-analysis by Marchand et al. indicated that pregnant women exposed to marijuana are at increased risk of experiencing preterm birth, delivering a low birth weight infant, and having their infant admitted to the NICU [12].
Prenatal marijuana exposure may also impact cognitive development in offspring. Several longitudinal studies have found increased deficits in several cognitive domains for marijuana exposed offspring, such as attention, language comprehension [13], memory [13, 14], visual perception [13], and visual reasoning [15]. Chakraborty et al. detailed enhanced performance on a global motion perception task in children at age four and five who experienced in-utero marijuana exposure compared to controls, with dose-dependent effects, a finding that suggests a hyperactivity of visual function in the visual cortex of children exposed to in-utero marijuana [16]. Studies examining individuals with Autism Spectrum Disorder (ASD), a developmental condition characterized by social impairments and repetitive behavior [17], have found links between visual sensory hyperactivity and overactivity in the visual cortex [18, 19]. Additionally, irregular visual sensory responsiveness has been shown to be associated with increased symptoms of Attention-Deficit Hyperactivity Disorder (ADHD) [20], a neurodevelopmental disorder defined by inattention and hyperactive/impulsive behavior [21]. Both disorders are thought to embody aspects of attention dysregulation, such as failing to habituate task irrelevant stimuli [22, 23]. However, the functional implications of visual sensory hyperactivity in offspring from in-utero marijuana exposure are less understood. Therefore, the goal of the present study was to investigate the relationship between in-utero marijuana exposure and sensory processing capabilities in toddlers using the Infant/Toddler Sensory Profile (ITSP). We hypothesized that in-utero marijuana exposure throughout pregnancy would specifically predict visual sensory hyperactivity in children as young as 15 months of age.
## Participants
Study participants were a subset of those who participated in a longitudinal study focused on pregnancy health and child outcomes. Characteristics of the parent study have been described elsewhere [24, 25]. Briefly, women were recruited at their first prenatal visit from several prenatal practices in Tennessee and Virginia, some for a pregnancy smoking intervention and others as non-smoking controls. The final prenatal sample (over $95\%$ of those approached for study consent) eligible for inclusion in the current study were those who completed at least two pregnancy research interviews ($92\%$) and gave birth to a newborn who survived to delivery hospitalization discharge at one of the study hospitals ($95\%$). This resulted in 1,063 maternal-child dyads. Of these, 250 were selected for follow-up at child age 15 months based on current age of child and representation of different prenatal exposures. Of the 250 selected, 225 ($90\%$) were located, agreed to participate in the developmental follow-up phase of the study, and completed all components of the 15-month assessment session.
## Procedures
Families were contacted by phone, email, and standard mail 4–6 weeks prior to the child reaching 15 months of age and invited to a research session for developmental follow-up. Families chose time of day for the 2-h session to avoid feeding and nap times, and were provided a small monetary incentive for participation. Transportation was arranged and on-site child care was provided for siblings free of charge as needed. During the session, parents completed demographic, substance use, parenting, and family environment surveys, along with standardized parent-assessments of child health and behavior. Developmental assessments were conducted with the child by a single masters’ level trained examiner blinded to prenatal drug exposure status, who also weighed and measured the child. Consistent order of assessment administration across all participants was observed, with breaks for rest permitted as needed. Both the parent study and the developmental follow-up were approved by the IRB at the affiliated university, and new informed consent was obtained for the 15-month assessment.
## Measures
The primary outcome of interest for the current study was performance on the Infant Toddler Sensory Profile [26]. This parent report measure evaluates five domains of sensory processing: visual, auditory, tactile, vestibular, oral. To complete the ITSP, the parent indicates the frequency of the child’s responses (Almost Always, Frequently, Occasionally, Seldom, or Almost Never) to various sensory experiences. Scores can be grouped based on established thresholds. For this study, we grouped total responses on each domain based on these established cut-points to compare those “More than Others” and “Much More than Others” (i.e., greater than one standard deviation above the mean for the norming sample) with all those “Just Like the Majority of Others,” “Less than Others,” and “Much Less than Others” (i.e., equal to or less than 1 standard deviation above the mean). Higher scores indicate seeking out or preferring a high level of stimulation in that domain, and may indicate failure to habituate and over-attention leading to failure to attend to or discriminate other sensory information.
Prenatal substance exposure information was collected prospectively during pregnancy via both self-report and biochemical assessment during the initial phase of the study. All participating mothers completed urine drug screens at entry to prenatal care, at least one additional time during pregnancy, and at delivery. Urine drug screens assessed cotinine, marijuana, opioids, benzodiazepines, stimulants, and hallucinogens, and standard laboratory cut-off values were used to indicate positive tests. Additionally, exhaled carbon monoxide levels, a marker of tobacco smoking, were assessed through expired air samples and were considered positive based on established cut-points for pregnant women [27]. Women were also asked to self-report any drug use via standardized tools including the gold-standard timeline follow-back method [28]. Finally, most ($93\%$) newborns had urine drug screens completed on their first urine for all substances listed above, and more than two thirds ($71\%$) had either meconium or cord blood testing for drugs. A woman was considered positive for use of a substance if any of the methods of detection were positive. Additionally, use was classified based on timing during pregnancy when it occurred, with each substance use grouped based on whether a participant used the substance only in the first trimester, or engaged in continued use beyond that. With respect to marijuana, the drug exposure of interest in this study, for those who continued to use past the first trimester, all but one had definitive evidence of still using marijuana at delivery. Thus, use beyond the first trimester was considered to indicate use throughout pregnancy.
Additional data collected included standard demographics and medical history, which were collected via maternal self-report throughout the study, and from medical chart review.
## Data analysis
Children who had prenatal exposure to marijuana were compared with those who did not on background factors, including other substance exposure, using t-tests and chi-square analysis. Bivariate group differences on the five sensory processing domains were examined using chi square analysis. Analyses controlling for significant background factors and other exposures utilized logistic regression, with odds ratios and $95\%$ confidence intervals reported. All analyses were conducted using IBM SPSS ver 28.
## Results
At the time of the developmental assessment, all children ranged in age from 14 months 2 weeks to 15 months 2 weeks. Of the 225 participants, 65 ($29\%$) had prenatal marijuana exposure, with more than half of these ($$n = 40$$) exposed throughout gestation.
Comparison of those with and without marijuana exposure on background factors is presented in Table 1. As shown, the two groups differed significantly on several characteristics. Compared to those without exposure, those with in-utero marijuana exposure had mothers that were significantly younger with lower levels of education, were less likely to be married and more likely to have a family income below the federal poverty level. In addition, they were significantly more likely to have had prenatal tobacco exposure.
**Table 1**
| Unnamed: 0 | Non-exposed (n = 160) | Exposed (n = 66) | p |
| --- | --- | --- | --- |
| Maternal age (years) | 26.4 (5.8) | 24.4 (5.0) | .011 |
| Maternal education (years) | 13.3 (2.5) | 12.3 (1.6) | .002 |
| Maternal marital status (% married) | 62.9% | 33.3% | <.001 |
| Family income (% below federal poverty level) | 46.3% | 64.2% | <.001 |
| Child age (months) | 14.9 (.2) | 14.9 (.2) | .993 |
| Child gender (% male) | 59.1% | 53.1% | .243 |
| Child care (% in any kind of non-parental care) | 57.0% | 54.5% | .740 |
| Child current second hand smoke exposure (%) | 26.3% | 28.2% | .626 |
| Prenatal alcohol exposure (%) | 7.5% | 7.3% | .942 |
| Prenatal tobacco exposure (%) | 20.6% | 58.5% | <.001 |
The relationships between prenatal marijuana exposure and the sensory processing domains are shown in Table 2. In-utero marijuana exposure significantly increased the odds of seeking out high levels of visual stimulation, but not when that exposure only occurred during the first trimester. Indeed, exposure to marijuana throughout gestation increased the risk for visual processing differences nearly two-fold after control for potentially confounding factors including gestational exposure to tobacco. Prenatal marijuana exposure did not significantly predict high scores on any of the other sensory domains, although high levels of tactile sensation seeking approached significance in this modestly sized sample.
**Table 2**
| Unnamed: 0 | No Marijuana exposure | 1st Trimester marijuana | Marijuana throughout gestation | Adjusted OR for ANY marijuana exposurea | Adjusted OR for marijuana exposure THROUGHOUT gestationb |
| --- | --- | --- | --- | --- | --- |
| Auditory | 28.9% | 26.3% | 29.8% | .87 (.30–2.58) | 1.04 (.51–2.13) |
| Visual | 39.0% | 47.4% | 55.3% | 1.41 (.54–3.66) | 1.94 (1.01–3.72) |
| Tactile | 25.2% | 27.7% | 36.8% | 1.14 (.55–2.37) | 1.74 (.94–2.71) |
| Vestibular | 37.1% | 31.6% | 38.3% | .78 (.28–2.17) | 1.05 (.54–2.06) |
| Oral | 13.2% | 10.5% | 17.0% | .77 (.17–3.59) | 1.35 (.55–3.78) |
## Discussion
The purpose of our study was to examine the relationship between in-utero marijuana exposure and sensory processing capabilities in toddlers using the ITSP. Our prediction, that in-utero marijuana exposure (throughout pregnancy) would specifically predict visual sensory hyperactivity in children as young as 15 months of age, was supported. Overall, we had three notable findings. First, we demonstrated (as others have) that pregnant women exposed to marijuana during pregnancy differ in many ways from non-exposed pregnant women. Specifically, pregnant women with marijuana exposure were generally younger in age, less educated, less likely to be married, more impoverished, and more likely to also engage in prenatal tobacco use. This clearly shows the need for extensive control for confounding when examining the impact of gestational exposure to marijuana.
Our second and primary finding was that toddlers who experienced prenatal marijuana exposure were two times more likely to seek out high levels of visual stimulation compared to non-exposed toddlers. We did not find increased odds for any other sensory domain in relation to prenatal marijuana exposure, suggesting specificity to the visual domain. Our third notable finding was that timing of marijuana exposure matters, and that visual processing is primarily impacted when exposure occurs late/throughout gestation. This had implications for intervention, suggesting that significant adverse outcomes related to visual processing may be avoided if pregnant marijuana users cease use by the end of the first trimester.
## Does prenatal marijuana exposure predict a habituation deficit?
Understanding in-utero exposure to marijuana in relation to visual sensory processing and visual attention is an important area of study as this relationship might underlie higher level issues in habituation and over-attention to irrelevant information in the external environment. Being able to successfully discriminate between relevant and irrelevant visual stimuli is necessary for daily life functioning in order to focus and avoid external distractions.
Research findings on the relationship between prenatal marijuana exposure and visual sensory processing are mixed. The present study suggests that seeking out high-level visual stimulation begins as early as 15 months of age following marijuana exposure throughout gestation. This finding aligns with previous studies such as that by Chakraborty et al. who uncovered enhanced performance on a global motion perception task for 4–5 year-olds exposed prenatally to marijuana [16], suggesting in-utero marijuana exposure is related to overactive visual function. Along similar lines, Leech et al. [ 29] found evidence for enhanced attention capabilities in 6-year-olds exposed to prenatal marijuana (through the second trimester) based on significantly fewer omission errors on a continuous performance task (CPT)—a common visual attention task [30].
Interestingly, the Ottawa Prospective Prenatal Study (OPPS) reported increased deficits in attention based on significantly more omission errors as measured by a visual attention vigilance task in 5–6 year olds [13]. Parental reports from another longitudinal cohort also indicated prenatal marijuana exposure correlates with symptoms of inattention at 10 years old [31], as measured by the Swanson, Noland, and Pelham (SNAP) questionnaire—a common survey used to assess symptoms of inattention and hyperactivity in relation to ADHD [32]. Follow-up studies of OPPS participants in adulthood indicate no evidence of executive function deficits, behaviorally, as measured by classic tasks such as the counting Stroop task, which is used to examine susceptibility to distractor interference. However, neurologically, follow-up studies indicated that OPPS participants exposed prenatally to marijuana had significantly more brain activity while completing executive functioning tasks inside a fMRI scanner compared to unexposed controls [33].
Although the research literature appears to be mixed in terms of results, a common theme is that children exposed to marijuana prenatally perform significantly “differently” in some way on tasks of visual perception and attention than those not exposed. In some cases this results in classically “worse” performance, while in others the performance appears to be “better.” This may not be as contradictory as it seems, and likely reflects the specific characteristics of the way in which visual perception is assessed in each study. We propose the showcased “better” visual processing performance, especially in the ranges described in the published studies, may be less of an advantage and more of an indication of a habituation deficit. For example, in the current study we found seeking out high levels of visual stimulation to be related to prenatal marijuana exposure at 15 months of age. This significantly increased desire for visual stimulation may underlie issues of habituating irrelevant stimuli. Thus, children exposed to marijuana in-utero may have an affinity toward visual stimulation, which in turn may make it difficult to disengage from irrelevant stimuli, or simply discriminate between relevant and irrelevant visual stimuli. Further study grounded in theories of visual processing development is needed to better understand this issue, and to test this proposed understanding and potential relationship between prenatal marijuana exposure and visual attention and habituation deficits.
The current study has many strengths including prospective and detailed assessment of a wide range of variables including substance exposure. However, several limitations are present. First, sensory processing was assessed here via parent report, which can be subject to error related to lack of attention to these issues in their own children and item interpretation differences across parents. Second, while at 225 the sample was reasonably sized for a prospective longitudinal study with comprehensive developmental assessment, investigation of timing of marijuana exposure did reduce the number of participants in each of the marijuana exposure groups to fewer than 50. This resulted in statistical power to find only moderately sized or larger effects. Thus, with a larger sample, it is possible that first trimester marijuana exposure may have significantly predicted visual processing differences, or that gestational marijuana exposure may have significantly predicted processing differences in other domains. A third limitation, related to this issue, is the way in which we examined timing of exposure. Given the patterns of use in this sample, we were only able to examine first trimester use only, or use through to delivery. It is unknown whether use that continues beyond the first trimester but ends at some point prior to delivery predicts processing deficits. It is also unclear whether the patterns of use here are actually proxies for amount of use in that those who quit use by the end of the first trimester may actually be using less marijuana or using less frequently. Thus, use in the first trimester may actually predict sensory processing issues if use is at higher levels than what occurred for our first trimester use only participants. Unfortunately, we did not have reliable data on amount and frequency of use for all of our participants. A fourth limitation of the current study is that while many background differences were controlled for, the number and nature of these differences suggests that there may be other ways, not measured in this study, that women who did and did not use marijuana during pregnancy differed. If these differences, such as continued substance use while parenting, impact sensory processing, it is possible that these factors may partially or even wholly explain the relationships found in this study. Related to this point, due to our sample size, we were also unable to examine the potential impact of prenatal marijuana exposure on sensory processing separately for boys and girls given that gender differences in sensory processing are sometimes evident at this age. While this does not diminish our findings of a global effect regardless of gender, it is a potential avenue of exploration for future studies. A final limitation of this investigation is that the study sample was comprised primarily of disadvantaged rural participants from a region spanning only three states. It is uncertain whether the associations we found generalize beyond this population.
In conclusion, this study provides further evidence of visual processing differences in children prenatally exposed to marijuana, effects not present for other processing domains. This study adds to what is known by demonstrating these effects in children as young as 15 months of age, and also suggests that exposure to marijuana throughout gestation, or at least in the latter stages of gestation, is the primary driver for these effects. Finally, this report suggests that previous seemingly contradictory findings on the association between prenatal marijuana exposure and visual processing may be the result of how this outcome is specifically tested, and that all findings of differences may suggest higher level and later emerging issues with visual attention. Thus, this study adds to the growing body of evidence of the harms of gestational marijuana exposure, and provides further support for clinical recommendations for women to avoid marijuana use during pregnancy.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by East Tennessee State University Medical IRB. Written informed consent to participate in this study was provided by the participants' legal guardian.
## Author contributions
BB: study design, study implementation, data analysis, writing original draft, writing review and editing. JO: writing original draft, writing-review and editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. 1.Substance Abuse and Mental Health Services Administration. Key substance use and mental health indicators in the United States: Results from the 2018 National Survey on Drug Use and Health (HHS Publication No. PEP19-5068, NSDUH Series H-54). Rockville, MD: Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration (2019). Available at:
https://www.samhsa.gov/data/.
2. Lee CM, Neighbors C, Woods BA. **Marijuana motives: young adults’ reasons for using marijuana**. *Addict Behav* (2007.0) **32** 1384-94. DOI: 10.1016/j.addbeh.2006.09.010
3. Hendin H, Haas AP. **The adaptive significance of chronic marijuana use for adolescents and adults**. *Adv Alcohol Subst Abuse* (1985.0) **4** 99-115. DOI: 10.1300/J251v04n03_05
4. Crutchfield RD, Gove WR. **Determinants of drug use: a test of the coping hypothesis**. *Soc Sci Med* (1984.0) **18** 503-9. DOI: 10.1016/0277-9536(84)90008-x
5. Hyman SM, Sinha R. **Stress-related factors in cannabis use and misuse: implications for prevention and treatment**. *J Subst Abuse Treat* (2009.0) **36** 400-13. DOI: 10.1016/j.jsat.2008.08.005
6. Bachhuber M, Arnsten JH, Wurm G. **Use of cannabis to relieve pain and promote sleep by customers at an adult use dispensary**. *J Psychoactive Drugs* (2019.0) **51** 400-4. DOI: 10.1080/02791072.2019.1626953
7. **Marijuana use during pregnancy and lactation committee opinion no. 722**. *Obstet Gynecol* (2017.0) **130** e205-9. DOI: 10.1097/AOG.0000000000002354
8. Ko JY, Coy KC, Haight SC, Haegerich TM, Williams L, Cox S. **Characteristics of marijuana use during pregnancy—eight states, pregnancy risk assessment monitoring system, 2017**. *Morb Mortal Wkly Rep* (2020.0) **69** 1058-63. DOI: 10.15585/mmwr.mm6932a2
9. Crume TL, Juhl AL, Broosk-Russell A, Hall KE, Wymore E, Borgelt LM. **Cannabis use during the perinatal period in a state with legalized recreational and medical marijuana: the association between maternal characteristics, breastfeeding patterns, and neonatal outcomes**. *J Pediatr* (2018.0) **197** 90-6. DOI: 10.1016/j.jpeds.2018.02.005
10. Ko JY, Tong VT, Bombard JM, Hayes DK, Davy J, Perham-Hester KA. **Marijuana use during and after pregnancy and association of prenatal use on birth outcomes: a population-based study**. *Drug Alcohol Depend* (2018.0) **187** 72-8. DOI: 10.1016/j.drugalcdep.2018.02.017
11. Bailey BA, Wood DL, Shah D. **Impact of pregnancy marijuana use on birth outcomes: results from two matched population-based cohorts**. *J Perinatol* (2020.0) **40** 1477-82. DOI: 10.1038/s41372-020-0643-z
12. Marchand G, Masoud AT, Govindan M, Ware K, King A, Ruther S. **Birth outcomes of neonates exposed to marijuana in utero: a systematic review and meta-analysis**. *JAMA Netw Open* (2022.0) **5** e2145653. DOI: 10.1001/jamanetworkopen.2021.45653
13. Fried PA. **The Ottawa prenatal prospective study (OPPS): methodological issues and findings–it's easy to throw the baby out with the bath water**. *Life Sci* (1995.0) **56** 2159-68. DOI: 10.1016/0024-3205(95)00203-i
14. Day NL, Richardson GA, Goldschmidt L, Robles N, Taylor PM, Stoffer DS. **Effect of prenatal marijuana exposure on the cognitive development of offspring at age three**. *Neurotoxicol Teratol* (1994.0) **16** 169-75. DOI: 10.1016/0892-0362(94)90114-7
15. Griffith DR, Azuma SD, Chasnoff IJ. **Three-year outcome of children exposed prenatally to drugs**. *J Am Acad Child Adolesc Psychiatry* (1994.0) **33** 20-7. DOI: 10.1097/00004583-199401000-00004
16. Chakraborty A, Anstice NS, Jacobs RJ, LaGasse LL, Lester BM, Wouldes TA. **Prenatal exposure to recreational drugs affects global motion perception in preschool children**. *Sci Rep* (2015.0) **5** 16921. DOI: 10.1038/srep16921
17. 17.American Psychological Association. Diagnosing and managing autism spectrum disorder (ASD). Available at:
https://www.apa.org/topics/autism-spectrum-disorder/diagnosing.
18. Takarae Y, Sablich SR, White SP, Sweeney JA. **Neurophysiological hyperresponsivity to sensory input in autism spectrum disorders**. *J Neurodev Disord* (2016.0) **8** 29. DOI: 10.1186/s11689-016-9162-9
19. Samson F, Mottron L, Soulières I, Zeffiro TA. **Enhanced visual functioning in autism: an ALE meta-analysis**. *Hum Brain Mapp* (2012.0) **33** 1553-81. DOI: 10.1002/hbm.21307
20. Panagiotidi M, Overton PG, Stafford T. **The relationship between ADHD traits and sensory sensitivity in the general population**. *Compreh Psychiatr* (2018.0) **80** 179-85. DOI: 10.1016/j.comppsych.2017.10.008
21. 21.American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 5th ed. Arlington, VA: American Psychiatric Association (2013).. *Diagnostic and statistical manual of mental disorders* (2013.0)
22. Jamal W, Cardinaux A, Haskins AJ, Kjelgaard M, Sinha P. **Reduced sensory habituation in autism and its correlation with behavioral measures**. *J Autism Dev Disord* (2021.0) **51** 3153-64. DOI: 10.1007/s10803-020-04780-1
23. Jansiewicz EM, Newschaffer CJ, Denckla MB, Mostofsky SH. **Impaired habituation in children with attention deficit hyperactivity disorder**. *Cogn Behav Neurol* (2004.0) **17** 1-8. DOI: 10.1097/00146965-200403000-00001
24. Bailey BA. **Effectiveness of a pregnancy smoking intervention: the Tennessee intervention for pregnant smokers (TIPS) program**. *Health Educ Behav* (2015.0) **42** 824-31. DOI: 10.1177/1090198115590780
25. Morrison C, McCook JG, Bailey BA. **First trimester depression scores predict development of gestational diabetes mellitus in pregnant rural Appalachian women**. *J Psychosom Obstet Gynecol* (2016.0) **37** 21-5. DOI: 10.3109/0167482X.2015.1106473
26. Dunn W, Daniels DB. **Initial development of the infant/toddler sensory profile**. *J Early Interv* (2002.0) **25** 27-41. DOI: 10.1177/105381510202500104
27. Bailey B. **Using expired air carbon monoxide to determine smoking status during pregnancy: preliminary determination of an appropriately sensitive and specific cut-point**. *Addict Behav* (2013.0) **38** 2547-50. DOI: 10.1016/j.addbeh.2013.05.011
28. Hjorthøj CR, Hjorthøj AR, Nordentoft M. **Validity of timeline follow-back for self-reported use of cannabis and other illicit substances–systematic review and meta-analysis**. *Addict Behav* (2012.0) **37** 225-33. DOI: 10.1016/j.addbeh.2011.11.025
29. Leech SL, Richardson GA, Goldschmidt L, Day NL. **Prenatal substance exposure: effects on attention and impulsivity of 6-year-olds**. *Neurotoxicol Teratol* (1999.0) **21** 109-18. DOI: 10.1016/s0892-0362(98)00042-7
30. Riccio CA, Reynolds CR, Lowe PA. *Clinical applications of continuous performance tests: Measuring attention and impulsive responding in children and adults* (2001.0)
31. Goldschmidt L, Day NL, Richardson GA. **Effects of prenatal marijuana exposure on child behavior problems at age 10**. *Neurotoxicol Teratol* (2000.0) **22** 325-36. DOI: 10.1016/S0892-0362(00)00066-0
32. Swanson J, Nolan W, Pelham WE. (1981.0)
33. Smith AM, Mioduszewski O, Hatchard T, Byron-Alhassan A, Fall C, Fried PA. **Prenatal marijuana exposure impacts executive functioning into young adulthood: an fMRI study**. *Neurotoxicol Teratol* (2016.0) **58** 53-9. DOI: 10.1016/j.ntt.2016.05.010
|
---
title: 'Lymphocyte-to-C reactive protein ratio as novel inflammatory marker for predicting
outcomes in hemodialysis patients: A multicenter observational study'
authors:
- Xinpan Chen
- Wang Guo
- Zongli Diao
- Hongdong Huang
- Wenhu Liu
journal: Frontiers in Immunology
year: 2023
pmcid: PMC10017876
doi: 10.3389/fimmu.2023.1101222
license: CC BY 4.0
---
# Lymphocyte-to-C reactive protein ratio as novel inflammatory marker for predicting outcomes in hemodialysis patients: A multicenter observational study
## Abstract
### Background
Patients undergoing hemodialysis experience inflammation, which is associated with a higher risk of mortality. The lymphocyte-to-C reactive protein ratio (LCR) is a novel marker of inflammation that has been shown to predict mortality in patients with malignant cancer. However, the utility of LCR has not been evaluated in patients undergoing hemodialysis.
### Methods
We performed a multi-center cohort study of 3,856 patients who underwent hemodialysis as part of the Beijing Hemodialysis Quality Control and Improvement Project between 1 January 2012 and December 2019. The relationship between LCR and all-cause mortality was assessed using a restricted cubic spline model and a multivariate Cox regression model. An outcome-oriented method was used to determine the most appropriate cut-off value of LCR. Subgroup analysis was also performed to evaluate the relationships of LCR with key parameters.
### Results
Of the 3,856 enrolled patients, 1,581 ($41\%$) were female, and their median age was 62 [53, 73] years. Over a median follow-up period of 75.1 months, 1,129 deaths occurred. The mortality rate for the patients after 60 months was $38.1\%$ ($95\%$ confidence interval (CI) $36\%$–$40.1\%$), resulting in a rate of 93.41 events per 1,000 patient-years. LCR showed an L-shaped dose-response relationship with all-cause mortality. The optimal cut-off point for LCR as a predictor of mortality in hemodialysis patients was 1513.1. An LCR of ≥1513.1 could independently predict mortality (hazard ratio 0.75, $95\%$ CI 0.66–0.85, $P \leq 0.001$).
### Conclusions
Baseline LCR was found to be an independent prognostic biomarker in patients undergoing hemodialysis. Implying that it should be a useful means of improving patient prognosis and judging the timing of appropriate interventions in routine clinical practice.
## Introduction
Persistent inflammation has been shown to facilitate the development of various diseases and increase the associated mortality rates, including for chronic kidney disease, chronic obstructive pulmonary disease, cardiovascular disease, and diabetes (1–5). For patients with end-stage renal disease who are undergoing hemodialysis, the complications of anemia, malnutrition, and vascular calcification are considered to increase morbidity and mortality (6–8), and the underlying mechanisms are inextricably linked to the persistent inflammatory state [9, 10]. This persistent inflammatory state can also lead to protein-energy wasting, resulting in malnutrition, which is associated with a high risk of mortality in patients undergoing hemodialysis [11, 12]. Therefore, paying close attention to the inflammatory status of such patients is important.
Although markers such as PCT, IL-6, TNF-α, and MMP-9 have been shown to be sensitive and accurate means of predicting inflammation and mortality, such assays are either expensive or unavailable in clinical laboratories. Therefore, an alternative prognostic marker that is simple, inexpensive, and convenient to measure, and that is available in most dialysis laboratories, is still required. Complete blood counts are easy and commonly performed measurements that can help predict inflammation, and several combinations of hematological indices, such as the neutrophil-to-lymphocyte ratio (NLR) and the platelet-to-lymphocyte ratio (PLR), have been developed for use as prognostic markers in patients undergoing hemodialysis (13–15). In addition, other makers of systemic inflammation, such as the prognostic nutritional index (PNI) and the Glasgow prognostic score (GPS), can be used to predict outcomes in such patients by quantifying their nutritional and immunological statuses [16, 17].
Recently, the lymphocyte-to-C reactive protein ratio (LCR), which can be easily calculated using the lymphocyte count and serum CRP concentration, has been shown to be associated with the severity of inflammation in and the mortality of patients with malignant disease. Specifically, low LCR has been shown to be an independent predictor of overall survival in patients with colorectal cancer, gastric cancer, hepatocellular carcinoma, bladder cancer, or rectal cancer (18–23). Thus, LCR has shown potential as a predictor of inflammation and mortality. However, to date, no studies have evaluated the relationship between LCR and mortality in patients undergoing hemodialysis. We hypothesized that LCR may also be a useful predictor of mortality in these patients, and in the present study, we performed pooled analyses of clinically relevant variables to determine whether LCR is associated with the overall survival of patients undergoing hemodialysis and whether it interacts with other clinical variables.
## Participants and study design
We performed a retrospective multi-center cohort study of individuals selected from the 6,126 participants in the Beijing Hemodialysis Quality Control and Improvement Project, during which data was collected at 138 dialysis centers between 1 January 2012 and 31 December 2019. The inclusion criteria were age ≥ 18 years and hemodialysis three times a week for at least 3 months. The exclusion criteria were as follows [1]: duration of hemodialysis < 3 months; [2] previous treatment by peritoneal dialysis; [3] organ transplantation; [4] malignant disease; [5] autoimmune disease, or chronic or acute infectious disease; and [6] missing baseline data. The existence of acute or chronic infection was determined based on the admission diagnosis of patients that explicitly stated infection (for example, pneumonia) or that provided evidence of infection (detected bacteria or virus). After the application of these criteria, 3,856 patients were eligible for enrolment in the present study (Figure S1). The study was performed in accordance with the principles of the Declaration of Helsinki and was approved by the Human Ethics Committee of Beijing Friendship Hospital.
## Clinical data
Variables and potential confounders were selected for study by considering clinical guidelines and the results of previously published studies. Only data collected during the first assessment made after the initiation of hemodialysis were included. The baseline demographic data (age and sex) and biochemical/hematological data (hemoglobin, albumin, platelet count, neutrophil count, lymphocyte count, C-reactive protein, creatinine, urea, calcium, parathormone, phosphorus, Fe, ferritin and UIBC) were obtained within the first month of hemodialysis. The biochemical/hematological data were standardized before being recorded in the Beijing Hemodialysis Quality Control and Improvement Project database to minimize variability in the measurements made by the various dialysis laboratories. All of these data were collected from the project database. The patients were followed from the initiation of hemodialysis until they were transferred to another hemodialysis center, they were changed to peritoneal dialysis, they underwent kidney transplantation, they were lost to follow-up, they died, or the end of the study period on 31 December 2019. The LCR was calculated as lymphocyte count (109/L)/C-reactive protein (mg/L) × 104. For the subsequent stratified analyses, on the basis of the laboratory reference ranges and the Kidney Disease Outcome Quality Initiative (KDOQI) guidelines, the biochemical data were defined as normal or abnormal. The normal values were as follows: hemoglobin 100–130 g/L, platelet count 125–350×109/L, neutrophil count 1.8–6.3×109/L, albumin >35 g/L, calcium 2.1–2.52 mmol/L, parathormone 150–300 pg/μl, phosphorus 1.13–1.78 mmol/L, ferritin 200–500 ng/ml.
## Outcome
The overall survival time was defined as the interval between the initiation of hemodialysis and the date of death, transfer to another dialysis center, the change to peritoneal dialysis, kidney transplantation, withdrawal from the study, or the end of the study period (31 December 2019). We aimed to evaluate the relationship between LCR and overall survival and to identify the most appropriate cut-off value of LCR for prognostic use in patients undergoing regular hemodialysis.
## Statistical analysis
Continuous datasets with skewed distributions are presented as median (interquartile range) and were compared using the Mann-Whitney test. Categorical data are expressed as numbers (percentages) and were compared using Pearson’s χ2 test. Spearman correlation analysis was used to identify linear relationships between LCR and selected variables. Univariate and multivariate Cox proportional hazards regression models were used to identify the risk factors for mortality and to provide hazard ratios (HRs) for each variable. Sex and age were included in the multivariate Cox regression model and other potentially confounding variables were selected based on the results of univariate analysis ($p \leq 0.1$ was defined as indicating a significant association with all-cause mortality). For the time-to-event analysis, survival curves were generated using the Kaplan-Meier method and compared using the log-rank test.
The non-linear relationship between LCR and HR was visualized using restricted cubic splines. The optimal cut-off value for LCR was determined using the ‘surv_cutpoint’ formula in the ‘survminer’ R package, which is an outcome-oriented method that provides a cut-off value with the closest relationship with the outcome. The participants were divided into high- and low-LCR groups based on the optimal cut-off value. Trend tests were performed by assigning a median value to each quartile of the LCR, which was then modeled as a continuous variable, and the Wald test was used to assess statistical significance. Forest plots were used to visualize the results of the analysis of the effects of the interactions of LCR with other variables on the overall survival of the participants. Calibration curves were used to visualize the differences between the predicted and observed probabilities. The accuracy of predictions made using the LCR risk score was assessed using the area under the ROC curve. All the tests performed were two-sided and $P \leq 0.05$ was considered to represent statistical significance. Statistical analyses were performed using R software version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria), and the packages ‘survminer’, ‘survival’, ‘rms’,’ timeROC’, ‘forestplot’, ‘ggrisk’, ‘ggplot2’, ‘ggsci’, ‘tableone’, and ‘dplyr’.
## Participant characteristics and the relationships between LCR and other variables
From the original total of 6,126 patients, we excluded 2,270 for the reasons listed above, leaving a total of 3,856 participants in the cohort. They had a median age of 62 [53, 73] years and 1,581 ($41\%$) were female. The overall characteristics of the participants and the results of their stratification according to the calculated LCR cut-off value are presented in Table 1. In summary, 1,129 deaths were recorded over a median follow-up period of 75.1 months. The overall mortality rate for the participants 60 months after their initial assessment was $38.1\%$ ($95\%$ CI $36\%$–$40.1\%$), resulting in a rate of 93.41 events per 1,000 patient-years. The log LCR was compared among the various sex and age groups comprising the sample. The results showed that young (<65 years old) participants had a significantly higher LCR than older (≥65 years old) participants ($P \leq 0.001$) (Figure S2). Moreover, *Spearman analysis* of the relationships of LCR with various parameters that are clinically relevant for patients undergoing hemodialysis showed significant correlations with age, hemoglobin level, neutrophil count, albumin concentration, and calcium concentration. On the basis of these results, we also conducted a stratified analysis of participants according to their age (≥65 years old vs. <65 years old) and sex (female vs. male), which showed that hemoglobin level, albumin concentration, and calcium concentration positively correlated with LCR in individuals of differing age and sex, while the neutrophil count negatively correlated (Figure 1).
## Relationship of LCR with overall survival
Potential risk factors for all-cause mortality were identified using univariate and multivariate Cox proportional hazards regression analyses (Table 2). Univariate analysis revealed that age, platelet count, lymphocyte count, CRP, creatinine, urea, albumin, phosphorus, ferritin, and LCR were associated with the overall survival of the participants. The multivariate analysis revealed that age, platelet count, albumin, phosphorus, and LCR were independent predictors of mortality. Then, according to the results of the multivariate analysis, we evaluated the prognostic value of combinations of LCR with the other independent prognostic factors, and the 1, 3, 5, and 7-year calibration curves showed that these combinations would be extremely useful for the prediction of survival (Figure S3). The cut-off point of LCR determined by using Kaplan-Meier curves was 1513.1 (Figure 2A), of which hemodialysis patients with an LCR <1513.1 were found to be associated with higher mortality.
To further evaluate the relationship between LCR and mortality in patients undergoing hemodialysis, four categories of LCR were defined. First, as a continuous variable, the restricted cubic spline plot showed that LCR had an L-shaped dose-response relationship with the risk of all-cause mortality risk in the participants (Figure 3). Second, the Cox regression models of the relationship between LCR and OS showed that LCR positively correlated with prognosis (HR 0.85 per SD increase, $95\%$ CI 0.75–0.95, $$P \leq 0.006$$) after adjustment for sex, age, hemoglobin level, platelet count, neutrophil count, creatinine, urea, albumin, calcium, parathormone, phosphorus, Fe, ferritin, and unsaturated iron-binding capacity (UIBC) (Table 3). Third, using the cut-off value calculated for LCR, the participants were divided into two groups: a low-LCR group and a high-LCR group. Compared with the low-LCR group, participants in the high-LCR group had a consistently better prognosis (HR 0.75, $95\%$ CI 0.66–0.85, $P \leq 0.001$) after adjustment for the variables listed above. Finally, the participants were divided into quartiles according to their LCR; and compared with the first quartile (Q1, <1,054.3), the second (1,054.3–3,144.9), third (3,144.9–9,068.6), and fourth quartiles (≥9,068.6) all had a better prognosis (P for trend <0.05). After adjustment for the potential confounding factors, the HRs for all-cause mortality were 0.78 ($95\%$ CI 0.66–0.92, $$P \leq 0.003$$), 0.73 ($95\%$ CI 0.62–0.87, $P \leq 0.001$), and 0.76 ($95\%$ CI 0.64–0.91, $$P \leq 0.003$$) for the second, third, and fourth quartiles, respectively.
**Figure 3:** *Relationships between LCR, as a continuous variable, and the hazard ratio for overall survival. Restricted cubic splines (RCSs) were used. (A) Unadjusted restricted cubic spline for LCR; (B) RCS adjusted for sex, age, hemoglobin level, platelet count, neutrophil count, creatinine, urea, albumin, calcium, parathormone, phosphorus, Fe, ferritin, and UIBC.* TABLE_PLACEHOLDER:Table 3
## Demographics and disease characteristics of the participants after stratification according to LCR
According to the cut-off point calculated for LCR, the 3,856 participants were divided into two groups: a low-LCR group (LCR <1513.1, $$n = 1$$,223) and a high-LCR group (LCR ≥1513.1, $$n = 2633$$). The Kaplan-Meier curves and log-rank test results revealed that the high-LCR group had a better prognosis than the low-LCR group (Figure 2B). Table 1 presents a comparison of the demographics and clinical characteristics of the low- and high-LCR groups. Briefly, the participants in the low-LCR group were older; and had lower hemoglobin, higher neutrophil counts, lower lymphocyte counts, higher CRP, lower creatinine, lower urea, lower albumin, lower calcium, lower phosphorus, lower Fe, higher ferritin, and lower UIBC than those in the high-LCR group.
## Results of the stratification analysis
Stratified analyses were conducted to evaluate the relationships between LCR and the HR for overall mortality in various subgroups (Figure 4 and Table S1). Overall, high LCR was consistently associated with a lower risk of death in the participants, irrespective of their subgrouping. However, although the same trend was also present in participants with a normal calcium concentration, it was not statistically significant ($$P \leq 0.234$$). Moreover, an analysis was performed to explore the interactions between high LCR and the other variables, but this showed no associations between high LCR and low overall risk of mortality in the participants (P for interaction >0.05 in all instances). In addition, the LCR and covariates were then cross-classified to better understand the effects of each variable (Table S2). This analysis showed that an abnormal status with regard to any of a number of variables and a low LCR had an additive effect to increase the risk of mortality. Kaplan-Meier curves also showed that combinations of low LCR and abnormalities in other variables had a deleterious effect on mortality. Specifically, participants with an LCR of <1513.1 had the worst survival rate when they were ≥ 65 years old and had an abnormal albumin concentration (Figure S4).
**Figure 4:** *Relationships between LCR category, based on the calculated cut-off value, and the hazard ratios for overall survival in the subgroups. The model was adjusted for sex, age, hemoglobin level, platelet count, neutrophil count, creatinine, urea, albumin, calcium, parathormone, phosphorus, Fe, ferritin, and UIBC. The normal ranges for each parameter were as follows: hemoglobin 100–130 g/L, platelet count 125–350×109/L, neutrophil count 1.8–6.3×109/L, albumin >35 g/L, calcium 2.1–2.52 mmol/L, parathormone 150–300 pg/μl, phosphorus 1.13–1.78 mmol/L, and ferritin 200–500 ng/ml.*
## Results of the sensitivity analysis and internal validation
To validate the finding that LCR is a useful predictor of mortality in patients undergoing hemodialysis, a sensitivity analysis and internal validation were performed to assess the robustness of the results (Table 4). An analysis performed after excluding the participants who died within 6 months of the first assessment showed that LCR remained an independent predictor of mortality (adjusted HR 0.85, $95\%$ CI 0.75–0.96, $$P \leq 0.007$$ for a high LCR per SD). Subsequently, the full cohort was randomly assigned at a 7:3 ratio to validation cohort A ($$n = 2$$,681) or validation cohort B ($$n = 1$$,175) using computer-generated random numbers (Table S3). Similar results were obtained for cohort A (adjusted HR 0.87, $95\%$ CI 0.78–0.98, $$P \leq 0.026$$ for high LCR per SD) and cohort B (adjusted HR 0.69, $95\%$ CI 0.49–0.97, $$P \leq 0.034$$ for high LCR per SD). Moreover, Kaplan-Meier curves showed that participants with a high LCR had a better prognosis when in either validation cohort A or B (Figures S5A, B).
**Table 4**
| LCRSensitive analysis | Crude model | Crude model.1 | Model A | Model A.1 | Model B | Model B.1 |
| --- | --- | --- | --- | --- | --- | --- |
| LCRSensitive analysis | HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value |
| Excluding patients dying within 6 months | Excluding patients dying within 6 months | Excluding patients dying within 6 months | Excluding patients dying within 6 months | Excluding patients dying within 6 months | Excluding patients dying within 6 months | Excluding patients dying within 6 months |
| As continuous (per SD) | 0.80 (0.70-0.91) | <0.001 | 0.84 (0.75-0.95) | 0.006 | 0.85 (0.75-0.96) | 0.007 |
| By LCR cut-off | By LCR cut-off | By LCR cut-off | By LCR cut-off | By LCR cut-off | By LCR cut-off | By LCR cut-off |
| Low (<1513.1) | Ref | | Ref | | Ref | |
| High (≥1513.1) | 0.66 (0.58,0.75) | <0.001 | 0.70 (0.62-0.80) | <0.001 | 0.74 (0.65-0.85) | <0.001 |
| Interquartile | Interquartile | Interquartile | Interquartile | Interquartile | Interquartile | Interquartile |
| Q1 (<1054.3) | Ref | | Ref | | Ref | |
| Q2 (1054.3-3144.9) | 0.73 (0.62-0.86) | <0.001 | 0.73 (0.62-0.86) | <0.001 | 0.77 (0.65-0.91) | 0.002 |
| Q3 (3144.9-9068.6) | 0.68 (0.58-0.81) | <0.001 | 0.73 (0.62-0.86) | <0.001 | 0.76 (0.64-0.80) | 0.002 |
| Q4 (≥9068.6) | 0.61 (0.52-0.72) | <0.001 | 0.71 (0.60-0.84) | <0.001 | 0.77 (0.64-0.92) | 0.004 |
| P for trend | | <0.001 | | <0.001 | | 0.007 |
| Validation cohort A | Validation cohort A | Validation cohort A | Validation cohort A | Validation cohort A | Validation cohort A | Validation cohort A |
| As continuous (per SD) | 0.83 (0.73-0.94) | 0.04 | 0.87 (0.77-0.98) | 0.024 | 0.87 (0.78-0.98) | 0.026 |
| By LCR cut-off | By LCR cut-off | By LCR cut-off | By LCR cut-off | By LCR cut-off | By LCR cut-off | By LCR cut-off |
| Low (<1513.1) | Ref | | Ref | | Ref | |
| High (≥1513.1) | 0.7 (0.6-0.81) | <0.001 | 0.74 (0.64-0.86) | <0.001 | 0.76 (0.65-0.88) | 0.008 |
| Interquartile | Interquartile | Interquartile | Interquartile | Interquartile | Interquartile | Interquartile |
| Q1 (<1054.3) | Ref | | Ref | | Ref | |
| Q2 (1054.3-3144.9) | 0.75 (0.61-0.9) | 0.003 | 0.75 (0.61-0.91) | 0.003 | 0.76 (0.62-0.93) | 0.008 |
| Q3 (3144.9-9068.6) | 0.7 (0.58-0.85) | <0.001 | 0.74 (0.61-0.9) | 0.002 | 0.75 (0.61-0.92) | 0.005 |
| Q4 (≥9068.6) | 0.68 (0.56-0.83) | <0.001 | 0.78 (0.64-0.95) | 0.014 | 0.81 (0.66-0.99) | 0.045 |
| P for trend | | <0.001 | | 0.018 | | 0.061 |
| Validation cohort B | Validation cohort B | Validation cohort B | Validation cohort B | Validation cohort B | Validation cohort B | Validation cohort B |
| As continuous (per SD) | 0.63 (0.45-0.89) | 0.009 | 0.67 (0.47-0.95) | 0.025 | 0.69 (0.49-0.97) | 0.034 |
| By LCR cut-off | | | | | | |
| Low (<1513.1) | Ref | | Ref | | Ref | |
| High (≥1513.1) | 0.57 (0.46-0.72) | <0.001 | 0.64 (0.51-0.81) | <0.001 | 0.7 (0.55-0.88) | 0.003 |
| Interquartile | Interquartile | Interquartile | Interquartile | Interquartile | Interquartile | Interquartile |
| Q1 (<1054.3) | Ref | | Ref | | Ref | |
| Q2 (1054.3-3144.9) | 0.74 (0.56-0.98) | 0.037 | 0.73 (0.55-0.97) | 0.028 | 0.85 (0.63-1.13) | 0.263 |
| Q3 (3144.9-9068.6) | 0.56 (0.41-0.76) | <0.001 | 0.62 (0.45-0.83) | 0.002 | 0.65 (0.47-0.89) | 0.008 |
| Q4 (≥9068.6) | 0.46 (0.34-0.63) | <0.001 | 0.56 (0.41-0.77) | <0.001 | 0.66 (0.47-0.94) | 0.019 |
| P for trend | | <0.001 | | <0.001 | | 0.006 |
## Construction of a risk score model
LCR values were naturally log-transformed to make it easier to assess the prognosis of patients undergoing hemodialysis, and then a risk score model was constructed using the β regression risk coefficient derived from the multivariate Cox regression model and the log LCR. The formula for the risk score was determined to be −0.1669 × log LCR. Using the calculated risk scores, a risk plot heatmap and a time-dependent ROC curve were created. The results, shown in Figure 5A, indicate that a high LCR is associated with a low risk score and a better prognosis. Figure 5B shows AUCs of 62.0, 56.5, 58.3, and 54.9 for 1, 3, 5, and 7 years, respectively.
**Figure 5:** *Risk score model for the participants, based on LCR. Prognostic risk score model for the participants, based on LCR. (A) High LCR is associated with a lower risk score and a better prognosis. (B) AUCs of 62.0, 56.5, 58.3, and 54.9 were calculated for 1, 3, 5, and 7 years, respectively. AUC, area under the curve.*
## Discussion
In the present study, we have demonstrated the prognostic utility of LCR for patients undergoing hemodialysis in a number of ways. First, we have shown that the baseline LCR is an independent predictor and has an L-shaped relationship with all-cause mortality in these patients. Second, we have calculated a specific cut-off value of LCR for use with patients undergoing maintenance hemodialysis and demonstrated that baseline LCR values above this level are associated with better survival. Third, we have constructed a prognostic risk model using LCR that is an excellent predictor of prognosis. Finally, we have performed sensitivity and internal validation analyses in which we obtained similar results.
Inflammation is a physiological response to various deleterious stimuli under normal conditions and represents the first stage of healing. Cohen et al. measured the circulating pro-inflammatory cytokine (IL-1, IL-6, and TNF-α) concentrations of 231 patients undergoing hemodialysis and found that those with high concentrations had shorter survival times than the others, which demonstrates that biomarkers of the severity of inflammation are independent and reliable prognostic indicators in such patients [24]. In the context of chronic kidney disease, and especially in patients undergoing hemodialysis, unmanaged and sustained systemic inflammation caused by uremic toxins, oxidative stress, fluid overload, and/or artificial materials have been identified to be critical in patients with cardiovascular disease, malnutrition, and anemia, and are associated with high mortality [25], and this is consistent with the results of the *Spearman analysis* in the present study, which showed that hemoglobin level, albumin, and calcium positively correlate with LCR in individuals of various ages and sexes.
The concentrations of specific pro-inflammatory cytokines, such as IL-1, IL-6, MMP-9, and TNF-α, can often not be measured routinely, and such assays are often not affordable for use with patients with hemodialysis, rendering them of limited use in clinical practice. Identifying useful and readily measurable biomarkers of inflammation and identifying patients at high risk would therefore be extremely valuable to permit early intervention (lifestyle, medication, or the modification of dialysis). A number of markers of inflammation, such as NLR, PLR, PNI, and GPS, have been reported to be a useful means of evaluating the prognosis of patients with hemodialysis, but none have become established as gold standards for the evaluation of the inflammatory status of patients. In the present study, we studied the utility of LCR, a newly developed marker of inflammation.
Lymphocytes play an important role in the cytotoxic immune response, and the loss of circulating lymphocytes is associated with a loss of immunological specificity, resulting in higher all-cause mortality in older individuals [26]. A previous study of the relationship between white blood cell count and mortality in 44,114 patients undergoing hemodialysis showed that a high lymphocyte count is associated with a lower risk of mortality (HR 0.86, $95\%$ CI 0.83–0.89, $P \leq 0.001$) [27]. CRP has served as a useful marker of infection and tissue inflammation for several decades, and has been shown to be regulated by the pro-inflammatory cytokines TNF-α and IL-6 in an in vitro study [28]. In addition, linear risk relationships of CRP with coronary heart disease, stroke, and mortality in the healthy population have been found. In a large international multi-center cohort study of patients undergoing hemodialysis, CRP was found to be a better predictor of mortality 1 year later than circulating ferritin concentration and white blood cell count [29]. Consistent with the results of a previous study, both the lymphocyte count and the C-reactive protein concentration were found to be predictors of mortality in the present study. LCR was calculated using the lymphocyte count and the CRP concentration, which is less variable and a better predictor than either lymphocyte count or CRP alone, and was first studied and shown to have prognostic value with respect to colorectal cancer by Okugawa et al. in 2020, after various combinations of pro-inflammatory marker concentrations in preoperative blood samples, including neutrophil count, lymphocyte count, CRP concentration, albumin concentration, and platelet count, were evaluated [18].
In recent years, the use of a high LCR as a predictor of overall survival has been validated in patients with stomach cancer, hepatocellular carcinoma, bladder cancer, or rectal cancer (19–23). The results of the present study are consistent with the published literature because high LCR was found to be consistently associated with better overall survival in patients undergoing hemodialysis and in the various subgroups. To the best of our knowledge, this is the first study to demonstrate the value of LCR for the prediction of overall survival in patients undergoing hemodialysis. Thus, LCR represents a valid means of assessing the prognosis of patients at the initiation of hemodialysis. In future studies, the prognostic utility of LCR, PLR, NLR, and PNI in such patients would be worth comparing.
In the present study cohort, hemodialysis patients with an LCR <1513.1 was found to be associated with higher mortality, which differs from that previously reported for use in patients with cancer. The cut-off value calculated for colorectal cancer was 6,000 and that for intrahepatic cholangiocarcinoma was 7,873.1 [18, 30], both of which are higher than that calculated in the present study. This may be explained by the LCR cut-off value being associated with differing levels of inflammation or comorbidities under the various disease conditions. The stratified analysis performed in the present study shows that LCR has prognostic value when used alongside various clinical parameters, low LCR patients with an abnormal clinical parameter always have an additive effect in increasing the risk of mortality, but that there are no significant interactions between them. We also found that there is a positive correlation between LCR and circulating albumin concentration, which reflects nutritional status relatively well. Moreover, when the participants were categorized according to the cut-off value for LCR, those in the low-LCR group had low creatinine, urea, and albumin concentrations. This may be explained by the presence of malnutrition-inflammation-cachexia syndrome in these patients with chronic kidney disease, but multiple mechanisms are likely to be involved. Increases in the concentrations of pro-inflammatory cytokines can induce anorexia, which is accompanied by chronic fatigue and the breakdown of muscle proteins, ultimately leading to a reduction in nutrient intake, greater resting energy expenditure, and muscle atrophy [31, 32]. Patients with low LCR are more likely to develop malnutrition-inflammation-cachexia syndrome, involving low circulating creatinine, urea, and albumin concentrations.
Thus, considering the results of our study that hemodialysis patients with low LCR levels are associated with a poor outcome and sustained inflammatory status, early intervention in improving patient’s outcome could be done in the below three aspects. First, building a healthy lifestyle with a balanced diet, regular physical exercise and quitting smoking [33, 34]. Second, choosing the right medication which has proved to have a positive effect on inflammation when dealing with the complications of CKD or other comorbidities, such as angiotensin-converting enzyme inhibitors in the treatment of hypertension and sevelamer in the treatment of hyperphosphatemia [35, 36]. Third, formulating appropriate dialysis strategies, such as increasing the frequency of dialysis, the application of hemodiafiltration (HDF) or longer dialysis sessions [37].
To the best of our knowledge, this is one of the largest studies to evaluate the relationship between markers of inflammation and the survival of patients undergoing hemodialysis, and the only study to assess whether LCR is independently associated with their survival. However, the present study had several limitations. First, other conventional markers of systemic inflammation, such as the platelet-to-lymphocyte ratio and the neutrophil-to-lymphocyte ratio, were not analyzed in this study, and additional parameters, such as IL-6 and TGF-β, should be included in more comprehensive evaluations. Second, whether variability in serial LCRs over the course of hemodialysis in individual patients is associated with clinical outcomes should be evaluated, to determine whether LCR values calculated at time-points other than baseline also have prognostic value. Third, the present study was a multi-center retrospective study, and there would have been some unidentified confounders that could have contributed to bias in the data obtained. Further well-designed prospective trials are necessary to circumvent this limitation. Finally, external validation of the findings should be performed using large samples in multiple geographical regions, to permit the generalization of the findings to all patients undergoing hemodialysis.
In conclusion, baseline LCR is an independent prognostic marker in patients who are undergoing hemodialysis, with high LCR being associated with superior outcomes. This implies that the calculation of LCR may be a useful means of improving patient prognosis and identifying the appropriate timing for interventions in routine clinical practice.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by Human Ethics Committees of Beijing Friendship Hospital. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
XC and WG conceived the study. WG provided the data. XC and WG analyzed the results and wrote the manuscript. XC, WG, ZD, WL, and HH analyzed the results and reviewed the manuscript. ZD, WL, and HH had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. All the authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1101222/full#supplementary-material
## References
1. Cobo G, Lindholm B, Stenvinkel P. **Chronic inflammation in end-stage renal disease and dialysis**. *Nephrol Dial Transplant* (2018.0) **33**. DOI: 10.1093/ndt/gfy175
2. Dicker AJ, Huang JTJ, Lonergan M, Keir HR, Fong CJ, Tan B. **The sputum microbiome, airway inflammation, and mortality in chronic obstructive pulmonary disease**. *J Allergy Clin Immunol* (2021.0) **147**. DOI: 10.1016/j.jaci.2020.02.040
3. Speer T, Dimmeler S, Schunk SJ, Fliser D, Ridker PM. **Targeting innate immunity-driven inflammation in CKD and cardiovascular disease**. *Nat Rev Nephrol.* (2022.0) **18**. DOI: 10.1038/s41581-022-00621-9
4. Rohm TV, Meier DT, Olefsky JM, Donath MY. **Inflammation in obesity, diabetes, and related disorders**. *Immunity* (2022.0) **55** 31-55. DOI: 10.1016/j.immuni.2021.12.013
5. Lawler PR, Bhatt DL, Godoy LC, Luscher TF, Bonow RO, Verma S. **Targeting cardiovascular inflammation: next steps in clinical translation**. *Eur Heart J* (2021.0) **42**. DOI: 10.1093/eurheartj/ehaa099
6. Maruyama Y, Kanda E, Kikuchi K, Abe M, Masakane I, Yokoo T. **Association between anemia and mortality in hemodialysis patients is modified by the presence of diabetes**. *J Nephrol.* (2021.0) **34**. DOI: 10.1007/s40620-020-00879-x
7. Sahathevan S, Khor BH, Ng HM, Gafor AHA, Mat Daud ZA, Mafra D. **Understanding development of malnutrition in hemodialysis patients: A narrative review**. *Nutrients* (2020.0) **12**. DOI: 10.3390/nu12103147
8. Fitzpatrick J, Kim ED, Sozio SM, Jaar BG, Estrella MM, Monroy-Trujillo JM. **Calcification biomarkers, subclinical vascular disease, and mortality among multiethnic dialysis patients**. *Kidney Int Rep* (2020.0) **5**. DOI: 10.1016/j.ekir.2020.07.033
9. Viegas C, Araujo N, Marreiros C, Simes D. **The interplay between mineral metabolism, vascular calcification and inflammation in chronic kidney disease (CKD): challenging old concepts with new facts**. *Aging (Albany NY).* (2019.0) **11**. DOI: 10.18632/aging.102046
10. Ganz T. **Anemia of inflammation**. *N Engl J Med* (2019.0) **381**. DOI: 10.1056/NEJMra1804281
11. Koppe L, Fouque D, Kalantar-Zadeh K. **Kidney cachexia or protein-energy wasting in chronic kidney disease: facts and numbers**. *J Cachexia Sarcopenia Muscle.* (2019.0) **10**. DOI: 10.1002/jcsm.12421
12. Rahman T, Khor BH, Sahathevan S, Kaur D, Latifi E, Afroz M. **Protein energy wasting in a cohort of maintenance hemodialysis patients in Dhaka, Bangladesh**. *Nutrients* **14**. DOI: 10.3390/nu14071469
13. Zhang J, Lu X, Wang S, Li H. **High neutrophil-to-Lymphocyte ratio and platelet-to-Lymphocyte ratio are associated with poor survival in patients with hemodialysis**. *BioMed Res Int* (2021.0) **2021**. DOI: 10.1155/2021/9958081
14. Zeng Y, Chen Z, Chen Q, Zhan X, Long H, Peng F. **Neutrophil to lymphocyte ratio predicts adverse cardiovascular outcome in peritoneal dialysis patients younger than 60 years old**. *Mediators Inflamm* (2020.0) **2020**. DOI: 10.1155/2020/4634736
15. Wang J, Huang L, Xu M, Yang L, Deng X, Li B. **Study on the clinical implications of NLR and PLR for diagnosing frailty in maintenance hemodialysis patients and their correlations with patient prognosis**. *J Healthc Eng.* (2022.0) **2022**. DOI: 10.1155/2022/1267200
16. Maraj M, Kusnierz-Cabala B, Dumnicka P, Gala-Bladzinska A, Gawlik K, Pawlica-Gosiewska D. **Malnutrition, inflammation, atherosclerosis syndrome (MIA) and diet recommendations among end-stage renal disease patients treated with maintenance hemodialysis**. *Nutrients* (2018.0) **10**. DOI: 10.3390/nu10010069
17. Miyasato Y, Hanna RM, Morinaga J, Mukoyama M, Kalantar-Zadeh K. **Prognostic nutritional index as a predictor of mortality in 101,616 patients undergoing hemodialysis**. *Nutrients* (2023.0) **15**. DOI: 10.3390/nu15020311
18. Okugawa Y, Toiyama Y, Yamamoto A, Shigemori T, Ide S, Kitajima T. **Lymphocyte-c-reactive protein ratio as promising new marker for predicting surgical and oncological outcomes in colorectal cancer**. *Ann Surg* (2020.0) **272**. DOI: 10.1097/SLA.0000000000003239
19. Miyatani K, Sawata S, Makinoya M, Miyauchi W, Shimizu S, Shishido Y. **Combined analysis of preoperative and postoperative lymphocyte-c-reactive protein ratio precisely predicts outcomes of patients with gastric cancer**. *BMC Cancer.* (2022.0) **22** 641. DOI: 10.1186/s12885-022-09716-9
20. Zhang YF, Lu LH, Zhong C, Chen MS, Guo RP, Wang L. **Prognostic value of the preoperative lymphocyte-C-Reactive protein ratio in hepatocellular carcinoma patients treated with curative intent: A Large-scale multicentre study**. *J Inflammation Res* (2021.0) **14**. DOI: 10.2147/JIR.S311994
21. Zhang H, Wang Y, Ni J, Shi H, Zhang T, Zhang Y. **Prognostic value of lymphocyte-C-Reactive protein ratio in patients undergoing radical cystectomy for bladder cancer: A population-based study**. *Front Oncol* (2021.0) **11**. DOI: 10.3389/fonc.2021.760389
22. Okugawa Y, Toiyama Y, Fujikawa H, Ide S, Yamamoto A, Omura Y. **Prognostic potential of lymphocyte-C-Reactive protein ratio in patients with rectal cancer receiving preoperative chemoradiotherapy**. *J Gastrointest Surg* (2021.0) **25** 492-502. DOI: 10.1007/s11605-019-04495-4
23. Ni HH, Lu Z, Huang X, Ning SW, Liang XL, Zhang SY. **Combining pre- and postoperative lymphocyte-C-Reactive protein ratios can better predict hepatocellular carcinoma prognosis after partial hepatectomy**. *J Inflammation Res* (2022.0) **15**. DOI: 10.2147/JIR.S359498
24. Cohen SD, Phillips TM, Khetpal P, Kimmel PL. **Cytokine patterns and survival in haemodialysis patients**. *Nephrol Dial Transplant.* (2010.0) **25**. DOI: 10.1093/ndt/gfp625
25. Manabe I. **Chronic inflammation links cardiovascular, metabolic and renal diseases**. *Circ J* (2011.0) **75**. DOI: 10.1253/circj.cj-11-1184
26. Ferrando-Martinez S, Romero-Sanchez MC, Solana R, Delgado J, de la Rosa R, Munoz-Fernandez MA. **Thymic function failure and c-reactive protein levels are independent predictors of all-cause mortality in healthy elderly humans**. *Age (Dordr).* (2013.0) **35**. DOI: 10.1007/s11357-011-9341-2
27. Reddan DN, Klassen PS, Szczech LA, Coladonato JA, O’Shea S, Owen WF. **White blood cells as a novel mortality predictor in haemodialysis patients**. *Nephrol Dial Transplant.* (2003.0) **18**. DOI: 10.1093/ndt/gfg066
28. Mortensen RF. **C-reactive protein, inflammation, and innate immunity**. *Immunol Res* (2001.0) **24**. DOI: 10.1385/IR:24:2:163
29. Bazeley J, Bieber B, Li Y, Morgenstern H, de Sequera P, Combe C. **C-reactive protein and prediction of 1-year mortality in prevalent hemodialysis patients**. *Clin J Am Soc Nephrol.* (2011.0) **6**. DOI: 10.2215/CJN.00710111
30. Yugawa K, Itoh S, Yoshizumi T, Morinaga A, Iseda N, Toshima T. **Lymphocyte-c-reactive protein ratio as a prognostic marker associated with the tumor immune microenvironment in intrahepatic cholangiocarcinoma**. *Int J Clin Oncol* (2021.0) **26**. DOI: 10.1007/s10147-021-01962-4
31. Graterol Torres F, Molina M, Soler-Majoral J, Romero-Gonzalez G, Rodriguez Chitiva N, Troya-Saborido M. **Evolving concepts on inflammatory biomarkers and malnutrition in chronic kidney disease**. *Nutrients* (2022.0) **14**. DOI: 10.3390/nu14204297
32. Stenvinkel P. **Can treating persistent inflammation limit protein energy wasting**. *Semin Dial* (2013.0) **26**. DOI: 10.1111/sdi.12020
33. Ikizler TA, Robinson-Cohen C, Ellis C, Headley SAE, Tuttle K, Wood RJ. **Metabolic effects of diet and exercise in patients with moderate to severe CKD: A randomized clinical trial**. *J Am Soc Nephrol.* (2018.0) **29**. DOI: 10.1681/ASN.2017010020
34. Kaesler N, Baid-Agrawal S, Grams S, Nadal J, Schmid M, Schneider MP. **Low adherence to CKD-specific dietary recommendations associates with impaired kidney function, dyslipidemia, and inflammation**. *Eur J Clin Nutr* (2021.0) **75**. DOI: 10.1038/s41430-020-00849-3
35. Ateya AM, El Hakim I, Shahin SM, El Borolossy R, Kreutz R, Sabri NA. **Effects of ramipril on biomarkers of endothelial dysfunction and inflammation in hypertensive children on maintenance hemodialysis: the SEARCH randomized placebo-controlled trial**. *Hypertension* (2022.0) **79**. DOI: 10.1161/HYPERTENSIONAHA.122.19312
36. Rodriguez-Osorio L, Zambrano DP, Gracia-Iguacel C, Rojas-Rivera J, Ortiz A, Egido J. **Use of sevelamer in chronic kidney disease: beyond phosphorus control**. *Nefrologia* (2015.0) **35**. DOI: 10.1016/j.nefro.2015.05.022
37. Calo LA. **Hemodiafiltration and reduction of inflammation in dialysis patients**. *Kidney Int* (2014.0) **86** 651. DOI: 10.1038/ki.2014.157
|
---
title: Determination of the postprandial cut-off value of triglyceride after a daily
meal corresponding to fasting optimal triglyceride level in Chinese subjects
authors:
- Yingying Xie
- Liling Guo
- Hao Chen
- Jin Xu
- Peiliu Qu
- Liyuan Zhu
- Yangrong Tan
- Miao Zhang
- Tie Wen
- Ling Liu
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC10017968
doi: 10.3389/fnut.2023.1037270
license: CC BY 4.0
---
# Determination of the postprandial cut-off value of triglyceride after a daily meal corresponding to fasting optimal triglyceride level in Chinese subjects
## Abstract
### Background
According to the 2021 consensus statement about triglyceride (TG)-rich lipoproteins and their remnants from the European Atherosclerosis Society (EAS), fasting TG level < 1.2 mmol/L is regarded as optimal, otherwise considered as non-optimal TG (NoTG). However, the postprandial cut-off value after a daily meal corresponding to a fasting TG level of 1.2 mmol/L has not been explored.
### Materials and methods
Six hundred and eighteen inpatients aged 18 to 70 were recruited in this study. Among them, 219 subjects had fasting TG levels < 1.2 mmol/L (i.e., OTG group), and 399 subjects had fasting TG levels ≥ 1.2 mmol/L (i.e., NoTG group). Serum levels of blood lipids, including calculated non-high-density lipoprotein cholesterol (non-HDL-C) and remnant cholesterol (RC), were monitored at 0, 2, and 4 h after a daily Chinese breakfast according to their dietary habits. Receiver operating characteristic (ROC) curve analysis was used to determine the postprandial cut-off value corresponding to the fasting TG level of 1.2 mmol/L. Kappa statistics were performed to determine the consistency between fasting and postprandial cut-off values in determining whether TG was optimal. Univariate and multivariate logistic regression analyses were conducted to evaluate the associations between NoTG and potential confounders. Subgroup analyses were performed to explore the association between postprandial TG levels at 4h (pTG4h) and NoTG in greater detail.
### Results
Postprandial levels of TG and RC significantly elevated and peaked at 4h after a daily breakfast in two groups ($P \leq 0.05$). The optimal cut-off value at 4h corresponding to fasting TG of 1.2 mmol/L was 1.56 mmol/L. According to the fasting cut-off value, the percentage of patients with NoTG was $64.6\%$ in the fasting state while increasing obviously to 73.3–$78.4\%$ at 2 and 4h, respectively, after a daily Chinese breakfast. According to the postprandial cut-off value, the percentage of patients with NoTG at 4h after a daily Chinese breakfast was $62.6\%$ which was close to $64.6\%$ in the fasting state. The Kappa coefficient was 0.551, indicating a moderate consistency between the fasting and postprandial cut-off values in the diagnosis of NoTG. Moreover, the subjects with NoTG determined by the postprandial TG cut-off value had an obviously higher postprandial level of RC (1.2 vs. 0.8 mmol/L) and percentage of HRC (37.1 vs. $32.1\%$) than those determined by the fasting TG cut-off value. Multivariate logistic regression analyses demonstrated that except for BMI, pTG4h emerged as an independent predictor of not. Subgroup analyses revealed that the association between pTG4h and NoTG was consistent across subgroups.
### Conclusion
Taken together, we for the first time determined TG 1.56 mmol/L as the postprandial cut-off value corresponding to fasting TG 1.2 mmol/L in Chinese subjects. This could make it more convenient to determine whether TG is optimal or not in the fasting or postprandial state.
## Introduction
The elevated level of low-density lipoprotein cholesterol (LDL-C) is regarded as an independent risk factor for atherosclerotic cardiovascular disease (ASCVD). Controlling low-density lipoprotein cholesterol (LDL-C) to the target level is the primary goal for patients with ASCVD to reduce clinical cardiovascular events [1]. Different from fasting LDL-C level, the relationship between fasting triglyceride (TG) level and ASCVD was controversial [2]. Since people are in a postprandial (i.e., non-fasting) state most of the day, more attention is focused on the relationship between postprandial blood lipids and ASCVD.
Strong evidence supports that elevated postprandial levels of triglyceride, as well as LDL-C, can independently predict the risk of ischemic heart disease, including that of myocardial infarction [3]. Based on these new findings, the testing of postprandial blood lipids has been recommended in routine clinical practice in Europe since 2016 [4]. The 2019 ESC/EAS guideline for the management of dyslipidemias classifies fasting TG (fTG) of < 1.7 mmol/L (150 mg/dL) as desirable, noting that fTG ≥ 1.7 mmol/L (150 mg/dL) is associated with the increased risk of ASCVD [5]. Moreover, controlling non-high-density-lipoprotein cholesterol (non-HDL-C) to the target level is the secondary goal to reduce the risk of ASCVD. Non-HDL-C is the total amount of cholesterol contained in lipoproteins other than high-density lipoprotein (HDL), which includes not only LDL-C but also cholesterol in other atherogenic lipoproteins, such as TG-rich lipoproteins (TRLs) and their hydrolyzed products, i.e., remnant lipoproteins [6]. The cholesterol in remnant lipoproteins (i.e., remnant cholesterol, RC) is an important part of non-HDL-C, especially in patients with hypertriglyceridemia. Hypertriglyceridemia represents the increased number of remnant lipoproteins in circulation. Compared with nascent TRLs, remnant lipoproteins contain more cholesterol ester, have smaller diameters, and thus are regarded as atherogenic as LDL.
Hypertriglyceridemia has been defined as fTG levels of 1.7 mmol/L (150 mg/dL) or higher [5, 7]. For example, the 2018 ACC/AHA guideline classifies moderate hypertriglyceridemia as 1.7–5.59 mmol/L (150–499 mg/dL) and severe hypertriglyceridemia as 5.6 mmol/L (500 mg/dL) or more [7]. The 2016 Chinese guideline classifies appropriate fTG level as < 1.7 mmol/L (150 mg/dL), borderline hypertriglyceridemia as 1.7–2.29 mmol/L (150–199 mg/dL) and hypertriglyceridemia as ≥ 2.3 mmol/L (200 mg/dL) [8]. The prevalence of hypertriglyceridemia is $16.9\%$ [8], supporting that hypertriglyceridemia is the most common form of dyslipidemia in the Chinese population. Recently, the definition of TG elevation has been updated again [9]. According to the 2021 consensus statement about TRLs and their remnants from the EAS, optimal TG level is defined as fasting TG < 1.2 mmol/L (100 mg/dL), borderline elevation as 1.2 mmol/L ≤ fTG < 1.7 mmol/L and elevation as fTG ≥ 1.7 mmol/L (150 mg/dL). To recommend postprandial blood lipids testing in routine clinical practice, the 2016 consensus statement from ESC proposes that the cut-off values for high TG and high RC (HRC) in the fasting state are 1.7 and 0.8 mmol/L, respectively, and those in the postprandial state are 2.0 and 0.9 mmol/L, respectively [4]. However, there was no idea about the postprandial cut-off value of TG after a daily meal corresponding to the fasting optimal TG cut-off value of 1.2 mmol/L.
This study aimed to determine the postprandial cut-off value after a daily meal corresponding to fTG level of 1.2 mmol/L in Chinese subjects, and to compare the roles of fasting and postprandial cut-off values in determining TG is optimal or not.
## Study subjects
Six hundred and eighteen inpatients of Chinese Han nationality aged 18 to 70 were enrolled in this study in the Department of Cardiovascular Medicine of the Second Xiangya Hospital from March 2017 to July 2020. Among them, 219 patients had optimal TG (OTG group: fTG < 1.2 mmol/L) and 399 patients had non-optimal TG (NoTG group: fTG ≥ 1.2 mmol/L). All of them were excluded from a history of digestive disease, autoimmune disease, hepatic and renal diseases, mental diseases, cancer or other severe medical diseases, or NYHA heart function class III-IV before getting involved. This study was approved by the Ethics Committee of the Second Xiangya Hospital of Central South University and informed consent was obtained from all participants.
## Specimen collection
After at least 8 h of overnight fasting, all subjects finished breakfast according to their own dietary habits within 15–20 min. The traditional Chinese breakfast that most Chinese people are used to usually includes the following categories: Meal 1 mainly included soybean milk, and fried dough sticks, which contain about 200–300 kcal. Meal 2 mainly included milk, bread, or eggs, which contain about 170–250 kcal. Meal 3 mainly included noodles or porridge, which contains about 220–270 kcal. Meal 4 mainly included steamed buns or rolls, which contain about 250–310 kcal. There is little difference in calories among different types of Chinese breakfast. All individuals were asked about their habitual dietary types and subjects who preferred one of the above types of breakfasts were included. It is not recommended that the subjects smoke, drink wine or beer, eat any food, or do strenuous exercise within 4h after breakfast, except for a little water and a slow walk. Venous blood samples were collected at fasting state, 2 and 4h after breakfast.
## Laboratory assays
The serum was separated from the venous blood samples. Serum levels of TG and total cholesterol (TC) were measured by automated enzymatic assays, and those of LDL-C and high-density lipoprotein cholesterol (HDL-C) were determined by the direct method, and were measured by laboratory technicians who are unaware of this study through a HITACHI 7170A analyzer (Instrument Hitachi Ltd., Tokyo, Japan) The detection kits were provided by Japan and Pure Pharmaceutical Industry Co., (Wako). Levels of RC and non-HDL-C were estimated by the following formulas, RC = TC–(HDL-C)–(LDL-C) and non-HDL-C = TC–(HDL-C) [10], respectively.
## Estimated sample size
The sample size was estimated based on the following calculation formula: n1 = n2 = 2 × [(tα + tβ)s/δ]2 (n1 and n2 are the required contents of the two samples respectively; tα and tβ are the t values corresponding to inspection level α and type II error probability β, respectively; s is the estimated value of the overall standard deviation; δ is the difference between the two means). Using a two-sample t-test, the test efficiency is $90\%$ at the test level of $5\%$ on both sides. The required estimated value of sample size is calculated based on the pre-experimental data of 2 and 4h after a meal, and the larger value is taken as the sample size required for this study. According to the difference between NoTG groups at 2 and 4h after a meal, the calculated sample size is 18 and 45, respectively. Taking the maximum value of each group, about 45 people are needed in each group, and a total of 90 people need to be included in this study. This study included 618 individuals, 219 in the OTG group and 399 in NoTG group. So, the sample size is appropriate for this present study, and sample size estimation was performed as further validation after the study design.
## Statistical analysis
Quantitative variables of normal distribution were expressed as mean ± standard deviation (SD), and Qualitative variables were expressed as numbers and percentages. One-way analysis of variance (ANOVA) for repeated measures was used to evaluate the difference between the mean values of variables within one group. ANOVA for completely randomized measures was used to evaluate the difference between the mean values of variables between the two groups. Qualitative variables were compared using the Chi-square test for R x C. The area under the curve (AUC) and AUC increment (iAUC) that represent the increase in the area after a daily meal relative to the fasting level of TG and RC were estimated by the trapezoid method. Receiver operating characteristic (ROC) curve analysis was used to determine the postprandial cut-off value corresponding to the fasting TG level of 1.2 mmol/L. Reliability analysis was performed with the Kappa statistic to determine the consistency between fasting and postprandial cut-off values in determining whether TG was optimal. Univariate and multivariate logistic regression analysis was conducted to examine associations between NoTG and several potential confounders including age, gender, BMI, current smoking, history of hypertension, DM and CHD. Subgroup analyses were performed after stratification by age (< 55 or ≥ 55 years), gender (male or female), BMI (< 24 or ≥ 24 kg/m2), current smoking (yes or no), history of hypertension (yes or no), DM (yes or no) and CHD (yes or no) to identify any modification caused by these variables. All statistical analyses were performed with SPSS version 26.0. All P values were 2-tailed, and $P \leq 0.05$ was considered statistically significant.
## Clinical characteristics and fasting blood lipids of two groups
There were 618 participants aged 18 to 70 recruited in this study population, and fasting NoTG (TG ≥ 1.2 mmol/L) was found in 399 ($64.6\%$) of the individuals. *The* general characteristics of the OTG group ($$n = 219$$) and the NoTG group ($$n = 399$$) were presented in Table 1. There was no significant difference in age, gender, systolic or diastolic blood pressure, heart rate, the percentages of smoking and recent use of lipid-lowering drugs between the two groups. Body mass index in the NoTG group was significantly higher than that in the OTG group ($P \leq 0.05$). The proportion of overweight and obese individuals and that of patients with CHD or diabetes mellitus in the NoTG group were significantly higher than those in the OTG group ($P \leq 0.05$).
**TABLE 1**
| Unnamed: 0 | OTG (n = 219) | NoTG (n = 399) | P-value |
| --- | --- | --- | --- |
| Age (year, SD) | 55.3 ± 12.6 | 53.9 ± 10.0 | NS |
| Men, n (%) | 140 (64.0) | 275 (68.9) | NS |
| BMI, kg/m2 | 23.4 ± 3.2 | 25.3 ± 3.6 | <0.0001 |
| OWand OB, n (%) | 85 (38.8) | 260 (65.2) | <0.0001 |
| Current smoking, n (%) | 79 (36.1) | 171 (43.1) | NS |
| Systolic pressure (mmHg) | 130.0 ± 19.1 | 132.0 ± 20.0 | NS |
| Diastolic pressure (mmHg) | 80.2 ± 13.5 | 82.0 ± 12.2 | NS |
| Heart rate (bpm) | 78.7 ± 17.1 | 78.5 ± 15.0 | NS |
| DM, n (%) | 31 (14.2) | 83 (20.9) | <0.05 |
| CHD, n (%) | 101 (46.1) | 217 (54.4) | <0.05 |
| Taking statins, n (%) | 78 (35.6) | 165 (41.8) | NS |
| Taking statins ≥ 3 months, n (%) | 37 (16.9) | 71 (18) | NS |
| Ezetimibe, n (%) | 1 (1.2) | 2 (1.4) | NS |
| TC, mmol/L | 3.78 ± 0.80 | 4.12 ± 0.77 | <0.0001 |
| LDL-C, mmol/L | 2.20 ± 0.68 | 2.62 ± 0.67 | <0.0001 |
| HDL-C, mmol/L | 1.17 ± 0.30 | 1.0 ± 0.21 | <0.0001 |
| Non-HDL-C, mmol/L | 2.60 ± 0.70 | 3.14 ± 0.70 | <0.0001 |
| TG, mmol/L | 0.90 ± 0.21 | 2.06 ± 0.78 | <0.0001 |
| RC, mmol/L | 0.37 ± 0.19 | 0.51 ± 0.18 | <0.0001 |
## Postprandial changes of blood lipids in two groups
In the fasting state, levels of TG, TC, LDL-C, non-HDL-C, and RC were significantly higher while fasting HDL-C level was markedly lower in the NoTG group ($P \leq 0.05$, Table 1).
After a daily breakfast, levels of TG and RC significantly increased, while those of TC, LDL-C and non-HDL-C significantly decreased in the two groups ($P \leq 0.05$). Both TG and RC levels peaked at 4h after a daily breakfast in both groups. Postprandial levels of TC, LDL-C, non-HDL-C, TG, and RC at 2 and 4h after a daily meal in the NoTG group were significantly higher than those in the OTG group ($P \leq 0.05$, Figures 1A–F).
**FIGURE 1:** *Postprandial changes in blood lipids after a daily meal in two groups. (A,B,D,E) The changes in serum concentrations of total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and triglyceride (TG). (C,F) The changes in serum concentrations of non-HDL-C and remnant cholesterol (RC) were determined by calculated methods. (G) Comparison of the total area under the curve (AUC) for TG and RC between two groups. (H) Comparison of increase in AUC (iAUC) for TG and RC between two groups. Optimal triglyceride (OTG) group: fasting TG < 1.2 mmol/L. Non-optimal triglyceride (NoTG) group: fasting TG ≥ 1.2 mmol/L. Values are mean ± standard error (SE). One-way analysis of variance (ANOVA) for repeated measures within a group or ANOVA for completely randomized measures between groups was used to assess any differences between the means of the variables, as appropriate. *P < 0.05 when compared with the patients in the OTG group. #P < 0.05 when compared with the fasting level in the same group.*
Both AUC and iAUC of TG or RC levels in the NoTG group were significantly higher than those in the OTG group ($P \leq 0.05$, Figures 1G, H).
## Evaluation of NoTG according to the new postprandial cut-off value
Correlation analysis showed that the correlation between fTG and postprandial TG level at 4h (pTG4h) was the strongest ($r = 0.668$, $P \leq 0.0001$, Figure 2A), followed by that between fTG and total postprandial TG (postprandial TG at 2 and 4h) ($r = 0.632$, $P \leq 0.0001$, Supplementary Figure 1A), and then by that between fTG and postprandial TG level at 2 h (pTG2h) ($r = 0.626$, $P \leq 0.0001$, Supplementary Figure 1C).
**FIGURE 2:** *Comparisons of the percentages of non-optimal triglyceride (NoTG) at different time-points. (A) Correlation between the level of fasting TG (fTG) and pTG4h. (B) Receiver operating characteristic (ROC) analysis and Youden’s index determined a cut-off value for postprandial TG level at 4h (pTG4h) after a daily meal; the cut-off value was indicated by a solid arrow. (C) Comparisons of the percentages of NoTG at different time-points according to the different cut-off values. (D) Distribution of the levels of fTG and pTG4h in all subjects. *P < 0.05 when compared with the fasting state. Optimal triglyceride (OTG) group: fasting TG < 1.2 mmol/L. Non-optimal triglyceride (NoTG) group: fasting TG ≥ 1.2 mmol/L. Values are represented as n (%) as appropriate.*
Receiver operating characteristic (ROC) curve analysis showed that the optimal cut-off value for TG at 4h corresponding to fasting TG 1.2 mmol/L was 1.56 mmol/L (sensitivity $83\%$, specificity $74\%$, and AUC 0.8438, Figure 2B).
According to the fasting cut-off value, the percentage of patients with NoTG was $64.60\%$ at the fasting state while increasing obviously to 73.3–$78.4\%$ at 2 and 4h, respectively, after a daily Chinese breakfast (Supplementary Figure 2). According to the postprandial cut-off value, the percentage of patients with NoTG at 4h after a daily Chinese breakfast was $62.6\%$ which was close to that in the fasting state (Figure 2C).
Subsequently, we evaluated the distribution of the fTG and pTG4h according to the fasting and postprandial cut-off values in all subjects. The number of individuals who were double-high (i.e., fTG ≥ 1.2 mmol/L and pTG 4h ≥ 1.56 mmol/L) reached 320 in 6l8, accounting for $51.8\%$. The number of individuals who were double-optimal (i.e., fTG < 1.2 mmol/L and pTG 4h < 1.56 mmol/L) was 167 in 618, accounting for $27.0\%$. Some patients with fasting TG levels ≥ 1.2 mmol/L showed optimal postprandial TG levels (74, $11.9\%$), while others with fasting TG levels < 1.2 mmol/L were found with non-optimal postprandial TG levels (57, $9.3\%$) (Figure 2D).
Kappa analysis showed that the Kappa coefficient was 0.551, indicating a moderate consistency between the fasting and postprandial cut-off values in the diagnosis of NoTG.
## Comparison of the distribution of NoTG and HRC according to different cut-off values
Correlation analysis between levels of TG and RC showed that the correlation between pTG4h and postprandial RC level at 4h (pRC4h) ($r = 0.714$, $P \leq 0.0001$) was stronger than that between fTG and fasting RC (fRC) ($r = 0.538$, $P \leq 0.0001$) (Figures 3A, B) and that between pTG2h and pRC2h ($r = 0.525$, $P \leq 0.0001$, Supplementary Figure 1B), then by that pTG2&4h and pRC2&4h ($r = 0.392$, $P \leq 0.0001$, Supplementary Figure 1D).
**FIGURE 3:** *Correlation analysis between the levels of triglyceride (TG) and remnant cholesterol (RC). (A) Correlation between fasting TG (fTG) and fasting RC (fRC) levels. (B) Correlation between postprandial TG level at 4h (pTG4h) and postprandial RC level at 4h (pRC4h). (C) Distribution of the fTG and fRC levels in all subjects. (D) Distribution of the pTG4h and pRC4h levels in all subjects.*
The distribution of TG and RC was observed in both fasting and postprandial states. Notably, the number of individuals who were double-high in the fasting state (i.e., fTG ≥ 1.2 mmol/L and fRC ≥ 0.8 mmol/L) was only 20 in 618, accounting for $3.2\%$ (Figure 3C), while that in the postprandial state (i.e., pTG4h ≥ 1.56 mmol/L and pRC4h ≥ 0.9 mmol/L) reached 139 in 618, accounting for $22.5\%$ (Figure 3D). The number and proportion of individuals who were double-optimal in the fasting state (i.e., fTG < 1.2 mmol/L and fRC < 0.8 mmol/L) and postprandial state (i.e., pTG4h < 1.56 mmol/L and pTG4h < 0.9 mmol/L) was similar, i.e., 222 in 6l8 ($36.9\%$) vs. 235 in 618 ($38.0\%$) (Figures 3C, D).
## Comparison of RC level and the percentage of HRC according to different cut-off values of TG
Compared with the OTG group, the NoTG group had significantly higher fasting and postprandial RC levels as well as the percentage of fasting HRC (fasting RC, i.e., fRC, ≥ 0.8 mmol/L) (Figures 4A, B). According to the postprandial cut-off value of optimal TG (i.e., 1.56 mmol/L), all subjects were divided into patients with postprandial OTG (pOTG: pTG4h < 1.56 mmol/L) and those with postprandial NoTG (pNoTG: pTG4h ≥ 1.56 mmol/L). Similarly, patients with pNoTG also had significantly higher postprandial RC levels as well as the percentage of postprandial HRC (postprandial RC, i.e., pRC ≥ 0.9 mmol/L) than those with pOTG (Figures 4C, D).
**FIGURE 4:** *Comparison of RC levels and percentages of HRC according to different cut-off values. (A) Comparison of levels of RC and percentages of high RC (HRC) between the OTG group (n = 219) and NoTG group (n = 399) classified by the fasting cut-off value of TG 1.2 mmol/L. (B) Comparison of percentages of high RC (HRC) between the OTG group (n = 219) and NoTG group (n = 399) classified by the fasting cut-off value of TG 1.2 mmol/L. (C) Comparison of levels of RC and percentage of HRC between the pOTG group (n = 231) and pNoTG group (n = 387) classified by the postprandial cut-off value of TG 1.56 mmol/L. (D) Comparison of percentages of HRC between the pOTG group (n = 231) and pNoTG group (n = 387) classified by the postprandial cut-off value of TG 1.56 mmol/L. Optimal triglyceride (OTG) group: fasting TG < 1.2 mmol/L. Non-optimal triglyceride (NoTG) group: fasting TG = 1.2 mmol/L. Postprandial optimal TG (pOTG, i.e. pTG4h < 1.56 mmol/L), and postprandial non-optimal TG (pNoTG, i.e. pTG4h = 1.56 mmol/L). *P < 0.05 when compared with the OTG group. #P < 0.05 when compared with the fasting state within the same group. P < 0.05 when compared with the pOTG group. Values are mean ± standard error (SE) or n (%) as appropriate.*
More importantly, patients with pNoTG showed higher postprandial RC levels (1.2 vs. 0.8 mmol/L) and the percentage of postprandial HRC (37.1 vs. $32.1\%$) than those with fasting NoTG (fTG ≥ 1.2 mmol/L), although their fasting RC levels and HRC percentages seemed similar (Figure 4B).
## Logistic regression and subgroup analyses
To investigate the association between all variables and NoTG, univariate and multivariate logistic regression analyses were conducted (Table 2). Univariate logistic regression analyses showed that age [odds ratio (OR): 1.008 ($95\%$ CI: 0.993–1.023); $$P \leq 0.325$$], being male [OR: 1.996 ($95\%$ CI: 1.402–2.766); $P \leq 0.001$], BMI [OR: 1.174 ($95\%$ CI: 1.113–1.239); $P \leq 0.001$], current smoking [OR: 1.341 ($95\%$ CI: 0.954–1.884); $$P \leq 0.091$$], history of hypertension [OR: 1.269 ($95\%$ CI: 0.906–1.778); $$P \leq 0.166$$], DM [OR: 1.593 ($95\%$ CI: 1.015–2.499); $$P \leq 0.043$$], CHD [OR: 1.393 ($95\%$ CI: 1.001–1.939); $$P \leq 0.049$$], medication [OR: 1.262 ($95\%$ CI: 0.900–1.771); $$P \leq 0.177$$], and pTG4h [OR: 5.133 ($95\%$ CI: 3.765–6.996); $P \leq 0.001$] were all significantly associated with NoTG. Subsequent multivariate regression analysis revealed that except for BMI [OR: 5.133 ($95\%$ CI: 3.765–6.996); $P \leq 0.001$], only pTG4h [OR: 4.490 ($95\%$ CI: 3.590–6.802); $P \leq 0.001$] was found to be an independent predictor of NoTG. Additionally, we performed subgroup analyses stratified by age, gender, BMI, smoke, hypertension, DM and medication to explore the association between pTG4h and NoTG in greater detail. It was found that the association of pTG4h and NoTG was consistent across subgroups (Figure 5).
## Discussion
In this study, significant postprandial hyperlipidemia mainly characterized by elevated TG and RC levels after a daily meal was observed in patients with NoTG, particularly at 4h. What’s more, There were higher AUC and iAUC of TG and RC in the NoTG group than in the OTG group, which indicated the more intensive and enduring postprandial reactive increases in TG and RC in those with fasting non-optimal TG levels. Through ROC curve analysis, the cut-off value of postprandial optimal TG level in Chinese subjects was first determined as 1.56 mmol/L, which corresponded to the fasting one recommended by the 2021 ESA consensus statement, i.e., 1.2 mmol/L [9]. And there was a moderate agreement between fasting and postprandial cut-off values in the diagnosis of optimal TG level with a Kappa coefficient of 0.551. Multivariate logistic regression analyses demonstrated that pTG4h was an independent predictor of NoTG, and subgroup analyses revealed the association between pTG4h and NoTG was consistent across subgroups. Hence, this could make it more convenient to determine whether TG is optimal or not, either in the fasting state or after a daily meal.
The postprandial state is a critical period in the progress of atherosclerosis [11, 12]. Two large population-based studies, the CORonary Diet Intervention with Olive Oil and Cardiovascular PREVention (CORDIOPREV) study [13] and the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) study [14] have proposed that an oral-fat tolerance test (OFTT) in clinical practice can be conducted to identify postprandial hyperlipidemia in subjects with fasting TG between 1 and 2 mmol/L (89–180 mg/dL) because approximately half of them have hidden postprandial hyperlipidemia, which may influence treatment [15]. And there were 318 subjects (about $51.50\%$) diagnosed with coronary heart disease. However, given the current popularization of healthy diet education, a high-fat diet is unacceptable, especially for patients with ASCVD. Hence, the subjects are more likely to accept the daily breakfast according to their own dietary habits, other than a high-fat diet. Additionally, there is no unified standard high-fat meal scheme in the clinic, even though OFTT has received a lot of attention and recommendations recently (13–15). More importantly, the 2016 European expert consensus recommended that blood lipids should be detected after daily meals rather than standard high-fat meals [4]. This is one of the main reasons why we paid more attention to postprandial levels of blood lipids after a daily meal.
Emerging studies have recommended that 4h after an OFTT was the most representative time-point to measure TG concentrations owing to the biggest difference between TG levels at 4h after a high-fat diet and that in a fasting state [16]. Similarly, we found that TG levels reached a peak at 4h in two groups after a daily breakfast. More importantly, the correlation between fasting and postprandial TG levels was the strongest at 4h after a daily breakfast. That’s why we chose 4h after a daily meal as the time-point to evaluate the cut-off value to determine the postprandial optimal TG cut-off value.
According to the new postprandial cut-off value determined by ROC curve analysis, the percentage of NoTG at 4h was close to that in the fasting state according to the cut-off value of TG 1.2 mmol/L recommended by the 2021 EAS consensus statement. The number of patients with fasting TG ≥ 1.2 mmol/L and postprandial TG ≥ 1.56 mmol/L as well as fasting TG < 1.2 mmol/L and postprandial TG < 1.56 mmol/L reached 489 in 618, accounting for about $80\%$ in this study. It suggested that there was at least moderate coincidence and consistency in the determination of optimal TG between the two cut-off values. Moreover, it indicates that in addition to patients who are in a fasting state, those who visit their doctors after meals can also obtain information on whether their TG levels are optimal or not.
It has been known that increased RC level can independently predict the residual risk of ASCVD [17]. Although TG level is closely associated with RC level, high RC is a more direct risk factor for atherosclerosis compared with high TG because remnant lipoproteins can enter the subendothelial area of the artery and promote the formation of foam cells [18]. It is recommended by the 2016 consensus statement from ESC that RC levels should not exceed 0.8 mmol/L in the fasting state and 0.9 mmol/L in the postprandial state [4], otherwise, it may indicate an increased risk of atherosclerosis. In this study, more subjects with postprandial HRC were found in those without optimal TG levels according to the postprandial cut-off value than those according to the fasting cut-off value. It supports that postprandial TG increase could expose the artery to more remnant lipoproteins in circulation. Hence, the postprandial cut-off value to determine whether TG is optimal or not could be conducive to improving the detection rate of HRC patients and promoting timely lifestyle intervention.
Notably, some patients with fasting TG levels ≥ 1.2 mmol/L had optimal postprandial TG levels (74, $11.9\%$), which may be due to individual differences in postprandial TG metabolism and/or dietary habits. However, other patients with fasting TG levels < 1.2 mmol/L were found with NoTG in the postprandial state (57, $9.0\%$), which indicated that a considerable number of subjects with NoTG will be omitted if postprandial blood lipids were not evaluated. As we all know, some patients with diabetes only show elevated postprandial blood glucose while their fasting blood glucose is normal. It can be speculated that monitoring postprandial TG and RC levels may be more important than their testing in the fasting state in some individuals, just like the importance of postprandial blood glucose detection for diabetes diagnosis and treatment.
Consistent with previous findings, BMI was considered as an independent predictor of elevated TG levels [19]. In this present study, the NoTG group was also found to have more individuals with higher BMI. These data demonstrated that elevated TG levels were significantly associated with overweight and/or obesity [20].
There were several limitations in this study. First, compared with similar clinical studies [21], the sample size of this study is relatively small. Second, the levels of blood lipids, especially TG, in those inpatients may be affected by underlying diseases and/or lipid-lowering drugs. Third, the exact amount of nutrition consumed in the breakfast by the patients was unknown. Compared with the undefined meals, the nutrient content of the standard meals is relatively constant and uniform, which makes it more convincing to compare the postprandial TG levels among individuals after standard meals. However, on the one hand, the scope of application of standard meals is still limited. For example, for patients with diabetes, the nutrition of standard meals may be excessive, and the type of nutrients may not be suitable. Therefore, standard meals may not be “standard,” and diverse kinds of nutritious meals should be formulated considering the specific energy needs of different disease populations. On the other hand, in the real world, undefined meals are more accessible, especially for outpatients. Additionally, emerging evidence showed that the dining places may also have some influence on postprandial lipid levels [22], and undefined meals are usually obtained at home, which seems to have little effect on postprandial lipid levels. Taken together, in an ideal state or among healthy people, the postprandial lipid levels after a standard meal are more convincing and suitable. While in the real world, due to their availability, the advantages of dining places and the concern for patients with diverse diseases, undefined meals seem to be more recommended. However, further studies on the sensitivity and specificity of the diagnosis of dyslipidemia through comparison between standardized and daily meals are still needed.
## Conclusion
Taken together, we for the first time determined TG 1.56 mmol/L as the postprandial cut-off value corresponding to fasting TG 1.2 mmol/L in Chinese subjects. This could make it more convenient to determine whether TG is optimal or not in the fasting or postprandial state.
## Data availability statement
The original contributions presented in this study are included in this article/Supplementary material, further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by the Ethics Committee of the Second Xiangya Hospital of Central South University and informed consent was obtained from all participants. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
LL and TW were the primary investigators and designers of this study. YX, LG, HC, JX, PQ, LZ, YT, MZ, and TW participated in the design of this study. All authors contributed to the article and approved the submitted version, accepted responsibility for the entire content of this manuscript, and approved its submission.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1037270/full#supplementary-material
## References
1. Clark J, Montori V. **In patients with ASCVD and elevated LDL-C with maximal statin therapy, inclisiran reduced LDL-C levels at 18 months.**. (2020) **173**
2. Zhang B, Yin F, Qiao Y, Guo S. **Triglyceride and triglyceride-rich lipoproteins in atherosclerosis.**. (2022) **9**. DOI: 10.3389/fmolb.2022.909151
3. Jørgensen AB, Frikke-Schmidt R, West AS, Grande P, Nordestgaard BG, Tybjærg-Hansen A. **Genetically elevated non-fasting triglycerides and calculated remnant cholesterol as causal risk factors for myocardial infarction.**. (2013) **34** 1826-33. DOI: 10.1093/eurheartj/ehs431
4. Nordestgaard BG, Langsted A, Mora S, Kolovou G, Baum H, Bruckert E. **Fasting is not routinely required for determination of a lipid profile: clinical and laboratory implications including flagging at desirable concentration cut-points-a joint consensus statement from the European atherosclerosis society and European federation of clinical chemistry and laboratory medicine.**. (2016) **37** 1944-58. PMID: 27122601
5. Mach F, Baigent C, Catapano AL, Koskinas KC, Casula M, Badimon L. **2019 ESC/EAS guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk.**. (2020) **41** 111-88. PMID: 31504418
6. Tushuizen ME, Pouwels PJ, Bontemps S, Rustemeijer C, Matikainen N, Heine RJ. **Postprandial lipid and apolipoprotein responses following three consecutive meals associate with liver fat content in type 2 diabetes and the metabolic syndrome.**. (2010) **211** 308-14. DOI: 10.1016/j.atherosclerosis.2010.02.002
7. Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS. **2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol: a report of the American college of cardiology/American heart association task force on clinical practice guidelines.**. (2019) **139** e1082-143. PMID: 30586774
8. **2016 Chinese guidelines for the management of dyslipidemia in adults.**. (2018) **15** 1-29. PMID: 29434622
9. Ginsberg HN, Packard CJ, Chapman MJ, Borén J, Aguilar-Salinas CA, Averna M. **Triglyceride-rich lipoproteins and their remnants: metabolic insights, role in atherosclerotic cardiovascular disease, and emerging therapeutic strategies-a consensus statement from the European atherosclerosis society.**. (2021) **42** 4791-806. DOI: 10.1093/eurheartj/ehab551
10. Kolovou GD, Anagnostopoulou KK, Pavlidis AN, Salpea KD, Iraklianou SA, Tsarpalis K. **Postprandial lipemia in men with metabolic syndrome, hypertensives and healthy subjects.**. (2005) **4**. DOI: 10.1186/1476-511X-4-21
11. Zilversmit DB. **Atherogenesis: a postprandial phenomenon.**. (1979) **60** 473-85. PMID: 222498
12. Kakuda H, Kobayashi J, Kakuda M, Yamakawa J, Takekoshi N. **The effect of anagliptin treatment on glucose metabolism and lipid metabolism, and oxidative stress in fasting and postprandial states using a test meal in Japanese men with type 2 diabetes.**. (2015) **48** 1005-9. DOI: 10.1007/s12020-014-0376-x
13. Delgado-Lista J, Perez-Martinez P, Garcia-Rios A, Alcala-Diaz JF, Perez-Caballero AI, Gomez-Delgado F. **CORonary Diet Intervention with Olive oil and cardiovascular PREVention study (the CORDIOPREV study): rationale, methods, and baseline characteristics: a clinical trial comparing the efficacy of a Mediterranean diet rich in olive oil versus a low-fat diet on cardiovascular disease in coronary patients.**. (2016) **177** 42-50. DOI: 10.1016/j.ahj.2016.04.011
14. Irvin MR, Zhi D, Aslibekyan S, Claas SA, Absher DM, Ordovas JM. **Genomics of post-prandial lipidomic phenotypes in the genetics of lipid lowering drugs and diet network (GOLDN) study.**. (2014) **9**. DOI: 10.1371/journal.pone.0099509
15. Perez-Martinez P, Alcala-Diaz JF, Kabagambe EK, Garcia-Rios A, Tsai MY, Delgado-Lista J. **Assessment of postprandial triglycerides in clinical practice: validation in a general population and coronary heart disease patients.**. (2016) **10** 1163-71. DOI: 10.1016/j.jacl.2016.05.009
16. Mihas C, Kolovou GD, Mikhailidis DP, Kovar J, Lairon D, Nordestgaard BG. **Diagnostic value of postprandial triglyceride testing in healthy subjects: a meta-analysis.**. (2011) **9** 271-80. DOI: 10.2174/157016111795495530
17. Gao S, Xu H, Ma W, Yuan J, Yu M. **Remnant cholesterol predicts risk of cardiovascular events in patients with myocardial infarction with nonobstructive coronary arteries.**. (2022) **11**. DOI: 10.1161/JAHA.121.024366
18. Whitman SC, Miller DB, Wolfe BM, Hegele RA, Huff MW. **Uptake of type III hypertriglyceridemic VLDL by macrophages is enhanced by oxidation, especially after remnant formation.**. (1997) **17** 1707-15. DOI: 10.1161/01.atv.17.9.1707
19. Hansen S, Madsen C, Varbo A, Nordestgaard B. **Body mass index, triglycerides, and risk of acute pancreatitis: a population-based study of 118 000 individuals.**. (2020) **105**. DOI: 10.1210/clinem/dgz059
20. Vekic J, Zeljkovic A, Stefanovic A, Jelic-Ivanovic Z, Spasojevic-Kalimanovska V. **Obesity and dyslipidemia.**. (2019) **92** 71-81. DOI: 10.1016/j.metabol.2018.11.005
21. Nordestgaard B, Langsted A, Mora S, Kolovou G, Baum H, Bruckert E. **Fasting is not routinely required for determination of a lipid profile: clinical and laboratory implications including flagging at desirable concentration cutpoints-a joint consensus statement from the European atherosclerosis society and European federation of clinical chemistry and laboratory medicine.**. (2016) **62** 930-46. PMID: 27235445
22. Schwedhelm C, Iqbal K, Schwingshackl L, Agogo G, Boeing H, Knüppel S. **Meal analysis for understanding eating behavior: meal- and participant-specific predictors for the variance in energy and macronutrient intake.**. (2019) **18**
|
---
title: Prognostic values of the prognostic nutritional index, geriatric nutritional
risk index, and systemic inflammatory indexes in patients with stage IIB–III cervical
cancer receiving radiotherapy
authors:
- Hong-Bing Wang
- Xin-Tian Xu
- Meng-Xing Tian
- Chen-Chen Ding
- Jing Tang
- Yu Qian
- Xin Jin
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC10017984
doi: 10.3389/fnut.2023.1000326
license: CC BY 4.0
---
# Prognostic values of the prognostic nutritional index, geriatric nutritional risk index, and systemic inflammatory indexes in patients with stage IIB–III cervical cancer receiving radiotherapy
## Abstract
### Background
Growing evidence suggests that nutritional status and inflammation are associated with survival in various cancers. This study aimed to evaluate the prognostic value of the prognostic nutritional index (PNI), geriatric nutritional risk index (GNRI), and systemic inflammatory indexes (neutrophil/lymphocyte ratio [NLR], monocyte/lymphocyte ratio [MLR], and platelet/lymphocyte ratio [PLR]) in patients with stage IIB–III cervical cancer receiving radiotherapy.
### Results
The ideal cutoff values for the PNI, GNRI, NLR, MLR, and PLR were 48.3, 97.04, 2.8, 0.41, and 186.67, respectively. Low PNI and GNRI scores were associated with poor OS and PFS. High NLR, MLR, and PLR also predicted inferior 5-year OS and PFS rates in patients with stage IIB–III cervical cancer. Multivariate Cox regression analysis identified tumor size, histological type, stage, number of metastatic lymph nodes, PNI, GNRI, NLR, PLR, and MLR as significant prognostic factors for OS and PFS.
### Conclusions
The current findings suggest that the PNI, GNRI, NLR, PLR, and MLR are essential parameters for predicting prognosis in patients with stage IIB–III cervical cancer receiving radiotherapy.
## 1. Introduction
Although largely preventable, cervical cancer is the fourth most common cancer in women in the USA and worldwide [1]. In 2020, approximately 604,000 new cases and 341,000 deaths were reported due to cervical cancer [2]. Unfortunately, more than two-thirds of women with cervical cancer are diagnosed at advanced stages in developing countries [3, 4]. In patients with locally advanced cervical cancer, survival is worse, and the recurrence rate is higher than that in those with early-stage cancer. The 5-year survival rate ranges from 31 to $55\%$ in patients with locally advanced cervical cancer undergoing optimal treatment such as chemoradiotherapy [5]. Staging, nodal involvement, and human papillomavirus infection affect local control and survival and have been used to predict treatment outcomes in patients with cervical cancer (6–8). However, the existing staging systems and other prognostic factors are not perfect to predict prognosis [9]. For example, although some patients have the same International Federation of Gynecology and Obstetrics (FIGO) stage, their prognosis is disparate because of their different pathological types [10, 11]. In addition, nutrition status is recognized as a critical determinant of quality of life in patients with cancer [12]. It is inherently inaccurate to predict the prognosis using only the existing system if the patient is malnourished. Accordingly, several novel prognostic parameters, a model with the existing system, and novel markers are required to predict life expectancy.
Nutritional status is recognized as a critical determinant of quality of life in patients with cancer [12]. Several studies have verified that malnutrition, sarcopenia, and cancer cachexia are associated with higher rates of post-treatment complications, lower rates of clinical response, longer hospital stays, and shorter survival times (13–17). In recent studies, several parameters, including nutritional and inflammatory indicators, have been shown to predict the prognosis of different tumors (18–20). PNI, an easily obtained index for evaluating nutritional status by calculating serum albumin levels and absolute lymphocyte counts, was first introduced to predict operative risk in gastrointestinal surgery [21]. Several retrospective studies have indicated that the prognostic nutritional index (PNI) is associated with clinical outcomes in many types of cancer [22, 23]. The geriatric nutritional risk index (GNRI) is calculated using serum albumin levels and ideal body weight. A low GNRI has also been verified as an independent prognostic factor affecting overall survival (OS) in patients with cancer [24].
Many studies have demonstrated the value of inflammatory cells in the blood and systemic inflammatory responses in the prognosis of patients with various types of tumors [25]. A series of systemic inflammatory indexes, such as the neutrophil/lymphocyte ratio (NLR), platelet/lymphocyte ratio (PLR), and monocyte/lymphocyte ratio (MLR), can be obtained in an easily available and inexpensive manner. The prognostic roles of NLR, PLR, and MLR have been verified in lung cancer, colorectal cancer, and hepatocellular carcinoma (26–28). For patients with operable cervical cancer, the prognostic value of NLR, PLR, and MLR has been investigated after surgery (29–32). Some studies have also reported the prognostic value of systemic inflammatory indexes in patients with non-surgical cervical cancer. One study reported that NLR and MLR predicted poor OS in patients with cervical cancer; however, only patients with stage IIB cancer were analyzed [33]. A retrospective study found that pretreatment NLR and PNI were significant predictors of prognosis in patients with cervical cancer treated with concurrent chemoradiotherapy [34]. However, many patients with stage I and IV disease were also included in the aforementioned study, and the prognosis of these patients was significantly different from that of patients with stage II–III disease. Moreover, survival curves and log-rank tests for different PNI/NLR/PLR values were not performed in Haraga et al. 's research. To date, there have been no reports on the impact of PNI, GNRI, NLR, PLR, and MLR on predicting survival time in patients with stage IIB–III disease undergoing radiotherapy (RT). Therefore, this study aimed to retrospectively analyze whether these factors are significantly associated with the prognosis of patients with stage IIB–III disease treated with RT.
## 2.1. Study population
Data from patients with cervical cancer who underwent RT were collected at the Hubei Cancer Hospital of Huazhong University of Science and Technology. A total of 178 patients were enrolled in this retrospective study from September 2013 to September 2015. Patients with incomplete medical records were excluded. As this was a retrospective study and the data were anonymous, the requirement for informed consent was waived. This study was approved by the Ethics Committee of Hubei Cancer Hospital of Huazhong University of Science and Technology (LLHBCH2021YN-049).
## 2.2. Data collection
The demographic characteristics, clinical characteristics, and laboratory results of the 178 patients were obtained from medical records. Data on age, body weight, tumor size, tumor stage, serum levels of squamous cell carcinoma (SCC) antigen, number of metastatic lymph nodes, serum albumin, and platelet, neutrophil, lymphocyte, and monocyte counts were collected. The International Federation of Gynecology and Obstetrics (FIGO) 2009 clinical staging system was used for tumor staging. Blood samples were collected before RT. Routine blood tests were performed using the Sysmex XN-9000 Hematology System (Sysmex Corporation, Shanghai, China). Biochemical tests were performed using an ADVIA 2400 Clinical Chemistry System (Siemens Healthineers, Erlangen, Germany). Serum SCC antigen tests were performed using a Cobas e 801 analytical unit (Roche Diagnostics International AG, Rotkreuz, Switzerland) and body weight was measured before treatment. The PNI and GNRI were calculated using the following formulas: PNI = serum albumin (g/L) + 5 × absolute lymphocyte count (109/L) and GNRI = [14.89 × serum albumin level (g/dL)] + [41.7 × actual body weight/ideal body weight]. NLR, PLR, and MLR were calculated as neutrophil/lymphocyte, platelet/lymphocyte, and monocyte/lymphocyte ratios, respectively.
## 2.3. RT procedures
Patients with cervical cancer (FIGO stages IIB–III) were treated with RT. If possible, after the initiation of RT, cisplatin at a dose of 40 mg/m2 on the body surface was also administered. A total of 105 patients underwent intensity-modulated RT (IMRT). The gross, clinical, and planned tumor volumes for patients who received IMRT were defined according to the Radiation Therapy Oncology Group guidelines. The prescribed dose was 45.0–50.4 Gy. IMRT was delivered at 1.8 Gy per fraction once daily for 5 days per week. A total of 73 patients underwent conventional RT (CRT). CRT was planned using the Eclipse Planning System and was conducted using a Varian 23EX. Conventional RT was delivered using anterior and posterior opposing techniques at a dose of 45.0–50.4 Gy (1.8 Gy per day, 5 days per week). All patients underwent a high dose of 192Ir brachytherapy after whole-pelvic irradiation at a dose of up to 36 Gy.
## 2.4. Follow-up strategy
Patients were followed up via outpatient examinations or telephone calls. The deadline for follow up was December 2019. OS was defined as the time from the start of RT to the date of death or last follow up. Progression-free survival (PFS) was defined as the initiation of RT, occurrence of tumor progression, death from any cause, or the last follow up.
## 2.5. Statistical analysis
Receiver operating characteristic (ROC) curves were used to determine the optimal PNI, GNRI, PLR, MLR, and NLR cutoff points using MedCalc (MedCalc Software Ltd., Belgium). R software version 4.1.3 (The R Foundation, Vienna, Austria) was used for statistical analysis. For the baseline characteristics of the patients, means and standard deviations are used to express continuous variables. Numbers and percentages are used to express categorical variables. Descriptive analysis using the chi-square test or Fisher's exact test was performed to compare differences between the two groups. Survival curves were calculated using the Kaplan–Meier method, and the log-rank test was used for comparison. Univariate and multivariate analyses were performed for each marker using the Cox proportional hazards model. Variables that were significant in the univariate analysis with P-values < 0.20 were included in multivariate analysis. We applied the nomogram in this study and visualized the prognostic strengths of different factors in predicting OS.
## 3.1. Patient characteristics
The patient characteristics are presented in Table 1. A total of 178 patients with cervical cancer were enrolled in this retrospective study. The mean age was 53.85. Thirty patients out of 178 ($16.9\%$) had more than two positive metastatic lymph nodes. Ninety-four patients ($52.8\%$) had stage II tumors and 84 ($47.2\%$) had stage III tumors, according to the FIGO 2009 clinical staging system. The mean body mass index (BMI) was 23.19 ± 2.88 kg/m2 with $3.4\%$ of patients being underweight. By setting survival status as an endpoint, ROC curves were used to determine the cutoff values. The cutoff values for the PNI, GNRI, NLR, MLR, and PLR were 48.3, 97.04, 2.8, 0.41, and 186.67, respectively (Figures 1, 2). The mean PNI, GNRI, NLR, MLR, and PLR were 49.37, 102.74, 2.77, 0.3, and 159.26, respectively. Low PNI and GNRI scores were observed in 78 ($43.8\%$) and 37 ($20.8\%$) patients, respectively. Low NLR, MLR, and PLR values were observed in 110 ($61.8\%$), 141 ($79.2\%$), and 136 ($76.4\%$) patients, respectively (Table 1).
## 3.2. Prognostic value of PNI and GNRI
In this retrospective cohort study, the 5-year OS rate of the entire population was $75.7\%$. The effect of nutritional status, as determined using the PNI and GNRI, on the prognosis of patients with cervical cancer was evaluated. Kaplan–*Meier analysis* showed that patients with a low PNI had shorter OS and PFS (low PNI vs. high PNI, 5-year OS, $64.1\%$ vs. $84.9\%$, $P \leq 0.001$; 5-year PFS, $62.8\%$ vs. $84.9\%$, $P \leq 0.001$) (Figures 3A, 4A). Similar results were obtained for the relationship between a low GNRI and the survival time of patients with cervical cancer (5-year OS, 48.5 vs. $82.2\%$, $P \leq 0.001$; 5-year PFS, 53.3 vs. $80.9\%$, $P \leq 0.001$) (Figures 3B, 4B). Survival analysis stratified by chemoradiotherapy (CRT) showed that patients with low PNI and GNRI values had shorter OS (low PNI vs. high PNI, $P \leq 0.01$; low GNRI vs. high GNRI, $P \leq 0.001$) and PFS (low PNI vs. high PNI, $P \leq 0.001$; low GNRI vs. high GNRI, $P \leq 0.001$) (Supplementary Figures 1, 2). Patients with a low GNRI had shorter OS ($P \leq 0.05$) and PFS ($P \leq 0.05$) than patients with a high GNRI in the survival analysis stratified by RT alone. There was no significant association between low PNI and OS/PFS in the survival analysis stratified by RT alone (Supplementary Figures 3, 4).
**Figure 3:** *Kaplan–Meier curves of overall survival according to the nutritional indicators. (A) Low prognostic nutritional index (PNI) vs. high PNI (low PNI: ≤ 48.3, high PNI: > 48.3) and (B) low geriatric nutritional risk index (GNRI) vs. high GNRI (low GNRI: ≤ 97.04, high GNRI: > 97.04). The Kaplan–Meier method was used to calculate the survival rate, and the log-rank test was used to compare survival distributions between the groups.* **Figure 4:** *Kaplan–Meier curves of progression-free survival according to the nutritional indicators. (A) Low prognostic nutritional index (PNI) vs. high PNI (low PNI: ≤ 48.3, high PNI: > 48.3) and (B) low geriatric nutritional risk index (GNRI) vs. high GNRI (low GNRI: ≤ 97.04, high GNRI: > 97.04). The Kaplan–Meier method was used to calculate the survival rate, and the log-rank test was used to compare survival distributions between the groups.*
## 3.3. Prognostic value of NLR, MLR, and PLR
The Kaplan–Meier results indicated that survival time differed depending on the NLR, MLR, and PLR. Patients with low NLR, MLR, and PLR had higher OS than patients in the other groups (5-year OS, low NLR vs. high NLR, 85.4 vs. $59.9\%$, $P \leq 0.001$; low MLR vs. high MLR, 82.9 vs. $49.9\%$, $P \leq 0.001$; low PLR vs. high PLR: 81.5 vs. $57.5\%$, $P \leq 0.001$) (Figure 5). We also analyzed the prognostic relationship between the systemic inflammatory indexes and PFS. Similar results were obtained (5-year PFS: low NLR vs. high NLR, 85.3 vs. $59.0\%$, $P \leq 0.001$; low MLR vs. high MLR, $82.8\%$ vs. $47.4\%$, $P \leq 0.001$; low PLR vs. high PLR, 81.5 vs. $5.6\%$, $P \leq 0.001$) (Figure 6). A significant association between low NLR/MLP/PLR and higher OS/FPS was also found in the survival analysis stratified by CRT (OS, $P \leq 0.001$; PFS $P \leq 0.001$) (Supplementary Figures 1, 2). In the survival analysis stratified by RT alone, there was no significant association between low NLR/MLP/PLR and OS/PFS (Supplementary Figures 3, 4).
**Figure 5:** *Kaplan–Meier curves of overall survival according to the inflammatory indicators. (A) Low neutrocyte/lymphocyte ratio (NLR) vs. high NLR (low NLR: ≤ 2.8, high NLR: > 2.8), (B) low monocyte/lymphocyte ratio (MLR) vs. high MLR (low MLR: ≤ 0.41, high MLR: > 0.41), and (C) low platelet lymphocyte ratio (PLR) vs. high PLR (low PLR: ≤ 186.67, high PLR: > 186.67). The Kaplan–Meier method was used to calculate the survival rate, and the log-rank test was used to compare survival distributions between the groups.* **Figure 6:** *Kaplan–Meier curves of progression-free survival according to the inflammatory indicators. (A) Low neutrocyte/lymphocyte ratio (NLR) vs. high NLR (low NLR: ≤ 2.8, high NLR: > 2.8), (B) low monocyte/lymphocyte ratio (MLR) vs. high MLR (low MLR: ≤ 0.41, high MLR: > 0.41), and (C) low platelet/lymphocyte ratio (PLR) vs. high PLR (low PLR: ≤ 186.67, high PLR: > 186.67). The Kaplan–Meier method was used to calculate the survival rate, and the log-rank test was used to compare survival distributions between the groups.*
## 3.4. Univariate and multivariate analyses for patients with cervical cancer
Univariate and multivariate analyses of the baseline characteristics of OS and PFS are shown in Tables 2, 3. In univariate analysis, tumor size, histological type, stage, number of metastatic lymph nodes, PNI, GNRI, NLR, PLR, and MLR were significantly associated with poor OS and PFS. Other factors, including age, type of RT, SCC antigen levels, and body weight, had no effect on cervical cancer prognosis. In univariate Cox regression analysis, the number of metastatic lymph nodes, tumor size, histological type, stage, GNRI, NLR, and MLR were the most significant predictors of OS and PFS, with hazards ratios (HR) higher than 3 or < 0.33.
In the multivariate Cox regression analysis, histological type remained the most significant predictor of OS (HR = 3.33; $95\%$ confidence interval [CI], 1.59–7.00; $$P \leq 0.001$$) The multivariate analysis identified that PNI (HR = 0.47; $95\%$ CI, 0.25–0.88; $P \leq 0.01$), GNRI (HR = 0.35; $95\%$ CI, 0.18–0.68; $$P \leq 0.002$$), NLR (HR = 2.60; $95\%$ CI, 1.36–4.97; $$P \leq 0.004$$), PLR (HR = 2.12; $95\%$ CI, 1.09–4.13; $$P \leq 0.028$$), and MLR (HR = 3.21; $95\%$ CI, 1.66–6.23, $P \leq 0.001$) were also significantly associated with OS. When the follow-up period was changed to PFS, PNI (HR = 0.47; $95\%$CI, 0.28–0.87; $$P \leq 0.017$$), GNRI (HR = 0.34; $95\%$CI, 0.17–0.65; $$P \leq 0.001$$), NLR (HR = 2.66; $95\%$CI, 1.42–4.97; $$P \leq 0.002$$), PLR (HR = 2.05; $95\%$CI, 1.10–3.80; $$P \leq 0.023$$), and MLR (HR = 3.36; $95\%$CI, 1.76–6.41; $P \leq 0.001$) were prognostic indicators for PFS, according to the multivariate analyses. In univariate and multivariate Cox regression analyses stratified by CRT, the GNRI, NLR, PLR, and MLR were also prognostic indicators for OS and PFS (Supplementary Tables 1, 2).
## 3.5. Prognostic nomograms of PNI, GNRI, and systemic inflammatory indexes
To predict the 3-year and 5-year OS of patients with cervical cancer, nomograms were constructed. Based on the results of the multivariate Cox analysis, the prognostic nomogram included tumor size, histological type, stage, number of metastatic lymph nodes, and PNI/GNRI/systemic inflammatory indexes (Figures 7, 8).
**Figure 7:** *Prognostic nomograms for overall survival prediction according to the prognostic nutritional index (PNI) (A) and geriatric nutritional risk index (GNRI) (B). Points were assigned for age before treatment, and for tumor size, histological type, stage, number of metastatic lymph nodes, and nutritional indicators. The score of each predictor was determined by drawing a vertical line from the value to the score scale. The total score was summed up by the scores of these predictors, which correspond to overall survival rate.* **Figure 8:** *Prognostic nomograms for overall survival prediction according to the neutrophil/lymphocyte ratio (NLR) (A), monocyte/lymphocyte ratio (MLR) (B), and platelet/lymphocyte ratio (PLR) (C). Points were assigned for age before treatment, and for tumor size, histological type, stage, number of metastatic lymph nodes, and inflammatory indicators. The score of each predictor was determined by drawing a vertical line from the value to the score scale. The total score was summed up by the scores of these predictors, which correspond to overall survival rate.*
## 4. Discussion
For patients with stage IIB–III cervical cancer, RT and a combination of chemotherapy and RT are the suggested treatment options. The present study demonstrated that a low PNI, low GNRI, high NLR, high MLR, and high PLR were negative prognostic factors for survival in patients with stage IIB–III disease treated with RT.
Similar to other types of cancers, there is a high prevalence of malnutrition among patients with cervical cancer [35]. The incidence of malnutrition was reported as high as $38.79\%$ in patients undergoing cervical cancer surgery before treatment [36]. Additionally, a higher stage grade indicates a higher incidence of malnutrition in cervical cancer [37]. Poor nutritional status at baseline is also associated with poor quality of life and chemotherapy interruption in patients with cervical cancer [38]. In clinical practice, the GNRI and PNI are easily obtained, objective, simple, efficient, and applicable tools to reflect nutritional status compared with other methods, such as patient-generated subjective global assessment and mini nutritional assessment. Our results also showed that poor status, as determined by the PNI and GNRI, was associated with shorter OS and PFS. Robust and consistent evidence has shown that cancer-related malnutrition plays a negative role in the prognosis of patients (39–42). Studies have shown that the prevalence of malnutrition in patients with cancer is as high as $80.4\%$ before treatment, and that nutritional status worsens with the progression of anticancer therapies [43, 44]. Due to clinically distinct causes, such as dysphagia, stomatitis, bowel obstruction caused by the tumor, and gastrointestinal disorders induced by anticancer therapies, the nutrient intake of patients with cancer is generally reduced [45]. In addition, altered metabolism-induced by excess catabolism, anabolic resistance, inflammation caused by tumors, and cancer therapy significantly affect nutritional status [46]. These factors lead to weight loss and skeletal muscle depletion in patients with cancer, which are independent risk factors for an unfavorable prognosis. Studies have demonstrated that unintentional weight loss is associated with poor post-operative survival and increased mortality risk in patients with cancer (47–49). The patients with locally advanced cervical cancer receiving primary chemoradiation who had unintentional weight loss ≥$10\%$ also had a higher risk of death (HR = 2.37) [50]. Decreased skeletal muscle mass, widely known as sarcopenia, has also been closely associated with a poor quality of life and short life expectancy [51]. Additionally, the common side effects of cytotoxic chemotherapy and RT directly affect the nutritional status of patients, and a poor nutritional status may aggravate these side effects [52]. Moreover, the decreased clearance of antitumor drugs in the tissues of patients with malnutrition with a higher drug concentration in the tissue may also lead to a higher rate of treatment toxicity [53]. The deterioration of nutritional status can lead to decreased treatment completion [54]. Furthermore, loss of body weight with a specific loss of skeletal muscle combined with systemic inflammation caused by tumors results in cancer cachexia [55]. Patients with cancer with cachexia have an impaired quality of life, high mortality, and increased treatment costs [46] and currently no effective medical intervention has been confirmed to completely reverse cachexia [56].
An increasing number of studies have shown that cancer-associated systemic inflammatory markers, such as NLR, PLR, and MLR, can be useful in predicting tumor progression. These markers are easily obtained, noninvasive, and inexpensive. Recently, three studies have demonstrated that systemic inflammatory markers are novel independent prognostic factors for predicting post-operative survival in patients with cervical cancer. High NLR, PLR, and MLR are closely related to poor prognosis (29–31). Similarly, our results showed that patients with stage IIB–III cervical cancer who underwent RT with high NLR, PLR, and MLR had shorter OS times. The close relationship between NLR/MLR and tumor prognosis involves tumor-induced inflammation and immune function changes. The systemic inflammatory response in patients with tumors is often accompanied by an increase in circulating neutrophil counts [57]. Recent studies have found that neutrophils not only exert an anti-tumor effect, but also promote tumor progression [58]. Most studies suggest that elevated neutrophil levels lead to tumor progression. The possible mechanisms by which neutrophils promote tumor progression include changes in the microenvironment shaped by cancer cells and release of some growth factors, such as epidermal growth factor and hepatocyte growth factor [59]. Monocytes also have diverse functions in different types and stages of the tumor [60]. The direct tumoricidal functions of monocytes result from cytokine-mediated induction of cell death and phagocytosis and effects on the components of the tumor microenvironment [61]. Interestingly, our study also suggests that low PLR is associated with cervical cancer prognosis. This result was inconsistent with Li's finding that PLR was not a significant independent prognostic factor in patients with stage IIB cervical cancer [33]. Another study also found that PLR was not associated with OS in gynecological cancer [62]. The inconsistent results may be due to the different stages of patients included in the different studies, which could affect the prognosis of cervical cancer. As an essential component of the blood, platelets play an important role in the inflammatory response in patients with cancer with chronic inflammation [63]. Angiogenesis is facilitated by the release of pro-angiogenic proteins, such as vascular epidermal growth factor and transforming growth factor, in the tumor microenvironment. The cytokines released by platelets can induce cancer-related inflammation and promote tumor growth and invasion [57].
Many studies have demonstrated that concurrent chemoradiotherapy provides therapeutic benefits over RT alone [64]. To explore the prognostic value of the PNI, GNRI, and systemic inflammatory indexes in patients who underwent CRT and RT alone, we performed survival analyses, univariate and multivariate analyses stratified by RT or CRT. The results showed that low GNRI, high NLR, high MLR, and high PLR predicted worse prognosis in patients treated with CRT. However, similar results were not observed in the patients who received RT alone. These inconsistent results may be explained by the small number of patients who underwent RT alone. Although there was an association between low PNI and poor OS/PFS in the multivariate cox analysis for all the patients, this association was not statistically significant in the multivariate analyses stratified by CRT. The possible reason is that patients who can only receive radiotherapy alone have poorer nutritional status than those who can receive concurrent chemotherapy.
Our study has several limitations. First, this was a retrospective study, and all data were collected from a single center. Second, the inflammatory state induced by infection before treatment may have an impact on the outcome. Third, we were not able to evaluate all covariates that might have affected prognosis, even though we included all likely covariates. Moreover, the sample size in this study was small. Additional prospective cohort studies are needed to determine the effects of GNRI, PNI, and systemic inflammatory indexes in patients with stage IIB–III cervical cancer.
## 5. Conclusions
Pretreatment GNRI, PNI, and systemic inflammatory indexes might be novel prognostic predictors for patients with stage II–III cervical cancer treated with RT. Low PNI, low GNRI, high NLR, high MLR, and high PLR predicted a worse prognosis. These markers can be incorporated into pretreatment evaluations and act as factors for decision-making in patients with cervical cancer receiving radiotherapy.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by Ethics Committee of Hubei Cancer Hospital of Huazhong University of Science and Technology. The Ethics Committee waived the requirement of written informed consent for participation.
## Author contributions
XJ: conceptualization, methodology, software, investigation, and writing-original draft. H-BW: data collection and writing-review and editing. X-TX: methodology, software, and investigation. M-XT: resources, data curation, and investigation. C-CD, JT, and YQ: writing-review and editing. All authors revised and approved the final manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1000326/full#supplementary-material
## References
1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. **Global cancer statistics 2018: globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA Cancer J Clin.* (2018) **68** 394-424. DOI: 10.3322/caac.21492
2. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A. **Global cancer statistics 2020: globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA Cancer J Clin.* (2021) **71** 209-49. DOI: 10.3322/caac.21660
3. Vizcaino AP, Moreno V, Bosch FX, Munoz N, Barros-Dios XM, Borras J. **International trends in incidence of cervical cancer: Ii**. *Squamous-Cell Carcinoma Int J Cancer.* (2000) **86** 429-35. DOI: 10.1002/(SICI)1097-0215(20000501)86:3<429::AID-IJC20>3.0.CO;2-D
4. Shrivastava S, Mahantshetty U, Engineer R, Tongaonkar H, Kulkarni J, Dinshaw K. **Treatment and outcome in cancer cervix patients treated between 1979 and 1994: a single institutional experience**. *J Cancer Res Ther.* (2013) **9** 672-9. DOI: 10.4103/0973-1482.126480
5. Manders DB, Moron A, McIntire D, Miller DS, Richardson DL, Kehoe SM. **Locally advanced cervical cancer: outcomes with variable adherence to treatment**. *Am J Clin Oncol.* (2018) **41** 447-51. DOI: 10.1097/COC.0000000000000300
6. Ditto A, Martinelli F, Lo Vullo S, Reato C, Solima E, Carcangiu M. **The role of lymphadenectomy in cervical cancer patients: the significance of the number and the status of lymph nodes removed in 526 cases treated in a single institution**. *Ann Surg Oncol.* (2013) **20** 3948-54. DOI: 10.1245/s10434-013-3067-6
7. Huang Y, Zou D, Guo M, He M, He H, Li X. **Hpv and radiosensitivity of cervical cancer: a narrative review**. *Ann Transl Med.* (2022) **10** 1405. DOI: 10.21037/atm-22-5930
8. Bhatla N, Aoki D, Sharma DN, Sankaranarayanan R. **Cancer of the Cervix Uteri**. *Int J Gynaecol Obstet.* (2018) **2** 22-36. DOI: 10.1002/ijgo.12611
9. Cho O, Chun M. **Management for locally advanced cervical cancer: new trends and controversial issues**. *Radiat Oncol J.* (2018) **36** 254-64. DOI: 10.3857/roj.2018.00500
10. Williams NL, Werner TL, Jarboe EA, Gaffney DK. **Adenocarcinoma of the cervix: should we treat it differently?**. *Curr Oncol Rep.* (2015) **17** 17. DOI: 10.1007/s11912-015-0440-6
11. Hu K, Wang W, Liu X, Meng Q, Zhang F. **Comparison of treatment outcomes between squamous cell carcinoma and adenocarcinoma of cervix after definitive radiotherapy or concurrent chemoradiotherapy**. *Radiat Oncol.* (2018) **13** 249. DOI: 10.1186/s13014-018-1197-5
12. Cong M, Zhu W, Wang C, Fu Z, Song C, Dai Z. **Nutritional status and survival of 8247 cancer patients with or without diabetes mellitus-results from a prospective cohort study**. *Cancer Med.* (2020) **9** 7428-39. DOI: 10.1002/cam4.3397
13. Fujiya K, Kawamura T, Omae K, Makuuchi R, Irino T, Tokunaga M. **Impact of malnutrition after gastrectomy for gastric cancer on long-term survival**. *Ann Surg Oncol.* (2018) **25** 974-83. DOI: 10.1245/s10434-018-6342-8
14. Santos I, Mendes L, Mansinho H, Santos CA. **Nutritional status and functional status of the pancreatic cancer patients and the impact of adjacent symptoms**. *Clin Nutr.* (2021) **40** 5486-93. DOI: 10.1016/j.clnu.2021.09.019
15. Kubrak C, Martin L, Gramlich L, Scrimger R, Jha N, Debenham B. **Prevalence and prognostic significance of malnutrition in patients with cancers of the head and neck**. *Clin Nutr.* (2020) **39** 901-9. DOI: 10.1016/j.clnu.2019.03.030
16. Zhang X, Tang M, Zhang Q, Zhang KP, Guo ZQ, Xu HX. **The glim criteria as an effective tool for nutrition assessment and survival prediction in older adult cancer patients**. *Clin Nutr.* (2021) **40** 1224-32. DOI: 10.1016/j.clnu.2020.08.004
17. Xu XT, He DL, Tian MX, Wu HJ, Jin X. **Prognostic value of sarcopenia in patients with diffuse large B-Cell lymphoma treated with R-chop: a systematic review and meta-analysis**. *Front Nutr.* (2022) **9** 816883. DOI: 10.3389/fnut.2022.816883
18. Kim SI, Kim SJ, Kim SJ, Cho DS. **Prognostic nutritional index and prognosis in renal cell carcinoma: a systematic review and meta-analysis**. *Urol Oncol.* (2021) **39** 623-30. DOI: 10.1016/j.urolonc.2021.05.028
19. Bullock AF, Greenley SL, McKenzie GAG, Paton LW, Johnson MJ. **Relationship between markers of malnutrition and clinical outcomes in older adults with cancer: systematic review, narrative synthesis and meta-analysis**. *Eur J Clin Nutr.* (2020) **74** 1519-35. DOI: 10.1038/s41430-020-0629-0
20. Wang D, Hu X, Xiao L, Long G, Yao L, Wang Z. **Prognostic nutritional index and systemic immune-inflammation index predict the prognosis of patients with Hcc**. *J Gastrointest Surg.* (2021) **25** 421-7. DOI: 10.1007/s11605-019-04492-7
21. Buzby GP, Mullen JL, Matthews DC, Hobbs CL, Rosato EF. **Prognostic nutritional index in gastrointestinal surgery**. *Am J Surg.* (1980) **139** 160-7. DOI: 10.1016/0002-9610(80)90246-9
22. Yan L, Nakamura T, Casadei-Gardini A, Bruixola G, Huang YL, Hu ZD. **Long-term and short-term prognostic value of the prognostic nutritional index in cancer: a narrative review**. *Ann Transl Med.* (2021) **9** 1630. DOI: 10.21037/atm-21-4528
23. Park S, Ahn HJ, Yang M, Kim JA, Kim JK, Park SJ. **The prognostic nutritional index and post-operative complications after curative lung cancer resection: a retrospective cohort study**. *J Thorac Cardiovasc Surg.* (2020). DOI: 10.1016/j.jtcvs.2019.10.105
24. Karayama M, Inoue Y, Yasui H, Hozumi H, Suzuki Y, Furuhashi K. **Association of the geriatric nutritional risk index with the survival of patients with non-small-cell lung cancer after platinum-based chemotherapy**. *BMC Pulm Med.* (2021) **21** 409. DOI: 10.1186/s12890-021-01782-2
25. Kumarasamy C, Sabarimurugan S, Madurantakam RM, Lakhotiya K, Samiappan S, Baxi S. **Prognostic significance of blood inflammatory biomarkers Nlr, Plr, and Lmr in cancer-a protocol for systematic review and meta-analysis**. *Medicine.* (2019) **98** e14834. DOI: 10.1097/MD.0000000000014834
26. Diem S, Schmid S, Krapf M, Flatz L, Born D, Jochum W. **Neutrophil-to-lymphocyte ratio (Nlr) and platelet-to-lymphocyte ratio (Plr) as prognostic markers in patients with non-small cell lung cancer (Nsclc) treated with nivolumab**. *Lung Cancer.* (2017) **111** 176-81. DOI: 10.1016/j.lungcan.2017.07.024
27. Kang Y, Zhu X, Lin Z, Zeng M, Shi P, Cao Y. **Compare the diagnostic and prognostic value of mlr, Nlr and Plr in Crc patients**. *Clin Lab.* (2021) **67** 9. DOI: 10.7754/Clin.Lab.2021.201130
28. Cho SH, Hwang JE, Bae WK, Chung IJ. **The prognostic role of Pd-L1 expression according to Msi status in stage Iii colon cancer after curative resection**. *J Clin Oncol.* (2019) **37** 555. DOI: 10.1200/JCO.2019.37.4_suppl.555
29. Huang H, Liu Q, Zhu L, Zhang Y, Lu X, Wu Y. **Prognostic value of preoperative systemic immune-inflammation index in patients with cervical cancer**. *Sci Rep.* (2019) **9** 3284. DOI: 10.1038/s41598-019-39150-0
30. Chao B, Ju X, Zhang L, Xu X, Zhao Y. **Novel prognostic marker systemic inflammation response index (Siri) for operable cervical cancer patients**. *Front Oncol.* (2020) **10** 766. DOI: 10.3389/fonc.2020.00766
31. Deng Q, Long Q, Liu Y, Yang Z, Du Y, Chen X. **Prognostic value of preoperative peripheral blood mean platelet volume/platelet count ratio (Mpv/Pc) in patients with resectable cervical cancer**. *BMC Cancer.* (2021) **21** 1282. DOI: 10.1186/s12885-021-09016-8
32. Jiang Y, Gu H, Zheng X, Pan B, Liu P, Zheng M. **Pretreatment C-reactive protein/albumin ratio is associated with poor survival in patients with 2018 Figo Stage Ib-Iia Hpv-positive cervical cancer**. *Pathol Oncol Res.* (2021) **27** 1609946. DOI: 10.3389/pore.2021.1609946
33. Li YX, Chang JY, He MY, Wang HR, Luo DQ Li FH. **Neutrophil-to-lymphocyte ratio (Nlr) and monocyte-to-lymphocyte ratio (Mlr) predict clinical outcome in patients with stage Iib cervical cancer**. *J Oncol.* (2021) **2021** 2939162. DOI: 10.1155/2021/2939162
34. Haraga J, Nakamura K, Omichi C, Nishida T, Haruma T, Kusumoto T. **Pretreatment prognostic nutritional index is a significant predictor of prognosis in patients with cervical cancer treated with concurrent chemoradiotherapy**. *Mol Clin Oncol.* (2016) **5** 567-74. DOI: 10.3892/mco.2016.1028
35. Bossi P, Delrio P, Mascheroni A, Zanetti M. **The spectrum of malnutrition/cachexia/sarcopenia in oncology according to different cancer types and settings: a narrative review**. *Nutrients.* (2021) **13** 1980. DOI: 10.3390/nu13061980
36. Tian M, Fu H, Du J. **Application value of Nrs2002 and Pg-Sga in nutritional assessment for patients with cervical cancer surgery**. *Am J Transl Res.* (2021) **13** 7186-92. PMID: 34306480
37. Flores-Cisneros L, Cetina-Perez L, Castillo-Martinez L, Jimenez-Lima R, Luvian-Morales J, Fernandez-Loaiza M. **Body composition and nutritional status according to clinical stage in patients with locally advanced cervical cancer**. *Eur J Clin Nutr.* (2021) **75** 852-5. DOI: 10.1038/s41430-020-00797-y
38. Aredes MA, Garcez MR, Chaves GV. **Influence of chemoradiotherapy on nutritional status, functional capacity, quality of life and toxicity of treatment for patients with cervical cancer**. *Nutr Diet.* (2018) **75** 263-70. DOI: 10.1111/1747-0080.12414
39. Laviano A, Di Lazzaro L, Koverech A. **Nutrition support and clinical outcome in advanced cancer patients**. *Proc Nutr Soc.* (2018) **77** 388-93. DOI: 10.1017/S0029665118000459
40. Arends J, Bachmann P, Baracos V, Barthelemy N, Bertz H, Bozzetti F. **Espen guidelines on nutrition in cancer patients**. *Clin Nutr.* (2017) **36** 11-48. DOI: 10.1016/j.clnu.2016.07.015
41. Aaldriks AA, van der Geest LG, Giltay EJ. **le Cessie S, Portielje JE, Tanis BC, et al. Frailty and malnutrition predictive of mortality risk in older patients with advanced colorectal cancer receiving chemotherapy**. *J Geriatr Oncol.* (2013) **4** 218-26. DOI: 10.1016/j.jgo.2013.04.001
42. Fukuda Y, Yamamoto K, Hirao M, Nishikawa K, Maeda S, Haraguchi N. **Prevalence of malnutrition among gastric cancer patients undergoing gastrectomy and optimal preoperative nutritional support for preventing surgical site infections**. *Ann Surg Oncol.* (2015) **22** S778-85. DOI: 10.1245/s10434-015-4820-9
43. Baracos VE. **Cancer-associated malnutrition**. *Eur J Clin Nutr.* (2018) **72** 1255-9. DOI: 10.1038/s41430-018-0245-4
44. Guo ZQ Yu JM, Li W, Fu ZM, Lin Y, Shi YY. **Survey and analysis of the nutritional status in hospitalized patients with malignant gastric tumors and its influence on the quality of life**. *Support Care Cancer.* (2020) **28** 373-80. DOI: 10.1007/s00520-019-04803-3
45. Martin L, Kubrak C. **How much does reduced food intake contribute to cancer-associated weight loss?**. *Curr Opin Support Palliat Care.* (2018) **12** 410-9. DOI: 10.1097/SPC.0000000000000379
46. Baracos VE, Martin L, Korc M, Guttridge DC, Fearon KCH. **Cancer-associated cachexia**. *Nat Rev Dis Primers.* (2018) **4** 17105. DOI: 10.1038/nrdp.2017.105
47. Hue JJ, Sugumar K, Kyasaram RK, Shanahan J, Lyons J, Ocuin LM. **Weight loss as an untapped early detection marker in pancreatic and periampullary cancer**. *Ann Surg Oncol.* (2021) **28** 6283-92. DOI: 10.1245/s10434-021-09861-8
48. Paixao EMS, Gonzalez MC, Nakano EY, Ito MK, Pizato N. **Weight loss, phase angle, and survival in cancer patients undergoing radiotherapy: a prospective study with 10-year follow-up**. *Eur J Clin Nutr.* (2021) **75** 823-8. DOI: 10.1038/s41430-020-00799-w
49. Hendifar AE, Petzel MQB, Zimmers TA, Denlinger CS, Matrisian LM, Picozzi VJ. **Pancreas cancer-associated weight loss**. *Oncologist.* (2019) **24** 691-701. DOI: 10.1634/theoncologist.2018-0266
50. Jou J, Coulter E, Roberts T, Binder P, Saenz C, McHale M. **Assessment of malnutrition by unintentional weight loss and its implications on oncologic outcomes in patient with locally advanced cervical cancer receiving primary chemoradiation**. *Gynecol Oncol.* (2021) **160** 721-8. DOI: 10.1016/j.ygyno.2020.12.009
51. Muscaritoli M, Anker SD, Argiles J, Aversa Z, Bauer JM, Biolo G. **Consensus definition of sarcopenia, cachexia and pre-cachexia: joint document elaborated by special interest groups (Sig) “cachexia-anorexia in chronic wasting diseases” and “nutrition in geriatrics”**. *Clin Nutr.* (2010) **29** 154-9. DOI: 10.1016/j.clnu.2009.12.004
52. Laura FC, Lucely CP, Tatiana GC, Roberto JL, Dulce GI, Arturo PS. **Handgrip strength, overhydration and nutritional status as a predictors of gastrointestinal toxicity in cervical cancer patients. A prospective study**. *Nutr Cancer.* (2022) **74** 2444-50. DOI: 10.1080/01635581.2021.2012209
53. Prado CM, Lima IS, Baracos VE, Bies RR, McCargar LJ, Reiman T. **An exploratory study of body composition as a determinant of epirubicin pharmacokinetics and toxicity**. *Cancer Chemother Pharmacol.* (2011) **67** 93-101. DOI: 10.1007/s00280-010-1288-y
54. Hamaker ME, Oosterlaan F, van Huis LH, Thielen N, Vondeling A, van den Bos F. **Nutritional status and interventions for patients with cancer—a systematic review**. *J Geriatr Oncol.* (2021) **12** 6-21. DOI: 10.1016/j.jgo.2020.06.020
55. Fearon K, Strasser F, Anker SD, Bosaeus I, Bruera E, Fainsinger RL. **Definition and classification of cancer cachexia: an international consensus**. *Lancet Oncol.* (2011) **12** 489-95. DOI: 10.1016/S1470-2045(10)70218-7
56. Jin X, Xu XT, Tian MX Dai Z. **Omega-3 polyunsaterated fatty acids improve quality of life and survival, but not body weight in cancer cachexia: a systematic review and meta-analysis of controlled trials**. *Nutr Res.* (2022) **107** 165-78. DOI: 10.1016/j.nutres.2022.09.009
57. Yamamoto T, Kawada K, Obama K. **Inflammation-related biomarkers for the prediction of prognosis in colorectal cancer patients**. *Int J Mol Sci.* (2021) **22** 8002. DOI: 10.3390/ijms22158002
58. Kargl J, Busch SE, Yang GH, Kim KH, Hanke ML, Metz HE. **Neutrophils dominate the immune cell composition in non-small cell lung cancer**. *Nat Commun.* (2017) **8** 14381. DOI: 10.1038/ncomms14381
59. Xiong S, Dong L, Cheng L. **Neutrophils in cancer carcinogenesis and metastasis**. *J Hematol Oncol.* (2021) **14** 173. DOI: 10.1186/s13045-021-01187-y
60. Olingy CE, Dinh HQ, Hedrick CC. **Monocyte heterogeneity and functions in cancer**. *J Leukoc Biol.* (2019) **106** 309-22. DOI: 10.1002/JLB.4RI0818-311R
61. Robinson A, Han CZ, Glass CK, Pollard JW. **Monocyte regulation in homeostasis and malignancy**. *Trends Immunol.* (2021) **42** 104-19. DOI: 10.1016/j.it.2020.12.001
62. Abu-Shawer O, Abu-Shawer M, Hirmas N, Alhouri A, Massad A, Alsibai B. **Hematologic markers of distant metastases and poor prognosis in gynecological cancers**. *BMC Cancer.* (2019) **19** 141. DOI: 10.1186/s12885-019-5326-9
63. Wagner DD. **New Links between inflammation and thrombosis**. *Arterioscler Thromb Vasc Biol.* (2005) **25** 1321-4. DOI: 10.1161/01.ATV.0000166521.90532.44
64. Datta NR, Stutz E, Liu M, Rogers S, Klingbiel D, Siebenhuner A. **Concurrent chemoradiotherapy vs. radiotherapy alone in locally advanced cervix cancer: a systematic review and meta-analysis**. *Gynecol Oncol.* (2017) **145** 374-85. DOI: 10.1016/j.ygyno.2017.01.033
|
---
title: The role of ADAM17 in cerebrovascular and cognitive function in the APP/PS1
mouse model of Alzheimer’s disease
authors:
- Yanna Tian
- Katie Anne Fopiano
- Vadym Buncha
- Liwei Lang
- Hayden A. Suggs
- Rongrong Wang
- R. Daniel Rudic
- Jessica A. Filosa
- Zsolt Bagi
journal: Frontiers in Molecular Neuroscience
year: 2023
pmcid: PMC10018024
doi: 10.3389/fnmol.2023.1125932
license: CC BY 4.0
---
# The role of ADAM17 in cerebrovascular and cognitive function in the APP/PS1 mouse model of Alzheimer’s disease
## Abstract
### Introduction
The disintegrin and metalloproteinase 17 (ADAM17) exhibits α-secretase activity, whereby it can prevent the production of neurotoxic amyloid precursor protein-α (APP). ADAM17 is abundantly expressed in vascular endothelial cells and may act to regulate vascular homeostatic responses, including vasomotor function, vascular wall morphology, and formation of new blood vessels. The role of vascular ADAM17 in neurodegenerative diseases remains poorly understood. Here, we hypothesized that cerebrovascular ADAM17 plays a role in the pathogenesis of Alzheimer’s disease (AD).
### Methods and results
We found that 9-10 months old APP/PS1 mice with b-amyloid accumulation and short-term memory and cognitive deficits display a markedly reduced expression of ADAM17 in cerebral microvessels. Systemic delivery and adeno-associated virus (AAV)-mediated re-expression of ADAM17 in APP/PS1 mice improved cognitive functioning, without affecting b-amyloid plaque density. In isolated and pressurized cerebral arteries of APP/PS1 mice the endothelium-dependent dilation to acetylcholine was significantly reduced, whereas the vascular smooth muscle-dependent dilation to the nitric oxide donor, sodium nitroprusside was maintained when compared to WT mice. The impaired endothelium-dependent vasodilation of cerebral arteries in APP/PS1 mice was restored to normal level by ADAM17 re-expression. The cerebral artery biomechanical properties (wall stress and elasticity) and microvascular network density was not affected by ADAM17 re-expression in the APP/PS1 mice. Additionally, proteomic analysis identified several differentially expressed molecules involved in AD neurodegeneration and neuronal repair mechanisms that were reversed by ADAM17 re-expression.
### Discussion
Thus, we propose that a reduced ADAM17 expression in cerebral microvessels impairs vasodilator function, which may contribute to the development of cognitive dysfunction in APP/PS1 mice, and that ADAM17 can potentially be targeted for therapeutic intervention in AD.
## Introduction
Cardiovascular disease is a common comorbidity in Alzheimer’s disease (AD; Gorelick et al., 2011; Iadecola, 2013; Corriveau et al., 2016). Notably, $50\%$ of clinically diagnosed AD patients display a mixed vascular and AD pathology (Snowdon et al., 1997; Santisteban and Iadecola, 2018). Cardio-and cerebrovascular diseases are multifactorial conditions, increasingly prevalent in older adults with hypertension and type 2 diabetes that are risk factors for dementia (Udelson, 2011; Goyal et al., 2016; Silverman et al., 2016). It is now believed that cerebrovascular changes not only accompany but are also mechanistically linked to the development of AD. In support, human autopsy findings demonstrate that a significant portion of clinically diagnosed AD patients have histopathology-defined microvascular brain injury (Bagi et al., 2018; Park et al., 2018). Consistent with the contribution of cerebral microvascular dysfunction to human AD, we recently described selective vasodilator dysfunction of cerebral parenchymal arterioles, which appears to be associated with AD neuropathological changes and MRI-defined cerebral microstructural changes in brain donors with no pathologically described macroscopic infarcts or hemorrhages (Bagi et al., 2022). Presently unresolved is the extent to which cerebrovascular dysfunction is associated with memory deficits and cognitive decline in AD patients. Furthermore, the molecular underpinnings through which cerebrovascular dysfunction contributes to the development of the AD pathomechanism is incompletely understood, and therefore preventative therapeutic measures cannot be adapted.
A disintegrin and metalloproteinase 17, ADAM17 (also known as tumor necrosis factor (TNF)-α converting enzyme or TACE), was initially described as a sheddase for cell membrane-bound TNF (Gooz, 2010). ADAM17 was later identified as a highly promiscuous enzyme regulating multiple substrates by proteolytic cleavage, such as TNF-α receptors, pro-inflammatory and inflammation resolution mediators, as well as vascular cell membrane-bound adhesion molecules, growth and angiogenic factors (Gooz et al., 2009). Interestingly, a prior report has shown that a rare genetic variant, leading to a loss-of-function in ADAM17, is associated with the pathogenesis of AD in humans (Hartl et al., 2020). It is known that pathological amyloidogenic amyloid beta (Aβ) accumulation, due to abnormal processing of the amyloid precursor protein (APP), promotes AD development via neurotoxic effects (Allinson et al., 2003; Sastre et al., 2008), and also, in part, by causing cerebrovascular impairments (Sastre et al., 2008; Iadecola, 2017). ADAM17, owing to its α-secretase activity, has been implicated in the shedding of the APP, promoting a soluble, non-amyloidogenic fragment formation (Pietri et al., 2013). However, the role of ADAM17 in the development of cerebrovascular and cognitive impairments in AD remains incompletely understood. In this study, we set out to examine the role of ADAM17 in contributing to vascular and cognitive impairments by using an established mouse model of AD, the APP/PS1 mice.
## Animals
The work involving experimental animals was conducted under the protocol approved by the Institutional Animal Care and Use Committee at the Medical College of Georgia, Augusta University. All experimental animal procedures performed in this study were in compliance with the European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes. Experiments were carried out in nine to 10-month-old male and female APP/PS1 mice, which are double transgenic mice expressing a chimeric mouse/human amyloid precursor protein (Mo/HuAPP695swe) and a mutant human presenilin 1 (PS1-dE9), both directed to CNS neurons (MMRRC strain: 034832-JAX, The Jackson Laboratory; Jankowsky et al., 2004). The mice are on the C57BL/6J genetic background. The mice were housed in the animal care facility and accessed rodent chow and tap water ad libitum with a 12-h light:dark cycle.
## Animal behavioral tests
Animal behavioral tests were conducted in the small animal behavior core lab in Augusta University.
## Morris water maze
The Morris Water Maze was used to test spatial learning and memory (Vorhees and Williams, 2006; Bromley-Brits et al., 2011). The water-maze apparatus consisted of a circular pool of 150 cm in diameter and 60 cm in height, filled with water made opaque with non-toxic white paint, kept at a temperature of 22°C. Direct light to the pool was avoided. Different signs were placed around the pool to help the mice to memorize direction. Individual mice received a five-day acquisition training, consisting of four trials per day. In each training trial, a plastic transparent platform was placed 1.5 cm below the water surface; the mice were placed into the pool facing the wall and were allowed to explore the pool to look for the platform until finding the platform, for 90 s. If the mice failed to find the platform within 90 s, they were guided to the platform. The mice were allowed to stay on the platform for 20 s. If the mouse failed to swim for the trial duration, they were removed and allowed a second trial. If upon the second trial, the mouse again failed to swim, it was excluded from the data set results. Velocity of mice swim times were analyzed to assess differences in movement and locomotion. On each trial, the mouse started from a different location across trials and days. After 5 days of acquisition training sessions, the platform was removed from the pool and each mouse was tested in a single 60 s probe trial. Two days after the probe trial, to exclude the mice with vision problems, each mouse was tested in 3 trials with a visible red flag on the platform. The mice which were not be able to find the platform within 60 s were excluded. The swimming path of the mice was recorded by a computer-based video-tracking system and the time spent to find the platform and the target quadrant were recorded.
## Spontaneous Y-maze
The spontaneous Y-maze test is used to measure spatial working memory (Kraeuter et al., 2019). The maze consists of three arms (41 cm long, 16 cm high, and 9 cm wide, labeled A, B, or C) diverging at a 120̊ angle from the central point. The experiments were performed in a dimly illuminated room and the floor of the maze was cleaned with hypochlorous water-soaked paper after each mouse to avoid olfactory cues. Each mouse was placed at the end of the starting arm and allowed to move freely through the maze during a 5-min session. The sequence of arm entries was recorded and the number of arm entries was manually counted in a blinded manner; a mouse was considered to have entered an arm when all four paws were positioned in the arm runway. An actual alternation was defined as entries into all of the three arms on consecutive occasions. The maximum alternation was subsequently calculated by measuring the total number of arm entries minus 2 and the percentage of alternation was calculated as (actual alternation/maximum alternation)*100. Total number of arms entered during the sessions (reflecting locomotor activity) was also recorded. Mice that entered arms less than eight times during the test were eliminated because the data obtained from those mice were not considered to be representative.
## Novel objective recognition
The Novel Objective Recognition (NOR) test was performed similarly as described before (Lueptow, 2017). Mouse behavior is recorded with a video camera. On each day before the experiment, the mice were brought to the testing room 30 mins before starting testing to pre-habituate with the environment. One day preceding the test, after 30 mins habituation, each mouse was allowed to freely explore the box with no objects for 8 mins. On the testing day, two identical objects were placed at two opposite positions within the box at the same distance from the nearest corner. Each mouse was allowed to explore the identical objects for 10 mins, and then returned to their home cage for 4 h. Then, one familiar object and one novel object were placed in the box. Four hours later, each mouse was allowed to explore the objects for 10 mins. Novel objects were identical in size but different in shape and color to the familiar objects. All objects were fixed on the bottom of the box to avoid movement. Box and objects were cleaned with hypochlorous water-soaked paper between each animal. The time mice spent exploring each object was recorded using two stop watches and measured manually. Animals with a total exploration time of less than 3 s during the testing session were excluded from analysis. The location preference in the training phase and recognition index (RI) for (N/F) in the testing phase were calculated using the following formula: Recognition index (RI; N/F) = time exploring novel object / (time exploring novel object + time exploring familiar object) × $100\%$. Delta Novel – Familiar = (time exploring the novel object – time exploring the familiar objects) / (time exploring novel object + time exploring familiar object) × $100\%$.
## Adeno-associated virus injection
Recombinant adeno-associated virus 9 (AAV9) was constructed and purchased from VectorBuilder. Mouse ADAM17-mCherry-AAV9 and eGFP-AAV9 were used in the following experiments. Mouse ADAM17-mCherry-AAV9, or eGFP-AAV9 controls, were used to overexpress ADAM17 in the APP/PS1 mice through intravenous tail-vein injections (particle number: 1 × 1011 of AAV9 diluted in 200 μL sterile PBS, total volume of 300 μL injected). Mice were assessed for functional and morphological endpoints 3 months after injections.
## Purification of brain microvessels
The method used for mice brain microvessel isolation was performed as described previously (Paraiso et al., 2020). Briefly, the brain tissue was placed in a glass tube containing 8 mL of ice-cold HBSS solution (14025-076, Gibco) and homogenized. The brain homogenate was mixed with an equal volume of $40\%$ Ficoll (17030010, Cytiva) to a final concentration of $20\%$ Ficoll in HBSS solution. The tubes were shaken before centrifugation and the resulting homogenate was centrifuged at 5,800g for 20 mins at 4°C. The pellet was gently washed with 5 mL of $1\%$ BSA/HBSS (BP1600, Fisher BioReagents). It was then filtered with 300 μm nylon mesh and the flow through was washed with $1\%$ BSA/HBSS. The mixture was filtered through a 30 μm strainer, with a speed of 1 drop every 2 s. The brain microvessels were captured on top of the strainer. The 30 μm strainer was inverted and rinsed with 10 mL of $1\%$ BSA/HBSS, to wash the vessels down. The wash-through was centrifuged at 5,800g for 20 mins at 4°C. The supernatant was discarded and the vessels pellet was saved for further analysis.
## Western blot
Whole brain samples and the isolated brain microvasculature were homogenized in radio-immunoprecipitation assay (RIPA, R0278, Sigma) buffer mixed with $1\%$ protease inhibitor cocktail (P8340, Sigma). Protein concentration was measured by Bradford assay. Equal amounts of protein were loaded for gel electrophoresis. After blotting, membranes (Hybond-P, GE Healthcare) were probed with a rabbit polyclonal anti-ADAM17 antibody (1:1000, ab2051, Abcam), followed by incubation with a HRP linked secondary antibody (anti-rabbit IgG, 7074S, Cell Signaling). Enhanced chemiluminescence was visualized autoradiographically by ChemiDoc XRS+ (170-5061, BioRad). Protein expression was normalized to β-actin (1:1000, 8H10D10, Cell Signaling).
## Videomicroscopic assessment of isolated cerebral arteries of the mouse
Mice, under deep anesthesia, were decapitated and the brain was removed. The bottom section of the brain was placed in ice-cold artificial cerebral spinal fluid (ACSF) solution, equilibrated with a gaseous mixture of $10\%$ O2-$5\%$ CO2 – balanced nitrogen, at a pH of 7.4. Mice were euthanized by exsanguination and removal of the aorta; other vital organs were saved and used for further assessments. With the use of microsurgical instruments and an operating microscope, the posterior cerebral artery was isolated and transferred into an organ chamber containing two glass micropipettes, filled with ACSF solution. The artery was cannulated at both ends and the micropipettes were connected with silicone tubing to a hydrostatic reservoir to set the intraluminal pressure to 50 mmHg. Artery preparations were incubated for a 1-h period; spontaneous arterial tone developed in response to 50 mmHg pressure during the incubation period, if spontaneous tone did not develop during the incubation period, U46619 (~10-8 M) was applied to the arteries to induce an approximately $25\%$ tone. Arteries that did not develop a spontaneous tone or could not be preconstricted were excluded from the analyses. After the incubation period, diameter changes were measured with a videocaliper (Colorado; Bagi and Koller, 2003; Erdei et al., 2006) in response to cumulative concentrations of the endothelium-dependent agonist, acetylcholine (ACh, 10-9–10-6 M, Sigma), and to the direct vascular smooth muscle acting, nitric oxide (NO) donor, sodium nitroprusside (SNP, 10-9–10-5 M, Sigma).
At the end of the experiments, the internal and outer artery diameter was measured in response to incremental increases in intraluminal pressure from 10 to 70 mmHg, in 20 mmHg increments, in the absence of extracellular calcium (calcium-free ACSF solution). The normalized artery diameter was calculated (as measured at 10 mmHg intraluminal pressure). The wall stress (s) was calculated as follows: s = P × D/2WT, where P is pressure and 1 mmHg = 1334 dyne/cm2, and as described (Ohanian et al., 2014). The incremental elastic modulus (Einc) was calculated as follows: Einc = (Δp/Δro)2rix2ro/(ro2−ri2), where ri is the inner, ro is the outer radius, Δro is the change in outer radius in response to intraluminal pressure change of Δp, as described (Mericli et al., 2004).
## Immunohistochemistry
Mice were anesthetized using isoflurane (1-$3\%$) and were perfused transcardially with 20 mL of room temperature 0.01 M PBS followed by 20 mL of cold $4\%$ formaldehyde (sc-281692, Santa Cruz Biotechnology). Then, the mouse brains were dissected and post-fixed in $4\%$ formaldehyde overnight at 4°C. Brains were cryoprotected with $30\%$ sucrose in 0.01 M PBS for a minimum of 72 h. The brains were cut into 40 μm thick cryosections and saved in a cryoprotectant solution (50 mM PBS, ethylene glycol, and glycerol) at −20°C. For immunohistochemical staining, brain slices located at +1.32, −0.82 (cortex), and −1.64 −2.92 (hippocampus and cortex) were used. After being removed from the cryoprotectant solution, brain slices were washed with 0.01 M PBS three times at room temperature, 45 mins per wash. Brain slices were permeabilized with 0.01 M PBS containing $0.5\%$ Triton X-100 (T8532, Sigma) for 10 mins at room temperature. Brain slices were blocked with 0.01 M PBS containing $10\%$ normal horse serum for 1 h at room temperature. Primary antibodies were applied in 0.01 M PBS containing $0.3\%$ Triton X-100 for 48 h at 4°C. To identify β-amyloid plaques, the primary chicken anti-amyloid beta antibody was used (1:1000, AP31802PU-N, Origene) followed by incubation with secondary antibody (1: 3000, anti-chicken, AlexaFluor®555) for 4 h at room temperature. DAPI (H-2000, Vector Laboratories) was used for nuclear staining. Structured illumination microscopy (SIM-Apotome, AxioImagerM2, CarlZeiss) was used for immunofluorescent detection. The analysis of positive β-amyloid staining was quantified using ImageJ.
## Automatic 3-dimensional vessel network reconstruction and quantification
For cerebral vascular network reconstruction, the thickly cut brain slice sections (40 μm) described in the previous section were immuno-labeled with Tomato-Lectin DyLight 594 Antibody (Vector Laboratories, DL-1177, 1:150, overnight, 4°C). DAPI was used for nuclear staining. Structured illumination microscopy (SIM-Apotome, AxioImagerM2, CarlZeiss) was used for immunofluorescent detection. Images were taken using z-stack imaging, with images taken every 0.5 μm, using 20X magnification for vascular reconstruction. Per animal, three fields of view were taken using z-stack imaging and averaged. After obtaining z-stack images via immunofluorescence microscopy, image stacks were uploaded into Vesselucida360® software for unbiased reconstruction and consequent analysis. Vessels were reconstructed using the Automatic Rayburst Crawl tracing option with vessel tracing set at a sensitivity of 70. Refinement of tracing seeds was between 1 and 3 based on the sample. User-guided manual tracing was used following automatic tracing, with the Rayburst Crawl tracing option, followed by manual tracing of any remaining vessels. Reconstructions were exported to the Vesselucida Explorer® software for automatic quantification of vessel network parameters.
## Liquid chromatography and mass spectrometry
For Mass Spectrometry, whole brain samples were homogenized in radio-immunoprecipitation assay (RIPA, Sigma) buffer mixed with $1\%$ protease inhibitor cocktail (Sigma). Afterwards, 50 μg of protein per sample was sent to the Proteomics Core Laboratory at the Medical College of Georgia for protein analysis via liquid chromatography mass spectrometry analysis (LC-MS/MS). Samples were digested for 16 h with trypsin at 37°C, trifluoroacetic acid was added to a final concentration of $0.1\%$ (v/v) to stop the digestion. Samples were centrifuged at 15000g for 5 mins and supernatants were used for the LC-MS analysis. Digested peptide samples were analyzed on an Orbitrap Fusion tribrid mass spectrometer (Thermo Scientific), with an Ultimate 3000 nano-UPLC system (Thermo Scientific) connected. Peptide samples were first trapped on a Pepmap100 C18 peptide trap (5 μm, 0.3 × 5 mm). Cleaned peptides were further separated on a Pepman 100 RSLC C18 column (2.0 μm, 75 μm × 150 mm) at 40°C, using a gradient of $2\%$ to $40\%$ acetonitrile with $0.1\%$ formic acid over 120 min at a flow rate of 300 nL/min. Eluted peptides were introduced into the Orbitrap Fusion MS via nano-electrospray ionization source at a temperature of 300 °C and a spray voltage of 2000 V. The peptides were then analyzed in the MS using data-dependent acquisition in positive mode with the Orbitrap MS analyzer for precursor scans at 120,000 FWHM from 400 to 1600 m/z (with quad isolation) and the ion-trap MS analyzer for MS/MS scans at top-speed mode (3-s cycle time), with dynamic exclusion settings (repeat count 1 and exclusion duration 15 s). Higher-energy C-trap dissociation (HCD) was used to fragment the precursor peptides with a normalized energy level of $30\%$. Raw MS data were processed via the Proteome Discoverer software (ver 1.4) and submitted for SequestHT algorithm against the SwissProt mouse protein database. SequestHT search parameters were set as 10 ppm precursor and 0.6 Da product ion mass tolerance, with static carbidomethylation (+57.021 Da) for cysteine and dynamic oxidation for methionine (+15.995 Da). The percolator peptide spectrum matching (PSM) validator algorithm was used for PSM validation. Proteins unable to be distinguished based on the database search results were grouped to satisfy the principles of parsimony. Proteomic data was analyzed for further analysis if the ΣPSM value ≥2 per group analyzed (WT, eGFP, ADAM17) using unpaired student t-test between two group comparisons. Principal component analysis was run using AtlAnalyze software. Statistically significant ($p \leq 0.05$) proteins identified were used in heat-map analysis (-log10(p-value) vs WT) and for further pathway analysis. Gene Ontology pathway analysis with statistically significant proteins was performed (Panther Classification System). For GO pathway analysis the following parameters were used: Analysis Type: PANTHER Overrepresentation Test; Annotation Version: GO Ontology database; Reference List: *Mus musculus* (all genes in database); Annotation Data Set Used: GO Biological Process complete, GO Molecular Function complete, GO Cellular Component complete; Test Type: Fisher’s Exact; Correction: Calculated False Discovery Rate (FDR). GO pathways identified with an FDR < 0.05 were used for further analysis. GO pathway analysis was visualized in scatterplot format using multidimensional scaling (MSD) whereby semantically similar GO pathways can be visualized in close proximity (Revigo).
## Statistics
All statistical analyses were performed using GraphPad Prism Software. Data were drawn to analyze after being tested for normality using Kolmogorov-Smirnov test. When meeting normality data comparisons between groups were analyzed by two-way repeated-measures ANOVA followed by Sidak’s multiple comparisons test, or with a two-tailed, unpaired Student t-test, as appropriate. A non-parametric Kruskal-Wallis test was used when normality assumptions were not met. Data are expressed either as mean±SEM or box-and-whisker plots, in which the minimum, the 25th percentile, the median, the 75th percentile, and the maximum are presented, and + indicates the mean of values. $P \leq 0.05$ was considered statistically significant.
## APP/PS1 mice display impairment in short term memory
To confirm impaired short term spatial memory and learning function in the APP/PS1 AD mouse model, 9–10-month-old, female and male, APP/PS1 and WT mice were subjected to a Morris Water maze (MWM) test. We found that the individual latency curves of APP/PS1 and WT mice were similar on day 1 and 2 in the learning phase, indicating that the APP/PS1 mice did not differ in swimming ability, or with motivation to escape from the pool. From day 3 to 5, however, acquisition trials revealed a progressive and statistically significant decrease in latencies in APP/PS1 mice compared to WT mice (Figure 1A). During the probe trial, APP/PS1 mice spent longer times to reach the platform area, spent less time in the platform area upon finding it, and had a trend of less platform line crossings (Figures 1B–D). There were no differences between WT and APP/PS1 mice for the percentage of spontaneous alternations or in velocity (locomotion; Figures 1E,F).
**Figure 1:** *Impaired learning capacities of APP/PS1 mice in the Morris Water Maze test (MWM) compared to WT mice. (A) Increase in the latency (seconds) to reach the target reflected impaired learning capacities of APP/PS1 mice (n=22) compared to WT mice (n=24) in the 5 day training phase in MWM. *P < 0.05 vs WT; #P < 0.05 vs day 1. (B,C) APP/PS1 mice spent a reduced time in the platform (B) and in the target area (C). (D) APP/PS1 mice had less platform line crossing compared to WT mice on the probe test. (E) The percent of spontaneous alternations in the Spontaneous Y-Maze test were not significantly different between groups. (F) There was no difference in velocity between WT and APP/PS1 mice during the testing. *P < 0.05 vs WT. Unpaired t test was used.*
## APP/PS1 mice exhibit a reduced protein expression of ADAM17 in cerebral microvessels
ADAM17 has been implicated in various cardiovascular pathologies, yet there have been only limited number of studies elucidating the role of vascular ADAM17 in neurogenerative diseases. Here, we found that while the protein expression of ADAM17 was similar in WT and APP/PS1 mouse whole brain lysates (Figure 2A), the expression of ADAM17 was remarkably reduced in purified cerebral microvessel preparations obtained from the APP/PS1 mice, when compared to those of WT mice (Figures 2B,C).
**Figure 2:** *Reduced ADAM17 expression in cerebral microvessels of APP/PS1 mice is restored by ADAM17-AAV9. (A) ADAM17 expression was measured in whole brain lysates and (B) in isolated, purified cerebral microvascular preparations in WT and APP/PS1 mice. (C) Representative micrograph depicts a cerebral microvascular segment after centrifugation-based purification. (D,E) Representative image and quantification data of western immunoblots showing ADAM17 expression in cerebral microvessels of WT or APP/PS1 mice 3 months after injections with either eGFP-AAV9 or ADAM17-AAV9 virus. n=3−4. *P < 0.05. Two way ANOVA was used.*
## ADAM17 re-expression improves short-term memory and cognitive function in APP/PS1 mice
In order to delineate the role of vascular ADAM17 in impaired short-term memory in the APP/PS1 mice, we used a systemic, AAV9-mediated genetic delivery approach to increase ADAM17 expression. We found that 3 months after the delivery of the ADAM17-AAV9 construct, the protein expression of ADAM17 in cerebral microvessels of APP/PS1 mice was augmented compared to eGFP-AAV9 injected mice and that the expression level was similar to that measured in the cerebral microvessels of age-matched WT mice (Figures 2D,E).
We then determined the effect of ADAM17 re-expression on short-term memory and cognitive dysfunction in the APP/PS1 mice by performing multiple, independent behavioral tests. WT mice and APP/PS1 mice receiving eGFP-AAV9 served as controls. First, the ADAM17-AAV9 or eGFP-AAV9 injected mice were repeatedly subjected to the Morris Water Maze (MWM) test. We found that, 3 months after ADAM17 re-expression, APP/PS1 mice spent shorter times to reach the platform area (A-40) and spent longer times in the platform area, while APP/PS1 mice that received the eGFP-AAV9 injection displayed increased latency curves, similar to the level of APP/PS1 mice before the AAV injections (Figures 3A,B).
**Figure 3:** *Increased ADAM17 expression improves cognitive function of APP/PS1 mice. (A,B) ADAM17 overexpressed APP/PS1 mice show a reduced latency (seconds) to reach the target area and increased total time in the target area compared to before ADAM17-AAV (adeno-associated virus) injection in a MWM probe test. There was no difference in APP/PS1 mice after receiving an eGFP-AAV9 injection. (C,D) The number of arm entries was lower in APP/PS1 mice with eGFP-AAV9 injected mice compared to the WT mice. This number was improved in the APP/PS1 mice with ADAM17-AAV9 virus injection. There was no difference in the percent of spontaneous alternation between the WT mice, eGFP, or ADAM17 overexpressed APP/PS1 mice in a spontaneous Y-Maze test. *P < 0.05. n=6−8. Unpaired t test was used.*
Mice in the experimental groups were additionally subjected to spontaneous Y-maze and Novel Objective Recognition (NOR) tests. In the spontaneous Y-maze, the eGFP-AAV9 injected APP/PS1 mice had a reduced number of arm-entries compared with the WT and ADAM17-AAV9 injected APP/PS1 mice (Figure 3C). There was no difference observed in the percent of spontaneous alternations between the three groups (Figure 3D). In the NOR test, compared to the age-matched WT mice, the APP/PS1 eGFP-AAV9 injected mice had a reduced delta novel-familiar score (Figure 4A), recognition index (novel/familiar, N/F) (Figure 4B) and discrimination index (d2 ratio) (Figure 4C). These parameters were similar in WT and the ADAM17-AAV9 injected APP/PS1 mice (Figures 4A–C). There was no significant difference in the total exploration time among the experimental groups (Figure 4D). Taken together, these observations indicates that re-expression of ADAM17 in APP/PS1 mice improved some of the indices of short-term memory and cognitive function to a level comparable to the aged-matched WT control mice.
**Figure 4:** *Increased ADAM17 expression in APP/PS1 mice improves performance in Novel Object Recognition test. (A,B) There was reduced time to explore the novel object and a reduced recognition index in the eGFP-AAV9 injected, but not in the ADAM17-AAV9 injected, APP/PS1 mice in a NOR test [Delta Novel – Familiar score, Recognition Index (Novel/Familiar, N/F)]. (C) There was decrease in the discrimination index (d2 ratio). (D) There was no change in Total Exploration Time between the three experimental groups. n=6−7. Two way ANOVA was used.*
## Deposition of amyloid beta is not affected by ADAM17 re-expression in APP/PS1 mice
Brain sections of experimental mice were immunostained for amyloid beta (Aβ) for semi-quantitative assessment of cortical amyloidosis. When compared to age-matched WT mice, we found that there was a significant increase in the number of Aβ plaques in the APP/PS1 mouse cortex receiving the control eGFP-AAV9. Whereas we found no significant effect, i.e. decrease in the number of Aβ plaques, in the APP/PS1 mice injected with the ADAM17-AAV9 construct (Figures 5A,B).
**Figure 5:** *Amyloid-β plaque density was unchanged after ADAM17 re-expression in APP/PS1 mice. (A,B) Amyloid-beta (Aβ) plaques were identified via immunofluorescent staining and quantified. The number of Aβ plaques was significantly increased in the APP/PS1 mice and was unchanged with the ADAM17 injection. n=6−11. *p < 0.05 **p < 0.01. Two way ANOVA was used.*
## ADAM17 re-expression improves cerebral artery vasodilator function in APP/PS1 mice
We assessed cerebral artery vasodilator function and vascular biomechanics in WT, as well as eGFP-AAV9 and ADAM17-AAV9 injected APP/PS1 mice using isolated and pressurized cerebral arteries. We found that the endothelium-dependent vasodilation in response to acetylcholine (ACh) was significantly reduced in eGFP overexpressed APP/PS1 mice, when compared to WT controls (Figure 6A). ADAM17 re-expressed APP/PS1 mice showed an improved vasodilatory response to acetylcholine (ACh) compared to the eGFP overexpressed APP/PS1 mice; no significant difference was observed between the ADAM17 re-expressed APP/PS1 mice and WT mice (Figure 6A). There were no differences in the nitric oxide donor, sodium nitroprusside (SNP)-induced, vascular smooth muscle acting vasodilation in the isolated cerebral arteries of WT, eGFP, or ADAM17 overexpressed APP/PS1 mice (Figure 6B).
**Figure 6:** *ADAM17 re-expression improves cerebral arteriole vasodilator function in APP/PS1 mice. (A) ADAM17 overexpressed APP/PS1 mice show an improved vasodilation in response to acetylcholine (ACh) in isolated basilar arteries compared to eGFP overexpressed APP/PS1 mice, which have a decreased dilatory response compared to the WT control mice. (B) There were no differences seen in the dilatory responses to sodium nitroprusside (SNP) in isolated basilar arteries of WT, eGFP, or ADAM17 overexpressed APP/PS1 mice. (C) Normalized (to 10 mmHg) diameter of arterioles in calcium free solution in response to incremental increases to intraluminal pressure (10 to 70 in 20 mmHg increments). (D–F) Calculated circumferential wall stress, incremental elastic modules and elastic modulus wall stress relationships in WT, eGFP-or ADAM17-overexpressed APP/PS1 mice. n=3−6 mice per group. *p < 0.05 WT, ADAM17 vs eGFP APP/PS1. Two way ANOVA was used.*
We also found that cerebral arteries of APP/PS1 (eGFP-AAV9) mice displayed a reduced passive diameter (in calcium free PSS and normalized to 10 mmHg) to incremental increases to intraluminal pressure (from 10 to 70 mmHg), when compared to WT controls (Figure 6C). These changes were accompanied by increases in artery wall circumferential stress and elastic modulus, as well as elastic modulus wall stress relationships in the APP/PS1 (eGFP-AAV9) mice (Figures 6D–F). Delivery and re-expression of ADAM17 in the APP/PS1 mice did not significantly affect these biomechanical properties of the cerebral arteries (Figures 6D–F). Taken together, these results indicate that re-expression of ADAM17 was associated with selective improvement of endothelial-dependent vasodilator function in the cerebral arteries of APP/PS1 mice.
## ADAM17 re-expression in APP/PS1 mice had no effect on the vascularization of cerebral cortex
In order to quantify vascularization and microvessel density of the cerebral cortex, the vascular tree was immunofluorescently labelled and in thick sections (40 μm), z-stack images were used for 3-dimensional reconstruction followed by unbiased quantification of microvascular networks (Figure 7A). The microvascular networks in the WT, eGFP control, and ADAM17 re-expressed APP/PS1 groups showed no differences in the measured parameters, including, total vessel length, surface area, vessel volume, or the number of branching nodes (Figures 7B–E).
**Figure 7:** *ADAM17 re-expression had no effect on the vascularization of the cerebral cortex in APP/PS1 mice. (A) Representative 3-dimensional reconstructions of the cerebral small vessel networks in WT, eGFP, and ADAM17 treated APP/PS1 mice. (B–E) No differences in the number of branching nodes, total vessel length, vessel surface area, or vessel volume was found between the three groups. Images were taken at 20X magnification, 0.5 μm slices in the z-plane. n=3 fields of view per mouse, 3 mice per group. *p < 0.05. Kruskal-Wallis test was used.*
## Proteomic profiling after ADAM17 re-expression in APP/PS1 mice
We performed mass spectrometry analysis to identify proteins that were differentially expressed in the brain of the APP/PS1 mice, which were potentially affected by the AAV-mediated re-expression of ADAM17. Liquid chromatography mass spectroscopy (LC-MS) was performed in triplicate per group. First, we wanted to identify the general trend in the proteomic profile between WT, APP/PS1 + eGFP, and APP/PS1 + ADAM17. Via a principal component analysis (PCA), we saw a differential trend in the protein profile in APP/PS1 + eGFP mice versus the WT control mice. Whereby in APP/PS1 mice with ADAM17 re-expression, we saw an overall trend in the protein profile shifted towards the WT mice (Figure 8A). To further elucidate microvascular proteomic changes due to ADAM17 re-expression, we analyzed individual proteins identified via LC-MS, with a cut-off of a ΣPSM (peptide-spectrum match) of 2 per group (Figure 8B). Through these proteomic analyses, multiple proteins were identified as being significantly downregulated in the APP/PS1 + eGFP mice, that were augmented via ADAM17 re-expression, including numerous proteins that have been studied in relation to neurodegenerative diseases, such as Septin, Ankyrin-2 (Ank2), and Moesin (Msn), among others (Figure 8C).
**Figure 8:** *Proteomic profile of the brain of APP/PS1 mice. (A) A principal component analysis of proteins with ΣPSM per animal group ≥2, showing a difference in overall protein profiles between WT and eGFP mice, which was slightly corrected in the ADAM17 APP/PS1 mice. (B) Volcano plots of the protein profiles between WT/eGFP and eGFP/ADAM17 show the differentially up-and downregulated individual proteins. (C) Heatmap of the –log10(p-value) per group compared to WT show the up-and downregulated, significant proteins. 3 mice per group. *p < 0.05. Unpaired t-test was used.*
Furthermore, Gene Ontology (GO) pathway analyses (Panther®) were completed in order to identify significant differential pathways that ADAM17 may impose its beneficial effects through. Using the significant identified proteins, biological processes, molecular functions, and cellular components were identified using an FDR of <0.05. The top 15 significant pathways per analysis are listed with the corresponding enrichment scores (Figure 9A). Moreover, these GO biological processes identified were mapped (Revigo®) to identify connected pathways, showing significant differences occur in pathways connected to amyloid precursor protein as expected, as well as in many metabolic pathways and in cellular organizational pathways (Figure 9B).
**Figure 9:** *Pathway analysis in the brain of APP/PS1 mice. (A) Using the significantly differentiated proteins identified via proteomic analysis, a Gene Ontology pathway analysis was performed. Pathways with an FDR <0.05 were included, the top 15 significant pathways per category are graphed by enrichment scoring. (B) GO pathway analysis of biological processes was visualized by functional grouping. 3 mice per group.*
## Discussion
The results from the present study indicate that cerebrovascular ADAM17 plays a role in the pathogenesis of Alzheimer’s disease (AD). This conclusion is supported by our results showing that short-term memory and cognitive dysfunction in the APP/PS1 mouse model of AD is associated with a reduced ADAM17 expression in cerebral microvessels, whereas re-expression of ADAM17 restored endothelium-dependent vasodilator function in cerebral arteries and improved memory and cognitive function in the APP/PS1 mice.
In this study, first we confirmed that 9-10 months old APP/PS1 mice (carrying transgenes for both APP bearing the Swedish mutation and PSEN1 containing an L166P mutation, both under the control of the Thy1 promoter) display short-term memory and cognitive deficits. Prior studies, including works from our laboratory on human autopsy findings, showed impaired microvascular vasodilator function in the presence of low and high AD neuropathological changes (Snowdon et al., 1997; Bagi et al., 2018; Park et al., 2018; Santisteban and Iadecola, 2018; Bagi et al., 2022). It is known that vascular impairments are highly prevalent in older adults with hypertension and type 2 diabetes (Prasad et al., 2017), which are risk factors for dementia and AD (Udelson, 2011; Goyal et al., 2016; Silverman et al., 2016). While clinical and preclinical studies argue that cerebrovascular endothelial dysfunction contributes to AD pathologies, the molecular mechanisms leading to cognitive decline remains largely unknown.
ADAM17 (also known as tumor necrosis factor (TNF)-α converting enzyme or TACE) exhibits α-secretase activity and has been implicated in APP cleavage, which results in a soluble, non-amyloidogenic fragment formation (Pietri et al., 2013). ADAM17 has multiple substrates, including mediators of vascular inflammation, inflammation resolution, and angiogenic factors (Gooz et al., 2009; Gooz, 2010). A previous study has found that a loss-of-function genetic variant in ADAM17 is associated with the pathogenesis of AD in humans (Hartl et al., 2020), whereas in rodent models genetic ADAM17 deletion caused an impaired collateral circulation formation and vascular growth in the cerebral surface arterioles (Lucitti et al., 2012). Thus, we hypothesized that in the microvasculature, ADAM17 could play an important role in the development of AD neuropathological changes and cognitive decline. Abnormal processing of APP is thought to promote AD development (Allinson et al., 2003; Sastre et al., 2008), among others, by causing cerebrovascular impairments (Sastre et al., 2008; Iadecola, 2017). In this study, we confirmed that β-amyloid plaque density is increased in the cerebral cortex of the APP/PS1 mice. However, we found that systemic delivery and AAV-mediated re-expression of ADAM17 in APP/PS1 mice improved cognitive functioning and endothelium-dependent vasodilator function (but not vascular wall biomechanics and cerebral vascular density), without affecting β-amyloid plaque density. These findings suggested beneficial effects of restoring microvascular ADAM17 expression in the APP/PS1 mice displaying already accumulated β-amyloid (Aβ) plaques. Based on our results we cannot confirm or exclude the possibility that a reduced microvascular ADAM17 expression via a reduced α-secretase activity and blunted non-amyloidogenic APP cleavage may have contributed to Aβ accumulation in the APP/PS1 mice, which has yet to be confirmed in longitudinal studies.
A common denominator in cerebrovascular disease and AD, and a likely contributor to microvascular dysfunction is chronic, low-grade inflammation (Borlaug and Paulus, 2011; Paulus and Tschope, 2013; Franssen et al., 2016). It is plausible that the role of ADAM17 can be also attributed to its ability to affect cerebrovascular inflammatory processes in AD mice. In this context, previous studies showed that deletion of TNFα reduces remyelination repair process in multiple sclerosis and cerebral ischemia mouse models, indicating a protective role for TNFα (Arnett et al., 2001; Lambertsen et al., 2009). ADAM17 not only sheds the membrane bound TNFα, but also cleaves both type I and type II TNFα receptors (TNFRI and TNFRII). Interestingly, deletion of TNFRI inhibits Aβ generation and prevents learning and memory deficits in AD mice (He et al., 2007; Lourenco et al., 2013). TNFRII, which has a higher affinity to the transmembrane TNFα and shows some anti-inflammatory effects, is considered to be protective in AD and ablation of TNFRII impairs cognition in AD mice (Naude et al., 2014). On the other hand and in contrast to the aforementioned studies, deletion of both TNFRI and TNFRII accelerated AD (Montgomery et al., 2011), whereas TNFα inhibitors and an anti-TNF antibody (infliximab) improved cognitive decline in AD animal models and human patients (McAlpine et al., 2009; Shi et al., 2011). Collectively, these studies along with the results from our present findings suggest that a maintained ADAM17 activity, via regulating substrate availability, such as TNFα or TNFRI and TNFRII plays a role in AD pathogenesis. Future studies are needed to determine if reduced expression or re-expression of ADAM17 alters the proteolytic cleavage of TNFα or TNFRI and TNFRII, whereby mediating the adverse or beneficial microvascular and cognitive effects in the APP/PS1 mice.
In addition, in this study we performed mass spectroscopy proteomic analysis in order to identify differentially expressed molecules involved in AD pathogenies, neurodegeneration, and repair; and test whether specific molecules and/or pathways were selectively reversed by ADAM17 re-expression. Through proteomic analysis, we identified multiple proteins that were significantly up or downregulated in the APP/PS1 mice, that were also changed, some back to normal levels, after ADAM17 re-expression. These proteins included Septin, Ankyrin-2, and Moesin, which have been previously implicated and studied in relation to neurodegenerative diseases, including AD. Furthermore, pathway analyses identified several candidate molecules that are associated with regulation of APP and β-amyloid formation and other biological and morphological quality regulatory proteins as well as neuronal morphogenesis and development. In this regard, further studies are needed to confirm whether these candidates can be associated or mechanistically involved in ADAM17-mediated regulation of microvascular and cognitive functioning in the APP/PS1 mice.
## Limitations
There are several limitations to our study. It should be noted that our results provide no direct evidence for the mechanistic, causative link between impaired/improved vasodilation and reduced/restored ADAM17 expression in cerebral microvessels, which has yet to be elucidated and will be part of future studies. We did not measure overall, basal, or stimulated cerebral blood flow changes in experimental groups, which could also impact and determine abnormal cognitive functioning and the role of ADAM17 in the APP/PS1 mice. In addition, possible mechanisms independent from that of a microvascular mechanism in leading to impaired or improved cognitive functioning, which can also be related to ADAM17 in APP/PS1 mice, cannot be excluded. For example, we cannot exclude the role of astrocytes, microglia, or neuronal mechanisms, which we did not evaluate in this study. In this regard, it should be noted that we used systemic delivery and re-expression of ADAM17 in the APP/PS1 mice. Although we measured changes in ADAM17 protein expression in purified microvascular preparations, the expressional changes in other aforementioned cell types could have an impact on AD related neuropathological changes and memory and cognitive function in the APP/PS1 mice, which has yet to be delineated in future studies.
In conclusion, our study identifies a reduced expression of microvascular ADAM17 as a novel mechanism by which microvascular dysfunction occurs in an AD mouse model and by which it could contribute to the development of cognitive decline and memory deficits. We propose that targeting and potentially restoring normal activation of ADAM17 to improve microvascular endothelial function could be an effective approach, as it appears to improve some of the key molecular abnormalities previously implicated in AD pathogenesis.
## Data availability statement
We declare that the data supporting the findings of this study are available within the article and from the corresponding authors upon request. The mass spectrometry proteomics data have been deposited to the ProteomeXchange *Consortium via* the PRIDE partner repository with the dataset identifier PXD040026.
## Ethics statement
The animal study was reviewed and approved by IACUC at Augusta University.
## Author contributions
ZB conceptualized the project. ZB, YT, KF, HS, VB, LL, RW, and JF performed the experiments and analyzed data. YT, KF, RR, JF, and ZB wrote, edited, and approved the manuscript. JF and ZB supervised the research. All authors contributed to the article and approved the submitted version.
## Funding
This work was supported by awards from the National Institutes of Health, National Institute on Aging R01AG054651 to ZB and the T32HL155011 to KF.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Allinson T. M. J., Parkin E. T., Turner A. J., Hooper N. M.. **ADAMs family members as amyloid precursor protein alpha-secretases**. *J Neurosci Res.* (2003) **74** 342-352. DOI: 10.1002/jnr.10737
2. Arnett H. A., Mason J., Marino M., Suzuki K., Matsushima G. K., Ting J. P. Y.. **TNF alpha promotes proliferation of oligodendrocyte progenitors and remyelination**. *Nat Neurosci.* (2001) **4** 1116-1122. DOI: 10.1038/nn738
3. Bagi Z., Brandner D. D., Le P., McNeal D. W., Gong X., Dou H.. **Vasodilator dysfunction and oligodendrocyte dysmaturation in aging white matter**. *Ann Neurol.* (2018) **83** 142-152. DOI: 10.1002/ana.25129
4. Bagi Z., Koller A.. **Lack of NO-Mediation of Flow-Dependent Arteriolar Dilation in Diabetes is Restored by Sepiapterin**. *J of Vascular Research.* (2003) **40** 47-57. DOI: 10.1159/000068938
5. Bagi Z., Kroenke C. D., Fopiano K. A., Tian Y., Filosa J. A., Sherman L. S.. **Association of cerebral microvascular dysfunction and white matter injury in Alzheimer's disease**. *Geroscience.* (2022) **44** 1-14. DOI: 10.1007/s11357-022-00585-5
6. Borlaug B. A., Paulus W. J.. **Heart failure with preserved ejection fraction: pathophysiology, diagnosis, and treatment**. *Eur Heart J.* (2011) **32** 670-679. DOI: 10.1093/eurheartj/ehq426
7. Bromley-Brits K., Deng Y., Song W.. **Morris water maze test for learning and memory deficits in Alzheimer's disease model mice**. *J Vis Exp.* (2011) **20** 2920. DOI: 10.3791/2920-v
8. Corriveau R. A., Bosetti F., Emr M., Gladman J. T., Koenig J. I., Moy C. S.. **The Science of Vascular Contributions to Cognitive Impairment and Dementia (VCID): A Framework for Advancing Research Priorities in the Cerebrovascular Biology of Cognitive Decline**. *Cell Mol Neurobiol.* (2016) **36** 281-288. DOI: 10.1007/s10571-016-0334-7
9. Erdei N., Toth A., Pasztor E. T., Papp Z., Edes I., Koller A.. **High-fat diet-induced reduction in nitric oxide-dependent arteriolar dilation in rats: role of xanthine oxidase-derived superoxide anion**. *Am J Physiol Heart Circ Physiol.* (2006) **291** H2107-H2115. DOI: 10.1152/ajpheart.00389.2006
10. Franssen C., Chen S., Unger A., Korkmaz H. I., De Keulenaer G. W., Tschope C.. **Myocardial Microvascular Inflammatory Endothelial Activation in Heart Failure With Preserved Ejection Fraction**. *JACC Heart Fail.* (2016) **4** 312-324. DOI: 10.1016/j.jchf.2015.10.007
11. Gooz M.. **ADAM-17: the enzyme that does it all**. *Critical reviews in biochemistry and molecular biology.* (2010) **45** 146-169. DOI: 10.3109/10409231003628015
12. Gooz P., Gooz M., Baldys A., Hoffman S.. **ADAM-17 regulates endothelial cell morphology, proliferation, and in vitro angiogenesis**. *Biochem Biophys Res Commun.* (2009) **380** 33-38. DOI: 10.1016/j.bbrc.2009.01.013
13. Gorelick P. B., Scuteri A., Black S. E., Decarli C., Greenberg S. M., Iadecola C.. **Vascular contributions to cognitive impairment and dementia: a statement for healthcare professionals from the american heart association/american stroke association**. *Stroke.* (2011) **42** 2672-2713. DOI: 10.1161/STR.0b013e3182299496
14. Goyal P., Almarzooq Z. I., Horn E. M., Karas M. G., Sobol I., Swaminathan R. V.. **Characteristics of Hospitalizations for Heart Failure with Preserved Ejection Fraction**. *Am J Med.* (2016) **129** e15-e26. DOI: 10.1016/j.amjmed.2016.02.007
15. Hartl D., May P., Gu W., Mayhaus M., Pichler S., Spaniol C.. **A rare loss-of-function variant of ADAM17 is associated with late-onset familial Alzheimer disease**. *Mol Psychiatr.* (2020) **25** 629-639. DOI: 10.1038/s41380-018-0091-8
16. He P., Zhong Z. Y., Lindholm K., Berning L., Lee W., Lemere C.. **Deletion of tumor necrosis factor death receptor inhibits amyloid beta generation and prevents learning and memory deficits in Alzheimer's mice**. *Journal of Cell Biology.* (2007) **178** 829-841. DOI: 10.1083/jcb.200705042
17. Iadecola C.. **The pathobiology of vascular dementia**. *Neuron.* (2013) **80** 844-866. DOI: 10.1016/j.neuron.2013.10.008
18. Iadecola C.. **The Neurovascular Unit Coming of Age: A Journey through Neurovascular Coupling in Health and Disease**. *Neuron.* (2017) **96** 17-42. DOI: 10.1016/j.neuron.2017.07.030
19. Jankowsky J. L., Fadale D. J., Anderson J., Xu G. M., Gonzales V., Jenkins N. A.. **Mutant presenilins specifically elevate the levels of the 42 residue beta-amyloid peptide in vivo: evidence for augmentation of a 42-specific gamma secretase**. *Hum Mol Genet.* (2004) **13** 159-170. DOI: 10.1093/hmg/ddh019
20. Kraeuter A. K., Guest P. C., Sarnyai Z.. **The Y-Maze for Assessment of Spatial Working and Reference Memory in Mice**. *Methods Mol Biol.* (2019) **1916** 105-111. DOI: 10.1007/978-1-4939-8994-2_10
21. Lambertsen K. L., Clausen B. H., Babcock A. A., Gregersen R., Fenger C., Nielsen H. H.. **Microglia Protect Neurons against Ischemia by Synthesis of Tumor Necrosis Factor**. *Glia.* (2009) **29** 1319-S59. DOI: 10.1523/JNEUROSCI.5505-08.2009
22. Lourenco M. V., Clarke J. R., Frozza R. L., Bomfim T. R., Forny-Germano L., Batista A. F.. **TNF-alpha Mediates PKR-Dependent Memory Impairment and Brain IRS-1 Inhibition Induced by Alzheimer's beta-Amyloid Oligomers in Mice and Monkeys**. *Cell Metab.* (2013) **18** 831-843. DOI: 10.1016/j.cmet.2013.11.002
23. Lucitti J. L., Mackey J. K., Morrison J. C., Haigh J. J., Adams R. H., Faber J. E.. **Formation of the collateral circulation is regulated by vascular endothelial growth factor-A and a disintegrin and metalloprotease family members 10 and 17**. *Circ Res.* (2012) **111** 1539-1550. DOI: 10.1161/CIRCRESAHA.112.279109
24. Lueptow L. M.. **Novel Object Recognition Test for the Investigation of Learning and Memory in Mice**. *J Vis Exp.* (2017) **30** 55718. DOI: 10.3791/55718
25. McAlpine F. E., Lee J. K., Harms A. S., Ruhn K. A., Blurton-Jones M., Hong J.. **Inhibition of soluble TNF signaling in a mouse model of Alzheimer's disease prevents pre-plaque amyloid-associated neuropathology**. *Neurobiology of Disease.* (2009) **34** 163-177. DOI: 10.1016/j.nbd.2009.01.006
26. Mericli M., Nadasy G. L., Szekeres M., Varbiro S., Vajo Z., Matrai M.. **Estrogen replacement therapy reverses changes in intramural coronary resistance arteries caused by female sex hormone depletion**. *Cardiovasc Res.* (2004) **61** 317-324. DOI: 10.1016/j.cardiores.2003.11.022
27. Montgomery S. L., Mastrangelo M. A., Habib D., Narrow W. C., Knowlden S. A., Wright T. W.. **Ablation of TNF-RI/RII Expression in Alzheimer's Disease Mice Leads to an Unexpected Enhancement of Pathology Implications for Chronic Pan-TNF-alpha Suppressive Therapeutic Strategies in the Brain**. *Am J Pathol.* (2011) **179** 2053-2070. DOI: 10.1016/j.ajpath.2011.07.001
28. Naude P. J. W., Dobos N., van der Meer D., Mulder C., Pawironadi K. G. D., den Boer J. A.. **Analysis of cognition, motor performance and anxiety in young and aged tumor necrosis factor alpha receptor 1 and 2 deficient mice**. *Behav Brain Res.* (2014) **258** 43-51. DOI: 10.1016/j.bbr.2013.10.006
29. Ohanian J., Liao A., Forman S. P., Ohanian V.. **Age-related remodeling of small arteries is accompanied by increased sphingomyelinase activity and accumulation of long-chain ceramides**. *Physiol Rep.* (2014) **28** e12015. DOI: 10.14814/phy2.12015
30. Paraiso H. C., Wang X., Kuo P. C., Furnas D., Scofield B. A., Chang F. L.. **Isolation of Mouse Cerebral Microvasculature for Molecular and Single-Cell Analysis**. *Front Cell Neurosci.* (2020) **14** 84. DOI: 10.3389/fncel.2020.00084
31. Park M., Moon Y., Han S. H., Kim H. K., Moon W. J.. **Myelin loss in white matter hyperintensities and normal-appearing white matter of cognitively impaired patients: a quantitative synthetic magnetic resonance imaging study**. *Eur Radiol.* (2018) **29** 4914-4921. DOI: 10.1007/s00330-018-5836-x
32. Paulus W. J., Tschope C.. **A novel paradigm for heart failure with preserved ejection fraction: comorbidities drive myocardial dysfunction and remodeling through coronary microvascular endothelial inflammation**. *J Am Coll Cardiol.* (2013) **62** 263-271. DOI: 10.1016/j.jacc.2013.02.092
33. Pietri M., Dakowski C., Hannaoui S., Alleaume-Butaux A., Hernandez-Rapp J., Ragagnin A.. **PDK1 decreases TACE-mediated alpha-secretase activity and promotes disease progression in prion and Alzheimer's diseases**. *Nat Med.* (2013) **19** 1124-1131. DOI: 10.1038/nm.3302
34. Prasad M., Matteson E. L., Herrmann J., Gulati R., Rihal C. S., Lerman L. O.. **Uric Acid Is Associated With Inflammation, Coronary Microvascular Dysfunction, and Adverse Outcomes in Postmenopausal Women**. *Hypertension.* (2017) **69** 236-242. DOI: 10.1161/HYPERTENSIONAHA.116.08436
35. Santisteban M. M., Iadecola C.. **Hypertension, dietary salt and cognitive impairment**. *J Cereb Blood Flow Metab.* (2018) **38** 2112-2128. DOI: 10.1177/0271678X18803374
36. Sastre M., Heneka M., Walter J., Klockgether T.. **Stimulation with noradrenaline induces a reduction of beta-amyloid levels in neuronal cells and in APP/PS1 transgenic mice**. *Fund Clin Pharmacol.* (2008) **22** 98
37. Shi J. Q., Wang B. R., Jiang W. W., Chen J., Zhu Y. W., Zhong L. L.. **Cognitive Improvement with Intrathecal Administration of Infliximab in a Woman with Alzheimer's Disease**. *Journal of the American Geriatrics Society.* (2011) **59** 1142-1144. DOI: 10.1111/j.1532-5415.2011.03445.x
38. Silverman M. G., Patel B., Blankstein R., Lima J. A., Blumenthal R. S., Nasir K.. **Impact of Race, Ethnicity, and Multimodality Biomarkers on the Incidence of New-Onset Heart Failure With Preserved Ejection Fraction (from the Multi-Ethnic Study of Atherosclerosis)**. *Am J Cardiol.* (2016) **117** 1474-1481. DOI: 10.1016/j.amjcard.2016.02.017
39. Snowdon D. A., Greiner L. H., Mortimer J. A., Riley K. P., Greiner P. A., Markesbery W. R.. **Brain infarction and the clinical expression of Alzheimer disease**. *The Nun Study. JAMA.* (1997) **277** 813-817. DOI: 10.1001/jama.1997.03540340047031
40. Udelson J. E.. **Heart failure with preserved ejection fraction**. *Circulation.* (2011) **124** e540-e543. DOI: 10.1161/CIRCULATIONAHA.111.071696
41. Vorhees C. V., Williams M. T.. **Morris water maze: procedures for assessing spatial and related forms of learning and memory**. *Nat Protoc.* (2006) **1** 848-858. DOI: 10.1038/nprot.2006.116
|
---
title: 'Association between dietary minerals and glioma: A case-control study based
on Chinese population'
authors:
- Weichunbai Zhang
- Yongqi He
- Xun Kang
- Ce Wang
- Feng Chen
- Zhuang Kang
- Shoubo Yang
- Rong Zhang
- Yichen Peng
- Wenbin Li
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC10018027
doi: 10.3389/fnut.2023.1118997
license: CC BY 4.0
---
# Association between dietary minerals and glioma: A case-control study based on Chinese population
## Abstract
### Background
As one of the essential nutrients for the human body, minerals participate in various physiological activities of the body and are closely related to many cancers. However, the population study on glioma is not sufficient.
### Objective
The purpose of this study was to evaluate the relationship between five dietary minerals and glioma.
### Methods
A total of 506 adult patients with glioma and 506 healthy controls were matched 1:1 according to age (±5 years) and sex. The food intake of the subjects in the past year was collected through the food frequency questionnaire, and the intakes of calcium, magnesium, iron, zinc, and copper in the diet were calculated. The logistic regression model was used to estimate the odds ratio (OR) and $95\%$ confidence interval ($95\%$ CI) for dietary minerals to gliomas.
### Results
After adjusting for confounders, higher intakes of calcium (OR = 0.65, $95\%$ CI: 0.57–0.74), magnesium (OR = 0.18, $95\%$ CI: 0.11–0.29), iron (OR = 0.04, $95\%$ CI: 0.02–0.11), zinc (OR = 0.62, $95\%$ CI: 0.54–0.73), and copper (OR = 0.22, $95\%$ CI: 0.13–0.39) were associated with a significantly decreased risk of glioma. Similar results were observed in gliomas of different pathological types and pathological grades. The restriction cubic spline function suggested significant linear dose-response relationships between intakes of five minerals and the risk of glioma. When the dietary minerals exceeded a particular intake, the risk of glioma stabilized.
### Conclusion
Our study suggests that higher dietary intakes of calcium, magnesium, iron, zinc, and copper are associated with a decreased risk of glioma. However, the results of this study require further exploration of potential mechanisms in the future better to elucidate the effects of mineral intake on gliomas.
## Introduction
Gliomas are the most common type of primary central nervous system tumor, accounting for about $80.7\%$ of malignant brain tumors [1]. Although the incidence of glioma is very low, only $\frac{5.47}{100}$,000 [1], due to the high mortality rate of glioma and the large population base in China, it caused serious disease burden and economic burden to families. Therefore, exploring modifiable factors in the etiology of glioma was essential to provide substantial scientific support for primary prevention.
Compared with other cancers, the etiology of gliomas was largely uncertain and complex. Although gliomas have long been reported to be related to head trauma [2], allergies [3], use of mobile phones [4], and occupational exposure [5]. Ionizing radiation was still the only apparent environmental risk factor [6]. In recent years, the relationship between diet and glioma has attracted more and more attention [7]. Existing studies have found that dietary patterns [8], food groups [9], and nutrients [10] all had certain effects against gliomas. In particular, the vitamins and phytochemicals in these foods had certain antioxidant effects, protecting healthy tissues from oxidative stress-induced damage and inhibiting the occurrence and development of glioma (7, 11–13). However, previous studies on diet and glioma ignored minerals with similar effects. These elements also played an antioxidant [14], anti-inflammatory [15], and anti-tumor [16] effect on the body. Therefore, several common minerals, such as calcium (Ca), iron (Fe), and zinc (Zn), may also affect gliomas. Yekta et al. found a significant negative association between dietary Ca intake and glioma (OR = 0.23, $95\%$ CI: 0.08–0.65) in an Iranian hospital-based case-control study [17]. This association was also found in the San Francisco Bay Area Adult Glioma Study but was only significant in women [18]. Chen et al. also found that the consumption of intracellular Ca ions affected the abnormal growth of C6 glioma cells by affecting the signal transduction of the endoplasmic reticulum [19]. Studies on Fe and gliomas had similar findings. Ward et al. followed up for 14.1 years in a European prospective cohort study to explore the effect of meat and heme Fe on gliomas but found no association between them [Hazard ratio (HR) = 0.96, $95\%$ CI: 0.73–1.26] [20]. However, in vivo exposure, a higher concentration of toenail Fe was found to have a protective effect against gliomas (OR = 0.42, $95\%$ CI: 0.19–0.95) [21]. Although cell experiments found that low Zn can inhibit the cell growth of rat glioma C6 cells [22], Dimitropoulou et al. did not find any significant association between dietary Zn and glioma in the nutritional epidemiological study [23]. The influence of magnesium (Mg) and copper (Cu) on glioma was far less than that of other elements, and their related studies mainly focused on metalloproteins. Bioinformatics studies have found differential expressions of various copper-related proteins in gliomas and normal tissues [24]. The overexpression of Mg transporter 1 was also associated with the occurrence and progression of gliomas [25].
Although experimental data on the role of minerals in the prevention of glioma were promising, the vast majority of studies have focused on in vitro assays. Epidemiological studies on minerals and gliomas were insufficient. On the one hand, no studies have reported their dose-response relationship. On the other hand, relevant studies were mainly based on the population of European and American populations, with geographical limitations. Therefore, to further explore the association between dietary minerals and glioma, we investigated the association of five typical dietary mineral intakes with the risk of glioma in the case-control study based on a Chinese population and attempted to delineate the dose-response relationship between the two aim to provide some epidemiological evidence for the prevention of glioma by five minerals.
## Study population
One thousand and twelve subjects (506 cases and 506 controls) participated in the diet and glioma case-control study at Beijing Tiantan Hospital, Capital Medical University, between 2021 and 2022. The case group consisted of patients with glioma who were recently diagnosed by neuro-oncologists and pathologists according to the 2021 neuro-tumor diagnostic criteria [26]. Among them, there were 104 cases of astrocytoma, 67 cases of oligodendroglioma, 237 cases of glioblastoma, and 98 cases of other gliomas (including 18 cases of diffuse midline glioma). The control groups were recruited from healthy individuals in the community and matched 1:1 with cases by age and sex. All participants were ≥18 years of age. Among them, 15 people in the case group refused to participate, with a response rate of $97.2\%$, and 41 people in the control group refused to participate, with a response rate of $92.7\%$ (Supplementary Figure 1). On this basis, they were excluded according to certain conditions, including suffering from digestive, neurological, and endocrine system diseases, suffering from other cancers, significant dietary behavior changes such as dieting before the investigation, abnormal energy intake (>5,000 or <400 kcal/d), pregnant women and nursing mothers, and taking drugs such as hormones. All participants provided informed consent, and the study protocol was approved by the Institutional Review Board of Beijing Tiantan Hospital, Capital Medical University (No. KY2022-203-02).
## Data collection
The required information was obtained through questionnaires. Through face-to-face interviews, investigators collected data on demographics, lifestyle habits, disease history, and dietary intake and measured some anthropometric indicators. Demographic data included date of birth, sex, occupation, education level, and household income. Lifestyle habits included living conditions in high-risk areas, smoking status, drinking status, and physical activity. Living near electromagnetic fields and broadcast antennas in the past 10 years have been defined as high-risk residential areas [4]. Physical activity was assessed using the International Physical Activity Questionnaire [27]. Disease histories were collected for diseases potentially associated with glioma, including allergies, head trauma, and other cancers.
Anthropometric data mainly consisted of height and weight, which were collected by trained staff using standardized techniques and calibrated equipment. Body mass index (BMI) was calculated by dividing weight (in kilograms) by the square of height (in meters), and the result was accurate to two decimal places.
Dietary intakes were assessed through the 114-item food frequency questionnaire (Supplementary Table 1). The questionnaire has been validated in previous studies [28]. According to the foods reported in the literature that may affect the risk of glioma, several foods were added and deleted on this basis to make the food frequency questionnaire more suitable for the research needs. To improve the accuracy of the dietary survey, the investigators collected the dietary intake information of the study subjects in the past year through face-to-face interviews by providing pictures of different food volumes and qualities. The study subjects need to fill in the intake of each food according to their conditions, including whether the food was consumed, the frequency of intake (number of intakes per day/week/month), and the average intake per time. In order to further verify the reproducibility and validity of the questionnaire in this study, after about 1 year, we investigated 30 healthy controls again, collected the dietary information of the subjects through the food frequency questionnaire and 24-h recall (two working days and one rest day), calculated the food consumption and nutrient intake, and evaluated the reproducibility and validity of the questionnaire by the mean and correlation coefficient. For reproducibility, the correlation coefficients of food group were 0.502–0.847, and that of nutrients were 0.437–0.807. For validity, the correlation coefficients of food group were 0.381–0.779, and the correlation coefficients of nutrients were 0.380–0.804 (Supplementary Tables 2–5).
The study involved five common minerals, including Ca, Mg, Fe, Zn, and Cu. The intakes of all minerals were calculated based on the information of each food item and the Chinese Food Composition Table [29]. The daily intakes of various foods were calculated according to the frequency of food intake and the amount of each intake filled in by the study subjects. The “Chinese Food Composition Table” provided the content of five minerals per unit of food. It was calculated by multiplying the daily intake of various foods and the unit content of minerals in the food. Then the sum of the intakes of all minerals in different foods was calculated as the total intake. Energy intake polyunsaturated fatty acids (PUFAs) were calculated similarly.
## Statistical analysis
Demographics, lifestyle habits, and disease history were characterized using descriptive analysis. The t-test was used for normally distributed continuous variables, and the chi-square test was used for categorical variables to compare general characteristics between pathological subtypes and controls. The Mann-Whitney U-test was used to compare mineral intakes between case and control groups, and Spearman's correlation coefficient was used to evaluate the correlation between the five minerals. We used logistic regression models to estimate ORs and $95\%$ CI between mineral intake and the risk of glioma, adjusting for potential confounders. In this analysis, each mineral intake was divided into tertiles, with the lowest tertile as the reference group. In addition, mineral intake was also brought into the model as a continuous variable.
Potential confounding variables included age, BMI, occupation (manual workers, mental workers, or others), educational levels (primary school and below, middle school, or university and above), household income (below 3,000 ¥/month, 3,000–10,000 ¥/month, or above 10,000 ¥/month), high-risk residential areas (yes or no), smoking status (never smoking, former smoking, or current smoking), drinking status (non-drinker, occasional drinker, or frequent drinker), history of allergies (yes or no), history of head trauma (yes or no), family history of cancer (yes or no), physical activity (low, moderate, or violent), PUFAs intake, and energy intake.
Age, sex, BMI, occupation, education level, household income, smoking status, history of allergies, family history of cancer, and physical activity were also used as the basis for subgrouping, and subgroup analyses were performed by logistic regression after adjusting for confounding factors. In addition, to overcome the inherent limitations of elemental analysis as a grade variable, the dose-response relationship was analyzed using the restricted cubic spline function in the logistic regression model after adjusting for confounders, with nodes distributed in the 20th, 40th, 60th, and 80th percentiles, the reference value (OR = 1) was set at the 10th percentile [30].
All statistical analyses were performed using SPSS 26.0 and R 4.1.1. A two-sided P-value < 0.05 was used to determine the statistical significance.
## Study population and mineral characteristics
The patients with glioma of different pathological subtypes and their corresponding control groups were completely identical in sex composition and similar in age distribution. Compared with controls, glioma patients differed in BMI ($P \leq 0.001$), occupation ($$P \leq 0.024$$), education levels ($P \leq 0.001$), household income ($P \leq 0.001$), smoking status ($$P \leq 0.039$$), drinking status ($P \leq 0.001$), physical activity ($P \leq 0.001$), history of allergies ($P \leq 0.001$), and family history of cancer ($$P \leq 0.001$$). Patients with various pathological subtypes of glioma had higher BMI, slightly lower education levels, lower household income, more drinkers, and more physical activity, which was consistent with the overall population. In addition, only the population with glioblastoma had a higher family history of cancer ($$P \leq 0.006$$). Other glioma populations had more manual workers ($$P \leq 0.002$$) and a lower history of allergies ($$P \leq 0.004$$). In other respects, there were no significant differences (Table 1).
**Table 1**
| Unnamed: 0 | Astrocytoma | Astrocytoma.1 | Unnamed: 3 | Oligodendroglioma | Oligodendroglioma.1 | Unnamed: 6 | Glioblastoma | Glioblastoma.1 | Unnamed: 9 | Others | Others.1 | Unnamed: 12 | P b |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Case | Control | P a | Case | Control | P a | Case | Control | P a | Case | Control | P a | |
| Age (years) | 39.32 ± 13.43 | 38.01 ± 13.01 | 0.477 | 39.52 ± 9.75 | 37.66 ± 9.67 | 0.268 | 45.27 ± 13.07 | 43.75 ± 12.89 | 0.202 | 41.83 ± 13.58 | 40.58 ± 13.27 | 0.517 | 0.072 |
| Sex (%) | | | 1.000 | | | 1.000 | | | 1.000 | | | 1.000 | 1.000 |
| Male | 58.7 | 58.7 | | 61.2 | 61.2 | | 55.3 | 55.3 | | 52.0 | 52.0 | | |
| Female | 41.3 | 41.3 | | 38.8 | 38.8 | | 44.7 | 44.7 | | 48.0 | 48.0 | | |
| BMI | 24.02 ± 3.05 | 22.89 ± 3.34 | 0.012 | 24.37 ± 2.91 | 23.27 ± 3.64 | 0.056 | 23.99 ± 3.30 | 23.13 ± 3.12 | 0.004 | 23.90 ± 3.56 | 22.89 ± 3.33 | 0.042 | <0.001 |
| High-risk residential area (%) | | | 0.222 | | | 0.662 | | | 0.729 | | | 0.319 | 0.534 |
| Yes | 23.1 | 16.3 | | 17.9 | 20.9 | | 19.0 | 20.3 | | 27.6 | 21.4 | | |
| No | 76.9 | 83.7 | | 82.1 | 79.1 | | 81.0 | 79.7 | | 72.4 | 78.6 | | |
| Occupation (%) | | | 0.406 | | | 0.931 | | | 0.119 | | | 0.002 | 0.024 |
| Manual workers | 23.1 | 17.3 | | 32.8 | 31.4 | | 22.4 | 21.1 | | 35.7 | 14.3 | | |
| Mental workers | 61.5 | 61.5 | | 55.2 | 58.2 | | 51.0 | 59.5 | | 43.9 | 63.3 | | |
| Others | 15.4 | 21.2 | | 12.0 | 10.4 | | 26.6 | 19.4 | | 20.4 | 22.4 | | |
| Education level (%) | | | 0.013 | | | 0.009 | | | <0.001 | | | <0.001 | <0.001 |
| Primary school and below | 2.9 | 4.8 | | 6.0 | 0 | | 6.3 | 2.5 | | 1.3 | 2.0 | | |
| Middle school | 43.3 | 24.0 | | 40.3 | 23.9 | | 41.8 | 26.2 | | 39.8 | 24.5 | | |
| University and above | 53.8 | 71.2 | | 53.7 | 76.1 | | 51.9 | 71.3 | | 46.9 | 73.5 | | |
| Household income (%) | | | 0.007 | | | <0.001 | | | <0.001 | | | 0.024 | <0.001 |
| <3,000 ¥/month | 11.5 | 18.3 | | 10.4 | 19.4 | | 7.2 | 19.8 | | 13.3 | 13.3 | | |
| 3000–10,000 ¥/month | 72.1 | 51.0 | | 83.6 | 47.8 | | 77.6 | 46.8 | | 70.4 | 54.0 | | |
| >10,000 ¥/month | 16.3 | 30.8 | | 6.0 | 32.8 | | 15.2 | 33.4 | | 16.3 | 32.7 | | |
| Smoking status (%) | | | 0.615 | | | 0.405 | | | 0.382 | | | 0.079 | 0.039 |
| Never smoking | 70.2 | 73.1 | | 64.2 | 74.6 | | 73.0 | 74.7 | | 66.3 | 79.6 | | |
| Former smoking | 10.6 | 6.7 | | 16.4 | 10.4 | | 13.1 | 9.3 | | 12.3 | 5.1 | | |
| Current smoking | 19.2 | 20.2 | | 19.4 | 15.0 | | 13.9 | 16.0 | | 21.4 | 15.3 | | |
| Drinking status (%) | | | 0.010 | | | 0.003 | | | <0.001 | | | 0.003 | <0.001 |
| Non-drinker | 61.5 | 58.7 | | 59.7 | 58.3 | | 64.6 | 55.3 | | 68.4 | 55.1 | | |
| Occasional drinker | 10.6 | 25.0 | | 11.9 | 31.3 | | 13.9 | 30.8 | | 13.2 | 33.7 | | |
| Frequent drinker | 27.9 | 16.3 | | 28.4 | 10.4 | | 21.5 | 13.9 | | 18.4 | 11.2 | | |
| History of allergies (%) | | | 0.122 | | | 0.171 | | | 0.167 | | | 0.004 | <0.001 |
| Yes | 7.7 | 14.4 | | 7.5 | 14.9 | | 8.0 | 11.8 | | 7.1 | 21.4 | | |
| No | 92.3 | 85.6 | | 92.5 | 85.1 | | 92.0 | 88.2 | | 92.9 | 78.6 | | |
| History of head trauma (%) | | | 0.675 | | | 0.069 | | | 0.451 | | | 0.345 | 0.474 |
| Yes | 13.5 | 11.5 | | 13.4 | 4.5 | | 9.3 | 11.4 | | 12.2 | 8.2 | | |
| No | 86.5 | 88.5 | | 86.6 | 95.5 | | 90.7 | 88.6 | | 87.8 | 91.8 | | |
| Family history of cancer (%) | | | 0.080 | | | 0.395 | | | 0.006 | | | 0.616 | 0.001 |
| Yes | 30.8 | 20.2 | | 23.9 | 17.9 | | 33.3 | 21.9 | | 25.5 | 22.4 | | |
| No | 69.2 | 79.8 | | 76.1 | 82.1 | | 66.7 | 78.1 | | 74.5 | 77.6 | | |
| Physical activity, (%) | | | 0.003 | | | <0.001 | | | <0.001 | | | <0.001 | <0.001 |
| Low | 16.3 | 35.6 | | 13.4 | 50.7 | | 14.3 | 47.7 | | 9.2 | 49.0 | | |
| Moderate | 40.4 | 37.5 | | 44.8 | 40.3 | | 41.8 | 35.4 | | 38.8 | 34.7 | | |
| Violent | 43.3 | 26.9 | | 41.8 | 9.0 | | 43.9 | 16.9 | | 52.0 | 16.3 | | |
In terms of dietary intakes, compared with controls, cases had higher intakes of refined grains ($P \leq 0.001$) and alcohol ($P \leq 0.001$), and lower intakes of whole grains ($P \leq 0.001$), legume and products ($P \leq 0.001$), tubers ($P \leq 0.001$), vegetables ($P \leq 0.001$), fungi and algae ($P \leq 0.001$), fruits ($P \leq 0.001$), fish and seafood ($P \leq 0.001$), and dairy products ($$P \leq 0.025$$). For other food groups, there was no significant difference (Supplementary Table 6).
In terms of mineral intakes, as shown in Table 2, the intakes of Ca, Mg, Fe, Zn, and Cu in the control group were all significantly higher than those in the case group. In addition, there were significant correlations between individual mineral intakes (Spearman coefficients ranged from 0.709 to 0.919) (Supplementary Table 7).
**Table 2**
| Minerals | Unnamed: 1 | Q1 | Q2 | Q3 | Q4 | P-value |
| --- | --- | --- | --- | --- | --- | --- |
| Ca (mg/d) | Case | 233.07 ± 61.02 | 411.04 ± 52.37 | 573.68 ± 51.32 | 941.45 ± 290.89 | <0.001 |
| Ca (mg/d) | Control | 203.90 ± 79.60 | 406.71 ± 51.36 | 592.04 ± 55.34 | 1,028.21 ± 358.82 | |
| Mg (mg/d) | Case | 161.51 ± 35.32 | 241.55 ± 19.45 | 318.91 ± 31.33 | 502.50 ± 140.23 | <0.001 |
| Mg (mg/d) | Control | 145.72 ± 43.22 | 241.06 ± 20.13 | 331.08 ± 31.10 | 529.23 ± 135.73 | |
| Fe (mg/d) | Case | 9.30 ± 1.81 | 13.29 ± 1.11 | 17.27 ± 1.37 | 27.01 ± 8.74 | <0.001 |
| Fe (mg/d) | Control | 8.24 ± 2.45 | 13.11 ± 1.04 | 17.40 ± 1.35 | 27.00 ± 7.37 | |
| Zn (mg/d) | Case | 5.09 ± 1.09 | 7.39 ± 0.65 | 9.79 ± 0.68 | 14.58 ± 4.23 | 0.002 |
| Zn (mg/d) | Control | 4.50 ± 1.33 | 7.54 ± 0.66 | 9.84 ± 0.72 | 15.07 ± 3.83 | |
| Cu (mg/d) | Case | 0.76 ± 0.17 | 1.14 ± 0.11 | 1.57 ± 0.17 | 2.77 ± 0.91 | <0.001 |
| Cu (mg/d) | Control | 0.66 ± 0.21 | 1.15 ± 0.11 | 1.62 ± 0.15 | 2.80 ± 0.87 | |
## Association between dietary minerals and glioma
The results of the association between minerals with glioma are shown in Table 3. After adjustment for confounding variables (Model 2), the results for the mineral categorical variable showed that individuals with the highest Ca intake was associated with a $90\%$ decreased risk of glioma compared with the first tertile (OR = 0.11, $95\%$ CI: 0.05–0.25), individuals with the highest Mg intake was associated with a $95\%$ decreased risk of glioma (OR = 0.06, $95\%$ CI: 0.02–0.16), and individuals with the highest Fe intake was associated with a $93\%$ decreased risk of glioma (OR = 0.07, $95\%$ CI: 0.03–0.17), and individuals with the highest Zn intake was associated with an $89\%$ decreased risk of glioma (OR = 0.07, $95\%$ CI: 0.02–0.18), and individuals with the highest Cu intake was associated with an $87\%$ decreased risk of glioma (OR = 0.09, $95\%$ CI: 0.04–0.22).
**Table 3**
| Unnamed: 0 | T1 | T2 | T3 | Continuousc | P−trend |
| --- | --- | --- | --- | --- | --- |
| Ca | ≤381.39 | 381.39–611.85 | >611.85 | | |
| Case/control | 211/127 | 187/157 | 108/222 | | |
| Model 1a | 1 | 0.70 (0.51–0.97) | 0.27 (0.19–0.39) | 0.84 (0.80–0.88) | <0.001 |
| Model 2b | 1 | 0.46 (0.23–0.90) | 0.11 (0.05–0.25) | 0.65 (0.57–0.74) | <0.001 |
| Mg | ≤229.28 | 229.28–341.01 | >341.01 | | |
| Case/control | 191/147 | 197/140 | 118/219 | | |
| Model 1a | 1 | 1.01 (0.74–1.38) | 0.41 (0.30–0.57) | 0.77 (0.70–0.84) | <0.001 |
| Model 2b | 1 | 0.45 (0.22–0.90) | 0.06 (0.02–0.16) | 0.18 (0.11–0.29) | <0.001 |
| Fe | ≤12.56 | 12.56–18.05 | >18.05 | | |
| Case/control | 184/154 | 200/137 | 122/215 | | |
| Model 1a | 1 | 1.19 (0.87–1.63) | 0.50 (0.37–0.68) | 0.67 (0.57–0.80) | <0.001 |
| Model 2b | 1 | 0.35 (0.18–0.71) | 0.07 (0.03–0.17) | 0.04 (0.02–0.11) | <0.001 |
| Zn | ≤7.06 | 7.06–10.14 | >10.14 | | |
| Case/control | 191/147 | 176/161 | 139/198 | | |
| Model 1a | 1 | 0.82 (0.61–1.11) | 0.53 (0.39–0.73) | 0.95 (0.93–0.98) | <0.001 |
| Model 2b | 1 | 0.32 (0.16–0.64) | 0.07 (0.02–0.18) | 0.62 (0.54–0.73) | <0.001 |
| Cu | <1.08 | 1.08–1.67 | >1.67 | | |
| Case/control | 194/144 | 179/158 | 133/204 | | |
| Model 1a | 1 | 0.83 (0.61–1.15) | 0.49 (0.36–0.67) | 0.76 (0.65–0.88) | <0.001 |
| Model 2b | 1 | 0.36 (0.18–0.71) | 0.09 (0.04–0.22) | 0.22 (0.13–0.39) | <0.001 |
The results of the analysis of the continuous variables showed that for each 100 mg/d increase in Ca intake, the risk of glioma decreased by $35\%$ (OR = 0.65, $95\%$ CI: 0.57–0.74), and for each 100 mg/d increase in Mg intake, the risk of glioma decreased by $82\%$ (OR = 0.18, $95\%$ CI: 0.11–0.29), and for each 10 mg/d increase in Fe intake, the risk of glioma decreased by $96\%$ (OR = 0.04, $95\%$ CI: 0.02–0.11), and for each 1 mg/d increase in Zn intake, the risk of glioma decreased by $38\%$ (OR = 0.62, $95\%$ CI: 0.54–0.73), and for each 1 mg/d increase in Zn intake, the risk of glioma decreased by $78\%$ (OR = 0.22, $95\%$ CI: 0.13–0.39).
## Minerals and pathological classification and grade of glioma
The analysis of pathological classifications of glioma showed that all five minerals were associated with decreased significantly risks of glioblastoma. The results were consistent with those of the overall population of gliomas. But for astrocytoma, the results of Fe and Cu were significant. Due to the small sample size of oligodendroglioma, no further analysis was carried out (Table 4).
**Table 4**
| Pathological classificationc | Model 1a | P-value | Model 2b | P-value.1 |
| --- | --- | --- | --- | --- |
| Astrocytoma | Astrocytoma | Astrocytoma | Astrocytoma | Astrocytoma |
| Ca | 0.83 (0.75–0.92) | <0.001 | 0.01 (0.001–1.47) | 0.072 |
| Mg | 0.76 (0.63–0.92) | 0.004 | 0.002 (0.001–2.46) | 0.086 |
| Fe | 0.62 (0.43–0.89) | 0.011 | –* | 0.014 |
| Zn | 0.95 (0.89–1.01) | 0.070 | 0.15 (0.01–1.67) | 0.124 |
| Cu | 0.71 (0.52–0.98) | 0.037 | 0.02 (0.001–0.43) | 0.014 |
| Glioblastoma | Glioblastoma | Glioblastoma | Glioblastoma | Glioblastoma |
| Ca | 0.85 (0.80–0.92) | <0.001 | 0.69 (0.53–0.89) | 0.004 |
| Mg | 0.75 (0.65–0.87) | <0.001 | 0.14 (0.05–0.41) | <0.001 |
| Fe | 0.72 (0.57–0.92) | 0.009 | 0.19 (0.05–0.64) | 0.008 |
| Zn | 0.95 (0.91–0.99) | 0.034 | 0.69 (0.53–0.88) | 0.003 |
| Cu | 0.74 (0.59–0.92) | 0.008 | 0.12 (0.03–0.47) | 0.002 |
The results of minerals and different grades of gliomas showed that Mg, and Zn significantly were associated with a significantly decreased risk of low-grade gliomas. In contrast, the results of Ca, Fe, and Cu were not statistically significant. For high-grade gliomas, Ca, Mg, Fe, Zn, and Cu were associated with a significantly decreased risk (Table 5).
**Table 5**
| Glioma gradingc | Model 1a | P-value | Model 2b | P-value.1 |
| --- | --- | --- | --- | --- |
| Low grade | Low grade | Low grade | Low grade | Low grade |
| Ca | 0.80 (0.70–0.90) | <0.001 | 0.02 (0.001–2.44) | 0.111 |
| Mg | 0.77 (0.62–0.95) | 0.014 | 0.01 (0.001–0.22) | 0.006 |
| Fe | 0.58 (0.39–0.87) | 0.008 | –* | 0.085 |
| Zn | 0.94 (0.88–1.01) | 0.080 | 0.30 (0.12–0.77) | 0.012 |
| Cu | 0.84 (0.61–1.14) | 0.254 | 0.53 (0.13–2.22) | 0.386 |
| High grade | High grade | High grade | High grade | High grade |
| Ca | 0.86 (0.81–0.91) | <0.001 | 0.67 (0.55–0.81) | <0.001 |
| Mg | 0.78 (0.70–0.88) | <0.001 | 0.13 (0.06–0.30) | <0.001 |
| Fe | 0.74 (0.60–0.91) | 0.003 | 0.09 (0.03–0.30) | <0.001 |
| Zn | 0.96 (0.93–0.99) | 0.037 | 0.67 (0.55–0.82) | <0.001 |
| Cu | 0.79 (0.66–0.94) | 0.009 | 0.12 (0.04–0.32) | <0.001 |
## Subgroup analysis
In the subgroup analysis by age, sex, BMI, occupation, education level, household income, smoking status, history of allergies, family history of cancer, and physical activity, we observed that most of the results in the subgroup analysis were consistent with the main results. Very few subgroups had no significant results due to the small sample size (Supplementary Table 8).
## Dose-response relationship
In Figure 1, we used restricted cubic splines to describe the relationship between minerals and the risk of glioma. There were linear dose-response relationships between the intakes of five minerals and the risk of glioma. For Ca, when the intake exceeded 398.02 mg/d, the risk of glioma decreased significantly with the increase in intake. When the intake exceeded 870.29 mg/d, the risk of glioma was relatively stable (P−nonlinearity = 0.6182). For Mg, when the intake exceeded 151.29 mg/d, the risk of glioma decreased significantly with the increase in intake. When the intake exceeded 310.40 mg/d, the risk of glioma was relatively stable (P−nonlinearity = 0.5374). For Fe, when the intake exceeded 8.80 mg/d, the risk of glioma decreased significantly with the increase in intake. When the intake exceeded 17.55 mg/d, the risk of glioma was relatively stable (P−nonlinearity = 0.0974). For Zn, when the intake exceeded 7.46 mg/d, the risk of glioma decreased significantly with the increase in intake. When the intake exceeded 12.55 mg/d, the risk of glioma was relatively stable (P−nonlinearity = 0.2470). For Cu, when the intake exceeded 1.28 mg/d, the risk of glioma decreased significantly with the increase in intake. When the intake exceeded 2.65 mg/d, the risk of glioma was relatively stable (P−nonlinearity = 0.0636).
**Figure 1:** *The restricted cubic spline for the associations between dietary minerals and glioma. The lines represent adjusted odds ratios based on restricted cubic splines for the intake in the regression model. Knots were placed at the 20th, 40th, 60th, and 80th percentiles of the dietary minerals intake, and the reference value was set at the 10th percentile. The adjusted factors were the same as in Model 2.*
## Discussion
Our study assessed the relationships between five common minerals and gliomas in the Chinese population. The results showed that the intakes of Ca, Mg, Fe, Zn, and Cu were significantly negatively associated with the risk of glioma. Similar results were observed in several subgroups, indicating that the association was relatively robust, especially in different pathological subtypes of gliomas and different grades of gliomas for the first time. It showed that this association was unlikely to be confused between different glioma subtypes. The restricted cubic spline model further confirmed a significant linear dose-response relationship between the five minerals and the risk of glioma, and with the increase in intake, the risk of glioma tended to be stable.
Ca is the most abundant mineral element in the human body, of which about $99\%$ is concentrated in bones and teeth and plays a crucial role in bone mineralization and a wide range of biological functions [31]. As an essential element for the human body, people can only get it from Ca-rich food sources, including milk and soybeans [31]. Based on the physiological effects of Ca, studies showed that it was closely related to osteoporosis and cardiovascular disease, and cancer was no exception [32]. In contrast, studies on dietary Ca and glioma were rare. Yekta et al. found a significant negative association between dietary Ca intakes in 128 glioma patients and 256 healthy individuals in a case-control study based on an Iranian hospital (OR = 0.23, $95\%$ CI: 0.08–0.65) [17]. Due to differences in dietary Ca sources between the Chinese population and the Middle East, although Ca intake in this study was higher than ours, our study found similar results. Higher dietary Ca intake was associated with a significantly decreased risk of glioma (OR = 0.11, $95\%$ CI: 0.05–0.25) with a significant linear-dose-response relationship, which complemented the evidence for low Ca intake. For glioma subtypes, dietary Ca also had the same protective effect against glioblastoma (OR = 0.69, $95\%$ CI: 0.53–0.89). This was similar to earlier results from the San Francisco Bay Area Adult Glioma Study. Tedeschi-Blok et al. compared dietary Ca intakes in 337 astrocytoma patients and 450 controls and found that dietary Ca intake was inversely associated with astrocytomas in a female-only population. Ca in our study population was inversely associated with astrocytoma, and the Ca intake of this population was closer to our study [18]. In addition, a meta-analysis of the dose-response relationship showed that each 100 mg/d increase in Ca intake reduced the risk of glioma by $7\%$ (OR = 0.93, $95\%$ CI: 0.88–0.98), which was similar to our results. However, this meta-analysis included only four studies with high heterogeneity, and the results still needed to be further confirmed in the future [33]. Most of its mechanisms were currently considered to be related to the regulation of parathyroid hormone. Increased Ca levels in the body can reduce the release of parathyroid hormone [34], which was thought to play a promoting role in the development of cancer [35, 36]. In gliomas, parathyroid hormone-related proteins were also found to regulate the transcriptional activation of glioma-related oncogenes [37], and immunohistochemical results showed that parathyroid hormone-related proteins were present in astrocytomas, suggesting that parathyroid hormone-related proteins may be related to the imbalance of growth or differentiation of astrocytoma cells [38]. In addition, intracellular Ca and Ca signal pathways were closely related to gliomas [19]. Elevated intracellular Ca2+ can activate nitric oxide synthase to generate nitric oxide, which impacted tumorigenesis [39], but it was difficult to directly link dietary Ca intake with Ca signaling channels.
Fe is involved in various metabolic processes in the body, including oxygen transport, DNA synthesis, and electron transport, and is an essential element in almost all organisms [40]. There are two main dietary Fe forms—heme Fe and non-heme Fe [41]. Among them, heme Fe mainly comes from meat. In the western diet, heme Fe accounted for $10\%$ of the total dietary Fe, but because the body more easily absorbed it, it accounted for nearly $\frac{2}{3}$ of Fe absorption [42]. Non-heme Fe exists mainly in plants. Nutritional disorders caused by Fe deficiency, such as anemia, infection, liver disease, and nervous system disease, have become public health issues of great concern [43]. There have been many reports of dietary Fe and cancer in recent years. But studies on Fe and glioma were rare. Parent et al. found a non-significant association between Fe in occupational exposure and glioma in a population-based multicenter case-control study (OR = 1.10, $95\%$ CI: 0.80–1.50). Still, the exposure route in this study was mainly the respiratory system [44]. Ward et al. followed up for 14.1 years in the European Prospective Investigation into Cancer and Nutrition and found that total dietary Fe (HR = 0.94, $95\%$ CI: 0.71–1.24) and heme Fe (HR = 0.96, $95\%$ CI: 0.73–1.26) were not associated with the risk of glioma [20]. This was not consistent with our results. We found that higher dietary Fe intake had a protective effect against gliomas (OR = 0.07, $95\%$ CI: 0.03–0.17), and similar results were observed in astrocytomas and glioblastomas as well as in high-grade gliomas. In addition, because Fe was not only an essential element of the body but also potentially toxic to cells, it was significant to describe its dose-response relationship. We found a linear dose-response relationship between dietary Fe and glioma, and when the intake exceeded 17.55 mg/d, the risk of glioma did not change (P−nonlinearity = 0.0974). Although dietary Fe has not shown a protective effect against gliomas in previous studies, studies on internal exposure seemed to support our results. Anic et al. determined the toenail Fe concentration by neutron activation analysis and found a significant negative association between toenail Fe and the risk of glioma (OR = 0.42, $95\%$ CI: 0.19–0.95) [21]. The relationship and mechanism between Fe and glioma were relatively complex. Bioinformatics studies have found that Fe metabolism-related genes can be used as prognostic indicators of low-grade gliomas [45]. Some studies also found that Fe played an essential role in treating gliomas. In animal experiments, it was found that after intravenous injection of Fe complex into male nude mice with glioma, the tumor growth of nude mice was significantly inhibited after 3 weeks, and the possible mechanism was apoptosis [46]. Eales et al. also found that verteporfin can radically and selectively kill anoxic glioma cells by binding free Fe [47].
Zn is the second most abundant transition metal ion in organisms, second only to Fe, and is indispensable to the growth and development of plants, animals, and microorganisms [48, 49]. It can not only be used as a cofactor of more than 300 enzymes [50] but also play a key role in oxidative stress, immunity, and aging [49], and it has been reported that dietary Zn has protective effects against depression, type 2 diabetes and some cancers [51]. Dietary Zn reduced the risk of cancer, mainly in the digestive system [52, 53]. There have been few studies on Zn and gliomas. Dimitropoulou et al. observed only a slight protective effect of dietary Zn against meningioma (OR = 0.62, $95\%$ CI: 0.39–0.99) in the study of adult brain tumors in the UK, but the association with glioma was not significant (OR = 0.92, $95\%$ CI: 0.66–1.28) [23]. However, in our study, higher dietary Zn intake significantly reduced the risk of glioma (OR = 0.07, $95\%$ CI: 0.02–0.18), and there was a significant linear-dose response relationship. When the intake was 7.46–12.55 mg/d, the risk of glioma decreased with increased intake. The risk did not change beyond 12.55 mg/d. The mechanism of Zn involved in the development of glioma may be various. On the one hand, Zn, as a component of superoxide dismutase [54], had a strong antioxidant effect and played an essential role in oxidative stress and repairing DNA damage [49, 55]. Along with the depletion of Zn in the body, this can lead to DNA damage and the production of free radicals, which can lead to the formation of tumors [56, 57]. This was no exception in glioma [58]. On the other hand, appropriate Zn can induce apoptosis of glioma cells. Haase et al. found in C6 rat glioma cells that Zn can promote proliferation and growth at a concentration of 50–100 μM, but too low (<50 μM) or too high (>200 μM) can induce apoptosis, especially when it exceeded 300 μM, it seemed to cause the necrosis of glioma cells [59]. In addition, Zn may act as an epigenetic regulator of gliomas by promoting proper DNA folding, protecting genetic material from oxidative damage, and controlling the activation of enzymes involved in epigenetic regulation [60].
Mg is the fourth most abundant mineral in the body and the second most abundant intracellular divalent cation [61, 62]. Whole grains, green vegetables, and nuts are rich dietary sources of Mg, but the loss of Mg during cooking and processing partly explains Mg deficiency [63]. Because *Mg is* involved in many biological processes in the body, including energy production, glycolysis, oxidative phosphorylation, nucleic acid, and protein synthesis [62, 64], it was closely related to muscle health, asthma, cardiovascular disease, and mental illness [62]. But the impact on cancer was currently inconsistent. In animal models, low *Mg status* was found to have a dual effect against tumors—inhibiting primary tumor growth and promoting metastatic tumor engraftment [65]. However, epidemiological evidence still suggested that Mg deficiency may increase the risk of certain cancers. The results of an earlier meta-analysis showed that higher dietary Mg intake was associated with a significant reduction in overall cancer risk (RR = 0.80, $95\%$ CI: 0.66–0.97) [66]. Our study provided some evidence of the association between dietary Mg and gliomas, with higher dietary Mg having a protective effect against gliomas (OR = 0.06, $95\%$ CI: 0.02–0.16), especially in low-grade gliomas. Mg has shown some anti-inflammatory effects in preclinical and epidemiological studies. Both dietary Mg intake and serum Mg levels were associated with increased levels of low-grade systemic inflammation, pro-inflammatory factors, and inflammatory markers (62, 67–69). This may create a microenvironment that was conducive to tumor invasion and metastasis [70]. Since the presence of inflammatory cells and the release of inflammatory mediators also promoted glioma proliferation, angiogenesis, and attack, it seemed impossible to ignore the effect of Mg on gliomas [71]. In addition, it may also be related to Mg transporter 1, which was highly selective for Mg transport [72]. Recent studies have also found that the overexpression of Mg transporter 1 promoted the growth of glioma cells through the up-regulation of PD-L1 expression mediated by the ERK/MAPK signaling pathway [73], but whether dietary Mg was involved remains to be further explored.
Although *Cu is* also an essential trace element for the body, compared with the previous minerals, the demand for *Cu is* deficient, with only about 100 mg of Cu in the human body [74]. Animal offal, corn products, certain vegetables, and individual fruits are good sources of dietary Cu [74]. Since Cu was a cofactor for many oxidoreductases, it was involved in the body's antioxidant defense, neuropeptide synthesis, and immune function [75, 76]. It also played an important role in fetal development [77], cardiovascular disease, and cognitive function [74]. Studies on Cu and cancer were rare, and most studies have found no significant association between dietary Cu and some cancers. However, our study found that higher dietary Cu significantly reduced the risk of glioma (OR = 0.09, $95\%$ CI: 0.04–0.22). Still, its effect was not as significant as that of the other four minerals. Moreover, dietary Cu seemed to have a protective effect only on high-grade gliomas (OR = 0.12, $95\%$ CI: 0.04–0.32). No similar result was observed in low-grade gliomas (OR = 0.53, $95\%$ CI: 0.13–2.22). Since both Cu deficiency and Cu excess were harmful to health, some scholars proposed that the dose-response curve between Cu and health was U-shaped [78]. This was similar to the results of our study. Although there was a linear dose-response relationship between dietary Cu and glioma in this study, the risk leveled off when the intake exceeded a certain level. Due to the relatively narrow range of dietary Cu intake, we suspected that the current dose-response curve might be the left half of the U-shaped curve. In addition, some studies also found that Cu played a role in the treatment of glioma. In vitro studies, Trejo-Solis et al. found that Cu compounds induced autophagy and apoptosis of glioma cells by increasing the production of intracellular reactive oxygen species and the activity of c-junNH2-terminal kinase [79]. Castillo-Rodriguez et al. found anti-proliferation, pro-apoptosis, and anti-invasion effects of Cu coordination compounds on U373 human glioma cells and significantly reduced tumor volume, cell proliferation, and mitotic index in mice transplanted with U373 glioma cells, and apoptosis index was increased [80].
The limitation of this study was that we could not explore the association between different forms and valence minerals in diet and glioma. Since the food composition table only provided the total amount of minerals without differentiating their form and valence, different forms and valence of minerals may have different effects on the body, we cannot analyze the results of the form and valence of these minerals on glioma in detail. Secondly, we only evaluated dietary sources for the relationship between minerals and glioma, and we could not comprehensively assess other sources, such as air. However, the sources of minerals were mainly dietary, so the results of this study were still of certain significance. In addition, because the study was a case-control study, we could not verify the causal relationship between the two and avoid inherent bias. In order to reduce the impact of information bias on this study, all questionnaires were completed face to face by investigators with medical education background. These investigators can participate in the survey only after receiving unified training and strict assessment before conducting the survey. In addition, in order to improve the accuracy of the dietary survey, the investigators assisted participants in estimating the amount of food in detail through food picture flip books containing different food volumes and qualities. However, the study still had some advantages. First, we explored the association between five common dietary minerals and gliomas. The results were consistent with existing in-vitro studies, especially for Cu and Mg, which lacked clinical studies and explored the association between gliomas of different pathological subtypes and pathological grades and minerals. Moreover, this was the first time that these dose-response relationships between dietary mineral intake and the risk of glioma were described, and the significant linear dose-response relationships provided further population evidence for mineral prevention and treatment of glioma.
## Conclusion
In summary, we observe that higher intakes of Ca, Mg, Fe, Zn, and Cu were associated with a decreased risk of glioma. Therefore, we may not be able to ignore the influence of dietary minerals on glioma. In the future, further prospective studies should be conducted to verify their relationship.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by the Institutional Review Board of Beijing Tiantan Hospital, Capital Medical University (No. KY2022-203-02). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
WL and WZ contributed to the conception or design of the work and wrote the manuscript. WZ, YH, XK, CW, and FC contributed to data collection and analysis. ZK, SY, RZ, and YP contributed to data collection and management. All authors have read and approved the final manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1118997/full#supplementary-material
## References
1. Ostrom QT, Price M, Ryan K, Edelson J, Neff C, Cioffi G. **Cbtrus statistical report: pediatric brain tumor foundation childhood and adolescent primary brain and other central nervous system tumors diagnosed in the united states in 2014-2018**. *Neuro Oncol.* (2022) **24** i1-38. DOI: 10.1093/neuonc/noac161
2. Gurney JG, Preston-Martin S, McDaniel AM, Mueller BA, Holly EA. **Head injury as a risk factor for brain tumors in children: results from a multicenter case-control study**. *Epidemiology.* (1996) **7** 485-9. DOI: 10.1097/00001648-199609000-00006
3. Turner MC, Krewski D, Armstrong BK, Chetrit A, Giles GG, Hours M. **Allergy and brain tumors in the interphone study: pooled results from Australia, Canada, FRANCE, israel, and New zealand**. *Cancer Causes Control.* (2013) **24** 949-60. DOI: 10.1007/s10552-013-0171-7
4. Morgan LL, Miller AB, Sasco A, Davis DL. **Mobile phone radiation causes brain tumors and should be classified as a probable human carcinogen (2a) (review)**. *Int J Oncol.* (2015) **46** 1865-71. DOI: 10.3892/ijo.2015.2908
5. Krishnan G, Felini M, Carozza SE, Miike R, Chew T, Wrensch M. **Occupation and adult gliomas in the san francisco bay area**. *J Occup Environ Med.* (2003) **45** 639-47. DOI: 10.1097/01.jom.0000069245.06498.48
6. Mathews JD, Forsythe AV, Brady Z, Butler MW, Goergen SK, Byrnes GB. **Cancer risk in 680,000 people exposed to computed tomography scans in childhood or adolescence: data linkage study of 11 million australians**. *BMJ.* (2013) **346** f2360. DOI: 10.1136/bmj.f2360
7. Bielecka J, Markiewicz-Zukowska R. **The influence of nutritional and lifestyle factors on glioma incidence**. *Nutrients* (2020) 12. DOI: 10.3390/nu12061812
8. Mousavi SM, Shayanfar M, Rigi S, Mohammad-Shirazi M, Sharifi G, Esmaillzadeh A. **Adherence to the mediterranean dietary pattern in relation to glioma: a case-control study**. *Clin Nutr.* (2021) **40** 313-9. DOI: 10.1016/j.clnu.2020.05.022
9. Zhang W, Jiang J, Li X, He Y, Chen F, Li W. **Dietary factors and risk of glioma in adults: a systematic review and dose-response meta-analysis of observational studies**. *Front Nutr.* (2022) **9** 834258. DOI: 10.3389/fnut.2022.834258
10. Zhang W, Jiang J, He Y, Li X, Yin S, Chen F. **Association between vitamins and risk of brain tumors: a systematic review and dose-response meta-analysis of observational studies**. *Front Nutr.* (2022) **9** 935706. DOI: 10.3389/fnut.2022.935706
11. Priyadarsini RV, Nagini S. **Cancer chemoprevention by dietary phytochemicals: promises and pitfalls**. *Curr Pharm Biotechnol.* (2012) **13** 125-36. DOI: 10.2174/138920112798868610
12. Surh YJ. **Cancer chemoprevention with dietary phytochemicals**. *Nat Rev Cancer.* (2003) **3** 768-80. DOI: 10.1038/nrc1189
13. Dadgostar E, Fallah M, Izadfar F, Heidari-Soureshjani R, Aschner M, Tamtaji OR. **Therapeutic potential of resveratrol in the treatment of glioma: insights into its regulatory mechanisms**. *Mini Rev Med Chem.* (2021) **21** 2833-45. DOI: 10.2174/1389557521666210406164758
14. Torres-Arce E, Vizmanos B, Babio N, Marquez-Sandoval F, Salas-Huetos A. **Dietary antioxidants in the treatment of male infertility: counteracting oxidative stress**. *Biology.* (2021) 10. DOI: 10.3390/biology10030241
15. Rehou S, Shahrokhi S, Natanson R, Stanojcic M, Jeschke MG. **Antioxidant and trace element supplementation reduce the inflammatory response in critically ill burn patients**. *J Burn Care Res.* (2018) **39** 1-9. DOI: 10.1097/BCR.0000000000000607
16. Ilghami R, Barzegari A, Mashayekhi MR, Letourneur D, Crepin M, Pavon-Djavid G. **The conundrum of dietary antioxidants in cancer chemotherapy**. *Nutr Rev.* (2020) **78** 65-76. DOI: 10.1093/nutrit/nuz027
17. Yekta MF, Soltani S, Shayanfar M, Benisi-Kohansal S, Mohammad-Shirazi M, Sharifi G. **A case-control study on dietary calcium intake and risk of glioma**. *Eur J Cancer Prev.* (2021) **30** 322-7. DOI: 10.1097/CEJ.0000000000000629
18. Tedeschi-Blok N, Schwartzbaum J, Lee M, Miike R, Wrensch M. **Dietary calcium consumption and astrocytic glioma: the San Francisco bay area adult glioma study, 1991-1995**. *Nutr Cancer.* (2001) **39** 196-203. DOI: 10.1207/S15327914nc392_6
19. Chen YJ, Lin JK, Lin-Shiau SY. **Proliferation arrest and induction of CDK inhibitors p21 and p27 by depleting the calcium store in cultured c6 glioma cells**. *Eur J Cell Biol.* (1999) **78** 824-31. DOI: 10.1016/S0171-9335(99)80033-8
20. Ward HA, Gayle A, Jakszyn P, Merritt M, Melin B, Freisling H. **Meat and haem iron intake in relation to glioma in the European prospective investigation into cancer and nutrition study**. *Eur J Cancer Prev.* (2018) **27** 379-83. DOI: 10.1097/CEJ.0000000000000331
21. Anic GM, Madden MH, Thompson RC, Nabors LB, Olson JJ, LaRocca RV. **Toenail iron, genetic determinants of iron status, and the risk of glioma**. *Cancer Causes Control.* (2013) **24** 2051-8. DOI: 10.1007/s10552-013-0281-2
22. Ho E, Ames BN. **Low intracellular zinc induces oxidative DNA damage, disrupts p53, NF kappa B, and AP1 DNA binding, and affects DNA repair in a rat glioma cell line**. *Proc Natl Acad Sci USA.* (2002) **99** 16770-5. DOI: 10.1073/pnas.222679399
23. Dimitropoulou P, Nayee S, Liu JF, Demetriou L, van Tongeren M, Hepworth SJ. **Dietary zinc intake and brain cancer in adults: a case-control study**. *Br J Nutr.* (2008) **99** 667-73. DOI: 10.1017/S0007114507831692
24. Wang X, Han MZ, Chen SY, Sun YF, Tan RR, Huang B. **The copper-associated protein steap2 correlated with glioma prognosis and immune infiltration**. *Front Cell Neurosci* (2022) **16** 944682. DOI: 10.3389/fncel.2022.944682
25. Wang G, Li Y, Li J, Zhang DX, Luo C, Zhang BQ. **Microrna-199a-5p suppresses glioma progression by inhibiting magt1**. *J Cell Biochem.* (2019) **120** 15248-54. DOI: 10.1002/jcb.28791
26. Louis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D. **The 2021 WHO classification of tumors of the central nervous system: a summary**. *Neuro Oncol.* (2021) **23** 1231-51. DOI: 10.1093/neuonc/noab106
27. Craig CL, Marshall AL, Sjostrom M, Bauman AE, Booth ML, Ainsworth BE. **International physical activity questionnaire: 12-country reliability and validity**. *Med Sci Sports Exerc.* (2003) **35** 1381-95. DOI: 10.1249/01.MSS.0000078924.61453.FB
28. Zhao WH, Huang ZP, Zhang X, He L, Willett W, Wang JL. **Reproducibility and validity of a Chinese food frequency questionnaire**. *Biomed Environ Sci.* (2010) **23** 1-38. DOI: 10.1016/S0895-3988(11)60014-7
29. Yang YX. *China Food Composition Tables* (2018)
30. Zhang W, Du J, Li H, Yang Y, Cai C, Gao Q. **Multiple-element exposure and metabolic syndrome in Chinese adults: a case-control study based on the Beijing population health cohort**. *Environ Int.* (2020) **143** 105959. DOI: 10.1016/j.envint.2020.105959
31. Peacock M. **Calcium metabolism in health and disease**. *Clin J Am Soc Nephrol.* (2010) **5** S23-30. DOI: 10.2215/CJN.05910809
32. Peterlik M, Kallay E, Cross HS. **Calcium nutrition and extracellular calcium sensing: relevance for the pathogenesis of osteoporosis, cancer and cardiovascular diseases**. *Nutrients.* (2013) **5** 302-27. DOI: 10.3390/nu5010302
33. Guo X, Piao HZ. **A meta-analysis of calcium intake and risk of glioma**. *Nutr Cancer.* (2022) **74** 3194-201. DOI: 10.1080/01635581.2022.2067336
34. Goltzman D, Mannstadt M, Marcocci C. **Physiology of the calcium-parathyroid hormone-vitamin D axis**. *Vitamin D Clin Med* (2018) **50** 1-13. DOI: 10.1159/000486060
35. McCarty MF. **Parathyroid hormone may be a cancer promoter - an explanation for the decrease in cancer risk associated with ultraviolet light, calcium, and vitamin D**. *Med Hypotheses.* (2000) **54** 475-82. DOI: 10.1054/mehy.1999.0880
36. Martin TJ, Johnson RW. **Multiple actions of parathyroid hormone-related protein in breast cancer bone metastasis**. *Br J Pharmacol.* (2021) **178** 1923-35. DOI: 10.1111/bph.14709
37. Alman BA, Wunder JS. **Parathyroid hormone-related protein regulates glioma-associated oncogene transcriptional activation lessons learned from bone development and cartilage neoplasia**. *Ann N Y Acad Sci* (2008) **1144** 36-41. DOI: 10.1196/annals.1418.006
38. de Miguel F, Sarasa JL, Lopez-Ferro O, Esbrit P. **Immunohistochemical detection of parathyroid hormone-related protein in human astrocytomas**. *J Histochem Cytochem.* (1998) **46** 277-9. DOI: 10.1177/002215549804600218
39. Pei Z, Lee KC, Khan A, Erisnor G, Wang HY. **Pathway analysis of glutamate-mediated, calcium-related signaling in glioma progression**. *Biochem Pharmacol* (2020) 176. DOI: 10.1016/j.bcp.2020.113814
40. Abbaspour N, Hurrell R, Kelishadi R. **Review on iron and its importance for human health**. *J Res Med Sci.* (2014) **19** 164-74. PMID: 24778671
41. Zhang CL, Rawal S. **Dietary iron intake, iron status, and gestational diabetes**. *Am J Clin Nutr.* (2017). DOI: 10.3945/ajcn.117.156034
42. Han O. **Molecular mechanism of intestinal iron absorption**. *Metallomics.* (2011) **3** 103-9. DOI: 10.1039/c0mt00043d
43. Lal A. **Iron in health and disease: an update**. *Indian J Pediatr.* (2020) **87** 58-65. DOI: 10.1007/s12098-019-03054-8
44. Parent ME, Turner MC, Lavoue J, Richard H, Figuerola J, Kincl L. **Lifetime occupational exposure to metals and welding fumes, and risk of glioma: a 7-country population-based case-control study**. *Environ Health* (2017) 16. DOI: 10.1186/s12940-017-0300-y
45. Xu SB, Wang ZF, Ye J, Mei SH, Zhang JM. **Identification of iron metabolism-related genes as prognostic indicators for lower-grade glioma**. *Front Oncol* (2021) **11** 729103. DOI: 10.3389/fonc.2021.729103
46. Lin H, Wang YF, Lai HQ, Li XL, Chen TF. **Iron(ii)-polypyridyl complexes inhibit the growth of glioblastoma tumor and enhance trail-induced cell apoptosis**. *Chem Asian J.* (2018) **13** 2730-8. DOI: 10.1002/asia.201800862
47. Eales KL, Wilkinson EA, Cruickshank G, Tucker J, Tennant DA. **Verteporfin selectively kills hypoxic glioma cells through iron-binding and increased production of reactive oxygen species**. *Sci Rep* (2018) 8. DOI: 10.1038/s41598-018-32727-1
48. Vasak M, Hasler DW. **Metallothioneins: new functional and structural insights**. *Curr Opin Chem Biol.* (2000) **4** 177-83. DOI: 10.1016/S1367-5931(00)00082-X
49. Chasapis CT, Loutsidou AC, Spiliopoulou CA, Stefanidou ME. **Zinc and human health: an update**. *Arch Toxicol.* (2012) **86** 521-34. DOI: 10.1007/s00204-011-0775-1
50. Rink L, Gabriel P. **Zinc and the immune system**. *Proc Nutr Soc.* (2000) **59** 541-52. DOI: 10.1017/S0029665100000781
51. Li J, Cao DH, Huang Y, Chen B, Chen ZY, Wang RY. **Zinc intakes and health outcomes: an umbrella review**. *Front Nutr* (2022) **9** 798078. DOI: 10.3389/fnut.2022.798078
52. Li PW, Xu JM, Shi Y, Ye Y, Chen K, Yang J. **Association between zinc intake and risk of digestive tract cancers: a systematic review and meta-analysis**. *Clin Nutr.* (2014) **33** 415-20. DOI: 10.1016/j.clnu.2013.10.001
53. Li L, Gai XS. **The association between dietary zinc intake and risk of pancreatic cancer: a meta-analysis**. *Biosci Rep* (2017) 37. DOI: 10.1042/BSR20171121
54. Zelko IN, Mariani TJ, Folz RJ. **Superoxide dismutase multigene family: a comparison of the CuZn-SOD (SOD1), Mn-SOD (SOD2), and EC-SOD (SOD3) gene structures, evolution, and expression**. *Free Radic Biol Med.* (2002) **33** 337-49. DOI: 10.1016/S0891-5849(02)00905-X
55. Song Y, Leonard SW, Traber MG, Ho E. **Zinc deficiency affects DNA damage, oxidative stress, antioxidant defenses, and DNA repair in rats**. *J Nutr.* (2009) **139** 1626-31. DOI: 10.3945/jn.109.106369
56. Ladeira C, Carolino E, Gomes MC, Brito M. **Role of macronutrients and micronutrients in DNA damage: results from a food frequency questionnaire**. *Nutr Metab Insights.* (2017) **10** 1006414458. DOI: 10.1177/1178638816684666
57. Klaunig JE. **Oxidative stress and cancer**. *Curr Pharm Des.* (2018) **24** 4771-8. DOI: 10.2174/1381612825666190215121712
58. Olivier C, Oliver L, Lalier L, Vallette FM. **Drug resistance in glioblastoma: the two faces of oxidative stress**. *Front Mol Biosci.* (2020) **7** 620677. DOI: 10.3389/fmolb.2020.620677
59. Haase H, Watjem W, Beyersmann D. **Zinc induces apoptosis that can be suppressed by lanthanum in c6 rat glioma cells**. *Biol Chem.* (2001) **382** 1227-34. DOI: 10.1515/BC.2001.153
60. Balaji EV, Kumar N, Satarker S, Nampoothiri M. **Zinc as a plausible epigenetic modulator of glioblastoma multiforme**. *Eur J Pharmacol.* (2020) **887** 173549. DOI: 10.1016/j.ejphar.2020.173549
61. Volpe SL. **Magnesium in disease prevention and overall health**. *Adv Nutr.* (2013). DOI: 10.3945/an.112.003483
62. Barbagallo M, Veronese N, Dominguez LJ. **Magnesium in aging, health and diseases**. *Nutrients* (2021) 13. DOI: 10.3390/nu13020463
63. Barbagallo M, Dominguez LJ. **Magnesium and aging**. *Curr Pharm Des.* (2010) **16** 832-9. DOI: 10.2174/138161210790883679
64. Saris NE, Mervaala E, Karppanen H, Khawaja JA, Lewenstam A. **Magnesium. An update on physiological, clinical and analytical aspects**. *Clin Chim Acta.* (2000) **294** 1-26. DOI: 10.1016/S0009-8981(99)00258-2
65. Leidi M, Wolf F, Maier J. **Magnesium and cancer: more questions than answers**. *Magn Cent Nerv Syst* (2011). DOI: 10.1017/UPO9780987073051.017
66. Ko HJ, Youn CH, Kim HM, Cho YJ, Lee GH, Lee WK. **Dietary magnesium intake and risk of cancer: a meta-analysis of epidemiologic studies**. *Nutr Cancer.* (2014) **66** 915-23. DOI: 10.1080/01635581.2014.922203
67. Mazur A, Maier JA, Rock E, Gueux E, Nowacki W, Rayssiguier Y. **Magnesium and the inflammatory response: potential physiopathological implications**. *Arch Biochem Biophys.* (2007) **458** 48-56. DOI: 10.1016/j.abb.2006.03.031
68. Song Y, Li TY, van Dam RM, Manson JE, Hu FB. **Magnesium intake and plasma concentrations of markers of systemic inflammation and endothelial dysfunction in women**. *Am J Clin Nutr.* (2007) **85** 1068-74. DOI: 10.1093/ajcn/85.4.1068
69. King DE, Mainous AR, Geesey ME, Woolson RF. **Dietary magnesium and C-reactive protein levels**. *J Am Coll Nutr.* (2005) **24** 166-71. DOI: 10.1080/07315724.2005.10719461
70. Trapani V, Arduini D, Cittadini A, Wolf FI. **From magnesium to magnesium transporters in cancer: trpm7, a novel signature in tumour development**. *Magnes Res.* (2013) **26** 149-55. DOI: 10.1684/mrh.2014.0354
71. Sowers JL, Johnson KM, Conrad C, Patterson JT, Sowers LC. **The role of inflammation in brain cancer**. *Adv Exp Med Biol.* (2014) **816** 75-105. DOI: 10.1007/978-3-0348-0837-8_4
72. Wu N, Veillette A. **Immunology: magnesium in a signalling role**. *Nature.* (2011) **475** 462-3. DOI: 10.1038/475462a
73. Wu Y, Wang H, Wei D. **Oncogenic magnesium transporter 1 upregulates programmed death-1-ligand 1 expression and contributes to growth and radioresistance of glioma cells through the ERK/MAPK signaling pathway**. *Bioengineered.* (2022) **13** 9575-87. DOI: 10.1080/21655979.2022.2037214
74. Bost M, Houdart S, Oberli M, Kalonji E, Huneau JF, Margaritis I. **Dietary copper and human health: current evidence and unresolved issues**. *J Trace Elem Med Biol.* (2016) **35** 107-15. DOI: 10.1016/j.jtemb.2016.02.006
75. Bonham M, O'Connor JM, Hannigan BM, Strain JJ. **The immune system as a physiological indicator of marginal copper status?**. *Br J Nutr.* (2002) **87** 393-403. DOI: 10.1079/BJN2002558
76. Uriu-Adams JY, Keen CL. **Copper, oxidative stress, and human health**. *Mol Aspects Med.* (2005) **26** 268-98. DOI: 10.1016/j.mam.2005.07.015
77. Georgieff MK. **Nutrition and the developing brain: nutrient priorities and measurement**. *Am J Clin Nutr.* (2007). DOI: 10.1093/ajcn/85.2.614S
78. Stern BR, Solioz M, Krewski D, Aggett P, Aw TC, Baker S. **Copper and human health: biochemistry, genetics, and strategies for modeling dose-response relationships**. *J Toxicol Environ Health B Crit Rev.* (2007) **10** 157-222. DOI: 10.1080/10937400600755911
79. Trejo-Solis C, Jimenez-Farfan D, Rodriguez-Enriquez S, Fernandez-Valverde F, Cruz-Salgado A, Ruiz-Azuara L. **Copper compound induces autophagy and apoptosis of glioma cells by reactive oxygen species and jnk activation**. *BMC Cancer* (2012) 12. DOI: 10.1186/1471-2407-12-156
80. Castillo-Rodriguez RA, Palencia G, Anaya-Rubio I, Perez J, Jimenez-Farfan D, Escamilla-Ramirez A. **Anti-proliferative, pro-apoptotic and anti-invasive effect of the copper coordination compound Cas III-la through the induction of reactive oxygen species and regulation of wnt/beta-catenin pathway in glioma**. *J Cancer.* (2021) **12** 5693-711. DOI: 10.7150/jca.59769
|
---
title: 'Sacubitril–valsartan versus enalapril for the treatment of acute decompensated
heart failure in Chinese settings: A cost-effectiveness analysis'
authors:
- Tianyang Hu
- Yiting Liu
- Yake Lou
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC10018029
doi: 10.3389/fphar.2023.925375
license: CC BY 4.0
---
# Sacubitril–valsartan versus enalapril for the treatment of acute decompensated heart failure in Chinese settings: A cost-effectiveness analysis
## Abstract
Background: The episode of acute decompensated heart failure (ADHF) is the main cause of hospitalization for heart failure (HF). Sacubitril–valsartan has been proven to be effective in reducing the risks of hospitalization for HF in ADHF. When to initiate sacubitril–valsartan in ADHF to make it the most cost-effective in China remains unclear.
Methods: A lifetime Markov model with a 1-month cycle length was developed to evaluate the cost-effectiveness of early or late initiation of sacubitril–valsartan versus enalapril in ADHF. Early initiation of sacubitril–valsartan meant that it was initiated after stabilization from ADHF, and late initiation of sacubitril–valsartan meant that it was initiated after stabilization from HF, which includes no hospitalization for at least three consecutive months. The primary outcome was the incremental cost-effectiveness ratio (ICER), expressed as the ratio of incremental cost to incremental effectiveness. The secondary outcomes were total costs and total effectiveness. Three times of per capita GDP of China in 2021 was set as the willingness-to-pay threshold. One-way sensitivity analysis and probabilistic sensitivity analysis were employed to test the robustness of the results.
Results: The early initiation of sacubitril–valsartan treatment resulted in an ICER of 3,662.4 USD per quality-adjusted life year, lower than the willingness-to-pay threshold, and the late initiation of sacubitril–valsartan treatment gained an ICER of 4,444.4 USD/QALY, still lower than the willingness-to-pay threshold. One-way sensitivity analysis showed that our results were robust, and probabilistic sensitivity analysis suggested that early initiation of sacubitril–valsartan in ADHF was cost-effective under a $97.4\%$ circumstance.
Conclusion: Early initiation of sacubitril–valsartan after stabilization of ADHF is highly cost-effective compared with the use of enalapril; late initiation of sacubitril–valsartan after stabilization of HF is still cost-effective but not as cost-effective as early initiation of sacubitril–valsartan in ADHF. For Chinese ADHF patients, the time to initiate sacubitril–valsartan should be when the patient is stabilized from ADHF rather than when stabilized from HF, from the perspective of economic evaluation.
## Introduction
Heart failure (HF) is a terminal manifestation of many heart diseases. It is estimated that about 38 million patients suffer from HF worldwide (Braunwald, 2015). The incidence of HF increases dramatically with age. For the population aged over 40 years old, the incidence is about $1\%$–$2\%$, but it increases to $10\%$ in those over 70 years old (McDonagh et al., 2021). The reason for hospitalization for HF is mainly due to the episode of acute decompensated heart failure (ADHF) (Zhang et al., 2017). ADHF may cause serious consequences, including deterioration of heart function, repeated hospitalizations, and death (Cook et al., 2014). About $4.1\%$ of ADHF patients with HF die during hospitalization (Zhang et al., 2017; Ma et al., 2021), and $20\%$ of patients will be subsequently readmitted to a hospital 1 month after hospitalization for ADHF (Reddy and Borlaug, 2019; Lan et al., 2021). In China, there are about 8.9 million patients suffering from HF, and annually, 2.58 million Chinese patients die of HF (Lan et al., 2021).
Sacubitril–valsartan, as an angiotensin–neprilysin inhibition agent, has been proven to be superior to enalapril in reducing the risks of cardiovascular death and hospitalization for HF in HF patients with reduced ejection fraction (HFrEF) (McMurray et al., 2014). In HFrEF patients with ADHF, sacubitril–valsartan has also been demonstrated to reduce the risk of hospitalization for HF, but it has not proved to reduce mortality within 2 months after hospitalization (Velazquez et al., 2019). Although sacubitril–valsartan is superior to enalapril in HFrEF treatment, whether sacubitril–valsartan should be added into the standard treatment remains controversial (Liu et al., 2021) because some studies showed that sacubitril–valsartan is not cost-effective in their settings. In a study conducted in America, investigators found that adding sacubitril–valsartan into the standard treatment resulted in an incremental cost-effectiveness ratio of 143,891 USD per quality-adjusted life year, which is higher than the willingness-to-pay threshold, indicating that sacubitril–valsartan is not cost-effective in HFrEF (Zueger et al., 2018), but a study performed in Chinese settings suggested that sacubitril–valsartan is cost-effective in HFrEF (Wu et al., 2020). For the cost-effectiveness of sacubitril–valsartan in ADHF patients, there are still different conclusions. A study conducted in Australia proved that adding sacubitril–valsartan into standard treatment is not cost-effective in ADHF (Perera et al., 2021). However, Krittayaphong’s study found that sacubitril–valsartan is cost-effective in Thailand (Krittayaphong and Permsuwan, 2021). There is still no study performing the economic evaluation of sacubitril–valsartan in Chinese ADHF patients.
Cost-effectiveness analysis is a way of balancing the costs and benefits of new and traditional therapies. A new therapy is often associated with higher costs but better effectiveness. Cost-effectiveness analysis could answer the question of whether a new therapy is worth it or not. Considering the huge economic burden of HF in China, it is necessary for us to perform an economic evaluation to investigate the cost-effectiveness of sacubitril–valsartan versus enalapril in ADHF.
## Methods
The present study was reported in accordance with the Consolidated Health Economic Evaluation Reporting Standards 2022 (CHEERS 2022) Statement (Husereau et al., 2022).
## Population
The target population of the present study was a hypothesis cohort in China with similar baseline characteristics to those in the PIONEER-HF study (Velazquez et al., 2019). In the PIONEER-HF study, the patients had a median age of 62 years old, with an interquartile range of 51–72 years, a left ventricular ejection fraction (EF) of $40\%$ or less, and an N-terminal pro–B-type natriuretic peptide (NT-proBNP) concentration of 1,600 pg per milliliter or more, or a B-type natriuretic peptide (BNP) concentration of 400 pg per milliliter or more, and had received a primary diagnosis of ADHF, including signs and symptoms of the fluid overload. The baseline characteristics of PIONEER-HF are shown in Table 1 and details of inclusion and exclusion criteria are shown in Table 1 in the supplementary materials.
**TABLE 1**
| Variable | Value |
| --- | --- |
| Age | 62 (51–72) |
| Female sex (%) | 28 |
| Body mass index (kg/m2) | 30.3 (25.8–37.1) |
| Previous heart failure (%) | 65.4 |
| Previous use of medication (%) | |
| ACE inhibitor or ARB | 47.9 |
| Beta blocker | 59.6 |
| MRA | 10 |
| Loop diuretic | 57 |
| Hydralazine | 7.2 |
| Nitrate | 9.5 |
| Digoxin | 8.6 |
| NYHA class (%) | |
| I | 1 |
| II | 25.2 |
| III | 62.7 |
| IV | 8.6 |
| Not assessed | 2.5 |
| Systolic blood pressure | 118 (109–133) |
| Left ventricular ejection fraction (%) | 24.5 (18–30) |
| NT-proBNP at randomization (pg/ml) | 2710 (1,363–5403) |
| Medical history (%) | |
| Myocardial infarction | 7 |
| Atrial fibrillation | 35.4 |
| ICD only | 19.8 |
| CRTD | 8.7 |
| Comorbidities (%) | |
| Hypertension | 85.5 |
| Previous stroke | 9.9 |
| Diabetes mellitus | 19.1 |
| Hyperlipidemia | 37.1 |
There were two comparators and one control; of the two comparators, one had an early initiation of sacubitril–valsartan, in which the sacubitril–valsartan treatment was initiated after stabilization from ADHF (comparator 1), defined by the maintenance of a systolic blood pressure of at least 100 mm Hg for the preceding 6 h, with no increase in the dose of intravenous diuretics, no use of intravenous vasodilators during the preceding 6 h, and no use of intravenous inotropes during the preceding 24 h. The other comparator had a late initiation of sacubitril–valsartan treatment, in which the treatment was initiated after stabilization from HF, defined as not being hospitalized for at least three consecutive months (comparator 2). The control was given enalapril from the start, and the same treatment was continued after discharge from hospitalization. All the subsequent sensitivity analyses were based on comparator 1. The research process is shown in Figure 1.
**FIGURE 1:** *Abbreviation: ADHF, acute decompensated heart failure; Sac–val, sacubitril–valsartan; C-E. cost-effectiveness; HF, heart failure. Chinese ADHF patients were randomly allocated to receive early initiation of sacubitril–valsartan, late initiation of sacubitril–valsartan, or no initiation. After a lifetime simulation, the cost-effectiveness analysis was performed.*
## Model construction
A lifetime horizon Markov model with a 1-month cycle was developed to evaluate the cost-effectiveness of early or late initiation of sacubitril–valsartan versus enalapril in ADHF. Considering the fact that the median age of patients in the PIONEER-HF study was 62 years old and that studies showed that the mean age of Chinese HFrEF patients was 60 years old, we took 60 years old as the starting age of the Markov simulation (Zhang et al., 2017; Velazquez et al., 2019). In the Markov model, hospitalized ADHF patients could receive sacubitril–valsartan 200 mg (97 mg of sacubitril plus 103 mg of valsartan) twice a day or enalapril 10 mg twice a day plus standard treatment, and the same treatment would continue after their discharge until the lifetime horizon (Velazquez et al., 2019). In our simulation, considering that readmission and cardiovascular death tended to occur within 3 months after discharge, we assumed that patients who had not been hospitalized for three consecutive months were in a stable state (Greene et al., 2015). Patients who had been hospitalized within 3 months had a higher incidence of rehospitalization and cardiovascular death, and a death occurring within 2 months after discharge was considered cardiovascular death. Patients stabilized for at least three consecutive months had a lower incidence of death and hospitalization, and they may experience cardiovascular death or non-cardiovascular death. Based on this assumption, there were four transition states and two absorbed states in our model, which were “Hospitalized HF,” “Non-hospitalized HF month 1,” “Non-hospitalized HF month 2,” “Non-hospitalized HF month 3,” “Cardiovascular death,” and “Non-cardiovascular death”. Patients who entered this model would begin with the “Hospitalized HF” state, and those who did not experience cardiovascular death would enter the transition state of “Non-hospitalized HF month 1.” Patients in “Non-hospitalized HF month 1” may experience cardiovascular death or readmission. Patients who did not experience such events would enter “Non-hospitalized HF month 2.” Patients in “Non-hospitalized HF month 2” may still experience cardiovascular death or readmission, and those who did not experience such events for at least three consecutive months would enter “Non-hospitalized HF month 3”. Patients in “Non-hospitalized HF month 3” were considered stable, and they may experience cardiovascular death, non-cardiovascular death, or hospitalization. Those who did not experience such events would continue the cycle in “Non-hospitalized HF month 3”. The model was validated by other studies, and the corresponding state transition diagram can be seen in Figure 2 (Gaziano et al., 2016; Krittayaphong and Permsuwan, 2021).
**FIGURE 2:** *Abbreviation: HF, heart failure. Diagram of the Markov model using the state transition.*
The study was performed from a Chinese healthcare system perspective. Only direct costs (drugs and hospitalization costs) were accounted for in our model. All the costs were converted to the price in 2021 in China, according to the consumer price indexes of healthcare (CPI). The CPIs in 2015–2021 were 1.027, 1.038, 1.06, 1.043, 1.024, 1.018, and 1.004, respectively. Future costs, life year (LY), and quality-adjusted life-years (QALYs) were discounted at a rate of 0.03 per year, which was the geometric mean value of the aforementioned figures, but non-monetary outcomes (hospitalization incidence, readmission incidence, and mortality rate) were not discounted. The discount rate ranged from 0 to 0.06 in the one-way sensitivity analysis.
## Parameter input
All input parameters are listed in Table 2.
**TABLE 2**
| Parameter | Base case | Range low | Range high | SD | Distribution | Source |
| --- | --- | --- | --- | --- | --- | --- |
| Transition probability for cardiovascular death | Transition probability for cardiovascular death | Transition probability for cardiovascular death | Transition probability for cardiovascular death | Transition probability for cardiovascular death | Transition probability for cardiovascular death | Transition probability for cardiovascular death |
| Sac–valsartan (≤2 months) a | 0.0114 | 0.0044 | 0.0185 | 0.0035 | β | Velazquez et al. (2019) |
| Enalapril (≤2 months) | 0.0172 | 0.0086 | 0.0258 | 0.0044 | β | Velazquez et al. (2019) |
| Sac–valsartan (≥3 months) b | 0.0053 | 0.0048 | 0.0057 | 0.0002 | β | McMurray et al. (2014) |
| Enalapril (≥3 months) | 0.0066 | 0.0061 | 0.0071 | 0.0003 | β | McMurray et al. (2014) |
| Transition probability for non-cardiovascular death | Transition probability for non-cardiovascular death | Transition probability for non-cardiovascular death | Transition probability for non-cardiovascular death | Transition probability for non-cardiovascular death | Transition probability for non-cardiovascular death | Transition probability for non-cardiovascular death |
| 60–64 years | 0.0004 | — | — | — | — | Ma et al. (2022) |
| 65–69 years | 0.0007 | — | — | — | — | Ma et al. (2022) |
| 70–74 years | 0.001 | — | — | — | — | Ma et al. (2022) |
| 75–79 years | 0.0017 | — | — | — | — | Ma et al. (2022) |
| 80–84 years | 0.0026 | — | — | — | — | Ma et al. (2022) |
| ≥85 years | 0.0054 | — | — | — | — | Ma et al. (2022) |
| Transition probability for hospitalization | Transition probability for hospitalization | Transition probability for hospitalization | Transition probability for hospitalization | Transition probability for hospitalization | Transition probability for hospitalization | Transition probability for hospitalization |
| Sac–valsartan (≤2 months) | 0.04060 | 0.0275 | 0.0539 | 0.0067 | β | Velazquez et al. (2019) |
| Enalapril (≤2 months) | 0.0717 | 0.0545 | 0.0893 | 0.0089 | β | Velazquez et al. (2019) |
| Sac–valsartan (≥3 months) c | 0.0256 | 0.0228 | 0.0285 | 0.0014 | β | McMurray et al. (2014) |
| Enalapril (≥3 months) d | 0.0339 | 0.0306 | 0.0373 | 0.0017 | β | McMurray et al. (2014) |
| Costs (USD/month) | | | | | | |
| Sac–valsartan e | 50.5 | 25.2 | 180.7 | 18 | γ | Local institution |
| Enalapril + standard f | 37.9 | 18.9 | 75.7 | 3.8 | γ | Huang et al. (2017) |
| Hospitalization g | 2361.5 | 1180.7 | 4722.9 | 236 | γ | Huang et al. (2017) |
| Utilities | | | | | | |
| Sac–valsartan (per month) h | 0.0698 | 0.0628 | 0.0768 | 0.0036 | β | Krittayaphong and Permsuwan (2021) |
| Enalapril (per month) | 0.0691 | 0.0622 | 0.076 | 0.0035 | β | Krittayaphong and Permsuwan (2021) |
| Hospitalization | -0.1 | -0.08 | -0.13 | 0.0128 | β | Krittayaphong and Permsuwan (2021) |
| Discount rate i | 0.03 | 0 | 0.06 | — | — | — |
## Transition probability input
For those who were in the transition states of “Hospitalized HF,” “Non-hospitalized HF month 1,” and “Non-hospitalized HF month 2,” the transition probabilities were calculated based on the PIONEER-HF study (comparison of sacubitril/valsartan versus enalapril on the effect on NT-proBNP in patients stabilized from an acute heart failure episode), which reported the cardiovascular mortalities and rehospitalization incidence for hospitalized ADHF patients. The 2-month cardiovascular mortality rate in sacubitril–valsartan was 0.023 = $\frac{10}{440}$, and the 1-month cardiovascular mortality rate was 0.0115 = -ln (1–$\frac{10}{440}$)/2; then, the transition probability for 1-month cardiovascular death in sacubitril–valsartan was 0.0114 = 1-exp (-0.0115). Using the same formula, we calculated that the transition probability for 1-month cardiovascular death in enalapril was 0.0172, and the corresponding 1-month transition probabilities for rehospitalization in sacubitril–valsartan and enalapril were 0.0406 and 0.0717, respectively. Those who entered “Non-hospitalized HF month 3” were regarded stable, and the transition probabilities for cardiovascular death were derived from the PARADIGM-HF study (efficacy and safety of LCZ696 compared to enalapril on morbidity and mortality of patients with chronic heart failure). Using the aforementioned formula, the transition probabilities obtained for 1-month cardiovascular death in sacubitril–valsartan and enalapril were 0.0053 and 0.0066, respectively. The transition probability for hospitalization of patients in enalapril was obtained from a Chinese study investigating the economic burden of HF in China; the 1-year hospitalization rate was 0.3388 = $\frac{392}{1}$,157, and the 1-month transition probability was 0.0339. For transition probability in sacubitril–valsartan, we used the hazard ratio (HR) and hospitalization rate in enalapril to calculate the rehospitalization rate in sacubitril–valsartan, and we found that the 1-month transition probability in sacubitril–valsartan was 0.0256.
The transition probability for non-cardiovascular death was accessed from the China Health Statistical Yearbook 2021 and was age-dependent (Ma et al., 2022). It was calculated using the total mortality rate minus the cardiovascular mortality rate as there is no non-cardiovascular mortality reported in the Yearbook. The yearly mortality was converted to a 1-month mortality rate by dividing it by 12. The 1-month non-cardiovascular mortality rates for those aged 60–64, 65–69, 70–74, 75–79, 80–84, and ≥85 were 0.0004, 0.0007, 0.0010, 0.0017, 0.0026, and 0.0054. Both cohorts adopted the same non-cardiovascular mortality rate.
## Cost input
The cost of sacubitril–valsartan was as per the collective purchasing price of the Chinese government, which was 38 Chinese yuan (CNY)/7 tablets (200 mg/tablets) (equal to 5.9 USD). Based on taking 200 mg twice a day, the monthly cost for sacubitril–valsartan was 325.7 CNY (equal to 50.5 USD). The lower interval of the cost of sac–valsartan was obtained, assuming that the cost of sacubitril–valsartan could reduce to $50\%$ of its current price. For the upper interval, we adopted the price before sacubitril–valsartan was included in the collective purchasing list, namely 1,165.7 Chinese yuan/month (180.7 USD). The cost for enalapril plus standard treatment was obtained from Huang et al. [ 2017]. The cost for enalapril plus standard treatment was 198 CNY (equal to 29.1 USD) per month in 2014, and it was 244.2 CNY (equal to 37.9 USD) per month in 2021, when taking the CPI into consideration. The cost for hospitalization is also derived from Huang et al. [ 2017]; using the same method, we can conclude that the cost for hospitalization was 15,235 CNY (2361.5 USD) per event in 2021. To enable the reader an easy understanding of the cost-effectiveness of sacubitril–valsartan versus enalapril in Chinese ADHF patients, all costs were converted from CNY to USD, at a ratio of 6.4515, which was the average value of the exchange rate in 2021.
The adverse events in the PIONEER-HF study incurred low treatment costs, and it was not included in the analysis.
## Utility input
According to a published study, the utilities in sacubitril–valsartan and enalapril were 0.838 and 0.829, respectively. Every hospitalization event would result in a reduction of 0.1 in utility (-0.1/time).
## Outcomes
The primary outcome of the present study was the incremental cost-effectiveness ratio (ICER), expressed as the ratio of incremental cost to incremental effectiveness. Secondary outcomes are total costs and total effectiveness (life-years and quality-adjusted life-years (QALYs)). According to the recommendation of the China Guidelines for Pharmacoeconomic Evaluations (Hu et al., 2020), the willingness-to-pay (WTP) threshold was three times the current per capita GDP in China, which was 242,928 CNY = 80,976 CNY*3 (equal to 37,654.5 USD) (Wang et al., 2021). If the ICER calculated was lower than that threshold, it would be thought to be cost-effective; otherwise, it would not be cost-effective.
## Sensitivity analysis
One-way sensitivity and probabilistic sensitivity analyses were employed to validate the impacts of these parameters on outcomes and the robustness of our results. In the one-way sensitivity analysis, the parameters fluctuated in their $95\%$ confidence interval (CI) or given interval, and a tornado diagram was drawn to display our results. In the probabilistic sensitivity analysis, 10,000 times of Monte Carlo simulations based on probabilistic sensitivity sampling were performed, and the results were illustrated in cost-effectiveness acceptability curves and scatter plots.
## Statistical analysis
All the statistical analyses were performed using TreeAge Pro 2011 software (Williamstown, MA. United States) and EXCEL software (Redmond, Washington, United States), and a half-cycle correction was applied in the model to prevent overestimation of the costs and effectiveness.
## Base case analysis
After a simulation of the lifetime horizon, the early initiation of sacubitril–valsartan treatment resulted in a higher cost than enalapril treatment but gained a higher QALY and life year, which incurred an ICER of 3,662.4 USD/QALY. The costs of sacubitril–valsartan and enalapril were 17,515.2 and 12,189.7 USD, respectively, and the incremental cost was 5325.4 USD. The QALYs in both groups were 7.28 and 5.82, respectively. For life years, sacubitril–valsartan still got higher life years than enalapril, which were 9.12 and 7.51 life years, respectively (Table 3).
**TABLE 3**
| Intervention | Total cost (USD) | Total effectiveness (QALY) | Total effectiveness (LY) | Incremental cost (USD) | Incremental effectiveness (QALY) | ICER (USD/QALY) |
| --- | --- | --- | --- | --- | --- | --- |
| Enalapril | 12189.7 | 5.82 | 7.51 | - | - | - |
| HF hospitalization | 8920.9 | -0.4 | 0.0 | - | - | - |
| Stable state | 3268.9 | 6.22 | 7.51 | - | - | - |
| Sac–val (early initiation) | 17515.2 | 7.28 | 9.12 | 5325.4 | 1.45 | 3662.4 |
| HF hospitalization | 8140.3 | -0.36 | 0.0 | -780.6 | 0.04 | - |
| Stable state | 9374.9 | 7.64 | 9.12 | 6106 | 1.42 | - |
| Sac–val (late initiation) | 16483.6 | 6.79 | 8.55 | 4293.9 | 0.97 | 4444.4 |
| HF hospitalization | 8085.7 | -0.37 | 0.0 | -835.2 | 0.03 | - |
| Stable state | 8397.9 | 7.16 | 8.55 | 5129 | 0.94 | - |
The late initiation of sacubitril–valsartan treatment still gained higher costs and higher QALY than enalapril treatment. The costs were 16,483.6 USD and 12,189.7 USD, respectively, and the corresponding effectiveness were 6.79 and 5.82 QALY, thus resulting in an ICER of 4,444.4 USD/QALY.
## One-way sensitivity analysis
As could be seen in Figure 3, the cost of sacubitril–valsartan had the largest impact on the ICER. When costs of sacubitril–valsartan fluctuated from 25.2 to 180.7 USD/month, the ICER ranged from 1762.5 to 13,462.8 USD/QALY, still lower than three times the per capita GDP in China in 2021. Other factors had little impact on the ICER fluctuation.
**FIGURE 3:** *Abbreviation: ICER, incremental cost-effectiveness ratio. Tornado diagram based on the one-way sensitivity analysis. Costs of sacubitril–valsartan impact the most on the ICER fluctuation; other input parameters have little impact on ICERs.*
## Probabilistic sensitivity analysis
Probabilistic sensitivity analysis using Monte Carlo simulations based on probabilistic sensitivity sampling was conducted to validate the robustness of the results. In Figure 4, the scatter plot illustrated that under $97.4\%$ of circumstances, sacubitril–valsartan was cost-effective or superior to enalapril when the WTP was 37,654.5 USD/QALY. Sacubitril–valsartan was not cost-effective or inferior to enalapril only in $2.6\%$ of circumstances. The cost-effectiveness acceptability curve suggested that when the WTP was 3,681.3 USD (0.293 times the per capita GDP in China in 2021), sacubitril–valsartan and enalapril shared the similar acceptability, and when the WTP was higher than that value, sacubitril–valsartan gained higher acceptability than enalapril. When the WTP was 37,654.5 USD/QALY, the acceptability of sacubitril–valsartan was over $97\%$ (Figure 5).
**FIGURE 4:** *Scatter plot based on probabilistic sensitivity analysis. The probability that sacubitril–valsartan is cost-effective or superior to enalapril is over 97%.* **FIGURE 5:** *Abbreviation: CE, cost effectiveness. Cost-effectiveness acceptability curve of sacubitril–valsartan versus enalapril in acute decompensated heart failure in Chinese settings. When the willingness-to-pay threshold is 3,681.3 USD/QALY (0.293 times the per capita GDP in China in 2021), sacubitril–valsartan and enalapril shared the similar acceptability.*
## Discussion
A previous economic evaluation of sacubitril–valsartan in Chinese settings has demonstrated that sacubitril–valsartan is cost-effective in stable HFrEF patients (Wu et al., 2020). Our study is the first to investigate sacubitril–valsartan in Chinese ADHF patients and found that early initiation of sacubitril–valsartan after stabilization of ADHF is cost-effective compared with enalapril; late initiation of sacubitril–valsartan after stabilization of HF is still cost-effective, even though not as cost-effective as early initiation of sacubitril–valsartan.
Sacubitril–valsartan is a combination of sacubitril and valsartan in equal proportions (Entresto, 2015). Sacubitril works by inhibiting neprilysin and enhancing the effect of natriuretic peptide, causing vasodilation, and the effects of diuretic and natriuretic peptides, ultimately reducing ventricular preload and remodeling (Mangiafico et al., 2013). Valsartan is a classical angiotensin Ⅱ receptor blocker that inhibits angiotensin II by blocking angiotensin Ⅱ receptor 1, causing vasodilation, and diuretic and natriuretic peptides, inhibiting aldosterone release (Wang et al., 2015; von Lueder et al., 2015). In addition to the abovementioned effects, sacubitril–valsartan could also function by improving endothelial dysfunction and arterial stiffness and by reducing oxidative stress, platelet activation, and inflammation circulating biomarkers (Cassano et al., 2022). Other drugs that could improve biomarkers of endothelial dysfunction and inflammation in hypertension might also improve the clinical prognosis of HF patients (Alshahawey et al., 2017; Ateya and Sabri, 2017; Alshahawey et al., 2019).
Sacubitril–valsartan has been proven effective in HFrEF patients in a large randomized controlled trial (RCT) (McMurray et al., 2014; Pascual-Figal et al., 2021). Wu’s study suggested that sacubitril–valsartan was cost-effective for Chinese HFrEF patients from the patient’s perspective (Wu et al., 2020), which may partly be due to the drug collective purchase policy and reimbursement policy. In 2017, the Chinese government launched the drug collective purchase policy to improve the healthcare quality (China National Healthcare Security Administration, 2023). Drugs only with cost-effectiveness could be included in the collective purchase lists, and drugs in the lists could be widely used in Chinese public hospitals, which provide over $80\%$ of healthcare in China. The costs of sacubitril–valsartan (200 mg/tablet) have decreased from 19.43 CNY (3 USD) to 5.43 CNY (0.84 USD) since it was included in the list. On the other hand, the $80\%$ reimbursement policy in Wu’s study also contributed to the cost-effectiveness of sacubitril–valsartan (Wu et al., 2020). Although sacubitril–valsartan was cost-effective for HFrEF patients from the patients’ perspective, we had no knowledge whether it was still cost-effective from the healthcare provider’s perspective, without consideration of any reimbursement policy. In addition, we also did not know whether sacubitril–valsartan should be used in ADHF patients as early as we can. In our study, there are two comparators; one had an early initiation of sacubitril–valsartan in the ADHF hospitalization period, and the other had an initiation of sacubitril–valsartan after the stabilization from HF, defined as not hospitalized for at least three consecutive months. Our results indicate that early initiation of sacubitril–valsartan can additionally gain 1.45 QALY and 1.61 life years, and the ICER is 23,628 CNY (3,662.4 USD)/QALY, far lower than the WTP of 37,654.5 USD. In addition, even though not equal to early initiation of sacubitril–valsartan, the initiation of sacubitril–valsartan after the stabilization from the HF event still could gain more benefit with less costs; the late initiation of sacubitril–valsartan gains 0.97 QALY, and the ICER is 28,673 CNY (4,444.4 USD)/QALY, still far lower than the WTP. Our results indicate that early initiation of sacubitril–valsartan is most cost-effective, and the late initiation of sacubitril–valsartan is still cost-effective. Chinese ADHF patients should initiate the sacubitril–valsartan treatment early in their hospitalization period to get better clinical outcomes and higher cost-effectiveness.
From the point of view of cost-effectiveness, the early initiation of sacubitril–valsartan remains controversial. A study conducted in the US showed that initiation of sacubitril–valsartan during hospitalization could reduce hospitalization incidence, increase quality-adjusted life years, and was cost saving compared with no initiation or initiation after HF stabilization (Gaziano et al., 2016). Another study conducted in Thailand confirmed this conclusion in their settings (Krittayaphong and Permsuwan, 2021). However, the study performed in Australia revealed that the current acquisition price could not make sacubitril–valsartan cost-effective (Perera et al., 2021). The WTP thresholds in China were lower than those in the US and Australia (Gaziano et al., 2016; Chin et al., 2020; Perera et al., 2021), but we still found that early initiation of sacubitril–valsartan was cost-effective, which may partly be attributed to the lower costs of sacubitril–valsartan in China. As mentioned previously, the collective purchase policy has made the drug costs decrease from 19.43 CNY (3 USD) to 5.43 for each tablet (equal to 0.84 USD), lower than that in the US and Australia, even lower than that in Thailand (Krittayaphong and Permsuwan, 2018). Another reason may be that the absolute value is great in reducing the events’ incidence (Packer et al., 2015). For a 30-day HF readmission, the incidence is $13.4\%$ for enalapril but $9.7\%$ for sacubitril–valsartan, which results in an absolute reduction of $3.7\%$ (Desai et al., 2016). The reduction of cardiovascular mortality is still significant, and the cardiovascular mortality rate in the 2-month follow-up period after ADHF hospitalization is $3.4\%$ and $2.3\%$, respectively. In our simulation, early initiation of sacubitril–valsartan could lead to an additional 1.45 QALY (or 1.61 life years). Even late initiation of sacubitril–valsartan still gained 0.97 QALY compared with the use of enalapril. The benefit in QALY and life years is almost consistent with that in Gaziano et al. [ 2016]. To validate the robustness of our study, sensitivity analysis was performed; when the higher range of sacubitril–valsartan of 1,165.7 CNY (180.7 USD)/month was employed, which was the price of sacubitril–valsartan before including in the collective purchase lists, the ICER obtained was still lower than the WTP. Other factors had little impact on the ICER fluctuation. The probabilistic sensitivity analysis also showed that under $97.4\%$ of circumstances, sacubitril–valsartan is cost-effective, indicating that our results are robust.
To improve healthcare quality, many programs have been established in China. In the China Heart Failure (China-HF) Registry launched in 2012, the in-hospital mortality was 4.1 ± $0.3\%$ (Zhang et al., 2017), but it decreased to $2.8\%$ in the latest Heart Failure Report in China (Working Group on Heart Failure NCfCQIN, 2021). We noticed that in 2017, sacubitril–valsartan accounted for about $2.3\%$ of the overall oral RAAS inhibitors, but in 2020, it had risen to $63.7\%$, partly due to the acceptable costs. The wide use of sacubitril–valsartan has improved clinical outcomes to some extent, along with the use of other novel drugs. In addition, the indications for the treatment of hypertension with sacubitril–valsartan have been proven by China’s National Medical Products Administration, which may further improve the quality of Chinese HF patients.
In addition to sacubitril–valsartan, vericiguat and sodium-dependent glucose transporters 2 inhibitors (SGLT2i) have also been proven effective in ADHF or acute HF treatment (Armstrong et al., 2020; Bhatt et al., 2021). In the VICTORIA study, researchers investigated the efficacy of vericiguat on patients who had worsening HF and found that the incidence of cardiovascular death or hospitalization for HF was lower among those who received vericiguat than among those who received a placebo (Armstrong et al., 2020). The EMPULSE study and SOLOIST-WHF study demonstrated that initiation of SGLT2i in patients who had worsening HF or were hospitalized for acute HF could result in significant clinical benefits. Currently, vericiguat and several SGLT2i agents have been approved to treat HF in China, and SGLT2i and sacubitril–valsartan have been included in the collective purchasing list. The use of SGLT2i and sacubitril–valsartan in Chinese HF patients has climbed in the past few years. The real-world study of vericiguat, SGLT2i, and sacubitril–valsartan is warranted.
There are some limitations to our study. First, the data in our study were derived from large RCTs, which may not completely represent the patients in China, but a study investigated the efficacy of sacubitril–valsartan in HF and found that the differences in races did not modify the benefit of sacubitril–valsartan (Kristensen et al., 2016). Second, the costs in our study were derived from China’s local data, and whether this conclusion could be extended to other regions remains unclear. Third, we only used direct medical costs and direct non-medical costs, and indirect costs were not included in our costs; this limited us to analyzing it from the society’s perspective, which is the most comprehensive perspective. Fourth, the transition probability of clinical outcomes was derived from published studies rather than from the raw data, which limited us to perform subgroup analysis. Lastly, this study was conducted using a mathematical model. The costs and effectiveness of the model were obtained from published studies, and research based on real-world data is needed to confirm our conclusion.
## Conclusion
Early initiation of sacubitril–valsartan after stabilization of ADHF is of high cost-effectiveness compared with the use of enalapril. Late initiation of sacubitril–valsartan after stabilization of HF is still cost-effective but not as cost-effective as the early initiation of sacubitril–valsartan in ADHF. For Chinese ADHF patients, the time to initiate sacubitril–valsartan should be when the patient is stabilized from ADHF rather than stabilized from HF, from the perspective of economic evaluation.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material; further inquiries can be directed to the corresponding author.
## Author contributions
YKL came up with the idea and designed the protocol. TH and YTL synthetized the data and drafted the manuscript. YKL, TH, and YTL participated in the data collection and data analysis. All authors approved the final version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors, and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Alshahawey M., Shaheen S. M., Elsaid T., Sabri N. A.. **Effect of febuxostat on oxidative stress in hemodialysis patients with endothelial dysfunction: A randomized, placebo-controlled, double-blinded study**. *Int. urology Nephrol.* (2019) **51** 1649-1657. DOI: 10.1007/s11255-019-02243-w
2. Alshahawey M., Shahin S. M., Elsaid T. W., Sabri N. A.. **Effect of febuxostat on the endothelial dysfunction in hemodialysis patients: A randomized, placebo-controlled, double-blinded study**. *Am. J. Nephrol.* (2017) **45** 452-459. DOI: 10.1159/000471893
3. Armstrong P. W., Pieske B., Anstrom K. J., Ezekowitz J., Hernandez A. F., Butler J.. **Vericiguat in patients with heart failure and reduced ejection fraction**. *N. Engl. J. Med.* (2020) **382** 1883-1893. DOI: 10.1056/NEJMoa1915928
4. Ateya A. M., Sabri N. A.. **Effect of omega-3 fatty acids on serum lipid profile and oxidative stress in pediatric patients on regular hemodialysis: A randomized placebo-controlled study**. *J. Counc. Ren. Nutri. Natl. Kidney Found.* (2017) **27** 169-174. DOI: 10.1053/j.jrn.2016.11.005
5. Bhatt D. L., Szarek M., Steg P. G., Cannon C. P., Leiter L. A., McGuire D. K.. **Sotagliflozin in patients with diabetes and recent worsening heart failure**. *N. Engl. J. Med.* (2021) **384** 117-128. DOI: 10.1056/NEJMoa2030183
6. Braunwald E.. **The war against heart failure: The lancet lecture**. *Lancet* (2015) **385** 812-824. DOI: 10.1016/S0140-6736(14)61889-4
7. Cassano V., Armentaro G., Magurno M., Aiello V., Borrello F., Miceli S.. **Short-term effect of sacubitril/valsartan on endothelial dysfunction and arterial stiffness in patients with chronic heart failure**. *Front. Pharmacol.* (2022) **13** 1069828. DOI: 10.3389/fphar.2022.1069828
8. Chin K. L., Zomer E., Wang B. H., Liew D.. **Cost-effectiveness of switching patients with heart failure and reduced ejection fraction to sacubitril/valsartan: The Australian perspective**. *Heart, Lung Circulation* (2020) **29** 1310-1317. DOI: 10.1016/j.hlc.2019.03.007
9. **China national healthcare security administration**. (2023)
10. Cook C., Cole G., Asaria P., Jabbour R., Francis D. P.. **The annual global economic burden of heart failure**. *Int. J. Cardiol.* (2014) **171** 368-376. DOI: 10.1016/j.ijcard.2013.12.028
11. Desai A. S., Claggett B. L., Packer M., Zile M. R., Rouleau J. L., Swedberg K.. **Influence of sacubitril/valsartan (LCZ696) on 30-day readmission after heart failure hospitalization**. *J. Am. Coll. Cardiol.* (2016) **68** 241-248. DOI: 10.1016/j.jacc.2016.04.047
12. Entresto F. L.. **Entresto (Sacubitril/Valsartan): First-in-Class angiotensin receptor neprilysin inhibitor FDA approved for patients with heart failure**. *Am. Health & Drug Benefits* (2015) **8** 330-334. PMID: 26557227
13. Gaziano T. A., Fonarow G. C., Claggett B., Chan W. W., Deschaseaux-Voinet C., Turner S. J.. **Cost-effectiveness analysis of sacubitril/valsartan vs enalapril in patients with heart failure and reduced ejection fraction**. *JAMA Cardiol.* (2016) **1** 666-672. DOI: 10.1001/jamacardio.2016.1747
14. Greene S. J., Fonarow G. C., Vaduganathan M., Khan S. S., Butler J., Gheorghiade M.. **The vulnerable phase after hospitalization for heart failure**. *Nat. Rev. Cardiol.* (2015) **12** 220-229. DOI: 10.1038/nrcardio.2015.14
15. Hu S. L., Wu J. H., Wu J., Dong C. H., Li H. C., Liu G. E.. *China Guidelines for pharmacoeconomic evaluations: Chinese-English version* (2020)
16. Huang J., Yin H. J., Zhang M. L., Ni Q., Xuan J. W.. **Understanding the economic burden of heart failure in China: Impact on disease management and resource utilization**. *J. Med. Econ.* (2017) **20** 549-553. DOI: 10.1080/13696998.2017.1297309
17. Husereau D., Drummond M., Augustovski F., de Bekker-Grob E., Briggs A. H., Carswell C.. **Consolidated health economic evaluation reporting standards 2022 (CHEERS 2022) statement: Updated reporting guidance for health economic evaluations**. *PharmacoEconomics* (2022) **40** 601-609. DOI: 10.1007/s40273-021-01112-8
18. Kristensen S. L., Martinez F., Jhund P. S., Arango J. L., Bĕlohlávek J., Boytsov S.. **Geographic variations in the PARADIGM-HF heart failure trial**. *Eur. heart J.* (2016) **37** 3167-3174. DOI: 10.1093/eurheartj/ehw226
19. Krittayaphong R., Permsuwan U.. **Cost-effectiveness analysis of sacubitril-valsartan compared with enalapril in patients with heart failure with reduced ejection fraction in Thailand**. *Drugs, Devices, Other Interventions* (2018) **18** 405-413. DOI: 10.1007/s40256-018-0288-x
20. Krittayaphong R., Permsuwan U.. **Cost-utility analysis of sacubitril-valsartan compared with enalapril treatment in patients with acute decompensated heart failure in Thailand**. *Clin. Drug Investig.* (2021) **41** 907-915. DOI: 10.1007/s40261-021-01079-6
21. Lan T., Liao Y-H., Zhang J., Yang Z-P., Xu G-S., Zhu L.. **Mortality and readmission rates after heart failure: A systematic review and meta-analysis**. *Ther. Clin. Risk Manag.* (2021) **17** 1307-1320. DOI: 10.2147/TCRM.S340587
22. Liu X-Q., He L-S., Huang J-Q., Xiong L-J., Xia C., Lao H-Y.. **Cost-effectiveness analyses of sacubitril-valsartan for heart failure**. *Heart Fail. Rev.* (2021) **26** 1119-1130. DOI: 10.1007/s10741-020-09956-6
23. Ma L. Y., Wang Z. W., Fan J., Hu S. S.. **Report on cardiovascular health and diseases in China(in Chinese)**. *Chin. Circulation J.* (2021) **36** 521-545. DOI: 10.3969/j.issn.1000-3614.2021.06.001
24. Ma X. W., Yu X. J., Wang H. D., Wang B., Mao Q. A., Liu J. F.. *China health statistics Yearbook (2021)* (2022)
25. Mangiafico S., Costello-Boerrigter L. C., Andersen I. A., Cataliotti A., Burnett J. C.. **Neutral endopeptidase inhibition and the natriuretic peptide system: An evolving strategy in cardiovascular therapeutics**. *Eur. Heart J.* (2013) **34** 886-893. DOI: 10.1093/eurheartj/ehs262
26. McDonagh T. A., Metra M., Adamo M., Gardner R. S., Baumbach A., Böhm M.. **2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure**. *Eur. Heart J.* (2021) **42** 3599-3726. DOI: 10.1093/eurheartj/ehab368
27. McMurray J. J. V., Packer M., Desai A. S., Gong J. J., Lefkowitz M. P., Rizkala A. R.. **Angiotensin-neprilysin inhibition versus enalapril in heart failure**. *N. Engl. J. Med.* (2014) **371** 993-1004. DOI: 10.1056/NEJMoa1409077
28. Packer M., McMurray J. J. V., Desai A. S., Gong J., Lefkowitz M. P., Rizkala A. R.. **Angiotensin receptor neprilysin inhibition compared with enalapril on the risk of clinical progression in surviving patients with heart failure**. *Circulation* (2015) **131** 54-61. DOI: 10.1161/CIRCULATIONAHA.114.013748
29. Pascual-Figal D., Bayés-Genis A., Beltrán-Troncoso P., Caravaca-Pérez P., Conde-Martel A., Crespo-Leiro M. G.. **Sacubitril-valsartan, clinical benefits and related mechanisms of action in heart failure with reduced ejection fraction. A review**. *Front. Cardiovasc. Med.* (2021) **8** 754499. DOI: 10.3389/fcvm.2021.754499
30. Perera K., Ademi Z., Liew D., Zomer E.. **Sacubitril-valsartan versus enalapril for acute decompensated heart failure: A cost-effectiveness analysis**. *Eur. J. Prev. Cardiol.* (2021) **28** 966-972. DOI: 10.1177/2047487319878953
31. Reddy Y. N. V., Borlaug B. A.. **Readmissions in heart failure: It's more than just the medicine**. *Mayo Clin. Proc.* (2019) **94** 1919-1921. DOI: 10.1016/j.mayocp.2019.08.015
32. Velazquez E. J., Morrow D. A., DeVore A. D., Duffy C. I., Ambrosy A. P., McCague K.. **Angiotensin-neprilysin inhibition in acute decompensated heart failure**. *N. Engl. J. Med.* (2019) **380** 539-548. DOI: 10.1056/NEJMoa1812851
33. von Lueder T. G., Wang B. H., Kompa A. R., Huang L., Webb R., Jordaan P.. **Angiotensin receptor neprilysin inhibitor LCZ696 attenuates cardiac remodeling and dysfunction after myocardial infarction by reducing cardiac fibrosis and hypertrophy**. *Circ. Heart Fail.* (2015) **8** 71-78. DOI: 10.1161/CIRCHEARTFAILURE.114.001785
34. Wang B. H., von Lueder T. G., Kompa A. R., Huang L., Webb R., Jordaan P.. **Combined angiotensin receptor blockade and neprilysin inhibition attenuates angiotensin-II mediated renal cellular collagen synthesis**. *Int. J. Cardiol.* (2015) **186** 104-105. DOI: 10.12010/j.issn.1673-5846.2021.03.001
35. Wang L., Peng L., Peng Y., Li S., Wan X., Zeng X.. **Comparative analysis between 2020 version and 2011 version on China Guidelines for pharmacoeconomic evaluation**. *China J. Pharm. Econ.* (2021) **16** 5-8+15
36. **2020 clinical performance and quality measures for heart failure in China**. *Chin. Circulation J.* (2021) **36** 221-238. DOI: 10.3969/j.issn.1000-3614.2021.03.002
37. Wu Y., Tian S., Rong P., Zhang F., Chen Y., Guo X.. **Sacubitril-valsartan compared with enalapril for the treatment of heart failure: A decision-analytic Markov model simulation in China**. *Front. Pharmacol.* (2020) **11** 1101. DOI: 10.3389/fphar.2020.01101
38. Zhang Y., Zhang J., Butler J., Yang X., Xie P., Guo D.. **Contemporary epidemiology, management, and outcomes of patients hospitalized for heart failure in China: Results from the China heart failure (China-HF) Registry**. *J. Card. Fail.* (2017) **23** 868-875. DOI: 10.1016/j.cardfail.2017.09.014
39. Zueger P. M., Kumar V. M., Harrington R. L., Rigoni G. C., Atwood A., DiDomenico R. J.. **Cost-effectiveness analysis of sacubitril/valsartan for the treatment of heart failure with reduced ejection fraction in the United States**. *Pharmacotherapy* (2018) **38** 520-530. DOI: 10.1002/phar.2108
|
---
title: Bayesian hierarchical models and prior elicitation for fitting psychometric
functions
authors:
- Maura Mezzetti
- Colleen P. Ryan
- Priscilla Balestrucci
- Francesco Lacquaniti
- Alessandro Moscatelli
journal: Frontiers in Computational Neuroscience
year: 2023
pmcid: PMC10018033
doi: 10.3389/fncom.2023.1108311
license: CC BY 4.0
---
# Bayesian hierarchical models and prior elicitation for fitting psychometric functions
## Abstract
Our previous articles demonstrated how to analyze psychophysical data from a group of participants using generalized linear mixed models (GLMM) and two-level methods. The aim of this article is to revisit hierarchical models in a Bayesian framework. Bayesian models have been previously discussed for the analysis of psychometric functions although this approach is still seldom applied. The main advantage of using Bayesian models is that if the prior is informative, the uncertainty of the parameters is reduced through the combination of prior knowledge and the experimental data. Here, we evaluate uncertainties between and within participants through posterior distributions. To demonstrate the Bayesian approach, we re-analyzed data from two of our previous studies on the tactile discrimination of speed. We considered different methods to include a priori knowledge in the prior distribution, not only from the literature but also from previous experiments. A special type of Bayesian model, the power prior distribution, allowed us to modulate the weight of the prior, constructed from a first set of data, and use it to fit a second one. Bayesian models estimated the probability distributions of the parameters of interest that convey information about the effects of the experimental variables, their uncertainty, and the reliability of individual participants. We implemented these models using the software Just Another Gibbs Sampler (JAGS) that we interfaced with R with the package rjags. The Bayesian hierarchical model will provide a promising and powerful method for the analysis of psychometric functions in psychophysical experiments.
## 1. Introduction
Psychophysical methods are largely used in behavioral neuroscience to investigate the functional basis of perception in humans and other animals (Pelli and Farell, 1995). Using a model called the psychometric function, it is possible to test the quantitative relation between a physical property of the stimulus and its perceptual representation provided by the senses. This model has a typical sigmoid shape and relates the actual stimulus intensity (“physics”) on the abscissa to the probability of the response of the observer (i.e., perceptual response and “psychology”) on the ordinate, as collected with a forced-choice experiment. It is possible to summarize the performance of an observer by the parameters that are computed by the psychometric function: the Point of Subjective Equality (PSE), the slope, and the Just Noticeable Difference (JND) (Knoblauch and Maloney, 2012). The PSE estimates the accuracy of the response and corresponds to the stimulus value associated with a probability of response at chance level ($$p \leq 0.50$$). In two-interval forced-choice experiments, a deviation of the PSE from the value of the reference stimulus may indicate a bias, for example, in perceptual illusions (Moscatelli et al., 2016, 2019). The JND measures the noise of the response; the higher the JND, the higher the perceptual noise (Prins, 2016). The JND is an inverse function of the slope parameter of the psychometric function that is a measurement of the precision of the response. It is possible to test the slope or the JND of the function to evaluate the precision (or the noise) of the response.
Typically, generalized linear models (GLMs) are applied to estimate the parameters of the psychometric functions for each individual participant (Knoblauch and Maloney, 2012). In our previous study, we showed the advantages of using generalized linear mixed models (GLMMs) to estimate the responses of multiple participants at the population level (Moscatelli et al., 2012). A fairly comprehensive literature on fitting GLM and GLMM exists in different programming languages, including R, Python, and Matlab (Linares and López-Moliner, 2016; Schütt et al., 2016; Moscatelli and Balestrucci, 2017; Prins and Kingdom, 2018; Balestrucci et al., 2022).
In GLMM, we distinguish between fixed- and random-effect parameters (Stroup, 2012). The former, akin to the parameters of the psychometric function, estimates the effects of the experimental variables. Typically, the random-effect parameters estimate the variability across individual participants. In more complex data-sets, it is possible to account for other sources of unobserved variation by means of random-effect parameters. Blocking or batch effects are common examples of other random-effects parameters. The addition of this random component is the distinguishing feature of mixed models. For GLMMs, we assume that the random-effect parameters are normally distributed variables. The goal is to estimate the variance of that distribution. The larger the variance, the larger the heterogeneity across participants for a given parameter. However, the mean (or other central tendencies) of that distribution can be treated as if fixed effects have been applied to standard models. The conditional modes of the model estimating the response of individual participants can be treated as the fixed effects in standard psychometric functions. For example, in Balestrucci et al. [ 2022], we used conditional modes to plot the model estimates for individual participants.
A natural reinterpretation of the mixed model is the Bayesian approach, where all parameters are naturally considered as random variables, each having its own probability distribution (Zhao et al., 2006; Fong et al., 2010). Bayesian models provide not only a point estimate but also a probability distribution of the population parameter. Therefore, a Bayesian approach allows a natural assessment of the uncertainty in the parameter estimation. The advantages of the hierarchical Bayesian framework have been established in different fields in experimental psychology (Gelman et al., 1995; Rouder et al., 2003) and item response (Fox and Glas, 2001; Wang et al., 2002). To the best of our knowledge, only a few studies evaluate the use of Bayesian inference for fitting psychometric functions (Alcalá-Quintana and Garćıa-Pérez, 2004; Kuss et al., 2005; Schütt et al., 2016; Houpt and Bittner, 2018). In addition to estimating the intercept and the slope of the model, the flexibility of a Bayesian approach allows the study of uncertainties of the PSE.
This article is organized as follows. In Section 2, the two-stage Bayesian hierarchical model is proposed and discussed. Section 2.1 focuses on the description of prior distributions and Section 2.2 is dedicated to the discussion of the computational aspects. In Section 3, the data from two published experiments are considered. In Section 3.1, a Bayesian hierarchical model is fitted and compared with the results of Dallmann et al. [ 2015], while in Section 3.2, a Bayesian hierarchical model is fitted and compared with the results of Picconi et al. [ 2022]. In Section 4, the two studies considered in Section 3 are jointly analyzed. Two alternative approaches are proposed. The first one considers the combination of the two studies with the parameters from the first study used as a prior distribution. The second approach introduces a parameter, a0, to quantify the uncertainty (or weight) of the first study that is considered as historical data—as detailed in Section 2.1. Finally, in Section 5, a discussion of the model is proposed and the results obtained are discussed.
## 2. Model
A typical data-set from a psychophysical experiment includes repeated responses from more than one participant. Fitting these types of data with ordinary generalized linear models (GLM) would produce invalid standard errors of the estimated parameters because they would treat the errors within the subject in the same manner as the errors between subjects. A viable approach to overcome this problem consists of applying a multilevel model (Morrone et al., 2005; Steele and Goldstein, 2006; Pariyadath and Eagleman, 2007; Johnston et al., 2008). First, the parameters of the psychometric function are estimated for each subject. Next, the individual estimates are pooled to perform the second-level analysis for statistical inference. Alternatively, it is possible to use the generalized linear mixed model (GLMM) that accounts separately for the experimental effects and the variability between participants using random- and fixed-effect parameters (Moscatelli et al., 2012).
Bayesian methods provide a viable solution for fitting models of the GLM and GLMM families (Gelman et al., 1995; Rouder and Lu, 2005). In particular, Kuss et al. [ 2005] have applied Bayesian methods for estimating a psychometric function, based on a binomial mixture model. A Bayesian hierarchical model is a statistical model written in multiple levels (hierarchical form) and estimates the parameters using Markov chain Monte Carlo (MCMC) sampling. Applying a Bayesian hierarchical model consists of the following processes: (i) model definition, including specification of parameters and prior distributions in different levels, (ii) update of the posterior distributions given the data, (iii) and Bayesian inference to analyze the parameters' posterior distributions (McElreath, 2020).
In the current study, we considered data from two-interval forced-choice discrimination tasks, as mentioned in the two example data-sets detailed in Sections 3.1 and 3.2. A two-stage Bayesian hierarchical model has been applied to these data-sets, with a probit model for each individual subject at the first stage. Let X denote the experimental variable (or variables), and let Y be the response variable that consists of binary responses. Thus Yij = 1 if subject i in trial j perceived a comparison stimulus with value xij as larger in magnitude (e.g., depending on the specific task, faster, stiffer, heavier, brighter, etc.) than a reference stimulus. As for the example data analyzed in this article (speed discrimination task), Yij = 1 if the subject perceived the comparison as “faster” than a reference one. The relationship between the response variable and the experimental variables is defined as: The model assumed that the forced-choice responses Yij are independent and identically distributed (i.i.d.) conditional on the individual parameters (αi, βi). In case of repeated measurement, for each subject and conditions, Equation [1] can easily be substituted by where Yij represents, the number of “faster” responses for subject i at condition xij.
The function Φ−1 in Equation [2] establishes a linear relationship between the response probability and the predictor that is fully described by two parameters αi and βi. The probit link function Φ−1 is the inverse of the cumulative distribution function of the standard normal distribution Z. That is: For more details on probit link function refer to Agresti [2002] and Moscatelli et al. [ 2012]. Other link functions like Logit and Weibull are also often used in psychophysics (Agresti, 2002; Foster and Zychaluk, 2009).
In the first stage, the model characterized the behavior of each individual participant i. The second level defines the model across all participants, similar to the GLMM described by Moscatelli et al. [ 2012]. To this end, the second level estimates the overall effects across subjects by combining individual-specific effects. The parameters (a, b) describe the overall model and results from the combination of the subject-specific parameters, taking into account their uncertainties. Through a Bayesian hierarchical approach, the second level takes into account the uncertainties of the subject-specific parameters. It assumes the following distributions: Appropriate hyperprior distributions for (τα, τβ, σa, and σb) need to be specified. The precision of the overall model and the between-subjects variability are gained by the posterior estimates of the parameters τα and τβ, respectively. In the application in Section 3.1, we will discuss different prior distributions for τα and τβ, which may be different for each subject or depend on other covariates. The proposed framework provides a reliable approach to account for the uncertainty of the fixed effects parameters.
The precision and the accuracy of the response are estimated by the parameters of the model. The slope parameters βi link the inverse probit of the expected probability and the covariates x (i.e., the stimulus). Therefore, this parameter estimates the precision of the response, the higher is the estimated value of βi, the more precise is the response. The interpretation of the location parameter of the psychometric function depends on the nature of the psychophysical task. In forced-choice discrimination tasks, as mentioned in the two examples detailed in Sections 3.1 and 3.2, the PSE estimates the accuracy of the response. The response is accurate if the PSE is equal to the value of the reference stimulus. The value of the PSE relative to observer i, psei is computed from intercept and slope in Equation [2] as follows: The PSE corresponds to the stimulus value yielding a response probability of 0.5, that is, the point at which participants are equally likely to choose the standard or the comparison stimulus in response to the task. In the examples mentioned later the PSE participants are equally likely to choose one stimulus or the other to the question “which stimulus moved faster?”.
## 2.1. Prior distribution
According to the Bayesian paradigm, prior distributions and likelihood constitute a whole decision model. Ideally, a prior distribution describes the degree of belief about the true model parameters held by the scientists. If empirical data are available, then new information can coherently be incorporated via statistical models, through Bayesian learning. This process begins by documenting the available expert knowledge and uncertainty. A subjective prior describes the informed opinion of the value of a parameter before the collection of data.
Prior distributions as described in the previous paragraph are non-informative prior distributions. The flexibility of the Bayesian model allows to modify (Equations 4, 5) by considering, for example, partition or group of subjects between historical and current data. We assume that there is one relevant historical study available. However, the approaches proposed here can in principle be extended to multiple historical studies. Here, we recall the method based on the power prior proposed by Ibrahim and Chen [2000]. This has emerged as a useful class of informative priors for a variety of situations in which historical data are available (Eggleston et al., 2017).
The power prior is defined as follows Ibrahim and Chen [2000]. Suppose we have two data-sets from the current study and from a previous study that is similar to the current one, labeled as the current and the historical data, respectively. The historical data are indicated as D0 = (n0, y0, x0), while the current data are indicated as D = (n, y, x), n, and n0 are the sample size, y and y0 are the response vectors, respectively n × 1 and n0 × 1 vectors. Finally, x and x0 are (either n × p matrix or n0 × p matrix) the covariates. Let indicate θ as the vector of parameters, π0(θ) represents the initial prior distribution for θ before the historical data D0 are observed. The parameter L(θ|D) indicates a general likelihood function for an arbitrary model, such as for linear models, generalized linear model, random-effects model, non-linear model, or a survival model with censored data. Given the parameter a0, between 0 and 1, the power prior distribution of θ for the current study is defined as: This way, a0 represents the weights of the historical data relative to the likelihood of the current study. According to this definition, the parameter a0 represents the impact of the historical data on L(θ|D).
Depending on the agreement between the historical and current data, the historical data may be down-weighted, reducing the value of a0. The main question is what value of a0 to use in the analysis, which means how to assess agreement between historical and current data and how to incorporate the historical data into the analysis of a new study. The easiest solution is to establish a hierarchical power prior by specifying a proper prior distribution for a0. A uniform prior on a0 might be a good choice, or a more informative prior would be to take a Beta distribution with moderate to large parameters. Although a prior for a0 is attractive, it is much more computationally intensive than the a0 fixed case. The a0 random case has been extensively discussed (Ibrahim et al., 1999, 2015; Ibrahim and Chen, 2000; Chen and Ibrahim, 2006). Another approach, computationally more feasible, is to take a0 as fixed and elicit a specific value for it and conduct several sensitivity analyzes about this value or to take a0 as fixed and proceed, for example, with a model selection criterion.
## 2.2. Computational aspects
The large improvements in the availability of computational packages for implementing Bayesian analyzes have allowed the growth of applications of hierarchical Bayesian models. Many of the available packages permit the implementation of the Monte Carlo Markov Chain (MCMC) algorithm which saves time by avoiding technical coding. MCMC sampling is a simulation technique to generate samples from Markov chains that allow the reconstruction of the posterior distributions of the parameters. Once the posterior distributions are obtained, then the accurate and unbiased point estimates of model parameters are gained. Software for the application of Bayesian models is currently applied in several different fields (Palestro et al., 2018; Myers-Smith et al., 2019; Zhan et al., 2019; Dal'Bello and Izawa, 2021; Mezzetti et al., 2022). Gibbs sampling is an MCMC algorithm that can be implemented with the software Just Another Gibbs Sampler (JAGS), (Plummer, 2017). It is possible to interface JAGS with R using the CRAN package rjags developed by Plummer [2003]. The reader may refer to the following tutorials for fitting hierarchical Bayesian models using JAGS (or STAN) and R (Plummer, 2003; Kruschke, 2014).
Once the model is defined in JAGS, it is possible to sample from the joint posterior distributions. The mean of samples from the posterior distribution of the parameters provides the posterior estimates of the parameters of interest. From the samples of the posterior distribution, it is also possible to extract the percentile and provide the corresponding $95\%$ credible intervals.
As a diagnostic tool to assess whether the chains have converged to the posterior distribution, we use the statistic R^ (Gelman and Rubin, 1992). Each parameter has the R^ statistic associated with it (Gelman and Rubin, 1992), in the recent version (Vehtari et al., 2021); this is essentially the ratio of between-chain variance to within-chain variance (analogous to ANOVA). The R^ statistic should be approximately 1 ± 0.1 if the chain has converged.
To compare Bayesian models, different indicators can be adopted (Gelfand and Dey, 1994; Wasserman, 2000; Gelman et al., 2014). The sum of squared errors is a reasonable measure proposed. Although log-likelihood plays an important role in statistical model comparison, it also has some drawbacks, for example, the dependence on the number of parameters and on the sample size. A reasonable alternative is to evaluate a model through the log predictive density and its accuracy. Log pointwise predictive density (lppd) for a single value yi is defined as Vehtari et al. [ 2017]; The log pointwise predictive density (lppd) is defined as the sum and can be computed using results from the posterior simulation
## 3. Fitting hierarchical bayesian models to the experimental data
Studies from our research group shed light on the interplay between slip motion and high-frequency vibrations (masking vibration) in the discrimination of velocity by touch (Dallmann et al., 2015; Picconi et al., 2022; Ryan et al., 2022). These and similar results are discussed in our recent review (Ryan et al., 2021). Using Bayesian hierarchical models, we combined two of these studies and evaluated the coherence of our findings across experiments. The two studies are summarized in Sections 3.1 and 3.2, respectively. Examples of the R and JAGS files for fitting our data are available in the following Github repository https://github.com/moskante/bayesian_models_psychophysics.
## 3.1. First data-set: The role of vibration in tactile speed perception
The data-set touch-vibrations was first published by Dallmann et al. [ 2015] and it is provided within the CRAN package MixedPsy. It consists of the forced-choice responses (i.e., the comparison stimulus is “faster” or “slower” than a reference) collected in a psychophysical study from nine human observers and the corresponding predictor variables. The task is as follows: In two separate intervals, participants were requested to compare the motion speed of a moving surface by touching it and reported whether it moved faster in the reference or the comparison stimulus. The speed of the comparison stimulus was chosen among seven values of speed ranging between 1.0 and 16.0 cm/s. In two separate blocks, participants performed the task either with masking vibrations (sinusoidal wave signal at 32 Hz) or without (control condition). Each speed and vibration combination was repeated 40 times in randomized order, resulting in a total of 560 trials for each participant.
According to Dallmann et al. [ 2015], GLMM with a probit link function was fitted to the data and the results presented in Supplementary Tables S1, S2 were obtained. Next, the data were fitted with a hierarchical Bayesian model in JAGS. Let Yijh indicates the number of “faster” responses for subject i at speed xj. Superscript h indicates the presence or absence of masking vibrations. That is, $h = 0$ masking vibrations were not present while $h = 1$ masking vibrations were present. nijh is the total number of trials for subject i, speed xj and vibration condition h. The model is the following: The following set of priors are assumed: The model in Equation [10] can be parameterized as follows to allow focus on parameter PSE and the slope βih: We used the Greek letter βih and the Latin letter bh for the slope of subject i and the conditional value of slope common to all subjects, respectively. Similarly, we used the term pseih and PSEh for the estimate of the PSE in subject i and the conditional estimate.
In this first example, non-informative prior distributions were adopted and the hierarchical Bayesian model confirmed the results obtained with the GLMM, as expected. Supplementary Table S3 presents the posterior estimates of ah and bh as defined in Equations [10]–[19], while Supplementary Table S4 presents posterior estimates of PSEh as defined (Equations 20–29). Comparing Supplementary Table S2 (GLMM) and Supplementary Table S4 (Bayesian model), the PSE estimates result very close and the uncertainty is very similar with the two model approaches. Figures 1, 2 show the posterior distribution of the two parameters of the model bh and PSEh as defined in Equations [22], [23] that are common to all the subjects. The slope of the model is slightly higher without masking vibrations (b0, in blue in the figure) as compared to masking vibrations (b1, in red in the figure). The difference in PSE is negligible.
**Figure 1:** *Posterior estimates of parameters bh (slope). Experiment in Section 3.1.* **Figure 2:** *Posterior estimates of parameters PSEh. Experiment in Section 3.1.*
We considered the overlap between the posterior distributions as a measure of similarities and differences between parameters, where overlapping is defined as the area intersected by the two distributions. Overlapping was computed as the proportion of the areas of the histograms belonging to the region shared by the two distributions. The idea of overlapping as a measure of similarity among data-sets or clusters is frequently used in different fields (Pastore and Calcagǹı, 2019; Mezzetti et al., 2022).
An effect of vibration is present for the intercept. The overlap between the distribution of b0 and b1, the slope of the model, is 0.04. The overlap of the posterior distributions of PSE, in presence of vibration versus absence of vibration, is 0.58. This is consistent with our GLMM analysis where we found a small (yet significant) difference in slope but no differences in PSE.
Figures 3, 4 illustrate the posterior distributions of the parameters of the individual psychometric function, as specified in Equations [11], [22]. It is interesting to notice that between-subject variability is present for the slope (parameter βih), while subjects show similar behavior in posterior distribution respect to PSE (parameter pseih). In fact in Figure 3, the between individual variability of PSE is quite negligible. Finally, Figures 5, 6 compare the predictions of the GLMM and of the hierarchical Bayesian model across the nine participants. The predictions of the two models are almost identical. To conclude, since we used a non-informative prior, the outcome of the Bayesian model does not differ substantially from the GLMM that was used in the original study.
**Figure 3:** *Posterior estimates of individual parameters of pseih. The (left) figure illustrated with red lines represents conditions with masking vibrations, while the (right) figure illustrated with blue lines represents conditions without masking vibrations. Experiment in Section 3.1.* **Figure 4:** *Posterior estimates of individual parameters of βih. The (left) figure illustrated with red lines represents conditions with masking vibrations while the (right) figure illustrated with blue lines represents conditions without masking vibrations. Experiment in Section 3.1.* **Figure 5:** *Psychometric functions of individual participants from Experiment 1 in conditions without masking vibrations. The scatter plot shows the observed (dots) versus predicted responses (solid lines) with data from individual participants illustrated in each panel. Blue lines correspond to the prediction by GLMM, while red lines correspond to predictions by the Bayesian model. Experiment in Section 3.1.* **Figure 6:** *Psychometric functions of individual participants from Experiment 1 in conditions with 32 Hz masking vibrations. The scatter plot shows the observed (dots) versus predicted responses (solid lines) with data from individual participants illustrated in each panel. Blue lines correspond to the prediction by GLMM, while red lines correspond to predictions by the Bayesian model. Experiment in Section 3.1.*
Different specifications of the prior distributions in Equations [24], [25] and in Equations [28], [29] were considered, based on the sum of squared errors and the uncertainties of parameters, measured with the length of credible intervals. In particular, alternative specification of Equations [22]–[25] was considered: Specifically, in the model earlier, each subject can have a different precision in the two parameters of PSE and slope—i.e., τPSEi and τbi may have different values depending on the participant. The previous choice of prior distributions assumed higher variability between subjects and evidenced a different outcome in the subject NI as compared to the others with respect to the intercept and the slope. The alternative specifications of prior distributions in Equations [30]–[33] provide similar values with respect to the sum of squared errors, and the length of credible intervals for the PSE was slightly lower than the model in Equations [28], [29]. Table 1 shows the frequentist approach (GLMM) and the different specifications of the Bayesian model. Comparing the models with respect to the uncertainties in PSE estimation and model fitting, we justify the choice of the model proposed.
**Table 1**
| Model | Effects | Log likelihood | LPPD | Sum errors | 95% CI of PSE | Width CI |
| --- | --- | --- | --- | --- | --- | --- |
| GLMM | Individual | - | - | - | | |
| | Overall | –284.42 | - | 0.62 | (0.52, 0.59) | 0.07 |
| Bayesian 1 | Individual | –276.23 | –14231.6 (1081.1) | 0.42 | | |
| | Overall | | | 0.61 | (0.49, 0.55) | 0.06 |
| Bayesian 2 | Individual | –278.03 | –14323.0 (3.6) | 0.41 | | |
| | Overall | | | 0.63 | (0.49, 0.50) | 0.01 |
| Bayesian 3 | Individual | –276.29 | 14163.2 (2.0) | 0.40 | | |
| | Overall | | | 0.61 | (0.57, 0.61) | 0.04 |
| Bayesian 4 | Individual | –275.86 | -14155.8 (2.3) | 0.39 | | |
| | Overall | | | 0.61 | (0.58, 0.63) | 0.05 |
## 3.2. Second data-set: Tactile speed discrimination in people with type 1 diabetes
The second data-set, touch-diabetes, includes data from 60 human participants that were tested in a speed discrimination task similar to the one described in Section 3.1. The experimental procedure and the results are detailed by Picconi et al. [ 2022]. Participants were divided into three groups, with 20 participants per group: healthy controls, participants with diabetes with mild tactile dysfunction, and participants with diabetes with moderate tactile dysfunction. The three groups were labeled as controls, mild, and moderate, respectively. As in touch-vibration, this experiment consisted of a force-choice, speed discrimination task. In each of the 120 trials, participants were requested to indicate whether a contact surface moved faster during a comparison or a reference stimulus interval. For this experiment, a smooth surface consisting of a glass plate was used. The motion speed of the comparison stimuli were as chosen pseudo-randomly from a set of five values ranging from 0.6 to 6.4 cm/s, with the speed of the reference stimulus equal to 3.4 cm/s. Participants performed the task with and without masking vibrations, with masking stimuli consisting of sinusoidal vibrations at 100 Hz.
As in the original study, we used the GLMM in Equations [34]–[36] to fit the data across groups and across masking vibration conditions: The response variable *Yijh is* the number of “faster” responses for subject i at speed xj. The suprascript $h = 0$ represents conditions without masking vibrations and $h = 1$ represents conditions with masking vibration. The variable nijh is the total number of trials. Considering two dummy variables for the two groups of participants with diabetes, mild (indicated with subscript 2) and moderate (indicated with subscript 3) patients with diabetes, the individual model with fixed effects is rewritten as: We used the packages MixedPsy (Balestrucci et al., 2022) and lme4 (Bates et al., 2015) for model fitting. Supplementary Tables S5, S6 report results for the frequentist approach (GLMM). The slope of the model (referred to as tactile sensitivity in the study) was different across the three groups, with controls performing significantly better in the task than people with mild and moderate tactile dysfunctions. The difference between groups was larger without masking vibrations. As in the first data-set, masking vibrations reduced the values of the slope across all groups. We computed the values of PSE for all groups and conditions, see Supplementary Table S6. We expected no significant change in PSE, both between masking vibration conditions and between groups. This is because, in this task, the cues and the sensory noise are the same in the reference and comparison stimulus.
As in the previous example, we re-analyzed the data with a Bayesian hierarchical model. Let i indicates subject, j speed, h masking or no masking, and k indicates group.
Similar to the analysis of the first data-set, the model was parameterized with respect to the PSE and the slope: The following prior and hyper-prior distributions are assumed: The mean and the credible intervals of the parameters of the models bkh (slope) and PSEkh, as defined in Equations [34]–[46], are reported in Supplementary Table S7. The results confirmed the difference in slopes between the groups and between conditions. In conditions without masking vibrations, the slope was the highest in controls followed by the mild and moderate groups. The mean of the slope in controls is higher than the credible intervals of the mild group. Similarly, the mean of the slope of the mild group is higher than the credible intervals of the moderate group. The same effect can be observed in the masking vibration conditions, although the difference in slope is smaller between the control and mild groups. In Figure 7, the posterior distributions of the slope of the model are shown. We can observe the two effects of group (ordered from controls to moderate) and masking conditions. In particular, the group with moderate tactile dysfunction (illustrated in blue) is the one with the lowest values of slope.
**Figure 7:** *Posterior distributions of parameters bkh from the second stage of the hierarchical model. Experiment in Section 3.2.*
In Figure 8, the posterior distributions of the PSE values, as specified in Equations [37]–[46] are shown. Uncertainties in the parameters PSEkh were comparable between the frequentist and the Bayesian models. This was expected because in this Bayesian model, we used a non-informative prior. Masking vibrations had a large effect on the slope and a much smaller effect on the PSE. Within the control group, the overlap between the posterior distributions of PSE with masking versus no masking is 0.04, and the overlap between the posterior distribution of the slope between masking and no masking is <0.01. This supports our finding that masking vibration reduced tactile sensitivity. In Figures 9, 10, the posterior distributions of the individual parameters βi and psei are shown. Again, it is interesting to notice that the posterior estimates of PSE have low subject variability. The individual posterior distributions show a higher overlapping, refer to Figure 10 for an almost perfect overlapping. Within groups, variability is lower for PSE compared to posterior distributions of the parameters representing the slopes.
**Figure 8:** *Posterior distributions of parameters of the second stage of the hierarchical model PSEkh. Experiment in Section 3.2.* **Figure 9:** *Posterior distributions of parameters of the first stage of the hierarchical model βih, by group and masking condition. Experiment in Section 3.2.* **Figure 10:** *Posterior distributions of parameters of the first stage of the hierarchical model PSEih, by group and masking condition. Experiment in Section 3.2.*
## 4. Combined analysis of the two experiments
In this section, we propose two different approaches for the joint analysis of the two studies. In Section 4.1, the prior distributions of the parameters relative to the second study are defined from the data of the first study. In Section 4.2, a model approach based on the power prior distribution explained in Section 2.1 was applied to combine the two data-sets touch-vibrations and touch-diabetes.
The data-set touch-vibrations is considered historical data and indicated a D0 = (n0, y0, x0), where n0 is the sample size of the historical data, y0 is the number of “faster” responses the n0 × 1 response vector, in this case number of, x0 is a n0 × 1 vector of speed. The data-set touch-diabetes indicated the current study, we restrict the analysis only to the control group, we discarded the two diabetic groups because of their reduced tactile sensitivity. Data are denoted by D = (n, y, x), where n denotes the sample size, y denotes the n × 1 response vector, the number of “faster” responses, and x the n × 2 matrix of covariates, indicator of cluster and speed.
## 4.1. Prior distribution defined on the first experiment
The two data-sets are jointly analyzed. Equations [34]–[46] are rewritten incorporating model (Equations 10, 11) in order to combine the two studies as follows: Because of Weber's Law, the sensitivity to speed and, therefore, the slope depends on the value of the stimulus. To address this issue, to combine the two experiments, we used the conversion factor in Equation [55].
From the posterior estimates of parameters σah and σbh, we can gain information about whether the combination of two studies is appropriate for the same model. The posterior distributions of the precision parameters indicate a good agreement between the two studies and confirm the suitability of the choice for the prior distribution. High-posterior estimates of the precision of the prior distribution indicate good agreement between prior distribution and data.
## 4.2. Power prior model
Recalling Section 2.1, the prior distribution of parameters θ = (α, β) is defined as follows: The power parameter a0 represents the weight of the historical data relative to the likelihood of the current study. The parameters represent how much data from the previous study is to be used in the current study. There are two special cases for a0, the first case a0 = 0 results in no incorporation of the data from the previous study relative to the current study. The second case a0 = 1 results in full incorporation of the data from the previous study relative to the current study. Therefore, a0 controls the influence of the data gathered from previous studies that is similar to the current study. This control is important when the sample size of the current data is quite different from the sample size of historical data or where there is heterogeneity between two studies (Ibrahim and Chen, 2000).
In Table 2, a comparison between all the models obtained by varying the parameter a0 is shown. The choice of the value for a0 is implemented by model comparison, taking into account the log-likelihood, the log point-wise predictive density, the sum of squared errors, of both the level of the model, that are the individual and overall model. Moreover, a comparison of the uncertainty in PSE estimation is computed. The uncertainty decreases as a0 increases indicating that we are updating our informative knowledge for the correct model use. The likelihood increases as the value of a0 increases. The measures of goodness of fit of the models are very similar increasing the value of a0. We decide to favor the model that lowers the uncertainties in the estimation, that is the model with a0 = 0.7.
**Table 2**
| a 0 | Effects | Log likelihood | LPPD | Sum errors | CI of PSE | Width CI |
| --- | --- | --- | --- | --- | --- | --- |
| 0.0 | Individual | –291.76 | –3169.67 (9.91) | 2.24 | | |
| | Overall | | | 2.68 | (0.32, 0.37) | 0.05 |
| 0.1 | Individual | –292.85 | –3183.66 (8.66) | 2.24 | | |
| | Overall | | | 2.64 | (0.27, 0.26) | 0.01 |
| 0.2 | Individual | –293.47 | –3202.41 (7.18) | 2.21 | | |
| | Overall | | | 2.64 | (0.24, 0.27) | 0.03 |
| 0.3 | Individual | –294.07 | –3221.91 (8.45) | 2.21 | | |
| | Overall | | | 2.64 | (0.22, 0.27) | 0.05 |
| 0.4 | Individual | –295.08 | –3243.93 (7.02) | 2.2 | | |
| | Overall | | | 2.65 | (0.25, 0.24) | 0.01 |
| 0.5 | Individual | -295.48 | -3242.41 (8.75) | 2.2 | | |
| | Overall | | | 2.66 | (0.21, 0.25) | 0.04 |
| 0.6 | Individual | -296.00 | –3253.47 (13.02) | 2.2 | | |
| | Overall | | | 2.66 | (0.19, 0.27) | 0.08 |
| 0.7 | Individual | –296.02 | –3256.63 (8.33) | 2.23 | | |
| | Overall | | | 2.68 | (0.18, 0.22) | 0.04 |
| 0.8 | Individual | –296.84 | –3276.64 (7.70) | 2.1 | | |
| | Overall | | | 2.68 | (0.18, 0.25) | 0.07 |
| 0.9 | Individual | -296.77 | –3268.68 (12.03) | 2.23 | | |
| | Overall | | | 2.68 | (0.18, 0.2) | 0.02 |
| 1.0 | individual | –297.27 | –3268.46 (10.7) | 2.22 | | |
| | Overall | | | 2.7 | (0.17, 0.24) | 0.07 |
In Table 3, three different prior distributions are compared. On one hand, an informative prior is assumed following Section 3.2; on the other hand, the first experiment is used to improve the understanding of experiment 2. A combination of the two studies [as in Equations [47]–[59]] illustrated in Section 4.1 is compared with power prior as in Section (4.2). In Figures 11, 12, a comparison of the posterior distributions of PSE and β, in the control group, obtained according to the three different prior distributions is shown. Again we favor the model that lowers the uncertainties of posterior estimates. Overall, combining the two studies with the power prior approach reduced the posterior estimate of the model parameters as can be clearly seen by comparing the three distributions in the figures.
## 5. Conclusion
In this study, we compared the outcome of a Bayesian approach to a frequentist mixed model (GLMM) approach. The comparison showed the importance of incorporating informative prior knowledge from previous studies for data analysis.
We re-analyzed data from two studies using GLMM and Bayesian models. First, we applied GLMM and four different Bayesian models to the data-set described by Dallmann et al. [ 2015]. We compared the log-likelihood, LPPD, the sum of errors between the different models, and confidence interval of the two parameters of slope and PSE. The Bayesian approach allowed for more flexibility in the model fitting (see Table 1). Next, we applied Bayesian models to the second data-set for re-analysis of the results described by Picconi et al. [ 2022]. With a non-informative prior, the Bayesian approach confirmed the estimation of the parameters of the frequentist model. Finally, we ran a joint analysis of the two data-sets using two different approaches, either by using the first data-set to choose the parameters of the prior or by using the power prior method. The informative prior in the power prior method reduced the credible intervals of the PSE and justified the choice of the model, as shown in Tables 2, 3.
The Bayesian approach provides useful features for the in-depth analysis of psychophysical data. Through a Bayesian approach, the random effects are estimated parameters, like the fixed effects, with the advantage of obtaining credible intervals for both the quantities. This allowed to estimate the effect of individual participants and the reliability of each of them. For example, in Figure 4, it is possible to identify a single participant with increased variability and higher slope as compared to the rest of the group. Potentially, this will simplify the identification of outliers or sources of unobserved variability. Another advantage of the hierarchical Bayesian approach is the possibility to incorporate information from past studies to reduce the uncertainty of the estimate. For example, compare the width of the three distributions in Figures 11, 12, with the non-informative prior having the larger width, i.e., the higher variance. This will increase the power of the analysis. Finally, this approach allowed quantifying the coherence of multiple studies on a related topic through the parameter a0. The greater the value of a0, the higher the coherence across the studies.
Hierarchical modeling is a natural tool for combining several data-sets or incorporating prior information. In the current study, the method presented by Chen and Ibrahim [2006] has been used that provides a formal connection between the power prior and hierarchical models for the class of generalized linear models. Understanding the impact of priors on the current data and subsequently making decisions about these priors is fundamental for the interpretation of data (Koenig et al., 2021). One of the assumptions of the power prior approach is the existence of a common set of parameters for the old and current data and this assumption may not be met in practice. An alternative approach to incorporate historical data has been proposed by Neuenschwander et al. [ 2010] and van Rosmalen et al. [ 2018]. This other method is based on meta-analytic techniques (MAP) and assumes exchangeability between old and current parameters.
Incorporating previous knowledge and insight into the estimation process is a promising tool (Van de Schoot et al., 2017) that is particularly relevant in studies with small sample sizes, as is often in psychophysical experiments. In our case, the sample size of the first data-set differed from the sample size of the second data-set. To take this into account, the power prior approach allowed us to assign a different weight to the historical data and the current data. It is possible to purposefully choose the hyperparameters of the prior, τ, to increase the precision of the posterior estimate. Zitzmann et al. [ 2015] suggested to specify a slightly informative prior to the group-level variance. As shown in Section 4, diffuse priors produce results that are aligned with the likelihood. On the other hand, using an informative prior that is relatively far from the likelihood, produces a shift in the posterior. It is possible to conduct a prior sensitivity analysis to fully understand its influence on posterior estimates (Van de Schoot et al., 2017).
Uncertainty quantification is an important issue in psychophysics. Hierarchical Bayesian models allow the researcher to estimate the uncertainty at a group level and the one specific to individual participants. This model approach will have an important impact on the evaluation of psychometric functions in psychophysical data.
## Data availability statement
Publicly available datasets were analyzed in this study. This data can be found at: https://github.com/moskante/bayesian_models_psychophysics.
## Author contributions
MM: conceptualization, methodology, visualization, software, formal analysis, writing—original draft, and writing—reviewing and editing. CR: conceptualization, data curation, visualization, software, writing—original draft preparation, and writing—review and editing. PB: conceptualization, software, and writing—reviewing and editing. FL: conceptualization, data curation, and writing—reviewing and editing. AM: conceptualization, data curation, visualization, software, formal analysis, writing—original draft, and writing—reviewing and editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fncom.2023.1108311/full#supplementary-material
## References
1. Agresti A.. *Categorical Data Analysis, Vol. 482* (2002)
2. Alcalá-Quintana R., García-Pérez M. A.. **The role of parametric assumptions in adaptive bayesian estimation**. *Psychol. Methods* (2004) **9** 250. DOI: 10.1037/1082-989X.9.2.250
3. Balestrucci P., Ernst M. O., Moscatelli A.. **Psychophysics with R: the R Package MixedPsy**. *bioRxiv* (2022). DOI: 10.1101/2022.06.20.496855
4. Bates D., Mächler M., Bolker B., Walker S.. **Fitting linear mixed-effects models using lme4**. *J. Stat. Softw* (2015) **67** 1-48. DOI: 10.18637/jss.v067.i01
5. Chen M.-H., Ibrahim J. G.. **The relationship between the power prior and hierarchical models**. *Bayesian Anal* (2006) **1** 551-574. DOI: 10.1214/06-BA118
6. Dal'Bello L. R., Izawa J.. **Task-relevant and task-irrelevant variability causally shape error-based motor learning**. *Neural Netw* (2021) **142** 583-596. DOI: 10.1016/j.neunet.2021.07.015
7. Dallmann C. J., Ernst M. O., Moscatelli A.. **The role of vibration in tactile speed perception**. *J. Neurophysiol* (2015) **114** 3131-3139. DOI: 10.1152/jn.00621.2015
8. Eggleston B. S., Ibrahim J. G., Catellier D.. **Bayesian clinical trial design using markov models with applications to autoimmune disease**. *Contemp Clin. Trials* (2017) **63** 73-83. DOI: 10.1016/j.cct.2017.02.004
9. Fong Y., Rue H., Wakefield J.. **Bayesian inference for generalized linear mixed models**. *Biostatistics* (2010) **11** 397-412. DOI: 10.1093/biostatistics/kxp053
10. Foster D. H., Zychaluk K.. **Model-free estimation of the psychometric function**. *J. Vis* (2009) **9** 30-30. DOI: 10.1167/9.8.30
11. Fox J.-P., Glas C. A.. **Bayesian estimation of a multilevel irt model using gibbs sampling**. *Psychometrika* (2001) **66** 271-288. DOI: 10.1007/BF02294839
12. Gelfand A. E., Dey D. K.. **Bayesian model choice: asymptotics and exact calculations**. *J. R. Stat. Soc. B* (1994) **56** 501-514. DOI: 10.1111/j.2517-6161.1994.tb01996.x
13. Gelman A., Carlin J. B., Stern H. S., Rubin D. B.. *Bayesian Data Analysis* (1995)
14. Gelman A., Hwang J., Vehtari A.. **Understanding predictive information criteria for bayesian models**. *Stat. Comput* (2014) **24** 997-1016. DOI: 10.1007/s11222-013-9416-2
15. Gelman A., Rubin D. B.. **Inference from iterative simulation using multiple sequences**. *Stat. Sci* (1992) **7** 457-472. DOI: 10.1214/ss/1177011136
16. Houpt J. W., Bittner J. L.. **Analyzing thresholds and efficiency with hierarchical bayesian logistic regression**. *Vision Res* (2018) **148** 49-58. DOI: 10.1016/j.visres.2018.04.004
17. Ibrahim J. G., Chen M.-H.. **Power prior distributions for regression models**. *Stat. Sci* (2000) **15** 46-60. DOI: 10.1214/ss/1009212673
18. Ibrahim J. G., Chen M.-H., Gwon Y., Chen F.. **The power prior: theory and applications**. *Stat. Med* (2015) **34** 3724-3749. DOI: 10.1002/sim.6728
19. Ibrahim J. G., Chen M.-H., MacEachern S. N.. **Bayesian variable selection for proportional hazards models**. *Can. J. Stat* (1999) **27** 701-717. DOI: 10.2307/3316126
20. Johnston A., Bruno A., Watanabe J., Quansah B., Patel N., Dakin S.. **Visually-based temporal distortion in dyslexia**. *Vision Res* (2008) **48** 1852-1858. DOI: 10.1016/j.visres.2008.04.029
21. Knoblauch K., Maloney L. T.. *Modeling Psychophysical Data in R* (2012)
22. Koenig C., Depaoli S., Liu H., Van De Schoot R.. **Moving beyond non-informative prior distributions: achieving the full potential of bayesian methods for psychological research**. *Front. Psychol* (2021). DOI: 10.3389/fpsyg.2021.809719
23. Kruschke J.. *Doing Bayesian data analysis: A tutorial with R, JAGS, and Stan* (2014)
24. Kuss M., Jäkel F., Wichmann F. A.. **Bayesian inference for psychometric functions**. *J. Vis* (2005) **5** 8-8. DOI: 10.1167/5.5.8
25. Linares D., López-Moliner J.. **quickpsy: an R package to fit psychometric functions for multiple groups**. *R J* (2016) **8** 122-131. DOI: 10.32614/RJ-2016-008
26. McElreath R.. *Statistical Rethinking: A Bayesian Course With Examples in R and Stan* (2020)
27. Mezzetti M., Borzelli D., d'Avella A.. **A bayesian approach to model individual differences and to partition individuals: case studies in growth and learning curves**. *Stat. Methods Appl* (2022) **31** 1245-1271. DOI: 10.1007/s10260-022-00625-6
28. Morrone M. C., Ross J., Burr D.. **Saccadic eye movements cause compression of time as well as space**. *Nat. Neurosci* (2005) **8** 950-954. DOI: 10.1038/nn1488
29. Moscatelli A., Balestrucci P.. *Psychophysics with R: the R package MixedPsy* (2017)
30. Moscatelli A., Bianchi M., Serio A., Terekhov A., Hayward V., Ernst M. O.. **The change in fingertip contact area as a novel proprioceptive cue**. *Curr. Biol* (2016) **26** 1159-1163. DOI: 10.1016/j.cub.2016.02.052
31. Moscatelli A., Mezzetti M., Lacquaniti F.. **Modeling psychophysical data at the population-level: the generalized linear mixed model**. *J. Vis* (2012) **12** 26-26. DOI: 10.1167/12.11.26
32. Moscatelli A., Scotto C. R., Ernst M. O.. **Illusory changes in the perceived speed of motion derived from proprioception and touch**. *J. Neurophysiol* (2019) **122** 1555-1565. DOI: 10.1152/jn.00719.2018
33. Myers-Smith I. H., Grabowski M. M., Thomas H. J., Angers-Blondin S., Daskalova G. N., Bjorkman A. D.. **Eighteen years of ecological monitoring reveals multiple lines of evidence for tundra vegetation change**. *Ecol. Monogr* (2019). DOI: 10.1002/ecm.1351
34. Neuenschwander B., Capkun-Niggli G., Branson M., Spiegelhalter D. J.. **Summarizing historical information on controls in clinical trials**. *Clin. Trials* (2010) **7** 5-18. DOI: 10.1177/1740774509356002
35. Palestro J. J., Bahg G., Sederberg P. B., Lu Z.-L., Steyvers M., Turner B. M.. **A tutorial on joint models of neural and behavioral measures of cognition**. *J. Math. Psychol* (2018) **84** 20-48. DOI: 10.1016/j.jmp.2018.03.003
36. Pariyadath V., Eagleman D.. **The effect of predictability on subjective duration**. *PLoS ONE* (2007) **2** e1264. DOI: 10.1371/journal.pone.0001264
37. Pastore M., Calcagnì A.. **Measuring distribution similarities between samples: a distribution-free overlapping index**. *Front. Psychol* (2019). DOI: 10.3389/fpsyg.2019.01089
38. Pelli D. G., Farell B.. **Psychophysical methods**. *Handbook Optics* (1995) **1** 29-21
39. Picconi F., Ryan C., Russo B., Ciotti S., Pepe A., Menduni M.. **The evaluation of tactile dysfunction in the hand in type 1 diabetes: a novel method based on haptics**. *Acta Diabetol* (2022) **59** 1073-1082. DOI: 10.1007/s00592-022-01903-1
40. Plummer M.. **“Jags: a program for analysis of bayesian graphical models using gibbs sampling,”**. *Proceedings of the 3rd International Workshop on Distributed Statistical Computing, Vol. 124* (2003) 1-10
41. Plummer M.. *Jags Version 4.3. 0 User Manual [computer software manual]* (2017)
42. Prins N.. *Psychophysics: A Practical Introduction* (2016)
43. Prins N., Kingdom F. A.. **Applying the model-comparison approach to test specific research hypotheses in psychophysical research using the palamedes toolbox**. *Front. Psychol* (2018). DOI: 10.3389/fpsyg.2018.01250
44. Rouder J. N., Lu J.. **An introduction to bayesian hierarchical models with an application in the theory of signal detection**. *Psychon. Bull. Rev* (2005) **12** 573-604. DOI: 10.3758/BF03196750
45. Rouder J. N., Sun D., Speckman P. L., Lu J., Zhou D.. **A hierarchical bayesian statistical framework for response time distributions**. *Psychometrika* (2003) **68** 589-606. DOI: 10.1007/BF02295614
46. Ryan C. P., Bettelani G. C., Ciotti S., Parise C., Moscatelli A., Bianchi M.. **The interaction between motion and texture in the sense of touch**. *J. Neurophysiol* (2021) **126** 1375-1390. DOI: 10.1152/jn.00583.2020
47. Ryan C. P., Ciotti S., Cosentino L., Ernst M. O., Lacquaniti F., Moscatelli A.. **Masking vibrations and contact force affect the discrimination of slip motion speed in touch**. *IEEE Trans. Haptics* (2022) **15** 693-704. DOI: 10.1109/TOH.2022.3209072
48. Schütt H. H., Harmeling S., Macke J. H., Wichmann F. A.. **Painfree and accurate bayesian estimation of psychometric functions for (potentially) overdispersed data**. *Vision Res* (2016) **122** 105-123. DOI: 10.1016/j.visres.2016.02.002
49. Steele F., Goldstein H.. **“12 multilevel models in psychometrics,”**. *Psychometrics, Volume 26 of Handbook of Statistics* (2006) 401-420
50. Stroup W. W.. *Generalized Linear Mixed Models: Modern Concepts, Methods and Applications* (2012)
51. Van de Schoot R., Winter S. D., Ryan O., Zondervan-Zwijnenburg M., Depaoli S.. **A systematic review of bayesian articles in psychology: the last 25 years**. *Psychol. Methods* (2017) **22** 217. DOI: 10.1037/met0000100
52. van Rosmalen J., Dejardin D., van Norden Y., Löwenberg B., Lesaffre E.. **Including historical data in the analysis of clinical trials: Is it worth the effort?**. *Stat. Methods Med. Res* (2018) **27** 3167-3182. DOI: 10.1177/0962280217694506
53. Vehtari A., Gelman A., Gabry J.. **Practical bayesian model evaluation using leave-one-out cross-validation and waic**. *Stat. Comput* (2017) **27** 1413-1432. DOI: 10.1007/s11222-016-9696-4
54. Vehtari A., Gelman A., Simpson D., Carpenter B., Bürkner P.-C.. **Rank-normalization, folding, and localization: an improved r for assessing convergence of mcmc (with discussion)**. *Bayesian Anal* (2021) **16** 667-718. DOI: 10.1214/20-BA1221
55. Wang X., Bradlow E. T., Wainer H.. **A general bayesian model for testlets: Theory and applications**. *Appl. Psychol. Meas* (2002) **26** 109-128. DOI: 10.1177/0146621602026001007
56. Wasserman L.. **Bayesian model selection and model averaging**. *J. Math. Psychol* (2000) **44** 92-107. DOI: 10.1006/jmps.1999.1278
57. Zhan P., Jiao H., Man K., Wang L.. **Using jags for bayesian cognitive diagnosis modeling: a tutorial**. *J. Educ. Behav. Stat* (2019) **44** 473-503. DOI: 10.3102/1076998619826040
58. Zhao Y., Staudenmayer J., Coull B. A., Wand M. P.. **General design bayesian generalized linear mixed models**. *Statist. Sci* (2006) **21** 35-51. DOI: 10.1214/088342306000000015
59. Zitzmann S., Lüdtke O., Robitzsch A.. **A bayesian approach to more stable estimates of group-level effects in contextual studies**. *Multivariate Behav. Res* (2015) **50** 688-705. DOI: 10.1080/00273171.2015.1090899
|
---
title: Global, regional, and national burden of chronic kidney disease attributable
to high sodium intake from 1990 to 2019
authors:
- Wei Liu
- Lingyun Zhou
- Wenjun Yin
- Jianglin Wang
- Xiaocong Zuo
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC10018037
doi: 10.3389/fnut.2023.1078371
license: CC BY 4.0
---
# Global, regional, and national burden of chronic kidney disease attributable to high sodium intake from 1990 to 2019
## Abstract
### Background
High sodium intake is a crucial risk factor for the development and progression of chronic kidney disease (CKD). However, the latest global spatiotemporal patterns of CKD burden attributable to high sodium intake still remain unclear. We aimed to evaluate the level and trends of the CKD burden associated with high sodium intake according to sex, age, socio-demographic index (SDI), region, and country from 1990 to 2019.
### Methods
Data on CKD burden attributable to high sodium intake from 1990 to 2019 were extracted from the Global Burden of Disease (GBD) Study 2019. The CKD-related deaths, disability-adjusted life years (DALYs), age-standardized mortality rate (ASMR), and age-standardized DALYs rate (ASDR) attributable to high sodium intake were estimated by age, sex, SDI, region, and country. The estimated annual percentage change (EAPC) was calculated to evaluate the secular trends of ASMR and ASDR of CKD attributable to high sodium intake from 1990 to 2019. We further explored the associations of SDI with the ASMR and ASDR of CKD attributable to high sodium intake.
### Results
Globally, the number of CKD-related deaths and DALYs attributable to high sodium intake were 45,530 ($95\%$ UI: 12,640 to 93,830) and 1.32 million ($95\%$ UI: 0.43 to 2.8) in 2019, both twice as many as those in 1990. However, the ASMR and ASDR slightly grew, with an EAPC of 0.22 ($95\%$ CI: 0.16 to 0.28) and 0.10 ($95\%$ CI: 0.04 to 0.16), respectively. The age-specific numbers and rates of deaths, as well as DALYs of CKD attributable to high sodium intake, rose with age and were greater in males than in females. The rates of deaths and DALYs peaked in the >95 age group for both females and males in 2019. From 1990 to 2019, the trends of both age-specific rates of mortality and DALYs of CKD attributable to high sodium intake were down in people under 60, while in people over 60, the trends were the opposite. The burden of CKD attributable to high sodium intake in 2019 and its temporal trends from 1990 to 2019 varied greatly by SDI quintile and geographic location. The ASMR or ASDR showed a non-linear negative correlation with SDI at the regional level. The EAPC in ASMR or ASDR showed a markedly negative correlation with ASMR or ASDR in 1990, with a coefficient of −0.40. Nevertheless, the EAPC in ASMR rather than ASDR was positively correlated with SDI in 2019, with a coefficient of 0.18.
### Conclusion
Our findings suggest that there are significant sexual and geographic variations in the burden of CKD attributable to high sodium intake and its temporal trends. Globally, the high sodium intake-caused CKD burden continues to elevate, posing a major challenge to public health. In response to this, strengthened and tailored approaches for CKD prevention and sodium intake management are needed, especially for elderly populations, males, and the population in the middle SDI regions.
## Introduction
Chronic kidney disease (CKD) has been recognized as a major global public health concern, affecting nearly one in ten individuals. The Global Burden of Disease (GBD) Study 2019 estimated that there are approximately 697 million CKD cases globally, with 41.5 million disability-adjusted life years (DALYs) resulting from CKD [1]. As the eleventh leading cause of morbidity and mortality worldwide in 2019, CKD accounted for more than 1.43 million deaths, and this number was projected to reach 4.0 million in 2040 in the worst-case scenario [1, 2]. From birth, the overall lifetime risk for developing CKD stage 3a or higher, stage 3b or higher, stage 4 or higher, and end-stage renal disease (ESRD) is almost 59.1, 33.6, 11.5, and $3.6\%$, respectively [3]. The lifetime risk of stage 3 CKD or more advanced CKD among patients aged 45 is $35.8\%$ for women and $21.3\%$ for men [4]. Furthermore, the risk of developing stage 3a or more advanced CKD increases dramatically with age [3]. A potential outcome of CKD is ESRD, which will result in many complications. Patients with CKD or ESRD experience worse clinical outcomes, including impaired life quality, increased medical costs, and a greater economic burden (5–7). In the USA, the total annual direct costs attributable to CKD are approximately USD 49 billion [8]. In China, according to the China Kidney Disease Network 2015 Annual Data Report, hospital admissions for CKD resulted in approximately USD 3430 million in medical costs, $6.3\%$ of the overall costs [9].
In recent decades, modifiable risk factors, e.g., impaired fasting plasma glucose, hypertension, obesity, and smoking, have been regarded as the major causes of CKD DALYs and deaths [10, 11]. Effective monitoring and management of modifiable risk factors, such as unhealthy lifestyle habits, have been identified as being cost-effective for the prevention and intervention of CKD or ESRD. High sodium intake is widely acknowledged as a major modifiable risk factor for the development and progression of CKD (12–14). Accumulating evidence suggests that excess sodium intake may cause CKD or ESRD through multiple pathways such as increased blood pressure, fluid retention, proteinuria, triggering inflammatory responses, causing oxidative damage, and endothelial dysfunction [15, 16]. CKD patients are particularly sensitive to excess sodium due to decreased sodium excretion with progressive kidney decline. Therefore, limiting the sodium intake of CKD patients is important. The Kidney Disease: Improving Global Outcomes (KDIGO) guidelines recommend that individuals with CKD consume less than 5 g of salt (2 g or 87 mmol sodium) per day [17].
Although behavior changes interventions such as awareness campaigns and health education programs are implemented to reduce sodium intake, the average sodium intake in CKD patients from salt per day [approximately 9–12 g (150–200 mmol sodium)] far exceeds the recommended level [13, 18]. Growing studies have reported the association between high sodium intake and the risk of the development and progression of CKD (12, 19–22). However, the epidemiological characteristics and dynamic trend of the global CKD burden attributable to high sodium intake at global, regional, and national levels are still unclear. Although the overall disease burden attributable to high sodium intake at global, regional, and national levels has been previously reported using data from the GBD study 2019, the temporal and spatial patterns of specific-disease burden attributable to high sodium intake are not considered in this study, and these may differ substantially from the overall disease burden due to high sodium intake patterns [23]. Understanding the epidemiological characteristics of the burden of CKD attributable to high sodium intake and its spatiotemporal changes worldwide can contribute to target actions for the global prevention and control of CKD. On this basis, we used the latest GBD Study 2019, a multi-national collaborative and updated worldwide epidemiological research estimating the disease burden for 204 countries and territories worldwide, to conduct a comprehensive analysis of the epidemiological characteristics and trend of CKD attributable to high sodium intake.
## Study data
Data on the burden of CKD attributable to high sodium intake were retrieved from the Global Health Data Exchange GBD Results Tool [24] by GBD collaborators for a systematic assessment of the age-and sex-specific mortality of 286 causes, the prevalence and DALYs of 369 diseases and injuries, and the comparative risks of 87 risk factors in 204 countries and territories, from 1st January 1990 to 31th December 2019. Previous studies have described the GBD research methods of analysis [1, 11]. We retrieved data on annual deaths, DALYs, age-standardized mortality rate (ASMR), and age-standardized DALY rate (ASDR) of CKD attributable to high sodium intake based on sex, age (5-year age groups of patients aged 25–94 and ≥ 95) in 204 countries, and region from 1990 to 2019. A total of 204 involved countries and regions were divided into five super regions based on the socio-demographic index (SDI) quintile, namely, the low SDI region, the low-middle SDI region, the middle SDI region, the high-middle SDI region, and the high SDI region. SDI is a composite indicator of the development status of a geographical location. It was calculated by combining the lag-distributed income per capita, educational attainment for those aged 15 and above, and the total fertility rate among females aged below 25 [1, 10, 11, 25]. In addition, the world was further categorized into 21 geographic regions based on epidemiological similarity and geographical proximity in GBD 2019 [1, 10, 11, 25].
## Definitions of CKD and high sodium intake exposure
CKD was defined as a permanent abnormality of kidney function, indicated by a decreased estimated glomerular filtration rate (eGFR) based on a serum creatinine measurement and/or an elevated urine albumin-to-creatinine ratio (ACR) [10]. This definition was different from that presented in the KDIGO 2012 Clinical Practice Guidelines [26]. In GBD, since CKD was defined only based on one measurement of eGFR and ACR, the requirement in KDIGO that abnormalities should last at least 3 months was not fulfilled. In addition, markers of kidney damage other than ACR were not considered in the GBD definition since they are not often reported in epidemiological studies on disease occurrence estimation. ICD 9 codes (250.4, 403–404.9, 581–583.9, 585–585.9, 589–589.9, 753–753.3) mapped to CKD and ICD 10 codes (D63.1, E10.2, E11.2, E12.2, E13.2, E14.2, I12-I13.9, N02-N08.8, N15.0, N18-N18.9, Q61-Q62.8) mapped to CKD were used to model CKD mortality. High sodium intake was defined as an average 24-h urinary sodium excretion (in grams per day) greater than 1–5 g [27], and its assessment, such as detailed information about the data selection and data inputs, has been described in previous studies [23, 27, 28].
## Estimation of high sodium intake-attributed CKD burden
*The* general methods of the GBD and the specific methods for estimating the burden of CKD attributable to high sodium intake have been elucidated elsewhere [1, 10, 11, 25, 27]. Here, we briefly described specific approaches to the estimation of CKD burden due to high sodium intake. The raw data from surveys, censuses, vital statistics, and other health-related data sources were processed and standardized, then the processed and standardized data were modeled using three primary standardized tools (disease model-Bayesian meta-regression (DisMod-MR) 2.1, Cause of Death Ensemble model (CODEm), and spatial–temporal Gaussian process regression (ST-GPR)) to produce the estimates of prevalence, incidence, remission, and excess mortality by age, sex, location, and year. The estimates of deaths attributable to CKD were multiplied by the estimates of standard life expectancy by age to generate the estimates of years of life lost (YLLs) for CKD. The years lived with disability (YLDs) were estimated by multiplying the prevalence of each CKD sequela by its corresponding disability weight. YLDs and YLLs for each CKD cause were summed to estimate DALYs.
Data on high sodium intake exposure were extracted from multiple sources, including nationally and subnationally representative nutrition surveys, household budget surveys, accounts of national sales from the Euromonitor, and availability data from the United Nations Food and Agriculture Organization Supply and Utilization Accounts. Ninety two original data sources from 53 countries of high sodium intake exposure were used in this dietary risk modeling in GBD 2019.
A comparative risk assessment framework was used to calculate the fraction of disease burden of CKD attributable to high sodium intake exposure [1, 11]. The following primary steps were included in the framework. The first was to determine the risk factor for high sodium intake that had convincing or probable evidence for a causal association. The second was to summarize the relative risks for the high sodium intake-CKD outcome pair as a function of exposure based on the systematic reviews and meta-regression. The third was to model the high sodium intake exposure levels and distributions for each age, sex, location, and year. Either Bayesian meta-regression modeling or spatial–temporal Gaussian process regression was used to model the high sodium intake exposure levels. The fourth was to identify the theoretical minimum risk exposure level as the exposure level associated with minimum risk, which was determined from published observational studies and trials. Fifth, the population attributable fraction (PAF) by age, sex, location, and year were calculated to assess the burden of CKD attributable to high sodium intake. This calculation considered the risk function (i.e., relative risk), exposure level, and the theoretical minimum risk exposure level. The standard GBD PAF equation is defined as follows: PAFasgt=∑x=luRRasg(x)Pasgt(x)−RRasg(TMRELas)∑x=luRRas(x)Pasgt(x) where PAFasgt was the PAF for CKD burden attributable to high sodium intake for age group a, sex s, location g, and year t. RRast (x) was the relative risks between exposure level x (from l to u) of high sodium intake and CKD for age group a, sex s, and year t; and Pasgt(x) was the proportion of the population exposed to high sodium intake at the level x for age group a, sex s, location g, and year t. TMRELas is the TMREL for age group a, and sex s. The sixth was to account for the potential mediating effect. Last, the YLLs, YLDs, and deaths for CKD were multiplied by the high sodium intake risk factor PAF to estimate the CKD burden attributable to high sodium intake.
## Statistical analyses
Data on deaths, DALYs, ASMR, and ASDR were reported as numbers with $95\%$ uncertainty intervals (UIs) based on the 2.5th and 97.5th percentiles of the ordered 1,000 estimations [1]. The number of deaths, DALYs, and ASMR and ASDR were computed to quantify the burden of CKD attributable to high sodium intake by age, sex, year, and location. The secular trends of ASMR and ASDR of high sodium intake-attributed CKD from 1990 to 2019 were calculated using an estimated annual percentage change (EAPC), which was widely accepted to reflect the trend of age-standardized rates (ASR) over a time interval [29]. The ASR could be fitted in a regression model ln(ASR)=α+βx+ε, where x denoted the calendar year. Then, EAPC could be obtained from the model 100×(exp(β)−1), and its $95\%$ CI [29]. If the lower limit of the $95\%$ CI of the corresponding EAPC estimation was greater than zero, the ASR (i.e., ASMR and ASDR) was considered to be increased. Conversely, if the upper limit of the $95\%$ CI of the corresponding EAPC estimation was lower than zero, the ASR represented a decreasing trend. The ASR would be regarded as stable if the $95\%$ CI included zero. Smoothing spline models were performed to examine the shape of the correlation between SDI and CKD burden attributable to high sodium intake measured as ASMR and ASDR for 21 regions. To explore the influential factors for the EAPC of the burden rate of CKD attributable to high sodium intake, we estimated the relationships between ASMR or ASDR in 1990, the SDI in 2019, and the EAPC in ASMR or ASDR using the Spearman rank test [30, 31]. All statistical analyses were conducted using R software (version 4.1.2). A two-sided p value of less than 0.05 was considered statistically significant.
## Global burden of CKD attributable to high sodium intake from 1990 to 2019
In 2019, the global number of deaths and DALYs attributable to high sodium were estimated at 45.53 × 103 and 1318.81 × 103, representing $6.6\%$ (1.4 to $15.7\%$) and $6.1\%$ (1.5 to $14.0\%$) of all CKD-related deaths and DALYs, respectively (Tables 1, 2). From 1990 to 2019, the number of global deaths and DALYs attributable to high sodium have more than doubled for both females and males, with a male-to-female ratio of approximately 1.4 in CKD deaths and DALYs (Tables 1, 2). The global ASMR of CKD attributable to high sodium intake decreased from 0.89 ($95\%$ UI: 0.20 to 2.08) per 100,000 population in 1990 to 0.87 ($95\%$ UI: 0.13 to 2.29) per 100,000 population in 2019 for females; in contrast, it increased from 1.58 ($95\%$ UI: 0.47 to 3.43) per 100,000 population in 1990 to 1.64 ($95\%$ UI: 0.38 to 3.71) per 100,000 population in 2019 for males. The global ASDR of CKD attributable to high sodium intake was lessened from 25.27 ($95\%$ UI: 6.95 to 56.47) per 100,000 population in 1990 to 23.20 ($95\%$ UI: 4.46 to 57.98) per 100,000 population in 2019 for females; conversely, it was increased from 40.57 ($95\%$ UI: 13.79 to 82.87) per 100,000 population in 1990 to 40.81 ($95\%$ UI: 11.14 to 87.39) per 100,000 population in 2019 for males. The percentage change in the global CKD-related ASMR and ASDR attributable to high sodium intake from 1990 to 2019 are presented in Supplementary Tables 1, 2 by region and country. In addition, from 1990 to 2019, the global CKD-related ASMR due to high sodium intake remained stable for females but showed an average increase of $0.30\%$ ($95\%$ CI: 0.24 to $0.37\%$) per year for males. The global ASDR of high sodium intake-caused CKD showed an average annual decline of $0.18\%$ ($95\%$ CI: –0.23 to −$0.13\%$) for females and an average annual increase of $0.24\%$ ($95\%$ CI: 0.17 to $0.32\%$) for males. However, it was worth noting that both the global CKD-associated ASMR and ASDR attributable to high sodium intake slightly increased, with an EAPC of 0.22 ($95\%$ CI: 0.16 to 0.28) and 0.10 ($95\%$ CI: 0.04 to 0.16), respectively.
At the SDI regional level, the middle SDI regions had the largest number of deaths and DALYs related to CKD attributable to high sodium intake, while these indicators were the lowest in the low SDI regions in both 1990 and 2019 (Tables 1, 2). In 2019, the proportion of all CKD-related deaths and DALYs attributable to high sodium intake ranged from 5.2 to $7.9\%$. The high-middle SDI regions showed the largest PAF of all CKD-related deaths and DALYs due to high sodium intake, followed by the low-middle SDI, the middle SDI, the high SDI, and the low SDI regions (Tables 1, 2). The middle SDI regions had the largest ASMR and ASDR of CKD attributable to high sodium intake in both 1990 and 2019, whereas the high SDI regions had the lowest ASMR and ASDR (Tables 1, 2). However, the annual trends of changes in ASMR and ASDR differed by SDI quintile. Between 1990 and 2019, the ASMR in the middle and high-middle SDI regions remained stable, while in the high and low-middle SDI regions increased; however, the low SDI regions showed a decrease in ASMR (Tables 1, 2). Regarding ASDR, the high and high-middle SDI regions showed a stable trend, the middle SDI and low SDI regions displayed a decreasing trend, and the increases during this period were found in the low-middle SDI regions (Tables 1, 2).
At the regional level, the heaviest burden of CKD attributable to high sodium intake occurred in East Asia from 1990 to 2019, with deaths and DALYs accounting for 41.58 and $36.82\%$ in 2019, respectively. The proportion of all CKD-related deaths and DALYs attributable to high sodium intake in 2019 ranged from 1.4 to $15.0\%$ and from 1.4 to $14.2\%$, respectively (Tables 1, 2). North Africa and Middle East (deaths: 1.4, $95\%$ UI: 0.3 to $5.8\%$; DALYs: 1.4, $95\%$ UI: 0.3 to $5.4\%$), Australasia (deaths: 2.2, $95\%$ UI: 0.2 to $8.9\%$; DALYs: 2.3, $95\%$ UI: 0.2 to $8.6\%$), Central Sub-Saharan Africa (deaths: 2.7, $95\%$ UI: 0.2 to $10.9\%$; DALYs: 2.2, $95\%$ UI: 0.2 to $9.4\%$) had the three lowest PAFs of CKD-related deaths and DALYs due to high sodium intake, while the highest were found in Central Europe (deaths: 15.0, $95\%$ UI: 5.0 to $26.8\%$; DALYs:11.8, $95\%$ UI: 3.8 to $21.6\%$), East Asia (deaths: 13.7, $95\%$ UI: 4.9 to $24.9\%$; DALYs:14.2, $95\%$ UI: 6.1 to $24.1\%$). Moreover, among GBD 2019 regions, the ASMR and ASDR were the highest in Southeast Asia (ASMR: 2.57 per 100,000 population, $95\%$ UI: 0.39 to 5.77; ASDR: 61.76 per 100,000 population, $95\%$ UI: 10.11 to 135.43) and Central Latin America (ASMR: 2.81 per 100,000 population, $95\%$ UI: 0.29 to 7.53; ASDR: 68.32 per 100,000 population, $95\%$ UI: 7.16 to 180.66). In contrast, Eastern Europe (ASMR: 0.22 per 100,000 individuals, $95\%$ UI: 0.02 to 0.64; ASDR: 8.08 per 100,000 individuals, $95\%$ UI: 0.88 to 21.36), Western Europe (ASMR: 0.36 per 100,000 individuals, $95\%$ UI: 0.03 to 1.20; ASDR: 6.66 per 100,000 individuals, $95\%$ UI: 0.55 to 21.37) and Australasia (ASMR: 0.24 per 100,000 individuals, $95\%$ UI: 0.02 to 0.96; ASDR: 4.90 per 100,000 individuals, $95\%$ UI: 0.46 to 18.42) had the three lowest ASMR and ASDR. From 1990 to 2019, the percentage change in the ASMR and ASDR of CKD attributable to high sodium intake differed substantially among regions (Supplementary Tables 1, 2). High-income Asia-*Pacific area* showed the largest decrease in the ASMR and ASDR of CKD attributable to high sodium intake, followed by Eastern Sub-Saharan Africa, Southeast Asia, and Central Europe. In contrast, high-income North America showed the largest increase over the same measurement period, followed by Central Latin America (Supplementary Tables 1, 2). In addition, the annual changing trends in ASMR and ASDR from 1990 to 2019 were not similar across the GBD 2019 regions (Tables 1, 2). High-income North America presented the most significant annual increasing trends in ASMR and ASDR, with an EAPC of 3.28 ($95\%$ CI: 3.01 to 3.55) in ASMR and an EAPC of 3.17 ($95\%$ CI: 2.85 to 3.49) in ASDR. Moreover, the largest decline in the annual changing trends in ASMR and ASDR was found in the high-income Asia-Pacific region, with an EAPC of −3.05 ($95\%$ CI: −3.23 to −2.88) in ASMR and −3.06 ($95\%$ CI, −3.25 to −2.88) in ASDR (Tables 1, 2).
At a national level, China, followed by India, had the greatest number of deaths and DALYs of CKD attributable to high sodium intake in both 1990 and 2019—nearly half of the global level; in contrast, Nauru, Niue, San Marino, Tokelau, and Monaco showed the lowest level (Supplementary Tables 3–6). In 2019, the proportion of all CKD-related deaths and DALYs attributable to high sodium intake differed substantially by country. Serbia, Slovenia, Hungary had the three highest PAFs of CKD-related deaths attributable to high sodium intake. While China, Serbia, and Hungary showed the highest PAFs of CKD-related DALYs attributable to high sodium intake. In contrast, the lowest PAFs of CKD-related deaths and DALYs attributable to high sodium intake were found in Turkey (Supplementary Tables 1, 2). In 1990, the Maldives and Lao People’s Democratic Republic had the largest and second-largest ASMR and ASDR of CKD attributable to high sodium intake; in 2019, the top two countries in terms of ASMR and ASDR were Mauritius and Nicaragua (Figures 1A,B; Supplementary Tables 7–10). The countries or territories with the lowest ASMR and ASDR of CKD attributable to high sodium intake in 1990 were Estonia, Ukraine, and Belarus; they became Belarus and Ukraine in 2019 (Figures 1A,B; Supplementary Tables 7–10). Increases in the percentage changes in the ASMR and ASDR of CKD attributable to high sodium intake were found in several countries and territories throughout the study. From 1990 to 2019, Estonia and El Salvador experienced the greatest increases in the ASMR and ASDR (Supplementary Tables 1, 2). In contrast, the high-income Asia-Pacific, Ethiopia, Japan, the Maldives, and Mongolia presented the largest reduction over this period (Supplementary Tables 1, 2). Between 1990 and 2019, the largest increase in the annual changing trends in ASMR of CKD attributable to high sodium intake was found in Estonia, Austria, El Salvador, Latvia, and the United States of America (Figure 1C; Supplementary Table 11); El Salvador, Pakistan, the United States of America, Austria, and Estonia had the largest increase in the annual changing trends in ASDR (Figure 1D; Supplementary Table 12). In contrast, the sharpest declines in the annual changing trends in ASMR and ASDR of CKD attributable to high sodium intake over the study period were found in Mongolia, Japan, Maldives, Rwanda, and Ethiopia (Figures 1C,D; Supplementary Tables 11, 12).
**Figure 1:** *The global disease burden of chronic kidney disease attributable to high sodium intake for both sexes combined in 204 countries and territories. (A) The spatial distribution of chronic kidney disease ASMR attributable to high sodium intake in 2019. (B) The spatial distribution of chronic kidney disease ASDR attributable to high sodium intake in 2019. (C) The EAPC in chronic kidney disease ASMR attributable to high sodium intake from 1990 to 2019. (D) The EAPC in chronic kidney disease ASDR attributable to high sodium intake from 1990 to 2019. ASMR age-standardized mortality rate; DALYs, disability-adjusted life years; ASDR, age-standardized DALYs rate; EAPC, estimated annual percentage.*
## Global burden of CKD attributable to high sodium intake by age and sex
In 2019, the number of CKD deaths attributable to high sodium intake peaked in males aged 65–69 and in females aged 75–79 (Figure 2A). The number of CKD-related DALYs attributable to high sodium intake followed a normal distribution and peaked in groups aged 65–69 in both sexes (Figure 2B). Furthermore, the number of CKD-related deaths and DALYs attributable to high sodium intake were higher in males than in females up to the ages of 85–89 years, whereas for those aged 90 and above, the number of deaths and DALYs were higher in females than in males (Figures 2A,B). Additionally, age-specific rates of CKD-related deaths and DALYs due to high sodium intake showed a non-linear increase with age for females and males (Figures 2A,B). Both CKD-related death and DALY rates attributable to high sodium intake were greater in males than in females across all age groups, and the difference in the death rates between males and females intensified with age (Figures 2A,B).
**Figure 2:** *Age-specific numbers and rates of chronic kidney disease deaths and DALYs attributable to high sodium intake by sex, in 2019. (A) Deaths. (B) DALYs. DALYs, disability-adjusted life years.*
CKD-related deaths and DALYs attributable to high sodium intake in the globe or regions with different SDI from 1990 to 2019 were mainly found in individuals aged 60–79 and 50–75, respectively (Supplementary Figures S1A,B) and were sharply increasing (Supplementary Figures S1A,B). From 1990 to 2019, the global EAPCs in age-specific mortality rate and DALYs rate showed an approximately linear increase with age; the trend of age-specific mortality rate and DALYs rate was decreasing in groups aged below 59 and increasing in groups aged above 60 for both females and males (Supplementary Figures S2A,B). In the high-middle SDI and the middle SDI regions, the EAPCs in age-specific mortality rate and DALYs rate were also approximately linearly increased with age; the trend of age-specific mortality rate was downward in individuals aged 25–64 and upward in individuals aged above 65 from 1990 to 2019; and the trend of age-specific DALYs rate was declining in individuals aged 25–59, while in people over 60, the trend were the opposite (Figures 3A,B). From 1990 to 2019, the trend of age-specific mortality rate and DALYs rate in low-middle SDI regions was increasing in individuals aged above 50 but decreasing in those aged 25–54, among which the fastest increase and reduction occurred in those aged 95 and 25–29, respectively (Figures 3A,B). In the low SDI regions, the trend of age-specific mortality rate and DALYs rate was down in groups aged 25–89 from 1990 to 2019, except the group aged 30–34. In contrast, the age-specific mortality rate and DALYs rate rose in those aged above 90 (Figures 3A,B). From 1990 to 2019, the trend of age-specific mortality rate in the high SDI regions was decreasing in individuals under 75 and then increasing in people over 75 (Figure 3A). The EAPC in age-specific DALYs rate showed a similar pattern to that in age-specific mortality rate (Figure 3B).
**Figure 3:** *The age distribution of the trends in chronic kidney disease-related mortality rate and DALYs rate attributable to high sodium intake from 1990 to 2019 by location. (A) EAPC in mortality rate. (B) EAPC in DALYs rate. DALYs, disability-adjusted life years; EAPC, estimated annual percentage change.*
## Factors associated with the burden of CKD attributable to high sodium intake
At the regional level, ASMR and ASDR had non-linear associations with SDI between 1990 and 2019. The ASMR and ASDR showed an approximately linear decrease with SDI improvement at SDI < 0.4 and SDI > 0.6. In contrast, the ASMR and ASDR remained stable at 0.4 < SDI < 0.6. Despite a declining trend of ASMR and ASDR in high-income Asia Pacific, Southeast Asia, East Asia, Central Europe, and Eastern Sub-Saharan Africa, these regions still showed a higher ASMR and ASDR than expected values during the measurement period (Figures 4A,B). At the global level, in Eastern Europe, Western Europe, Australasia, North Africa, the Middle East, tropical Latin America, South Asia, as well as the Western and Central Sub-Saharan Africa regions, the observed burden estimates of CKD attributable to high sodium intake were stable and lower than expected levels based on the SDI between 1990 and 2019 (Figures 4A,B). The ASMR and ASDR in the rest regions showed an intermittent increase and decrease with SDI improvement (Figures 4A,B).
**Figure 4:** *Chronic kidney disease-related ASMR and ASDR attributable to high sodium intake across 21 Global Burden of Disease regions by socio-demographic index for both sexes combined, 1990–2019. For each region, points right depict estimates from each year from 1990 to 2019. (A) The correlation between chronic kidney disease-related ASMR attributable to high sodium intake and socio-demographic index. (B) The correlation between chronic kidney disease-related ASDR attributable to high sodium intake and socio-demographic index. ASMR, age-standardized mortality rate; DALY, disability-adjusted life year; ASDR, age-standardized DALY rate.*
The ASMR, or ASDR, of CKD attributable to high sodium intake in 1990 reflected the disease reservoir at baseline. Moreover, the SDI in 2019 served as an indicator of each country’s improvement level. As shown in Figures 5A,B, EAPC had significant associations with ASMR or ASDR (in 1990), and there was also a significant association between EAPC and SDI (in 2019). The EAPC in ASMR and ASDR showed a significant negative correlation with corresponding ASMR (ρ = −0.414, $p \leq 0.001$) and ASDR (ρ = −0.406, $p \leq 0.001$) of CKD attributable to high sodium intake in 1990, respectively (Figures 5A,B). However, the EAPC in ASMR showed a slightly positive correlation with SDI in 2019 (ρ = 0.18, $$p \leq 0.02$$) and the EAPC in ASDR presented no correlation with SDI in 2019 (ρ = 0.12, $$p \leq 0.12$$) (Figures 5C,D).
**Figure 5:** *The influential factors for EAPC. (A) The correlation between EAPC in ASMR and ASMR in 1990. (B) The correlation between EAPC in ASDR and ASDR in 1990. (C) The correlation between EAPC in ASMR and sociodemographic index in 2019. (D) The correlation between EAPC in ASDR and sociodemographic index in 2019. The circles represent countries and the size of circle is increased with the number of chronic kidney disease-related deaths and DALYs attributable to high sodium intake. The ρ indices and p values presented in (C,D) were derived from Pearson correlation analysis. ASMR, age-standardized mortality rate; DALYs, disability-adjusted life years; ASDR, age-standardized DALYs rate; EAPC, estimated annual percentage change.*
## Discussion
This study provides up-to-date estimates of the global spatial and temporal trends of the high sodium intake-caused CKD burden, covering 204 countries and territories worldwide from 1990 to 2019. Our results showed that in 2019, the global number of deaths and DALYs from CKD attributable to high sodium intake nearly doubled in both sexes compared to 1990. However, after age-standardization, the deaths and DALYs showed only a slight increase in males, while the corresponding ASMR was stable and the corresponding ASDR slightly decreased in females. The trends in the absolute number of CKD deaths and DALYs attributable to high sodium intake can be partly attributed to population growth and aging. Additionally, the global CKD deaths and DALYs attributable to high sodium intake were greater in males than in females aged below 90, which was contrary to the condition in groups aged above 90. Furthermore, the spatiotemporal trends of the burden of CKD attributable to high sodium intake were not homogeneous, showing a complex association with sociodemographic factors.
Sodium is an essential dietary mineral and nutrient and is significant for cell function normalization, nerve impulse transmission, acid–base balance, and plasma volume maintenance. The daily minimum sodium intake for maintaining normal physiological function is about 200–500 mg [32, 33]. However, global data suggested that the average daily sodium intake for CKD patients was above this range, and three out of four CKD patients had a sodium intake of more than 100 mmol/day [34], greater than the recommended level (87 mmol/day) by the KDIGO 2021 Clinical Practice Guideline for the Management of Blood Pressure in Chronic Kidney Disease [35]. Numerous randomized trials and observational studies have demonstrated a positive association between high dietary sodium intake and CKD development or progression (12, 19–22). The potential mechanisms of high sodium exposure leading to renal function deterioration include oxidative stress, inflammation, fibrosis, endothelial dysfunction, high salt-induced blunted renal autoregulation, and tissue remodeling (36–41). Moreover, high sodium intake is also relevant to increased blood pressure, albuminuria, obesity, insulin resistance, and the metabolic syndrome, which are critical risk factors for renal function decline [13, 41, 42].
We found males were more affected by high sodium intake-caused CKD than females worldwide, as the age-specific mortality and DALY rates of the CKD due to high sodium intake were universally higher in males than in females across all age groups. However, a higher age-standardized prevalence of CKD was observed in females [10], which was not completely explained. The global age-standardized CKD mortality rate and age-standardized death attributable to high sodium intake among males were 1.39 and 2 times greater than those among females, respectively [10, 23]. Similar to our results, previous research found that males were more exposed to high sodium intake than females, thus suffering more global high sodium intake-caused burdens [23]. This result might partly explain the sex-related differences in the high sodium intake-caused CKD burden. Another important reason was that females could more efficiently maintain Na+ homeostasis during acclimation to high sodium intake challenges, making them much less vulnerable to the adverse influence of high sodium intake exposure [43, 44]. Hormones in women were also reported to have a protective effect on CKD progression [45]. Furthermore, men were more likely to have poor lifestyles, especially smoking and alcoholism [46, 47]. Therefore, the sex differences concerning the global burden of CKD attributable to high sodium intake should be addressed by designing and adopting gender-based sodium reduction policies and programs in future studies.
In addition, we observed the increasing trend in the burden of CKD attributable to high sodium intake with age. The age-specific mortality rate and DALYs rate decreased in people aged below 59 and increased in people aged above 60 in both sexes from 1990 to 2019. The absolute number of CKD deaths and DALYs attributable to high sodium intake showed a similar pattern. These results suggested that the high sodium intake-related CKD burden was heavier in the elderly than in the young population. This phenomenon might be attributed to several factors. First, vascular compliance and renal filtration decrease with age, potentially making older people more susceptible to the adverse effects of high sodium intake. Second, many countries have implemented effective sodium-reduction measures, which may reduce the disease burden in younger individuals. A study in the United Kingdom identified that salt reduction policies have prevented or postponed 57,000 new cases and 12,000 deaths from CVD from 2003 to 2015 [48]. Another publication also suggested that reducing dietary salt by 3 g per day would reduce the annual number of new cases of coronary heart disease by 60,000 to 120,000, reduce stroke by 32,000 to 66,000, reduce myocardial infarction by 54,000 to 99,000 and reduce annual all-cause mortality by 44,000 to 92,000 [49]. However, the decreasing burden of CKD attributable to high sodium intake may not offset the effects of more and more serious population aging in many countries, leading to a net increase in the high sodium intake-induced CKD burden.
SDI is a summary measure of socioeconomic development. Regions with different SDI quintiles show discrepancies in the burden of CKD attributable to high sodium intake, reflecting socio-spatial inequalities in CKD prevention, health care, and sodium reduction measures. In the present study, the ASMR and ASDR of CKD attributable to high sodium intake were the lowest in high SDI regions between 1990 and 2019 (Supplementary Figures S3, S4). Residents of high SDI regions have more access to better education, health care, and social security, as well as more effective implementation of prevention strategies, all of which contribute to the reduction of the burden of CKD attributable to high sodium intake. It is worth noting that the ASMR and ASDR of middle SDI regions have surpassed those of low-middle SDI and low SDI regions since 1990. One possible reason was that it was difficult to conduct CKD examinations, and CKD data were underreported in these regions, where medical institutions and advanced laboratory diagnostic services were lacking. This further demonstrates that the association between the burden of CKD attributable to high sodium intake and SDI should not be assumed to be simplistic and linear. More specifically, the high-income Asia-Pacific region saw the largest decrease in the ASMR and ASDR of CKD attributable to high sodium intake from 1990 to 2019, while high-income North America has witnessed the greatest rise in ASMR and ASDR. The main reason was that high-income Asia-Pacific and high-income North America showed the most pronounced decrease and increase in high sodium intake, respectively [23]. This discrepancy partially reflected large differences in the effectiveness of global sodium reduction measures.
The CKD-related burden attributable to high sodium intake also varied substantially across countries. In 2019, the ASMR and ASDR of CKD attributable to high sodium intake varied nearly 80-fold and 58-fold across countries, respectively. Of note, although China and India have witnessed a significant decrease in ASR of CKD attributable to high sodium intake, the absolute number of CKD deaths and DALYs attributable to high sodium intake between 1990 and 2019 were the highest in these two countries. This is mainly because almost one-third of the world’s CKD patients live in the two most populous countries worldwide [10]. The heaviest absolute burden in China and India was also caused by demographic factors, such as improved life expectancy, population growth, and population aging. The higher sodium intake in China and India, especially China, than that in other countries was also a vital reason [50]. Although China and India have taken a series of actions to reduce salt intake recently, the salt intakes of Chinese and Indians were still high, with an average of 9.3 and 10.98 g per day [51, 52], approximately twice the amount recommended by the Chinese Dietary Guidelines (<6 g per day) [53] or the World Health Organization (<5 g per day) [54]. However, the good news is that the average salt intake of Chinese residents has been decreasing in recent years [55], which may lead to a net decrease in the ASMR and ASDR of CKD attributable to high sodium intake. In response to this, salt intake reduction measures are still needed, especially in countries with high sodium intake. To date, salt intake reduction measures can be broadly divided into two categories: supply reduction measures (e.g., reduction of salt content in commercialized foods) and demand reduction measures (e.g., increasing the price of salt products and raising the risk awareness of consumers for high sodium intake) [56]. Nonetheless, targeted, flexible, gender-based, and geographic salt reduction measures should be considered because of differences in diets, economic conditions, and demography. In addition, we should spare no efforts to prevent and treat CKD to reduce the burden of CKD attributable to high sodium intake.
## Limitation
This is the first and most comprehensive epidemiological study to analyze the burden of CKD attributable to high sodium intake at the global, regional, and national levels according to SDI and elucidate its trend from 1990 to 2019. The limitations of this study are as follows: Since the data for this study were collected from the GBD 2019, the limitations of the GBD 2019 methods mentioned in previous studies [1, 10, 11] cannot be avoided in the present study. First, in GBD 2019, the data on the CKD burden in many countries and regions, especially less developed countries, were lacking. Therefore, the burden of CKD attributable to high sodium intake in these countries and territories must be inferred by performing covariate-driven modeling using GBD collaborators, which might result in the uncertainty of the data. Although statistically robust methods have been applied in GBD 2019 to overcome this problem, a greater investment is needed to improve vital registration and data collection in developing countries. Moreover, this study did not report the burden of the different causes of CKD attributable to high sodium intake. Second, the severity of CKD is not considered in this analysis. Third, the misclassification of CKD death cannot be fully avoided because of conditions that coexist with CKD. Fourth, since the risk outcome was minimal for the population aged 0–24, only adults aged 25 and above were included in our study. Fifth, the dietary sodium intake was estimated based on 24-h urine collections, which might introduce inaccuracies and biases. Finally, as this study was an analysis of the available data from GBD 2019, we have no detailed data to further control bias from other important risk factors for CKD, including lifestyle, occupation, ethnicity, and air pollution.
## Conclusion
In conclusion, this study systematically evaluated the temporal and spatial changes in the burden of CKD attributable to high sodium intake from 1990 to 2019, and large regional and national variations in the burden were also observed. Although the ASMR and ASDR of CKD attributable to high sodium intake slightly increased globally, the absolute number of deaths and DALYs showed a substantial increase from 1990 to 2019 with population growth and aging. High sodium intake remains an important dietary risk factor for the global CKD burden, particularly in males, the elderly, and the population in the middle SDI regions. Flexible, integrated, gender-based, and geo-specific sodium reduction policies and programs are encouraged in the future. In addition, the prevention, assessment, and treatment strategies for CKD and risk factor management are of great significance. Our findings could provide valuable information for policymakers to develop targeted interventions, plans, and policies for future CKD prevention and sodium intake management in different regions.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
JW and WL designed the study and provided overall guidance, analyzed the data, and performed the statistical analysis. WY and LZ double-checked all the data. JW, WL, and XZ drafted the initial manuscript. All authors reviewed the drafted manuscript for critical content and approved the submitted version of the manuscript.
## Funding
This work was supported by the Hunan Province Natural Science Foundation (Grant Nos. 2021JJ40939 and 2022JJ80043), the Scientific research project of Hunan Health Commission (Grant Nos. 202102041763 and 202203014949), the Changsha Municipal Natural Science Foundation (No. kq2014267), and the Hunan Engineering Research Center of intelligent prevention and control for drug induced organ injury (No. 40).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1078371/full#supplementary-material
## References
1. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990-2019: a systematic analysis for the global burden of disease study 2019**. *Lancet* (2020.0) **396** 1204-22. DOI: 10.1016/S0140-6736(20)30925-9
2. Foreman KJ, Marquez N, Dolgert A, Fukutaki K, Fullman N, McGaughey M. **Forecasting life expectancy, years of life lost, and all-cause and cause-specific mortality for 250 causes of death: reference and alternative scenarios for 2016-40 for 195 countries and territories**. *Lancet* (2018.0) **392** 2052-90. DOI: 10.1016/S0140-6736(18)31694-5
3. Grams ME, Chow EK, Segev DL, Coresh J. **Lifetime incidence of CKD stages 3-5 in the United States**. *Am J Kidney Dis* (2013.0) **62** 245-52. DOI: 10.1053/j.ajkd.2013.03.009
4. Inker LA, Tighiouart H, Aspelund T, Gudnason V, Harris T, Indridason OS. **Lifetime risk of stage 3-5 CKD in a community-based sample in Iceland**. *Clin J Am Soc Nephrol* (2015.0) **10** 1575-84. DOI: 10.2215/CJN.00180115
5. Brown EA, Zhao J, McCullough K, Fuller DS, Figueiredo AE, Bieber B. **Burden of kidney disease, health-related quality of life, and employment among patients receiving peritoneal dialysis and in-center hemodialysis: findings from the DOPPS program**. *Am J Kidney Dis* (2021.0) **78** 489-500.e1. DOI: 10.1053/j.ajkd.2021.02.327
6. Saran R, Pearson A, Tilea A, Shahinian V, Bragg-Gresham J, Heung M. **Burden and cost of caring for US veterans with CKD: initial findings from the VA renal information system (VA-REINS)**. *Am J Kidney Dis* (2021.0) **77** 397-405. DOI: 10.1053/j.ajkd.2020.07.013
7. Legrand K, Speyer E, Stengel B, Frimat L, Ngueyon SW, Massy ZA. **Perceived health and quality of life in patients with CKD, including those with kidney failure: findings from national surveys in France**. *Am J Kidney Dis* (2020.0) **75** 868-78. DOI: 10.1053/j.ajkd.2019.08.026
8. Honeycutt AA, Segel JE, Zhuo X, Hoerger TJ, Imai K, Williams D. **Medical costs of CKD in the medicare population**. *J Am Soc Nephrol* (2013.0) **24** 1478-83. DOI: 10.1681/ASN.2012040392
9. Wang F, Yang C, Long J, Zhao X, Tang W, Zhang D. **Executive summary for the 2015 annual data report of the China kidney disease network (CK-NET)**. *Kidney Int* (2019.0) **95** 501-5. DOI: 10.1016/j.kint.2018.11.011
10. **Global, regional, and national burden of chronic kidney disease, 1990-2017: a systematic analysis for the global burden of disease study 2017**. *Lancet* (2020.0) **395** 709-33. DOI: 10.1016/S0140-6736(20)30045-3
11. **Global burden of 87 risk factors in 204 countries and territories, 1990-2019: a systematic analysis for the global burden of disease study 2019**. *Lancet* (2020.0) **396** 1223-49. DOI: 10.1016/S0140-6736(20)30752-2
12. He J, Mills KT, Appel LJ, Yang W, Chen J, Lee BT. **Urinary sodium and potassium excretion and CKD progression**. *J Am Soc Nephrol* (2016.0) **27** 1202-12. DOI: 10.1681/ASN.2015010022
13. McMahon EJ, Campbell KL, Bauer JD, Mudge DW, Kelly JT. **Altered dietary salt intake for people with chronic kidney disease**. *Cochrane Database Syst Rev* (2021.0) **2021** CD010070. DOI: 10.1002/14651858.CD010070.pub3
14. Malta D, Petersen KS, Johnson C, Trieu K, Rae S, Jefferson K. **High sodium intake increases blood pressure and risk of kidney disease. From the science of salt: a regularly updated systematic review of salt and health outcomes (August 2016 to March 2017)**. *J Clin Hypertens* (2018.0) **20** 1654-65. DOI: 10.1111/jch.13408
15. Shin J, Lee CH. **The roles of sodium and volume overload on hypertension in chronic kidney disease**. *Kidney Res Clin Pract* (2021.0) **40** 542-54. DOI: 10.23876/j.krcp.21.800
16. Martin K, Tan SJ, Toussaint ND. **Total body sodium balance in chronic kidney disease**. *Int J Nephrol* (2021.0) **2021** 7562357. DOI: 10.1155/2021/7562357
17. **KDIGO 2021 clinical practice guideline for the management of glomerular diseases**. *Kidney Int* (2021.0) **100** S1-S276. DOI: 10.1016/j.kint.2021.05.021
18. de Borst MH, Navis G. **Sodium intake, RAAS-blockade and progressive renal disease**. *Pharmacol Res* (2016.0) **107** 344-51. DOI: 10.1016/j.phrs.2016.03.037
19. Burnier M. **Sodium intake and progression of chronic kidney disease-has the time finally come to do the impossible: a prospective randomized controlled trial?**. *Nephrol Dial Transplant* (2021.0) **36** 381-4. DOI: 10.1093/ndt/gfaa120
20. Sugiura T, Takase H, Ohte N, Dohi Y. **Dietary salt intake is a significant determinant of impaired kidney function in the general population**. *Kidney Blood Press Res* (2018.0) **43** 1245-54. DOI: 10.1159/000492406
21. Vegter S, Perna A, Postma MJ, Navis G, Remuzzi G, Ruggenenti P. **Sodium intake, ACE inhibition, and progression to ESRD**. *J Am Soc Nephrol* (2012.0) **23** 165-73. DOI: 10.1681/ASN.2011040430
22. Kang M, Kang E, Ryu H, Hong Y, Han SS, Park SK. **Measured sodium excretion is associated with CKD progression: results from the KNOW-CKD study**. *Nephrol Dial Transplant* (2021.0) **36** 512-9. DOI: 10.1093/ndt/gfaa107
23. Chen X, Du J, Wu X, Cao W, Sun S. **Global burden attributable to high sodium intake from 1990 to 2019**. *Nutr Metab Cardiovas* (2021.0) **31** 3314-21. DOI: 10.1016/j.numecd.2021.08.033
24. 24.
http://ghdx.healthdata.org/gbd-results-tool
25. **Global, regional, and national age-sex-specific mortality for 282 causes of death in 195 countries and territories, 1980-2017: a systematic analysis for the global burden of disease study 2017**. *Lancet* (2018.0) **392** 1736-88. DOI: 10.1016/S0140-6736(18)32203-7
26. **KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease**. *Kid Int Suppl* **2013** 1-150
27. Afshin A, Sur PJ, Fay KA, Cornaby L, Ferrara G, Salama JS. **Health effects of dietary risks in 195 countries, 1990-2017: a systematic analysis for the global burden of disease study 2017**. *Lancet* (2019.0) **393** 1958-72. DOI: 10.1016/S0140-6736(19)30041-8
28. Wang L, Du J, Cao W, Sun S. **Trends of stroke attributable to high sodium intake at the global, regional, and national levels from 1990 to 2019: a population-based study**. *Neurol Res* (2021.0) **43** 474-81. DOI: 10.1080/01616412.2020.1867950
29. Hankey BF, Ries LA, Kosary CL, Feuer EJ, Merrill RM, Clegg LX. **Partitioning linear trends in age-adjusted rates**. *Cancer Causes Control* (2000.0) **11** 31-5. DOI: 10.1023/A:1008953201688
30. Zhang T, Yin X, Chen H, Li Y, Chen J, Yang X. **Global magnitude and temporal trends of idiopathic developmental intellectual disability attributable to lead exposure from 1990 to 2019: results from global burden of disease study**. *Sci Total Environ* (2022.0) **834** 155366. DOI: 10.1016/j.scitotenv.2022.155366
31. Man J, Zhang T, Yin X, Chen H, Zhang Y, Zhang X. **Spatiotemporal trends of colorectal cancer mortality due to low physical activity and high body mass index from 1990 to 2019: a global, regional and national analysis**. *Front Med* (2021.0) **8** 800426. DOI: 10.3389/fmed.2021.800426
32. Holbrook JT, Patterson KY, Bodner JE, Douglas LW, Veillon C, Kelsay JL. **Sodium and potassium intake and balance in adults consuming self-selected diets**. *Am J Clin Nutr* (1984.0) **40** 786-93. DOI: 10.1093/ajcn/40.4.786
33. He FJ, MacGregor GA. **A comprehensive review on salt and health and current experience of worldwide salt reduction programmes**. *J Hum Hypertens* (2009.0) **23** 363-84. DOI: 10.1038/jhh.2008.144
34. Borrelli S, Provenzano M, Gagliardi I, Michael A, Liberti ME, De Nicola L. **Sodium intake and chronic kidney disease**. *Int J Mol Sci* (2020.0) **21** 4744. DOI: 10.3390/ijms21134744
35. **KDIGO 2021 clinical practice guideline for the management of blood pressure in chronic kidney disease**. *Kidney Int* (2021.0) **99** S1-S87. DOI: 10.1016/j.kint.2020.11.003
36. Zhang WC, Zheng XJ, Du LJ, Sun JY, Shen ZX, Shi C. **High salt primes a specific activation state of macrophages, M(Na)**. *Cell Res* (2015.0) **25** 893-910. DOI: 10.1038/cr.2015.87
37. Hijmans RS, van Londen M, Sarpong KA, Bakker S, Navis GJ, Storteboom T. **Dermal tissue remodeling and non-osmotic sodium storage in kidney patients**. *J Transl Med* (2019.0) **17** 88. DOI: 10.1186/s12967-019-1815-5
38. Yu HC, Burrell LM, Black MJ, Wu LL, Dilley RJ, Cooper ME. **Salt induces myocardial and renal fibrosis in normotensive and hypertensive rats**. *Circulation* (1998.0) **98** 2621-8. DOI: 10.1161/01.CIR.98.23.2621
39. Mihai S, Codrici E, Popescu ID, Enciu AM, Albulescu L, Necula LG. **Inflammation-related mechanisms in chronic kidney disease prediction, progression, and outcome**. *J Immunol Res* (2018.0) **2018** 2180373. DOI: 10.1155/2018/2180373
40. Fellner RC, Cook AK, O'Connor PM, Zhang S, Pollock DM, Inscho EW. **High-salt diet blunts renal autoregulation by a reactive oxygen species-dependent mechanism**. *Am J Physiol Renal Physiol* (2014.0) **307** F33-40. DOI: 10.1152/ajprenal.00040.2014
41. Oh SW, Han KH, Han SY, Koo HS, Kim S, Chin HJ. **Association of sodium excretion with metabolic syndrome, insulin resistance, and body fat**. *Medicine* (2015.0) **94** e1650. DOI: 10.1097/MD.0000000000001650
42. Lanaspa MA, Kuwabara M, Andres-Hernando A, Li N, Cicerchi C, Jensen T. **High salt intake causes leptin resistance and obesity in mice by stimulating endogenous fructose production and metabolism**. *Proc Natl Acad Sci U S A* (2018.0) **115** 3138-43. DOI: 10.1073/pnas.1713837115
43. Gohar EY, De Miguel C, Obi IE, Daugherty EM, Hyndman KA, Becker BK. **Acclimation to a high-salt diet is sex dependent**. *J Am Heart Assoc* (2022.0) **11** e20450. DOI: 10.1161/JAHA.120.020450
44. Veiras LC, Girardi A, Curry J, Pei L, Ralph DL, Tran A. **Sexual dimorphic pattern of renal transporters and electrolyte homeostasis**. *J Am Soc Nephrol* (2017.0) **28** 3504-17. DOI: 10.1681/ASN.2017030295
45. Valdivielso JM, Jacobs-Cacha C, Soler MJ. **Sex hormones and their influence on chronic kidney disease**. *Curr Opin Nephrol Hypertens* (2019.0) **28** 1-9. DOI: 10.1097/MNH.0000000000000463
46. Safiri S, Nejadghaderi SA, Abdollahi M, Carson-Chahhoud K, Kaufman JS, Bragazzi NL. **Global, regional, and national burden of cancers attributable to tobacco smoking in 204 countries and territories, 1990-2019**. *Cancer Med* (2022.0) **11** 2662-78. DOI: 10.1002/cam4.4647
47. Hernandez-Vasquez A, Chacon-Torrico H, Vargas-Fernandez R, Grendas LN, Bendezu-Quispe G. **Gender differences in the factors associated with alcohol binge drinking: a population-based analysis in a Latin American country**. *Int J Environ Res Public Health* (2022.0) **19** 4931. DOI: 10.3390/ijerph19094931
48. Kypridemos C, Guzman-Castillo M, Hyseni L, Hickey GL, Bandosz P, Buchan I. **Estimated reductions in cardiovascular and gastric cancer disease burden through salt policies in England: an IMPACTNCD microsimulation study**. *BMJ Open* (2017.0) **7** e13791. DOI: 10.1136/bmjopen-2016-013791
49. Bibbins-Domingo K, Chertow GM, Coxson PG, Moran A, Lightwood JM, Pletcher MJ. **Projected effect of dietary salt reductions on future cardiovascular disease**. *N Engl J Med* (2010.0) **362** 590-9. DOI: 10.1056/NEJMoa0907355
50. Powles J, Fahimi S, Micha R, Khatibzadeh S, Shi P, Ezzati M. **Global, regional and national sodium intakes in 1990 and 2010: a systematic analysis of 24 h urinary sodium excretion and dietary surveys worldwide**. *BMJ Open* (2013.0) **3** e3733. DOI: 10.1136/bmjopen-2013-003733
51. Johnson C, Praveen D, Pope A, Raj TS, Pillai RN, Land MA. **Mean population salt consumption in India: a systematic review**. *J Hypertens* (2017.0) **35** 3-9. DOI: 10.1097/HJH.0000000000001141
52. 52.The State Council Information Office. The state council information office (SCIO) holds a press briefing on the “report on the nutrition and chronic disease status of Chinese residents 2020”. Available at: http://www.gov.cn/xinwen/2020-12/24/content_5572983.htm
53. 53.Chinese Nutrition Society. Dietary guidelines for Chinese residents 2016. Beijing, China: People’s Medical Publishing House (2016).. *Dietary guidelines for Chinese residents 2016* (2016.0)
54. 54.World Health Organization. Guideline: sodium intake for adults and children. Geneva, Switzerland: WHO (2012).. *Guideline: sodium intake for adults and children* (2012.0)
55. Xu A, Ma J, Guo X, Wang L, Wu J, Zhang J. **Association of a province-wide intervention with salt intake and hypertension in Shandong Province, China, 2011-2016**. *JAMA Intern Med* (2020.0) **180** 877-86. DOI: 10.1001/jamainternmed.2020.0904
56. 56.World Health Organization. Reducing salt intake in populations: report of a WHO forum and technical meeting, 5–7 October 2006, Paris, France. 924159537X (2007).. (2007.0)
|
---
title: Meteorin-like levels are associated with active brown adipose tissue in early
infancy
authors:
- Cristina Garcia-Beltran
- Artur Navarro-Gascon
- Abel López-Bermejo
- Tania Quesada-López
- Francis de Zegher
- Lourdes Ibáñez
- Francesc Villarroya
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC10018039
doi: 10.3389/fendo.2023.1136245
license: CC BY 4.0
---
# Meteorin-like levels are associated with active brown adipose tissue in early infancy
## Abstract
### Introduction
Meteorin-like (METRNL) is a hormonal factor released by several tissues, including thermogenically active brown and beige adipose tissues. It exerts multiple beneficial effects on metabolic and cardiovascular systems in experimental models. However, the potential role of METRNL as brown adipokine in humans has not been investigated previously, particularly in relation to the metabolic adaptations taking place in early life, when brown adipose tissue (BAT) is particularly abundant.
### Methods and materials
METRNL levels, as well as body composition (DXA) and circulating endocrine-metabolic variables, were assessed longitudinally in a cohort of infants at birth, and at ages 4 and 12 months. BAT activity was measured by infrared thermography at age 12 months. METRNL levels were also determined cross-sectionally in adults; METRNL gene expression (qRT-PCR) was assessed in BAT and liver samples from neonates, and in adipose tissue and liver samples form adults. Simpson-Golabi-Behmel Syndrome (SGBS) adipose cells were thermogenically activated using cAMP, and METRNL gene expression and METRNL protein released were analysed.
### Results
Serum METRNL levels were high at birth and declined across the first year of life albeit remaining higher than in adulthood. At age 4 and 12 months, METRNL levels correlated positively with circulating C-X-C motif chemokine ligand 14 (CXCL14), a chemokine released by thermogenically active BAT, but not with parameters of adiposity or metabolic status. METRNL levels also correlated positively with infrared thermography-estimated posterior-cervical BAT activity in girls aged 12 months. Gene expression analysis indicated high levels of METRNL mRNA in neonatal BAT. Thermogenic stimulus of brown/beige adipocytes led to a significant increase of METRNL gene expression and METRN protein release to the cell culture medium.
### Conclusion
Circulating METRNL levels are high in the first year of life and correlate with indices of BAT activity and with levels of an established brown adipokine such as CXCL14. These data, in addition with the high expression of METRNL in neonatal BAT and in thermogenically-stimulated brown/beige adipocytes, suggest that METRNL is actively secreted by BAT and may be a circulating biomarker of BAT activity in early life.
## Introduction
Meteorin-like protein (METRNL), also known as Meteorin-β, interleukin-41 and subfatin, is a recently identified hormone involved in metabolic regulation and considered a candidate biomarker of metabolic syndrome [1]. In rodent models, METRNL is highly expressed in brown adipose tissue (BAT) upon thermogenic activation and also in skeletal muscle after exercise [2]. METRNL was found to promote energy expenditure and glucose tolerance through the induction of alternatively activated macrophages at adipose depots and by promoting the browning of adipose tissue. Further research showed that peroxisome proliferator–activated receptor-γ (PPARγ) enhances the capacity of METRNL to antagonize insulin resistance in adipose tissue [3]. METRNL also attenuates inflammation and insulin resistance in skeletal muscle via AMP-activated protein kinase and PPARδ-dependent pathways [4, 5]. The beneficial effects of METRNL have been associated with innate immunity [6, 7], and protection against cardiac dysfunction [8, 9]. In adult humans, METRNL levels are low in patients with obesity and diabetes and correlate negatively with glucose levels and markers of insulin resistance (10–13).
Metabolic and nutritional alterations in the early postnatal life are not only relevant for health during infancy but may also contribute to the development of metabolic syndrome in later life. In recent years, the activity of thermogenic (brown/beige) adipose tissues in adult humans has gained attention as a protective factor against obesity, type 2 diabetes and cardiovascular disease [14]. This is attributed to the capacity of BAT to both drain glucose and lipids for adaptive thermogenesis and to secrete adipokines with healthy effects on metabolism [15]. However, despite the existing awareness that BAT size and activity are particularly relevant in infants [16], the pathophysiological consequences of distinct BAT activities early after birth have not been studied. The identification of BAT-derived adipokines in infants and their capacity to be used as biomarkers of metabolic health has also been scarcely undertaken, and only a few circulating molecules, such as bone morphogenetic protein-8B (BMP8B) and C-X-C motif chemokine ligand 14 (CXCL14), have respectively been associated with BAT activity in newborns and in one-year-old infants (17–19).
Here we determined for the first time the circulating levels of METRNL across the first year of life and disclosed a significant association between this variable and the extent of BAT activity.
## Study population and ethics
The primary study cohort consisted of 50 infants (27 girls and 23 boys) who were enrolled prenatally during the customary third trimester visit among Caucasian pregnant mothers consecutively seen in the outpatient clinics of Hospital Sant Joan de Déu and Hospital de Sant Boi – Parc Sanitari Sant Joan de Déu (Barcelona, Spain) (Supplementary Figure 1). These infants had previously participated in a longitudinal study assessing BAT activity and circulating levels of CXCL14 and BMP8B in the first year of life [18, 19].
Inclusion criteria were: maternally uncomplicated, singleton pregnancy with delivery at term (37-42 weeks), exclusive breastfeeding or formula-feeding in the first 4 months, postnatal follow-up completed (at 15 days, 4 and 12 months) and written informed consent. Exclusion criteria were maternal disease, alcohol or drug abuse, congenital malformations and complications at birth. Birth weight was not considered as inclusion or exclusion criterium; accordingly, the study population included infants with a wide range of birth weight Z-scores (between −2.9 and +1.0).
Circulating METRNL was exclusively measured in a subset of infants who had spare serum sample available at birth (20 girls and 18 boys), and at age 4 and 12 months (26 girls and 16 boys). Serum METRNL was also measured in 30 mothers of those infants (age, 33.6 ± 0.9 years) during the third trimester of pregnancy (Supplementary Figure 1). In addition, serum METRNL concentrations were analyzed cross-sectionally in healthy adult women ($$n = 10$$; age, 38.7 ± 1.9 years) METRNL mRNA gene expression was assessed in dorso-interscapular BAT ($$n = 5$$) and liver ($$n = 6$$) post-mortem samples obtained on occasion of autopsies (2–3 h after the death) of Caucasian newborns with a gestational age of 28–36 weeks who survived, at most, 3 days post-partum, supplied by the Academy of Sciences of the Czech Republic as previously described [20] (Supplementary Table 1). For comparison, METRNL mRNA gene expression was also determined in adult liver samples (obtained from hepatic biopsies performed when a hepatic tumor was suspected, with a negative ultimate result), deltoid muscle samples (from adult individuals who underwent skeletal muscle biopsy because of muscle complaint in whom skeletal muscle histology was thereafter normal) and subcutaneous adipose tissue samples (obtained from volunteers), as described [8, 21].
The study was approved by the Institutional Review Board of the University of Barcelona, Sant Joan de Déu University Hospital; all participating mothers signed the informed consent at recruitment.
## Clinical and endocrine-metabolic assessments
Maternal data were retrieved from hospital clinical records. Gestational age was calculated according to the last menses and validated by first-trimester ultrasound. Weight and length of the newborns were measured immediately after delivery, and again at age 4 months and 12 months.
Maternal venous samples were obtained during the third trimester of gestation, between week 28 and delivery. Neonatal blood samples were obtained at birth from the umbilical cord before placenta separation [22]. At age 4 and 12 months, venous samples were obtained during the morning in the fasting state. Adult venous samples were also obtained after overnight fasting. The serum fraction of the samples was separated by centrifugation and stored at -80°C until analysis.
Serum glucose, insulin, insulin-like growth factor (IGF)-I, high-molecular-weight (HMW) adiponectin, CXCL14 and BMP8B were assessed as previously reported [18, 19]. Serum METRNL levels were determined in serum and cell culture media with a specific human enzyme-linked immunosorbent assay kit [R&D Systems, Minneapolis, MN, USA; sensitivity: 0.64 ng/mL; intra-assay coefficient of variation (CV) <$10\%$; inter-assay CV < $12\%$].
## Body composition and BAT activity assessment
Body composition was assessed at age 15 days, 4 months and 12 months by dual-energy X-ray absorptiometry (DXA) with a Lunar Prodigy and Lunar software (version $\frac{3.4}{3.5}$; Lunar Corp., Madison, WI, USA) adapted for infants [22].
As previously described [18], BAT activity at age 12 months was estimated through the infrared thermography-based measurement of the skin temperature overlying BAT depots. The parameters assessed included the maximal temperature at the posterior cervical (TPCR) and supraclavicular (TSCR) regions, and the extent of active BAT in these regions (AreaPCR and AreaSCR).
## Cell cultures of neonatal beige adipocytes
Pre-adipocyte cells obtained post-mortem from a 3-month-old infant with Simpson Golabi Behmel Syndrome (SGBS cells) [23], capable to differentiate into adipocytes bearing a beige phenotype [24, 25] were used. SGBS pre-adipocytes were maintained in Dulbecco’s modified Eagle’s (DMEM)/F12 medium, $10\%$ fetal bovine serum (FBS). Beige adipogenic differentiation was initiated by incubating confluent cell cultures for 4 days in serum-free medium plus 20 nM insulin, 0.2 nM triiodothyronine, 100 nM cortisol, 25 nM dexamethasone, 500 µM 3-isobutyl-1-methyl-xanthine, and 2 µM rosiglitazone. Subsequently, cells were switched to DMEM/F12, 20 nM insulin, 0.2 nM triiodothyronine, and 100 nM cortisol and maintained for up to 10 days, when more than $90\%$ cells have acquired differentiated adipocyte morphology. To induce thermogenic activation of adipocytes, differentiated cells were treated with 1mM dibutyril-cAMP for 24 hours. All cell culture reagents and drugs were from Sigma-Aldrich (St Louis, Missouri, USA). Cells were collected for RNA isolation and the cell culture medium, corresponding to 24 h before harvest, was also collected for measurement of METRNL levels.
## RNA isolation and qRT-PCR analyses
RNA was extracted from tissues and cells using an affinity-based method (NucleoSpin, Macherey-Nagel, Germany). METRNL, UCP1, PPARGC1A, DIO2 and BMP8B transcript levels were determined by qRT-PCR using TaqMan technology (Thermo Fisher Scientific, Waltham, MA, USA). 0.5 μg RNA were retrotranscribed using random hexamer primers (Thermo Fisher Scientific, Waltham, MA, USA). For qRT-PCR, the METRNL (Hs00417150), UCP1 (Hs00222453), PPARGC1A (Hs00173304), DIO2 (Hs00255341), BMP8B (Hs01629120) TaqMan Gene Expression assay probes were used, with reaction mixtures containing 1 μL cDNA, 10 μL TaqMan Universal PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, USA), 250 nM probes and 900 nM of primers from the Assays-on-Demand Gene Expression Assay Mix (Thermo Fisher Scientific, Waltham, MA, USA). The 18S rRNA transcript (Hs99999901) was measured as housekeeping reference gene. The mRNA level of METRNL and UCP11: PGC1a, Dio2, Bmp8b in tissues and cells sample was normalized to that of the reference control using the comparative (2–ΔCT) method.
## Statistics
Statistical analyses were implemented in SPSS version 27.0 (SPSS software, IBM, Armonk, NY, USA), GraphPad Prism 5 (GraphPad Software, CA, USA) and R Project version 4.2.2 (RStudio, MA, USA). Results are shown as mean ± standard error of the mean (SEM). Variables with normal distribution were compared with two-tailed Student’s t-test. Chi-square test was used to compare qualitative variables. Correlation and stepwise multi-regression analysis were used to study associations between circulating METRNL levels and the assessed variables; outliers were detected using a studentized residual outlier test and excluded from further analyses; this approach did not modify the statistical significance of any analysis. Covariance analysis was used to adjust for ponderal index and breastfeeding. A P-value < 0.05 was considered statistically significant.
## METRNL levels in the first year of life
Supplementary Table 2 shows the longitudinal data from infants over the first year of life and from their mothers in late pregnancy in the cohort in which serum METRNL assessment was performed. As previously reported [18, 19], girls had less lean mass, higher levels of circulating CXCL14 and higher posterior BAT activity.
When splitting METRNL levels by sex, or by type of early feeding, no differences were found at any study time; accordingly, the results were pooled. Circulating METRNL concentrations in infants at birth were higher than at the postnatal ages of 4 and 12 months, and higher than in non-pregnant women (Figure 1).
**Figure 1:** *Serum Meteorin-like (METRNL) concentrations in human infants at birth and at age 4 and 12 months, in the mothers of those infants during the third trimester of pregnancy, and in healthy adult women. *P<0.05, **P<0.01 vs at birth. P values are adjusted for ponderal index and breastfeeding.*
## Correlations between METRNL levels and clinical, endocrine-metabolic and body composition variables
The associations between circulating levels of METRNL and anthropometric, adiposity-related, and endocrine-metabolic parameters, including some putative brown adipokines (CXCL14, BMP8B), throughout follow-up are shown in Supplementary Table 3. At birth, circulating METRNL levels were negatively related to abdominal fat only in girls (R= -0.678; $$P \leq 0.013$$, Supplementary Table 3). At age 4 months, circulating METRNL showed a strong positive correlation with circulating CXCL14 concentrations in the entire population ($R = 0.648$; $$P \leq 0.002$$) (Figure 2A). At age 12 months, METRNL concentrations were also positively correlated with CXCL14 levels ($R = 0.693$, $$P \leq 0.001$$) (Figure 2B). This correlation was maintained when girls were analyzed separately ($R = 0.698$; $$P \leq 0.012$$).
**Figure 2:** *Correlation between circulating Meteorin-like (METRNL) and C-X-C motif chemokine ligand 14 (CXCL14) at age 4 months (A) and 12 months (B). Grey dots correspond to girls and black dots represent boys. P values are adjusted for ponderal index and breastfeeding.*
## Correlations between METRNL concentrations and parameters of BAT activity
BAT activity at the posterior cervical and supraclavicular regions at age 12 months was analyzed in a subset of infants of the study cohort by using infrared thermography-based procedures, as previously described [18]. Correlations between circulating METRNL levels at all time points and indicators of BAT activity at age 12 months are summarized in Supplementary Table 4. Significant positive correlations between posterior-cervical BAT activity and METRNL levels at 4 months ($R = 0.400$; $$P \leq 0.047$$; Figure 3A) and at 12 months ($R = 0.432$; $$P \leq 0.006$$; Figure 3B) were disclosed. Separate analyses by sex revealed a significant correlation between BAT activity and METRNL levels at 12 months only in girls ($R = 0.426$; $$P \leq 0.004$$).
**Figure 3:** *Correlation between the area of active brown adipose tissue at the posterior cervical region, as determined by infrared thermography (AreaPCR), at age 12 months, and circulating Meteorin-like (METRNL) concentrations at age 4 months (A) and at age 12 months (B). Grey dots correspond to girls and black dots represent boys. P values are adjusted for ponderal index and breastfeeding.*
## METRNL expression in neonatal and adult tissues
METRNL expression levels in human neonatal post-mortem samples were significantly higher in dorso-interscapular BAT than in the liver. In addition, METRNL expression in neonatal BAT was much higher than in adult adipose tissue, skeletal muscle, and liver (Figure 4).
**Figure 4:** *Meteorin-like (METRNL) gene expression levels in dorso-interscapular brown adipose tissue (BAT, N= 5) and liver (N= 6) obtained from post-mortem autopsies of newborns with a gestational age of 28–36 weeks who survived, at most, 3 days post-partum, and samples from subcutaneous adipose tissue (AT, N= 5), skeletal muscle (N = 5) and liver (N= 5) from healthy adults. Data are mean ± SEM of relative levels of METRNL mRNA (METRNL mRNA/18S rRNA). *P<0.05; **P<0.01 vs neonatal BAT.*
## Activation of neonatal brown/beige adipocytes leads to increased METRNL expression and secretion of METRNL
Neonatal adipocytes differentiated into beige phenotype (SGBS cells) were thermogenically activated using cAMP [26]. METRNL gene expression was dramatically induced, similarly to the thermogenic biomarker UCP1 (Figure 5A) and other marker genes of brown/beige phenotype such as peroxisome proliferator-activated receptor-γ coactivator-1α (PPARGC1A), iodothyronine 5’-deiodinase (DIO2) and BMP8B (Supplementary Figure 2). Moreover, thermogenic activation of the cells also induced a significant increase in the release of METRNL protein to the culture medium (Figure 5B).
**Figure 5:** *Meteorin-like (METRNL) gene expression (A) and METRNL protein secretion (B) by neonatal SGBS beige adipocytes. METRNL and uncoupling protein-1 (UCP1) transcript level (A) and METRNL protein levels (B) accumulating in cell culture medium after 24 h treatment of cell with 1 mM dibutyril-cAMP and untreated controls. Data mean ± SEM, N= 3 each condition. *P<0.05; **P<0.01; ***P<0.001 vs control.*
## Discussion
The present study is, to our knowledge, the first to have assessed the circulating concentrations of METRNL in human infants, and to have related METRNL levels to BAT activity.
Our study demonstrates that circulating levels of METRNL – a novel adipokine that promotes browning of white adipocytes upon thermogenic stimulus [2] – is high at birth as compared to adult values and decreases over the first year of human life, although remaining higher than in adults. These findings are in line with those recently reported on circulating BMP8B – another brown adipokine involved in thermoregulation and metabolic homeostasis – that shows a similar decreasing trend from birth to age 12 months [19]. Moreover, our data also disclosed a high expression of METRNL in neonatal human BAT and in thermogenic activated neonatal brown/beige adipocytes, as well as an increased secretion of METRNL in thermogenically activated cells, which confirms the capacity of human neonatal brown adipocytes to secrete this adipokine. Altogether, these data highlight the elevated activity of BAT after birth, when the demands for thermogenesis and risks for hypothermia can be especially high [16], and the concomitant production of BAT-secreted adipokines such as METRNL.
Circulating METRNL concentrations displayed a positive association with posterior-cervical BAT activity, as well as with circulating levels of CXCL14 – a chemokine secreted by active brown/beige adipose tissue [27]. These data fit well with previous studies reporting a positive correlation between circulating CXCL14 levels and BAT activity in early life [18], and also between CXCL14 and METRNL expression in adult adipose tissue [28]. Interestingly, CXCL14 and METRNL have emerged as circulating factors that modulate M2 macrophage activation playing a role in brown/beige thermogenic regulation [2, 27].
There is evidence of an interplay between BAT and skeletal muscle development in large mammalian species, which is characterized by a progressive decline in BAT after birth concomitant with skeletal muscle maturation, and this may affect BAT and muscle secretome [29]. In rodents, skeletal muscle is a relevant site of METRNL gene expression [2] whereas in adult humans METRNL expression is low [8] but induced after exercise [2, 30]. Lack of availability of muscle samples -or tissues other than BAT and liver- from neonates and young infants is a limitation of our study on METRL expression, and thus we cannot exclude a role of muscle or other tissues in influencing systemic METRNL levels in early development. Given the developmental overlap between BAT and muscle [29], it can’t be excluded that the correlation between METRNL levels and BAT activity in early infancy is indeed an indirect reflection of METRNL release by muscle. In any case, data retrieved from the transcriptomics database in muscle from infants in the first year of life and elderly adults does not indicate relevant differences in METRNL gene expression [31]. Further studies would be required to establish the relative contribution of BAT and muscle to METRNL level changes in the first year of life.
Although METRNL levels were not different between girls and boys, the above-mentioned associations of METRNL levels in relation to BAT activity and CXCL14 levels at age 12 months were only maintained in girls. This finding may fit with the previously reported observation that BAT activity at that age is higher in girls than in boys [18] and with prior data reporting sex-based differences in the levels of other putative batokines such as CXCL14 and BMP8B [18, 19]. There is extensive evidence of sex-based differences in BAT thermogenic activity due to direct and indirect hormonal mechanisms [32] and it is likely that sex-based differences occur also for the BAT secretome.
The distinct prevalence of “classic brown” versus “beige” adipocytes at specific anatomical BAT depots in humans [33, 34] may explain why circulating METRNL levels correlate with measures of posterior-cervical – but not supraclavicular – BAT activity. Differential secretory properties of brown-versus-beige cells have not yet been reported (even in experimental models) but a distinct capacity for METRNL secretion by different types of thermogenic adipocytes could account for the preferential association between METRNL levels and posterior-cervical BAT. On the other hand, high METRNL levels in early life, released by BAT and perhaps also by other tissues may promote a “browned” phenotype in white adipose depots in infants, given the known effects of METRNL in inducing the browning of adipose tissue [8], which would be especially adaptive to the thermally challenging conditions occurring in early infancy.
Circulating METRNL levels did not show significant correlations with systemic parameters of endocrine-metabolic status or adiposity in our cohort. This indicates that, although METRNL concentrations appear to be a potential indicator of the extent of BAT activity in one-year-old infants, they are poorly informative about their general endocrine-metabolic status. Possibly, the fact that our cohort involved apparently healthy children exhibiting no major differences in metabolic or adiposity parameters among individuals, precluded the identification of meaningful associations. Along these lines, the only significant correlation was the negative association between METRNL levels and abdominal fat in girls, present only at birth. Although this finding is reminiscent of the negative correlations between METRNL levels and visceral adiposity found in adults with obesity and/or type 2 diabetes [12, 13], the fact that it occurs only at birth indicates the need for future studies to explore a possible involvement of METRNL in the fat accretion occurring during late fetal development, something totally unknown to date.
Our study has several limitations, among them the relatively low number of serum samples available for METRNL assessment, the lack of tissue samples for METRNL mRNA gene expression from infants of the studied cohort due to obvious ethical reasons, and the lack of follow-up beyond age 1 year. Moreover, the lack of availability of neonatal samples from additional tissues (e.g. muscle) for gene expression analysis limited our capacity to infer whether, in addition to BAT, other tissue sources may be relevant contributors to systemic METRNL levels in infants, as in rodent models. Moreover, high levels of METRNL in blood from pregnant mothers, which may be caused by the high METRNL gene expression in placenta [data accessible at GEO profile database; GEO accession GDS3113, symbol 197624 [35]], may influence the high levels of METRNL in neonates at birth. On the other hand, further studies would be particularly interesting to assess in early infancy the potential relationship of METRNL levels with those of other secreted factors for which there is experimental evidence of involvement in BAT development, such as fibroblast growth factor (FGF)-9 or FGF21 [36]. It should be also mentioned that our data on BAT activity were obtained by infrared methodology, which is minimally invasive but does not allow to assess the actual BAT mass which would require water-fat magnetic resonance imaging or quantification of the proton density fat fraction using magnetic resonance imaging (MRI) [37, 38]. The strengths of our study include being the first assessment of METRNL levels in humans in early life and the co-availability of a large set of endocrine-metabolic, body composition and BAT activity data.
In summary, early life is associated with higher levels of circulating METRNL. The progressive reduction of METRNL concentrations in the first year of life -albeit maintained above those in adults- might reflect overall changes in BAT activity during early development. In addition, circulating METRNL concentrations associate with BAT activity and with CXCL14 levels, particularly in girls, supporting a role for METRNL as a brown adipokine and novel biomarker for BAT activity in early life.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by Institutional Review Board of the University of Barcelona, Sant Joan de Déu University Hospital, Spain. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.
## Author contributions
CG-B contributed to literature research, design of figures and tables, data collection, data analysis and interpretation. AN-G contributed to the analysis of circulating parameters and interpretation of data. TQ-L performed gene expression and cell culture-based studies. AL-B and FZ contributed to data interpretation, and reviewed/edited the manuscript. LI and FV contributed to study design, data interpretation, reviewed/edited the manuscript and wrote the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1136245/full#supplementary-material
## References
1. Alizadeh H. **Meteorin-like protein (Metrnl): A metabolic syndrome biomarker and an exercise mediator**. *Cytokine* (2022) **157**. DOI: 10.1016/j.cyto.2022.155952
2. Rao RR, Long JZ, White JP, Svensson KJ, Lou J, Lokurkar I. **Meteorin-like is a hormone that regulates immune-adipose interactions to increase beige fat thermogenesis**. *Cell* (2014) **157**. DOI: 10.1016/j.cell.2014.03.065
3. Li ZY, Song J, Zheng SL, Fan MB, Guan YF, Qu Y. **Adipocyte metrnl antagonizes insulin resistance through PPARγ signaling**. *Diabetes* (2015) **64**. DOI: 10.2337/db15-0274
4. Jung TW, Lee SH, Kim HC, Bang JS, Abd El-Aty AM, Hacımüftüoğlu A. **METRNL attenuates lipid-induced inflammation and insulin resistance**. *Exp Mol Med* (2018) **50** 1-11. DOI: 10.1038/s12276-018-0147-5
5. Lee JO, Byun WS, Kang MJ, Han JA, Moon J, Shin MJ. **The myokine meteorin-like (metrnl) improves glucose tolerance in both skeletal muscle cells and mice by targeting AMPKα2**. *FEBS J* (2020) **287**. DOI: 10.1111/febs.15301
6. Ushach I, Arrevillaga-Boni G, Heller GN, Pone E, Hernandez-Ruiz M, Catalan-Dibene J. **Meteorin-like/Meteorin-β is a novel immunoregulatory cytokine associated with inflammation**. *J Immunol* (2018) **201**. DOI: 10.4049/jimmunol.1800435
7. Ushach I, Burkhardt AM, Martinez C, Hevezi PA, Gerber PA, Buhren BA. **METEORIN-LIKE is a cytokine associated with barrier tissues and alternatively activated macrophages**. *Clin Immunol* (2015) **156**. DOI: 10.1016/j.clim.2014.11.006
8. Rupérez C, Ferrer-Curriu G, Cervera-Barea A, Florit L, Guitart-Mampel M, Garrabou G. **Meteorin-like/Meteorin-β protects heart against cardiac dysfunction**. *J Exp Med* (2021) **218** e20201206. DOI: 10.1084/jem.20201206
9. Reboll MR, Klede S, Taft MH, Cai CL, Field LJ, Lavine KJ. **Meteorin-like promotes heart repair through endothelial KIT receptor tyrosine kinase**. *Science* (2022) **376**. DOI: 10.1126/science.abn3027
10. Lee JH, Kang YE, Kim JM, Choung S, Joung KH, Kim HJ. **Serum meteorin-like protein levels decreased in patients newly diagnosed with type 2 diabetes**. *Diabetes Res Clin Pract* (2018) **135** 7-10. DOI: 10.1016/j.diabres.2017.10.005
11. Pellitero S, Piquer-Garcia I, Ferrer-Curriu G, Puig R, Martínez E, Moreno P. **Opposite changes in meteorin-like and oncostatin m levels are associated with metabolic improvements after bariatric surgery**. *Int J Obes (Lond)* (2018) **42**. DOI: 10.1038/ijo.2017.268
12. Du Y, Ye X, Lu A, Zhao D, Liu J, Cheng J. **Inverse relationship between serum metrnl levels and visceral fat obesity (VFO) in patients with type 2 diabetes**. *Diabetes Res Clin Pract* (2020) **161**. DOI: 10.1016/j.diabres.2020.108068
13. Schmid A, Karrasch T, Schäffler A. **Meteorin-like protein (Metrnl) in obesity, during weight loss and in adipocyte differentiation**. *J Clin Med* (2021) **10**. DOI: 10.3390/jcm10194338
14. Becher T, Palanisamy S, Kramer DJ, Eljalby M, Marx SJ, Wibmer AG. **Brown adipose tissue is associated with cardiometabolic health**. *Nat Med* (2021) **27** 58-65. DOI: 10.1038/s41591-020-1126-7
15. Villarroya F, Cereijo R, Villarroya J, Giralt M. **Brown adipose tissue as a secretory organ**. *Nat Rev Endocrinol* (2017) **13** 26-35. DOI: 10.1038/nrendo.2016.136
16. Lidell ME, Pfeifer A, Klingenspor M, Herzig S. **Brown adipose tissue in human infants**. *Brown adipose tissue. handbook of experimental pharmacology* (2019) **251**. DOI: 10.1007/164_2018_118
17. Urisarri A, González-García I, Estévez-Salguero Á, Pata MP, Milbank E, López N. **BMP8 and activated brown adipose tissue in human newborns**. *Nat Commun* (2021) **12** 5274. DOI: 10.1038/s41467-021-25456-z
18. Garcia-Beltran C, Cereijo R, Plou C, Gavaldà-Navarro A, Malpique R, Villarroya J. **Posterior cervical brown fat and CXCL14 levels in the first year of life: Sex differences and association with adiposity**. *J Clin Endocrinol Metab* (2022) **107**. DOI: 10.1210/clinem/dgab761
19. Garcia-Beltran C, Villarroya J, Plou C, Gavaldà-Navarro A, Casano P, Cereijo R. **Bone morphogenetic protein-8B levels at birth and in the first year of life: Relation to metabolic-endocrine variables and brown adipose tissue activity**. *Front Pediatr* (2022) **10**. DOI: 10.3389/fped.2022.869581
20. Hondares E, Gallego-Escuredo JM, Flachs P, Frontini A, Cereijo R, Goday A. **Fibroblast growth factor-21 is expressed in neonatal and pheochromocytoma-induced adult human brown adipose tissue**. *Metabolism* (2014) **63**. DOI: 10.1016/j.metabol.2013.11.014
21. Gallego-Escuredo JM, Gómez-Ambrosi J, Catalan V, Domingo P, Giralt M, Frühbeck G. **Opposite alterations in FGF21 and FGF19 levels and disturbed expression of the receptor machinery for endocrine FGFs in obese patients**. *Int J Obes (Lond)* (2015) **39**. DOI: 10.1038/ijo.2014.76
22. Díaz M, García C, Sebastiani G, de Zegher F, López-Bermejo A, Ibáñez L. **Placental and cord blood methylation of genes involved in energy homeostasis: Association with fetal growth and neonatal body composition**. *Diabetes* (2017) **66**. DOI: 10.2337/db16-0776
23. Wabitsch M, Brenner RE, Melzner I, Braun M, Möller P, Heinze E. **Characterization of a human preadipocyte cell strain with high capacity for adipose differentiation**. *Int J Obes Relat Metab Disord* (2001) **25** 8-15. DOI: 10.1038/sj.ijo.0801520
24. Yeo CR, Agrawal M, Hoon S, Shabbir A, Shrivastava MK, Huang S. **SGBS cells as a model of human adipocyte browning: A comprehensive comparative study with primary human white subcutaneous adipocytes**. *Sci Rep* (2017) **7** 4031. DOI: 10.1038/s41598-017-04369-2
25. Klusóczki Á, Veréb Z, Vámos A, Fischer-Posovszky P, Wabitsch M, Bacso Z. **Differentiating SGBS adipocytes respond to PPARγ stimulation, irisin and BMP7 by functional browning and beige characteristics**. *Sci Rep* (2019) **9** 5823. DOI: 10.1038/s41598-019-42256-0
26. Szatmári-Tóth M, Shaw A, Csomós I, Mocsár G, Fischer-Posovszky P, Wabitsch M. **Thermogenic activation downregulates high mitophagy rate in human masked and mature beige adipocytes**. *Int J Mol Sci* (2020) **21**. DOI: 10.3390/ijms21186640
27. Cereijo R, Gavaldà-Navarro A, Cairó M, Quesada-López T, Villarroya J, Morón-Ros S. **CXCL14, a brown adipokine that mediates brown-Fat-to-Macrophage communication in thermogenic adaptation**. *Cell Metab* (2018) **28** 750-763.e6. DOI: 10.1016/j.cmet.2018.07.015
28. Cereijo R, Quesada-López T, Gavaldà-Navarro A, Tarascó J, Pellitero S, Reyes M. **The chemokine CXCL14 is negatively associated with obesity and concomitant type-2 diabetes in humans**. *Int J Obes (Lond)* (2021) **45**. DOI: 10.1038/s41366-020-00732-y
29. Pani S, Dey S, Pati B, Senapati U, Bal NC. **Brown to white fat transition overlap with skeletal muscle during development of larger mammals: Is it a coincidence**. *J Endocr Soc* (2022) **6**. DOI: 10.1210/jendso/bvac151
30. Eaton M, Granata C, Barry J, Safdar A, Bishop D, Little JP. **Impact of a single bout of high-intensity interval exercise and short-term interval training on interleukin-6, FNDC5, and METRNL mRNA expression in human skeletal muscle**. *J Sport Health Sci* (2018) **7**. DOI: 10.1016/j.jshs.2017.01.003
31. Kang PB, Kho AT, Sanoudou D, Haslett JN, Dow CP, Han M. **Variations in gene expression among different types of human skeletal muscle**. *Muscle Nerve* (2005) **32**. DOI: 10.1002/mus.20356
32. Kaikaew K, Grefhorst A, Visser JA. **Sex differences in brown adipose tissue function: Sex hormones, glucocorticoids, and their crosstalk**. *Front Endocrinol (Lausanne)* (2021) **12**. DOI: 10.3389/fendo.2021.652444
33. Lidell ME, Betz MJ, Enerbäck S. **Two types of brown adipose tissue in humans**. *Adipocyte* (2014) **3**. DOI: 10.4161/adip.26896
34. Pilkington AC, Paz HA, Wankhade UD. **Beige adipose tissue identification and marker specificity-overview**. *Front Endocrinol (Lausanne)* (2021) **12**. DOI: 10.3389/fendo.2021.599134
35. Dezso Z, Nikolsky Y, Sviridov E, Shi W, Serebriyskaya T, Dosymbekov D. **A comprehensive functional analysis of tissue specificity of human gene expression**. *BMC Biol* (2008) **6** 49. DOI: 10.1186/1741-7007-6-4
36. Sahu B, Tikoo O, Pati B, Senapati U, Bal NC. **Role of distinct fat depots in metabolic regulation and pathological implications**. *Rev Physiol Biochem Pharmacol* (2023) **186**. DOI: 10.1007/112_2022_73
37. Andersson J, Roswall J, Kjellberg E, Ahlström H, Dahlgren J, Kullberg J. **MRI Estimates of brown adipose tissue in children - associations to adiposity, osteocalcin, and thigh muscle volume**. *Magn Reson Imaging* (2019) **58**. DOI: 10.1016/j.mri.2019.02.001
38. Drabsch T, Junker D, Bayer S, Wu M, Held C, Karampinos DC. **Association between adipose tissue proton density fat fraction, resting metabolic rate and FTO genotype in humans**. *Front Endocrinol (Lausanne)* (2022) **13**. DOI: 10.3389/fendo.2022.804874
|
---
title: 'Growth mindset in young people awaiting treatment in a paediatric
mental health service: A mixed methods pilot of a digital single-session
intervention'
authors:
- Brian CF Ching
- Sophie D Bennett
- Nicola Morant
- Isobel Heyman
- Jessica L Schleider
- Kate Fifield
- Sophie Allen
- Roz Shafran
journal: Clinical Child Psychology and Psychiatry
year: 2022
pmcid: PMC10018056
doi: 10.1177/13591045221105193
license: CC BY 4.0
---
# Growth mindset in young people awaiting treatment in a paediatric
mental health service: A mixed methods pilot of a digital single-session
intervention
## Abstract
### Background
Wait times are significant in child mental health services but may offer opportunity to promote growth mindsets in young people with physical and mental health needs. A digital growth mindset single-session intervention is effective in young people, but its use in paediatric settings has not been examined. This mixed methods pilot aimed to assess the intervention’s feasibility, acceptability, and impact in this population.
### Method
Patients aged 8–18 on waiting lists in a paediatric hospital’s specialist mental health service were offered the intervention remotely. Treatment completion and retention rates, symptoms of depression and anxiety, perceived control, and personality mindset were assessed at baseline, post-treatment, and follow-ups. Semi-structured interviews to explore the intervention’s acceptability were conducted post-treatment.
### Results
Twenty-five patients completed the intervention and 17 patients and three carers/parents were interviewed. Outcomes showed small to large improvements across time-points. Most patients reported finding the intervention enjoyable, accessible, and instilled a hope for change. They valued elements of the intervention but made suggestions for improvement.
### Conclusions
The digital growth mindset single-session intervention is feasible, acceptable, and potentially beneficial for young people with physical and mental health needs on waiting lists. Further research is warranted to examine its effectiveness and mechanism of change.
## Introduction
People’s beliefs include mental representations of the self (e.g., one’s personality, qualities, and traits) and are related to outcomes, such as adaptive responses to stressors (Erdley et al., 1997; Yeager et al., 2013), coping, and quality of life (Griggs & Walker, 2016) in young people. The belief that one’s attributes (such as feeling depressed or ability to cope with adversity) are malleable and can be developed is termed a ‘growth mindset’ and underlies adaptive functioning (Dweck, 2008). It is driven by perceived control, which comprises of primary control (control of objective events/conditions through behaviour) and secondary control (control of the psychological impact of such events/conditions) (Weisz et al., 2001, 2010). Young people with a growth mindset have a lower risk for depression and anxiety than those with beliefs that personality is unchangeable, a fixed mindset (Schleider et al., 2015; Schleider & Weisz, 2016a).
Young people with functional symptoms (a preferred term used to describe persistent physical symptoms without clear organic cause that impairs functioning; Marks & Hunter, 2015), are at particularly high risk of developing mental health problems, such as anxiety and depression; $71.7\%$ of children with mental health problems also had physical symptoms (NHS Digital, 2018). Although this specific group of young people benefit from psychological treatment, including brief interventions (Bennett et al., 2015; Catanzano et al., 2020; Moore et al., 2019), reduced access to evidence-based treatments due to long waits may have negative effects on mental health outcomes (British Medical Association, 2017). There is evidence from referral data to UK child and adolescent mental health services that children with combinations of physical health needs and emotional symptoms receive low rates of intervention and follow-up (Children’s Commissioner, 2016).
Single-session interventions (SSIs) may bridge this treatment gap and offer useful input during periods of waiting for longer term treatments. An online growth mindset SSI that aims to develop adaptive growth mindset beliefs has demonstrated effectiveness in improving perceived control, stress responses, and symptoms of depression and anxiety in youth, may enhance care, and potentially improve outcomes (Miu & Yeager, 2015; Schleider & Weisz, 2016b, 2018; Schleider et al. 2021a, 2021b). However, little is known about its effectiveness in young people with both complex physical and mental health needs. Such an intervention may be particularly beneficial as reduced perceived control is associated with psychological distress in children requiring medical care (Carpenter, 1992; Hoff et al., 2002).
This pilot aimed to:(a) Assess the feasibility and acceptability of a digital growth mindset SSI in young people on waiting lists for mental health assessment and/or treatment in a paediatric hospital, through recruitment, treatment completion, and retention rates, and qualitative interviews;(b) Preliminarily evaluate its impact on symptoms of depression and anxiety, perceived control, and personality mindset.
## Methods
A mixed methods case series design with the principal method being quantitative that is complemented by qualitative methods (see Morgan, 1998) was used to pilot the intervention. Quantitative data were collected at baseline, post-treatment, 1-month follow-up, and 3-month follow-up to assess feasibility and preliminary impact of the intervention. Qualitative data were collected using semi-structured interviews at post-treatment to enhance understanding of the quantitative data and explore acceptability of the intervention.
## Ethics
Approval was granted by the Great Ormond Street Hospital for Children NHS Foundation Trust Clinical Audit Team (reference number: 2689). Data were anonymised and no personally identifiable information were collected or described in this paper. We sought patient informed consent for publication.
## Sample
Patients aged 8–18 were recruited from waiting lists for assessment and/or treatment in a specialist mental health service in a paediatric hospital in London, United Kingdom, between March and June 2020. The patients seen in this service have complex physical and mental health needs, the majority requiring neuropsychiatric care, including Tourette syndrome (TS) and functional symptoms, as well as comorbid emotional symptoms, like depression and anxiety. Patients were excluded if they had active suicidal ideation, needed a translator, or had a chronological/developmental age below 8 years old, identified in their electronic patient records. See Figure 1 for the recruitment flowchart. Figure 1.Consort diagram.
## Intervention
The 20-to-30-minute digital growth mindset SSI was based on a pre-existing US-based intervention (Schleider & Weisz, 2018). The intervention was self-administered and accessed through the internet via the Qualtrics platform, where materials were read or listened to. The intervention covered the four B.E.S.T. elements of SSIs (Schleider et al., 2020):(a) Brain science to normalise concepts (teaching about neuroplasticity to highlight the malleability of thoughts, feelings, and behaviours, including a story about Phineas Gage (Macmillan, 2000));(b) Empower young people to a “helper/expert” role (opportunity to give advice to other young people);(c) Saying-is-believing exercises to solidify learning (internalisation of learning through open text);(d) Testimonials and evidence from valued others (research about and stories of young people who overcame difficulties).
We adapted the intervention to ensure it was appropriate for a UK sample. A focus group discussion was conducted with the research team, which included clinical psychologists, psychiatrists, and researchers ($$n = 11$$). The discussion explored four topics: (a) strengths of the intervention; (b) potential impact on patients with depression and anxiety; (c) adaptations needed to suit British patients; and (d) other improvements. We piloted the intervention with two patients who were accessing treatment in the service and sought feedback on their experiences of the intervention. Based on the focus group discussion and pilot, we added British language, audio narration, and stories of young people of different ages.
## Measures
Demographic and clinical information, including age, sex, ethnicity, physical health problems, and presenting mental health difficulties and neurodevelopmental disorders were collected from electronic patient records at baseline.
## (i) Primary outcome
Symptoms of depression and anxiety were assessed at baseline, 1-month follow-up, and 3-month follow-up using the young person self-report Revised Child Anxiety and Depression Scale (RCADS) (Chorpita et al., 2000). Comprised of 47-items, responses are rated on a four-point scale from 0 (‘never’) to 3 (‘always’).
## (ii) Secondary outcomes
Perceived primary control was assessed at baseline, post-treatment, 1-month follow-up, and 3-month follow-up using the Perceived Control Scale for Children (PCSC) (Weisz et al., 2001). Statements (e.g. ‘I can do well on tests at school if I study hard’) are rated on a four-point scale from 0 (‘very false’) to 3 (‘very true’).
The 20-item Secondary Control Scale for Children (SCSC) was used to measure perceived secondary control (Weisz et al., 2010) at baseline, post-treatment, 1-month follow-up, and 3-month follow-up. Positive and negative statements are rated on a four-point scale from 0 (‘very false’) to 3 (‘very true’).
Personality mindset was assessed using the 3-item Implicit Personality Theory Questionnaire (IPT-Q) which captures beliefs about the malleability of personality (Yeager et al., 2013) and used as a manipulation check. It was administered at baseline, post-treatment, 1-month follow-up, and 3-month follow-up. Responses are rated on a six-point scale from 1 (‘really disagree’) to 6 (‘really agree’).
## Interviews
Semi-structured interviews with patients with/without their carer/parent(s) based on patient preference were conducted and audio-recorded. The interview schedule (see Supplemental materials) was developed to explore: (a) what patients learnt; (b) what patients enjoyed and disliked; (c) perceived impact of the intervention; and (d) possible improvements. Interviews were conducted by BCFC and supervised by NM, a specialist qualitative researcher.
## Procedure
We telephoned carers/parents, and in discussion with the young person, offered an appointment to receive the intervention. Carers/parents were informed that it was optional and their decision regarding participation would not impact their child’s clinical care or position on the waiting list. Interested families were emailed information about the intervention and informed consent was sought. Consented families were emailed baseline measures to complete before the appointment.
The appointments were planned to be face-to-face in hospital. However, due to restrictions in response to coronavirus disease (COVID-19), appointments took place through telephone/video call. Prior to the appointment, families were emailed the link to the intervention. At the start of the appointment, we explained what would happen during the session and confirmed whether patients could access the intervention online. Patients were asked to complete the intervention with/without their carer/parent depending on their preference and/or need. The researcher (BCFC) was available throughout for support. Upon intervention completion, patients were asked to complete post-treatment measures that were emailed to families at the start of the appointment. We asked patients and carers/parents to participate in an optional 15-to-30 minute telephone/video call interview about their experiences of the intervention. Interested patients were given a 30-minute break before the interview.
The appointment lasted approximately one-to-two hours depending on whether an interview was conducted. All appointments were supervised by SDB and RS, clinical psychologists. We emailed families follow-up measures one- and 3-months after the appointment.
## (i) Statistical analysis
We conducted all analyses on SPSS statistical analysis software (V.25, IBM). We calculated the mean recruitment, treatment completion, and retention rates, mean subscale and total scores of the RCADS, and mean total scores of the PCSC, SCSC, and IPT-Q at baseline, post-treatment, 1-month follow-up, and 3-month follow-up. We conducted a Wilcoxon Sign-Rank Test to compare changes in scores between baseline and post-treatment/follow-ups. As a feasibility pilot, we present $95\%$ confidence interval (CI) estimations instead of p-values (Lancaster et al., 2004). We calculated standardised effect size (Cohen’s d) estimations with the formula used by G*Power (Faul et al., 2007).
## (ii) Qualitative analysis
Audio recordings of interviews were transcribed verbatim. We conducted thematic analysis (Braun & Clarke, 2006) to explore patients’ experiences of the intervention on NVivo (V.12, QSR International Pty). BCFC identified initial codes based on a sub-sample of transcripts and developed a coding frame to analyse further transcripts. This involved grouping related codes and developing themes to capture broader concepts. Themes were refined iteratively throughout the analytic process, and their conceptual coherence was discussed extensively amongst the research team (BCFC, SDB, NM, and RS). The team’s diverse perspectives (clinicians and researchers in paediatric mental health and a qualitative methodologist) were considered in these analytic discussions to enhance reflexivity (Barry et al., 1999). We used post-interview reflective field notes made by BCFC and frequent team discussions during data collection to enhance the validity of analysis (Miles & Huberman, 1994).
## Sample characteristics
Fourteen ($56\%$) patients were male and 21 ($84\%$) were White. Patients reported different physical health problems including pain and neurological conditions. The most common presenting mental health and neurodevelopmental difficulties included TS (17, $68\%$), generalised anxiety (16, $64\%$), autism spectrum disorder (ASD; 9, $36\%$), functional symptoms (8, $32\%$), and depression (7, $28\%$). Twenty-one patients ($84\%$) presented with co-occurring difficulties. See Table 1 for more details. Table 1.Baseline characteristics of patients who completed intervention and interviews. Completed intervention ($$n = 25$$)Completed interviews ($$n = 17$$)MedianMedianAge1413nnSex Male1412 Female115Ethnicity White2115 Asian31 Mixed (White and Asian)11Physical health problems Diabetes10 Respiratory22 Neurological31 Pain21 Other83Presenting mental health and neurodevelopmental difficulties* Generalised anxiety1610 Panic20 Social phobia52 Specific phobia22 Obsessions/compulsions54 Depression74 Tourette syndrome1715 Functional symptoms83 Attention deficit hyperactivity disorder35 *Autism spectrum* disorder94 Learning disability32 Multiple mental health and neurodevelopmental difficulties2114*Patients may have multiple presenting difficulties and so frequency may exceed total sample.
Thirteen patients scored above clinical threshold for the RCADS subscales and total scores at baseline (see Supplementary Table 1); the most prevalent being separation anxiety (13, $52\%$), depression (12, $48\%$), and panic (11, $44\%$). See Table 2 for the mean RCADS subscales and total scores, PCSC, SCSC, and IPT-Q.Table 2.Mean scores of outcome measures at each time-point, and change scores, standardised effect sizes, and $95\%$ confidence intervals between baseline and other time-points. Mean (SD)ChangenBaselinenPost-treatmentnOne-month follow-upnThree-month follow-upBaseline versus post-treatmentBaseline versus 1-month follow-upBaseline versus 3-month follow-upZd ($95\%$CI)Zd ($95\%$CI)Zd ($95\%$CI)RCADS separation anxiety2562.5 (14.5)1661.8 (17.1)561.4 (17.8)−.4020.04 (−0.59, 0.67)−1.0690.07 (−0.89, 1.03)RCADS generalised anxiety2553.0 (14.7)1654.1 (16.6)546.2 (15.7)−.670−0.07 (−0.70, 0.56)−.5350.45 (−0.52, 1.42)RCADS panic2558.6 (16.9)1656.6 (15.4)553.8 (16.5)−.2450.12 (−0.51, 0.75)−1.0690.29 (−0.67, 1.25)RCADS social phobia54.5 (14.0)1658.8 (13.3)555.0 (17.1)−.079−0.31 (−0.94, 0.32)−.184−0.03 (−0.99, 0.93)RCADS obsessions/compulsions2550.3 (16.2)1651.6 (15.7)547.4 (16.8)−.358−0.08 (−0.61, 0.55)−.5770.18 (−0.78, 1.14)RCADS depression2560.4 (14.8)1663.6 (16.1)558.0 (20.9)−.594−0.21 (−0.84, 0.42)−.2710.13 (−0.83, 1.09)RCADS total anxiety2557.0 (16.2)1658.4 (17.2)554.4 (20.5)−.315−0.08 (−0.71, 0.55)−.1360.14 (−0.82, 1.10)RCADS total anxiety and depression2558.1 (16.2)1659.9 (17.3)555.6 (22.1).000−0.11 (−0.74, 0.52)−.6740.13 (−0.83, 1.09)PCSC2450.3 (11.7)2551.3 (13.2)1647.5 (14.1)552.4 (10.9)−.732−0.08 (−0.64, 0.48)−.9090.22 (−0.42, 0.86)−.271−0.19 (−1.15, 0.77)SCSC2526.6 (13.6)2527.2 (15.01622.6 (14.5)535.6 (19.8)−.521−0.04 (−0.59, 0.51)−.1040.28 (−0.35, 0.91)−2.023−0.53 (−1.50, 0.44)IPT-Q2311.7 (4.6)249.2 (4.7)1610.8 (3.5)56.4 (3.8)−2.9430.54 (−0.04, 1.12)−1.2640.22 (−0.42, 0.86)−1.4611.26 (0.24, 2.28)
## (i) Recruitment, treatment completion, and retention rates
Thirty-nine patients were contacted and 29 ($74\%$) consented. Of these, 25 ($86\%$) completed the intervention. Seventeen patients and three carers/parents completed the interviews. Sixteen ($55\%$) and 5 ($17\%$) patients completed 1-month and 3-month follow-up measures, respectively. See Figure 1 for the recruitment flow.
## (i) Outcome measures
We found moderate improvement in IPT-Q ($d = 0.54$) but none in PCSC (d = −0.08) and SCSC (d = −0.04) at post-treatment. Negligible improvements were seen across measures at 1-month follow-up (see Table 2). We found large improvement in IPT-Q ($d = 1.26$), moderate improvements in generalised anxiety ($d = 0.45$) and SCSC ($d = 0.53$), and small improvements in panic ($d = 0.29$), obsessions/compulsions ($d = 0.18$), and PCSC ($d = 0.19$) at 3-month follow-up.
We found no difference in treatment effects between patients who met the clinical threshold on the RCADS scores at baseline and the full sample, and therefore did not report this.
## Acceptability
The thematic analysis produced findings about patients’ experiences of the intervention within three clusters presented below: Overall accessibility and interest; specific components of intervention; and potential perceived impact of intervention.
## (i) Overall accessibility and interest
Almost all patients reported completing the intervention independently within 10-to-30-minutes. Many patients described never having come across a similar intervention, and most reported enjoying completing it and found it clear and understandable. Some stated that the visually attractive slides, which included pictures and graphs, maintained their focus and motivation to complete the intervention. “I enjoyed how for each slide they gave you something you could see as well, like an actual image. It made it very easy to visualise and understand… I could actually pay attention to what I’m reading.” – PID 17 (male, 15, functional symptoms) Some adolescent male patients with varying mental health difficulties said they did not find the intervention interesting; no further detail was provided when prompted to elaborate. These patients were also less responsive overall in the interview.
## Audio narration
Many patients highlighted that having the option of reading or listening to a narration of the slides was helpful for sustaining attention, especially for those with attention or learning disabilities. The clear narration facilitated better understanding of the content and made their experience enjoyable.
## Research
Inclusion of research on other young people’s experiences was deemed helpful by many patients because it normalised their own experiences. Most said they were aware of the possibility of overcoming their difficulties but seeing it through research findings solidified their beliefs.
## Neuroplasticity
Many reported being particularly interested in learning about neurons and their link with personality, thoughts, feelings, and behaviours, especially the Phineas Gage story because of its gory nature. Patients expressed that although a complex topic, the content was digestible because of the clear explanations. One autistic patient who disliked human anatomy reported feeling discomfort in reading about neurons, which impaired their concentration.
## Other young people’s stories
Most patients described other young people’s stories as valuable and relatable which promoted identification, despite experiences not being identical. Many spoke about how the stories importantly emphasised how others experienced adversity too. “You know that you’re not the only one who has problems. That other people suffer too.” – PID 12 (male, 13, TS) Some reported learning strategies that others used to overcome their difficulties and manage their mood. Some patients noted that the strategies seemed easy to implement, while others wanted more clarity. A few patients noted the stories felt inauthentic. This, in addition to not being able to relate to stories from older adolescents, reduced the relatability of the stories for some. Many patients expressed wanting greater diversity in young people’s ages and difficulties to make the stories more relatable. Some felt the stories of peer difficulties at school did not capture the variety of problems they faced such as general interactions with friends or exams.
## Giving advice
Most patients reported feeling motivated to advise other young people going through difficulties in the open text; they felt proud they could potentially help others. Some described this process as helpful in consolidating their learning from the intervention. “I know that other kids would be seeing this and then they’d know how I felt about it. I felt pretty good because I know that they would be looking at it and some of them might even try it.” – PID 9 (male, 12, TS) However, some older autistic patients recounted giving advice as challenging and overwhelming; they struggled to understand the questions, comprehend what other young people may be thinking and feeling, relate to the example scenarios, and feel that their advice was sound.
## Wanting more
Some patients and their carers/parents thought the intervention could have included more on changing negative beliefs; they described seeing the value of learning about growth mindset but were unsure about how to change embedded beliefs. One carer/parent suggested including reflective journals and mind maps so patients could take them away after the intervention to practice.
## (iii) Potential impact of intervention
When asked if they would recommend the intervention to another young person going through similar difficulties to themselves, all patients endorsed the intervention and recognised its potential benefit irrespective of whether they found it helpful themselves.
## Hope of change
Many patients described feeling mistreated by others the past, which made them feel sad, confused, and angry. These patients reported that the intervention instilled hope that these young people could change. “At my college, when I first started, I had loads of friends. Gradually they all turned against me because of my illness. Before [the research], I would’ve stuck to the opinion that I don’t think people can change. Clearly, they can.” – PID 6 (female, 17, autistic, functional symptoms) Some patients also described being hopeful of change in themselves, reporting an enhanced recognition of the fluidity of their own predicament, thoughts, and feelings. Some referred to the fact that other young people could feel better as evidence for the possibility of their own change. “It gives you a sense of what other people are going through and that they’ve changed and that you can change too.” - PID 12 (male, 13, TS) However, a few patients spoke about feeling simultaneously hopeful and doubtful. Although encouraging, patients wondered if the impact of the intervention was more fleeting than permanent as they anticipated difficulty in applying their learning in daily life. “As much as I find it easy after reading it, when I actually face some situations like that, I won’t be able to hold onto it in the moment”. – PID 17 (male, 15, functional symptoms)
## New perspective
As a result of the intervention, some patients described acknowledging that their thoughts can be unhelpful, and problems can be framed positively. This extended to a deeper awareness that their thoughts, feelings, and behaviours are malleable, and may inform responses to future problems. One participant expressed that this shift in perspective made them confident in their own ability to tackle difficulties. “Say I go to someone’s party and I don’t really know anyone. I might feel more confident speaking to people now.” – PID 26 (male, 15, TS)
## Reflection
For many, hearing others’ stories brought back painful memories of their own difficulties. However, some saw this as an opportunity to reflect on past responses to problems and how they can respond adaptively in future situations. A few reflected on how they could apply their learning to different contexts, such as family conflict and the pandemic. “Is there really a problem or am I just being negative about the way I think about it? I think it would help me if I was talking to my parents or my brother or sister because we’re all stuck at home at the moment.” – PID 27 (female, 13, TS)
## Pathway to overcoming difficulties
A few patients described the intervention as a ‘first step’ in overcoming their difficulties; being cognizant of the possibility of change may promote recovery. Regardless of the presence of perceived immediate benefit, some expressed hope for long-term benefit. Another reported that completing the intervention may have made them more open to other treatments.
## Discussion
The findings from this pilot suggest that the adapted digital growth mindset SSI is feasible and acceptable for young people with physical and mental health needs on waiting lists in a paediatric hospital mental health service. High recruitment and treatment completion rates demonstrate patients are willing to receive the intervention as part of a remote appointment. Qualitative interviews suggest that most patients enjoyed completing the intervention because of the visuals, content, and computer-guided format. Most patients reported completing the intervention independently, which highlights the accessibility and feasibility of the intervention for young people. The option to complete the intervention with carer/parent support also demonstrates the possibility of flexible delivery based on individual patient needs. This supports previous findings that suggest digital interventions may be more accessible to young people than traditional treatments (Hollis et al., 2017).
We were unable to replicate published treatment effects, but this may be because our sample did not meet clinical threshold on the RCADS at baseline at a group level; this may have made it difficult to identify meaningful improvement. Previous trials identified strongest effects ($d = 0.32$–0.60 for depression, $d = 0.28$–0.33 for anxiety, and $d = 0.24$–0.27 for perceived control; Schleider & Weisz, 2018) among samples who had clinical levels of depression and anxiety (Schleider & Weisz, 2018; Schleider, Mullarkey, et al., 2021). However, the descriptive statistics indicate small to moderate effect size improvements in personality mindset, symptoms of depression and anxiety, and perceived control suggesting potential value for young people with complex needs. Facilitating the development of growth mindset that is driven by the possibility of change via self-determination and hope (Dweck, 2008) in young people with physical and mental health symptoms may improve long-term outcomes, as hope is a significant predictor of depression and anxiety in chronic illness (Rasmussen et al., 2017). As our qualitative interviews suggest, developing a growth mindset whilst on waiting lists may be an important precursor for preparing young people for psychological interventions through increased motivation.
There are limitations to this study. The self-administered nature of the intervention may have inadvertently restricted our sample to only including patients who had fewer impairing symptoms; three patients did not complete the intervention due to symptom interference like attention difficulties. Conducting interviews immediately after treatment allowed us to capture experiences of the intervention without recall problems but restricted our ability to explore the intervention’s perceived longer-term impacts. Social desirability may have influenced young people’s responses to questions about the intervention as the same researcher administered the interview and collected research data. There was low follow-up retention which may be explained by respondent burden from completing long measures such as the RCADS. Although we demonstrate positive intervention effects at follow-up, attrition may have skewed intervention effects and only captured responses from patients who experienced improvements in outcomes. The small sample at follow-up may have also limited analysis of outcomes and needs to be accounted for when interpreting effect sizes. However, this should be considered within the restraints of conducting paediatric clinical research during the pandemic (Stiles-Shields et al., 2020).
Future research should use larger samples and control groups to isolate and assess the intervention’s effectiveness in improving outcomes in paediatric samples, and nested qualitative studies in follow-ups of larger randomised controlled trials to evaluate the impact of the intervention after young people leave waiting lists (e.g. to start treatment). Continuing to conduct research using mixed methods may add further value beyond our findings and explore mechanisms of change and its implementation in child and paediatric mental health services.
Despite the limitations, the potential integration and use of digital SSIs in specialist paediatric mental health services is promising. The representative nature of the sample indicates that a brief intervention with little-to-no therapist input can be easily delivered remotely to paediatric patients. This pragmatic pilot suggests the highly accessible intervention can be offered to patients on waiting lists for mental health treatment. This is relevant to children services during COVID-19, as we have seen an uptake in technology use to maintain service provisions (Ching et al., 2021; Sharma et al., 2020). A recent trial found that a digital SSI improved mental health outcomes of students during COVID-19 (Wasil et al., 2021) suggesting that SSIs are especially useful when access to care is difficult.
The use of mixed methods provided rich data about important patient and intervention factors and areas for modification. Intervention ‘ingredients’ deemed important by patients and carers/parents were elicited by the interviews, such as the online interface, delivery options, and relatability of stories. Patients suggested varying the age and problems in the stories of young people, covering explicit strategies to identify and change negative thoughts, and providing takeaway materials to promote application of learning.
Deeper understanding of the acceptability and impact of the intervention in autistic patients is necessary as highlighted by differences in experiences identified in our qualitative findings. A recent study evaluating a longer growth mindset intervention in young people with mild to borderline intellectual disabilities reported high satisfaction in the intervention (Verberg et al., 2021), though no qualitative data was collected. This is vital as young people with communication difficulties may be more likely to endorse a fixed mindset (Brooks & Goldstein, 2013; Verberg et al., 2019).
## Conclusion
The digital growth mindset SSI is feasible, acceptable, and potentially useful for young people with complex physical and mental health needs on waiting lists for mental health treatment in a paediatric hospital. This pilot study integrates data on patients’ diverse experiences and views of the intervention, providing useful implications for clinical services and intervention modification. Further robust research is warranted to examine the intervention’s long-term effectiveness and mechanism of change.
## ORCID iD
Brian CF Ching https://orcid.org/0000-0002-2179-9793
## References
1. Barry C.
A., Britten N., Barber N., Bradley C., Stevenson F.. **Using reflexivity to optimize teamwork in qualitative research**. *Qualitative Health
Research* (1999) **9** 26-44. DOI: 10.1177/104973299129121677
2. Bennett S., Shafran R., Coughtrey A., Walker S., Heyman I.. **Psychological interventions for mental health disorders in children with chronic physical illness: A systematic review**. *Archives of Disease in Childhood* (2015) **100** 308-316. DOI: 10.1136/archdischild-2014-307474
3. Braun V., Clarke V.. **Using thematic analysis in psychology**. *Qualitative Research in
Psychology* (2006) **3** 77-101. DOI: 10.1191/1478088706qp063oa
4. British Medical Association.
(2017). Breaking down barriers – the challenge of
improving mental health outcomes. https://unitementalhealth.files.wordpress.com/2018/02/breaking-down-barriers-mental-health-briefing-apr2017.pdf. *Breaking down barriers – the challenge of
improving mental health outcomes* (2017)
5. Brooks R., Goldstein S., Goldstein S., Naglieri J.. **Changing the mindset of children and adolescents with autism spectrum disorders**. *Interventions for Autism spectrum disorders* (2013) 325-349
6. Carpenter P.
J.. **Perceived control as a predictor of distress in children undergoing invasive medical procedures**. *Journal of
Pediatric Psychology* (1992) **17** 757-773. DOI: 10.1093/jpepsy/17.6.757
7. Catanzano M., Bennett S., Sanderson C., Patel M., Manzotti G., Kerry E., Coughtrey A., Liang H., Heyman I., Shafran R.. **Brief psychological interventions for psychiatric disorders in young people with long term physical health conditions: A systematic review and meta-analysis**. *Journal of Psychosomatic Research* (2020) **136** 110187. DOI: 10.1016/j.jpsychores.2020.110187
8. Children’s Commissioner.
(2016). Lightning review: Access to child and
adolescent mental health services. https://www.childrenscommissioner.gov.uk/wp-content/uploads/2017/06/Childrens-Commissioners-Mental-Health-Lightning-Review.pdf. *Lightning review: Access to child and
adolescent mental health services* (2016)
9. Ching B. C.
F., Bennett S.
D., Heyman I., Liang H., Catanzano M., Fifield K., Berger Z., Gray S., Hewson E., Bryon M., Coughtrey A.
E., Shafran R.. *A survey of mental health professionals in a
paediatric hospital during COVID-19* (2021)
10. Chorpita B.
F., Yim L., Moffitt C., Umemoto L.
A., Francis S.
E.. **Assessment of symptoms of DSM-IV anxiety and depression in children: A revised child anxiety and depression scale**. *Behaviour Research and Therapy* (2000) **38** 835-855. DOI: 10.1016/s0005-7967(99)00130-8
11. Dweck C.
S.. **Can personality be changed? The role of beliefs in personality and change**. *Current Directions in
Psychological Science* (2008) **17** 391-394. DOI: 10.1111/j.1467-8721.2008.00612.x
12. Erdley C.
A., Loomis C.
C., Cain K.
M., Dumas-Hines F.. **Relations among children’s social goals, implicit personality theories, and responses to social failure**. *Developmental Psychology* (1997) **33** 263-272. DOI: 10.1037/0012-1649.33.2.263
13. Faul F., Erdfelder E., Lang A.
G., Buchner A.. **G* Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences**. *Behavior Research Methods* (2007) **39** 175-191. DOI: 10.3758/bf03193146
14. Griggs S., Walker R.
K.. **The role of hope for adolescents with a chronic illness: An integrative review**. *Journal of Pediatric
Nursing* (2016) **31** 404-421. DOI: 10.1016/j.pedn.2016.02.011
15. Hoff A.
L., Mullins L.
L., Chaney J.
M., Hartman V.
L., Domek D.. **Illness uncertainty, perceived control, and psychological distress among adolescents with type 1 diabetes**. *Research and Theory for Nursing
Practice* (2002) **16** 223-236. DOI: 10.1891/rtnp.16.4.223.53023
16. Hollis C., Falconer C.
J., Martin J.
L., Whittington C., Stockton S., Glazebrook C., Davies E.
B.. **Annual research review: Digital health interventions for children and young people with mental health problems–a systematic and meta‐review**. *Journal of Child Psychology and
Psychiatry* (2017) **58** 474-503. DOI: 10.1111/jcpp.12663
17. Lancaster G.
A., Dodd S., Williamson P.
R.. **Design and analysis of pilot studies: Recommendations for good practice**. *Journal of Evaluation in Clinical
Practice* (2004) **10** 307-312. DOI: 10.1111/j.2002.384.doc.x
18. Macmillan M.. **Restoring phineas gage: A 150th retrospective**. *Journal of the History of the
Neurosciences* (2000) **9** 46-66. DOI: 10.1076/0964-704x(200004)9:1;1-2;ft046
19. Marks E.
M., Hunter M.
S.. **Medically unexplained symptoms: An acceptable term?**. *British Journal of Pain* (2015) **9** 109-114. DOI: 10.1177/2049463714535372
20. Miles M.
B., Huberman A.
M.. *Qualitative data analysis: An expanded sourcebook* (1994)
21. Miu A.
S., Yeager D.
S.. **Preventing symptoms of depression by teaching adolescents that people can change: Effects of a brief incremental theory of personality intervention at 9-month follow-up**. *Clinical
Psychological Science* (2015) **3** 726-743. DOI: 10.1177/2167702614548317
22. Moore D.
A., Nunns M., Shaw L., Rogers M., Walker E., Ford T., Garside R., Ukoumunne O., Titman P., Shafran R., Heyman I., Anderson R., Dickens C., Viner R., Bennett S., Logan S., Lockhart F., Thompson
Coon J.. **Interventions to improve the mental health of children and young people with long-term physical conditions: Linked evidence syntheses**. *Health Technology
Assessment (Winchester, England)* (2019) **23** 1-164. DOI: 10.3310/hta23220
23. Morgan D.
L.. **Practical strategies for combining qualitative and quantitative methods: Applications to health research**. *Qualitative Health Research* (1998) **8** 362-376. DOI: 10.1177/104973239800800307
24. NHS Digital. (2018).
Mental health of children and young people in England.
https://digital.nhs.uk/data-and-information/publications/statistical/mental-health-of-children-and-young-people-in-england/2017/2017. *Mental health of children and young people in England* (2018)
25. Rasmussen H.
N., O’Byrne K.
K., Vandamente M., Cole B.
P., Lopez S.
J., Gallagher M.
W.. **Hope and physical health**. *The oxford
handbook of hope* (2017) 159-168
26. Schleider J., Weisz J.. **A single‐session growth mindset intervention for adolescent anxiety and depression: 9‐month outcomes of a randomized trial**. *Journal of Child Psychology and
Psychiatry* (2018) **59** 160-170. DOI: 10.1111/jcpp.12811
27. Schleider J.
L., Abel M.
R., Weisz J.
R.. **Implicit theories and youth mental health problems: A random-effects meta-analysis**. *Clinical Psychology
Review* (2015) **35** 1-9. DOI: 10.1016/j.cpr.2014.11.001
28. Schleider J.
L., Dobias M.
L., Sung J.
Y., Mullarkey M.
C.. **Future directions in single-session youth mental health interventions**. *Journal of Clinical Child &
Adolescent Psychology* (2020) **49** 264-278. DOI: 10.1080/15374416.2019.1683852
29. Schleider J.
L., Mullarkey M.
C., Fox K., Dobias M.
L., Shroff A., Hart E., Roulston C.
A.. **A randomized trial of online single-session interventions for adolescent depression during COVID-19**. *Nature Human
Behaviour* (2021) **6** 1-11. DOI: 10.1038/s41562-021-01235-0
30. Schleider J.
L., Sung J.
Y., Bianco A., Gonzalez A., Vivian D., Mullarkey M.
C.. **Open pilot trial of a single-session consultation service for clients on psychotherapy wait-lists**. *The Behavior
Therapist* (2021) **44** 8-15
31. Schleider J.
L., Weisz J.
R.. **Implicit theories relate to youth psychopathology, but how? A longitudinal test of two predictive models**. *Child
Psychiatry & Human Development* (2016) **47** 603-617. DOI: 10.1007/s10578-015-0595-2
32. Schleider J.
L., Weisz J.
R.. **Reducing risk for anxiety and depression in adolescents: Effects of a single-session intervention teaching that personality can change**. *Behaviour Research and Therapy* (2016) **87** 170-181. DOI: 10.1016/j.brat.2016.09.011
33. Sharma A., Sasser T., Gonzalez E.
S., Vander
Stoep A., Myers K.. **Implementation of a home-based telemental health in a large child psychiatry department during the COVID-19 crisis**. *Journal of Child and Adolescent
Psychopharmacology* (2020) **30** 404-413. DOI: 10.1089/cap.2020.0062
34. Stiles-Shields C., Plevinsky J.
M., Psihogios A.
M., Holmbeck G.
N.. **Considerations and future directions for conducting clinical research with pediatric populations during the COVID-19 pandemic**. *Journal of Pediatric Psychology* (2020) **45** 720-724. DOI: 10.1093/jpepsy/jsaa055
35. Verberg F., Helmond P., Otten R., Overbeek G.. **Mindset and perseverance of adolescents with intellectual disabilities: Associations with empowerment, mental health problems, and self-esteem**. *Research in Developmental
Disabilities* (2019) **91** 103426. DOI: 10.1016/j.ridd.2019.103426
36. Verberg F., Helmond P., Otten R., Overbeek G.. **Effectiveness of the online mindset intervention ‘the growth factory’for adolescents with intellectual disabilities**. *Journal of Applied Research in
Intellectual Disabilities* (2021) **35** 217-230. DOI: 10.1111/jar.12941
37. Wasil A.
R., Taylor M.
E., Franzen R.
E., Steinberg J.
S., DeRubeis R.
J.. **Promoting graduate student mental health during COVID-19: Acceptability, feasibility, and perceived utility of an online Single-Session intervention**. *Fronters in
Psychology* (2021) **12** 1167. DOI: 10.3389/fpsyg.2021.569785
38. Weisz J.
R., Francis S.
E., Bearman S.
K.. **Assessing secondary control and its association with youth depression symptoms**. *Journal of Abnormal Child
Psychology* (2010) **38** 883-893. DOI: 10.1007/s10802-010-9440-z
39. Weisz J.
R., Southam-Gerow M.
A., McCarty C.
A.. **Control-related beliefs and depressive symptoms in clinic-referred children and adolscents: Developmental differences and model specificity**. *Journal of Abnormal
Psychology* (2001) **110** 97-109. DOI: 10.1037/0021-843x.110.1.97
40. Yeager D.
S., Miu A.
S., Powers J., Dweck C.
S.. **Implicit theories of personality and attributions of hostile intent: A meta‐analysis, an experiment, and a longitudinal intervention**. *Child Development* (2013) **84** 1651-1667. DOI: 10.1111/cdev.12062
|
---
title: 'A Systematic Review of Black People Coping With Racism: Approaches,
Analysis, and Empowerment'
authors:
- Grace Jacob
- Sonya C. Faber
- Naomi Faber
- Amy Bartlett
- Allison J. Ouimet
- Monnica T. Williams
journal: Perspectives on Psychological Science
year: 2022
pmcid: PMC10018067
doi: 10.1177/17456916221100509
license: CC BY 4.0
---
# A Systematic Review of Black People Coping With Racism: Approaches,
Analysis, and Empowerment
## Body
Black people in Western societies experience mental and physical stress because of racialization. According to social norms in the United States and Canada, a person racialized as Black (which includes the subcategories African American, Black American, and Black Canadian) may originate from any country. Persons racialized as Black typically share darker skin shades; however, they may have any shade of skin. “ Black” here refers to racial grouping, which, in the United States, is defined by government census and is not the same as ethnic group and not synonymous with biological relatedness. Black is a social category, and a person racialized as Black in one country may not be considered Black at all in a different country. For the purposes of this article, Black refers to people in U.S. and Canadian society who are assumed to have African ancestry based on their appearance.
Racial discrimination occurs when a person is mistreated because of their perceived race or ethnic group (Haeny et al., 2021). A person who is regularly exposed to racial discrimination must integrate coping mechanisms into their everyday life to combat the many and ongoing adverse effects associated with race-based stress and trauma. There are many forms that racism can take, and it can occur on individual, community, and institutional levels (Harrell, 2000; J. M. Jones, 1972, 1999). Studies have identified a high prevalence of racist incidents faced by Black Americans, including Lee and colleagues [2019], who found that $69.5\%$ have experienced racial discrimination from time to time or regularly. In the United States, studies have found that Black adolescents cope with incidents of racial discrimination on average five times per day (English et al., 2020). Likewise, Cénat and colleagues [2021] recently demonstrated that at least four of every 10 Black Canadians experience racial discrimination on a weekly basis (Cénat et al., 2021).
Specific mechanisms are required by Black persons to cope with the onslaught of everyday racism, without which they would leave themselves open to significant stress and risk of traumatization, all of which can lead to self-destructive and psychologically taxing responses that may have serious ramifications on a mental and physical well-being (Paradies et al., 2015; Williams et al., 2021). Persistent experiences of racism can lead to an increase in depressive symptoms (English et al., 2020; Wheaton et al., 2018), posttraumatic stress disorder (Helms et al., 2012; Sibrava et al., 2019), and anxiety (Soto et al., 2011). This persistent exposure can also lead to an increased risk of long-term physical illness (Thames et al., 2019), obesity (Sewell, 2017; Stepanikova et al., 2017), diabetes (Bacon et al., 2017; Sewell, 2017), high blood pressure (Brondolo et al., 2011; Forde et al., 2020; Sewell, 2017), and poor birth outcomes (Alhusen et al., 2016; Mustillo et al., 2004).
A Black person may resort to various methods to continue functioning in a systemically anti-Black, racist society. A major mechanism utilized to cope with this aforementioned stress is a set of psychologically defined processes called “emotion regulation” and “coping.” Although these concepts are distinct, they do share some similar characteristics. This article describes the role of emotion regulation and coping as protective responses of Black Americans to race-related stress. Through a systematic review of the literature, we explore the gender differences in how Black people in racialized societies react to racist incidents and then provide recommendations and suggested guidelines for addressing these incidents.
## Abstract
This article reviews the current research literature concerning Black people in Western societies to better understand how they regulate their emotions when coping with racism, which coping strategies they use, and which strategies are functional for well-being. A systematic review of the literature was conducted, and 26 studies were identified on the basis of a comprehensive search of multiple databases and reference sections of relevant articles. Studies were quantitative and qualitative, and all articles located were from the United States or Canada. Findings demonstrate that Black people tend to cope with racism through social support (friends, family, support groups), religion (prayer, church, spirituality), avoidance (attempting to avoid stressors), and problem-focused coping (confronting the situation directly). Findings suggest gender differences in coping strategies. We also explore the relationship between coping with physical versus emotional pain and contrast functional versus dysfunctional coping approaches, underscoring the importance of encouraging personal empowerment to promote psychological well-being. Findings may help inform mental-health interventions. Limitations include the high number of American-based samples and exclusion of other Black ethnic and national groups, which is an important area for further exploration.
## Coping
Coping has been described as an individual’s changing cognitive and behavioral efforts to manage external and internal demands that are experienced as stressful or that exceed the person’s resources (Lazarus & Folkman, 1984). The coping process is enlisted to respond to stress, and these strategies can change over time and vary depending on the context. Furthermore, there are a multitude of different ways to cope. Skinner and colleagues [2003] identified and assembled 400 different ways of coping.
According to Bridges and Grolnick [1995], all emotion regulation is a form of coping; however, coping involves attempts to regulate one’s emotions specifically in response to a stressful event (for a review, see Skinner & Zimmer-Gembeck, 2007). Many times in a day, people are subjected to different types of stimuli that require them to regulate their emotions (Gross, 1998; Gross & Thompson, 2007; Mauss et al., 2007; Thompson, 1994); however, emotion regulation has generally been defined as the efforts people make to influence which emotions they have the moment they have them, as well as the manner in which the emotions are experienced and expressed (Gross et al., 2006). Typical strategies that are commonly used in emotion regulation include problem-solving, mindfulness, acceptance, distraction, reappraisal, rumination, worry, behavioral avoidance, expressive suppression, and experiential avoidance (Brockman et al., 2017; Naragon-Gainey et al., 2017).
An extensive body of research suggests that some strategies are more adaptive than others. In a large meta-analysis, Webb and colleagues [2012] found that expressive suppression often leads to more negative and less positive emotional experiences, as evidenced by both subjective report and physiological measures. Further, compared with women who smiled in response to distress, women who suppressed their anxiety were rated more negatively on interpersonal characteristics such as warmth and likeability (Bahl & Ouimet, 2022). Finally, people with mood and anxiety disorders report more avoidance, rumination, and expressive suppression along with less problem-solving and cognitive reappraisal (for a meta-analysis, see Aldao et al., 2010). Taken together, these findings suggest that although emotion regulation can have positive short-term benefits by enabling people to cope in the moment, the long-term negative consequences of some strategies may be severe. Understanding how Black people regulate their emotions to cope with race-related stress is thus vital to mitigate the potential downstream harmful effects not only of racism but also potentially the emotion-regulation strategies themselves.
Coping and emotion regulation have common elements that include controlled efforts, intentional efforts, and regulation processes, which have a specific temporal duration (Compas et al., 2014). One important difference between coping and emotion regulation, however, is that coping is uniquely for managing stress. In this systematic review and analysis, we examine how Black people use emotion regulation and other strategies to cope with race-related stress.
## Racial stress requires a unique approach to coping
In the Western world, life experiences differ vastly on the basis of appearance. One of the first well-known psychological experiments carried out through the lens of Black-White racism was by John Howard Griffin, a White journalist who dyed his skin Black and shaved his head to experience typical Black American life in the American South in 1959. He described his everyday experiences in his bestselling book Black Like Me (J. H. Griffin, 1961):As I walked down Mobile Street, a car full of white men and boys sped past. They yelled obscenities at me. A tangerine flew past my head and broke against a building. The street was loud and raw, with tension as thick as fog. I felt the insane terror of it. ( p. 66) *It is* this type of racial stress, described here by a White person experiencing it for the first time, that has been a part of the long history and experience of being racialized as Black in America; however, Jim Crow laws also extended into Canada, with the last racially segregated school (Nova Scotia) closing as late as 1983 (James et al., 2010; Maynard, 2017). The singularity of anti-Black sentiment is not just a relic of the past. Statistics show that even in 2018, Black Canadians were more likely than any other racial group in Canada to be the victims of a hate crime (Moreau, 2020; Taylor & Richards, 2019). Racial trauma in North America has required the cultural development of unique coping responses.
All Western nations appear to exhibit some level of anti-Black bias (Faber & Williams, 2019). Although stress reactions can vary from one person to the other and persons of all racial groups experience stress in one form or another, Black people are subject to a unique set of racial stresses that influence the way in which they wield emotion regulation and coping processes. This starts early, because the way in which children are racialized impacts the way they regulate their emotions (Gaylord-Harden & Cunningham, 2009). Studies that have been conducted on this topic explored how Black people specifically regulate their emotions to cope with racist incidents, which is a common occurrence among this population. They also considered how Black people can adapt to and cope with these stressful events.
Clark and colleagues [1999] studied the mechanism through which racism can act as a stressor for African Americans. They suggested a model that integrated biopsychosocial effects of perceived racism on these individuals, noting, however, that the extant research was insufficient. In addition, a selective review of individual-level coping strategies for combating interpersonal racism carried out by Brondolo and colleagues [2009] also emphasized the lack of research focusing on strategies people can use to cope with racism. Thus, more work is needed to understand the current ways that Black people cope with racism and ascertain the most advantageous means of coping. The most advantageous approaches would be defined as “functional” such that they contribute to and maintain well-being.
## Mapping racism-specific coping mechanisms
Faced with a dearth of concrete research, this article provides a comprehensive review of the literature to summarize the research documenting the various ways Black people in racialized societies regulate their emotions and cope when faced with racism. Therefore, we aim to determine how the deployment of these coping mechanisms can vary between men and women and discuss how they may be helpful or harmful. By gathering information from multiple articles from different databases, we compile the available literature including race-specific strategies and identify important gaps to stimulate further progress. We discuss how Black men and women tend to use different strategies to cope with racism and compare this with the experience of coping with physical pain. Finally, we contrast functional and dysfunctional coping responses and highlight the value of empowerment for psychological well-being and social change. We offer these recommendations with the hope that they may inform clinicians and researchers, paving the way toward improved mental health and well-being for all Black people who would be better protected and more empowered to act as agents of social change and address the root of the issues they face—namely racism itself.
## Method
An ongoing larger scoping review exploring cross-cultural differences in emotion regulation was initially used to find articles for the current article. The larger scoping review seeks to identify key cultural and emotion-regulation factors that are essential for building a culturally informed model of emotion regulation. However, very few of the 9,257 total articles originally identified for abstract and title screening were related to the topic of Black people’s coping response to racism. Therefore, a wide search for peer-reviewed articles was conducted using the following online databases: PsycInfo, PubMed, MEDLINE, Google Scholar, Microsoft Academic, and Scholars Portal.
Relevant articles were collected by using a combination of the following search terms: “emotion regulation,” “coping strategies,” “racial discrimination,” “racism,” “coping response,” “Black,” and “Black people.” The initial search resulted in 11,215 articles. The article’s title and abstract were screened to assess for the inclusion criteria. To be included, the article must have been (a) published in English or French, (b) published in a peer-reviewed journal, (c) related to coping or emotion regulation, (d) relevant to discrimination or racism, and (e) included Black people of any ethnicity or nationality. We also manually searched the reference sections of the initially identified articles to find additional relevant articles. After the initial search, 56 articles dated from 1996 to 2021 could be identified.
A full text review was subsequently carried out to confirm articles met the inclusion criteria. We did not include articles about racism and psychopathology if they did not also focus on coping. Articles that studied children (under 18 years old) were not included in the project because coping techniques in children would be expected to differ from coping in adults (e.g., S. C. T. Jones et al., 2020). Articles were excluded if the topic was not coping with racism or if they emphasized only biological functions such as blood pressure and heart rate as their measure of coping. Ultimately, 26 articles were included in the current review (for the literature flow diagram, see Fig. 1; Moher et al., 2009).
**Fig. 1.:** *Process of screening and selecting studies for inclusion in this
review.*
## Results
The 26 articles selected for review and analysis are listed in Tables 1 and 2 and are ordered by publication year. There were 18 quantitative and eight qualitative studies. Tables 1 and 2 also list the sample size, type of study, and key findings. Quantitative studies were also examined to help ascertain which strategies were connected to positive or helpful outcomes for participants. Notably, the majority of studies were cross-sectional, which limits our ability to draw cause-and-effect conclusions. There were, however, two daily diary studies and two longitudinal survey studies (Hoggard et al., 2012; D. M. Pittman & Kaur, 2018; Sanders Thompson, 2006; Swim et al., 2003), but of these only D. M. Pittman and Kaur [2018] and Sanders Thompson [2006] examined the utility of a coping strategy. Of the qualitative studies, two used focus groups and six used interviews, which is also noted.
These studies evaluated the common coping strategies used by Black Americans and Canadians, often with a specific focus on how they impact mental-health variables, such as depression and anxiety symptoms (Graham et al., 2013). None of the studies differentiated between sex and gender, and few differentiated between race and ethnicity. Most of them did not differentiate between Black people and African Americans; exceptions are Graham et al. [ 2013], Griffith et al. [ 2019], Spates et al. [ 2019], and Volpe et al. [ 2021]. However, surprisingly, none reported results separately by ethnic group or made subgroup comparisons. Of the 26 articles, 25 were from U.S. samples and one was from a Canadian sample; no other countries were represented. Given the predominance of American literature, the remaining findings will best apply to Black Americans.
The two most frequently mentioned coping mechanisms were religion and social support; however, there was variability in regard to the other coping mechanisms used and studied. Strategies that were presented for Black Americans and Canadians can be found in Table 3, which presents a complete overview of all the coping strategies mentioned in the articles reviewed.
**Table 3.**
| Category | Assessment |
| --- | --- |
| Avoidance (3) Disengagement, not responding (2) Distancing, escape avoidance (4) Cognitive emotional debriefing (avoidance and denial) (3) Self-blame/accepting responsibility (2) Ritual-centered coping (1) Assimilation (1) Shifting (1)Substance use (2) Smoking (1) Drugs (licit and illicit) (1) Alcohol (1) Food | Dysfunctional |
| Cognitive strategies Processing the event (1) Positive reframing/reappraisal (2) Acceptance (1) Mindfulness, meditation (2) Meaning making (1) Planning (2)Physical strategies Physical activity (2) John Henryism, working harder (4) Resisting retaliation (impulse/self-control) (2) | Ambiguous |
| Social support (10) (friends and family) Collective coping, support groups (1) Instrumental support (2) Venting (1) Humor (1) Therapy (1)Direct strategies (3) Problem-solving (3) Covert resistance (1) Agentic strategies (1) Speaking out (1) and confrontation (2) Active anger (1)Identity affirmation Positive self-statements (2) Africultural coping (2) Spirituality/religion (6) Art (1)Activism (1) Public resistance Educating others (2) Community/civic involvement (1) | Functional |
We found that the type of racist experience determined to a great extent which type of strategy people reported using. We identified three distinct levels of racism in the studies: cultural (i.e., assertion of Eurocentric Western values and practices resulting in the exclusion or denigration of other histories and traditions), interpersonal (i.e., biases that occur when an individual’s racialized beliefs affect their interactions with people of color), and institutional (i.e., differential access to goods, services, and opportunities based on perceived racial identity; see Fig. 2; Haeny et al., 2021; J. M. Jones, 1999).
**Fig. 2.:** *Coping-strategies model. Types of racism are depicted as a linked circle.
Categories of strategies used by Black individuals are wielded
preferentially in a specific way to cope with each of the three types of
racism.*
Of the 26 articles, four highlighted differences in the coping response to all or one of the types of racism listed. People reported using a combination of separate strategies to combat institutional racism, including active coping (Volpe et al., 2021) and problem-solving strategies (Joseph & Kuo, 2009). If faced with interpersonal racism, individuals emphasized spirituality-based strategies. In contrast, in the face of cultural racism, individuals chose collective coping, social support, and problem-solving (Utsey, Ponterotto, et al., 2000).
Although the cited articles show how universal many of these coping mechanisms are for Black Americans, there has been no general consensus on their efficacy as tools to combat racial stress and trauma. Observing how different forms of coping are habitually used for specific types of stressors is a useful insight for designing experiments and interventions that can inform researchers and clinicians as to which of these types of responses are most efficacious in which situations.
## Gender comparisons
The majority of studies found that Black people used a range of strategies in response to racism (Brown et al., 2011; Clark, 2004; E. K. Griffin & Armstead, 2020; Hoggard et al., 2012; Hudson et al., 2016; Joseph & Kuo, 2009; Lewis-Coles & Constantine, 2006; Pearson et al., 2014; C. T. Pittman, 2010; Shorter-Gooden, 2004; Spates et al., 2019; Utsey, Ponterotto, et al., 2000; Volpe et al., 2021). Within this variety exist several recurring coping mechanisms used widely by Black Americans regardless of gender or other factors. Multiple articles show they use religion, social support, and problem-focused coping to respond to racist experiences. Social-support mechanisms were outlined in seven publications (Griffith et al., 2019; Joseph & Kuo, 2009; Lewis-Coles & Constantine, 2006; Pearson et al., 2014; Sanders Thompson, 2006; Utsey, Ponterotto, et al., 2000; Volpe et al., 2021), five studies included religion (Brown et al., 2011; Greer et al., 2015; Joseph & Kuo, 2009; Lewis-Coles & Constantine, 2006; Pearson et al., 2014), and four included problem-focused coping (Greer et al., 2015; E. K. Griffin & Armstead, 2020; Joseph & Kuo, 2009; Plummer & Slane, 1996). In addition, the authors had similar interpretations of the usefulness of these strategies. Religion, for example, referred to attending church, prayer, and spirituality. Problem-focused coping encompassed active efforts made by an individual to directly confront the stressor to eliminate, modify, or reduce it, whereas social support described attending support groups such as Alcoholics Anonymous or talking with friends or family (Schoenmakers et al., 2015).
## Black women
Many of the studies reviewed here focused specifically on coping mechanisms used more frequently by Black women than by men. Seeking social support, for example, was found to be incredibly important for Black women in eight independent studies (Brown et al., 2011; Holder et al., 2015; Lewis-Coles & Constantine, 2006; Shorter-Gooden, 2004; Spates et al., 2019; Swim et al., 2003; Utsey, Ponterotto, et al., 2000; Volpe et al., 2021) and was described in one study as a “buffer” from “the sting of oppression” and a reminder that they were “never alone when adversity arises” (Shorter-Gooden, 2004, p. 417). Ultimately, Black women used social support to validate their difficult experiences and felt less alone in their struggles.
Religion and spirituality is another coping strategy that follows a similar pattern. It was categorized as significantly more popular for Black women in six different studies (Brown et al., 2011; Clark, 2004; Holder et al., 2015; Lewis-Coles & Constantine, 2006; Shorter-Gooden, 2004; Spates et al., 2019). Participants in Spates et al. ’s [2019] study categorized religion as a way to stay optimistic and joyful in spite of the hardships they face.
Finally, other types of strategies that were observed in Black women were “overt strategies” (or agentic strategies) that were mentioned in three studies (C. T. Pittman, 2010; Shorter-Gooden, 2004; Spates et al., 2019). This term encompasses observable responses such as confronting or speaking out (McCarty et al., 1999). In C. T. Pittman [2010], female faculty used assertive actions in the face of classroom racial stressors and to reestablish their authority. One of the Black female faculty members described how she spoke up for herself after a White student threw paper at her (C. T. Pittman, 2010). Likewise, Black women in Spates and colleagues’ [2019] study used overt strategies such as calling out discriminatory behavior. Furthermore, a participant in Shorter-Gooden’s [2004] study used active strategies to combat racism by filing a report against an officer after experiencing police abuse. Thus, as exemplified by numerous articles in this review, there are many ways overt strategies can be enacted.
In addition, “covert strategies,” which are intrapersonal actions not readily observed by others, were also observed (Aldao & Dixon-Gordon, 2014). Some Black women described trying to blend in and not stand out to avoid racism by assimilating, achieving integration through behavioral and attitude modification (Spates et al., 2019). Black American women will often adjust their behaviors and roles to reduce barriers (Hall et al., 2012). Avoidance strategies were used by Black women in four studies (Lewis-Coles & Constantine, 2006; Shorter-Gooden, 2004; Thomas et al., 2008; Utsey, Ponterotto, et al., 2000). Avoidance coping comprises avoidance of the stressors instead of becoming actively involved with them and includes minimizing or denying noxious behaviors (Holahan et al., 2005). The cognitive-emotional debriefing coping style (Utsey, Adams, & Bolden, 2000) was identified in three studies. This strategy entails forgetting about the situation, minimizing the negativity associated with the situation, and/or taking part in distracting activities (Joseph & Kuo, 2009; Lewis-Coles & Constantine, 2006; Thomas et al., 2008). Thomas et al. [ 2008] found that when more gendered racism is experienced by a Black woman, it will lead to more distress and more engagement in cognitive-emotional debriefing. Despite the misleading name ascribed to this strategy, it would be considered an avoidant approach (e.g., analogous to the concept of experiential avoidance).
Responses to the three different levels of racism identified in the previous section also differ by gender (Fig. 2). For interpersonal racism, African American women preferred avoidance strategies (Utsey, Ponterotto, et al., 2000), whereas they responded with spiritual-centered strategies, cognitive-emotional debriefing, and collective coping strategies when dealing with institutional racism (Lewis-Coles & Constantine, 2006).
In sum, Black women use a variety of strategies, both overt and covert, to cope with racism, the most common being social support and faith-based strategies. Although used more heavily by Black women, these two coping mechanisms are very common among the general *Black populous* regardless of sex.
## Black men
In regard to the different levels of racism (Fig. 2), African American men used collective coping strategies that included social support from the community, family, and friends when faced with cultural racism (Lewis-Coles & Constantine, 2006). Although this trend was identifiable, other trends relating to the coping styles of men were less clear-cut.
In contrast to literature detailing the coping mechanisms for Black women, we did not find uniform strategies among Black men. Strategies included seeking social support (Hudson et al., 2016), active anger (C. T. Pittman, 2011), substance use (Clark, 2004; Hudson et al., 2016), planning (Brown et al., 2011), religion (Hudson et al., 2016; Lewis-Coles & Constantine, 2006), not responding (Swim et al., 2003), various active strategies (Brown et al., 2011), and acceptance (Brown et al., 2011). On the basis of the literature to date, it is difficult to say with certainty which of the strategies used by Black men to cope with racist incidents are the most common or effective.
One distinct difference in coping strategies between Black men and women was that passive strategies such as ignoring (Table 4) were used more frequently by men than by women. These types of strategies are highlighted in Table 4, which compares coping strategies by gender and physical versus emotional pain caused by racism. The gender difference could be attributed to the ways in which our racialized society punishes Black men much more harshly for using strategies in which the agency is externally visible (e.g., Williams, 2020). Thus, speaking out and confrontation were primarily used as a coping strategy by Black women. The heightened use of physical activity as a coping mechanism by Black men compared with women can be seen as compensation in taking back their agency in a socially acceptable (but very circumscribed), external way. It is one of the only avenues that remains open to them because of the intensive scrutiny and surveillance many of their actions are subjected to.
**Table 4.**
| Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Coping Categorization | Coping Categorization.1 | Coping Categorization.2 |
| --- | --- | --- | --- | --- | --- | --- |
| Coping strategies for physical pain(Meints et al., 2016) | Coping strategies for emotional pain and stress of racism | Men | Women | Cognitive/behavioral | Active/ passive | Problem-/ emotion- focused |
| Relaxation | | | | Behavioral | Active | Problem |
| Catastrophizing | | | | Behavioral | Active | Emotional |
| Hoping/praying | Identity affirmation (11) | Religion, emotion-focused coping | Religion, emotion-focused coping | Cognitive | Active | Emotional |
| Seeking social support | Social support (16) | Problem-focused social support | Problem-focused social support | Behavioral | Active | Emotional |
| Ignoring pain sensations | Avoidance (9) | Avoidance, nonresponse | Avoidance, assimilation | Behavioral | Passive | Emotional |
| Diverting attention | Cognitive emotional debriefing/ritual (4) | Self-blame, denial | Self-blame, denial | Cognitive | Active | Problem |
| Coping self-statements | Acceptance (1) | Acceptance | | Cognitive | Active | Problem |
| Reinterpreting pain | Positive reframing/reappraisal (2) | Reappraisal | Reappraisal | Behavioral | Active | Problem |
| Increasing behavioral activity | Agentic strategies (8); activism (4) | Active anger, self-control | Speaking out, confrontation, civic engagement | Behavioral | Active | Problem |
| Exercising and stretching | Physical activity (2) | Physical activity | | Behavioral | Active | Problem |
| Task persistence | John Henryism (2) | Working harder | | Behavioral | Active | Problem |
| Guarding | Processing/mindfulness/meditation (6) | Planning | Rumination | Cognitive | Active | Problem |
| Not listed | Substance use (5) | Substance use | Substance use | Behavioral | Active | Problem |
## Pain and coping
Coping with racism is psychologically taxing and even painful. Although not identical, there are commonalities between how the brain processes physical versus emotional pain (Sturgeon & Zautra, 2016). Functional MRI (fMRI) studies have shown that physical pain and social rejection activate common brain regions (Eisenberger, 2012; Novembre et al., 2014). Furthermore, coping mechanisms defined as “active methods” can be seen successfully regulating pain in real time through fMRI studies (Emmert et al., 2017). A meta-analysis also demonstrated that cognitive and meditative therapies can alter the functioning of brain regions and reduce the affective experience of pain (Nascimento et al., 2018). These examples demonstrate a degree of usefulness in using active coping strategies to reduce and regulate pain. There is also an observed commonality between the coping strategies as reactions to both physical and mental pain, as reviewed by Sturgeon and Zautra [2016]. Armed with this knowledge, it is well within the realm of possibility to impute common resiliency mechanisms used by African Americans for coping with physical and emotional pain alike (Gorczyca et al., 2013; Sturgeon & Zautra, 2016).
## Racial differences in coping with pain
For this reason, it may be instructive to reassess the literature examining the differences between how White and Black Americans cope with physical pain and contrast this with data generated here in comparison. The first holistic publication to examine and quantify the relationship between race and the use of pain-coping strategies, a meta-analysis, found that for physical pain, Black individuals tend to use more types of coping strategies more frequently than White individuals (Meints et al., 2016). Specifically, Black individuals utilize five separate strategies—praying and hoping, diverting attention, catastrophizing, and reinterpreting pain sensations—all with more frequency than White individuals. White people, in most cases, made use of a single strategy (task persistence) and relied on this strategy more frequently than African Americans. The magnitude of the differences observed between racial groups was larger for passive versus active strategies, emotion-focused versus problem-focused strategies, and cognitive versus behavioral strategies (Meints et al., 2016).
A reconceptualized graph of data from Meints and colleagues [2016] focusing on effect sizes for pain-coping strategies used by Black Americans contrasted with White Americans outlines the wider array of strategies used (Fig. 3). The coping mechanisms used by Black participants in response to racism were somewhat similar to but also notably different from those utilized by Black participants in response to physical pain.
**Fig. 3.:** *Coping strategies for pain and effect sizes. As shown by the
reconceptualized graph of data from Meints et al.’s (2016)
meta-analysis of the effect sizes for coping strategies as utilized
differentially between Black and White Americans (a), Black people (blue
bars) used more types of coping strategies than White people (orange
bars), and although there were some commonalities, each race largely had
specific coping-strategy preferences. The asterisk indicates
significance. Black people (blue bars) used cognitive, emotion-focused,
problem-focused, and passive strategies more than White people (b). Mean
sizes below 0.1 (light blue bars) were not significant. As one of the
only significant empirical meta-analyses focused on responses of Black
individuals, there is a recognizable overlap between Black-specific
coping strategies for physical pain and emotional responses, providing
some indication for the validity of current findings, although it is not
divided by gender.*
Meints and colleagues [2016] identified 12 specific types of coping strategies for physical pain, which were subcategories of six broader categorizations (Table 4, column 1). When grouping the specific strategies into these categories (Table 4, column 2), the differences and similarities become clearer. Although Black individuals use catastrophizing or relaxing as coping strategies for physical pain, these strategies do not seem to be used at all to cope with racism. After creating this comparison, we noted an absence of any relaxation coping strategies in the face of the stress of racism, which is alarming. In therapeutic practice, it can be a hurdle to convince Black patients suffering from burnout to take even one day off, which may be due to the stigma of Western myths surrounding the work ethic of people racialized as Black (e.g., Quaye et al., 2019). The epidemic underutilization by Black people of self-care is in itself worthy of further research and an indication that more self-care may be warranted.
Black Americans are, however, using some of the same strategies used for coping with physical pain (Table 4) to cope with emotional pain, although these strategies differ somewhat by gender. When visualized in this way, a few interesting trends are immediately visible. First, religion is a widely utilized coping strategy for both physical and emotional pain for both genders. The use of social support among Black participants is also a common coping strategy for physical pain and for pain caused by racism. Ignoring pain is a strategy that is rarely used by Black people when the pain is physical. However, avoidance strategies, including ignoring, are used by both genders in response to racism. Physical activity, in contrast, is a coping strategy used for both physical and emotional stress, but it was only noted for Black men when coping with racism. Substance use, although not included as a coping category in Meints et al. [ 2016], has been documented in many other studies as a means of coping with physical and emotional pain (Garland et al., 2013; Heggeness et al., 2019; Novak et al., 2009).
Interestingly, women and men sometimes had opposite outcomes for the same strategies for coping with racism (Table 4, row 9); although both strategies are behavioral, active, and problem-focused, women tend to utilize activism as a response to racism, whereas Black men do not cope in this way but rather tend to choose to respond with self-control (Riley et al., 2021; Volpe et al., 2021). This may be indicative (again) of the very different societal punishment relegated specifically to Black men.
This kind of comparative mapping allows us to see coping trends from a bird’s-eye perspective to make useful inferences. Interestingly, in general, Black people rarely use passive coping strategies for emotional pain, although they do somewhat for physical pain. Passive coping strategies for racism (Table 4) include only nine examples, in contrast to a full 54 active strategies. The idea of having agency would therefore seem to be a common thread in coping strategies for racism; the ability to be able to do something (anything) links the responses of Black people and is notable in this context, although not all active strategies are positive (i.e., cognitive-emotional debriefing). Certain coping strategies that incorporate agency are also called “agentic” (Table 3) in the literature (Merluzzi et al., 2015). Here, it has been specifically noted that agentic coping strategies are effective but not always feasible for Black people because of racism backlash. In a trauma such as racism that is caused by the lack of agency, the determination of an individual to take up agency in any way no matter how small to reclaim a sense of autonomy is therapeutic. The frequent use of active coping identified here testifies to the potential usefulness of such strategies for future holistic-therapy approaches developed to mitigate trauma-induced racism, and such approaches should include teachings on how to find positive agency.
## Discussion
Men and women primarily use the same categories of coping strategies (identity affirmation and social support); any differences appear to be shaped by cultural forces and social latitude (agentic, activism, and physical activity), with activism being more dangerous for men than women and physical activity being more culturally accepted for men than women. The interesting commonality between coping strategies used by Black people for both pain and racism are seen in that the most utilized strategies for pain (hoping/praying) and for racism (identity affirmation) are connected by a spiritual thread, and again, these strategies seem to have been shaped by similar cultural forces; the spirituality developed by the *African diaspora* in America developed in part to cope with centuries of brutal pressures by a hateful society. Further research in this area is clearly needed to have a more in-depth understanding of the relative efficacy of these common strategies.
## Functional versus dysfunctional coping methods
It should be relatively clear that some coping methods for racism are more functional than others (Table 3). How functionality is defined will determine whether the goal is to stop acts of racism from occurring or to reduce the emotional distress caused by racist acts in the short term. The former may result in increased stress in the short term but could offer greater dividends in the long term (Jacob et al., 2021). Although not all racism can be prevented, individuals can reduce racism they experience through deliberate actions, such as changing their environment, confronting offenders when safe to do so, and reporting more dangerous perpetrators. Actions that are instead focused on changing one’s emotional or mental state may have a more immediate impact in terms of reducing distress, but actions that have greater utility in putting a stop to racism may end up being a more useful method of coping and facilitate a reduction in racism in the person’s overall social environment. That said, it is important to keep in mind that, at an individual level, even the most dysfunctional approaches have an important purpose for the person making use of it, and more beneficial approaches may not be available or apparent to them.
We expound on two distinct categories of coping strategies, emotion-focused coping (Table 4, rows 3–5) and problem-focused coping (Table 4, rows 6–13), and each of these has different ways of being effective. Emotion-focused coping involves all efforts made to decrease the emotional impacts of a stressful experience (Schoenmakers et al., 2015). The following strategies are included in this category: mindfulness, positive reframing, venting, acceptance, and processing the event. Problem-focused coping involves all active efforts made by an individual to handle stressful experiences by modifying or eliminating the stressor. Speaking out and confrontation are examples of problem-focused coping. This type of coping style can be helpful or harmful. A full list can be found in Table 4.
## Dysfunctional coping
Many frequently used coping strategies such as problem- and emotion-based coping can have detrimental effects on an individual’s psychosocial function. In turn, this can give rise to more dysfunction in the individual. One example of a harmful coping mechanism often used by Black *Americans is* John Henryism, a work ethic famously named after the fable of the Black man that worked himself to death competing against a steam engine. This coping style is particularly prevalent in Black middle-class families who believe that enough hard work will eventually lead to them being successful and recognized as equal to their White counterparts (Bennett et al., 2004). Some articles have identified this strategy as being positive in the short term because it can promote hard work and minimize conflict (Hoggard et al., 2012). However, in the long term, it is found to cause or accelerate debilitating physical ailments (Felix et al., 2019; S. Jones et al., 2019; Volpe et al., 2018). Other types of dysfunctional coping mechanisms include rumination and disengagement, which can result in or be caused by depression (Koppe & Rothermund, 2017; Nolen-Hoeksema, 2000); avoidance (including cognitive-emotional debriefing), which increases anxiety in the long term (Hofmann & Hay, 2018); unrealistic positive reframing or self-blame, which can intensify denial and may result in internalized racism and depression (Zahn et al., 2015); and problematic substance use, which can bring about negative outcomes such as dependence and health problems (D. M. Pittman & Kaur, 2018).
## Ambiguous coping
The manner in which coping strategies are used can determine whether they are helpful or harmful to individuals. Humor is a strategy used by many to cope with stressful situations. Research has demonstrated that positive (good-natured) humor can be a useful emotion-regulation strategy for negative emotions because it helps enable reappraisal of the situation in a less harmful way (Samson & Gross, 2012), whereas negative (mean-spirited) humor creates emotional distance from negative experiences more akin to avoidance strategies and engenders an overall hostile attitude.
Discerning when to avoid escalation by not responding is a critical skill for Black Americans and can also be an essential survival mechanism. An individual can evade negative or even dangerous confrontations that often can be deeply triggering or intensely traumatizing (e.g., Smith et al., 2007). This type of avoidance can ensure personal safety both on a physical and emotional level. However, not responding also causes several long-term psychological issues. For one, this coping strategy can lead to the internalization of racial trauma, which in turn can cause internalized self-hatred, repressed anger, and depression, which can later surface in self-destructive or unhealthy manifestations (e.g., Greer et al., 2015). It also depletes a person’s sense of autonomy and agency.
Confrontation coping has shown to be useful in creating good outcomes such as psychological forgiveness as well as decreased arousal and positive emotions (Hershcovis et al., 2018; Lohr et al., 2007). Likewise, speaking out in other forms can be equally beneficial because it facilitates agency and allows the individual to overcome feelings of powerlessness (Sue et al., 2019). In cases in which confrontation results in heightened risks of anger, persecution, or retaliation from others, speaking out can be dangerous and should be carefully considered (Bushman, 2002; Hershcovis et al., 2018).
Although anger can be a motivator for positive changes and is a normal response for those who feel wronged, active anger can be destructive in several ways. It can lead to impulsive actions or words that are emotionally abusive or violent, and suppressing one’s anger can lead to symptoms of depression (C. T. Pittman, 2011). The consequences of showing anger are higher for Black people than others (Hutchings & Haddock, 2008) because it can result in the involvement of law enforcement. Therefore, Black Americans must exercise more self-control and avoid public displays of anger.
Last, acceptance and mindfulness are two strategies that can have positive impacts for coping because they provide mechanisms for accepting the emotions caused by racist events and decrease anxiety (Brockman et al., 2017; Graham et al., 2013; Hwang & Chan, 2019); however, they can potentially become a net negative if they bring about the acceptance of recurrent racist mistreatment and allow for constant exposure to racism that leads to even greater racial stress or trauma (Sobczak & West, 2013; Williams, 2020). They are also not an effective means of stopping future occurrences of racism.
## Functional coping
Numerous coping strategies mentioned in this article have been identified as helpful for coping with general life stressors, in particular physical pain (Meints et al., 2016), which provides some evidence that they may be beneficial for the emotional pain of coping with racism as well. All of the helpful strategies are active in nature and can be explored on a spectrum from least active to more active coping strategies.
Planning is the least active coping strategy as a response to racism. This strategy involves preparing for the emotional impacts of racism by actively examining the manner in which to cope with a future event (Brown et al., 2011). Although it is a thought exercise, the focus of this strategy is on the future and not the past or the present. Studies have shown that people of color may begin coping before experiencing a racial event in anticipation of that situation to reduce the effects of this stressor (DeLapp & Williams, 2021). More research should be conducted to address the paucity of literature that exists in relation to the use of prestressor coping in response to racial discrimination.
Moving down the spectrum, the act of venting involves speaking to yourself or another person emotionally about a past experience, although venting can also be journaled. This strategy offers the possibility of processing racial incidents in a concrete way as well as externalizes the event (Brown et al., 2011). Venting can be carried out in the presence of another person; however, it essentially describes a process of speaking out critically about a past event.
Social support is one of the most beneficial and effective coping mechanisms used to counter racial trauma (Driscoll et al., 2015). Social-safety networks and communities give individuals a way to express themselves, provide self-care, obtain feedback, assume agency, and establish resilience. Further, social networks contribute to external positive affirmation. Positive affirmation from a source external to oneself is deeply therapeutic, which is why it is so effective (Stock et al., 2018). Other forms of coping such as the aforementioned venting prove more successful when an individual has strong social connections to fall back on (Seawell et al., 2014).
Racism results in an emotional insult to one’s identity. Therefore, forms of increasingly active and targeted coping mechanisms such as Africultural coping and religious practice that minister to and offer antidotes specifically for injury to the identity represent higher order coping strategies. We call these identity-affirming coping strategies (e.g., Anderson et al., 2019). Africultural coping examines a racist event through the lens of identity and may use community and social support in a safe and uplifting manner (Stock et al., 2018; Utsey, Adams, & Bolden, 2000). The result is essentially an act of reclaiming the value, beauty, and purpose of an individual from the attempt of racism to negate one’s identity. It is different, however, from other social-support networks such as religion, because the whole of one’s identity cannot mentally healthfully be defined through ethnonationalism, which is central to the concept of Africultural coping.
The most commonly used coping strategy is religious practice. Religion is among the most active strategies and is frequently utilized by women for several reasons. Religion can offer an established and extensive positive viewpoint on self-identity and provide meaning as to a person’s role in the world. For many Christian Black Americans, the broader perspective has roots in the undeniable intrinsic worth and beauty of the Black individual and the specificity of purpose for which they have been created (Ellison, 1993). This value lies outside of Western value systems and therefore even if disputed by racist thought supersedes its effects. As a coping strategy, through religious inquiry, Black people remind themselves that their mental and physical attributes serve a positive preordained purpose. This understanding of one’s higher purpose can replace the distorted image of oneself that is conjured by racism and likewise thwarts anti-Black racist indoctrination.
Positive affirmation of identity is, through religious inquiry, also linked with the concept of transformative meaning. Transformative meaning is a type of positive reframing that can be effective when used as a strategy to reframe suffering, as that occurring from a racist event. It is the well-founded concept that negative, and even events that are traumatic, are able to provide character-redeeming experience if they can be seen as a learning experience on a life-long journey (Miller et al., 2020; Peterson et al., 2008). This thinking, if grounded in the framework of a belief in a benevolent creator with a positive intent and purposeful identity, can also be effective.
Finding a positive purpose and reclaiming identity through transformative meaning is a powerful coping mechanism (Stock et al., 2018). Although the study of coping with racism is modern, these techniques have been in use since antiquity. In an example of how historical figures have used coping mechanisms to deal with racism, consider a famous poem created by the great Black Renaissance professor and linguist Juan Latino (1518–1594), the son of the Count of Cabra in Cordoba and an enslaved Ethiopian woman. He had renamed himself after winning his freedom, and he was active in noble society around the time Spain was on the cusp of a racial holy war—the extermination of Jews and people of color. In the following verse, he reflects on racial relations, among other contemporary events, writing: Pious kings often keep wonderful things at court, so they can show them off to other kings. Generations of rulers, the Power of Rome itself might rightly envy you, [newborn king] Philip, for having a Black poet. ( lines 35–38) Here, after taking up agency by renaming himself, he uses humor, venting, and a transformative cognitive process with self-affirmation of his identity to cope with the nascent anti-Black racism of his time. Both the variety of coping mechanisms and the variety within categories of coping testify to the creative resiliency of humans as demonstrated by the prose that suffering has left behind. The preference by Black individuals for identity-affirming coping strategies that allow individuals to reclaim their identity and reaffirm their self-worth and dignity can explain the frequent use of such strategies. These higher order identity-affirming methods may provide better results as effective coping strategies. Future studies should be designed to confirm these observations.
## Eliminating racism and empowerment
Although helpful in the short term and on a personal level, many of these strategies may not be effective in alleviating the fundamental cause of the pain being experienced or in making meaningful reductions in the lived experience of racism. Genuine social change requires a commitment to strategies that reduce racism everywhere. An understanding of the history, source, and nature of racism is a necessary prerequisite for remedying the stress experienced by Black individuals caused by racism, because it can be difficult to come up with solutions when the core nature of the issue is incompletely grasped. Black people will be best prepared to cope with racism when they understand how racism operates, feel secure in their identity, and are well equipped to address racism as it arises in the moment (Jacob et al., 2021). In this way forceful personal agency in the face of racist actions can serve a therapeutic purpose. It is empowering when an individual can choose how to respond in the moment to a racist confrontation, and it imparts an added measure of control over the discriminatory experience (e.g., Sue et al., 2019).
Problem-focused coping, including direct strategies such as active confrontation, agentic coping, or resolution of a specific racism event, are strategies that can be successful as coping mechanisms; however, they must be carefully navigated. The possibility of asymmetrical backlash from powerful individuals is a risk. If possible, active-confrontation situations should be designed with community (support networks) and not individual involvement in mind. Utilizing these types of coping activities such as activism, that address structural racism issues, is a more beneficial avenue. With such coping activities it is possible to address the source of racism; this can provide a satisfying and therapeutic coping experience that will reap long-term benefits not just for the individual involved but for society at large (e.g., Carlson et al., 2018). Structural racism, as embodied in housing, employment, and education, represent the major community burden of racism, and proper targets therefore for change, because their resolution can contribute to a more wide-ranging alleviation of racism. Activities as coping strategies that specifically address these community burdens can be therapeutic in ways that lift up both the individual and the community and therefore create a positive feedback loop and make the community itself more resilient against racist threats. Additionally targeted and thoughtful community involvement in itself increases the number and quality of outlets for social support. Such communal activities can be as varied as the creation of a community garden, after-school tutoring, removal of polluting/environmental eyesores, cataloguing Black-owned business, classes on homeownership, or how to build a resume (Haldane et al., 2019; Lam et al., 2016).
One helping behavior that can be an effective coping method is educating others about the way racism operates (Table 3). This behavior also provides a virtuous circle by decreasing the general level of racism in the community. Awareness of racism helps the community understand how they themselves commit acts and contribute to racist systems, and in turn this awareness allows them to more easily make antiracist choices.
Notably absent from coping strategies were self-focused self-care approaches—behaviors strictly for personal well-being. These would include things such as fitness, relaxation, enjoyable personal pursuits, time off work, shopping, massage, individual psychotherapy, aromatherapy, listening to music, and turning off the phone and unplugging from stressful social media (e.g., Hansson et al., 2005; Quaye et al., 2019). Only one study in our review mentioned exercise (Hudson et al., 2016). Many authors have noted the benefits of self-care in terms of recovering from racial stress and trauma (Bryant-Davis & Ocampo, 2006; Quaye et al., 2019). It could be that Black people feel they cannot relax or pursue pleasurable activity lest they be judged in accordance with negative stereotypes as being lazy. Taking time for one’s own personal wants and needs could be viewed as an act of empowerment. This is clearly an area in need of future study.
Empowerment is essential to enable racialized people to move toward eliminating racism, which is what is ultimately needed. Positive social change will occur as racism becomes increasingly unacceptable and society becomes more equitable. Making value-based contributions to anti-racist and social-justice causes that work to dismantle racism can be a coping act of agency and self-affirmation (Hope et al., 2018; Jacob et al., 2021).
## Clinical implications
For clinicians seeking ways to support Black clients with racial trauma, the successful coping strategies enumerated here can serve as model starting points and should provide clients with greater agency and better outcomes (Heard-Garris et al., 2021; Hope et al., 2018) than the use of an ambiguous strategy. Therapy should be palpable positive affirmation; clients should feel validated and empowered. If they are religious, finding purpose in their experience even if it was negative can have a positive therapeutic effect. Helping clients find a coping strategy that affirms their intrinsic worth and beauty can also be profoundly therapeutic. If clients do not have affirmative social-support networks, or have dysfunctional social support, helping them find positively affirming support can be highly beneficial. Encouraging clients to create and make art, music, or prose out of their racist experience through positive reframing can be a transformative and proactive coping mechanism (Miller et al., 2020; Stuckey & Nobel, 2010). Certain forms of activism furthermore seem to have specific mental-health benefits (Heard-Garris et al., 2021; Montagno & Garrett-Walker, 2022; Riley et al., 2021). Ensuring that the coping mechanism chosen allows clients to reclaim their identity and dignity is essential. It is important to keep in mind that activism comes in many forms and may or may not involve formal protests or a Black Lives Matter event (E. K. Griffin & Armstead, 2020). Black clients can look for opportunities to promote antiracist change in their personal environments as well (work, school, community) through any number of prosocial means. For a cognitive-behavioral approach to helping clients with racial stress and trauma, see Williams et al. ( in press).
## Limitations and future directions
After reviewing the literature, it is clear that much remains to be learned about the role of emotion regulation and coping as strategies for individuals navigating racism. As evidenced by the summary tables, nearly all of the articles reviewed here come from U.S. examples. The exception is Joseph and Kuo’s [2009] study that was done in Canada and focused on Black Canadians. It was rare to identify studies that addressed racialized Black populations other than those in the United States, which may limit the generalizability of the findings.
Racial discrimination based on skin shade is prevalent in many countries (e.g., Chen & Francis-Tan, 2021). The dearth of research in this area demonstrates a lack of attention to global contexts that may impact Black people in other countries, their trauma and subsequent coping strategies, and race-based experiences. Racialized individuals in American and Canadian society live in a unique context, with a violent and oppressive national history. Therefore, it is expected that different racial groups develop different coping responses in opposition to these maladaptive acculturation forces. Thus, it is reasonable to assume that there are social, cultural, and ethnic differences in regard to identifying the best strategies for coping with racism. For example, African Americans might not cope with racism in the same way as first-generation Black Caribbeans. However, these nuances are not taken into account in the current literature. For this reason, it is difficult to generalize these findings to persons racialized as Black globally because they do not include the majority of them. The topic of racially specific and effective coping strategies as an issue needs to be addressed in future research to provide a well-rounded and comprehensive view of how all people racialized as Black respond to racism.
Race and ethnicity were conflated in this study because most articles reviewed did not report these characteristics separately. Sex and gender were conflated for the same reasons. Relatedly, there is a lack of research that concentrates on the responses of Black LGBTQ+ people or other intersectionalities (i.e., class, disability, etc.) other than gender. Very few articles reviewed accounted for the challenge of living with intersectional identities. For example, the way queer Black Americans face racism might not be the same as cis-heterosexual African Americans people; therefore, this would be important to study and compare. Research demonstrates that the experienced reality for American Black women in regard to racism is different because it is also gendered. For this reason, it is very plausible that Black LGBTQ+ members’ responses to racist incidents have been influenced by their gender and/or sexuality (Spates et al., 2019). Without addressing these gaps in the literature, it will be impossible to quantify the magnitude, frequency, and range of responses of Black people to racism or the efficacy of their strategies or responses to racism. The global coping strategies of Black people are likely to be as varied as their experiences.
There were some limitations in distilling the data for the tables; specifically, in creating Table 3, the method used to sort the coping strategies into categories was based on the description provided in each article on that specific coping response. The articles were not always consistent in their use of terminology for various coping strategies, which made the process challenging, and there may certainly be room for refinement as more data are collected.
Although many studies drew connections between types of coping styles and well-being, most were descriptive, and even among those that were able to show statistical relationships between well-being and coping styles, almost none were designed such that directionality could be determined (for an exception, see D. M. Pittman & Kaur, 2018). This is clearly an area that would benefit from longitudinal and experimental research paradigms to quantify functionality.
Although seldom mentioned in the research literature, it is well-known that acts of artistic prowess, including but not limited to music, visual arts, and prose, can serve as positive, active problem-focused coping mechanisms (Stuckey & Nobel, 2010). This type of creative coping differs from the previously mentioned methods because it incorporates an act of creation. Here we refer to a self-affirmative, defiant act of beauty that serves to negate the message of worthlessness present in racist speech and thought patterns. Some of the major cultural treasures of history are creative acts that defy the message of racism with a message to the world reclaiming one’s own existence, proclaiming value, and ultimately making agency itself physical, with a tangible creation. In this way we have entire art forms collectively born out of anti-Black racism (e.g., jazz, blues, rap; Secundy, 1989; Stuckey & Nobel, 2010). Future studies should examine the efficacy of a broader repertoire of coping strategies.
Finally, this review excluded children. Although our literature search unearthed some research focused on responses of African American children and adolescents to racism that may be helpful in understanding the development of racism-related coping, this must be a topic for a separate review.
## Conclusion
A spectrum of coping strategies are being used by Black people to respond to their lived experiences of racism. These strategies encompass emotion-focused coping strategies such as religion and spirituality, as well as problem-focused coping strategies such as social support. Gender differences are evident in these coping responses, with Black women prioritizing spirituality and social support. There are commonalities between the coping strategies used by Black people for stressors that are emotional versus those that are physical, some of which may be race-specific, although self-care is clearly underutilized. Black individuals noticeably prefer to utilize active strategies when coping with racism, which helps to diminish the loss of agency that accompanies racism. Therapeutic methods should deemphasize coping strategies that reinforce notions of powerlessness in favor of more functional strategies that bring about growth and change. More research focused on outcomes for well-being is needed in this important area.
## Transparency
Action Editor: June Gruber Editor: Laura A. King
## References
1. Aldao A., Dixon-Gordon K. L.. **Broadening the scope of research on emotion regulation strategies and psychopathology**. *Cognitive Behaviour
Therapy* (2014.0) **43** 22-33. DOI: 10.1080/16506073.2013.816769
2. Aldao A., Nolen-Hoeksema S., Schweizer S.. **Emotion-regulation strategies across psychopathology: A meta-analytic review**. *Clinical Psychology Review* (2010.0) **30** 217-237. DOI: 10.1016/j.cpr.2009.11.004
3. Alhusen J. L., Bower K. M., Epstein E., Sharps P.. **Racial discrimination and adverse birth outcomes: An integrative review**. *Journal of Midwifery & Women’s Health* (2016.0) **61** 707-720. DOI: 10.1111/jmwh.12490
4. Anderson R. E., McKenny M. C., Stevenson H. C.. **EMBRace: Developing a racial socialization intervention to reduce racial stress and enhance racial coping among black parents and adolescents**. *Family
Process* (2019.0) **58** 53-67. DOI: 10.1111/famp.12412
5. Bacon K. L., Stuver S. O., Cozier Y. C., Palmer J. R., Rosenberg L., Ruiz-Narváez E. A.. **Perceived racism and incident diabetes in the Black Women’s Health Study**. *Diabetologia* (2017.0) **60** 2221-2225. DOI: 10.1007/s00125-017-4400-6
6. Bahl N., Ouimet A. J.. **Smiling won’t make you feel better, but it might make people like you more: Interpersonal and intrapersonal consequences of response-focused emotion regulation strategies**. *Journal of Social and Personal
Relationships* (2022.0) **39** 2262-2284. DOI: 10.1177/02654075221077233
7. Bennett G. G., Merritt M. M., Sollers J.
J., Edwards C. L., Whitfield K. E., Brandon D. T., Tucker R. D.. **Stress, coping, and health outcomes among African-Americans: A review of the John Henryism hypothesis**. *Psychology & Health* (2004.0) **19** 369-383. DOI: 10.1080/0887044042000193505
8. Bridges L. J., Grolnick W. S., Eisenberg N.. *Review of personality and social
psychology* (1995.0) **Vol. 15** 185-211
9. Brockman R., Ciarrochi J., Parker P., Kashdan T.. **Emotion regulation strategies in daily life: Mindfulness, cognitive reappraisal and emotion suppression**. *Cognitive Behaviour Therapy* (2017.0) **46** 91-113. DOI: 10.1080/16506073.2016.1218926
10. Brondolo E., Brady Ver Halen N., Pencille M., Beatty D., Contrada R. J.. **Coping with racism: A selective review of the literature and a theoretical and methodological critique**. *Journal of Behavioral Medicine* (2009.0) **32** 64-88. DOI: 10.1007/s10865-008-9193-0
11. Brondolo E., Love E. E., Pencille M., Schoenthaler A., Ogedegbe G.. **Racism and hypertension: A review of the empirical evidence and implications for clinical practice**. *American Journal of Hypertension* (2011.0) **24** 518-529. DOI: 10.1038/ajh.2011.9
12. Brown T. L., Phillips C. M., Abdullah T., Vinson E., Robertson J.. **Dispositional versus situational coping: Are the coping strategies African Americans use different for general versus racism-related stressors?**. *Journal of Black Psychology* (2011.0) **37** 311-335. DOI: 10.1177/0095798410390688
13. Bryant-Davis T., Ocampo C.. **A therapeutic approach to the treatment of racist-incident-based trauma**. *Journal of Emotional Abuse* (2006.0) **6** 1-22. DOI: 10.1300/J135v06n04_01
14. Bushman B. J.. **Does venting anger feed or extinguish the flame? Catharsis, rumination, distraction, anger and aggressive responding**. *Personality and Social
Psychology Bulletin* (2002.0) **28** 724-731. DOI: 10.1177/0146167202289002
15. Carlson M. D., Endsley M., Motley D., Shawahin L. N., Williams M. T.. **Addressing the impact of racism on veterans of color: A race-based stress and trauma intervention**. *Psychology of Violence* (2018.0) **8** 748-762. DOI: 10.1037/vio0000221
16. Cénat J. M., Hajizadeh S., Dalexis R. D., Ndengeyingoma A., Guerrier M., Kogan C.. **Prevalence and effects of daily and major experiences of racial discrimination and microaggressions among Black individuals in Canada**. *Journal of
Interpersonal Violence* (2021.0). DOI: 10.1177/08862605211023493
17. Chen J. M., Francis-Tan A.. **Setting the tone: An investigation of skin color bias in Asia**. *Race and
Social Problems* (2022.0) **14** 150-169. DOI: 10.1007/s12552-021-09329-0
18. Clark R.. **Interethnic group and intraethnic group racism: Perceptions and coping in Black university students**. *Journal of Black Psychology* (2004.0) **30** 506-526. DOI: 10.1177/0095798404268286
19. Clark R., Anderson N. B., Clark V., Williams D.. **Racism as a stressor for African Americans. A biopsychosocial model**. *The
American Psychologist* (1999.0) **54** 805-816. DOI: 10.1037//0003-066x.54.10.805
20. Compas B. E., Jaser S. S., Dunbar J. P., Watson K. H., Bettis A. H., Gruhn M. A., Williams E. K.. **Coping and emotion regulation from childhood to early adulthood: Points of convergence and divergence**. *Australian Journal of
Psychology* (2014.0) **66** 71-81. DOI: 10.1111/ajpy.12043
21. DeLapp R. C. T., Williams M. T.. **Preparing for racial microaggressions: The role of cognition and emotion in the proactive coping process of African American college students**. *New
Ideas in Psychology* (2021.0) **63**. DOI: 10.1016/j.newideapsych.2021.100897
22. Driscoll M. W., Reynolds J. R., Todman L. C.. **Dimensions of race-related stress and African American life satisfaction: A test of the protective role of collective efficacy**. *Journal of Black
Psychology* (2015.0) **41** 462-486. DOI: 10.1177/0095798414543690
23. Eisenberger N. I.. **Broken hearts and broken bones: A neural perspective on the similarities between social and physical pain**. *Current Directions in Psychological
Science* (2012.0) **21** 42-47. DOI: 10.1177/0963721411429455
24. Ellison C. G.. **Religious involvement and self-perception among Black Americans**. *Social
Forces* (1993.0) **71** 1027-1055. DOI: 10.2307/2580129
25. Emmert K., Breimhorst M., Bauermann T., Birklein F., Rebhorn C., Van De Ville D., Haller S.. **Active pain coping is associated with the response in real-time fMRI neurofeedback during pain**. *Brain Imaging and Behavior* (2017.0) **11** 712-721. DOI: 10.1007/s11682-016-9547-0
26. English D., Lambert S. F., Tynes B. M., Bowleg L., Zea M. C., Howard L. C.. **Daily multidimensional racial discrimination among Black U.S. American adolescents**. *Journal of Applied Developmental Psychology* (2020.0) **66**. DOI: 10.1016/j.appdev.2019.101068
27. Faber S. C., Williams M. T.. **Implicit racial bias across ethnic groups and cross-nationally: Mental health implications**. (2019.0)
28. Felix A. S., Shisler R., Nolan T. S., Warren B. J., Rhoades J., Barnett K. S., Williams K. P.. **High-effort coping and cardiovascular disease among women: A systematic review of the John Henryism Hypothesis**. *Journal of Urban Health* (2019.0) **96** 12-22. DOI: 10.1007/s11524-018-00333-1
29. Forde A. T., Sims M., Muntner P., Lewis T., Onwuka A., Moore K., Diez Roux A. V.. **Discrimination and hypertension risk Among African Americans in the Jackson Heart Study**. *Hypertension* (2020.0) **76** 715-723. DOI: 10.1161/HYPERTENSIONAHA.119.14492
30. Garland E. L., Froeliger B., Zeidan F., Partin K., Howard M. O.. **The downward spiral of chronic pain, prescription opioid misuse, and addiction: cognitive, affective, and neuropsychopharmacologic pathways**. *Neuroscience and Biobehavioral Reviews* (2013.0) **37** 2597-2607. DOI: 10.1016/j.neubiorev.2013.08.006
31. Gaylord-Harden N.
K., Cunningham J. A.. **The impact of racial discrimination and coping strategies on internalizing symptoms in African American youth**. *Journal of Youth and
Adolescence* (2009.0) **38** 532-543. DOI: 10.1007/s10964-008-9377-5
32. Gorczyca R., Filip R., Walczak E.. **Psychological aspects of pain**. *Annals of Agricultural and Environmental
Medicine* (2013.0) **20** 23-27
33. Graham J. R., West L. M., Roemer L.. **The experience of racism and anxiety symptoms in an African-American sample: Moderating effects of trait mindfulness**. *Mindfulness* (2013.0) **4** 332-341. DOI: 10.1007/s12671-012-0133-2
34. Greer T. M., Ricks J., Baylor A. A.. **The moderating role of coping strategies in understanding the effects of intragroup race-related stressors on academic performance and overall levels of perceived stress for African American students**. *Journal of Black
Psychology* (2015.0) **41** 565-585. DOI: 10.1177/0095798414560018
35. Griffin E. K., Armstead C.. **Black’s coping responses to racial stress**. *Journal of Racial and Ethnic Health
Disparities* (2020.0) **7** 609-618. DOI: 10.1007/s40615-019-00690-w
36. Griffin J. H.. *Black like me* (1961.0)
37. Griffith A. N., Hurd N. M., Hussain S. B.. **“I didn’t come to school for this”: A qualitative examination of experiences with race-related stressors and coping responses among Black students attending a predominantly White institution**. *Journal of Adolescent
Research* (2019.0) **34** 115-139. DOI: 10.1177/0743558417742983
38. Gross J. J.. **The emerging field of emotion regulation: An integrative review**. *Review of
General Psychology* (1998.0) **2** 271-299. DOI: 10.1037/1089-2680.2.3.271
39. Gross J. J., Richards J. M., John O. P., Snyder D. K., Simpson J., Hughes J. N.. **Emotion regulation in everyday life**. *Emotion regulation in couples and families:
Pathways to dysfunction and health* (2006.0) 13-35. DOI: 10.1037/11468-001
40. Gross J. J., Thompson R., Gross J. J.. *Handbook of emotion regulation* (2007.0) 3-24
41. Haeny A. M., Holmes S. C., Williams M. T.. **The need for shared nomenclature on racism and related terminology**. *Perspectives on Psychological Science* (2021.0) **16** 886-892. DOI: 10.1177/17456916211000760
42. Haldane V., Chuah F., Srivastava A., Singh S. R., Koh G., Seng C. K., Legido-Quigley H.. **Community participation in health services development, implementation, and evaluation: A systematic review of empowerment, health, community, process outcomes**. *PLOS ONE* (2019.0) **14**. DOI: 10.1371/journal.pone.0216112
43. Hall J. C., Everett J. E., Hamilton-Mason J.. **Black women talk about workplace stress and how they cope**. *Journal of Black
Studies* (2012.0) **43** 207-226. DOI: 10.1177/0021934711413272
44. Hansson A., Hilleras P., Forsell Y.. **What kind of self-care strategies do people report using and is there an association with well-Being?**. *Social Indicators Research* (2005.0) **73** 133-139. DOI: 10.1007/s11205-004-0995-3
45. Harrell S. P.. **A multidimensional conceptualization of racism-related stress: Implications for the well-being of people of color**. *American Journal of
Orthopsychiatry* (2000.0) **70** 42-57. DOI: 10.1037/h0087722
46. Heard-Garris N., Ekwueme P. O., Gilpin S., Sacotte K. A., Perez-Cardona L., Wong M., Cohen A.. **Adolescents’ experiences, emotions, and coping strategies associated with exposure to media-based vicarious racism**. *JAMA Network Open* (2021.0) **4**. DOI: 10.1001/jamanetworkopen.2021.13522
47. Heggeness L. F., Lechner W. V., Ciesla J. A.. **Coping via substance use, internal attribution bias, and their depressive interplay: Findings from a three-week daily diary study using a clinical sample**. *Addictive Behaviors* (2019.0) **89** 70-77. DOI: 10.1016/j.addbeh.2018.09.019
48. Helms J. E., Nicolas G., Green C. E.. **Racism and ethnoviolence as trauma: Enhancing professional and research training**. *Traumatology* (2012.0) **18** 65-74. DOI: 10.1177/1534765610396728
49. Hershcovis M. S., Cameron A.-F., Gervais L., Bozeman J.. **The effects of confrontation and avoidance coping in response to workplace incivility**. *Journal of Occupational Health Psychology* (2018.0) **23** 163-174. DOI: 10.1037/ocp0000078
50. Hill L. K., Hoggard L. S.. **Active coping moderates associations among race-related stress, rumination, and depressive symptoms in emerging adult African American women**. *Development and Psychopathology* (2018.0) **30** 1817-1835. DOI: 10.1017/S0954579418001268
51. Hofmann S. G., Hay A. C.. **Rethinking avoidance: Toward a balanced approach to avoidance in treating anxiety disorders**. *Journal of Anxiety Disorders* (2018.0) **55** 14-21. DOI: 10.1016/j.janxdis.2018.03.004
52. Hoggard L. S., Byrd C. M., Sellers R. M.. **Comparison of African American college students’ coping with racially and nonracially stressful events**. *Cultural Diversity and Ethnic Minority
Psychology* (2012.0) **18** 329-339. DOI: 10.1037/a0029437
53. Holahan C. J., Moos R. H., Holahan C. K., Brennan P. L., Schutte K. K.. **Stress generation, avoidance coping, and depressive symptoms: A 10-year model**. *Journal of Consulting and Clinical Psychology* (2005.0) **73** 658-666. DOI: 10.1037/0022-006X.73.4.658
54. Holder A. M. B., Jackson M. A., Ponterotto J. G.. **Racial microaggression experiences and coping strategies of Black women in corporate leadership**. *Qualitative Psychology* (2015.0) **2** 164-180. DOI: 10.1037/qup0000024
55. Hope E. C., Velez G., Offidani-Bertrand C., Keels M., Durkee M. I.. **Political activism and mental health among Black and Latinx college students**. *Cultural Diversity and Ethnic Minority Psychology* (2018.0) **24** 26-39. DOI: 10.1037/cdp0000144
56. Hudson D. L., Eaton J., Lewis P., Grant P., Sewell W., Gilbert K.. **“Racism?!? . . . Just look at our neighborhoods”: Views on racial discrimination and coping among African American men in Saint Louis**. *The Journal of
Men’s Studies* (2016.0) **24** 130-150. DOI: 10.1177/1060826516641103
57. Hutchings P. B., Haddock G.. **Look Black in anger: The role of implicit prejudice in the categorization and perceived emotional intensity of racially ambiguous faces**. *Journal of
Experimental Psychology* (2008.0) **44** 1418-1420. DOI: 10.1016/j.jesp.2008.05.002
58. Hwang W.-C., Chan C. P.. **Compassionate meditation to heal from race-related stress: A pilot study with Asian Americans**. *American Journal of
Orthopsychiatry* (2019.0) **89** 482-492. DOI: 10.1037/ort0000372
59. Jacob G., Williams M. T., Faber N., Faber S., Guerrero E.. **Gender differences in coping with racism: African American experience and empowerment**. *Effective elimination of structural
racism* (2021.0). DOI: 10.5772/intechopen.99930
60. James C., Este D., Thomas Bernard W., Lloyd B., Turner T.. *Race & well-being: The lives,
hopes, and activism of African Canadians* (2010.0)
61. Jones J. M.. *Prejudice and racism* (1972.0)
62. Jones J. M., Prentice D. A., Miller D. T.. *Cultural divides: Understanding and overcoming
group conflict* (1999.0) 465-490
63. Jones S., Brooks J. H., Milam A. J., Barajas C. B., LaVeist T. A., Kane E., Furr-Holden C.. **Racial discrimination, John Henryism coping, and behavioral health conditions among predominantly poor, urban African Americans: Implications for community-level opioid problems and mental health services**. *Journal of Community
Psychology* (2019.0) **47** 1032-1042. DOI: 10.1002/jcop.22168
64. Jones S. C. T., Anderson R. E., Gaskin-Wasson A.
L., Sawyer B. A., Applewhite K., Metzger I. W.. **From “crib to coffin”: Navigating coping from racism-related stress throughout the lifespan of Black Americans**. *American Journal of
Orthopsychiatry* (2020.0) **90** 267-282. DOI: 10.1037/ort0000430
65. Joseph J., Kuo B. C. H.. **Black Canadians’ coping responses to racial discrimination**. *Journal of Black
Psychology* (2009.0) **35** 78-101. DOI: 10.1177/0095798408323384
66. Koppe K., Rothermund K.. **Let it go: Depression facilitates disengagement from unattainable goals**. *Journal of Behavior Therapy and Experimental Psychiatry* (2017.0) **54** 278-284. DOI: 10.1016/j.jbtep.2016.10.003
67. Lam C. A., Sherbourne C., Tang L., Belin T. R., Williams P., Young-Brinn A., Miranda J., Wells K. B.. **The impact of community engagement on health, social, and utilization outcomes in depressed, impoverished populations: Secondary findings from a randomized trial**. *Journal of the American Board of Family
Medicine* (2016.0) **29** 325-338. DOI: 10.3122/jabfm.2016.03.150306
68. Latino Juan. *Ad Catholicum pariter et
inuictissimum Philippum dei gratia hispaniarum Regem, de foelicissima
serenissimi Ferdinandi Principis navitate, epigrammatum liber* (1571.0) 9-12
69. Lazarus R. S., Folkman S.. *Stress, appraisal and
coping* (1984.0)
70. Lee R. T., Perez A. D., Boykin C. M., Mendoza-Denton R.. **On the prevalence of racial discrimination in the United States**. *PLOS
ONE* (2019.0) **14**. DOI: 10.1371/journal.pone.0210698
71. Lewis-Coles M. E.
L., Constantine M. G.. **Racism-related stress, Africultural coping, and religious problem-solving among African Americans**. *Cultural Diversity and Ethnic Minority
Psychology* (2006.0) **12** 433-443. DOI: 10.1037/1099-9809.12.3.433
72. Lohr J. M., Olatunji B. O., Baumeister R. F., Bushman B. J.. **The psychology of anger venting and empirically supported alternatives that do no harm**. *The Scientific Review of Mental Health
Practice: Objective Investigations of Controversial and Unorthodox Claims in
Clinical Psychology, Psychiatry, and Social Work* (2007.0) **5** 53-64
73. Mauss I. B., Bunge S. A., Gross J. J.. **Automatic emotion regulation**. *Social and Personality Psychology
Compass* (2007.0) **1** 146-167. DOI: 10.1111/j.1751-9004.2007.00005.x
74. Maynard R.. *Policing Black lives: State
violence in Canada from slavery to the present* (2017.0)
75. McCarty C. A., Weisz J. R., Wanitromanee K., Eastman K. L., Suwanlert S., Chaiyasit W., Band E. B.. **Culture, coping and context: Primary and secondary control among Thai and American youth**. *Journal of Child Psychology and Psychiatry* (1999.0) **40** 808-818. DOI: 10.1111/1469-7610.00496
76. Meints S. M., Miller M. M., Hirsh A. T.. **Differences in pain coping between Black and White Americans: A meta-analysis**. *The Journal of Pain* (2016.0) **17** 642-653. DOI: 10.1016/j.jpain.2015.12.017
77. Merluzzi T. V., Philip E. J., Zhang Z., Sullivan C.. **Perceived discrimination, coping, and quality of life for African-American and Caucasian persons with cancer**. *Cultural Diversity and Ethnic Minority
Psychology* (2015.0) **21** 337-344. DOI: 10.1037/a0037543
78. Miller T. N., Matthie N., Best N. C., Price M. A., Hamilton J. B.. **Intergenerational influences on faith-based strategies used in response to racial discrimination among young African American adults**. *Journal of the
National Medical Association* (2020.0) **112** 176-185. DOI: 10.1016/j.jnma.2020.02.005
79. Moher D., Liberati A., Tetzlaff J., Altman D. G.. **Preferred reporting items for systematic reviews and meta-analyses: The PRISMA Statement**. *PLOS
Medicine* (2009.0) **6**. DOI: 10.1371/journal.pmed.1000097
80. Montagno M. J., Garrett-Walker J.
J.. **LGBTQ+ engagement in activism: An examination of internalized heterosexism and LGBTQ+ community connectedness**. *Journal of Homosexuality* (2022.0) **69** 911-924. DOI: 10.1080/00918369.2021.1898802
81. Moreau G.. *Police-reported hate crime in Canada, 2018* (2020.0)
82. Mustillo S., Krieger N., Gunderson E. P., Sidney S., McCreath H., Keife C. I.. **Self-reported experiences of racial discrimination and Black-White differences in preterm and low-birthweight deliveries: The CARDIA Study**. *American Journal of Public Health* (2004.0) **94** 2125-2131. DOI: 10.2105/ajph.94.12.2125
83. Naragon-Gainey K., McMahon T. P., Chacko T. P.. **The structure of common emotion regulation strategies: A meta-analytic examination**. *Psychological Bulletin* (2017.0) **143** 384-427. DOI: 10.1037/bul0000093
84. Nascimento S. S., Oliveira L. R., DeSantana J. M.. **Correlations between brain changes and pain management after cognitive and meditative therapies: A systematic review of neuroimaging studies**. *Complementary Therapies in Medicine* (2018.0) **39** 137-145. PMID: 30012384
85. Nolen-Hoeksema S.. **The role of rumination in depressive disorders and mixed anxiety/depressive symptoms**. *Journal of Abnormal Psychology* (2000.0) **109** 504-511. DOI: 10.1037/0021-843X.109.3.504
86. Novak S. P., Herman-Stahl M., Flannery B., Zimmerman M.. **Physical pain, common psychiatric and substance use disorders, and the non-medical use of prescription analgesics in the United States**. *Drug
and Alcohol Dependence* (2009.0) **100** 63-70. DOI: 10.1016/j.drugalcdep.2008.09.013
87. Novembre G., Zanon M., Silani G.. **Empathy for social exclusion involves the sensory-discriminative component of pain: A within-subject fMRI study**. *Social Cognitive and Affective
Neuroscience* (2014.0) **10** 153-164. DOI: 10.1093/scan/nsu038
88. Paradies Y., Ben J., Denson N., Elias A., Priest N., Pieterse A., Gupta A., Kelaher M., Gee G.. **Racism as a determinant of health: A systematic review and meta-analysis**. *PLOS
ONE* (2015.0) **10**. DOI: 10.1371/journal.pone.0138511
89. Pearson M. R., Derlega V. J., Henson J. M., Holmes K. Y., Ferrer R. A., Harrison S. B.. **Role of neuroticism and coping strategies in psychological reactions to a racist incident among African American university students**. *Journal of
Black Psychology* (2014.0) **40** 81-111. DOI: 10.1177/0095798412471682
90. Peterson C., Park N., Pole N., D’Andrea W., Seligman M. E.. **Strengths of character and posttraumatic growth**. *Journal of Traumatic
Stress* (2008.0) **21** 214-217. DOI: 10.1002/jts.20332
91. Pittman C. T.. **Exploring how African American faculty cope with classroom racial stressors**. *The Journal of Negro Education* (2010.0) **79** 66-78
92. Pittman C. T.. **Getting mad but ending up sad: The mental health consequences for African Americans using anger to cope with racism**. *Journal of Black Studies* (2011.0) **42** 1106-1124. DOI: 10.1177/0021934711401737
93. Pittman D. M., Kaur P.. **Examining the role of racism in the risky alcohol use behaviors of black female college students**. *Journal of American College
Health* (2018.0) **66** 310-316. DOI: 10.1080/07448481.2018.1440581
94. Plummer D. L., Slane S.. **Patterns of coping in racially stressful situations**. *Journal of Black
Psychology* (1996.0) **22** 302-315. DOI: 10.1177/00957984960223002
95. Quaye S., Karikari S. N., Allen C. R., Okello W., Carter K. D.. **Strategies for practicing self-care from racial battle fatigue**. *JCSCORE* (2019.0) **5** 94-131. DOI: 10.15763/issn.2642-2387.2019.5.2.94-131
96. Riley T. N., DeLaney E., Brown D., Lozada F. T., Williams C. D., Dick D. M.. **The associations between African American emerging adults’ racial discrimination and civic engagement via emotion regulation**. *Cultural Diversity and
Ethnic Minority Psychology* (2021.0) **27** 169-175. DOI: 10.1037/cdp0000335
97. Samson A. C., Gross J. J.. **Humour as emotion regulation: The differential consequences of negative versus positive humour**. *Cognition & Emotion* (2012.0) **26** 375-384. DOI: 10.1080/02699931.2011.585069
98. Sanders Thompson V.
L. **Coping responses and the experience of discrimination**. *Journal of Applied
Social Psychology* (2006.0) **36** 1198-1214. DOI: 10.1111/j.0021-9029.2006.00038.x
99. Schoenmakers E. C., van Tilburg T. G., Fokkema T.. **Problem-focused and emotion-focused coping options and loneliness: How are they related?**. *European Journal of Ageing* (2015.0) **12** 153-161. DOI: 10.1007/s10433-015-0336-1
100. Seawell A. H., Cutrona C. E., Russell D. W.. **The effects of general social support and social support for racial discrimination on African American women’s well-being**. *The Journal of Black
Psychology* (2014.0) **40** 3-26. DOI: 10.1177/0095798412469227
101. Secundy M. G.. **Coping with words and song: The New Orleans jazz funeral**. *Literature and
Medicine* (1989.0) **8** 100-105. DOI: 10.1353/lm.2011.0087
102. Sewell A. A.. **The illness associations of police violence: Differential relationships by ethnoracial composition**. *Sociological Forum* (2017.0) **32** 975-997. DOI: 10.1111/socf.12361
103. Shorter-Gooden K.. **Multiple resistance strategies: How African American women cope with racism and sexism**. *Journal of Black Psychology* (2004.0) **30** 406-425. DOI: 10.1177/0095798404266050
104. Sibrava N. J., Bjornsson A. S., Pérez Benítez A., Moitra E., Weisberg R. B., Keller M. B.. **Posttraumatic stress disorder in African American and Latinx adults: Clinical course and the role of racial and ethnic discrimination**. *The American
Psychologist* (2019.0) **74** 101-116. DOI: 10.1037/amp0000339
105. Skinner E. A., Edge K., Altman J., Sherwood H.. **Searching for the structure of coping: A review and critique of category systems for classifying ways of coping**. *Psychological Bulletin* (2003.0) **129** 216-269. DOI: 10.1037/0033-2909.129.2.216
106. Skinner E. A., Zimmer-Gembeck M.
J.. **The development of coping**. *Annual Review of Psychology* (2007.0) **58** 119-144. DOI: 10.1146/annurev.psych.58.110405.085705
107. Smith W. A., Allen W. R., Danley L. L.. **“Assume the position . . . you fit the description” psychosocial experiences and racial battle fatigue among African American male college students**. *American Behavioral Scientist* (2007.0) **51** 551-578
108. Sobczak L. R., West L. M.. **Clinical considerations in using mindfulness- and acceptance-based approaches with diverse populations: Addressing challenges in service delivery in diverse community settings**. *Cognitive and Behavioral
Practice* (2013.0) **20** 13-22. DOI: 10.1016/j.cbpra.2011.08.005
109. Soto J. A., Dawson-Andoh N. A., BeLue R.. **The relationship between perceived discrimination and Generalized Anxiety Disorder among African Americans, Afro Caribbeans, and non-Hispanic Whites**. *Journal of Anxiety Disorders* (2011.0) **25** 258-265. DOI: 10.1016/j.janxdis.2010.09.011
110. Spates K., Evans N. M., Watts B. C., Abubakar N., James T.. **Keeping ourselves sane: A qualitative exploration of Black women’s coping strategies for gendered racism**. *Sex Roles* (2019.0) **82** 513-524. DOI: 10.1007/s11199-019-01077-1
111. Stepanikova I., Baker E. H., Simoni Z. R., Zhu A., Rutland S. B., Sims M., Wilkinson L. L.. **The role of perceived discrimination in obesity among African Americans**. *American Journal of Preventive Medicine* (2017.0) **52**. DOI: 10.1016/j.amepre.2016.07.034
112. Stock M. L., Gibbons F. X., Beekman J. B., Williams K. D., Richman L. S., Gerrard M.. **Racial (vs. self) affirmation as a protective mechanism against the effects of racial exclusion on negative affect and substance use vulnerability among black young adults**. *Journal of Behavioral
Medicine* (2018.0) **41** 195-207. DOI: 10.1007/s10865-017-9882-7
113. Stuckey H. L., Nobel J.. **The connection between art, healing, and public health: A review of current literature**. *American Journal of Public Health* (2010.0) **100** 254-263. DOI: 10.2105/AJPH.2008.156497
114. Sturgeon J. A., Zautra A. J.. **Social pain and physical pain: Shared paths to resilience**. *Pain
Management* (2016.0) **6** 63-74. DOI: 10.2217/pmt.15.56
115. Sue D. W., Alsaidi S., Awad M. N., Glaeser E., Calle C. Z., Mendez N.. **Disarming racial microaggressions: Microintervention strategies for targets, White allies, and bystanders**. *American Psychologist* (2019.0) **74** 128-142. DOI: 10.1037/amp0000296
116. Swim J. K., Hyers L. L., Cohen L. L., Fitzgerald D. C., Bylsma W. H.. **African American college students’ experiences with everyday racism: Characteristics of and responses to these incidents**. *Journal of Black
Psychology* (2003.0) **29** 38-67. DOI: 10.1177/0095798402239228
117. Taylor D., Richards D.. **Triple jeopardy: Complexities of racism, sexism, and ageism on the experiences of mental health stigma among young Canadian Black women of Caribbean descent**. *Frontiers in Sociology* (2019.0) **4**. DOI: 10.3389/fsoc.2019.00043
118. Thames A. D., Irwin M. R., Breen E. C., Cole S. W.. **Experienced discrimination and racial differences in leukocyte gene expression**. *Psychoneuroendocrinology* (2019.0) **106** 277-283. DOI: 10.1016/j.psyneuen.2019.04.016
119. Thomas A. J., Witherspoon K. M., Speight S. L.. **Gendered racism, psychological distress, and coping styles of African American women**. *Cultural Diversity and Ethnic Minority
Psychology* (2008.0) **14** 307-314. DOI: 10.1037/1099-9809.14.4.307
120. Thompson R. A.. **Emotion regulation: A theme in search of definition**. *Monographs of the Society
for Research in Child Development* (1994.0) **59** 25-52. DOI: 10.2307/1166137
121. Utsey S. O., Adams E. P., Bolden M.. **Development and initial validation of the Africultural Coping Systems Inventory**. *Journal of Black Psychology* (2000.0) **26** 194-215. DOI: 10.1177/0095798400026002005
122. Utsey S. O., Ponterotto J. G., Reynolds A. L., Cancelli A.. **Racial discrimination, coping, life satisfaction, and self-esteem among African Americans**. *Journal of Counseling &
Development* (2000.0) **78** 72-80. DOI: 10.1002/j.1556-6676.2000.tb02562.x
123. Volpe V. V., Katsiaficas D., Neal A. J.. **“Easier said than done”: A qualitative investigation of Black emerging adults coping with multilevel racism**. *Cultural Diversity and Ethnic Minority
Psychology* (2021.0) **27** 495-504. DOI: 10.1037/cdp0000446
124. Volpe V. V., Rahal D., Holmes M., Rivera S. Z.. **Is hard work and high effort always healthy for Black college students?: John Henryism in the face of racial discrimination**. *Emerging Adulthood* (2018.0) **8** 245-252. DOI: 10.1177/2167696818804936
125. Webb T. L., Miles E., Sheeran P.. **Dealing with feeling: A meta-analysis of the effectiveness of strategies derived from the process model of emotion regulation**. *Psychological
Bulletin* (2012.0) **138** 775-808. DOI: 10.1037/a0027600
126. Wheaton F. V., Thomas C. S., Roman C., Abdou C. M.. **Discrimination and depressive symptoms among African American men across the adult lifecourse**. *The Journals of Gerontology: Series
B* (2018.0) **73** 208-218. DOI: 10.1093/geronb/gbx077
127. Williams M. T.. *Managing microaggressions:
Addressing everyday racism in therapeutic spaces* (2020.0)
128. Williams M. T., Holmes S., Zare M., Haeny A. H., Faber S. C.. **An evidence-based approach for treating stress and trauma due to racism**. *Cognitive
and Behavioral Practice*
129. Williams M. T., Osman M., Gran-Ruaz S., Lopez J.. **Intersection of racism and PTSD: Assessment and treatment of racism-related stress and trauma**. *Current Treatment Options in
Psychiatry* (2021.0) **8** 167-185. DOI: 10.1007/s40501-021-00250-2
130. Zahn R., Lythe K. E., Gethin J. A., Green S., Deakin J. F., Young A. H., Moll J.. **The role of self-blame and worthlessness in the psychopathology of major depressive disorder**. *Journal of Affective Disorders* (2015.0) **186** 337-341. DOI: 10.1016/j.jad.2015.08.001
|
---
title: Nursing students’ simulated home-visit learning experiences with dementia -a
qualitative research
authors:
- Youn-Joo Um
journal: BMC Nursing
year: 2023
pmcid: PMC10018068
doi: 10.1186/s12912-023-01232-w
license: CC BY 4.0
---
# Nursing students’ simulated home-visit learning experiences with dementia -a qualitative research
## Abstract
### Background
In response to the growing demand for community nursing, practical and dynamic changes in educational methods are essential to nurturing competent nurses. The aim of this study was to explore the learning experiences of nursing students’ simulation-based community visits and understand these experiences in detail.
### Methods
This study followed Colizzi’s phenomenological research method. Nineteen participants were divided into three teams and participated in focus group interviews. The research question was: “How was your experience with the simulated nursing home visit?”
### Results
Four essential themes were identified: “burden of community nursing simulation-based learning,” “solving the problems faced by patients with dementia through teamwork,” “home-visiting nursing skills learned through physical practice,” and “community nursing competency growth.”
### Conclusion
The study results provide a basis for developing a community nursing curriculum with effective evaluation and management of community nursing home-visit education using simulation.
## Background
Dementia is a disease that negatively affects the daily life and quality of life of the elderly due to the gradual occurrence of complex physical, mental, and behavioral symptoms due to cognitive decline [1]. Patients with dementia develop symptoms such as wandering, anxiety, depression, delusions, sleep disturbances, urinary problems, and memory loss.
According to the World Health Organization [2], the approximate number of people worldwide with dementia is 50 million and is expected to more than triple to 152 million by 2050. Additionally, 5–$8\%$ of the population over the age of 60 years suffers from dementia, and it is the fifth leading cause of death worldwide.
As the population with dementia gradually increases, the admission rate of old people with dementia or cognitive decline to senior medical welfare facilities tends to increase [3], but Korea lacks sufficient facilities for dementia patients [4]. Therefore, most dementia patients receive care at home from their families rather than from specialized institutions or facilities. Therefore, for those receiving care at home, it is important for experts to provide nursing care and management home visits according to the individual situations in each home and the patient’s case because each family has different challenges, situations, and environments [5].
Accordingly, in 2017 the Ministry of Health and Welfare introduced dementia as an issue of national responsibility rather than a problem of individuals and families and has established 256 dementia relief centers nationwide centered on the Central Dementia Center to strengthen support for dementia patients and their families [4]. With the announcement of the 4th Comprehensive Dementia Management Plan (2021–2025) in September 2021, various community resources centered on the National Dementia.
Support Center were linked to provide patients with specialized management according to their degree of dementia progression and suitable facilities or medical institutions as needed. The focus was on realizing a dementia-safe society where people can live with their families in the community [4].
Accordingly, the role of home-visiting nursing at the Dementia Support *Center is* to provide various health services such as dementia patient health care, dementia symptom management, medication management, health education, nursing, and the addressing of grievances. These services can provide nursing care tailored to the individual patient’s home situation and health condition, and are recognized as professional services that help patients live with their families at home without experiencing the difficulties associated with adapting to unfamiliar hospitals [6]. Additionally, through home-visiting nursing, the main caregiver of a family member with dementia often complains of health problems such as fatigue and disease along with life stressors such as restricted social activities and increased psychological burdens. From this perspective, caregiver education and training on health management methods suitable to the circumstances of the family members of patients with dementia, and education and training on stress management methods tailored to the home can also be provided by home-visiting nurses [7].
Home-visiting nurses play a role as a mediator in the relationship between health care professionals, patients, and caregivers in providing home-visiting nursing for patients with dementia, so high-level nursing skills are highly emphasized and necessary. Accordingly, home-visiting nurses are required to have the ability to communicate and cooperate with experts for consultation and diagnosis, preventive activities, and case management, as well as the ability to cooperate with dementia patients, their families, and the community [8]. However, due to COVID-19, opportunities for field practice among nursing students in home-visiting nursing for dementia patients have significantly decreased, and providing inexperienced nursing care to dementia patients can cause negative side effects, so sufficient prior education is needed. Therefore, education using simulation is required to develop nursing students’ home-visiting competency for caring for patients with dementia [9, 10].
Simulation-based learning is effective because it mimics actual clinical situations and practices in a safe environment that is tolerant of mistakes and allows practice and repetition [11]. Simulation-based home-visit nursing research involves directly analyzing various home environments through simulation to identify problems and nurture cultural competency to suit these environments [12]. The studies available on conducting home visits through simulation identify an improvement in the perception of self-efficacy, an increase in critical thinking [13, 14], emotional control skills, empowerment, and self-efficacy in nursing students when they develop this teaching methodology in a home visit environment [15]. Additionally, a simulation of receiving care support through Korea’s health insurance, etc., was conducted [16]. However, there is no study that has explored the simulation experience of nursing students targeting dementia patients, and it is necessary to explore nursing students’ experiences through this study.
This study is expected to contribute as literature for the development of practical nursing interventions that can enhance the coping ability and nursing competency of home-visiting nurses with patients with dementia by identifying nursing students’ experiences with home-visiting nursing simulation education for dementia patients and exploring their meaning.
It further aims to reveal the essence of home-visiting nursing simulation education for dementia patients by applying a phenomenological method to confirm the nature of human behavior and the meaning of experience derived from behavior. The study question is “what is the nursing student’s experience with home-visiting nursing simulation education for dementia patients?”
## Research design
This qualitative study applied a phenomenological method to assess the value of home-visiting nursing simulation practice sessions undertaken by nursing students. The study’s purpose was to explore the essence of nursing students’ simulation-based community home-visiting nurse learning experiences and to understand associated phenomena. The research question was, “How was your experience with home-visiting nurse simulation-based learning as a nursing student?”
## Participants
The study participants were students enrolled in the 4th year of training in the Nursing Department at D University located in Y city and had already completed the home-visiting nursing simulation class. Among 61 4th-year nursing students, students who took the home-visiting nursing simulation class posted an article on the department’s community bulletin board about recruiting interested students for an interview. The purpose and method of the study were explained face-to-face to the 19 nursing students who expressed their interest in participating in the study, and ethical considerations were explained. Next, the nursing students who voluntarily expressed their intention to participate in the study completed a research consent form and scheduled a focus group interview. The selection of participants was conducted through convenience sampling until the data were saturated, and 19 people finally participated. There were no dropouts. The sociodemographic characteristics of the participants are shown in Table 1.
Table 1General Characteristics of Participants in a Focus Group Interview ($$n = 19$$)CharacteristicsCategoriesn (%)SexMale3 (15.8)Female16 (84.2)Age20–2513 (67.4)25–304 (21.0)≥ 302 (10.5)JobCollege student19 (100.0)MajorNursing19 (100.0)Year4th19 (100.0)
## Data collection
Data collection occurred from November 18, 2021, to December 17, 2021. Research data were collected through face-to-face focus group interviews. Participants were interviewed by three focus groups of five to six people. This was based on the fact that the number of participants for focus group interviews is generally composed of three groups [17]. The qualifications of the researchers included those who had doctoral degrees in nursing and experience in conducting qualitative research. In this study, the professor in charge of the home-visiting nursing simulation class participated in the focus group interviews and analysis as a researcher who is female. The researcher had a relationship with the students through the major classes before the start of the research, so it was appropriate to conduct the interview. The researcher conducted qualitative research on nursing student education for many years, published several in top journals, and attended various qualitative research method workshops. The researcher worked as a school nurse for 12 years and is currently teaching nursing students at a university. The interview was conducted in a relaxed atmosphere at the university lecture hall that is frequently used by students. Participants fully understood the purpose of the study, participated voluntarily and were willing to share their experiences. The interviews were unstructured, which allows the researcher to interview the participant with a minimum of questions.
Participants were allowed to discuss their views freely and encouraged to give a full account of their experiences. Interviews continued until the participant’s story was repeated or the discussion produced no new stories. After the first interview was recorded, a second interview was conducted individually with 3 participants by phone to avoid incomplete interview content. After conducting 19 focus group interviews, the data were saturated and no further recruitment was conducted. The duration of one interview for each focus group was approximately 60 min. The interviews were recorded as they proceeded and transcribed by the person who conducted the interview.
## The home-visit nursing simulation process
This simulation program consisted of one 100-minute session. The session covered the scenario of a nurse home-visiting a dementia patient’s home to provide nursing care. The session consisted of pre-briefing (30 min), scenario (30 min), and debriefing (40 min) in this sequence. Scenarios were placed in simulated homes that replicated real-life situations, with 3–4 students attending the scenarios while others observed.
The first step, a pre-brief (30 min), introduced the students to learning objectives and the basic rules of respect and confidentiality, thus creating a safe learning environment. After a brief introduction of the scenario by the professor, the students decided in groups how to interact with the patient through discussion. The patient role was played by a student from another team.
During the second step, the scenario (30 min), rapport formation and assessment (home environment, patient symptoms, caregiver counseling), nursing planning and intervention (dementia symptom management education, caregiver education, and psychological support, guidance on how to take medications), and nursing evaluation (evaluation of educational content, greetings, and appointments for the next visit) were conducted.
During the third debriefing phase, the nursing students share their thoughts on the scenario.
Open-ended questions such as “What do you think happened to the patient?”, “ What did you want to achieve?”, and “Tell me more about it” elicited information and allowed time for self-assessment.
After that, the session was finalized after receiving feedback from the professor and other nursing students.
## Data analysis
In this study, data collection and analysis were performed according to the analysis methods and procedures of Colaizzi [18], and data were analyzed using Microsoft Excel (Microsoft Corp., Redmond, WA, USA). As a result of the analysis, 252 codes, eight sub-categories and four categories were analyzed.
The method utilized was as follows. In Step 1, the overall meaning was grasped by reading and re-reading the transcribed content. In Step 2, the researcher reviewed the collected data and extracted the sections that were judged to have occurred in the process of remembering and representing the experiences of the research participants to extract meaningful statements from individually collected data. These extraction results were verified by three experienced phenomenological researchers to prevent ambiguous or unreasonable meanings in extracting meaningful statements. Through this, data were extracted by selecting meaningful sentences or phrases representing home-visit nursing simulation-based learning. In Step 3, while carefully examining the meaningful statements, redundant expressions were excluded, and general and abstract statements were constructed. In Step 4, the meanings that were constructed by the researcher were verified to ensure they fit the intentions. In Step 5, the statements were categorized according to the subject and essential themes were collected based on the intentions. In Step 6, the participants’ experiences were described according to the essential themes of the data analyzed to that point and the fundamental structure was stated. In Step 7, the credibility, transferability, dependability, and confirmability suggested by Guba and Lincoln [19].
] were identified to determine the study’s rigor.
For data collection credibility, open-ended questions were asked during interviews to allow participants to freely express their thoughts and experiences and to minimize midway intervention. The researcher recorded the interviews according to the focus group and transcribed the recorded material. Additionally, whether it was transcribed was checked by comparison with the recorded file, and the researcher conducted data analysis. In order to verify the credibility of the analysis, the primary and assistant moderators reviewed and discussed the similarities and differences of the data belonging to the codes and sub-categories for each stage of the analysis, and modified the name of categories. The content of the analysis and naming were reviewed by a professor with extensive experience in qualitative research. To ensure applicability, the results of the analysis were shown to another university student who had a similar experience but did not participate in the study and confirmed that it was meaningful and applicable in light of his own experience. In order to ensure auditability, the interview questions, progress, and analysis process were described in as much detail as possible. In order to maintain neutrality, during the interview it was emphasized that the researcher only fulfilled the role of the interviewer, and efforts were made to minimize and objectify the influence of the teacher-student relationship. The research director who conducted and analyzed the interview obtained a doctorate in qualitative research and is constantly active in qualitative research-related societies.
Additionally, he has conducted a number of qualitative studies, including focus group studies, targeting various participants such as community nursing, nursing students, nurses, and education.
## Ethical considerations
This study was approved by the bioethics committee of Dongyang University before data collection and was assigned ethical approval number 1041495-202202-HR-02-01 on November 17, 2021. The researchers wrote to the participants and shared the aims and methods of the research and explained that confidentiality was protected and participation in the research was completely voluntary. Informed consent was obtained from all participants included in the study. All the steps and methods were performed in accordance with the relevant guidelines and regulations. Additionally, the participants were informed that they could leave the research at any time without providing a reason. All procedures in the study were conducted according to the Declaration of Helsinki. Participants voluntarily participated. Confidentiality and anonymity were guaranteed, and the recordings were not used for any purpose other than the specified research. Participants were informed that the recorded files were password protected, not connected to the internet, and would be destroyed.
## Results
This study explored the experiences of 4th-year nursing students who participated in simulation-based home nursing home-visits and that had the meaning extracted. The phenomenological analysis method suggested by Colaizzi [18] was used for data analysis. A total of 252 meaningful statements were analyzed. Of these, eight sub-categories were chosen based on repeated or similar statements, then grouping and verification of comparable data, and four categories were derived (Table 2). The subcategories of this study comprised four subcategories: “burden of community nursing simulation-based learning,” “solving the problems faced by patients with dementia through teamwork,” “home-visiting nursing skills learned through practice,” and “growth in community nursing competency”. Finally, Nursing students’ simulated home-visit learning experiences were abstracted into one category, “cooperating and growing together in a simulated situation”.
Table 2Nursing students’ simulated home-visit learning experiencesCategoriesSubcategoriesBurden of community nursing simulation-based learningFear of unfamiliar teaching methodsLearned the difference from simple nursing skills practiceSolving the problems faced by patients with dementia through teamworkHaving fun in team classInteraction through collaborationHome-visiting nursing skills learned through practiceChanged from a passive observer to an active nursing agentMotivate learning by acknowledging one’s shortcomingsCommunity nursing competency growthTake responsibility as a professionalImproving nursing confidence
## The burden of community nursing simulation-based learning
When nursing students encountered the new subject of simulation practice, they felt burdened about how the class was conducted and taught. They experienced the burden of classes in which communication skills and nursing skills were put into a room and nursing interventions were conducted according to the situation that was required at the moment.
## Fear of unfamiliar teaching methods
The participants felt awkward and afraid about the simulation practice since it was their first time.
“It was my first time doing a simulation, so there was a lot of pressure and awkwardness to act with my classmates.” ( Participant 10).
“I also doubted whether the nursing intervention was appropriate.” ( Participant 3).
“Initially, it wasn’t easy when I realized that it was a class where practical situations were presented, and practice was conducted.” ( Participant 14).
## Learning the difference between previous nursing lessons and simple nursing skills in practice
Initially, there was a misunderstanding that the core nursing competency test was being repeated. However, after the simulation-based learning exercise began, the nursing students realized it differed from their traditional practice. Participants felt embarrassed because they had to virtually respond to unexpected situations in a given scenario.
“I thought the experience would only go according to the previous framework; however, I was embarrassed by the unexpected situation that was staged.” ( Participant 15).
“*When a* dementia patient behaves unexpectedly, I can’t think of anything, so I don’t know how to respond or what kind of nursing care to provide. Although it was a simulation situation, the same atmosphere was as the actual clinical situation.” ( Participant 10).
## Solving the problems faced by patients with dementia through teamwork
Because the simulation class is not an individual practice, but a team practice involving various participatory roles, the nursing students had not developed close relationships, but they developed positive teamwork by sharing their opinions in order to fulfill their roles well.
## Having fun with teammates
The participants had the burden of performing the test alone; however, the team-based simulation practice, which consisted of four to five people, allowed them to have fun and reduced tension.
“It was fun to practice with the team members. It was awkward at first, but it was nice to get to know each other while doing the script.” ( Participant 13).
“As we progressed as a team rather than alone, the burden of making mistakes was reduced, and as each role was designated, the burden was reduced, and we became mutually reliant on one another. I was in charge of educating the caregivers, and was able to complete it safely by dividing the responsibilities of patient medication education, vital sign measurement, and emergency response.” ( Participant 16).
## Interacting through collaboration
Participants were pleased that they could help each other and cooperate through teamwork and that team members could supplement their shortcomings. They believed that better nursing behaviors could be implemented when they collected various opinions among the team members through passionate discussion on nursing care activities for patients with dementia.
“Because we studied together with the team members, we could engage in simulation practice more passionately.” ( Participant 3).
“While preparing with the team members, I was able to learn what I was lacking or not prepared for. It was nice to be able to choose better actions based on various opinions.” ( Participant 7).
## Home-visiting nursing skills learned through practice
The Nursing students constantly discussed with team members and tried to get answers in the process of solving the problem of visiting nursing care scenarios for dementia patients. Their interest in nursing knowledge increased through recognition of the difference between the theoretical knowledge of visiting nursing for patients with dementia and the actual situation and learning about the application of the actual nursing process.
## Changing from being a passive observer to an active nursing agent
While the simulation scenario was being implemented, the participants came to understand the patient’s nursing problems and considered the appropriate nursing intervention to be provided. In community-based nursing, nursing students only observe nurses providing care, and nurses mainly provided the care themselves. However, in the simulation-based learning experience, there was an opportunity to experience home-visit nursing practices because the participants had to judge and decide how the patient would be cared for on their own.
“During clinical practice, I went to a home-visit with a nurse at a public health center, but as a student, there was very little I could do, which was regretful. However, through this simulation-based learning, I was proud to have a more active experience with home-visit nursing.” ( Participant 3).
“Through the simulation activity, I understood the patient better. There was even time for the caregiver to learn how to empathize and educate the patient.” ( Participant 8).
“During clinical practice, I watched a nurse from an observational point of view. When I actually tried providing nursing care myself, I realized that, yes, this is home-visit nursing.” ( Participant 6).
## Learning is motivated by acknowledging one’s shortcomings
The participants felt their lack of nursing hands-on experience while participating in the simulation-based nursing scenario. They recognized that they could not easily answer the patient’s unexpected questions. Therefore, they learned that they should develop the ability to respond quickly based on thorough knowledge. Additionally, they felt that their nursing skills, including therapeutic communication, were lacking, and thought that they should correct this in the future.
“*When a* patient asks me a question, I think it is important to understand the patient’s situation and have accurate nursing information to answer the question properly.” ( Participant 11).
“I don’t know how to empathize with a caregiver when they are having a hard time. I think I should study more about therapeutic communication.” ( Participant 12).
“I think it was very helpful when the professor pointed out what was wrong and gave me feedback during the debriefing time to look back on the practice after finishing the simulation.” ( Participant 5).
## Community nursing competency growth
Nursing students were proud to be able to deal with situations they hadn’t experienced in clinical practice as if they were real situations through participating in nursing simulation education for dementia patients and gained confidence in their ability to cope with situations that occur in actual nursing visits improved.
## Take responsibility as a professional
Through simulation-based learning, the participants felt a sense of responsibility and mission to provide professional nursing care that truly helps patients. They felt that each nursing procedure they performed significantly impacted the patient. This helped them decide to develop their professionalism.
“*To a* nurse, any patient is just one of many patients, but from the patient’s point of view, they are unique. A sense of responsibility grew with the recognition that patients feel you are the only medical staff they can trust and rely on.” ( Participant 15).
“I felt that the quality of nursing visits is determined by the visiting nurse. So I thought of studying to gain more specialized knowledge and skills to care for people with dementia and their families.” ( Participant 2).
“When I approached the simulation as if it was a real situation, I felt the responsibility of the nurse position. Simultaneously, while dealing with patients and caregivers, I thought about how I would feel from the perspective of the patients and caregivers.” ( Participant 3).
## Improving nursing confidence
Participants thought that their ability to identify and cope with patient problems in various changing situations had improved through participation in this practical simulation-based learning.
“Contemplating what I did well during the debriefing class, I was able to gain a lot of confidence due to the generous praise of the professors and my classmates.” ( Participant 3).
“In clinical practice, to be honest, as a nursing student, there was little I could do for myself. However, although it is standardized nursing practice for patients, I was very proud to be able to think and judge comprehensively for each situation and provide nursing care. Ah, I thought that this is how I was doing it, and I felt confident that I hadn’t felt previously in clinical practice.” ( Participant 9).
## Discussion
In this study, in order to understand the educational experience of 4th year nursing students in a visiting nursing simulation for dementia patients, the meaning and nature of the educational simulation experience were described by applying a phenomenological method. Through focus group interviews, the results of this study derived experiences related to the “burden of community nursing simulation-based learning,” “solving the problems faced by patients with dementia through teamwork,” “home-visiting nursing skills learned through physical practice,” and “growth in community nursing competency”.
Nursing students’ experiences of home-visiting nursing practice education for patients with dementia are as follows. Although the researcher provided information on how to proceed with the class before it started, the nursing students did not fully understand the simulation education initially. Additionally, they were confused because they knew that the class was solving problems on their own without an established protocol and had to interact with colleagues whom they were not familiar with. Furthermore, they complained about the burden of performing post-surgery in response to the sudden situations of dementia patients. This is a similar result to those of previous studies [20, 21] that applied team-based simulation to nursing students, and it is thought that this is because students who are accustomed to lecture-style classes feel burdened by classes that they conduct more independently. Practical training in home-visiting nursing care for patients with dementia is difficult to learn and apply in a short time [22]. Therefore, to reduce the students’ burden from the home-visiting nursing simulation classes for patients with dementia, it is necessary to participate in this type of learning several times before class and to promote interaction among classmates by sharing opinions and feelings with team members. As the class progressed, the students were able to overcome the burden of the new class method and environment that they had initially and were able to confirm their theoretical knowledge once again by sharing opinions with each other through teamwork. Also, by realizing the necessity and effectiveness of teamwork, positive teamwork developed and the students became immersed in the dementia patient simulation. Akaike, et al. [ 23] stated that in the case of simulation training, it is necessary to understand the limits of teamwork, communication, and simulation situations, as well as the learners’ practical skills.
In order to relieve the burden of the primary class method and utilize the simulation learning method more efficiently and effectively, it is necessary for students to familiarize themselves with the simulation class method before the class begins and form teams by sharing opinions with team members. Additionally, nursing students shared their feelings after the simulation class, reanalyzed the contents of practice, and exchanged opinions with the instructor and students. Through this, new knowledge was obtained, and the relationship between team members and other teams’ practices was reviewed to improve situational nursing skills and capabilities. In particular, becoming more motivated resulted from comparing one’s own performance with their team or other team students, which is a similar result as seen in previous studies [24, 25]. It is thought that nursing students formed a cooperative learning structure with peers and a positive attitude among learners rather than one-sided knowledge transfer through simulation education. The participating nursing students realized the importance of communication with patients and recognized the problems and reactions of patients through the simulation class. Additionally, it provided an opportunity to apply clinical nursing knowledge that they knew theoretically, and their confidence in performing nursing visits for patients with dementia improved.
These findings are consistent with previous studies that showed that nursing skills and team collaboration abilities were improved after patient simulation education [22, 26, 27]. In particular, this study supports it as an effective educational method to increase confidence in nursing practice, and it became an opportunity to feel a sense of responsibility through the experience of directly affecting the health of patients through their nursing. Previous studies have shown that nursing students realize their lack of professional knowledge and preparation through simulation-based learning [28, 29] and improve their confidence in nursing practice [30]. Therefore, the practice education of home-visiting nursing care for dementia patients will supplement the limitations of limited education in clinical practice and contribute to improving practical clinical performance.
Additionally, as the number of dementia patients increases and care becomes more important in the community, home-visiting nursing simulation education applies more clinical expertise and skills to nursing students, and practices problem-solving processes by situation considering the patient’s environment and available resources. It is thought that this will be helpful in the nursing practice of future nurses.
This study has some limitations including that it analyzed data from only a small number of participants from one university, so it is difficult to generalize and validate the results. Additionally, this study was conducted targeting students who were unfamiliar with the simulation class method in order to reduce biases based on existing simulation education experience. In this process, the first simulation experience may have been biased rather than the nursing home-visiting simulation experience.
However, this study presented specific and realistic educational experiences of nursing students who completed home-visiting nursing simulation education for patients with dementia. Furthermore, it provides meaningful basic data for grasping the reality of home-visiting nursing education for patients with dementia and for practical application and dissemination.
## Conclusion
Simulation classes are offered as an educational and innovative tool that favors home learning for nursing students. The use of simulation brought positive benefits to nursing students through improved self-confidence, nursing skills, communication skills, and reflective thinking. Based on this information, simulation should be considered as an instructional methodology in university education programs in community nursing subjects. This will improve clinical practice by facilitating the training of future medical professionals. Despite these promising results, further research is needed in this area to evaluate the development of simulation training scenarios for dementia prevention, simulation facilitator training, and other training using simulation methods.
## References
1. 1.Prince MJ, Wimo A, Guerchet MM, Ali GC, Wu YT, Prina M. World Alzheimer Report 2015 - The Global Impact of Dementia: An analysis of prevalence, incidence, cost, and trends. Alzheimer’s Disease International. 2015. http://www.alz.co.uk/research/world-report-2015. Accessed on 10 October 2022.
2. 2.World Health Organization. Dementia [Internet]. Geneva: World Health Organization. 2022.: https://www.who.int/news-room/fact-sheets/detail/dementia. Accessed on 11August 2022.
3. Wang J, Monroe TB, Simning A, Conwell Y, Caprio TV, Cai X, Li Y. **Pain management in home health care: relationship with dementia and facility admissions**. *Pain Manage Nurs* (2021.0) **22** 36-43. DOI: 10.1016/j.pmn.2020.06.007
4. 4.Central Dementia CenterKorea Dementia Status 2021 Report2021SeoulCentral Dementia Center. *Korea Dementia Status 2021 Report* (2021.0)
5. Kitamura T, Tanimoto C, Oe S, Kitamura M, Hino S. **Familial caregivers’ experiences with home-visit nursing for persons with dementia who live alone**. *Psychogeriatrics* (2019.0) **19** 3-9. DOI: 10.1111/psyg.12352
6. Osakwe ZT, Ikhapoh I, Arora BK, Fleur-Calixte RS. **Perception of Home Healthcare Nurses toward persons with dementia**. *Home Health Care Management & Practice* (2021.0) **33** 65-71. DOI: 10.1177/1084822320963086
7. Coe NB, Boyd CM, Chodosh J. **Chronic care, dementia care management, and financial considerations**. *J Am Med Dir Assoc* (2021.0) **22** 1371-6. DOI: 10.1016/j.jamda.2021.05.012
8. 8.World Health Organization. (2018). Continuity and coordination of care: a practice brief to support implementation of the WHO Framework on integrated people-centred health services.
9. Lairamore C, Reed CC, Damon Z, Rowe V, Baker J, Griffith K. **A peer-led interprofessional simulation experience improves perceptions of teamwork**. *Clin Simul Nurs* (2019.0) **34** 22-9. DOI: 10.1023/A:10009875017951
10. Morse C, Fey M, Kardong-Edgren S, Mullen A, Barlow M, Barwick S. **The changing landscape of simulation-based education**. *Am J Nurs* (2019.0) **119** 42-8. DOI: 10.1097/01.NAJ.0000577436.23986.81
11. Phrampus PE. **Simulation and integration into patient safety systems**. *Simul Healthc* (2018.0) **13** 225-6. DOI: 10.1097/SIH.0000000000000332
12. Jack SM, Boyle M, McKee C. **Effect of addition of an intimate partner violence intervention to a nurse home visitation program on maternal quality of life: a randomized clinical trial**. *JAMA* (2019.0) **321** 1576-85. DOI: 10.1001/jama.2019.3211
13. Hwang WJ, Kim JA. **Development and evaluation of a home-visit simulation scenario for elderly people with diabetes mellitus who live alone**. *J Commun Health Nurs* (2020.0) **37** 89-102. DOI: 10.1080/07370016.2020.1736399
14. 14.Choi YJ, Um YJ. The effects of a home-visit nursing simulation for older people with dementia on nursing students’ communication skills, self-efficacy, and critical thinking propensity: quantitative research. Nurse Educ Today. 2020;119(105564). 10.1016/j.nedt.2022.105564. https://doi-org.proxy.cau.ac.kr/.
15. 15.Ruiz-Fernández MD, Alcaraz-Córdoba A, López-Rodríguez MM, Fernández-Sola C, Granero-Molina J, Hernández-Padilla JM. The effect of home visit simulation on emotional intelligence, self-efficacy, empowerment, and stress in nursing students. A single group pre-post intervention study. Nurse Educ Today. 2022;117(105487). 10.1016/j.nedt.2022.105487.
16. 16.Chang S, Yang W, Deguchi H. Care providers, access to care, and the long-term care nursing insurance in China: an agent-based simulation. Soc Sci Med. 2020;244(112667). 10.1016/j.socscimed.2019.112667. https://doi-org.proxy.cau.ac.kr/.
17. Lee M, Ko M, Sohn H, Kim J, Kang S, Oh S. *Qualitative research* (2018.0) 97-9
18. 18.Colaizzi P. Psychological research as phenomenologist views it. In: Valle RS, King M, editors. Existential phenomenological alternatives for psychology. Oxford University Press, Inc.; 1978. pp. 48–71.
19. Lincoln YS, Guba EG, Pilotta JJ. **Naturalistic inquiry**. *Int J Intercultural Relations* (1985.0) **9** 438-9. DOI: 10.1016/0147-1767(85)90062-8
20. Mauriz E, Caloca-Amber S, Córdoba-Murga L, Vázquez-Casares AM. **Effect of psychophysiological stress and socio-emotional competencies on the clinical performance of nursing students during a simulation practice**. *Int J Environ Res Public Health* (2021.0) **18** 5448. DOI: 10.3390/ijerph18105448
21. Judd BK, Currie J, Dodds KL, Fethney J, Gordon CJ. **Registered nurses psychophysiological stress and confidence during high-fidelity emergency simulation: Effects on performance**. *Nurse Educ Today* (2019.0) **78** 44-9. DOI: 10.1016/j.nedt.2019.04.005
22. 22.Kimzey M, Mastel-Smith B, Seale A. Effects of Dementia-Specific Education for Nursing Students. Nurse Educator 44(6):p 338–341, 11/12 2019. | DOI: 10.1097/NNE.0000000000000623
23. Akaike M, Fukutomi M, Nagamune M, Fujimoto A, Tsuji A, Ishida K. **Simulation based medical education in clinical skills laboratory**. *J Investig Med* (2012.0) **59** 28-35. DOI: 10.2152/jmi.59.28
24. Dahlin KB, Chuang YT, Roulet TJ. **Opportunity, motivation, and ability to learn from failures and errors: review, synthesis, and Ways to Move Forward [Review]**. *Acad Manag Ann* (2018.0) **12** 252-77. DOI: 10.5465/annals.2016.0049
25. Kim MS, Jeong HC. **A phenomenological research on simulation practice and video watching educational experience using dyspnea case in nursing students**. *J Digit Convergence* (2019.0) **17** 287-97. DOI: 10.14400/JDC.2019.17.6.287
26. Goolsarran N, Hamo CE, Lane S, Frawley S, Lu WH. **Effectiveness of an interprofessional patient safety team-based learning simulation experience on healthcare professional trainees**. *BMC Med Educ* (2018.0) **18** 192. DOI: 10.1186/s12909-018-1301-4
27. Ghufron MA, Ermawati S. **The strengths and weaknesses of cooperative learning and problem-based learning in EFL writing class: Teachers’ and students’ perspectives**. *Int J Instruction* (2018.0) **11** 657-72. DOI: 10.12973/iji.2018.11441a
28. Bruce R, Levett-Jones T, Courtney-Pratt H. **Transfer of learning from university-based simulation experiences to nursing students’ future clinical practice: an exploratory study**. *Clin Simul Nurs* (2019.0) **35** 17-24. DOI: 10.1016/j.ecns.2019.06.003
29. Coyne E, Needham J. **Undergraduate nursing students’ placement in speciality clinical areas: understanding the concerns of the student and registered nurse**. *Contemp Nurse* (2012.0) **42** 97-104. DOI: 10.5172/conu.2012.42.1.97
30. Kim MR, Kim SY. **A study on the experience of nursing student’ intensive care unit simulation class**. *J Korean Soc Wellness* (2019.0) **14** 121-34. DOI: 10.21097/ksw.2019.08.14.3.121
|
---
title: Aurantiamide suppresses the activation of NLRP3 inflammasome to improve the
cognitive function and central inflammation in mice with Alzheimer's disease
authors:
- Heping Shen
- Hongyan Pei
- Liping Zhai
- Qiaobing Guan
- Genghuan Wang
journal: CNS Neuroscience & Therapeutics
year: 2023
pmcid: PMC10018077
doi: 10.1111/cns.14082
license: CC BY 4.0
---
# Aurantiamide suppresses the activation of NLRP3 inflammasome to improve the cognitive function and central inflammation in mice with Alzheimer's disease
## Abstract
### Aim
This study was aimed at exploring the mechanism by which aurantiamide (Aur) targeted NLRP3 to suppress microglial cell polarization.
### Methods
The 7‐month‐old APP/PS1 mice and C57BL/6 mice were applied to be the study objects, and Aur was administered intragastrically to APP/PS1 mice at 10 mg/kg and 20 mg/kg. The changes in the neurocognitive function of mice were measured by Morris Water Maze (MWM) test. In the in vitro experiments, the mouse BV2 cells were employed as the study objects, which were subject to treatment with 10 μM and 20 μM Aur and induced with LPS and IFN‐γ in order to activate BV2 cells and induce their M1 polarization.
### Results
Aur was found to suppress the M1 polarization of mouse microglia, reduce central neuroinflammation, and improve the cognitive function in mice. Meanwhile, Aur suppressed the activation and the expression of NLRP3 inflammasome. The results of experiments in vitro demonstrated that Aur inhibited the activation and M1 polarization of BV2 cells.
### Conclusion
Aur targets NLRP3 and suppresses the activation of NLRP3 inflammasome.
## BACKGROUND
Aurantiamide (Aur) refers to a kind of small‐molecular compound that is extracted from pepper, 1 which is an important active substance in addition to piperine and pipercide. 2, 3 Currently, there are few reports on the pharmacological activity of Aur, while piperine and pipercide with similar structures have been extensively investigated. Piperine possesses favorable anti‐inflammatory and antitumor activities. 4 *It is* shown that piperine exhibits excellent antagonistic effects against experimental electrical stimulation in mice. 5 Piperine also resists against seizure and audiogenic seizure induced by pentetrazol, picrotoxin, strychnine, intraventricular injection of tubocurarine and glutamic acid to varying degrees. 6 In addition, it is also effective on certain types of seizure. These neuropharmacological effects are all related to anti‐inflammation. At present, anti‐inflammatory research indicates that Aur can inhibit the virus‐induced inflammatory response by inhibiting the NF‐B signal, 7 and exerts a good antagonistic effect on influenza A virus infection. Some scholars have also indicated that Aur can generate anti‐inflammatory effects through metabolic signaling and insulin‐like signaling based on network pharmacology research. 8 However, there is no report on the role and target of Aur in the study of neuroinflammation.
AD studies show that piperine can improve cognitive impairment in AD mice, and its impact is associated with the improvement of synaptic function and the inhibition of inflammation, 9 while the exact target remains unknown. Considering the similar structure of Aur to piperine, the current work attempted to show the anti‐inflammatory effect and targets of Aur with Alzheimer's disease (AD) as an example.
In the study of AD, it has been discovered that neuroinflammation is an important factor which can stimulate the occurrence and development of AD. 10, 11 The activation and polarization of microglial cells (MG) can be considered to be the main factors leading to neuroinflammation. Based on the existing studies, NLRP3 inflammasome can mediate the activation of MG as well as the release of inflammatory factors. 12 After the activation, the activated NLRP3 is capable of recruiting the ASC protein to activate the inflammasome, which can further form the protein complex with Caspase‐1. The activation of NLRP3 proves to be a vital hallmark of M1 polarization 13 and NLRP3 is thus one of the important targets suppressing neuroinflammation. In addition, this study also focused on exploring the association between Aur and NLRP3.
## Cell culture
After resuscitation, the mouse microglial BV2 cell line (Procell Biotechnology, Co., Ltd) was cultivated in the RPMI‐1640 + $10\%$ FBS complete medium and also subject to incubation based on the conditions of 37°C, $5\%$ CO2 and saturated humidity. In order to perform the experiments, the cells at logarithmic phase were harvested. BV2 cells were classified as DMSO, LPS + IFN‐γ (L/I), and Aur groups. LPS + IFN‐γ was regarded as the M1 cell induction group, 14 where cells were triggered with 200 ng/mL PMA (Sigma) for 6 h. Subsequently, for further induction, 1 μg/mL LPS (Sigma) and 20 ng/mL IFN‐γ (Sigma) were supplemented. In Aur group, BV2 cells were pretreated with 10 μM and 20 μM Aur for 6 h. Thereafter, their M1 polarization was induced based on the same method used in L/I group.
The NLRP3 knockdown cell line (BV2‐nlrp3 −/−, prepared in our laboratory) was adopted in the mechanism research. To be specific, cells were classified as DMSO, L/I, nlrp3 −/− ‐L/I and nlrp3 −/− ‐L/I + Aur groups. In DMSO group, BV2 cells were used as control, while in L/I group, the M1 polarization of BV2 cells was induced using the above‐mentioned method. As shown in nlrp3 −/− ‐L/I group, BV2‐nlrp3 −/− cells were treated as the objects for M1 polarization induction, while BV2‐nlrp3 −/− cells in nlrp3 −/− ‐L/I + Aur group were subject to pretreatment with 20 μM Aur for 6 h. Thereafter, the M1 polarization was induced.
## Flow cytometry (FCM) analysis
To measure the ratio of F$\frac{4}{80}$+CD11b+M1 cells, 15 BV2 cells were exposed to an inoculation into the 6‐well plate and M1 polarization was induced after adaptive culture for 12 h. Following LPS/IFN‐γ treatment for 2 days, BV2 cells were gathered, rinsed with pre‐chilled PBS twice, and fixed with methanol. Subsequently, cells were subject to incubation with 10 μl FITC‐F$\frac{4}{80}$ monoclonal antibody and PE‐CD11b monoclonal antibody (BD) for 20 min in dark. By rinsing twice with PBS, the cells were resuspended with 50 μl of liquid. After the machine detection, the obtained findings could be denoted to be %.
## Immunofluorescence staining
In addition, the expressions of CD11b and NLRP3 proteins in BV2 cells were identified. BV2 cells were inoculated on the glass slide to prepare the cell climbing films, treated with LPS and IFN‐γ to cause polarization for 24 h, and then stained. Afterwards, cells were rinsed with pre‐chilled PBS thrice, fixed with pre‐chilled methanol for 0.5 h, as well as permeabilized with $0.2\%$ Triton X‐100 for 5 min. Later, the CD11b and NLRP3 monoclonal antibodies (Abcam, Massachusetts, USA) were diluted with TBST at 1:300 and added to incubate cells at 4°C on the shaking table. By rinsing twice with PBS, cells were subject to incubation with fluorescence secondary antibody, and mounted with $95\%$ glycerin. In addition, we adopted the fluorescence microscope for observation.
## ELISA
As shown in BV2 cell experiments, the expressions of M1 cell marker cytokines IL‐1β, IL‐6 and TNF‐α were identified. In FCM, cell medium was gathered after the cell extraction, followed by 30 min centrifugation at 3000 rpm and preservation at −80°C. Then, following the specific instructions, the medium was detected for cytokines with the use of the ELISA kit (Jiancheng Institute of Biology). In addition, the standard curve was used to calculate the expression levels. The results were denoted to be pg/ml.
For animal experiments, the mouse brain tissues were separated, rinsed with PBS twice and grinded until no granule was observed after the removal of tissues including blood vessels and thin membranes. Afterwards, 1 ml NP‐40 lysate (Beyotime Biotechnology Co., Ltd) was supplemented to lyse cells on ice for half an hour. Meanwhile, using the same method in BV2 cell experiments, the supernatant protein solution was harvested for the identification.
## ROS detection
The DCFH‐DA probe (Green) and DHE probe (Red) were used to detect reactive oxygen species (ROS). BV2 cells were inoculated into the 6‐well plates and treated with LPS/IFN‐γ for a day. After rinsing with pre‐chilled PBS twice, the DCFH‐DA probe was diluted with serum‐free medium at 1:1000, the DHE probe was diluted at 1:1500, and 1 ml diluted fluorescence probe solution was added into each well to deeply incubate the cells for 30 min. Subsequently, cells were washed twice. Using the fluorescence microscope, the cell staining level was observed. Meanwhile, the fluorescence spectrophotometer was used to detect the absorbance (OD) value.
## Western‐blot (WB) assay
As shown in cell experiments, LPS/IFN‐γ were added to induce BV2 cells, followed by rinsing with pre‐chilled PBS twice. After washing twice by PBS, thin membranes and blood vessels were removed from mouse brain tissues, and the brain tissues were grinded until no granule was observed. Later, cells and tissues were digested with 1 ml NP‐40 lysate on ice for 30 min with the purpose of extracting the total proteins. The protein contents were identified. Next, proteins were separated through electrophoresis and transferred onto the PVDF membrane. Then, the PVDF membrane was blocked with $5\%$ defatted milk powder for 2 h. The expression levels of NLRP3, ASC, CD11b and Caspase‐1 were determined. Afterwards, the membrane was subject to incubation with monoclonal antibodies (Abcam) diluted with TBST at a volumetric ratio of 1:300–1:500 at 4°C overnight. Then, the membrane was deeply incubated with HRP‐IgG diluted with TBST at the volumetric ratio of 1:2000. After the incubation, the protein blots were identified with chemiluminescence (ECL), and the OD value was explored by adopting the Image Pro‐Plus 6.0 software. With GAPDH as the internal control, the findings were denoted to be the OD ratio of target protein to internal control protein.
## Molecule‐protein docking and pull‐down assays
The NLRP3 receptor protein (PDB ID: 6NPY) was retrieved from the Protein Data Bank database. The box centers (center_x = 88.446, center_y = 95.078, and center_z = 92.124) and the box lattice parameters (size_x = 50, size_y = 60, and size_z = 52) suitable for NLRP3 receptor protein were determined. Afterwards, the active pocket sites possibly bound by the small‐molecular ligand, including NLRP3 receptor protein and Aur ligand small molecule, were subject to molecular docking (AutoDock Vina 1.1.2). Thereafter, the PyMOL was applied to prepare the 3D diagram to display the hydrogen bond interaction between the receptor protein and the ligand small molecule. Meanwhile, the Ligplus software was employed to plot the 2D diagram for displaying the hydrophobic effect between the receptor protein and the ligand small molecule.
Thereafter, 15 μg recombinant NLRP3 protein was bound to the Biotin‐conjugated Aur (Biotin‐Aur). Cells were incubated with the recombinant G protein magnetic bead and NLRP3 antibody. After rinsing with Tris buffer, NLRP3 expression was identified by the aforementioned WB assay.
## Mouse grouping experiments
The wild‐type (WT) mice (Normal) and APP/PS1 double‐transgenic AD mice were raided in the Jiaxing University Animal Experimental Center. The mouse experiments were approved by the Ethics Committee of Jiaxing University and performed following the Guides for the Care and Use of Experimental Animals. The 7‐month‐old AD mice presented with senile plaques and symptoms of AD neurological disorder, consistent with the standards of AD research. Mice were classified into WT, AD and Aur groups, with 10 mice (5 females and 5 males) in each group. Mice in Gla group were given intragastric administration of Aur at 10 mg/kg (low‐concentration) and 20 mg/kg (high concentration) once a day for 30 consecutive days.
## Morris Water Maze (MWM) test
The MWM and video system were obtained from Feidi Biotechnology Co., Ltd. One day before the experiment, all animals underwent adaptive training. Briefly, they entered through the MWM entrance and swam freely for the 60 s. Mice stood on the platform for the 20 s if they could find it, and later they were returned to their cages. For a 5‐day navigation test, the platform was positioned in the fourth quadrant. Afterwards, we measured the time of mice moving to the entrance and recorded the time spent finding and climbing on the platform. For mice unable to find the platform in the 60 s period, they were trained and stood on the platform for a 20 s period. Escape latency (EL) was regarded as the duration between mice entering the water and searching for the platform. The platform was removed for the spatial probe test. After mice entered the entrance, we noted the frequency crossing the fourth quadrant within a 60 s period as well as retention time on that platform. The mice were tested every 5 days for 30 days consecutively.
## Hematoxylin and eosin (H&E) staining
The mouse brain tissues (cerebral cortex) were deparaffinized with xylene, dehydrated with gradient concentrations of methanol ($100\%$, $95\%$ and $80\%$ in succession), rinsed with tap water for 2 min, and stained with hematoxylin for 3 min. After washing with tap water for 2 min, sections were treated with $1\%$ hydrochloric acid alcohol for 2 s, and rinsed by tap water for 2 min. Next, the sections were treated with $1\%$ ammonia water for 20 s and stained with $0.5\%$ eosin alcohol. After gradient alcohol dehydration, the sections were subject to xylene permeabilization and neutral resin mounting. Finally, a microscope was used to observe section pathological changes. Fluorescence staining of tissues.
## Fluorescence staining of tissues
CD11b levels within mouse cerebral cortex tissues were examined. Briefly, brain tissues were dehydrated using sucrose solutions ($15\%$ and $30\%$), embedded in OCT, cut into 8‐μm sections using a freezing microtome, and then preserved at −20°C. Sections were washed with PBS, then $5\%$ serum was mixed to block them for 30 min, followed by overnight incubation with CD11b monoclonal antibody under 4°C. Sections were rinsed by PBS three times, with 1‐h incubation using fluorescence antibody in dark. Then, sections were washed with PBS thrice, and mounted by using an anti‐fluorescence quenching agent. Meanwhile, a microscope was used for observation.
## Statistical analysis
The measurement data were represented by (x¯±s). Data were analyzed and processed with SPSS17.0. All data conform to a normal distribution. After homogeneity test of variance, two independent sample t‐test was employed for comparison between two groups, whereas one‐way ANOVA was conducted for comparison among three groups. Two‐way ANOVA was performed for comparison of behavioral test at multiple time points, and the subsequent pairwise comparison between groups was completed by LSD. All the above tests were two‐sided, and the difference of $p \leq 0.05$ stood for statistical significance.
## Aur inhibited the M1 polarization of BV2 cells
LPS + IFN‐γ was found to trigger the M1 polarization of BV2 cells, and the proportion of F$\frac{4}{80}$+CD11b+ cells in L/I group was increased, which was notably higher when compared with that of DMSO group ($p \leq 0.05$). Aur suppressed the M1 polarization of BV2 cells, and the ratio of F$\frac{4}{80}$ + CD11b + cells lowered obviously in relative to L/I group ($p \leq 0.05$). The high‐dose (20 μM) Aur had a superior effect to 10 μM Aur (Figure 1A,B). Results of fluorescence staining suggested that, after LPS + IFN‐γ induced the M1 polarization of BV2 cells, the expression of CD11b and NLRP3 obviously elevated, and that the fluorescence intensity was higher when compared with that of DMSO group. Aur suppressed the expression of CD11b and NLRP3 and reduced the fluorescence intensity in a dose‐dependent manner (Figure 1C–F). According to ELISA results, the expressions of inflammatory factors IL‐6, TNF‐α and IL‐1β in DMSO were low, while those in L/I group significantly increased ($p \leq 0.05$), consistent with the feature of M1 cells. After the pretreatment of Aur, the expression of inflammatory factors was reduced in a dose‐dependent manner, and the difference was shown to be significant in relative to L/I group ($p \leq 0.05$) (Figure 2A–C). Based on the detection results of NLRP3 inflammasome‐related protein, the expressions of NLRP3, ASC and Caspase‐1 in DMSO were low, and NLRP3 was not significantly activated. In L/I group, NLRP3 was activated, its protein expression level was significantly elevated, and the difference was obvious in relative to DMSO group ($p \leq 0.05$). Aur was found to inhibit the activation of NLRP3 inflammasome and reduce the protein levels in a dose‐dependent manner. Additionally, the difference was of statistical significance in relative to L/I group ($p \leq 0.05$). The CD11b expression level was similar to NLRP3 inflammasome, and Aur inhibited CD11b expression (Figure 2D–H).
**FIGURE 1:** *Aur suppresses the M1 polarization of BV2 cells. (A, B) FCM (n = 3). Compared with DMSO, the proportion of F4/80+CD11b+ cells in L/I group increased, and Aur hindered the M1 polarization of BV2 cells. Relative to L/I group, the ratio of F4/80+CD11b+ cells obviously lowered, and the high‐dose Aur exhibited better impact than low‐dose Aur. *p < 0.05 in relative to DMSO group, #
p < 0.05 in relative to L/I group. B–F: Immunofluorescence staining (n = 3). After LPS + IFN‐γ induced the M1 polarization of BV2 cells, the expression of CD11b and NLRP3 notably increased, and the fluorescence intensity was found to be higher when compared with that in DMSO group. Aur inhibited the expression of CD11b and NLPR3 and reduced the fluorescence intensity in a dose‐dependent manner. *p < 0.05 in relative to DMSO group, #
p < 0.05 in relative to L/I group.* **FIGURE 2:** *Aur suppresses inflammatory factor expression and NLRP3 inflammasome activation in BV2 cells. (A–C) ELISA (n = 3). Detection of inflammatory factors indicated that, the expression levels of inflammatory factors IL‐6, TNF‐α and IL‐1β in DMSO were low, while those in L/I group presented significant upregulation compared with DMSO. The pretreatment of Aur reduced the expression of inflammatory factors in a dose‐dependent manner, and the difference was of statistical significance in comparison with L/I group. *p < 0.05 in relative to DMSO group, #
p < 0.05 in relative to L/I group. (D–H) WB assay (N = 3). The levels of NLRP3, ASC, Caspase‐1 and CD1b in DMSO were low, while NLRP3 was activated in L/I group, and the protein levels were significantly upregulated, with the significant difference compared with DMSO group. Aur inhibited the activation of NLRP3 inflammasome, reduced protein level in a dose‐dependent manner, and the difference was of statistical significance. *p < 0.05 in comparison with DMSO group, #
p < 0.05 in comparison with L/I group.*
## Aur inhibited ROS expression and bound to NLRP3
During the M1 polarization of BV2 cells, ROS was activated. The ROS expression in L/I group was notably upregulated. Both DCFH‐DA and DHE detection results revealed a significantly increased number of positive cells, whereas no positive cell was detected in DMSO. The pretreatment of Aur inhibited ROS expression and decreased the positive cell number (Figure 3A). The docking results of NLRP3 with Aur showed that there was hydrogen bonding between Aur and NLRP3. Meanwhile, Aur bound to SER in the form of hydrogen bonds, and bound to TYR, ALA, LYS, and PRO in the form of hydrophobic bonds. Results of pull‐down assay also indicated that Aur bound to NLRP3 (Figure 3B–E).
**FIGURE 3:** *Aur suppresses ROS and specifically binds to NLRP3. (A) DCFH‐DA and DHE (n = 3). ROS was not expressed in DMSO, the number of positive cells in L/I group was significantly elevated, ROS expression was upregulated, and the difference was of statistical significance compared with DMSO. The pretreatment of Aur inhibited ROS expression, and the number of positive cells in Aur evidently decreased. (C–E) Docking results of NLRP3 with Aur suggested that there was stable inter‐molecular force between Aur and NLRP3‐SER. Pull‐down assay suggested that Aur bound to NLRP3. Aur bound to SER in the form of hydrogen bonds, and bound to TYR, ALA, LYS, and PRO in the form of hydrophobic bonds.*
## NLRP3 knockdown suppressed the effects of Aur
NLRP3 was knocked down in BV2 cells. Inflammatory factor expression findings suggested that there was no significant difference between nlrp3 −/− ‐L/I and nlrp3 −/− ‐L/I + Aur groups, and that the inflammatory factor levels were significantly lower than those in L/I group ($p \leq 0.05$). This demonstrated that NLRP3 knockdown decreased the inflammatory factor levels whereas such effect was not related to Aur (Figure 4A–C). Fluorescence staining results indicated that NLRP3 knockdown inhibited CD11b expression, and there existed no significant difference between nlrp3 −/− ‐L/I and nlrp3 −/− ‐L/I + Aur groups (Figure 4D,E). Protein detection results also revealed no significant difference in ASC and CD11b expression between nlrp3 −/− ‐L/I and nlrp3 −/− ‐L/I + Aur groups, while their levels were significantly lower than those in L/I group ($p \leq 0.05$) (Figure 4F,G). According to the obtained results, after NLRP3 knockdown, Aur exerted no obvious influence on the M1 polarization of BV2 cells.
**FIGURE 4:** *NLRP3 knockdown suppresses the effects of Aur. (A–C) ELISA (n = 3). The levels of inflammatory factors in L/I group were significantly higher than those in DMSO group, while those in nlrp3
−/−
‐L/I and nlrp3
−/−
‐L/I + Aur groups were lower compared with those in L/I group, and the differences were not significantly different. *p < 0.05 compared with DMSO group, #
p < 0.05 in relative to L/I group. D, E: Immunofluorescence staining (n = 3). After NLRP3 knockdown, CD11b expression was suppressed, and no significant difference was detected between nlrp3
−/−
‐L/I and nlrp3
−/−
‐L/I + Aur groups. *p < 0.05 compared with DMSO group. F, G: WB assay (n = 3). The expression of ASC and CD11b between nlrp3
−/−
‐L/I and nlrp3
−/−
‐L/I + Aur groups was not obviously different but was significantly lower than that in L/I group. *p < 0.05 in relative to DMSO group, #
p < 0.05 in relative to L/I group.*
## Aur improved the cognitive function of AD mice
As presented in the MWM test, AD mice showed significant cognitive impairment in relative to Normal group, with fewer times across the platform, longer EL, and longer time to find the platform, conforming to the cognitive impairment characteristics of AD mice. Aur could improve the times across the platform, reduce the EL, and shorten the time to find the platform. In addition, it also indicated that Aur enhanced the cognitive function of AD mice, and high‐dose Aur exhibited a better impact than low‐dose Aur (Figure 5).
**FIGURE 5:** *Aur improves the cognitive function in AD mice. (A) Trajectory chart of MWM. (B–D) Cognitive function of mice (n=). In comparison with Normal mice, AD mice experienced significant cognitive impairment, less time across the platform, longer EL and longer time to find the platform. Aur could increase the times across platform, reduce the EL, and shorten the time to find the platform. *p < 0.05 in relative to Normal group, #
p < 0.05 in relative to AD group.*
## Aur suppressed the polarization of MG and activation of NLRP3 in AD mice
The results of H&E staining indicated no obvious cell injury and inflammatory response in mice of Normal group. In addition, normal brain tissue morphology was found. In AD mice, obvious tissue injury accompanied by inflammatory response was observed. The difference was of significance in relative to Normal group. Tissue inflammation was significantly alleviated in Aur group, and the degree of cell damage is lower than that in AD group, implying that Aur could inhibit AD‐related neural injury (Figure 6A). According to the results of immunofluorescence staining, CD11b was negatively expressed in Normal group, and no obvious polarization of MG was observed, whereas CD11b expression was notably upregulated in AD group, and the difference was significant compared with Normal group. Aur inhibited CD11b expression in a dose‐dependent way (Figure 6B). Based on the detection of inflammatory factors, the expressions of IL‐1β, IL‐6 and TNF‐α in AD were significantly elevated compared with those in Normal group, revealing the presence of obvious inflammatory response in AD. Aur reduced the levels of inflammatory factors in tissues, and significant differences were found compared with AD group. Notably, high‐dose Aur exhibited superior impact to that of low‐dose (Figure 6C–E).
**FIGURE 6:** *Aur inhibits the activation of MG and the level of inflammatory factors in AD mice. (A) H&E staining (n = 5). Normal group showed normal brain tissue morphology, while obvious tissue injury accompanied by inflammatory response was found in AD group, and there existed no obvious difference in comparison with Normal group. In Aur group, the tissue inflammation was obviously alleviated, and the cell injury degree was also mitigated in relative to AD group. (B) Fluorescence staining (n = 5). CD11b was not expressed in Normal group but presented significant upregulation in AD group. The difference was of significance in relative to Normal group. Aur inhibited CD11b expression in a dose‐dependent manner. The high‐dose Aur exerted superior effect to low‐dose one. (C–E) ELISA (n = 10). The levels of IL‐1β, IL‐6 and TNF‐α in AD group were notably higher than those in Normal group. Aur reduced the inflammatory factor levels in tissues, and the differences were significant in comparison with AD group. Meanwhile, high‐dose Aur exhibited better impact than low‐dose one. *p < 0.05 in relative to Normal group, #
p < 0.05 in relative to AD group.*
The detection of NLRP3 inflammasome indicated that the expression levels of NLRP3, ASC and Caspase‐1 in Normal group were low, and that NLRP3 was not activated. In AD group, the NLRP3 inflammasome and related protein expression levels significantly increased. Aur suppressed the activation of NLRP3, and the levels of NLRP3, ASC and Caspase‐1 were notably downregulated in relative to AD group ($p \leq 0.05$) (Figure 7).
**FIGURE 7:** *Aur inhibits the activation of NLRP3 inflammasome (n = 5). The levels of NLRP3, ASC and Caspase‐1 in Normal group were lower, and NLRP3 was not activated. By contrast, NLRP3 inflammasome and related protein expression levels significantly increased in AD. Aur suppressed the activation of NLRP3, and the expression levels of NLRP3, ASC and Caspase‐1 were notably lower than those in AD group. *p < 0.05 in relative to Normal group, #
p < 0.05 in relative to AD group.*
## DISCUSSION
MG are the immune cells in the central nervous system (CNS), which can regulate the immune homeostasis. 17 This study suggests that MG can phagocytize and eliminate the abnormal protein aggregates under physiological conditions, which can thus decrease β‐amyloid (Aβ) protein deposition and delay the decline of the cognitive function. 18, 19 However, when MG are abnormally activated, a large number of pro‐inflammatory factors are released, which can induce pro‐inflammatory cascade reaction and lead to neuronal injury. 20, 21 In the healthy brain, MG possesses morphological plasticity and pleiotropy features, and under different activation states, MG can rapidly change the morphology and function in response to the changes in intracerebral microenvironment. 22 Simulated by multiple factors, MG can transform into the activated M1 and M2 types, 23 with a larger cell body together with reduced and thickened processes. Different activation phenotypes exert distinct immune effects on the nervous system. 24 Upon stimulation by immune cells and molecules including LPS, IFN‐γ, β‐amyloid and α‐synuclein, 25, 26 MG are polarized into the classical activated M1 type, which secretes inflammatory mediators including TNF‐α, NO and ROS to damage the peripheral nerve cells. 27 In neurodegenerative diseases, the abnormal aggregation of Aβ, Tau and α‐synuclein proteins can induce the chronic activation of MG, causing M1/M2 MG cell dysfunction and immunologic inadequacy. Therefore, a large number of inflammatory mediators are released, accelerating Tau protein phosphorylation. The increase in ROS level is also the key factor which can induce the M1 polarization of MG, resulting in the release of pro‐inflammatory cytokines. 28 The effect of LPS + IFN‐γ is mainly mediated by the Toll‐like receptor (TLR). TLRs can activate inflammasomes, especially NLRP3. NLRP3 inflammasome is formed by NLRP3, ASC and Caspase‐1, exerting its effect by promoting the maturation of pro‐inflammatory factors including pro‐IL‐1β. 29 NLRP3 is also the key regulatory factor for MG polarization. The disturbance in NLRP3 expression or assembly can suppress the maturation and release of inflammatory factors and inhibit the MG polarization. Therefore, NLPR3 is the vital target for MG polarization and neuroinflammation.
Aur is a derivative of piperine and is also a small‐molecular compound extracted from piperaceae. However, there are currently few studies on the pharmacological effect of Aur whereas piperine, which exhibits a similar structure to Aur, has been extensively investigated. At present, it has been found that piperine exerts vital effects, including anti‐inflammation, antitumor and anti‐bacterium. This study concentrated on the anti‐inflammatory effect of Aur. CD11b is the cell surface antigen of M1 cells, and co‐expresses F$\frac{4}{80}$‐labeled M1 cells with macrophages. LPS + IFN‐γ treatment is the common method which can be applied to induce M1 polarization. In this study, LPS + IFN‐γ treatment successfully induced M1 polarization of BV2 cells, and upregulated the expression of inflammatory factors, consistent with the phenotypic features of M1 cells. During the polarization process, NLRP3 inflammasome was assembled and activated, and the expression of NLRP3, ASC and Caspase‐1 in the protein complex was also upregulated, indicating that NLRP3 played a vital role in BV2 polarization. After the pretreatment of Aur, the impact of LPS + IFN‐γ was antagonized, and the M1 polarization of BV2 cells was suppressed. In addition, Aur suppressed the maturation and release of inflammatory factors, and also decreased NLRP3 expression. The binding relation between Aur and NLRP3 was confirmed through small molecule‐protein docking and pull‐down assays. Therefore, Aur bound to the SER site, which blocked the assembly of ASC and Caspase‐1 with NLRP3. The blockade of inflammasome assembly further blocked pro‐IL‐β cleavage and polarization, and downregulated CD11b expression. BV2 polarization is mutually promoted by ROS promotes. ROS is excessively expressed during the polarization process, whereas ROS can promote NLRP3 formation and NF‐Κb activation. ROS is one of the important signals that mediate the activation of NLRP3. Two probes were used to detect ROS. The results demonstrated that Aur inhibited the generation of ROS and the activation of NLRP3, and thus ROS and NLRP3 were activated. Our results also suggested that Aur inhibited ROS expression and significantly decreased the positive cell number. To further determine the binding relation between Aur and NLRP3, this study knocked down NLRP3 expression in BV2 cells. Therefore, NLRP3 knockdown antagonized the effect of LPS + IFN‐γ and suppressed the M1 polarization of BV2 cells. After NLRP3 knockdown, Aur lost its activity, which did not further suppress polarization or inhibit the downregulation of inflammatory factors, and it exerted no distinct effect on ASC or NLRP3 expression. As a result, we determined that NLRP3 was the target for Aur.
After the polarization of MG, a large number of inflammatory factors will be released, generating AD progression and aggravation of cognitive impairment, 30 which are also the important sources of neuroinflammation. In this study, the 7‐month‐old mice with cognitive impairment were detected, which had apparent memory disorder, and the difference was of significance in relative to Normal group. Aur administration could improve the cognitive function, increase the times across the platform, and shorten the time to find the platform, indicating the improved cognitive function of mice. The obtained effect was of significant difference in comparison with AD. Similarly, in AD group, the brain tissue neuroinflammation in mice was improved, the expression of inflammatory factors was downregulated, and cell injury was suppressed. Moreover, these effects were related to the suppression of NLRP3 inflammasome. Experimental results in vitro were consistent with those obtained in vivo. The early efforts to delineate the versatility of microglia/macrophages in stroke brains categorize them into two conceptual phenotypes with pro‐inflammatory (M1) or anti‐inflammatory (M2) functional identities. Although a strict demarcation of M1/M2 polarities is currently known to be oversimplified, the concept of phenotypic diversity is extensively accepted and explored. 31 Therefore, even though our study finds that Aur can inhibit the polarization of M1 microglia, further exploration is needed for various differential and functional studies. It is also reported that the function of microglia is related to gender differences, 32, 33 Although our study used female and male mice, we did not further evaluate the role of Aur in mice of different sexes. This is also a limitation of our research. In order to further study the role of Aur in microglia of different genders, we need to expand the sample size and conduct research on female and male mice separately to further evaluate the role of Aur。.
## CONCLUSION
Aur can target NLRP3 and suppress its activation, thus regulating the M1 polarization of MG and neuroinflammatory response. Furthermore, Aur can also suppress the cognitive disorder in AD mice, which is a promising small molecule that deserves further investigation.
## AUTHOR CONTRIBUTIONS
Heping Shen and Hongyan Pei are primarily responsible for the operation of the experiment, the acquisition of relevant data and the statistical distraction of the data; Liping Zhai mainly accounts for literature review, project coordination, partial data analysis and article writing; Qiaobing Guan and Genghuan Wang are mainly responsible for project development, financial support, overall project proposal and experimental design.
## FUNDING INFORMATION
This study was funded by the ZheJiang provincial Natural Science Foundation [LGF20C090003] as well as the Science and technology planning project of JiaXing [2020AY30018].
## CONFLICT OF INTEREST
No competing interests.
## CONSENT FOR PUBLICATION
All authors approved the publication of the article.
## DATA AVAILABILITY STATEMENT
The data supporting the findings of the current work can be acquired from the corresponding author upon reasonable request.
## References
1. Banerji A, Ray R. **Aurantiamides: a new class of modified dipeptides from Piper aurantiacum**. *Phytochemistry* (1981) **20** 2217-2220
2. Talapatra SK, Mallik AK, Talapatra B. **Pongaglabol, a new hydroxyfuranoflavone, and aurantiamide acetate, a dipeptide from the flowers of Pongamia glabra**. *Phytochemistry* (1980) **19** 1199-1202
3. Suhas R, Gowda DC. **Structure‐based rationale design and synthesis of aurantiamide acetate analogues—towards a new class of potent analgesic and anti‐inflammatory agents**. *Chem Biol Drug des* (2012) **79** 850-862. PMID: 22251852
4. Mahran R, Pan S, Sun D, Brenner DE. **Abstract 5257: cellular pharmacology of curcumin cellular pharmacology of curcumin±piperine**. *Cancer Res* (2017) **77** 5257
5. Kumar S. **Piperine inhibits TNF‐alpha induced adhesion of neutrophils to endothelial monolayer through suppression of NF‐kappaB and IkappaB kinase activation**. *Eur J Pharmacol* (2007) **575** 177-186. PMID: 17764673
6. Pal A, Nayak S, Sahu PK, Swain T. **Piperine protects epilepsy associated depression: a study on role of monoamines**. *Eur Rev Med Pharmacol Sci* (2011) **15** 1288-1295. PMID: 22195361
7. Zhou B, Yang Z, Feng Q. **Aurantiamide acetate from baphicacanthus cusia root exhibits anti‐inflammatory and anti‐viral effects via inhibition of the NF‐κB signaling pathway in influenza a virus‐infected cells**. *J Ethnopharmacol* (2017) **199** 60-67. PMID: 28119097
8. Zhang Q, Li R, Peng W. **Identification of the active constituents and significant pathways of Guizhi‐Shaoyao‐Zhimu decoction for the treatment of diabetes mellitus based on molecular docking and network pharmacology**. *Comb Chem High Throughput Screen* (2019) **22** 584-598. PMID: 31642770
9. Nazifi M, Oryan S, Esfahani DE, Ashrafpoor M. **The functional effects of piperine and piperine plus donepezil on hippocampal synaptic plasticity impairment in rat model of Alzheimer's disease**. *Life Sci* (2021) **15**
10. Zakaria A, Hamdi N, Abdel‐Kader RM. **Methylene blue improves brain mitochondrial ABAD functions and decreases Aβ in a neuroinflammatory Alzheimer's disease mouse model**. *Mol Neurobiol* (2016) **53** 1220-1228. PMID: 25601181
11. Sachdeva AK, Chopra K. **Lycopene abrogates Aβ(1–42)‐mediated neuroinflammatory cascade in an experimental model of Alzheimer's disease**. *J Nutr Biochem* (2015) **26** 736-744. PMID: 25869595
12. Duewell P, Kono H, Rayner KJ. **NLRP3 inflammasomes are required for atherogenesis and activated by cholesterol crystals**. *Nature* (2010) **464** 1357-1361. PMID: 20428172
13. Tschopp J, Schroder K. **NLRP3 inflammasome activation: the convergence of multiple signalling pathways on ROS production?**. *Nat Rev Immunol* (2010) **10** 210-215. PMID: 20168318
14. Ohteki T, Fukao T, Suzue K. **Interleukin 12–dependent interferon γ production by CD8α+lymphoid dendritic cells**. *J Exp Med* (1999) **189** 1981-1986. PMID: 10377194
15. Schiechl G, Bauer B, Fuss I. **Tumor development in murine ulcerative colitis depends on MyD88 signaling of colonic F4/80+CD11bhighGr1low macrophages**. *J Clin Invest* (2011) **121** 1692-1708. PMID: 21519141
16. Mcnamara RK, Skelton RW. **The neuropharmacological and neurochemical basis of place learning in the Morris water maze**. *Brain Res Rev* (1993) **18** 33-49. PMID: 8467349
17. Streit WJ. **Microglia and neuroprotection: implications for Alzheimer's disease**. *Brain Res Brain Res Rev* (2005) **48** 234-239. PMID: 15850662
18. Parvathenani LK, Tertyshnikova S, Greco CR, Roberts SB, Robertson B, Posmantur R. **P2X7 mediates superoxide production in primary microglia and is up‐regulated in a transgenic mouse model of Alzheimer's disease**. *J Biol Chem* (2003) **278** 13309-13317. PMID: 12551918
19. Khoury JE, Hickman SE, Thomas CA, Loike JD. **Microglia, scavenger receptors, and the pathogenesis of Alzheimer's disease**. *Neurobiol Aging* (1998) **19** S81-S84. PMID: 9562474
20. Block ML, Hong JS. **Microglia and inflammation‐mediated neurodegeneration: multiple triggers with a common mechanism**. *Prog Neurobiol* (2005) **76** 77-98. PMID: 16081203
21. Graeber MB, Li W, Rodriguez ML. **Role of microglia in CNS inflammation**. *FEBS Lett* (2011) **585** 3798-3805. PMID: 21889505
22. Combs CK. **Inflammation and microglia actions in Alzheimer's disease**. *J Neuroimmune Pharmacol* (2009) **4** 380-388. PMID: 19669893
23. Herder V, Iskandar CD, Kegler K. **Dynamic changes of microglia/macrophage M1 and M2 polarization in Theiler's murine encephalomyelitis**. *Brain Pathol* (2015) **25** 712-723. PMID: 25495532
24. Zhang F, Zhong R, Song L. **Acute hypoxia induced an imbalanced M1/M2 activation of microglia through NF‐κB signaling in Alzheimer's disease mice and wild‐type littermates**. *Front Aging Neurosci* (2017) **9** 282-294. PMID: 28890695
25. Harms AS, Cao S, Rowse AL. **MHCII is required for α‐synuclein‐induced activation of microglia, CD4 T cell proliferation, and dopaminergic neurodegeneration**. *J Neurosci* (2013) **33** 9592-9600. PMID: 23739956
26. Ye M, Chung HS, Lee C. **Bee venom phospholipase A2 ameliorates motor dysfunction and modulates microglia activation in Parkinson's disease alpha‐synuclein transgenic mice**. *Exp Mol Med* (2016) **48**. PMID: 27388550
27. Meireles M, Marques C, Norberto S. **Anthocyanin effects on microglia M1/M2 phenotype: consequence on neuronal fractalkine expression**. *Behav Brain Res* (2016) **305** 223-228. PMID: 26965567
28. Stadio AD, Angelini C. **Microglia polarization by mitochondrial metabolism modulation: a therapeutic opportunity in neurodegenerative diseases**. *Mitochondrion* (2018) **43** 334-336
29. Toma C, Higa N, Koizumi Y. **Pathogenic vibrio activate NLRP3 inflammasome via cytotoxins and TLR/nucleotide‐binding oligomerization domain‐mediated NF‐κB signaling**. *J Immunol* (2012) **184** 5287-5297
30. Felipo V, Urios A, Montesinos E. **Contribution of hyperammonemia and inflammatory factors to cognitive impairment in minimal hepatic encephalopathy**. *Metab Brain Dis* (2012) **27** 51-58. PMID: 22072427
31. Hu X. **Microglia/macrophage polarization: fantasy or evidence of functional diversity?**. *J Cereb Blood Flow Metab* (2020) **40** S134-S136. PMID: 33023387
32. Chandra PK, Cikic S, Baddoo MC. **Transcriptome analysis reveals sexual disparities in gene expression in rat brain microvessels**. *J Cereb Blood Flow Metab* (2021) **41** 2311-2328. PMID: 33715494
33. Han J, Fan Y, Zhou K, Blomgren K, Harris RA. **Uncovering sex differences of rodent microglia**. *J Neuroinflammation* (2021) **18** 74-81. PMID: 33731174
|
---
title: Ferulic acid alleviates sciatica by inhibiting neuroinflammation and promoting
nerve repair via the TLR4/NF‐κB pathway
authors:
- Di Zhang
- Bei Jing
- Zhen‐ni Chen
- Xin Li
- Hui‐mei Shi
- Ya‐chun Zheng
- Shi‐quan Chang
- Li Gao
- Guo‐ping Zhao
journal: CNS Neuroscience & Therapeutics
year: 2023
pmcid: PMC10018085
doi: 10.1111/cns.14060
license: CC BY 4.0
---
# Ferulic acid alleviates sciatica by inhibiting neuroinflammation and promoting nerve repair via the TLR4/NF‐κB pathway
## Abstract
Ferulic acid promotes sciatic nerve repair by inhibiting Schwann cell apoptosis and promoted the transformation of M1 GMI‐R1 microglia to M2 microglia to relieve inflammatory infiltration though the TLR4/NF‐κB pathway.
### Introduction
Sciatica causes intense pain. No satisfactory therapeutic drugs exist to treat sciatica. This study aimed to probe the potential mechanism of ferulic acid in sciatica treatment.
### Methods
Thirty‐two SD rats were randomly divided into 4 groups: sham operation, chronic constriction injury (CCI), mecobalamin, and ferulic acid. We conducted RNA sequencing, behavioral tests, ELISA, PCR, western blotting, and immunofluorescence analysis. TAK‐242 and JSH23 were administered to RSC96 and GMI‐R1 cells to explore whether ferulic acid can inhibit apoptosis and alleviate inflammation.
### Results
RNA sequencing showed that TLR4/NF‐κB pathway is involved in the mechanism of sciatica. CCI induced cold and mechanical hyperalgesia; destroyed the sciatic nerve structure; increased IL‐1β, IL‐6, TNF‐α, IL‐8, and TGF‐β protein levels and IL‐1β, IL‐6, TNF‐α, TGF‐β, TLR4, and IBA‐1 mRNA levels; and decreased IL‐10 and INF‐γ protein levels and IL‐4 mRNA levels. Immunohistochemistry showed that IBA‐1, CD32, IL‐1β, iNOS, nNOS, COX2, and TLR4 expression was increased while S100β and Arg‐1 decreased. CCI increased TLR4, IBA‐1, IL‐1β, iNOS, Myd88, p‐NF‐κB, and p‐p38MAPK protein levels. Treatment with mecobalamin and ferulic acid reversed these trends. Lipopolysaccharide (LPS) induced RSC96 cell apoptosis by reducing Bcl‐2 and Bcl‐xl protein and mRNA levels and increasing Bax and Bad mRNA and IL‐1β, TLR4, Myd88, p‐NF‐κB, and p‐p38MAPK protein levels, while ferulic acid inhibited cell apoptosis by decreasing IL‐1β, TLR4, Myd88, p‐NF‐κB, and p‐p38MAPK levels and increasing Bcl‐2 and Bcl‐xl levels. In GMI‐R1 cells, Ferulic acid attenuated LPS‐induced M1 polarization by decreasing the M1 polarization markers IL‐1β, IL‐6, iNOS, and CD32 and increasing the M2 polarization markers CD206, IL‐4, IL‐10 and Arg‐1. After LPS treatment, IL‐1β, iNOS, TLR4, Myd88, p‐p38MAPK, and p‐NF‐κB levels were obviously increased, and Arg‐1 expression was reduced, while ferulic acid reversed these changes.
### Conclusion
Ferulic acid can promote injured sciatic nerve repair by reducing neuronal cell apoptosis and inflammatory infiltration though the TLR4/NF‐κB pathway.
## INTRODUCTION
Sciatica, a common type of neuropathic pain attributed to impingement of the sciatic nerves or injury to the sciatic nerves, is experienced by up to $10\%$ of patients with chronic lower back pain, with a reported lifetime incidence ranging from $10\%$ to $40\%$ 1 or even up to $70\%$. 2 According to the National Institute for Health & Clinical Excellence (NICE) guidelines, noninvasive treatments, including specific exercises, psychological therapy, and nonsteroidal anti‐inflammatory drugs, can be used to treat sciatica. 3 Gabapentinoids, other antiepileptics, oral corticosteroids or benzodiazepines are not used to manage sciatica because there is no overall evidence of their benefit. 3 Although nonsteroidal anti‐inflammatory drugs can help relieve sciatica, it is necessary to take into account their potential gastrointestinal, liver and cardio‐renal toxicity. 3 Moreover, opioids have a risk of addiction in patients. 4 Thus, it is urgent to seek new therapeutic drugs for the efficacious treatment of sciatica.
Nerve injury and neuroinflammation play vital roles in sciatica. Following injury to the nerve, axonal breakdown is initiated, and the products of degenerated neural tissue stimulate microglia and resident macrophages to secrete chemokines and cytokines, which promote neuroinflammation and axonal breakdown. Activation of microglia leads to the progression of neuropathic pain by interfering with neuronal function. 5 Inhibiting microglial activation reduces hyperalgesia after nerve damage. 6 Similarly, Schwann cells undergo dramatic reprogramming from highly quiescent, mature, differentiated myelinating cells to proliferative, prorepair cells after nerve injury and exhibit Wallerian degeneration. In addition, the proliferation and migration of Schwann cells and the inhibition of cell aging and apoptosis can restore the structure of injured peripheral nerves.
TLR4 can induce neuroinflammation and neuralgia. TLR4 is expressed on the cell surface as well as in endosomes, mainly in immune and glial cells. 7 Myeloid differentiation primary response 88 (MyD88) is the most common adaptor protein that interacts with the intracellular domain of TLR4, which can activate transcription factors such as nuclear factor κ light‐chain enhancer of activated B cells (NF‐κB) and mitogen‐activated protein kinase (MAPK). 8 Following injury, TLR4 activation on microglia and macrophages contributes to their shift towards an inflammatory phenotype and thus their release of inflammatory factors, including IL‐1β, TNF‐α, and IL‐6. 9 Intrathecal administration of TLR4 antagonists and siRNA‐mediated suppression of TLR4 signaling prevents activation of the NF‐κB pathway and production of TNF and IL‐1β, which attenuates mechanical allodynia and thermal hyperalgesia in a chronic constriction injury (CCI)‐induced pain model. 10, 11 Ferulic acid exhibits a potential advantage in the treatment of sciatica. Ferulic acid can decrease the levels of oxidative stress, inflammation and apoptosis markers in the sciatic nerves of patients with diabetes. 12 Ferulic acid exerts a neuroprotective effect against radiation‐induced nerve damage by targeting the NLRP3 inflammasome to enhance learning and memory ability and ameliorate pathological changes in the hippocampal tissues of irradiated mice. 13 *In this* study, ferulic acid was found to relieve pain in CCI rats, and we aimed to identify the related mechanisms. The therapeutic effect of ferulic acid on CCI of the sciatic nerve was assessed via behavioral tests, pathological examination, and immunohistochemistry. Next, we investigated the underlying mechanism at the cellular level to provide additional experimental evidence supporting the application of ferulic acid for the treatment of sciatica.
## Subjects
A total of 32 male Sprague–Dawley rats weighing approximately 150–180 g were acquired from the Experimental Center of Beijing Huafukang Co., Ltd. All animal experiments were conducted in accordance with the guidelines established by the National Academy of Sciences of the National Institutes of Health (NIH) and in accordance with the guidelines of the Animal Ethical Committee of Jinan University. Ethics approval (No. IACUC‐20201223‐07) was obtained on December 23, 2020.
## Reagents
Ferulic acid (F103701; $99\%$ purity) was acquired from Shanghai Aladdin Biotechnology Co., Ltd. Mecobalamin (lot number: 1703098) was purchased from Eisai Pharmaceutical Co., Ltd. PageRμLer Prestained Protein Ladder and Marker (P12083) was acquired from Shanghai Bioscience Technology Co., Ltd. RIPA buffer (WB‐0071) was purchased from Beijing Dingguo Biological Co., Ltd. RNAiso Plus [9108] was acquired from Takara Biomedical Technology Co., Ltd. SYBR Green Premix qPCR, an RT–PCR Kit, and RNase‐free water (AG11701, AG11602, AG11012) were obtained from Accurate Biotechnology Co., Ltd. IL‐1β (MM‐0047R1), IL‐6 (MM‐0190R1), TNF‐α (MM‐0180R1), IL‐8 (MM‐0175R1), IL‐10 (MM‐195R1), TGF‐β (MM‐20594R1), and IFN‐γ (MM‐0198R1) antibodies were purchased from Jiangsu Meimian Industrial Co., Ltd. Lipopolysaccharide (LPS, L2880), JSH‐23 (M134534, an NF‐κB inhibitor) and TAK‐242 (S80562, a TLR4 inhibitor) were acquired from Guangzhou Yiyou Biotechnology Biological Co., Ltd. The following antibodies were used: anti‐nNOS (CPA5524, Cohesion), anti‐CD32 (40700, SAB), anti‐CD206 (DF4149, Affinity), anti‐COX2 (33345, SAB), anti‐IBA‐1 (A5595, Bimake), anti‐IL‐1β (511369S, CST), anti‐iNOS (2982S, CST), anti‐TLR4 (505258, Zen Bio), anti‐Myd88 (4283S, CST), anti‐NF‐κB (8242S, CST), anti‐p‐NF‐κB (3033, CST), anti‐p38MAPK (8690S, CST), anti‐p‐p38MAPK (4511S, CST), anti‐Bcl‐2 (ab59348, Abcam), anti‐Bcl‐xl (2764S, CST), anti‐S100β (GB13359, Servicebio), anti‐β‐actin (4970S, CST), and anti‐Arg‐1 (93668S, CST). All antibodies were diluted 1:1000. A phosphatase inhibitor cocktail was purchased from Servicebio Biological Technology. An Annexin V APC Apoptosis Detection Kit I (62700‐80) was purchased from Guangzhou Squirrel Biological Co., Ltd. An MRC1 polyclonal antibody (CD206) (PA5‐114370) was purchased from Thermo Fisher Scientific Co., Ltd. A PE‐conjugated anti‐mouse CD$\frac{16}{32}$ [156605] antibody was purchased from BioLegend Biotechnology Co., Ltd. Fixation buffer [420801] and intracellular staining perm wash buffer [421002] were purchased from Dakewe Biotech Co., Ltd. (Shenzhen, China). GMI‐R1 cells (microglia) were acquired from Huatuo Biotechnology Biological Co., Ltd. Schwann cells (RSC96 cells) were acquired from the Shanghai Institute of Cell Biology (GNR6).
## Sciatica model
The CCI model (sciatica model) was constructed as described in previous studies. 14, 15, 16, 17, 18, 19 Before the mice were anesthetized with pentobarbital sodium ($3\%$; 40 mg/kg) and fixed to the operation table, the rats were fasted for 12 h. After exposing the sciatic nerve under a microscope, the right sciatic nerve was tied 4 times with 4.0 sutures at intervals of approximately 1 mm. At this point, we observed a small twitch in the operated hind limb. The rats in the sham group did not undergo nerve ligation. Finally, gentamicin (10 mg/ml, i.m.) was injected.
## Treatment programs
Thirty‐two rats were randomly divided into four groups: the sham operation group, the CCI group, the mecobalamin group, and the ferulic acid group. The rats in the sham operation group and the CCI group were given saline ($0.9\%$, 12 ml/kg), the rats in the mecobalamin group received a gavage of mecobalamin (20 mg/kg), and the rats in the ferulic acid group received ferulic acid (100 mg/kg) by gavage. 20 The drugs were administered for 21 days.
## Behavioral tests
To assess cold hyperalgesia, we chose the acetone experiment. A total of 100 μl of acetone was dropped on the plantar surface of the right paw. Next, we noted the total number of times that the rat lifted or clutched its right hind paw within 120 s. The von Frey test was used to evaluate mechanical hyperalgesia ($50\%$ mechanical withdrawal threshold [MWT]). The hind paw was stimulated 10 times with each filament (2.0–26.0 g) beginning with the 2‐g filament, and paw lifting was considered a positive response. If we detected a positive response, we calculated the pain threshold ($50\%$ g threshold = 10[Xf+kδ]/10,000). We assessed hyperalgesia of the right paw on the 1st, 4th, 7th, 14th, and 21st days after surgery. All tests were repeated three times at 10‐min intervals for each paw, and the mean was calculated.
After CCI, rats exhibited cold and mechanical hyperalgesia from the 4th day to the 21st day (Figure 1A,B; $p \leq 0.05$). Treatment with ferulic acid and mecobalamin (positive control drug) relieved neuropathic pain but did not normalize sensitivity from the 4th day to the 21st day ($p \leq 0.05$). The analgesic effects of ferulic acid were not different from those of mecobalamin ($p \leq 0.05$). These results showed a lack of full functional recovery of the injured sciatic nerves.
**FIGURE 1:** *Ferulic acid attenuated CCI‐induced neuropathic pain. (A, B) Behavioral results (A, von Frey test; B, acetone experiment). (C) RNA sequencing of tissues from the sham operation group and CCI group. (D) Heatmap of the differentially expressed genes between the sham operation group and the CCI group. (E) KEGG pathway enrichment analysis. (F) Heatmap of the differentially expressed genes associated with both the “NF−kappa B signalling pathway” and the “Toll‐like receptor signalling pathway.” (G–I) H&E staining of the sciatic nerve, liver, and kidney. (J, K) Levels of serum inflammatory factors. (L) mRNA levels in the sciatic nerve. #Compared with the sham operation group; *compared with the CCI group; p < 0.05.*
## RNA sequencing
The sciatic nerves of rats in the sham operation group and CCI group were collected 21 days postinjury. Total RNA was isolated using RNAiso Plus. Subsequently, the concentration and quality of the total RNA were assessed using a Nano Drop and Agilent 2100 bioanalyzer (Thermo Fisher Scientific). After the mRNA was purified with oligo(dT)‐attached magnetic beads, it was fragmented into small pieces with fragment buffer at the appropriate temperature. Then, first‐strand cDNA was generated using random hexamer‐primed reverse transcription, followed by second‐strand cDNA synthesis. Afterwards, A‐Tailing Mix and RNA Index Adapters were added by incubation for end repair. The cDNA fragments obtained in the previous step were amplified by PCR, and the products were purified by Ampure XP Beads and then dissolved in EB solution. The products were validated with an Agilent Technologies 2100 bioanalyzer for quality control. The double‐stranded PCR products obtained in the previous step were heated, denatured and circularized by the splint oligo sequence to obtain the final library. Single‐strand circular DNA (ssCir DNA) was formatted as the final library. The final library was amplified with phi29 to make DNA nanoballs (DNBs), which had more than 300 copies of a single molecule. DNBs were loaded into the patterned nanoarray, and paired‐end 150‐base reads were generated on the DNBSEQ‐T7 platform by Tsingke Biotechnology Co., Ltd. The raw reads were filtered using the Trim Galore method (https://ccb.jhu.edu/software/hisat2/index.shtml) to obtain clean reads for subsequent analysis and to ensure the quality of the information analysis. The clean reads obtained after filtering were compared with the reference database annotations (the Rno6 version of the rat genome was selected) using HISAT2 software (https://ccb.jhu.edu/software/hisat2/index.shtml). Differentially expressed genes (DEGs) were screened using Sangerbox, and two criteria were used for screening the DEGs: a false discovery rate (FDR) ≤0.05 and |Log2‐fold change (FC) | ≥ 1. Then, we used DAVID to conduct KEGG enrichment analysis.
## H&E staining, immunohistochemistry, and ELISA
Liver, kidney, and sciatic nerve tissues were fixed in $4\%$ paraformaldehyde and then sliced at a thickness of 3 μm for H&E staining and 9 μm for immunohistochemistry. The sections were deparaffinized, washed with PBS for 5 min, stained with H&E, and washed with water. The sciatic nerve sections were incubated in sodium citrate antigen repair solution (1:1000 dilution; pH = 6). Next, the sections were incubated with primary antibodies diluted to 1:200, followed by the corresponding secondary antibody. The stained sections were viewed under an Olympus fluorescence microscope (BX53). We calculated the IOD/area ratio using Image‐Pro Plus software (Media Cybernetics, Inc.) and conducted statistical analysis.
To measure the concentrations of serum inflammatory factors, we collected 8 ml of blood from the abdominal aorta. The blood samples were centrifuged at 4032 g for 15 min at 4°C, and 800 μl of the supernatant was retained for measurement of inflammatory factor levels. The remaining steps were performed according to the manufacturer's instructions. We used a microplate reader (Bio Tek Instruments, Inc.) to measure the optical density.
## Cell viability and cytotoxicity assays
The viability of GMI‐R1 cells and RSC96 cells was determined by the CCK‐8 assay. According to the results of the cell viability assay, 10 μg/ml, 10 μM, 10 μM, and 2 μM were selected as the concentrations of LPS, TAK‐242, JSH23, and ferulic acid, respectively. 15, 20
## Grouping for the cell experiment
Cells were divided into the control group, LPS group (10 μg/ml LPS), TAK‐242 group (10 μg/ml LPS + 10 μM TAK‐242), JSH23 group (10 μg/ml LPS + 10 μM JSH23), LPS + ferulic acid group (10 μg/ml LPS + 2 μM ferulic acid), and ferulic acid group (2 μM ferulic acid). GMI‐R1 cells and RSC96 cells were cultured in 6‐well plates for 24 h. Next, the medium was discarded, and the cells were washed with PBS. Drugs and LPS were added to the medium at the same time, and then the cells were cultured for 24 h.
## Flow cytometry analysis
We collected and washed GMI‐R1 cells and RSC96 cells. We first measured the percentages of M1 and M2 microglia among LPS‐treated GMI‐R1 cells. The membrane protein CD32 was detected by direct staining. The cells were fixed using fixation buffer, permeabilized twice with intracellular staining perm wash buffer, and incubated with a PE‐conjugated monoclonal mouse CD$\frac{16}{32}$ antibody in the dark for 30 min. Then, the cells were fixed using fixation buffer, blocked and incubated with a polyclonal MRC1 antibody, followed by DyLight 638‐conjugated goat anti‐rabbit IgG for 30 min in the dark. The cells were washed twice with PBS and resuspended in 500 μl of PBS. Annexin V‐APC and PI staining were conducted to evaluate the rate of RSC96 cell death. After treatment, the cells were harvested with trypsin and washed with PBS. Then, the cells were incubated in binding buffer and double stained with Annexin V‐APC and PI in the dark for 20 min at 4°C. The light scattering properties of each sample (105 cells) were analyzed using a flow cytometer (CytoFLEX, Beckman Coulter) equipped with FlowJo software.
## Quantitative real‐time PCR
Total RNA was harvested using RNAiso Plus and synthesized into cDNA with an RT–PCR kit according to the manufacturer's instructions. The relative mRNA expression was calculated by the 2−ΔΔCq method after normalization to the level of β‐actin expression. 21 The Applied Biosystems 7900 real‐time PCR (qPCR) system, SYBR® Green Premix qPCR, and primers, which are shown in Table 1, were used for quantitative real‐time PCR.
**TABLE 1**
| Gene | Forward primer (5′–>3′) | Reverse primer (5′–>3′) |
| --- | --- | --- |
| Arg‐1 | CAGTATTCACCCCGGCTA | CCTCTGGTGTCTTCCCAA |
| Bad | GCAGCCAATAACAGTCAT | CTAAGCTCCTCCTCCATC |
| Bax | CTGGACAACAACATGGAG | AAGTAGAAAAGGGCAACC |
| Bcl‐2 | CAGGCTGGAAGGAGAAGAT | CGGGAGAACAGGGTATGA |
| Bcl‐xl | TAGGTGGTCATTCAGGTAGG | GTGGAAAGCGTAGACAAGG |
| IBA‐1 | ATCAACAAGCACTTCCTC | ATATCTCCATTGCCATTC |
| IL‐10 | AGGGTTACTTGGGTTGCC | GGGTCTTCAGCTTCTCTCC |
| IL‐1β | AGGAGAGACAAGCAACGACA | CTTTTCCATCTTCTTCTTTGGGTAT |
| IL‐4 | CAAGGAACACCACGGAGAA | AGCACGGAGGTACATCACG |
| IL‐6 | AGTTGCCTTCTTGGGACTGATGT | GGTCTGTTGTGGGTGGTATCCTC |
| TGF‐β | ACAGGGCTTTCGCTTCAGT | AGGTCACCTCGACGTTTGG |
| TLR4 | ATCAGTGTATCGGTGGTCAGT | AGCCAGCAATAAGTATCAGGT |
| TNF‐α | GCGTGTTCATCCGTTCTCTACC | TACTTCAGCGTCTCGTGTGTTTCT |
| β‐Actin | CCTAGACTTCGAGCAAGAGA | GGAAGGAAGGCTGGAAGA |
## Western blot analysis
The levels of TLR4, IBA‐1, Arg‐1, IL‐1β, iNOS, Myd88, NF‐κB, p‐NF‐κB, p38MAPK, p‐p38MAPK, Bcl‐2, Bcl‐xl, and β‐actin were measured by western blotting. Total protein was extracted as previously described (Zhang et al., 2021b). We acquired the membrane fraction with a cell membrane protein and cytoplasmic protein extraction kit. The proteins (10 μg) were separated by $12\%$ SDS‐PAGE and transferred onto PVDF membranes, which were blocked with $5\%$ skimmed milk powder for 1 h, incubated with primary antibody (1:1000) overnight at 4°C for 24 h and incubated with secondary antibody (1:30,000) for 1 h. Finally, a ChemiDoc XRS imager was used to visualize the bands. Each experiment was conducted in triplicate.
## Statistical analyses
We used GraphPad Prism 8 to generate all graphs and to perform all analyses. The values are expressed as the mean ± standard deviation. Formal tests for normality were used to assess data distributions. All data were subjected to tests for normality. The behavior data were analyzed by repeated‐measures ANOVA. The other data were analyzed by one‐way ANOVA. Tukey's multiple comparisons test was performed after ANOVA. The results with a p value of <0.05 were considered statistically significant.
## Analysis of DEGs and KEGG pathway analysis
In the present study, DEGs were identified by comparing gene expression in the sciatic nerve between the sham operation group and the CCI group. A total of 14,622 genes were found, and 3777 genes with an FDR ≤0.05 and |Log2(FC)| ≥ 1 (Figure 1C,D) were selected as DEGs. A heatmap (Figure 1D) was used to visualize the expression of the differentially expressed genes in each sample. Then, we performed KEGG pathway analysis of the 3777 DEGs via the DAVID database (Figure 1E; $p \leq 0.05$). KEGG pathway analysis (Figure 1E) revealed that the relationship between the “NF−kappa B signalling pathway” and the “Toll‐like receptor signalling pathway” was quite close, and the two pathways ranked in the top 20. The “NF‐kappa B signaling pathway” is ranked in the top 3, and the “TLR4/NF‐κB” pathway is also involved. The “NF‐kappa B signaling pathway” is involved in the “Toll‐like receptor signaling pathway.” We chose the TLR4/NF‐κB pathway to conduct experimental verification. The differentially expressed genes associated with these two pathways are shown in Figure 1E,F.
## H&E staining
The structures of the liver, kidney, and sciatic nerve were observed via H&E staining. The neural structure of the sciatic nerve (Figure 1G) was normal in the sham operation group but was destroyed after CCI. Ferulic acid and mecobalamin helped restore nerve structure. Liver and kidney structures (Figure 1H,I) were normal in all groups, which indicated that neither the drugs nor CCI had negative effects on the livers and kidneys of the rats.
## Serum inflammatory factor and mRNA levels in the sciatic nerve
The levels of IL‐1β, IL‐6, TNF‐α, IL‐8, IL‐10, TNF‐α, TGF‐β, and INF‐γ in the serum were measured via ELISA. The levels of IL‐1β, IL‐6, IL‐8, TNF‐α, and TGF‐β were significantly increased, and the expression levels of IL‐10 and INF‐γ were decreased after CCI (Figure 1J,K; $p \leq 0.05$). After treatments, ferulic acid and mecobalamin decreased the levels of IL‐1β, IL‐6, IL‐8, TNF‐α, and TGF‐β to nearly normal levels and increased the level of INF‐γ to normal levels ($p \leq 0.05$), but they could not restore the expression of IL‐10 to normal levels. To assess neuroinflammation in the sciatic nerve, we measured the mRNA levels of IL‐1β, IL‐6, IL‐4, TNF‐α, TGF‐β, TLR4, and IBA‐1 in the sciatic nerve and observed that the mRNA levels of IL‐1β, IL‐6, TNF‐α, TGF‐β, TLR4, and IBA‐1 were significantly increased after CCI, while the mRNA level of IL‐4 was reduced (Figure 1L; $p \leq 0.05$). Ferulic acid and mecobalamin lowered the levels of IL‐1β, IL‐6, TNF‐α, TLR4, and IBA‐1 to normal levels, and ferulic acid decreased the level of TGF‐β to normal levels ($p \leq 0.05$). However, they could not normalize the level of IL‐4.
## Immunohistochemical staining of the sciatic nerve
IBA‐1 (a microglial and macrophage marker), M1 polarization markers (IL‐1β, CD32, and iNOS), nNOS, COX2, and TLR4 were expressed at low levels in the normal sciatic nerve in the sham operation group but were expressed at higher levels after CCI (Figure 2A; $p \leq 0.05$). The levels of IBA‐1, IL‐1β, CD32, iNOS, nNOS, COX2, and TLR4 in the ferulic acid and mecobalamin groups were lower than those in the CCI group ($p \leq 0.05$). S100β (a Schwann cell marker) and Arg‐1 (an M2 polarization marker) were expressed at higher levels in the sham operation group and at lower levels in the CCI group (Figure 2A; $p \leq 0.05$). Ferulic acid and mecobalamin increased the expression of S100β and Arg‐1 ($p \leq 0.05$).
**FIGURE 2:** *Immunohistochemical analysis and effects of ferulic acid on the TLR4/NF‐κB pathway in the sciatic nerve. (A) Immunohistochemical staining of IBA‐1, IL‐1β, CD32, iNOS, nNOS, COX2, Arg‐1, S100β and TLR4. (A1) Statistical analysis. (B) Western blots. (B1) Statistical analysis. #Compared with the sham operation group; * compared with the CCI group; p < 0.05.*
## Effects of ferulic acid on the TLR4/NF‐κB pathway
After CCI, the levels of IBA‐1, IL‐1β, iNOS, TLR4, Myd88, p‐NF‐κB, and p‐p38MAPK were obviously increased (Figure 2B; $p \leq 0.05$). Ferulic acid and mecobalamin decreased the levels of these proteins. The levels of p38MAPK and NF‐κB were nearly equal among the groups ($p \leq 0.05$). In addition, ferulic acid and mecobalamin restored the levels of IBA‐1, IL‐1β, iNOS, TLR4, Myd88, p‐NF‐κB, and p‐p38MAPK to nearly normal levels ($p \leq 0.05$).
## Ferulic acid alleviated LPS‐induced apoptosis via the TLR4/NF‐κB pathway in RSC96 cells
Flow cytometry indicated that LPS induced RSC96 cell apoptosis (Figure 3A; $p \leq 0.05$). PCR showed that LPS increased the mRNA expression levels of Bax and Bad and decreased the mRNA levels of Bcl‐2 and Bcl‐xl (Figure 3B; $p \leq 0.05$). Ferulic acid, TAK‐242 or JSH23 inhibited LPS‐induced cell apoptosis ($p \leq 0.05$), but ferulic acid alone did not affect Schwann cell viability (Figure 3A). In addition, ferulic acid combined with TAK‐242 or JSH23 lowered the mRNA level of Bad to normal levels ($p \leq 0.05$), but it could not normalize the level of Bax or increase the levels of Bcl‐2 and Bcl‐xl to normal levels (Figure 3B; $p \leq 0.05$).
**FIGURE 3:** *Ferulic acid alleviated LPS‐induced apoptosis via the TLR4/NF‐κB pathway in RSC96 cells. (A) Flow cytometry. (A1) Statistical analysis. (B) mRNA levels of Bax, Bcl‐2, Bad, and Bcl‐xl. (C) Western blots. (C1) Statistical analysis. #Compared with the sham operation group; * compared with the CCI group; p < 0.05.*
LPS increased the protein levels of IL‐1β, TLR4, Myd88, p‐NF‐κB, and p‐p38MAPK and reduced the expression levels of Bcl‐2 and Bcl‐xl, and the levels of p38MAPK and NF‐κB were almost normal in all groups (Figure 3C; $p \leq 0.05$). After treatment with ferulic acid, TAK‐242 or JSH23, the expression of IL‐1β, TLR4, Myd88, p‐NF‐κB, and p‐p38MAPK was restored almost to normal levels (Figure 3C; $p \leq 0.05$). These results indicated that TAK‐242, JSH23, and ferulic acid alleviated LPS‐induced RSC96 cell apoptosis.
## Ferulic acid promoted the transformation of M1 GMI‐R1 microglia to M2 microglia following LPS treatment via the TLR4/NF‐κB pathway
To explore whether ferulic acid has anti‐inflammatory effects, we conducted PCR and flow cytometry to assess whether ferulic acid regulates the levels of inflammatory cytokines. Figure 4 shows that the mRNA levels of the M1 microglia‐related proinflammatory cytokines IL‐1β, IL‐6, iNOS, and CD32 were increased after LPS treatment but decreased after treatment with TAK‐242, JSH23, and ferulic acid (Figure 4A; $p \leq 0.05$). The levels of M2 microglia‐related anti‐inflammatory cytokines (IL‐4, IL‐10 and Arg‐1) were decreased by LPS stimulation but increased after treatment with TAK‐242, JSH23, and ferulic acid (Figure 4B; $p \leq 0.05$). Flow cytometry and immunofluorescence showed that LPS induced the M1 polarization of microglia (Figure 4C,E; $p \leq 0.05$); TAK‐242, JSH23, and ferulic acid had a repressive effect on the LPS‐induced M1 polarization of microglia; and ferulic acid alone had no effect on microglia (Figure 4D,F; $p \leq 0.05$). LPS reduced the proportion of CD206‐positive microglia (Figure 4D,F; $p \leq 0.05$); TAK‐242, JSH23, and ferulic acid increased the proportion of M2 microglia; and ferulic acid alone increased the proportion of M2 microglia (Figure 4D1,F1). These results indicated that TAK‐242, JSH23, and ferulic acid promoted the transformation of M1 microglia to M2 microglia.
**FIGURE 4:** *Ferulic acid promoted the transformation of M1 microglia to M2 microglia via the TLR4/NF‐κB pathway. (A) M1 microglial marker levels. (B) M2 microglial marker levels. (C) The proportion of M1 microglia (CD32) was determined via flow cytometry. (C1) Statistical analysis. (D) The proportion of M2 microglia (CD206) was determined via flow cytometry. (D1) Statistical analysis. (E) Immunofluorescence of CD32. (E1) Statistical analysis. (F) Immunofluorescence of CD206. (F1) Statistical analysis. (G) Western blots. (G1) Statistical analysis. #Compared with the sham operation group; * compared with the CCI group; p < 0.05.*
The protein levels of iNOS, IL‐1β, TLR4, Myd88, p‐NF‐κB and p‐p38MAPK were obviously increased, while the expression of Arg‐1 was reduced (Figure 4G; $p \leq 0.05$). TAK‐242, JSH23, and ferulic acid reduced the expression levels of iNOS, IL‐1β, TLR4, Myd88, p‐NF‐κB and p‐p38MAPK and increased the level of Arg‐1. The levels of p38MAPK and NF‐κB were almost equivalent among the groups. Ferulic acid reduced the levels of iNOS, IL‐1β, TLR4, Myd88, p‐NF‐κB and p‐p38MAPK to approximately normal levels (Figure 4G; $p \leq 0.05$). These results indicated that ferulic acid promoted the transformation of M1 microglia to M2 microglia via the TLR4/NF‐κB signaling pathway.
## DISCUSSION
In this study, the results of rat and cell experiments provide some convincing evidence that ferulic acid can treat sciatica. The major findings are as follows: [1] ferulic acid effectively alleviated neuroinflammation and promoted sciatic nerve repair; [2] the development of sciatica may be closely related to the “NF−kappa B signalling pathway” and the “Toll‐like receptor signalling pathway”; [3] CCI led to sciatic nerve injury and neuroinflammation with higher expression levels of TLR4, Myd88 and p‐NF‐κB in the sciatic nerve; [4] ferulic acid suppressed Schwann cell apoptosis induced by LPS in an inflammatory environment via the TLR4/NF‐κB pathway; and [5] ferulic acid promoted the transformation of M1 microglia to M2 microglia to suppress neuroinflammation via the TLR4/NF‐κB pathway.
We conducted RNA sequencing of sciatic nerves from the sham operation group and the CCI group. Given that the “NF−kappa B signalling pathway” ranked in the top 3 and the “Toll‐like receptor signalling pathway” ranked in the top 20, these two pathways are closely related to the mechanism of sciatica. Based on the heatmap (Figure 1F), the TLR4/NF‐κB pathway was selected for follow‐up research. The TLR4/NF‐κB pathway is closely related to apoptosis and inflammation. 22 Serum ELISA and PCR of the sciatic nerve showed that nerve injury can increase the levels of inflammatory factors, including IL‐1β, IL‐6, TNF‐α, and IL‐8, and western blotting showed that CCI can lead to higher protein levels of TLR4, MyD88, IL‐1β, and p‐NF‐κB, which verified the RNA sequencing results.
We evaluated the therapeutic effect of ferulic acid on the CCI model via the von Frey test and acetone experiment. Ferulic acid relieved cold and mechanical hyperalgesia, and H&E staining showed that ferulic acid promoted nerve repair. The results of ELISA and PCR demonstrated that ferulic acid decreased the levels of inflammatory factors (IL‐1β, IL‐6, IL‐8, TGF‐β, and TNF‐α) and increased the levels of anti‐inflammatory factors (IL‐4, IL‐10, and INF‐γ). Western blotting and immunohistochemistry indicated that ferulic acid decreased the expression levels of TLR4, IBA‐1, iNOS, IL‐1β, Myd88, p‐p38MAPK, and p‐NF‐κB. These results indicated that ferulic acid may reduce inflammatory factor levels and promote nerve repair. Therefore, we verified the anti‐inflammatory mechanism of ferulic acid in GMI‐R1 cells and the mechanism by which it inhibits apoptosis in RSC96 cells.
S100β is a marker of Schwann cells, which secrete neurotrophic factors and provide structural support and guidance to promote nerve regeneration. 23 *The autologous* transplantation of Schwann cells can promote human peripheral nerve repair in 7.5‐cm and 5‐cm sciatic nerve injuries. 24 After CCI, S100β distribution was altered, and CCI reduced the expression of S100β, indicating that the normal nerve structure was destroyed. However, ferulic acid increased the levels of S100β and promoted nerve repair. A TLR4 inhibitor (TAK‐242) and an NF‐kB inhibitor (JSH23) suppressed LPS‐induced apoptosis by increasing the mRNA and protein levels of Bcl‐2 and Bcl‐xl and decreasing the mRNA levels of Bax and Bad in Schwann cells (RSC96). Ferulic acid attenuated LPS‐induced Schwann cell apoptosis by decreasing the levels of TLR4, p‐NF‐κB, p‐p38MAPK, IL‐1β, Bcl‐2, and Bcl‐xl; increasing the mRNA levels of Bcl‐2 and Bcl‐xl; and decreasing the mRNA levels of Bax and Bad. These results indicated that ferulic acid reduces Schwann cell apoptosis via the TLR4/NF‐κB pathway.
IBA‐1 is a marker of activated microglia and macrophage. Activation of microglia and macrophage induces neuroinflammation. 25 Macrophages in the sciatic nerve had some characteristics of microglia. 26 Acute injury of the sciatic nerve led to a rapid infiltration of circulating monocytes, and the monocytes quickly adapted a macrophage phenotype. 27 The GMI‐R1 cell (microglia) was used to conduct the vitro experiments in this study. Immunohistochemical analysis of the sciatic nerve showed that CCI led to the activation of macrophage or microglia and that ferulic acid suppressed these activations. ELISA showed that CCI increased the expression levels of IL‐1β, IL‐6, TNF‐α, IL‐8, and TGF‐β and decreased the levels of IL‐4 and IL‐10 and that ferulic acid reversed these trends. PCR showed that CCI increased the mRNA levels of IL‐1β, IL‐6, TNF‐α, TLR4, IBA‐1, and TGF‐β and decreased the levels of IL‐4 and that ferulic acid reversed these trends. In cells experiments, M1 microglia had increased expression of several proteins and cytokines, including iNOS, CD32, IL‐1β, IL‐6, while M2 microglia had increased expression of several proteins and cytokines, such as Arg‐1, CD206, IL‐4, and IL‐10. 28 LPS increased the proportion of M1 microglia, while ferulic acid, TAK‐242, and JSH23 increased the proportion of M2 microglia, including increasing the expression of Arg‐1, IL‐4, and IL‐10 and decreasing the levels of IL‐1β, IL‐6, CD32, and iNOS. In addition, CCI increased the protein levels of TLR4, Myd88, IL‐1β, p‐p38MAPK, and p‐NF‐kB, and ferulic acid decreased the levels of these proteins in rats. Similarly, LPS increased the expression levels of TLR4, MyD88, IL‐1β, p‐p38 MAPK, and p‐NF‐κB, while TAK‐242, JSH23, and ferulic acid decreased the expression levels of these proteins. These results indicated that ferulic acid can inhibit neuroinflammation via the TLR4/NF‐κB pathway.
Since ferulic acid can alleviate sciatica by inhibiting neuroinflammation, promoting sciatic nerve repair and exerting an analgesic effect via the TLR4/NF‐κB pathway, ferulic acid might be developed as a novel therapeutic drug.
However, there were limitations to the present study; specifically, the number of samples for RNA sequencing was not adequate. Three samples in each group exhibited individual differences that could not be avoided. In addition, primary cells should be used for cell experiments in the future. For cell experiments, it is better to add the experiments of macrophages in vitro.
## CONCLUSION
Ferulic acid can alleviate sciatica in CCI rats and inhibit neuroinflammation, promote sciatic nerve repair and exert an analgesic effect via the TLR4/NF‐κB pathway.
## AUTHOR CONTRIBUTIONS
Di Zhang, Bei Jing, Zhenni Chen, and Guoping Zhao contributed substantially to the experimental design, data analysis and experimental procedures. Huimei Shi, Xin Li, Shiquian Chang, Zhenni Chen, Li Gao, and Yachun Zheng assisted with the English writing and partial experiments. We thank Yixuan Li for her valuable comments on the statistical analysis and English writing. Guoping *Zhao is* the corresponding author. All data were generated in‐house, and no paper mill was used. All authors agree to be accountable for all aspects of the work, ensuring its integrity and accuracy.
## CONFLICT OF INTEREST
The authors declare no conflicts of interest.
## DATA AVAILABILITY STATEMENT
The data used to support the findings of this study are available from the corresponding author upon request.
## References
1. Davis D, Maini K, Vasudevan A. *StatPearls* (2022)
2. Koes BW, van Tulder MW, Peul WC. **Diagnosis and treatment of sciatica**. *BMJ* (2007) **334** 1313-1317. PMID: 17585160
3. de Campos TF. **Low back pain and sciatica in over 16s: assessment and management NICE guideline [NG59]**. *J Physiother* (2017) **63** 120. PMID: 28325480
4. Pichini S, Solimini R, Berretta P, Pacifici R, Busardo FP. **Acute intoxications and fatalities from illicit fentanyl and analogues: an update**. *Ther Drug Monit* (2018) **40** 38-51. PMID: 29120973
5. Wei J, Su W, Zhao Y. **Maresin 1 promotes nerve regeneration and alleviates neuropathic pain after nerve injury**. *J Neuroinflammation* (2022) **19** 32. PMID: 35109876
6. Clark AK, Gentry C, Bradbury EJ, McMahon SB, Malcangio M. **Role of spinal microglia in rat models of peripheral nerve injury and inflammation**. *Eur J Pain* (2007) **11** 223-230. PMID: 16545974
7. Squillace S, Salvemini D. **Toll‐like receptor‐mediated neuroinflammation: relevance for cognitive dysfunctions**. *Trends Pharmacol Sci* (2022) **43** 726-739. PMID: 35753845
8. Chen CY, Shih YC, Hung YF, Hsueh YP. **Beyond defense: regulation of neuronal morphogenesis and brain functions via toll‐like receptors**. *J Biomed Sci* (2019) **26** 90. PMID: 31684953
9. Bruno K, Woller SA, Miller YI. **Targeting toll‐like receptor‐4 (TLR4)‐an emerging therapeutic target for persistent pain states**. *Pain* (2018) **159** 1908-1915. PMID: 29889119
10. Eidson LN, Murphy AZ. **Blockade of toll‐like receptor 4 attenuates morphine tolerance and facilitates the pain relieving properties of morphine**. *J Neurosci* (2013) **33** 15952-15963. PMID: 24089500
11. Wu FX, Bian JJ, Miao XR. **Intrathecal siRNA against toll‐like receptor 4 reduces nociception in a rat model of neuropathic pain**. *Int J Med Sci* (2010) **7** 251-259. PMID: 20714435
12. Dhaliwal J, Dhaliwal N, Akhtar A, Kuhad A, Chopra K. **Beneficial effects of ferulic acid alone and in combination with insulin in streptozotocin induced diabetic neuropathy in Sprague Dawley rats**. *Life Sci* (2020) **255**. PMID: 32473246
13. Liu G, Nie Y, Huang C. **Ferulic acid produces neuroprotection against radiation‐induced neuroinflammation by affecting NLRP3 inflammasome activation**. *Int J Radiat Biol* (2022) **98** 1442-1451. PMID: 35445640
14. Yoon C, Wook YY, Sik NH, Ho KS, Mo CJ. **Behavioral signs of ongoing pain and cold allodynia in a rat model of neuropathic pain**. *Pain* (1994) **59** 369-376. PMID: 7708411
15. Zhang D, Chang S, Li X. **Therapeutic effect of paeoniflorin on chronic constriction injury of the sciatic nerve via the inhibition of Schwann cell apoptosis**. *Phytother Res* (2022) **36** 2572-2582. PMID: 35499270
16. Zhang D, Chen D, Ma SS. **Effect of Danggui Sini decoction on the behaviour and dorsal root ganglion TRP Channel of neuropathic pain in CCI rats**. *Indian J Pharm Sci* (2019) **81** 922-932
17. Zhang D, Jing B, Li X. **Antihyperalgesic effect of Paeniflorin based on chronic constriction injury in rats**. *Rev Bras Farmacogn* (2022) **32** 375-385
18. Zhang D, Li X, Jing B. **α‐Asarone attenuates chronic sciatica by inhibiting peripheral sensitization and promoting neural repair**. *Phytother Res* (2022) 1-12
19. Zhang D, Sun J, Yang B, Ma S, Zhang C, Zhao G. **Therapeutic effect of Tetrapanax papyriferus and hederagenin on chronic neuropathic pain of chronic constriction injury of sciatic nerve rats based on KEGG pathway prediction and experimental verification**. *Evid Based Complement Alternat Med* (2020) **2020**. PMID: 32617100
20. Zhang D, Jing B, Chen ZN. **Ferulic acid alleviates sciatica by inhibiting peripheral sensitization through the RhoA/p38MAPK signalling pathway**. *Phytomedicine* (2022) **106**. PMID: 36115115
21. Zhang D, Li X, Jing B. **Identification of pathways and key genes in male late‐stage carotid atherosclerosis using bioinformatics analysis**. *Exp Ther Med* (2022) **24** 460. PMID: 35747144
22. El‐Sherbiny M, Eisa HN, El‐Magd FNA, Elsherbiny MN, Said E, Khodir EA. **Anti‐inflammatory/anti‐apoptotic impact of betulin attenuates experimentally induced ulcerative colitis: an insight into TLR4/NF‐kB/caspase signalling modulation**. *Environ Toxicol Phar* (2021) **88**
23. Castro R, Taetzsch T, Vaughan SK. **Specific labeling of synaptic schwann cells reveals unique cellular and molecular features**. *Elife* (2020) **9**. PMID: 32584256
24. Gersey ZC, Burks SS, Anderson KD. **First human experience with autologous Schwann cells to supplement sciatic nerve repair: report of 2 cases with long‐term follow‐up**. *Neurosurg Focus* (2017) **42** E2
25. Simoes JLB, Galvan ACL, da Silva ELM, Ignacio ZM, Bagatini MD. **Therapeutic potential of the purinergic system in major depressive disorder associated with COVID‐19**. *Cell Mol Neurobiol* (2022) 1-17
26. Amann L, Prinz M. **The origin, fate and function of macrophages in the peripheral nervous system‐an update**. *Int Immunol* (2020) **32** 709-717. PMID: 32322888
27. Ydens E, Amann L, Asselbergh B. **Profiling peripheral nerve macrophages reveals two macrophage subsets with distinct localization, transcriptome and response to injury**. *Nat Neurosci* (2020) **23** 676-689. PMID: 32284604
28. Ruytinx P, Proost P, Van Damme J, Struyf S. **Chemokine‐induced macrophage polarization in inflammatory conditions**. *Front Immunol* (2018) **9** 1930. PMID: 30245686
|
---
title: 'Association of serum uric acid to lymphocyte ratio, a novel inflammatory biomarker,
with risk of stroke: A prospective cohort study'
authors:
- Xue Tian
- Penglian Wang
- Shuohua Chen
- Yijun Zhang
- Xiaoli Zhang
- Qin Xu
- Yanxia Luo
- Shouling Wu
- Anxin Wang
journal: CNS Neuroscience & Therapeutics
year: 2023
pmcid: PMC10018086
doi: 10.1111/cns.14094
license: CC BY 4.0
---
# Association of serum uric acid to lymphocyte ratio, a novel inflammatory biomarker, with risk of stroke: A prospective cohort study
## Abstract
Inflammation plays an important role in the pathological progress associated with stroke. Serum uric acid (SUA) to lymphocyte ratio (ULR), a novel inflammatory biomarker, has been considered as a better risk stratification tool of adverse outcomes than SUA or lymphocyte alone. Based on the Kailuan study, we enrolled 93,023 participants to investigate whether ULR produced more predictive value for stroke and explore the potential mediators of the associations. We found ULR was significantly associated with the risk of HS, but not with IS. ULR outperformed SUA or lymphocytes alone in predicting stroke. Systolic blood pressure, diastolic blood pressure, and estimated glomerular filtration rate were potential mediators for the association.
### Main Problem
Inflammation plays an important role in the pathological progress associated with stroke. Serum uric acid (SUA) to lymphocyte ratio (ULR), a novel inflammatory biomarker, has been considered as a better risk stratification tool of adverse outcomes than SUA or lymphocyte alone. This study aimed to investigate whether ULR produced more predictive value for stroke and explore the potential mediators of the associations.
### Methods
This study enrolled 93,023 Chinese participants without stroke and myocardial infarction at baseline. Cox proportional hazard models were used to analyze the associations of ULR with stroke and subtypes. Mediation analyses were conducted to explore potential mediators of the associations.
### Results
During a median follow‐up of 13.00 years, 6081 cases of incident stroke occurred, including 5048 cases of ischemic stroke (IS) and 900 cases of hemorrhagic stroke (HS). After adjustment for confounders, the Q4 group was associated with a higher risk of HS (HR, 1.25; $95\%$ CI, 1.03–1.50), but not with total stroke (HR, 1.07; $95\%$ CI, 1.03–1.13) or IS (HR, 1.04; $95\%$ CI, 0.97–1.12). No significant associations were found between SUA or lymphocyte and any stroke. ULR outperformed SUA or lymphocytes alone in predicting stroke. Additionally, the significant association between ULR and HS was partially mediated by systolic blood pressure ($20.32\%$), diastolic blood pressure ($11.18\%$) and estimated glomerular filtration rate ($9.19\%$).
### Conclusions
ULR was significantly associated with the risk of HS, but not with IS. Systolic blood pressure, diastolic blood pressure and estimated glomerular filtration rate were potential mediators for the association.
## INTRODUCTION
Stroke is a disabling sequel of atherosclerosis with high morbidity and mortality rates worldwide, especially in East Asian countries. 1 Given the complexity of stroke, different mechanisms are thought to be involved in this pathophysiology. In fact, increasing evidence shows that inflammation is involved in all stages of stroke. 2, 3, 4, 5 Inflammation‐induced neuronal death is one of the key factors in stroke pathology. 6 Therefore, inflammation control is a new strategy for reducing the occurrence and damage from stroke.
Serum uric acid (SUA) has been reported to have both antioxidant and pro‐inflammatory properties. 7, 8 Elevated SUA is linked to inflammation and has been showed to be associated with metabolic syndrome, carotid atherosclerosis, endothelia dysfunction, oxidative stress and inflammation, which have adverse effects on platelet adhesiveness and aggregation. 9, 10 Systemic inflammation response is characterized by increased neutrophil counts and decreased lymphocyte counts in chronic diseases. 11, 12 And the neutrophil–lymphocyte ratio was reported to be associated with the prognosis of acute ischemic stroke in both clinical study and meta‐analysis. 13, 14 Decrease in lymphocyte count reflects an impairment of adaptive immune system and poor general health status, and immune system is contributed to all levels of stroke cascade. 15 However, it should be stated that the role of SUA or lymphocyte counts in stroke and subtypes has long been under debate. 16, 17 Currently, SUA to lymphocyte ratio (ULR), showing a joint effect of SUA and lymphocyte count, has been considered as a novel risk stratification tool to refine prognostic prediction for heart disease and cancer. Additionally, ULR exhibited a better predictive ability than SUA or lymphocyte alone in these diseases. 18, 19 However, whether ULR could be used as a novel inflammatory biomarker to produce better predictive value for stroke and subtypes has not been investigated up to now.
Therefore, based on a large community‐based population study, our present study sought to investigate the associations of ULR with stroke and its subtypes (ischemic stroke [IS] and hemorrhagic stroke [HS]), and to assess the potential mediators in the associations using mediation analysis.
## Study population
The data analyzed in this explorative study were retrieved from the Kailuan study, which is an ongoing prospective cohort study launched in Tangshan City, China. The detailed study designed and procedures have been described elsewhere. 20, 21 Briefly, from June 2006 to October 2007, a total of 101,510 participants aged 18–98 years were enrolled from 11 affiliated hospitals of the Kailuan community. All the participants underwent questionnaire interviews, clinical examinations, and laboratory tests at enrollment, and were followed up every 2 years since baseline [2006]. Participants meeting the following criteria were excluded1: had a history of stroke or MI ($$n = 3715$$)2; had missing data on SUA or lymphocytes at baseline ($$n = 4772$$). Finally, a total of 93,023 participants were enrolled in the current study (Figure S1). A comparison of baseline characteristics between excluded and included participants was presented in Table S1. The study was performed according to the guidelines of the Declaration of Helsinki and was approved by the Ethics Committee of Kailuan General Hospital and Beijing Tiantan Hospital. All participants provided written informed consent.
## Assessment of SUA, lymphocytes, and ULR
Fasting blood samples were collected in the morning after an 8‐ to 12‐h overnight fast. The concentration of SUA was examined with a commercial kit (Ke Hua Biological Engineering Corporation, Shanghai, China) using an automatic biochemical analyzer (Hitachi 7600, Tokyo, Japan), according to the manufacturer's instructions. Lymphocytes were determined using a full blood count analyzer (Sysmex XT‐1800i, Sysmex Corporation). ULR was calculated as SUA (mg/dl)/lymphocyte count (×109/L), as previously reported. 18, 19
## Assessment of outcomes
The primary outcome was the first occurrence of stroke (including IS, HS, and subarachnoid hemorrhage), either the first nonfatal stroke event, or stroke death without a preceding nonfatal event. The secondary outcomes were the individual endpoint of ischemic or HS. Subarachnoid hemorrhage events were not analyzed separately since the small sample size ($$n = 133$$). Participants were followed up via face‐to‐face interviews at every 2‐year interval, until 31 December 2019, or until the event of interest or death. Incident stroke events were ascertained by checking each year's discharge lists from the 11 local hospitals in Kailuan, medical records from medical insurance, or face‐to‐face interview on self‐reported history if disease. Vital status was collected from death certificates from the provincial vital statistic offices. The diagnosis of incident stroke was confirmed by medical review, using the World Health Organization criteria. 22 Information on imaging diagnoses (including brain computerized tomography, magnetic resonance, or lumbar puncture) were collected to identify the type of incident stroke cases.
## Assessment of covariates
Information on age, sex, education level, family income, smoking status, drinking status, physical activity, history of hypertension, diabetes, dyslipidemia, the use of antihypertensive, antidiabetic, or lipid‐lowering agents were collected using self‐reported questionnaires. Active physical activity was defined as ≥80 minutes of activity per week. Body mass index (BMI) was calculated as weight (kg)/ height (m)2. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured three times with the participants in the seated position using a mercury sphygmomanometer, and the average of three readings was used in the analyses.
Laboratory tests, including fasting blood glucose (FBG), total cholesterol (TC), triglycerides (TG), low‐density lipoprotein cholesterol (LDL‐C), high‐density lipoprotein cholesterol (HDL‐C), serum creatinine and high sensitivity C‐reactive protein (hs‐CRP) were assessed by an auto analyzer (Hitachi 747, Hitachi) at the central laboratory of Kailuan Hospital. Estimated glomerular filtration rate (eGFR) was calculated using the creatinine‐based Chronic Kidney Disease Epidemiological Collaboration (CKD‐EPI) equation. 23
## Statistical analysis
The baseline characteristics were described as the mean ± standard deviation for continuous variables, and frequencies with percentages for categorical variables. Participants were divided into four groups by quartiles of ULR. Person‐years was calculated from baseline to the first occurrence of stroke, mortality, or the end of the study (December 31, 2019), whichever came first. The incidence rate of stroke was calculated by dividing the number of incident cases by the total follow‐up duration (person‐years). A Cox proportional hazards model was used to estimate the associations of ULR with incident stroke and its subtypes. ULR was categorized in quartiles and was also modeled as a continuous variable in the analyses. Proportional hazards assumption was stratified by checking the Schoenfeld residual plots. Three models were built systematically, model 1 was unadjusted; model 2 was adjusted for age, sex, education, drinking status, smoking status, physical activity, BMI, SBP, DBP, FBG, TC, and HDL‐C; model 3 was further adjusted for history of hypertension, diabetes, dyslipidemia, medication on hypertension, diabetes, dyslipidemia, eGFR, and hs‐CRP. Restricted cubic spline with five knots (at the 5th, 25th, 50th, 75th, and 95th percentiles) was used to assess the shape of the associations between ULP and stroke.
A series of sensitivity analyses was performed to validate the robustness of our findings. First, the competing risk model was applied to with non‐stroke death being regarded as a competing risk event. Second, to explore the potential impact of reverse causality, we repeated the primary analysis using a 2‐year lagged period by excluding participants who developed stroke cases within the first 2 years of follow‐up. Third, we excluded population with an eGFR less than 30 ml/min/1.73 m2. Fourth, we excluded participants who used antihypertensive, antidiabetic, or lipid‐lowering agents to explore whether the association were confounded by medication use. Finally, we further excluded participants who had a history of cancer at baseline. Subgroup analyses stratified by age (<60 vs ≥60 years) and sex were performed, interaction of stratified variables with ULR was tested using likelihood test. The prognostic accuracy of ULR in predicting risk of stroke in terms of area under curve (AUC), sensitivity, and specificity was performed by using receiver operative characteristics curves, and the corresponding sensitivity and specificity were recorded with the largest Youden index. To explore whether ULR was better than SUA and lymphocyte count alone in predicting stroke, we compared the strength of the association of SUA, lymphocyte count, and ULR with the risk of stroke. Additionally, we used the C‐statistics, integrated discrimination improvement (IDI), and net reclassification index (NRI) to compare the incremental predictive value of SUA, lymphocyte count, and ULR beyond conventional risk factors.
Once the temporal relationships of ULR with stroke and its subtypes had been established, mediation analysis was performed to examine whether the associations between ULR (X) and stroke (Y) were mediated by other metabolic factors (M), using the method described by Valeri and VanderWeele. 24, 25, 26 *In* general, four steps are involved in the mediation analysis1: demonstrating that the predictor is associated with the outcome (Model Y = β Tol X, β Tol = total effect)2; demonstrating that the predictor is associated with the mediator (Model M = β 1 X, β 1 = indirect effect 1)3; demonstrating which part of the outcome is explained by controlling for the predictor (Model Y = β 2 M + β Dir X, β 2 = indirect effect 2, β Dir = direct effect); and4 calculating the proportion of mediation: mediation effect (%) = (β 1 × β 2/β Tol) × $100\%$. We adjusted age, sex, education, income, smoking status, and drinking status in the mediation analysis, as it is necessary for mediation models in which baseline covariates are sufficient to control for exposure‐outcome, mediator‐outcome, and exposure‐mediator confounding. 27, 28, 29 All statistical analyses were conducted using SAS version 9.4 software (SAS Institute, Cary, NC, USA). All reported p‐values were based on two‐sided tests of significance, and values of $p \leq 0.05$ were deemed statistically significant.
## Baseline characteristics
A total of 93,023 participants were enrolled in the current study, 73,778 ($79.31\%$) participants were men, and the mean age was 51.57 ± 12.55 years old. The baseline characteristics of participants according to quartiles of ULR was presented in Table 1. Compared with participants in the Q1 group, those with higher ULR tended to older, men, well‐educated, current smokers and drinkers, have a higher prevalence of hypertension, dyslipidemia, were more likely to take antihypertensive medication, have a higher level of BMI, SBP, DBP, TC, LDL‐C, hs‐CRP, but a lower prevalence of diabetes and a lower level of FBG, HDL‐C and eGFR.
**TABLE 1**
| Characteristics | Overall | Quartiles of ULR | Quartiles of ULR.1 | Quartiles of ULR.2 | Quartiles of ULR.3 | p value |
| --- | --- | --- | --- | --- | --- | --- |
| Characteristics | Overall | Q1 (<1.63) | Q2 (1.63–2.11) | Q3 (2.12–2.74) | Q4 (≥2.75) | p value |
| Participants, n (%) | 93023 | 23253 | 23244 | 23274 | 23252 | |
| Age, years | 51.57 ± 12.55 | 49.51 ± 11.77 | 50.55 ± 12.07 | 51.65 ± 12.49 | 54.59 ± 13.24 | <0.0001 |
| Males, n (%) | 73,778 (79.31) | 16,378 (70.43) | 17,789 (76.53) | 19,087 (82.01) | 20,524 (88.27) | <0.0001 |
| High school or above, n (%) | 12,665 (14.12) | 2829 (12.55) | 3125 (13.86) | 3341 (14.83) | 3370 (15.27) | <0.0001 |
| Income ≥800 yuan/month, n (%) | 6222 (6.93) | 1293 (5.73) | 1568 (6.95) | 1693 (7.51) | 1668 (7.55) | <0.0001 |
| Current smoker, n (%) | 30,424 (32.7) | 6780 (29.16) | 7673 (33.01) | 8192 (35.20) | 7776 (33.44) | <0.0001 |
| Current alcohol, n (%) | 33,110 (36.58) | 6553 (28.97) | 8028 (35.42) | 9049 (39.82) | 9477 (42.10) | <0.0001 |
| Active physical activity, n (%) | 13,499 (14.51) | 2823 (12.14) | 3231 (13.90) | 3531 (15.17) | 3914 (16.83) | <0.0001 |
| Hypertension, n (%) | 40,449 (43.48) | 9773 (42.03) | 9624 (41.40) | 9965 (42.82) | 11,085 (47.67) | <0.0001 |
| Diabetes mellitus, n (%) | 8404 (9.03) | 2541 (10.93) | 2097 (9.02) | 1836 (7.89) | 1929 (8.30) | <0.0001 |
| Dyslipidemia, n (%) | 32,061 (34.46) | 7757 (33.36) | 7830 (33.69) | 8121 (34.89) | 8352 (35.92) | <0.0001 |
| Antihypertensive agents, n (%) | 8766 (9.42) | 1649 (7.09) | 1965 (8.45) | 2307 (9.91) | 2845 (12.24) | <0.0001 |
| Hypoglycemic agents, n (%) | 1981 (2.13) | 569 (2.45) | 498 (2.14) | 437 (1.88) | 477 (2.05) | 0.0003 |
| Lipid‐lowering agents, n (%) | 614 (0.66) | 133 (0.57) | 161 (0.69) | 144 (0.62) | 176 (0.76) | 0.0699 |
| Body mass index, kg/m2 | 25.03 ± 3.50 | 24.87 ± 3.52 | 24.99 ± 3.48 | 25.08 ± 3.47 | 25.17 ± 3.51 | <0.0001 |
| Systolic blood pressure, mmHg | 130.81 ± 20.86 | 129.74 ± 20.60 | 129.88 ± 20.48 | 130.71 ± 20.84 | 132.92 ± 21.34 | <0.0001 |
| Diastolic blood pressure, mmHg | 83.48 ± 11.76 | 83.01 ± 11.56 | 83.11 ± 11.62 | 83.50 ± 11.78 | 84.30 ± 12.03 | <0.0001 |
| Fasting blood glucose, mmol/L | 5.46 ± 1.68 | 5.55 ± 1.92 | 5.45 ± 1.68 | 5.40 ± 1.53 | 5.43 ± 1.54 | <0.0001 |
| Total cholesterol, mmol/L | 4.94 ± 1.15 | 4.89 ± 1.21 | 4.94 ± 1.13 | 4.96 ± 1.13 | 4.97 ± 1.11 | <0.0001 |
| Triglyceride, mmol/L | 1.68 ± 1.38 | 1.66 ± 1.33 | 1.65 ± 1.34 | 1.67 ± 1.37 | 1.72 ± 1.47 | <0.0001 |
| LDL cholesterol, mmol/L | 2.34 ± 0.92 | 2.35 ± 0.94 | 2.33 ± 0.92 | 2.33 ± 0.90 | 2.33 ± 0.93 | 0.0040 |
| HDL cholesterol, mmol/L | 1.55 ± 0.41 | 1.57 ± 0.41 | 1.55 ± 0.40 | 1.55 ± 0.41 | 1.54 ± 0.41 | <0.0001 |
| eGFR, ml/min/1.73m2 | 81.81 ± 25.77 | 81.67 ± 26.69 | 82.01 ± 24.84 | 82.4 ± 24.56 | 81.16 ± 26.88 | <0.0001 |
| hs‐CRP, mg/L | 2.40 ± 6.54 | 2.34 ± 7.01 | 2.27 ± 6.02 | 2.30 ± 6.26 | 2.71 ± 6.79 | <0.0001 |
| SUA, mg/dl | 4.85 ± 1.41 | 3.74 ± 0.98 | 4.55 ± 1.00 | 5.15 ± 1.12 | 5.98 ± 1.43 | <0.0001 |
| lymphocyte count, *109/L | 2.37 ± 2.71 | 3.21 ± 5.25 | 2.43 ± 0.54 | 2.14 ± 0.47 | 1.71 ± 0.44 | <0.0001 |
## Associations of ULR with stroke and its subtypes
During a median follow‐up of 13.00 (interquartile range, 12.61–13.18; range from 0.16–14.93 years) years, a total of 6081 stroke cases ($6.54\%$) occurred, including 5048 cases IS and 900 cases of HS. The incidence per 1000 person‐years of total, ischemic, and HS increased substantially with increasing ULR quartiles, ranging from 4.76, 3.97, and 0.66 in the Q1 group to 6.38, 5.13 and 1.05 in the Q4 group, respectively (Table 2). After adjusted for potential variables, participants in the Q4 group remained having a $25\%$ higher risk of HS (adjusted HR, 1.25; $95\%$ CI, 1.03–1.50, p for trend = 0.0050), compared with those in the Q1 group. Nevertheless, the association of ULR with total stroke (HR, 1.04; $95\%$ CI, 0.97–1.12; p for trend = 0.3798) and IS (HR, 1.04; $95\%$ CI, 0.96–1.12; p for trend = 0.7358) attenuated to an insignificant level. Multivariable adjusted spline regression models showed that there was a non‐linear association between ULR and HS (p for non‐linear = 0.0053), but not with total stroke and IS (Figure 1).
The sensitivity analyses using competing risk model, excluding participants who developed stroke cases within the first 2 years of follow‐up ($$n = 1245$$), excluding those with eGFR less than 30 ml/min/1.73 m2 ($$n = 526$$), excluding those who used antihypertensive, antidiabetic, or lipid‐lowering agents ($$n = 22$$,530), excluded those with a history of cancer ($$n = 314$$) all generated similar findings with the primary analysis (Figure 2, Table S2). Subgroup analyses stratified by age and sex showed that the associations of ULR with stroke were similar to the full cohort and consistent across different subgroups, p values for tests of two‐way interaction effects of ULR by age and sex on total, ischemic, and HS were all >0.05, indicating no significant effect modification of the association between ULR and stroke (Table S3).
**FIGURE 2:** *Sensitivity analysis for the association of ULR with stroke and subtypes. Sensitivity 1: Taking non‐stroke related death as competing risk event rather than censoring. Sensitivity 2: Excluded person time and incident stroke cases from the first 2 years of follow‐up (n = 91,788 for analysis). Sensitivity 3: Excluded participants with estimated glomerular filtration rate < 30 ml/min/1.73m2 (n = 92,497 for analysis). Sensitivity 4: Excluded participants with medication on hypertension, diabetes, dyslipidemia (n = 70,493 for analysis). Sensitivity 5: Excluded participants with a history of cancer (n = 92,709 for analysis). Adjusted for age, sex, education, drinking status, smoking status, physical activity, body mass index, systolic blood pressure, diastolic blood pressure, fasting blood glucose, total cholesterol, high density lipoprotein cholesterol, history of hypertension, diabetes, dyslipidemia, medication on hypertension, diabetes, dyslipidemia, estimated glomerular filtration rate, and high‐sensitivity C‐reactive protein*
## Comparisons of the associations between SUA, lymphocytes, ULR, and stroke
We compared the associations of SUA, lymphocytes, and ULR with the risk of stroke, the results showed that the risk of total stroke, IS, and HS was not significantly increased with increasing SUA (p for trend = 0.2528, 0.4683, and 0.2258, respectively) or lymphocytes (p for trend = 0.2641, 0.8542, and 0.1534, respectively), whereas the risk of HS was significantly increased with increasing ULR (p for trend = 0.0050) (Table S4).
The prognostic accuracy of ULR in predicting risk of stroke in terms of area AUC, sensitivity, and specificity is presented in Figure S2. We also compared the incremental predictive value of SUA, lymphocytes, and ULR beyond the conventional risk factors. When HS was the outcome of interest, the C‐statistics by the conventional model significantly improved with the addition of ULR (from 0.701 to 0.706, $$p \leq 0.0028$$), but not significantly improved with the addition of SUA ($$p \leq 0.1000$$) and lymphocytes ($$p \leq 0.2384$$). The discriminatory power and risk reclassification appeared to substantially better with the addition of ULR (IDI, $0.02\%$; $95\%$ CI, $0.01\%$–$0.04\%$; $$p \leq 0.0045$$; NRI, 14.80; $95\%$ CI, 8.25–21.34; $p \leq 0.0001$), but not with the addition of either SUA ($$p \leq 0.1789$$ for IDI, and 0.3345 for NRI) or lymphocytes ($$p \leq 0.1651$$ for IDI, and 0.2564 for NRI). We also found that the C‐statistics ($$p \leq 0.0358$$), IDI ($$p \leq 0.0251$$), NRI ($$p \leq 0.0010$$) were significantly improved by the addition of ULR into the conventional model with SUA and lymphocyte count, indicating that the ULR had a higher predicting value than single SUA, lymphocyte count, and the combination of the two biomarkers (Table 3 and Figure S3). When total stroke or IS was the outcomes of interest, the addition of either SUA, lymphocytes, or ULR was not significantly improved the predictive value of conventional model (Table 3).
**TABLE 3**
| Outcomes | C statistics | C statistics.1 | IDI | IDI.1 | Category‐free NRI | Category‐free NRI.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Outcomes | Estimate (95% CI) | p | Estimate (95% CI), % | p | Estimate (95% CI), % | p |
| Total stroke | | | | | | |
| Conventional model a | 0.722 (0.716–0.728) | | Reference | | Reference | |
| Conventional model + SUA | 0.722 (0.716–0.728) | 0.5507 | 0.00 (0.00–0.00) | 0.7953 | 1.29 (−3.89–1.30) | 0.3292 |
| Conventional model +LY | 0.722 (0.716–0.728) | 0.3978 | 0.00 (0.00–0.00) | 0.6849 | 4.07 (−2.95–2.13) | 0.7585 |
| Conventional model +ULR | 0.723 (0.717–0.728) | 0.2875 | 0.01 (0.00–0.00) | 0.7813 | 2.60 (0.04–5.17) | 0.0494 |
| Ischemic stroke | | | | | | |
| Conventional model a | 0.721 (0.715–0.727) | | | | | |
| Conventional model + SUA | 0.721 (0.715–0.728) | 0.7625 | 0.00 (0.00–0.00) | 0.6537 | 2.44 (−0.38–5.28) | 0.0908 |
| Conventional model +LY | 0.721 (0.715–0.728) | 0.6967 | 0.00 (0.00–0.01) | 0.5607 | 0.36 (−0.24–3.13) | 0.8059 |
| Conventional model +ULR | 0.721 (0.715–0.728) | 0.3708 | 0.00 (0.00–0.01) | 0.4851 | 4.63 (1.83–7.43) | 0.0014 |
| Hemorrhagic stroke | | | | | | |
| Conventional model a | 0.701 (0.685–0.718) | | Reference | | Reference | |
| Conventional model + SUA | 0.704 (0.688–0.721) | 0.1000 | 0.01 (0.00–0.01) | 0.1789 | 2.03 (−1.03–5.15) | 0.3345 |
| Conventional model +LY | 0.703 (0.687–0.720) | 0.2384 | 0.00 (0.00–0.02) | 0.1651 | 2.29 (−0.42–18.74) | 0.2564 |
| Conventional model +ULR | 0.706 (0.690–0.723) | 0.0028 | 0.02 (0.01–0.04) | 0.0045 | 14.80 (8.25–21.34) | <0.0001 |
| Conventional model+SUA + LY | 0.704 (0.688–0.722) | | Reference | | Reference | |
| Conventional model +SUA + LY + ULR | 0.706 (0.691–0.723) | 0.0358 | 0.01 (0.00–0.01) | 0.0251 | 4.36 (1.80–6.93) | 0.0010 |
## Mediation analysis between ULR and HS
Since ULR was significantly associated with the risk of HS, mediation analysis was then used to assess the potential mechanisms. We tested BMI, SBP, DBP, FBG, TC, TG, LDL‐C, HDL‐C, eGFR, and hs‐CRP as potential mediators of the association between ULR and HS. The results showed that the total effect of ULR on HS was 0.0019 ($95\%$ CI, 0.0013–0.0026; $$p \leq 0.0010$$). The association between ULR and HS was partially mediated by SBP (β indir = 0.0004, mediation proportion [MP] = $20.32\%$), DBP (β indir = 0.0002, MP = $11.18\%$), and eGFR (β indir = 0.0002, MP = $9.19\%$) (Figure 3). The remaining factors: BMI, FBG, TC, TG, LDL‐C, HDL‐C, and hs‐CRP did not play significant mediating roles in the association (Table S5).
**FIGURE 3:** *Mediation analyses of the association of ULR with stroke and subtypes. (A) Contribution of SBP; (B) Contribution of DBP; (C) Contribution of eGFR. Adjusted for age, sex, education, drinking status, smoking status, physical activity, body mass index, systolic blood pressure, diastolic blood pressure, fasting blood glucose, total cholesterol, high density lipoprotein cholesterol, history of hypertension, diabetes, dyslipidemia, medication on hypertension, diabetes, dyslipidemia, estimated glomerular filtration rate, and high‐sensitivity C‐reactive protein. ULR, serum uric acid to lymphocyte count ratio; SBP, systolic blood pressure; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate. *p < 0.05; **p < 0.01*
## DISCUSSION
In our present study, we prospectively investigated the association between ULR, a novel biomarker of inflammation and the risk of incident stroke. The major findings are listed as follows1: ULR was positively associated with the risk of HS, but not with either total stroke or IS, after adjustment for confounding risk factors2; ULR was a better predictor of HS than SUA or lymphocytes alone, the addition of ULR to the conventional model significantly promoted the ability of risk stratification3; the association between ULR and HS was partially mediated by SBP, DBP, and eGFR.
The immune‐biomarkers of stroke have drawn a lot of attention recently, such as lymphocyte, 30 neutrophil, 31 leukocyte, 32 monocytes, 33 T cell, 34 and et al. Current literature on ULR is limited and restricted in patients after surgery. ULR was first proposed in the study of Wei et al., a retrospective study conducted in consecutive patients with rheumatic heart disease undergoing valve replacement surgery. The study with 949 elderly patients demonstrated that ULR, combining the effect of SUA and lymphocyte count, produced more prognostic value in elderly patients than SUA and lymphocyte count. Similarly, another study was carried out based on a prospectively‐maintained database and included 335 patients after video‐assisted thoracoscopic surgery lobectomy for early‐stage non‐small‐cell lung cancer, showing that elevated ULR could independently predict both unfavorable overall survival and disease‐free survival. The better predictive role of ULR was extended for primary prevention of stroke in our current study. Our results demonstrated that the risk of HS was significantly associated with a higher level of ULR, even adjustment for other potential confounders, but not with SUA or lymphocyte count. Furthermore, the addition of ULR into the conventional model yielded an incremental predictive value. Our findings suggested that application of ULR (a novel biomarker) could improve risk stratification ability in the primary prevention of HS.
We hypothesized that the biological reasons underlying the better predictive value of this ULR tool may be elucidated by a combination of the following three plausible mechanisms. First, SUA had both pro‐oxidant and antioxidant capabilities. 35 Experimental studies have shown that increased SUA as a pro‐oxidant is associated with endothelial dysfunction, increased oxidative stress, elevated plasma renin activity and systemic inflammation mediators, which may contribute to the development of HS. 36, 37 Population study found that elevated serum uric acid increases the risk of stroke recurrence. 38 On the other hand, SUA is also an abundant natural antioxidant capable of reducing cellular oxidation, a major cause of neurodegenerative disease. 39 The double‐edged properties may reduce the impact of SUA on the risk of HS, leading to the inconsistent association between SUA and HS.
Second, HS occurs when a blood vessel with the brain ruptures and releases blood. 40 A systemic inflammation response, frequently accompanied with a sharp decline in peripheral lymphocyte count, has been proposed to be involve in the weakening of vessel wall and lead to vessel rupture, thus may increase the risk of HS. 41 One study have found that the white blood cell count was significantly associated with deep intracerebral hemorrhage compared with small artery occlusion group. 41 On the other hand, it is stated that lymphocyte count can mediate the inflammatory response, where it may increase the level of anti‐inflammatory cytokines and suppress the production of pro‐inflammatory cytokines. 42 These may contribute to the controversial role of lymphocyte counts in the development of HS.
Finally, ULR, a combination of SUA and lymphocyte count, is elevated with the increase in SUA or the decrease in lymphocyte count, both of which were associated with a higher level of inflammation. An elevated SUA level has also been proved to show a strong relationship with a high circulating level of various pro‐inflammatory mediators, such as hs‐CRP. 43 An inflammatory stress may also be generated when SUA enters cells due to the impact of intracellular SUA on the generation of reactive oxygen species to the reactive nitrogen species and the Cox‐2 activation. 44 Decreased lymphocyte count implies a decline in immune regulation and an increase in stress. 45 Taken together, the above facts strongly supported the premise, suggesting that a higher ULR may serve as a better risk stratification tool for HS.
Another striking finding with the present study is that ULR, as a novel biomarker, is significantly associated with the risk of HS, but not the risk of total stroke and IS. To further explore the potential mechanisms underlying the different associations, mediation analysis was performed, and the results showed that the association between ULR and HS was partially mediated by SBP, DBP, and eGFR, which can explain the different associations. Previous studies demonstrated that IS and HS have distinct risk profiles, hypertension and a level of low eGFR have been reported to be more strongly linked to HS than IS. 46, 47 The main causes for the stronger correlation between hypertension and HS included micro‐aneurysm formed in cerebral arteriole at basal ganglion, weakened structure of external membrane and middle‐layer membrane of cerebral arterial wall, spasm of cerebral arteriole induced and fibrinoid necrosis of cerebral arteriole, high pulsating blood flow into the ruptured arteries in the brain can cause cerebral microvascular damage, and further lead to the occurrence of hemorrhage. 48, 49 Like the mechanisms of hypertension, low eGFR may induce platelet dysfunction leading to prolonged bleeding time and increasing risk of cerebral hemorrhages, or be correlated with cerebral small vessel disease, the mechanism behind most brain hemorrhages. 46 Elevated ULR was associated with a higher level of SBP, DBP, and a lower level of eGFR, these factors contributed more to the pathophysiology of HS than IS.
Our study has several strengths. First, we investigated a novel biomarker, ULR with the risk of stroke, and we compared the incremental predictive value of SUA, lymphocyte, and ULR to. Furthermore, we quantified the contribution of other risk factors in the pathways between ULR and stroke to explore the potential mechanisms. However, several limitations should also be noted. First, some information was not available in the current study, such as urate‐lowering agents and reperfusion therapy, which could be considered in future investigations. Second, we separately assessed the mediating effects of an indicator of obesity, blood pressure, blood lipids, blood glucose, and an indicator of inflammation on the associations of ULR with stroke. However, because of the complexity associated with the numerous permutations of mediators, it may not be feasible to mutually consider the combined mediating and interactive effects of these mediators. Third, our study recruited much lower number of women than men. However, subgroup analysis stratified by sex showed that the associations were consistent across women and men. Fourth, selection bias may exist because participants with missing data on SUA or lymphocyte count were excluded. Finally, the study was conducted among Chinese population, thus the findings could not be generalized to other ethnicities. However, the constitution of the population was complex by consisting of individuals from all levels of society and across various occupations. Study of such a geographically confined and controlled population greatly reduces the residual confounding factors because of diverse socioeconomic factors and lifestyle patterns.
## CONCLUSIONS
In conclusion, this explorative study found that ULR, serving as a novel inflammatory biomarker, was significantly associated with the risk of HS in Chinese adults, indicating that ULR can be employed as a simplified, effective, and routine risk stratification tool to provide readily objective information for the risk of HS. Additionally, the association between ULR and HS was partially mediated by SBP, DBP, and eGFR. These findings emphasized the important roles of ULR and metabolic factors as conjunctive intervention targets in the prevention of HS.
## AUTHOR CONTRIBUTIONS
YL, SW, and AW contributed to the conception and design of the study; XT contributed to manuscript drafting; XT, SC, YZ, XZ, and QX contributed to the statistics analysis; PW contributed to the acquisition of data; all authors contributed to critical revisions of the manuscript.
## FUNDING INFORMATION
This work was supported by National Key Research and Development Program of China (2022YFC3600600, 2022YFC3600603, 2018YFC1312400, and 2018YFC1312402), Training Fund for Open Projects at Clinical Institutes and Departments of Capital Medical University (CCMU2022ZKYXZ009), Beijing Natural Science Foundation Haidian original innovation joint fund (L222123), and Beijing Municipal Administration of Hospitals Incubating Program (PX2020021). The funder has no role in study design, data collection, data analysis, manuscript preparation, and/or publication decisions.
## CONFLICT OF INTEREST
None declared.
## DATA AVAILABILITY STATEMENT
Data are available to researchers upon request for purposes of reproducing the results or replicating the procedure by directly contacting the corresponding author.
## References
1. Feigin VL, Nguyen G, Cercy K. **Global, regional, and country‐specific lifetime risks of stroke, 1990 and 2016**. *N Engl J Med* (2018) **379** 2429-2437. PMID: 30575491
2. Vidale S, Consoli A, Arnaboldi M, Consoli D. **Postischemic inflammation in acute stroke**. *J Clin Neurol* (2017) **13** 1-9. PMID: 28079313
3. Libby P, Simon DI. **Inflammation and thrombosis: the clot thickens**. *Circulation* (2001) **103** 1718-1720. PMID: 11282900
4. Marnane M, Prendeville S, McDonnell C. **Plaque inflammation and unstable morphology are associated with early stroke recurrence in symptomatic carotid stenosis**. *Stroke* (2014) **45** 801-806. PMID: 24481971
5. Libby P. **Inflammation in atherosclerosis**. *Nature* (2002) **420** 868-874. PMID: 12490960
6. Philips T, Robberecht W. **Neuroinflammation in amyotrophic lateral sclerosis: role of glial activation in motor neuron disease**. *Lancet Neurol* (2011) **10** 253-263. PMID: 21349440
7. Yang X, Gu J, Lv H. **Uric acid induced inflammatory responses in endothelial cells via up‐regulating(pro)renin receptor**. *Biomed Pharmacother* (2019) **109** 1163-1170. PMID: 30551366
8. Gersch C, Palii SP, Imaram W. **Reactions of peroxynitrite with uric acid: formation of reactive intermediates, alkylated products and triuret, and in vivo production of triuret under conditions of oxidative stress**. *Nucleosides Nucleotides Nucleic Acids* (2009) **28** 118-149. PMID: 19219741
9. Lippi G, Montagnana M, Franchini M, Favaloro EJ, Targher G. **The paradoxical relationship between serum uric acid and cardiovascular disease**. *Clin Chim Acta* (2008) **392** 1-7. PMID: 18348869
10. Abeles AM. **Hyperuricemia, gout, and cardiovascular disease: an update**. *Curr Rheumatol Rep* (2015) **17** 13. PMID: 25740704
11. Imtiaz F, Shafique K, Mirza SS, Ayoob Z, Vart P, Rao S. **Neutrophil lymphocyte ratio as a measure of systemic inflammation in prevalent chronic diseases in Asian population**. *Int Arch Med* (2012) **5** 2. PMID: 22281066
12. Papa A, Emdin M, Passino C, Michelassi C, Battaglia D, Cocci F. **Predictive value of elevated neutrophil‐lymphocyte ratio on cardiac mortality in patients with stable coronary artery disease**. *Clin Chim Acta* (2008) **395** 27-31. PMID: 18498767
13. Sharma D, Spring KJ, Bhaskar SMM. **Role of neutrophil‐lymphocyte ratio in the prognosis of acute Ischaemic stroke after reperfusion therapy: a systematic review and meta‐analysis**. *J Cent Nerv Syst Dis* (2022) **14** 11795735221092518. PMID: 35492740
14. Sharma D, Spring KJ, Bhaskar SMM. **Neutrophil‐lymphocyte ratio in acute ischemic stroke: immunopathology, management, and prognosis**. *Acta Neurol Scand* (2021) **144** 486-499. PMID: 34190348
15. Uthamalingam S, Patvardhan EA, Subramanian S. **Utility of the neutrophil to lymphocyte ratio in predicting long‐term outcomes in acute decompensated heart failure**. *Am J Cardiol* (2011) **107** 433-438. PMID: 21257011
16. Li M, Hou W, Zhang X, Hu L, Tang Z. **Hyperuricemia and risk of stroke: a systematic review and meta‐analysis of prospective studies**. *Atherosclerosis* (2014) **232** 265-270. PMID: 24468137
17. Li J, Muraki I, Imano H. **Serum uric acid and risk of stroke and its types: the circulatory risk in communities study (CIRCS)**. *Hypertens Res* (2020) **43** 313-321. PMID: 31988479
18. Yang Z, Li S, Zhao L. **Serum uric acid to lymphocyte ratio: a novel prognostic biomarker for surgically resected early‐stage lung cancer. A propensity score matching analysis**. *Clin Chim Acta* (2020) **503** 35-44. PMID: 31926813
19. Wei XB, Chen WJ, Duan CY. **Joint effects of uric acid and lymphocyte count on adverse outcomes in elderly patients with rheumatic heart disease undergoing valve replacement surgery**. *J Thorac Cardiovasc Surg* (2019) **158** 420-7.e1. PMID: 30459109
20. Wang C, Yuan Y, Zheng M. **Association of age of onset of hypertension with cardiovascular diseases and mortality**. *J Am Coll Cardiol* (2020) **75** 2921-2930. PMID: 32527401
21. Zhao M, Song L, Sun L. **Associations of type 2 diabetes onset age with cardiovascular disease and mortality: the Kailuan study**. *Diabetes Care* (2021) **44** 1426-1432. PMID: 35239970
22. **Recommendations on stroke prevention, diagnosis, and therapy. Report of the WHO task force on stroke and other cerebrovascular disorders**. *Stroke* (1989) **20** 1407-1431. PMID: 2799873
23. Levey AS, Stevens LA, Schmid CH. **A new equation to estimate glomerular filtration rate**. *Ann Intern Med* (2009) **150** 604-612. PMID: 19414839
24. Valeri L, Vanderweele TJ. **Mediation analysis allowing for exposure‐mediator interactions and causal interpretation: theoretical assumptions and implementation with SAS and SPSS macros**. *Psychol Methods* (2013) **18** 137-150. PMID: 23379553
25. Nowlin S, Cleland C, Parekh N, Hagan H, Melkus G. **Racial and ethnic disparities in predictors of glycemia: a moderated mediation analysis of inflammation‐related predictors of diabetes in the NHANES 2007‐2010**. *Nutr Diabetes* (2018) **8** 56. PMID: 30348948
26. Li Y, Zhang T, Han T. **Impact of cigarette smoking on the relationship between body mass index and insulin: longitudinal observation from the Bogalusa heart study**. *Diabetes Obes Metab* (2018) **20** 1578-1584. PMID: 29446554
27. Zhu Y, Hedderson MM, Quesenberry CP, Feng J, Ferrara A. **Central obesity increases the risk of gestational diabetes partially through increasing insulin resistance**. *Obesity (Silver Spring)* (2019) **27** 152-160. PMID: 30461219
28. Pan WC, Wu CD, Chen MJ. **Fine particle pollution, alanine transaminase, and liver cancer: a Taiwanese prospective cohort study (REVEAL‐HBV)**. *J Natl Cancer Inst* (2016) **108**. DOI: 10.1093/jnci/djv341
29. Xie Y, Guo R, Li Z. **Temporal relationship between body mass index and triglyceride‐glucose index and its impact on the incident of hypertension**. *Nutr Metab Cardiovasc Dis* (2019) **29** 1220-1229. PMID: 31383505
30. Xie W, Li P. **Visualizing regulatory lymphocytic responses to predict neurological outcome after stroke**. *CNS Neurosci Ther* (2021) **27** 867-868. PMID: 34156147
31. Li JW, Xu YY, Chen YJ. **Early elevated neutrophil‐to‐lymphocyte ratio associated with remote diffusion‐weighted imaging lesions in acute intracerebral hemorrhage**. *CNS Neurosci Ther* (2020) **26** 430-437. PMID: 31651093
32. Zimmermann L, Pham M, März AG, Kollikowski AM, Stoll G, Schuhmann MK. **Defining cerebral leukocyte populations in local ischemic blood samples from patients with hyperacute stroke**. *J Cereb Blood Flow Metab* (2022) **42** 901-904. PMID: 35107055
33. Pedragosa J, Miró‐Mur F, Otxoa‐de‐Amezaga A. **CCR2 deficiency in monocytes impairs angiogenesis and functional recovery after ischemic stroke in mice**. *J Cereb Blood Flow Metab* (2020) **40** S98-s116. PMID: 32151226
34. Zhang Y, Li F, Chen C. **RAGE‐mediated T cell metabolic reprogramming shapes T cell inflammatory response after stroke**. *J Cereb Blood Flow Metab* (2022) **42** 952-965. PMID: 34910890
35. Mannarino MR, Pirro M, Gigante B. **Association between uric acid, carotid intima‐media thickness, and cardiovascular events: prospective results from the IMPROVE study**. *J Am Heart Assoc* (2021) **10**. PMID: 33998285
36. Ae R, Kanbay M, Kuwabara M. **The causality between the serum uric acid level and stroke**. *Hypertens Res* (2020) **43** 354-356. PMID: 31988480
37. Li P, Zhang L, Zhang M, Zhou C, Lin N. **Uric acid enhances PKC‐dependent eNOS phosphorylation and mediates cellular ER stress: a mechanism for uric acid‐induced endothelial dysfunction**. *Int J Mol Med* (2016) **37** 989-997. PMID: 26935704
38. Zhu HY, Zhao SZ, Zhang ML. **Elevated serum uric acid increases the risk of ischemic stroke recurrence and its inflammatory mechanism in older adults**. *Front Aging Neurosci* (2022) **14** 822350. PMID: 35350634
39. Kueider AM, An Y, Tanaka T. **Sex‐dependent associations of serum uric acid with brain function during aging**. *J Alzheimers Dis* (2017) **60** 699-706. PMID: 28922153
40. Goldstein LB, Adams R, Alberts MJ. **Primary prevention of ischemic stroke: a guideline from the American Heart Association/American Stroke Association stroke council: cosponsored by the atherosclerotic peripheral vascular disease interdisciplinary working group; cardiovascular nursing council; clinical cardiology council; nutrition, physical activity, and metabolism council; and the quality of care and outcomes research interdisciplinary working group: the American Academy of Neurology affirms the value of this guideline**. *Stroke* (2006) **37** 1583-1633. PMID: 16675728
41. Chen Z, Mo J, Xu J. **Risk profile of ischemic stroke caused by small‐artery occlusion vs**. *Deep Intracerebral Hemorrhage Front Neurol* (2019) **10** 1213. PMID: 31827458
42. Juli C, Heryaman H, Nazir A. **The lymphocyte depletion in patients with acute ischemic stroke associated with poor neurologic outcome**. *Int J Gen Med* (2021) **14** 1843-1851. PMID: 34017192
43. Chen YF, Li Q, Chen DT. **Prognostic value of pre‐operative serum uric acid levels in esophageal squamous cell carcinoma patients who undergo R0 esophagectomy**. *Cancer Biomark* (2016) **17** 89-96. PMID: 27314297
44. Fini MA, Monks J, Farabaugh SM, Wright RM. **Contribution of xanthine oxidoreductase to mammary epithelial and breast cancer cell differentiation in part modulates inhibitor of differentiation‐1**. *Mol Cancer Res* (2011) **9** 1242-1254. PMID: 21775420
45. Liu Q, Gao X, Xiao Q. **A combination of NLR and sST2 is associated with adverse cardiovascular events in patients with myocardial injury induced by moderate to severe acute carbon monoxide poisoning**. *Clin Cardiol* (2021) **44** 401-406. PMID: 33496356
46. Hatleberg CI, Ryom L, Kamara D. **Predictors of ischemic and hemorrhagic strokes among people living with HIV: the D:a:D international prospective multicohort study**. *EClinicalMedicine* (2019) **13** 91-100. PMID: 31517266
47. Tsai CF, Anderson N, Thomas B, Sudlow CL. **Comparing risk factor profiles between intracerebral hemorrhage and ischemic stroke in Chinese and white populations: systematic review and meta‐analysis**. *PLoS One* (2016) **11**. PMID: 26991497
48. Yang G, Shao G. **Clinical effect of minimally invasive intracranial hematoma in treating hypertensive cerebral hemorrhage**. *Pak J Med Sci* (2016) **32** 677-681. PMID: 27375713
49. Woo D, Haverbusch M, Sekar P. **Effect of untreated hypertension on hemorrhagic stroke**. *Stroke* (2004) **35** 1703-1708. PMID: 15155969
|
---
title: The independent impact of dementia in patients undergoing percutaneous coronary
intervention for acute myocardial infarction
authors:
- Afek Kodesh
- Tamir Bental
- Hana Vaknin‐Assa
- Yeela Talmor‐Barkan
- Pablo Codner
- Amos Levi
- Ran Kornowski
- Leor Perl
journal: Clinical Cardiology
year: 2023
pmcid: PMC10018096
doi: 10.1002/clc.23967
license: CC BY 4.0
---
# The independent impact of dementia in patients undergoing percutaneous coronary intervention for acute myocardial infarction
## Abstract
### Background
Although age and frailty are associated with worse prognoses for patients who undergo percutaneous coronary intervention (PCI), little is known regarding the independent impact of dementia.
### Hypothesis
The aim of this study was to evaluate the association between dementia and outcomes for patients with acute myocardial infarction (AMI).
### Methods
Consecutive patients with ST‐elevation or non‐ST elevation MI who had undergone PCI as part of our AMI registry were included in this study. We compared outcomes within the 1‐year period of their PCI, including death and major adverse cardiac events (MACE) and corrected for confounders using Cox regression.
### Results
Of 28 274 patients, 9167 patients who had undergone PCI for AMI were included in this study, 250 with dementia; Mean age (77.4 ± 9.4 in the dementia group vs. 63.6 ± 12.7 in the control), female gender (32.4 vs. $24.2\%$, $$p \leq .003$$), diabetes mellitus (54.0 vs. $42.4\%$, $p \leq .001$) and chronic kidney disease (44.4 vs. $19.3\%$, $p \leq .001$) were higher. At 12 months, unadjusted rates of death (25.5 vs. $9.8\%$, $p \leq .001$) and MACE (33.8 vs. $17.6\%$, $p \leq .001$) were higher for patients with dementia. After standardizing for confounding variables, dementia remained an independent risk factor for death (HR 1.90; CI 1.37–2.65; $p \leq .001$) and MACE (HR 1.73; CI 1.30–2.31; $p \leq .001$), as well as in propensity score matched analysis (HR 1.54; CI: 1.03–2.28; $p \leq .001$ and HR 1.49; CI: 1.09–2.02; $p \leq .001$, respectively).
### Conclusions
Dementia is an independent predictor of worse outcomes in patients undergoing PCI for AMI. Future intervention and specialized healthcare measures to mitigate this risk is warranted.
## INTRODUCTION
The rate of acute myocardial infarction (AMI) is a growing concern in the elderly population. The global prevalence of dementia more than doubled from 1990 to 2016 and is expected to increase from an estimated 57.4 million cases in 2019 to an estimated 152.8 million cases in 2050. 1, 2 Additionally, the rates of ischemic heart disease (IHD) are expected to increase from 1655 to 1845 cases per 100 000 people in 2030. 3 The elderly populations continues to grow as well, with predictions assuming that by 2050, 1 in 6 individuals will be over the age of 65, compared to 1 in 11 in 2019. 4 As the elderly population increases, the incidence of IHD and AMI in this cohort will rise as well. 3 In patients with acute myocardial infarction (AMI), dementia has an impact on whether patients receive percutaneous coronary intervention (PCI). Several studies have demonstrated that patients who have dementia are considerably less likely to receive PCI. 5, 6, 7, 8 This has been hypothesized to be due to a myriad of factors, including the preference for a higher quality of life over an attempt to decrease mortality by invasive measures in this patient population. Another explanation might stem from the vulnerability to side effects or drug–drug interactions of patients suffering from dementia. However, after addressing and standardizing for these clinical measures, patients with dementia still received lower rates of interventional procedures. 7, 8 Nevertheless, elderly patients, specifically octogenarians, have experienced similar benefits in ischemic complications when treated by PCI for AMI, when compared to younger patients. 9 *It is* therefore becoming increasingly important to address outcomes in this specific growing population of patients with dementia who experience an AMI and are treated by PCI. In this study, based on a large registry of consecutive PCI patients, we examined the rates of 1‐year mortality and major adverse cardiac events (MACE) following the procedure in patients suffering from dementia and AMI.
## Study design
The present study is based on a prospectively collected PCI registry from the Rabin Medical Center in Petach Tikva—Israel, which includes 2 campuses—Beilinson and Hasharon hospitals. The registry includes consecutive patients treated with PCI from January 2004 through December 2020. The data is continuously entered into an ongoing registry for purposes of recording and monitoring patient‐related parameters, clinical events, and angiographic findings.
Of the 28 274 patients in the registry, we included only patients who were treated for ST‐elevation (STEMI) or non‐ST‐elevation myocardial infarction (NSTEMI) (Figure 1). Myocardial infarction was defined as detection of myocardial injury along with a clinical presentation of myocardial infarction, which may include symptoms such as chest or epigastric discomfort during exertion or rest. 10 STEMI and NSTEMI differed from each other by extent of myocardial ischemia, evidenced by ECG findings. Patients were excluded if they presented with stable or unstable angina, if they were treated with thrombolysis instead of PCI (<$1\%$ of cases), in cases of periprocedural infarction or type‐2 myocardial infarction, or if they were ineligible for stent placement. We then collected 1‐year outcomes of patients with or without dementia. Dementia was defined by any of the following terms: “dementia” or “cognitive dysfunction” or “cognitive decline” or “cognitive impairment” or “Alzheimer's disease.” 11 The study protocol and data collection was approved by the local Institutional Review Board.
**Figure 1:** *Patients included in study. Methodology for patient inclusion. Of the 28 274 patients who received PCI for AMI, only those who had STEMI or NSTEMI were included. Those with stable or unstable angina, or with periprocedural or type II MI were excluded. AMI, acute myocardial infarction; NSTEMI, non‐ST‐elevation myocardial infarction; PCI, percutaneous coronary intervention; STEMI, ST‐elevation myocardial infarction.*
## Interventional procedure
All patients provided explicit written informed consent to undergo cardiac catheterization. Precatheterization treatment consisted of aspirin and unfractionated heparin (70 U/kg). Clopidogrel 300 or 600 mg, prasugrel 60 mg, or ticagrelor 180 mg was administered as a loading dose before or immediately after PCI. The utilization of glycoprotein IIb/IIIa inhibitors (GP2b3a) and choice of stent, as well as other therapeutic modalities such as mechanical thrombectomy and distal protection devices, were left to the discretion of the primary operator. All stents were implanted with moderate‐to‐high deployment pressure (12 to 16 atm). All patients received dual antiplatelet therapy with aspirin 100 mg daily and a thienopyridine (clopidogrel, prasugrel, or ticagrelor) for at least 12 months after PCI unless bleeding events caused premature cessation of dual antithrombotic treatment.
## Endpoints
Immediate and in‐hospital events were prospectively collected in the institutional database. During follow‐up, patients completed standardized questionnaires for clinical events either by telephone (e.g., with the patient or with a family member) or in the outpatient clinics at 6‐month intervals. When indicated, records from peripheral hospitals were acquired to verify the events in the follow‐up period. All events were further confirmed and adjudicated by the institutional clinical events adjudication committee. Survival status at follow‐up was assessed by review of municipal civil registries at 1 year. Clinical outcomes included all‐cause mortality and MACE, which comprised death, MI, target vessel revascularization, and subsequent coronary artery bypass graft surgery. Renal failure was defined as glomerular filtration rate below 50 ml/min/1.73 m2 (according to the Modification of Diet in Renal Disease formula), anemia was defined as hemoglobin levels lower than 13.0 g/dl for men and 12.0 g/dl for women. Findings were compared between patients with dementia and those without.
## Statistical analysis
Continuous data are summarized as mean and SD or median and interquartile range and were compared using Student t‐tests or analyses of variance. Categorical variables are presented as frequency and were compared by χ 2 or Fisher's exact tests. The normality of variable distributions was assessed using the Kolmogorov–Smirnov test. Time‐to‐event curves were constructed using the Kaplan–Meier method and compared using log‐rank test. Cox regression analyses were performed to identify independent predictors of the primary endpoint. Covariates for the Cox model were chosen according to their known association with dementia and outcomes, and included age, sex, diabetes mellitus, renal failure, peripheral artery disease, left ventricular ejection fraction (for each $1\%$ increase), previous oncological disease, ST‐elevation myocardial infarction, trans radial access and dementia. Finally, due to several differences in baseline characteristics, we compiled a cohort of propensity score matched patients with a 1:1 ratio between patients with dementia and controls. The propensity score was derived from a multivariate logistic regression model that included dementia, considered as the independent (outcome) variable, and all baseline clinical characteristics and procedural characteristics as covariates. The propensity score matched cohort was analyzed for the main combined outcome. Effect sizes are presented as odds ratios and $95\%$ confidence intervals. All statistical analyses were performed with IBM SPSS statistics V.28 software. A $p \leq .05$ was considered statistically significant.
## RESULTS
Our study consisted of 9167 patients, of which 8917 did not suffer from dementia and 250 did. Mean age was 77.4 ± 9.4 and 63.6 ± 12.7 for patients with and without dementia, respectively ($p \leq .001$). $32.4\%$ of the patients with dementia were female compared to $24.2\%$ for those without ($$p \leq .003$$). Other baseline characteristics, including the prevalence of previous coronary artery bypass graft, history of atrial fibrillation and prior peripheral vascular disease—did not differ between the groups (Table 1).
**Table 1**
| Parameter | Control (n = 8917) | Dementia (n = 250) | p Value |
| --- | --- | --- | --- |
| Age | 63.6 ± 12.7 | 77.4 ± 9.4 | <.001 |
| Female sex (%) | 24.2 | 32.4 | .003 |
| Diabetes mellitus (%) | 42.4 | 54.0 | <.001 |
| Hypertension (%) | 69.2 | 84.0 | <.001 |
| Prior smoking (%) | 46.2 | 17.2 | <.001 |
| Prior CHF (%) | 26.4 | 44.4 | <.001 |
| Prior COPD (%) | 8.2 | 16.8 | <.001 |
| Prior PVD (%) | 5.6 | 4.8 | .569 |
| Atrial fibrillation (%) | 7.2 | 8.4 | .452 |
| Stroke (%) | 6.7 | 24.8 | <.001 |
| Prior malignancy (%) | 10.0 | 20.8 | <.001 |
| CABG (%) | 8.2 | 10.8 | .141 |
| CHA2DS2VASc | 3.4 | 5.3 | <.001 |
| CKD (%) | 19.3 | 44.4 | <.001 |
Procedural characteristics are shown in Table 2. Importantly, there was no difference in rates of STEMI as opposed to NSTEMI between the two groups ($$p \leq .822$$). Patients with dementia were more likely to present in state of cardiogenic shock (4.8 vs. $2.5\%$, $$p \leq .026$$) and have a greater number of vessels involved (2.4 ± 0.7 vs. 2.2 ± 0.8, p ≤.001) as part of the acute presentation of AMI.
**Table 2**
| Parameter | Control (n = 8917) | Dementia (n = 250) | p Value |
| --- | --- | --- | --- |
| EF (%) | 52.8 | 49.6 | .001 |
| Unprotected LMCA (%) | 3.0 | 7.6 | <.001 |
| Shock (%) | 2.5 | 4.8 | .026 |
| No. of vessels involved | 2.2 ± 0.8 | 2.4 ± 0.7 | <.001 |
| No. of territories | 1.6 ± 0.8 | 1.7 ± 0.9 | .043 |
| STEMI (%) | 31.9 | 31.2 | .822 |
| Radial approach (%) | 46.1 | 33.2 | <.001 |
| Drug‐eluting stent (%) | 90.9 | 91.0 | .215 |
| Symptoms to admission (hours) | 3.81 ± 1.48 | 4.00 ± 1.51 | .022 |
| Admission to PCI, STEMI patients (hours) | 0.92 ± 0.39 | 1.02 ± 0.42 | .095 |
| Hemoglobin A1C (%) | 7.0 ± 1.8 | 7.2 ± 2.0 | .443 |
| Hemoglobin (g/dl) | 13.6 ± 1.9 | 12.4 ± 2.0 | <.001 |
| Platelet count (×103 mm3) | 244.1 ± 79.5 | 249.7 ± 85.4 | .277 |
| Total cholesterol (mg/dl) | 173.4 ± 46.1 | 158.9 ± 42.0 | <.001 |
| HDL (mg/dl) | 40.3 ± 11.8 | 43.0 ± 12.8 | .004 |
| LDL (mg/dl) | 102.5 ± 39.2 | 89.7 ± 33.0 | <.001 |
The unadjusted cumulative probability of reaching the endpoint of death for patients with dementia at a 1‐year follow‐up period was 25.5 versus $9.8\%$ for the control ($p \leq .001$). Comparatively, the MACE endpoint was reached 33.8 versus $17.6\%$ for the control ($p \leq .001$). Kaplan–Meier curves demonstrating these unadjusted risks are shown in in Figures 2 and 3. Following Cox regression analysis, adjusting for differences in baseline characteristics, patients with dementia were 1.73 ($95\%$ CI 1.30–2.31; $p \leq .001$) and 1.90 ($95\%$ CI 1.37–2.65; $p \leq .001$) times more likely to suffer MACE and death, respectively, 1‐year after PCI (Supporting Information: Tables 1 and 2, Figures 1 and 2).
**Figure 2:** *Kaplan–Meier Curves—Risk of death. Unadjusted rates of death dependent on patient's dementia status. In a 1‐year follow‐up period post‐PCI, patients with dementia had a 25.5% chance of death, compared to 9.8% for those without dementia.* **Figure 3:** *Kaplan–Meier Curves—Risk of MACE. Unadjusted rates of MACE dependent on patient's dementia status. In a 1‐year follow‐up period post‐PCI, patients with dementia had a 33.8% chance to encounter an event of MACE, compared to 17.6% for those without dementia. MACE, major adverse cardiac events; PCI, percutaneous coronary intervention.*
The propensity match score was able to form 171 matched pairs of patients with dementia and control patients, showing similar results. After propensity matched score analysis, mean age was 76.6 ± 9.2 and 76.4 ± 9.3 for patients with and without dementia, respectively. The cohort of patients with dementia were $32.3\%$ female, $53.9\%$ had diabetes, and $83.7\%$ had hypertension. Those without dementia were $32.1\%$ female, $53.8\%$ had diabetes, and $83.5\%$ had hypertension. Following Cox regression, patients presenting with dementia demonstrated higher rates of MACE (HR 1.49; CI: 1.09–2.02; $p \leq .001$) and death (HR 1.54; CI: 1.03–2.28; $p \leq .001$) than patients without dementia. Additional remaining baseline characteristics postpropensity matching as well as the KM curves demonstrating these results are included in Supporting Information: Table 3, Figures 3 and 4.
## DISCUSSION
The current study demonstrates a significant independent association in the rate of adverse outcomes for patients who have dementia following PCI for AMI. The study shows that patients with dementia are more than two times more likely to suffer death and nearly two times more likely to encounter MACE 1‐year after their PCI. Also, after correcting for confounders, dementia remains an independent risk for adverse events.
The incidence of IHD and dementia is closely tied. Those with a history of IHD are, on average, at a $45\%$ increased risk of developing cognitive impairment. 12 The direct effect of large artery atherosclerosis is one of the mechanisms responsible for this shared correspondence, as it is a significant component for the development of vascular dementia. In addition to this direct influence, IHD has been shown to be correlated with increased senile plaque formation and reduced hippocampal size. 13, 14, 15 Apart from the aforementioned correlations between IHD and dementia, both diseases share common risk factors, including obesity, type II diabetes mellitus, and hypercholesterolemia, among others. 12, 13, 16 Nevertheless, when adjusting for these various cofounders, the association between cardiovascular disease and cognitive decline is still prominent. 17 Several studies have examined the association between these two pathologies, but very few have examined the implications of cognitive impairment on the prognosis of patients with IHD who undergo PCI. While not equivalent to dementia, frailty is a marker that has been used to explore this topic.
In a study that assessed patients undergoing PCI for frail patients with stable angina or acute coronary syndrome, both mortality and length of hospital stay increased: 30‐day mortality was $4.9\%$ versus $1.1\%$ for nonfrail patients ($$p \leq .01$$) and length of hospital stay was 2.9 ± 5.6 versus 1.7 ± 3.1 days for nonfrail patients ($p \leq .001$). 18 Similarly, in a cohort of 62 patients who presented with STEMI, and after adjustment for common confounders, including BMI and troponin levels, higher frailty scores were associated with increased in‐hospital mortality and failure of discharge to home. Higher frailty scores were 6.28 times more likely to suffer in‐hospital mortality and 16.69 times more likely to not be discharged home. 19 In these studies, frailty was defined using the Canadian Study of Health and Aging—Clinical Frailty Score, which has been demonstrated to be correlated with cognitive impairment. 20 Hamonangan and colleagues defined frailty using the frailty phenotype criteria, which has also been demonstrated to be highly associated with cognitive status. 21 However, in this study that only looked at MACE for elderly patients who had undergone PCI for coronary artery disease, no significant association was found for frail patients who suffered MACE after PCI: $8.19\%$ of cases versus $5.12\%$ of cases for nonfrail patients ($p \leq .05$). 22 *While a* patient's frailty status is not equivocal to a patient's clinical dementia status, higher mortality rates, worse long‐term prognoses, and longer hospital stays were all associated with higher frailty scores. 18, 23, 24 Most of the studies investigating the association between higher levels of clinical frailty and worse outcomes for patients with acute coronary syndrome or CAD have found it to be significant. 18, 19, 23, 24, 25 However, it is also important to study the independent effects of cognitive dysfunction on outcomes in this cohort. To that extent, this study is the largest and first of its kind to look at the independent effects of dementia on outcomes for patients with STEMI or NSTEMI who undergo PCI.
Other independent associations that have been demonstrated to lead to higher‐all cause mortality in patients with AMI include older age, 26 lower hemoglobin levels, 27 and progressive kidney dysfunction. 28 Even in cases of therapy with PCI, complication rate is highly correlated with age. Older patients tend to have higher rates of in‐hospital death, restenosis, and postprocedural complications, as compared to younger patients. 29, 30 Nevertheless, PCI remains the preferred treatment over medical therapy for this aging population suffering from AMI. 29, 31 Similarly, lower levels of hemoglobin and progressive kidney dysfunction presume a worse prognosis. Our data demonstrated that the cohort of patients with dementia did have statistically significant differences in age, hemoglobin levels, and rates of chronic kidney disease. However, even after adjusting for these major factors, patients with dementia were 1.7 times more likely to suffer from MACE and 1.9 times more likely to die at the 1‐year interval.
An explanation for the increase in MACE and death for this elderly population could stem from a delay from initial symptoms to admission in the hospital. In our analysis, patients with dementia were admitted 4.00 ± 1.51 h after initial symptoms compared to 3.81 ± 1.48 h for patients without dementia ($p \leq .05$). Patients who suffer from dementia may not be as aware of their symptoms and requirement for medical attention as their controls. This decreased awareness may lead to delays in hospital arrivals and increase in the percentage of ischemic burden upon presentation. Even with PCI performed, the prognosis in this situation would be less favorable. In our study, there were no differences in door‐to‐balloon times in this combined STEMI/NSTEMI cohort, but there was a significant difference in patient delay before admission. Additionally, patients with cognitive impairment are also more likely to suffer from other comorbidities. The worse overall health that these patients initially present with may confer a worse overall outcome, even after PCI.
With a growing population of elderly patients and patients with dementia, along with a projected increase in cases of IHD, consideration should be placed on how to improve the prognosis for this population of patients with dementia suffering from AMI. Appropriate surveillance of this population will likely lead to earlier assessment and revascularization for patients with IHD. To increase awareness of symptoms and avoid delays in hospitalization, frequent screening via direct patient visits, as well as telemedicine, can be used. Implementing ECG transmission systems from home has been show to improve triage at the hospital and lead to decreased delays in treatment. 32 When medications and frequent surveillance are not enough to prevent the progression of IHD, the usage of PCI should be more effectively assessed. Compared to younger patients, elderly patients have a 2‐to‐4‐fold increase in the risk of morbidity and mortality. 33 When inappropriately used, PCI can lead to complications such as stent thrombosis and restenosis. 34 The development and increased usage of fractional flow reserve instead of anatomic angiography in the assessment of PCI usage in nonculprit AMI lesions may lead to a better prognosis for this cohort.
## LIMITATIONS
First, this is an observational study, meaning the generalizability of its findings are limited. This also includes the definition of dementia, derived from ICD‐9 and ICD‐10 diagnoses in the patient history. Therefore, correlations can be used on the population observed, but causality cannot be concluded. Second, there was a relatively small number of patients with dementia, which again limits the generalizability of such a study. Third, the diagnosis of dementia in our study is clinical. We do not have objective neuropsychological testing documentation to assess severity. Although it is likely that most patients with the diagnosis have cognitive impairment, it is not confirmed. Finally, our study was absent of a clinical frailty index. Past studies have used some variation of clinical frailty to demonstrate the association between PCI outcomes and frailty. The lack of an index limits the generalizability of our study. Nevertheless, our use of dementia as an associative factor provides a novel perspective on the outcomes for PCI in this elderly cohort and is the largest thus far to report of outcomes in this important and gradually growing patient population suffering from AMI.
## CONCLUSION
Our study demonstrates worse outcomes for patients with dementia treated by PCI for AMI than those without dementia. These findings provide clinicians with a better understanding of the prognosis of patients with dementia and emphasize the need for novel methods of specialized healthcare delivery to mitigate this added risk.
## CONFLICT OF INTEREST
The authors declare no conflict of interest.
## DATA AVAILABILITY STATEMENT
Original research data will be available upon request.
## References
1. Nichols E, Szoeke C, Vollset SE. **Global, regional, and national burden of Alzheimer's disease and other dementias, 1990‐2016: a systematic analysis for the Global Burden of Disease Study 2016**. *Lancet Neurol* (2019) **18** 88-106. PMID: 30497964
2. Nichols E, Vos T. **The estimation of the global prevalence of dementia from 1990‐2019 and forecasted prevalence through 2050: an analysis for the Global Burden of Disease (GBD) study 2019**. *Alzheimer's Dement* (2021) **17**
3. Khan MA, Hashim MJ, Mustafa H. **Global epidemiology of ischemic heart disease: results from the Global Burden of Disease Study**. *Cureus* (2020) **12**. PMID: 32742886
4. 4
UN
. World population prospects: highlights. 2019. https://population.un.org/wpp/Publications/Files/WPP2019_10KeyFindings.pdf. (2019)
5. Kimata T, Hirakawa Y, Uemura K, Kuzuya M. **Absence of outcome difference in elderly patients with and without dementia after acute myocardial infarction**. *Int Heart J* (2008) **49** 533-543. PMID: 18971565
6. Lin C‐F, Wu F‐LL, Lin S‐W. **Age, dementia and care patterns after admission for acute coronary syndrome**. *Drugs Aging* (2012) **29** 819-828. PMID: 23018581
7. Tehrani DM, Darki L, Erande A, Malik S. **In‐hospital mortality and coronary procedure use for individuals with dementia with acute myocardial infarction in the United States**. *J Am Geriatr Soc* (2013) **61** 1932-1936. PMID: 24219195
8. Sloan FA, Trogdon JG, Curtis LH, Schulman KA. **The effect of dementia on outcomes and process of care for medicare beneficiaries admitted with acute myocardial infarction**. *J Am Geriatr Soc* (2004) **52** 173-181. PMID: 14728624
9. Perl L, Franzé A, D'Ascenzo F. **Elderly suffering from ST‐segment elevation myocardial Infarction—results from a database analysis from two Mediterranean Medical Centers**. *J Clin Med* (2021) **10** 2435. PMID: 34070865
10. Thygesen K, Alpert JS, Jaffe AS. **Fourth universal definition of myocardial infarction (2018)**. *Eur Heart J* (2019) **40** 237-269. PMID: 30165617
11. Lappalainen L, Rajamaki B, Tolppanen A‐M, Hartikainen S. **Coronary artery revascularizations and cognitive decline—a systematic review**. *Curr Probl Cardiol* (2022) **47**. PMID: 34363848
12. Deckers K, Schievink SHJ, Rodriquez MMF. **Coronary heart disease and risk for cognitive impairment or dementia: systematic review and meta‐analysis**. *PLoS One* (2017) **12**. PMID: 28886155
13. Justin BN, Turek M, Hakim AM. **Heart disease as a risk factor for dementia**. *Clin Epidemiol* (2013) **5** 135-145. PMID: 23658499
14. Koschack J, Irle E. **Small hippocampal size in cognitively normal subjects with coronary artery disease**. *Neurobiol Aging* (2005) **26** 865-871. PMID: 15718045
15. Soneira CF, Scott TM. **Severe cardiovascular disease and Alzheimer's disease: senile plaque formation in cortical areas**. *Clin Anat* (1996) **9** 118-127. PMID: 8720786
16. Deckers K, van Boxtel MPJ, Schiepers OJG. **Target risk factors for dementia prevention: a systematic review and Delphi consensus study on the evidence from observational studies**. *Int J Geriatr Psychiatry* (2015) **30** 234-246. PMID: 25504093
17. Haring B, Leng X, Robinson J. **Cardiovascular disease and cognitive decline in postmenopausal women: results from the women's health initiative memory study**. *J Am Heart Assoc* (2013) **2**. PMID: 24351701
18. Murali‐Krishnan R, Iqbal J, Rowe R. **Impact of frailty on outcomes after percutaneous coronary intervention: a prospective cohort study**. *Open Heart* (2015) **2**. PMID: 26380099
19. Sujino Y, Tanno J, Nakano S. **Impact of hypoalbuminemia, frailty, and body mass index on early prognosis in older patients (≥85 years) with ST‐elevation myocardial infarction**. *J Cardiol* (2015) **66** 263-268. PMID: 25547740
20. Rockwood K. **A global clinical measure of fitness and frailty in elderly people**. *Can Med Assoc J* (2005) **173** 489-495. PMID: 16129869
21. Chu NM, Bandeen‐Roche K, Xue Q‐L, Carlson MC, Sharrett AR, Gross AL. **Physical frailty phenotype criteria and their synergistic association on cognitive functioning**. *J Gerontol A* (2021) **76** 1633-1642
22. Hamonangan R, Wijaya IP, Setiati S, Harimurti K. **Impact of frailty on the first 30 days of major cardiac events in elderly patients with coronary artery disease undergoing elective percutaneous coronary intervention**. *Acta Med Indones* (2016) **48** 91-98. PMID: 27550877
23. Tse G, Gong M, Nunez J. **Frailty and mortality outcomes after percutaneous coronary intervention: a systematic review and meta‐analysis**. *J Am Med Dir Assoc* (2017) **1** 1097.e1-1097.e10
24. Sanchis J, Bonanad C, Ruiz V. **Frailty and other geriatric conditions for risk stratification of older patients with acute coronary syndrome**. *Am Heart J* (2014) **168** 784-791.e2. PMID: 25440808
25. Sanchis J, Ruiz V, Bonanad C. **Prognostic value of geriatric conditions beyond age after acute coronary syndrome**. *Mayo Clin Proc* (2017) **92** 934-939. PMID: 28389067
26. Hovanesyan A, Rich MW. **Outcomes of acute myocardial infarction in nonagenarians**. *Am J Cardiol* (2008) **101** 1379-1383. PMID: 18471445
27. Bindra K, Berry C, Rogers J. **Abnormal haemoglobin levels in acute coronary syndromes**. *QJM* (2006) **99** 851-862. PMID: 17121766
28. Yamaguchi J, Kasanuki H, Ishii Y. **Serum creatinine on admission predicts long‐term mortality in acute myocardial infarction patients undergoing successful primary angioplasty**. *Circ J* (2007) **71** 1354-1359. PMID: 17721010
29. Graham MM, Ghali WA, Faris PD, Galbraith PD, Norris CM, Knudtson ML. **Survival after coronary revascularization in the elderly**. *Circulation* (2002) **105** 2378-2384. PMID: 12021224
30. Núñez J, Ruiz V, Bonanad C. **Percutaneous coronary intervention and recurrent hospitalizations in elderly patients with non ST‐segment acute coronary syndrome: the role of frailty**. *Int J Cardiol* (2017) **228** 456-458. PMID: 27870976
31. Chanti‐Ketterl M, Pathak EB, Andel R, Mortimer JA. **Dementia: a barrier to receiving percutaneous coronary intervention for elderly patients with ST‐elevated myocardial infarction**. *Int J Geriatr Psychiatry* (2014) **29** 906-914. PMID: 24523068
32. Ortolani P, Marzocchi A, Marrozzini C. **Usefulness of prehospital triage in patients with cardiogenic shock complicating ST‐elevation myocardial infarction treated with primary percutaneous coronary intervention**. *Am J Cardiol* (2007) **100** 787-792. PMID: 17719321
33. Batchelor WB, Anstrom KJ, Muhlbaier LH. **Contemporary outcome trends in the elderly undergoing percutaneous coronary interventions: results in 7,472 octogenarians**. *JACC* (2000) **36** 723-730. PMID: 10987591
34. Pfisterer M, Brunner‐La Rocca HP, Buser PT. **Late clinical events after clopidogrel discontinuation may limit the benefit of drug‐eluting stents**. *JACC* (2006) **48** 2584-2591. PMID: 17174201
|
---
title: Machine learning‐based prediction of 1‐year mortality in hypertensive patients
undergoing coronary revascularization surgery
authors:
- Amir Hossein Behnoush
- Amirmohammad Khalaji
- Malihe Rezaee
- Shahram Momtahen
- Soheil Mansourian
- Jamshid Bagheri
- Farzad Masoudkabir
- Kaveh Hosseini
journal: Clinical Cardiology
year: 2023
pmcid: PMC10018097
doi: 10.1002/clc.23963
license: CC BY 4.0
---
# Machine learning‐based prediction of 1‐year mortality in hypertensive patients undergoing coronary revascularization surgery
## Abstract
### Background
Machine learning (ML) has shown promising results in all fields of medicine, including preventive cardiology. Hypertensive patients are at higher risk of mortality after coronary artery bypass graft (CABG) surgery; thus, we aimed to design and evaluate five ML models to predict 1‐year mortality among hypertensive patients who underwent CABG.
### Hyothesis
ML algorithms can significantly improve mortality prediction after CABG.
### Methods
Tehran Heart Center's CABG data registry was used to extract several baseline and peri‐procedural characteristics and mortality data. The best features were chosen using random forest (RF) feature selection algorithm. Five ML models were developed to predict 1‐year mortality: logistic regression (LR), RF, artificial neural network (ANN), extreme gradient boosting (XGB), and naïve Bayes (NB). The area under the curve (AUC), sensitivity, and specificity were used to evaluate the models.
### Results
Among the 8,493 hypertensive patients who underwent CABG (mean age of 68.27 ± 9.27 years), 303 died in the first year. Eleven features were selected as the best predictors, among which total ventilation hours and ejection fraction were the leading ones. LR showed the best prediction ability with an AUC of 0.82, while the least AUC was for the NB model (0.79). Among the subgroups, the highest AUC for LR model was for two age range groups (50–59 and 80–89 years), overweight, diabetic, and smoker subgroups of hypertensive patients.
### Conclusions
All ML models had excellent performance in predicting 1‐year mortality among CABG hypertension patients, while LR was the best regarding AUC. These models can help clinicians assess the risk of mortality in specific subgroups at higher risk (such as hypertensive ones).
## INTRODUCTION
Cardiovascular diseases (CVDs) are responsible for approximately 17.9 million deaths annually. 1 Ischemic heart disease (IHD) is the most prevalent CVD in the general population, as $49.2\%$ of CVD deaths are among IHD patients. 2 Revascularization methods, including percutaneous coronary intervention (PCI) and coronary artery bypass grafting (CABG), are the primary therapies in IHD. 3 CABG is one of the most common cardiac surgeries, considered the preferable therapeutic approach in patients with multivessel or left main coronary artery disease (CAD) or in case of left ventricular dysfunction. 4 With the prevalence of one in every three adults in the United States, 5 hypertension is a major modifiable risk factor for CAD irrespective of sex and age. 6 Hypertensive patients tend to have different risk factor patterns from other CABG patients. 7 Moreover, increased postoperative complications, early mortality, and 2‐year mortality have been reported, compared to nonhypertensive patients. 7, 8 This was reported to be an up to $40\%$ increase in perioperative morbidity in hypertensive patients undergoing CABG. 9 Besides traditional risk scores, machine learning (ML)‐developed models are getting attention for outcome prediction after cardiac surgeries. 10 However, there are controversies about the accuracy of ML models compared to risk scores currently being used. 11 Knowing the greater need for mortality prediction in the hypertensive population, we aimed to use and compare different ML methods to predict 1‐year mortality of hypertensive patients after isolated CABG.
## Study design and data collection
We conducted this serial cross‐sectional study based on the Tehran Heart Center CABG registry among hypertensive patients between 2005 and 2015. Hypertension was defined as systolic blood pressure (SBP) ≥ 140 mmHg and/or diastolic blood pressure (DBP) ≥ 90 mmHg following two separate examinations in patients' history and/or taking antihypertensive medications. All the perioperative data of patients were collected and managed by expert nurses in our center. The ethics committee of Tehran Heart Center approved this study (IR.TUMS.THC.1401.023).
## Variables' definition
Baseline characteristics including demographic, preoperative, and intraoperative variables were used as potential predictors. Age, gender, weight, height, and body mass index (BMI) were demographics. Serum hemoglobin (Hb), high‐density lipoprotein cholesterol (HDL‐C), low‐density lipoprotein cholesterol (LDL‐C), total cholesterol, triglycerides (TG), and creatinine, in addition to left ventricular ejection fraction (EF) measured by echocardiogram, diabetes, opium consumption, smoking status, prior myocardial infarction (MI), preoperative heart failure (HF), and chronic obstructive pulmonary disease (COPD) were preoperative variables. Finally, hospitalization parameters and intraoperative variables were total ICU hours, total ventilation hours, and cardiopulmonary pump utilization (on‐pump or off‐pump). All these data were obtained from either past medical records or blood sample measurements during hospitalization episodes and before surgery.
## Main outcome
The study's main outcome was 1‐year mortality post‐CABG, for which we compared different ML‐based prediction models. This outcome included both in‐hospital and after‐discharge mortality events.
## Data cleaning
Exclusion criteria were missing data in addition to out‐of‐range values such as: [1] Hb > 25 g/dl or Hb < 5 g/dl, [2] LDL‐C > 400 mg/dl, [3] TG > 1200 mg/dl or TG < 20 mg/dl, [4] HDL‐C < 5 mg/dl or HDL‐C > 100 mg/dl, and [5] creatinine > 15 mg/dl or creatinine < 0.2 mg/dl were excluded. As we had sufficient data to develop and test models, excluding missing data were possible.
## Test/train split, feature selection, and oversampling
In a random assignment process, the total hypertensive population was divided into train and test cohorts ($70\%$ and $30\%$, respectively). The test cohort sample was used to evaluate and validate the ML models.
To select the best predictors for mortality in the total population and each of the subgroups, a feature selection process based on the random forest (RF) model was designed using 10‐fold cross‐validation. This technique investigates the effect of each predictor alone and in combination with other predictive variables. RF feature selection works based on mean decrease accuracy (MDA) and the mean decrease gini (MDG). The former shows how much accuracy is lost if a variable is excluded, while MDG represents the contribution of each variable to the homogeneity of the nodes and leaves in the resulting RF. The higher these scores, the higher the importance of variable. 12, 13, 14 Wherever there was a strong clinical and statistical correlation between two variables, the one with better prediction potential and/or clinical relevance was chosen, and the other was omitted.
Our study population was completely imbalanced in terms of mortality, where its rate was only $3.39\%$. To tackle this common challenge in ML models, we performed the synthetic minority oversampling technique (SMOTE) to balance our data in the training sample. SMOTE works by identifying the minority group's k‐nearest neighbors, and it selects a set of neighbors which then generates new data using them. 15 Ten‐fold cross‐validation with the SMOTE of $25\%$ (for the ratio of the minority to majority group) was used to tune this oversampling strategy and select the best minority to majority class ratio.
As the last step of preparation for the model development, the “standard scaler” (from the scikit‐learn package 16) was used to scale each variable by removing the mean and scaling to unit variance, which is the requirement for many ML algorithms.
## Model development
Predictive ML models used in this study were [1] logistic regression (LR), [2] extreme gradient boosting (XGB), [3] naïve Bayes (NB), [4] RF, and [5] Artificial Neural Network (ANN). The diagram for ANN and the number of layers are shown in Supporting Information: Figure 1. In all models, we used variables obtained by the feature selection method previously described. The "Grid Search” method was used to select the best parameters in each model to increase the accuracy of the model performance.
## Model performance evaluation
Performance evaluation was done using the following metrics: A) sensitivity and specificity; B) accuracy of prediction using 10‐fold cross‐validation; C) AUC score by plotting true positive against false positive rate. The threshold is the cut‐off to allocate a probability into a class label and is normally set at 0.5 ($50\%$). Due to the highly imbalanced outcome in our study, this rate of 0.5 was tuned by utilizing 10‐fold cross‐validation in the train data to adjust the sensitivity and specificity of models.
The primary metric for evaluating models was chosen as AUC (with a $95\%$ confidence interval [CI] using several random states) since it is independent of the threshold. To validate the findings, the best model in terms of AUC was implemented to measure the metrics for the most recent $30\%$ of cases in terms of admission time (2013–2015). This method assesses the temporal validity of findings over time. 17, 18
## Statistical analysis
Baseline characteristics are reported as mean ± standard deviation (SD) or proportion (percentage). The comparison was made using Pearson χ 2 test and Fisher's exact test for categorical variables, in addition to an independent sample t‐test for continuous variables. A two‐sided p value of less than.05 was considered statistically significant. Prediction models were designed and evaluated for 1‐year mortality for the whole hypertensive cohort of patients and subgroups based on gender, age group, BMI, diabetes, and smoking status. All statistical analyses and model development were performed using Python (version 3.10). LR, NB, and RF models were implemented using scikit‐learn (1.0.2) library, 16 ANN with TensorFlow (version 2), 19 and XGB using XGBoost (version 1.6.0) Python library. The methodological design of the study including all the mentioned stages performed is illustrated in Figure 1.
**Figure 1:** *Design of study, including all mentioned steps in ML models. ANN, artificial neural network; AUC, area under the curve; CABG, coronary artery bypass grafting; ML, machine learning; SMOTE, synthetic minority oversampling technique.*
## Baseline characteristics
Totally, 8,493 hypertensive patients with a mean age of 68.27 ± 9.27 years (mean ± SD) were assessed for model development and evaluation. Of the mentioned population, $63.86\%$ were male, $46.84\%$ had diabetes, and $38.61\%$ had a family history of CAD. Details of baseline characteristics of hypertensive patients in the CABG cohort are shown in Supporting Information: Table 1. Among all patients, 303 ($3.39\%$) died during a 1‐year follow‐up. Patients who died were significantly older than survivors (71.94 ± 9.56 vs. 68.41 ± 9.26 years; $p \leq .001$). Hb and EF were significantly lower in dead patients compared to alive ones. In addition, the prevalence of diabetes was higher in patients who died ($54.78\%$ vs. $45.65\%$; $p \leq .001$). Figure 2 represents the baseline characteristics of the dead and alive patients in the whole cohort measured before, during, or after the CABG.
**Figure 2:** *Comparison of baseline and hospitalization characteristics of survivors and nonsurvivors with 1‐year follow‐up; (A) dichotomous variables, (B) continuous variables. BMI, body mass index; CABG, coronary artery bypass grafting; COPD, chronic obstructive pulmonary disease; CVA, cerebrovascular accident; EF, ejection fraction; FBS, fasting blood glucose; Hb, hemoglobin; HDL‐C, high‐density lipoprotein cholesterol; HF, heart failure; LDL‐C, low‐density lipoprotein cholesterol; MI, myocardial infarction; PCI, percutaneous coronary intervention; PVD, peripheral vascular disease; TG, triglyceride.*
## Feature selection
The RF feature selector was used using a 10‐fold cross‐validation method to select top features given their AUC. Using Pearson correlation r, we determined correlations between the features. Total ventilation hours and BMI were used instead of total ICU hours and weight due to statistical correlation and more clinical acceptance. Figure 3A illustrates all sorted feature importance obtained by the RF feature selector and the feature selection cut‐off line. The top variables used in our models were total ventilation hours, EF, TG, age, creatinine, Hb, LDL‐C, total cholesterol, FBS, HDL‐C, and BMI, respectively.
**Figure 3:** *Main findings of the ML algorithm for prediction of 1‐year mortality in patients undergoing CABG; (A) feature importance based on the random forest model, (B) receiver operating characteristic curve for 1‐year mortality prediction in all five ML models. CABG, coronary artery bypass grafting; ML, machine learning.*
## Models evaluation
We designed five ML algorithms for the prediction of 1‐year mortality among hypertensive patients undergoing CABG. Table 1 compares the sensitivity, specificity, and AUC of prediction models. All the models had an acceptable performance with LR outperforming others [AUC ($95\%$ CI) = 0.82 (0.78–0.86)]. Considering AUC as the main metric for evaluation, LR was followed by XGB, ANN, RF, and NB. In addition, LR had the highest specificity and accuracy (specificity = $83\%$ and accuracy = $82.37\%$), while XGB had the best performance in terms of sensitivity ($88\%$). Figure 3B demonstrates the receiver operating characteristic curve (ROC) for all five models. Finally, the LR model showed an AUC of 0.77 (0.73–0.81) for the most recent $30\%$ of the total cohort.
**Table 1**
| Unnamed: 0 | Sensitivity (%) | Specificity (%) | Accuracy (%) | AUC [95% confidence Interval] |
| --- | --- | --- | --- | --- |
| Logistic regression | 66.0 | 83.0 | 82.37 | 0.819 [0.777–0.863] |
| Extreme gradient boosting | 88.0 | 59.23 | 60.31 | 0.812 [0.765–0.854] |
| Random forest | 76.0 | 69.34 | 69.59 | 0.804 [0.759–0.846] |
| Naïve Bayes | 67.0 | 81.49 | 80.95 | 0.791 [0.739–0.845] |
| Artificial neural network | 78.0 | 69.92 | 70.22 | 0.806 [0.759–0.853] |
## Subgroups
Models ran for each of the subgroups of hypertensive patients. Evaluation metrics for the LR model as the top prediction model in the whole population are illustrated in Figure 4. LR model had the highest prediction ability for females, the age range of 50–59 and 80–89, overweight patients, diabetic cases, and smokers with all having AUC > 0.8 (classified as excellent 20). The highest sensitivity was for the overweight subgroup of patients, while the highest specificity and accuracy were for the female subgroup. Details of the other four models' evaluations are illustrated in Supporting Information: Figures 2–5. Among all prediction models, the best performance in terms of AUC was for the age subgroup of 50–59 and NB [AUC ($95\%$ CI) = 0.863 (0.704–0.988)].
**Figure 4:** *Logistic regression (LR) model evaluation for prediction of mortality in different subgroups of patients*
## DISCUSSION
In the present study, we attempted to apply and compare five alternative ML algorithms (LR, XGB, ANN, RF, and NB) concerning the prediction of 1‐year mortality among hypertensive patients undergoing CABG. The findings of this study clearly illustrated the power of ML in improving the prediction of 1‐year mortality, with all five ML models found to be able to predict 1‐year mortality with most of them showing AUC > 0.8 which demonstrates excellent predictive ability, according to AUC interpretation. 20 LR model generally outperformed the remaining methods and exhibited the greatest discrimination (AUC = 0.82).
Accordingly, accurate mortality risk prediction systems can play an essential role in improving the continuity of care and management after cardiac surgery, leading to an increase in the survival of patients. Several risk scoring tools, such as Society of Thoracic Surgeons (STS), EURO‐Score I and II, have been developed to predict mortality and detect the involved factors. However, several limitations of these scores in some surgeries or patient subgroups have been reported. Therefore, it seems that these scoring tools need to be modified and improved. 21, 22, 23, 24 The recent advancement in electronic medical records and artificial intelligence resulted in an increasing interest in utilizing ML algorithms for individualized clinical decision‐making and risk prediction. 25 ML algorithms showed a notable ability to be trained to develop personalized risk prediction scoring systems for outcomes of specific therapeutic approaches such as surgeries by identifying complex patterns in the big data. 26, 27, 28 In addition, ML models allow for adjustment of the sensitivity and specificity of each model in different clinical settings in the context of risk predictions at the individual level. 29 Although some previous studies used the $50\%$ default threshold, which can lead to a plethora of missed cases, we modified it to achieve the optimum sensitivity and specificity on the ROC curve, the same as what was done earlier in other studies. 29, 30 It has been illustrated that ML could improve the forecasting quality of the traditional epidemiologic standard mortality models. 31, 32 These findings extend several studies where they demonstrated the superiority of ML models compared to classical tools for identifying patients at increased risk of mortality after CABG. ML models demonstrated that they could be more accurate in forecasting in‐hospital mortality after cardiac surgeries than EURO‐Score II. 33 The preoperative ML models also outperformed the conventional STS model concerning the prediction of mortality or major morbidity in patients who underwent isolated CABG, mainly using intraoperative parameters such as cross‐clamp and bypass times as additive predictive factors. 34 In agreement with these findings, a recent study demonstrated the better performance and validity of eight ML models for predicting 4‐year mortality after cardiac surgery compared to traditional statistical methods. Further analyses showed that adapting boosting (Ada) model had the highest predictive performance (AUC = 0.8). In this study, LR was found to be the second best‐performing ML model with a slight difference in accuracy from the Ada (AUC = 0.797). 10 As LR is sometimes considered a traditional model and not an ML algorithm, a meta‐analysis concluded that ML was superior to the LR model in terms of mortality prediction after cardiac surgery, with a nonsignificant trend toward the better predictive ability of each ML algorithm. Nevertheless, the clinical importance of such an enhancement remains challenging. 35 In contrast, another meta‐analysis reported that the discrimination of other alternative ML models for clinical prediction modeling was superior to the traditional LR in studies at high risk of bias, while this result was no longer in studies with low risk of bias. 36 The frequency of hypertension among patients who require CABG is notable, as a recent study reported that hypertension was present in $54.6\%$ of patients who underwent CABG. 37 Moreover, hypertension is a significant risk factor for mortality and worsened prognosis after CABG. 9, 38 Therefore, this study focused on the prediction of mortality after CABG in a hypertensive group of patients, that the LR model represented the best discriminative performance for predicting the 1‐year mortality. The simplicity of implementation and regularization, good efficiency from a training perspective, and not being affected by small data noise and multicollinearity constitute the advantage of LR. 39 It has been reported that LR performs as well as ML models in predicting the risk of CVDs, chronic kidney disease (CKD), diabetes, and hypertension. 40
Similar to our study, several studies aimed to investigate implementing ML models in certain groups of patients after cardiac surgery. For instance, Zhong et al. 41 revealed that the XGB was associated with overall better predictive ability in terms of AUC compared to other models for forecasting the 30‐day mortality in critically ill patients after cardiac surgery. Consistently, another study compared five ML algorithms for estimating the long‐term mortality risk in the older adults (>65 years old) group who underwent CABG. Based on their results, the XGB and multivariate adaptive regression spline (MARS) models yielded the best predictive performance before and after variable selection, respectively. 42 Altogether, there are controversies in selecting the best model for predicting clinical outcomes and mortality.
Feature selection is widely applied to removing irrelevant and unnecessary data, thereby could improve the accuracy and understanding of the ML models. 43 RF algorithm has been applied in many studies 44, 45, 46 and found to perform better in classification prediction modeling compared to other methods in ML techniques. 47, 48 Since the use of too many features can lead to a decrease in the model's performance, reducing the number of variables and taking the correlation of features into consideration are among the advantages of the RF model. 45 Likewise, in this study, we used the RF feature selector technique to determine the top features. Based on our results, the ventilation time after the surgery was recorded as the most influential variable for predicting mortality, followed by baseline EF.
Consistently, in another study, LR, RF, and XGB models selected the mechanical ventilation time as an important perioperative factor for predicting mortality after CABG. 42 Also, the prolonged mechanical ventilation requirement after cardiac surgery has been reported as a predictive factor for in‐hospital and long‐term mortality, with patients who were intubated for more than 21 days having significantly worsened long‐term survival compared to other patients in 1 year (88.9 vs. $70.9\%$, $$p \leq .03$$). 49 Fernandez‐Zamora et al. 50 also reported that prolonged mechanical ventilation (>24 h) postcardiac surgery was observed in $10\%$–$20\%$ of patients, and they represented most of the postoperative mortality. 50 *In a* meta‐analysis, He et al. reported that prolonged mechanical ventilation time (>48 h) could be associated with a higher risk of ventilator‐associated pneumonia (VAP), averaging $35.2\%$. Also, VAP after cardiac surgery is related to poor prognosis with high mortality and long ICU stays. 51 The notable reverse relationship between low EF and risk of post‐CABG mortality has also been frequently reported by other investigations, 52, 53, 54, 55 with a dose–response relationship between reducing EF and risk of death has been revealed. 53 So far, lots of previous studies reported the pivotal role of age, 56, 57 impaired glucose 58, 59, 60 and lipid profile, 61, 62 Hb levels, 63, 64 serum creatinine, 65, 66 and BMI 67, 68 in estimating the early or late post‐CABG survival and mortality.
Similarly, in this study, TG, age, creatinine, Hb, LDL‐C, total cholesterol, FBS, HDL‐C, and BMI were detected in order as other important features by RF. The importance of these factors has been shown in several studies using various ML methods or otherwise. Like our study, a recent survey reported the 25 important predictors selected by the RF algorithm for mortality after cardiac surgery, which include chronic HF, mechanical ventilation, sodium, blood pressure, Hb, age, creatinine, renal failure, dyslipidemia, and glucose. 10 In another study, XGB selected the serum creatinine, weight, age, and EF as the most important predictor for in‐hospital and 30‐day mortality of cardiac surgery. 69 Besides, age, renal disease, chronic heart failure, and hyperlipidemia were selected as influential factors of long‐term survival in elderly patients with CABG by various ML algorithms. 42 We applied each ML model for each of the subgroups of hypertensive patients, with the results implying that the LR model had the highest predictive ability for females, the age range of 50–59 and 80–89, overweight patients and diabetic cases, and smokers. This can elucidate the beneficiary effects of ML prediction models for each subgroup of patients and result in better utilization of such models in clinical settings. Although there are not many studies in this regard, the study predicting mortality in the elderly population undergoing CABG compared five ML models and found that XGB had the best predictive ability. 42 These findings might suggest that it should be considered a targeted approach to training the ML models for mortality prediction in each subgroup of patients. However, more studies are needed to confirm these primary findings.
Our study proposed a prediction model for the hypertensive population undergoing CABG. As it has been reported that hypertensive patients can contribute up to $80\%$ of patients scheduled for coronary revascularization surgery, 70, 71 our findings can have clinical applications in these highly susceptible cases. There is a need for regional models designed for specific populations based on their local demographic features. There have been studies conducted on this topic in CABG patients in Iran 72, 73; however, this is the first study focusing on hypertensive patients. Moreover, we designed these models with a combination of demographic, clinical, and laboratory characteristics which are widely available in clinical settings, while the use of a combination of variables has been shown to be beneficial in the prediction. 74, 75 While our study was based on the databank, which provides a large population's various demographic, preoperative, intra‐operative, and postoperative information, there were several limitations. First, we used data from a single heart center in Iran. Although we tried to address this issue by performing a separate analysis on the most recent $30\%$ of data to assess the model's validity over time, multicenter registry studies and assessment of the prediction models on other centers in the country are required to strengthen the generalizability of findings. Second, since the missing data of some relevant input features were high, we discarded them. Moreover, the potential effect of not including confounding variables should be considered. It should also be noted that there were no ECG data and follow‐up laboratory data available in this databank. Like usual real‐world clinical data sets, our data set is also composed of imbalanced data, which is a main methodological challenge in ML models. Several approaches have been suggested to resolve this issue, including oversampling the minority group, undersampling the majority group, and lowering the prediction threshold. 76, 77 To overcome the imbalance of mortality data, we modified the threshold and applied the SMOTE oversampling method, which is more frequently used for predicting meager outcomes such as mortality than undersampling methods due to retaining valuable data. 15, 78
## CONCLUSION
Five different predictions using ML models for 1‐year mortality after CABG in hypertensive patients were developed. After applying RF for feature selection, the 11 most important features for the mortality prediction were detected. Among them, the mechanical ventilation time and baseline EF were by far the most influential determinants. All ML models, including LR, XGB, ANN, RF, and NB, demonstrated acceptable predictive performance, with LR providing the greatest AUC. ML algorithms may pave the innovative way for early and accurate prediction of post‐CABG mortality in high‐risk groups, especially hypertensive patients. It could offer individualized tools for clinical decision‐making and management. However, the current study's findings warrant further validation by more studies.
## AUTHOR CONTRIBUTIONS
Amir Hossein Behnoush and Amirmohammad Khalaji: Design, manuscript drafting, data analysis, and revision. Malihe Rezaee, Shahram Momtahen, Soheil Mansourian, Jamshid Bagheri, and Farzad Masoudkabir: Data gathering and manuscript drafting. Kaveh Hosseini: Supervision, design, manuscript drafting, and critical revision.
## CONFLICT OF INTEREST
The authors declare no conflict of interest.
## DATA AVAILABILITY STATEMENT
The data set analyzed in this study, along with the codes used to develop and evaluate machine learning models, are available upon reasonable request from the corresponding author.
## References
1. 1
WHO
. Cardiovascular Diseases. World Health Organization; 2022. Accessed August 2022. https://www.who.int/health-topics/cardiovascular-diseases. *Cardiovascular Diseases* (2022)
2. Roth GA, Mensah GA, Johnson CO. **Global burden of cardiovascular diseases and risk factors, 1990–2019**. *JACC* (2020) **76** 2982-3021. PMID: 33309175
3. Gu D, Qu J, Zhang H, Zheng Z. **Revascularization for coronary artery disease: principle and challenges**. *Adv Exp Med Biol* (2020) **1177** 75-100. PMID: 32246444
4. Deb S, Wijeysundera HC, Ko DT, Tsubota H, Hill S, Fremes SE. **Coronary artery bypass graft surgery vs percutaneous interventions in coronary revascularization: a systematic review**. *JAMA* (2013) **310** 2086-2095. PMID: 24240936
5. Mozaffarian D, Benjamin EJ, Go AS. **Executive summary: heart disease and stroke Statistics‐2016 update a report from the American Heart Association**. *Circulation* (2016) **133** 447-454. PMID: 26811276
6. Rosendorff C, Lackland DT, Allison M. **Treatment of hypertension in patients with coronary artery disease: a scientific statement from the American Heart Association, American College of Cardiology, and American Society of Hypertension**. *Circulation* (2015) **131** e435-e470. PMID: 25829340
7. Herlitz J, Brandrup‐Wognsen G, Haglid M. **Original article 309 mortality and morbidity during a period of 2 years after coronary artery bypass surgery in patients with and without a history of hypertension**. *J Hypertens* (1996) **14** 309-314. PMID: 8723983
8. Zhou Z, Chen J, Fu G. **Association of post‐operative systolic blood pressure variability with mortality after coronary artery bypass grafting**. *Front Cardiovasc Med* (2021) **8**. PMID: 34458342
9. Aronson S, Boisvert D, Lapp W. **Isolated systolic hypertension is associated with adverse outcomes from coronary artery bypass grafting surgery**. *Anesth Analg* (2002) **94** 1079-1084. PMID: 11973166
10. Yu Y, Peng C, Zhang Z. **Machine learning methods for predicting long‐term mortality in patients after cardiac surgery**. *Front Cardiovasc Med* (2022) 9
11. Benedetto U, Sinha S, Lyon M. **Can machine learning improve mortality prediction following cardiac surgery?**. *Eur J Cardiothorac Surg* (2020) **58** 1130-1136. PMID: 32810233
12. Cutler DR, Edwards TC, Beard KH. **Random forests for classification in ecology**. *Ecology* (2007) **88** 2783-2792. PMID: 18051647
13. Calle ML, Urrea V. **Letter to the editor: stability of random forest importance measures**. *Brief Bioinform* (2011) **12** 86-89. PMID: 20360022
14. Strobl C. *Statistical issues in machine learning: towards reliable split selection and variable importance measures* (2008)
15. Chawla NV, Bowyer KW, Hall LO, Kegelmeyer WP. **SMOTE: synthetic minority over‐sampling technique**. *J Artif Intell Res* (2002) **16** 321-357
16. Pedregosa F, Varoquaux G, Gramfort A. **Scikit‐learn: machine learning in python**. *J Mach Learn Res* (2011) **12** 2825-2830
17. Steyerberg EW, Harrell FE. **Prediction models need appropriate internal, internal‐external, and external validation**. *JCE* (2016) **69** 245-247. PMID: 25981519
18. Justice AC. **Assessing the generalizability of prognostic information**. *Ann Intern Med* (1999) **130** 515-524. PMID: 10075620
19. Rampasek L, Goldenberg A. **TensorFlow: Biology's Gateway to deep learning?**. *Cell Systems* (2016) **2** 12-14. PMID: 27136685
20. Mandrekar JN. **Receiver operating characteristic curve in diagnostic test assessment**. *J Thorac Oncol* (2010) **5** 1315-1316. PMID: 20736804
21. Nashef SAM, Roques F, Michel P. **European system for cardiac operative risk evaluation (Euro SCORE)**. *Eur J Cardiothorac Surg* (1999) **16** 9-13. PMID: 10456395
22. Ii E, Nashef S, Roques F. **EuroSCORE II**. *Eur J Cardiothorac Surg* (2012) **41** 734-745. PMID: 22378855
23. Shahian DM, O'Brien SM, Filardo G. **The Society of Thoracic Surgeons 2008 cardiac surgery risk models: part 3—valve plus coronary artery bypass grafting surgery**. *Ann Thorac Surg* (2009) **88** S43-S62. PMID: 19559824
24. Chhor V, Merceron S, Ricome S. **Poor performances of EuroSCORE and CARE score for prediction of perioperative mortality in octogenarians undergoing aortic valve replacement for aortic stenosis**. *Eur J Anaesthesiol* (2010) **27** 702-707. PMID: 20520558
25. Fröhlich H, Balling R, Beerenwinkel N. **From hype to reality: data science enabling personalized medicine**. *BMC Med* (2018) **16** 150. PMID: 30145981
26. Bica I, Alaa AM, Lambert C, Van Der Schaar M. **From real‐world patient data to individualized treatment effects using machine learning: current and future methods to address underlying challenges**. *Clin Pharmacol Ther* (2021) **109** 87-100. PMID: 32449163
27. MacEachern SJ, Forkert ND. **Machine learning for precision medicine**. *Genome* (2021) **64** 416-425. PMID: 33091314
28. Vellido A. **The importance of interpretability and visualization in machine learning for applications in medicine and health care**. *Neural Comput Appl* (2020) **32** 18069-18083
29. Bihorac A, Ozrazgat‐Baslanti T, Ebadi A. **MySurgeryRisk: development and validation of a machine‐learning risk algorithm for major complications and death after surgery**. *Ann Surg* (2019) **269** 652-662. PMID: 29489489
30. Thorsen‐Meyer H‐C, Nielsen AB, Nielsen AP. **Dynamic and explainable machine learning prediction of mortality in patients in the intensive care unit: a retrospective study of high‐frequency data in electronic patient records**. *Lancet Digit Health* (2020) **2** e179-e191. PMID: 33328078
31. Levantesi S, Pizzorusso V. **Application of machine learning to mortality modeling and forecasting**. *Risks* (2019) **7** 26
32. Weng SF, Vaz L, Qureshi N, Kai J. **Prediction of premature all‐cause mortality: A prospective general population cohort study comparing machine‐learning and standard epidemiological approaches**. *PLoS One* (2019) **14**. PMID: 30917171
33. Allyn J, Allou N, Augustin P. **A comparison of a machine learning model with EuroSCORE II in predicting mortality after elective cardiac surgery: a decision curve analysis**. *PLoS One* (2017) **12**. PMID: 28060903
34. Zea‐Vera R, Ryan CT, Havelka J. **Machine learning to predict outcomes and cost by phase of care after coronary artery bypass grafting**. (2022) **114** 711-719
35. Benedetto U, Dimagli A, Sinha S. **Machine learning improves mortality risk prediction after cardiac surgery: systematic review and meta‐analysis**. *J Thorac Cardiovasc Surg* (2022) **163** 2075-2087.e9. PMID: 32900480
36. Christodoulou E, Ma J, Collins GS, Steyerberg EW, Verbakel JY, Van Calster B. **A systematic review shows no performance benefit of machine learning over logistic regression for clinical prediction models**. *JCE* (2019) **110** 12-22. PMID: 30763612
37. Ahmed Z, Kumar V, Kamal AA, Khatri A, Khushk S. **Frequency of hypertension and diabetes mellitus in patients undergoing coronary artery bypasses grafting surgery**. *Ann Romanian Soc Cell Biol* (2021) **25** 914-924
38. Karimi A, Ahmadi H, Davoodi S. **Factors affecting postoperative morbidity and mortality in isolated coronary artery bypass graft surgery**. *Surg Today* (2008) **38** 890-898. PMID: 18820863
39. Ray S. *2019 International conference on machine learning, big data, cloud and parallel computing (COMITCon)* (2019) 35-39
40. Nusinovici S, Tham YC, Chak Yan MY. **Logistic regression was as good as machine learning for predicting major chronic diseases**. *JCE* (2020) **122** 56-69. PMID: 32169597
41. Zhong Z, Yuan X, Liu S, Yang Y, Liu F. **Machine learning prediction models for prognosis of critically ill patients after open‐heart surgery**. *Sci Rep* (2021) **11** 3384. PMID: 33564090
42. Huang Y‐C, Li S‐J, Chen M, Lee T‐S, Chien Y‐N. **Machine‐learning techniques for feature selection and prediction of mortality in elderly CABG patients**. *Healthcare (Basel)* (2021) **9**. PMID: 34067148
43. Cai J, Luo J, Wang S, Yang S. **Feature selection in machine learning: a new perspective**. *Neurocomputing* (2018) **300** 70-79
44. Paul D, Su R, Romain M, Sébastien V, Pierre V, Isabelle G. **Feature selection for outcome prediction in oesophageal cancer using genetic algorithm and random forest classifier**. *Comput Med Imaging Graph* (2017) **60** 42-49. PMID: 28087102
45. Behnoush B, Bazmi E, Nazari S, Khodakarim S, Looha M, Soori H. **Machine learning algorithms to predict seizure due to acute tramadol poisoning**. *Hum Exp Toxicol* (2021) **40** 1225-1233. PMID: 33538187
46. Cui S, Luo Y, Tseng H‐H, Ten Haken RK, El Naqa I. **Combining handcrafted features with latent variables in machine learning for prediction of radiation‐induced lung damage**. *Med Phys* (2019) **46** 2497-2511. PMID: 30891794
47. Chen R‐C, Dewi C, Huang S‐W, Caraka RE. **Selecting critical features for data classification based on machine learning methods**. *J Big Data* (2020) **7** 52
48. Speiser JL, Miller ME, Tooze J, Ip E. **A comparison of random forest variable selection methods for classification prediction modeling**. *Expert Syst Appl* (2019) **134** 93-101. PMID: 32968335
49. Pappalardo F, Franco A, Landoni G, Cardano P, Zangrillo A, Alfieri O. **Long‐term outcome and quality of life of patients requiring prolonged mechanical ventilation after cardiac surgery**. *Eur J Cardiothorac Surg* (2004) **25** 548-552. PMID: 15037270
50. Fernandez‐Zamora MD, Gordillo‐Brenes A, Banderas‐Bravo E. **Prolonged mechanical ventilation as a predictor of mortality after cardiac surgery**. *Respir Care* (2018) **63** 550-557. PMID: 29382792
51. He S, Chen B, Li W. **Ventilator‐associated pneumonia after cardiac surgery: a meta‐analysis and systematic review**. *J Thorac Cardiovasc Surg* (2014) **148** 3148-3155. PMID: 25240522
52. Fallahzadeh A, Sheikhy A, Ajam A. **Significance of preoperative left ventricular ejection fraction in 5‐year outcome after isolated CABG**. *J Cardiothorac Surg* (2021) **16** 353. PMID: 34961534
53. Omer S, Adeseye A, Jimenez E, Cornwell LD, Massarweh NN. **Low left ventricular ejection fraction, complication rescue, and long‐term survival after coronary artery bypass grafting**. *J Thorac Cardiovasc Surg* (2022) **163** 111-119. PMID: 32327186
54. Kurniawaty J, Setianto BY, Supomo Y, Widyastuti CE. **The effect of low preoperative ejection fraction on mortality after cardiac surgery in Indonesia**. *Vasc Health Risk Manag* (2022) **18** 131-137. PMID: 35356550
55. Awan NI, Jan A, Rehman MU, Ayaz N. **The effect of ejection fraction on mortality in coronary artery bypass grafting (CABG) patients**. *Pak J Med Sci* (2020) **36** 1454. PMID: 33235556
56. Lemaire A, Soto C, Salgueiro L, Ikegami H, Russo MJ, Lee LY. **The impact of age on outcomes of coronary artery bypass grafting**. *J Cardiothorac Surg* (2020) **15** 158. PMID: 32611349
57. Nuru A, Weltzien JAH, Sandvik L, Tønnessen T, Bjørnstad JL. **Short‐and long‐term survival after isolated coronary artery bypass grafting, the impact of gender and age**. *Scand Cardiovasc J* (2019) **53** 342-347. PMID: 31321989
58. Djupsjo C, Sartipy U, Ivert T. **Preoperative disturbances of glucose metabolism and mortality after coronary artery bypass grafting**. *Open Heart* (2020) **7**. PMID: 32487771
59. Fallahzadeh A, Sheikhy A, Hosseini K. **Prognostic impact of prediabetes on patient outcomes after coronary artery bypass grafting: a single‐center cohort study**. *Crit Pathw Cardiol* (2021) **20** 220-225. PMID: 34570012
60. Anderson RE, Klerdal K, Ivert T, Hammar N, Barr G, Öwall A. **Are even impaired fasting blood glucose levels preoperatively associated with increased mortality after CABG surgery?**. *Eur Heart J* (2005) **26** 1513-1518. PMID: 15800018
61. Njoku P, Meka I, Mbadiwe N. **Pre‐operative lipid profile of cardiac surgery patients: implications of incidental findings**. *Int J Med Health Dev* (2022) **27** 131
62. Foody JM, Ferdinand FD, Pearce GL, Lytle BW, Cosgrove DM, Sprecher DL. **HDL cholesterol level predicts survival in men after coronary artery bypass graft surgery: 20‐year experience from the Cleveland Clinic Foundation**. *Circulation* (2000) **102** Iii‐90-Iii‐4
63. Bell ML, Grunwald GK, Baltz JH. **Does preoperative hemoglobin independently predict short‐term outcomes after coronary artery bypass graft surgery?**. *Ann Thorac Surg* (2008) **86** 1415-1423. PMID: 19049724
64. Zindrou D, Taylor KM, Bagger JP. **Preoperative haemoglobin concentration and mortality rate after coronary artery bypass surgery**. *Lancet* (2002) **359** 1747-1748. PMID: 12049866
65. Brown JR, Cochran RP, Dacey LJ. **Perioperative increases in serum creatinine are predictive of increased 90‐day mortality after coronary artery bypass graft surgery**. *Circulation* (2006) **114** I‐409-I‐13
66. Chikwe J, Castillo JG, Rahmanian PB, Akujuo A, Adams DH, Filsoufi F. **The impact of moderate–to–end‐stage renal failure on outcomes after coronary artery bypass graft surgery**. *J Cardiothorac Vasc Anesth* (2010) **24** 574-579. PMID: 20570181
67. Echahidi N, Pibarot P, Després JP. **Metabolic syndrome increases operative mortality in patients undergoing coronary artery bypass grafting surgery**. *JACC* (2007) **50** 843-851. PMID: 17719470
68. van Straten AHM, Bramer S, Soliman Hamad MA. **Effect of body mass index on early and late mortality after coronary artery bypass grafting**. *Ann Thorac Surg* (2010) **89** 30-37. PMID: 20103201
69. Kilic A, Goyal A, Miller JK. **Predictive utility of a machine learning algorithm in estimating mortality risk in cardiac surgery**. *Ann Thorac Surg* (2020) **109** 1811-1819. PMID: 31706872
70. Mehta RH, Bhatt DL, Steg PG. **Modifiable risk factors control and its relationship with 1 year outcomes after coronary artery bypass surgery: insights from the REACH registry**. *Eur Heart J* (2008) **29** 3052-3060. PMID: 18996953
71. Pivatto Júnior F, Valle FH, Pereira EMC. **Sobrevida em longo prazo de octogenários submetidos à cirurgia de revascularização miocárdica isolada**. *Braz J Cardiovasc Surg* (2011) **26** 21-26
72. Arian F, Amini M, Mostafaei S. **Myocardial function prediction after coronary artery bypass grafting using MRI radiomic features and machine learning algorithms**. *J Digit Imaging* (2022) **35** 1708-1718. PMID: 35995896
73. Khalaji A, Behnoush AH, Jameie M. **Machine learning algorithms for predicting mortality after coronary artery bypass grafting**. *Front Cardiovasc Med* (2022) **9**. PMID: 36093147
74. Chadaga K, Chakraborty C, Prabhu S, Umakanth S, Bhat V, Sampathila N. **Clinical and laboratory approach to diagnose COVID‐19 using machine learning**. *Interdiscip Sci* (2022) **14** 452-470. PMID: 35133633
75. Bazmi E, Behnoush B, Hashemi Nazari S, Khodakarim S, Behnoush AH, Soori H. **Seizure prediction model in acute tramadol poisoning; a derivation and validation study**. *Arch Acad Emerg Med* (2020) **8** 59
76. Esposito C, Landrum GA, Schneider N, Stiefl N, Riniker S. **GHOST: adjusting the decision threshold to handle imbalanced data in machine learning**. *J Chem Inf Model* (2021) **61** 2623-2640. PMID: 34100609
77. Tyagi S, Mittal S, Singh P, Kar A, Singh Y, Kolekar M, Tanwar S. *Proceedings of ICRIC 2019. Lecture Notes in Electrical Engineering* (2020) **597** 209-221
78. Karanasiou GS, Tripoliti EE, Papadopoulos TG. **Predicting adherence of patients with HF through machine learning techniques**. *Healthc Technol Lett* (2016) **3** 165-170. PMID: 27733922
|
---
title: 'Association between visceral adiposity index and heart failure: A cross‐sectional
study'
authors:
- Xinyu Zhang
- Yijun Sun
- Ying Li
- Chengwei Wang
- Yi Wang
- Mei Dong
- Jie Xiao
- Zongwei Lin
- Huixia Lu
- Xiaoping Ji
journal: Clinical Cardiology
year: 2023
pmcid: PMC10018101
doi: 10.1002/clc.23976
license: CC BY 4.0
---
# Association between visceral adiposity index and heart failure: A cross‐sectional study
## Abstract
### Background
Obesity is an important risk factor for heart failure (HF).
### Hypothesis
Visceral adiposity index (VAI) is a simple metric for assessing obesity; however, the association between VAI and risk for HF has not been studied.
### Methods
A cross‐sectional study involving 28 764 participants ≥18 years of age from the National Health and Nutrition Examination Survey (NHANES), 2009–2018, in the United States was performed. VAI was calculated using body mass index (BMI), waist circumference (WC), triglycerides (TG), and high‐density lipoprotein cholesterol. VAI was analyzed as a continuous and categorical variable to examine its association with HF. Subgroup analysis was also performed.
### Results
The highest VAI (fourth quartile [Q4]) was found among males, BMI, systolic and diastolic blood pressure, WC, hypertension, diabetes, liver disease, coronary heart disease, smoking, total cholesterol, and TG. More participants in Q4 took β‐receptor blockers, angiotensin‐converting enzyme inhibitors/angiotensin II receptor blockers/angiotensin receptor‐neprilysin inhibitor, calcium channel blockers, and antidiabetic and antihyperlipidemic medications. Participants with HF exhibited greater VAI. A per‐unit increase in VAI resulted in a $4\%$ increased risk for HF (odds ratio [OR] 1.04 [$95\%$ confidence interval (CI) 1.02–1.05]). After multivariable adjustment, compared with the lowest quartile, the OR for Q3 was 1.55 ($95\%$ CI 1.24–1.94). Subgroup analysis revealed no significant interactions between VAI and specific subgroups.
### Conclusion
VAI was independently associated with the risk for HF. As a noninvasive index of visceral adiposity, VAI could be used for a “one shot” assessment of HF risk and may serve as a novel marker.
## INTRODUCTION
Heart failure (HF) refers to a group of complex clinical syndromes caused by many factors (myocardial dysfunction, valvular diseases, pathological changes in the pericardium and endocardium, and dysfunction of heart rhythm and conduction), 1 which leads to ventricular systolic or diastolic dysfunction. 2 In the United States, approximately 6.2 million individuals ≥20 years of age experience HF, with approximately 1 million newly diagnosed cases of HF annually, and the prevalence continues to rise. 3, 4 Data from the European Society of Cardiology show that approximately $1\%$ of patients with HF are <55 years of age and approximately $10\%$ of those with HF are ≥70 years of age. 1 In developed countries, the incidence rate of HF may decrease after adjusting for age, which may reflect good management of cardiovascular disease (CVD); however, the overall incidence rate of HF is increasing due to aging. 1 This situation is similar to that observed in developing countries. According to the latest survey results of HF epidemiology released in 2019, the number of chronic HF cases in *China is* approximately 13.7 million, and its prevalence has increased by $44\%$ in the past 15 years. 5 Approximately $29\%$–$40\%$ of patients with HF are overweight (body mass index [BMI] 25.0–29.9 kg/m2), and $30\%$–$49\%$ are obese (BMI ≥ 30 kg/m2). It is noteworthy that obesity is more common in patients with HF with preserved ejection fraction (HFpEF) than in those with HF with reduced ejection fraction (HFrEF), and >$80\%$ of HFpEF patients exhibit a BMI in the range of overweight or obesity. 6, 7 However, the relationship between obesity and HF remains controversial. Dunlay et al. demonstrated that obesity, measured according to increased BMI, is a major risk factor for the development of HF. 8 However, some research appears to highlight the “obesity paradox,” 9 which means that overweight or grade 1 obesity can also result in a better survival rate for HF. In addition to BMI, visceral adipose tissue (VAT) is an indicator of obesity. Previous studies have shown that patients with HF, especially HFpEF, exhibit higher VAT. 10, 11, 12 These studies usually used computed tomography (CT) or magnetic resonance imaging (MRI) to evaluate VAT. The visceral adiposity index (VAI) was calculated using BMI, waist circumference (WC), triglycerides (TG), and high‐density lipoprotein cholesterol (HDL‐c). 13 Compared to CT or MRI, the calculation of VAI is easier, and more economical and convenient. Previous studies have shown that VAT is associated with diabetes, hyperuricemia, metabolic syndrome, hypertension, atherosclerosis, and vascular calcification. 13, 14, 15, 16, 17, 18 However, to our knowledge, the association between VAI and HF has not been studied.
As such, the purpose of this study was to evaluate the association between VAI and HF among middle‐age and elderly participants of the US National Health and Nutrition Examination Survey (NHANES), 2009–2018.
## Study population
The NHANES is a nationally representative cross‐sectional study that enrolls participants through a stratified multistage probability and oversampling design that enables weighted analysis that represents the noninstitutionalized, civilian population of the United States (US). Data are released every 2‐year cycle. Each participant represents approximately 50 000 US citizens. All participants provided informed consent before participation, and ethics approval for the study was obtained from the Research Ethics Review Board at the National Centre for Health Statistics, which consisted of a physician, medical and health technicians, and dietary and health interviewers, who conducted surveys through interviews, health measurements, and laboratory tests. An advanced computer system collects and processes all NHANES data. Findings of this survey can be used to determine the prevalence of diseases and their risk factors.
Data for the present study were derived from the 2009–2018 NHANES cycle. In this cohort, 49 693 participants completed the interviews. Participants who were <18 years of age ($$n = 19$$ 341) and those with missing data regarding HF status ($$n = 1588$$) were excluded. Ultimately, data from 28 764 participants were included in this cross‐sectional study. Participants with HF were defined as those who answered yes to the question, “*Has a* doctor or other health professional ever told you that you had congestive HF?.” A detailed flow‐diagram illustrating participant selection is presented in Supporting Information: Figure 1. The National Center for Health Statistics Ethics Review Board reviewed and approved the NHANES protocol and all participants provided written informed consent before data collection.
## VAI score
VAI score was calculated according to previously reported equations 13: For males, VAI=WC(cm)/(39.68+1.88×BMI kg/m2)×(TG[mmol/l]/1.03)×(1.31/HDL−c[mmol/l]).
For females, VAI=WC(cm)/(39.58+1.89×BMIkg/m2)×(TG[mmol/l]/0.81)×(1.52/HDL−c[mmol/l]).
NHANES researchers collected anthropometric data (i.e., height, weight, calculated BMI, and WC) and biochemical data (i.e., glycated hemoglobin, direct HDL‐c, and fasting TG) that were used to calculate VAI. A higher VAI score reflected a greater amount of estimated visceral adiposity.
## Variables of interest
Potential covariates, including demographics, comorbidities, lifestyle variables, BMI, TG, TC, serum uric acid (UA), estimated glomerular filtration rate (eGFR), and markers of inflammation, were selected based on clinical relevance and statistical significance. The baseline characteristics of the participants, including demographics, comorbidities, and lifestyle information, were obtained using a questionnaire. BMI, WC, and other biochemical parameters were obtained from medical examinations and laboratory assessments performed at the mobile examination center. BMI was calculated as weight (kg) divided by height (m) squared (kg/m2). Hypertension was defined as self‐reported physician‐diagnosed hypertension, use of antihypertensive medications, or blood pressure measurement of $\frac{140}{90}$ mmHg. 19 *Diabetes mellitus* was defined as self‐reported physician‐diagnosed diabetes, taking oral hypoglycemic agents or insulin, fasting glucose level of 126 mg/dl, or plasma glucose level of 200 mg/dl 2 h after an oral glucose tolerance test. 20 Participants with anemia were defined as those who answered yes to the question, “During the past 3 months, have you been on treatment for anemia?.” Participants with liver disease were defined as those who answered yes to the question, “*Has a* doctor or other health professional ever told you that you had any kind of liver condition?.” Participants with coronary heart disease were defined as those who answered yes to the question, “*Has a* doctor or other health professional ever told you that you had coronary heart disease?.” Participants with kidney disease were defined as those who answered yes to the question, “Have you ever been told by a doctor or other health professional that you had weak or failing kidneys? Do not include kidney stones, bladder infections, or incontinence?.” Participants with a history of heart attack were defined as those who answered yes to the question, “*Has a* doctor or other health professional ever told you that you had a heart attack (also called myocardial infarction)?.” Smokers were defined as those who had smoked at least 100 cigarettes in their lifetimes. eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration creatinine equation. 21 Dietary Inflammatory Index (DII) was calculated based on a 24 h dietary recall interview with each participant. 22, 23 The inflammatory indicator was the neutrophil‐lymphocyte ratio (NLR). 24, 25
## Statistical analysis
According to the NHANES analytic guidelines, descriptive results are expressed as weighted mean ± standard error (SE) or median (first quartile, third quartile) (Q2 [Q1, Q3)] for continuous variables and frequency (weighted percentage) for categorical variables. The VAI was analyzed as a continuous and categorical variable (quartiles). Differences in VAI between groups (with and without HF) were tested using the Student's t‐test. Differences in characteristics between the groups were tested using the Student's t‐test for continuous variables and χ2 tests for categorical variables. The odds ratio (OR) and corresponding $95\%$ confidence interval (CI) for HF per unit increase and each quartile, with the lowest quartile as the reference, was estimated using both univariate and multivariate logistic regression models. Tests for linear trends across the VAI categories were conducted using an independent ordinal variable [0, 1, 2, 3] in all models. In addition to the unadjusted model, potential covariates were progressively adjusted in the three models. Model 1 was adjusted for age, sex, and race; model 2 was additionally adjusted for hypertension, diabetes mellitus (DM), smoking, alcohol consumption, coronary heart disease (CHD), kidney disease, and liver disease; and model 3 was further adjusted for eGFR, systolic blood pressure (SBP), diastolic blood pressure (DBP), serum UA, albumin (Alb), hemoglobin (HGB), hematocrit (HCT), and NLR. The restricted cubic spline model was used for dose–response analysis. To explore whether the association between the VAI and HF was modified by sex, age, race, smoking status, and comorbidities, subgroup analyses was performed according to sex (male or female), age group (18–34, 35–54, 55–74 or ≥75 years of age), race, smoking (yes or no), hypertension (yes or no), diabetes (yes or no), CHD (yes or no), liver disease (yes or no), serum UA (< 350 or ≥350 μmol/L), eGFR (<90 or ≥90 ml/min/1.73 m2), and Alb (<40 or ≥40 g/L), and examined the interactions between the stratified variables and VAI using likelihood ratio tests. The “nhanesR” package version 0.9.1.9 was used for data extraction and processing. Free statistics software version 1.4 and the statistical software package R 4.0.1 (R Foundation for Statistical Computing, Vienna, Austria) were used for all analyses. Differences with a two‐tailed $p \leq .05$ were considered to be statistically significant.
## Characteristics of the study population
Basic characteristics of the study population are summarized in Table 1. In total, 28 764 participants, with a weighted average age of 50 years, were included in this study. According to VAI quartile, participants with the highest VAI (i.e., Q4) had higher values in males, Mexican Americans, other Hispanics, non‐Hispanic whites, BMI, SBP, DBP, WC, hypertension, DM, liver diseases, CHD, smoking, TC, TG, and serum UA. More participants in Q4 took β‐receptor blockers, angiotensin‐converting enzyme inhibitors (ACEIs)/angiotensin II receptor blockers (ARBs)/angiotensin receptor‐neprilysin inhibitor (ARNI), calcium channel blockers (CCBs), antidiabetic medication, and antihyperlipidemic medications. The opposite patterns were observed for non‐Hispanic Blacks, alcohol, and HDL‐c. A comparison of the four groups revealed that sex, race, BMI, SBP, DBP, WC, alcohol, smoking, several diseases (hypertension, DM, HF, liver disease, CHD, heart attack, and kidney disease), relevant test results (TC, HDL‐C, TG, serum UA, eGFR, NLR, DII, lymphocytes (Lym), Alb, HGB, and HCT), and the utilization rate of medications (β‐receptor blocker, ACEI/ARB/ARNI, MRA, CCB, diuretic, antidiabetic, and antihyperlipidemic) were significantly different ($p \leq .001$).
**Table 1**
| Variables | All (n = 28764) | Q1 (n = 7191) | Q2 (n = 7191) | Q3 (n = 7191) | Q4 (n = 7191) | p |
| --- | --- | --- | --- | --- | --- | --- |
| Age, years | 49 (34, 64) | 45 (30, 61) | 49 (34, 63) | 52 (36, 68) | 51 (38, 63) | <.001 |
| Gender, no.(%) | | | | | | <.001 |
| Male | 13910 (48.4) | 3661 (50.9) | 3354 (46.6) | 3343 (46.5) | 3552 (49.4) | |
| Female | 14854 (51.6) | 3530 (49.1) | 3837 (53.4) | 3848 (53.5) | 3639 (50.6) | |
| Race, no.(%) | | | | | | <.001 |
| Mexican American | 4157 (14.5) | 691 (9.6) | 1057 (14.7) | 1004 (14) | 1405 (19.5) | |
| Other Hispanic | 2994 (10.4) | 592 (8.2) | 762 (10.6) | 748 (10.4) | 892 (12.4) | |
| Non‐Hispanic White | 11266 (39.2) | 2671 (37.1) | 2829 (39.3) | 2700 (37.5) | 3066 (42.6) | |
| Non‐Hispanic Black | 6239 (21.7) | 2216 (30.8) | 1552 (21.6) | 1649 (22.9) | 822 (11.4) | |
| Other Race | 4108 (14.3) | 1021 (14.2) | 991 (13.8) | 1090 (15.2) | 1006 (14) | |
| BMI, kg/m2 | 28.5 (24.6, 32.5) | 25.1 (22.2, 29.1) | 28.0 (24.5, 32.5) | 29.3 (26.1, 32.3) | 30.4 (27.0, 34.9) | <.001 |
| SBP, mmHg | 124.0 (112.0, 132.0) | 118.0 (108.0, 130.0) | 122.0 (112.0, 134.0) | 124.3 (116.0, 132.0) | 124.3 (114.0, 134.0) | <.001 |
| DBP, mmHg | 70.2 (64.0, 78.0) | 70.0 (62.0, 76.0) | 70.2 (64.0, 78.0) | 70.2 (66.0, 76.0) | 72.0 (64.0, 80.0) | <.001 |
| WC, cm | 99.5 (89.0, 107.8) | 89.0 (80.3, 99.8) | 97.7 (88.5, 108.2) | 99.5 (97.1, 103.6) | 104.7 (95.9, 115.0) | <.001 |
| Hypertension, no.(%) | 11989 (41.7) | 2318 (32.2) | 2842 (39.5) | 3243 (45.1) | 3586 (49.9) | <.001 |
| Diabetes mellitus, no.(%) | 5470 (19.2) | 726 (10.2) | 1199 (16.9) | 1493 (21) | 2052 (28.8) | <.001 |
| HF, no.(%) | 958 (3.3) | 146 (2) | 189 (2.6) | 340 (4.7) | 283 (3.9) | <.001 |
| Anemia, no.(%) | 1292 (4.5) | 305 (4.2) | 304 (4.2) | 388 (5.4) | 295 (4.1) | <.001 |
| Liver disease, no.(%) | 1187 (4.1) | 216 (3) | 277 (3.9) | 295 (4.1) | 399 (5.5) | <.001 |
| CHD, no.(%) | 1173 (4.1) | 196 (2.7) | 263 (3.7) | 344 (4.8) | 370 (5.1) | <.001 |
| Heart attack, no.(%) | 1196 (4.2) | 202 (2.8) | 282 (3.9) | 356 (5) | 356 (5) | <.001 |
| Kidney disease, no.(%) | 997 (3.5) | 170 (2.4) | 202 (2.8) | 350 (4.9) | 275 (3.8) | <.001 |
| Alcohol, g/day | 0.0 (0.0, 15.4) | 0.0 (0.0, 15.4) | 0.0 (0.0, 15.4) | 15.4 (0.0, 15.4) | 0.0 (0.0, 15.4) | <.001 |
| Smoking, no.(%) | 12448 (43.3) | 2869 (39.9) | 3040 (42.3) | 3107 (43.2) | 3432 (47.7) | <.001 |
| TC, mmol/L | 5.0 (4.3, 5.5) | 4.6 (4.0, 5.3) | 4.8 (4.2, 5.5) | 5.0 (4.7, 5.1) | 5.2 (4.5, 5.9) | <.001 |
| HDL‐C, mmol/L | 1.4 (1.1, 1.6) | 1.7 (1.4, 2.0) | 1.4 (1.2, 1.6) | 1.4 (1.2, 1.4) | 1.0 (0.9, 1.2) | <.001 |
| TG, mmol/L | 1.5 (0.9, 2.0) | 0.7 (0.6, 0.9) | 1.2 (1.0, 1.4) | 1.7 (1.5, 1.7) | 2.7 (2.1, 3.5) | <.001 |
| Serum uric acid, mmol/L | 323.8 (267.7, 368.8) | 291.5 (243.9, 350.9) | 309.3 (261.7, 368.8) | 323.8 (309.3, 345.0) | 339.0 (279.6, 398.5) | <.001 |
| eGFR, ml/min/1.73 m2 | 94.0 (81.4, 109.4) | 100.0 (83.9, 114.8) | 96.6 (80.0, 111.7) | 94.0 (86.9, 97.6) | 94.2 (76.0, 108.2) | <.001 |
| NLR, % | 3.3 (2.8, 3.9) | 3.2 (2.7, 3.9) | 3.3 (2.8, 4.0) | 3.3 (3.1, 3.7) | 3.3 (2.8, 4.0) | <.001 |
| DII, mg/L | 0.9 (−0.1, 2.2) | 0.8 (−0.4, 2.1) | 1.0 (−0.3, 2.3) | 0.9 (0.0, 2.0) | 1.1 (−0.2, 2.3) | <.001 |
| Lym,109/L | 2.1 (1.7, 2.5) | 1.9 (1.5, 2.3) | 2.0 (1.6, 2.5) | 2.2 (1.9, 2.3) | 2.3 (1.8, 2.8) | <.001 |
| Alb, g/L | 42.2 (40.0, 44.0) | 43.0 (41.0, 45.0) | 42.0 (40.0, 45.0) | 42.2 (41.0, 43.0) | 42.0 (40.0, 44.0) | <.001 |
| HGB, g/L | 14.0 (13.1, 14.9) | 13.9 (12.9, 14.9) | 14.0 (13.0, 15.0) | 14.0 (13.4, 14.4) | 14.2 (13.2, 15.2) | <.001 |
| HCT, % | 41.3 (38.8, 43.9) | 41.2 (38.5, 43.9) | 41.3 (38.5, 44.2) | 41.3 (39.8, 42.5) | 41.9 (38.9, 44.7) | <.001 |
| Medications, no.(%) | Medications, no.(%) | Medications, no.(%) | Medications, no.(%) | Medications, no.(%) | Medications, no.(%) | Medications, no.(%) |
| β‐receptor blocker | 3208 (11.2) | 492 (6.8) | 697 (9.7) | 963 (13.4) | 1056 (14.7) | <.001 |
| ACEI/ARB/ARNI | 5843 (20.3) | 1046 (14.5) | 1404 (19.5) | 1558 (21.7) | 1835 (25.5) | <.001 |
| MRA | 235 (0.8) | 37 (0.5) | 48 (0.7) | 83 (1.2) | 67 (0.9) | <.001 |
| CCB | 2076 (7.2) | 404 (5.6) | 494 (6.9) | 586 (8.1) | 592 (8.2) | <.001 |
| Diuretic | 3719 (12.9) | 628 (8.7) | 840 (11.7) | 1133 (15.8) | 1118 (15.5) | <.001 |
| Antidiabetic | 3530 (12.3) | 440 (6.1) | 757 (10.5) | 1013 (14.1) | 1320 (18.4) | <.001 |
| Antihyperlipidemic | 5957 (20.7) | 1069 (14.9) | 1417 (19.7) | 1648 (22.9) | 1823 (25.4) | <.001 |
## Association between VAI and HF
Differences in VAI between the two groups with and without HF are shown in Figure 1. Results of analysis revealed that participants with HF exhibited a higher VAI than those without HF ($p \leq .001$).
**Figure 1:** *Comparison of VAI between patients with HF and non‐HF. HF, heart failure; VAI, visceral adiposity index.*
The relationship between VAI and HF as continuous and categorical variables is shown in Table 2. When VAI was analyzed as a continuous variable, a per unit increase in VAI resulted in a higher risk for HF in the univariate logistic regression model (OR 1.04 [$95\%$ CI 1.02–1.05]). The association remained statistically significant in all multivariate logistic regression models after adjusting for several covariates including sex, age, race, hypertension, DM, smoking, alcohol consumption, CHD, kidney disease, liver disease, eGFR, SBP, DBP, UA, Alb, HGB, HCT, and NLR (model 1, OR 1.05 [$95\%$ CI 1.03–1.07]; model 2, OR 1.02 [$95\%$ CI 1–1.05]; and model 3, OR 1.03 [$95\%$ CI 1–1.05]). When VAI was analyzed as a categorical variable, compared with the top VAI quartile, subjects in the third quartile (Q3) had the highest risk for HF (OR 1.55 [$95\%$ CI 1.24–1.94]), adjusting for age, sex, race, hypertension, DM, smoking, alcohol, CHD, kidney disease, liver disease, eGFR, SBP, DBP, UA, Alb, HGB, HCT, and NLR.
**Table 2**
| Unnamed: 0 | Unadjusted | Unadjusted.1 | Model 1a | Model 1a.1 | Model 2b | Model 2b.1 | Model 3c | Model 3c.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p |
| Continuous per unit increase | 1.04 (1.02~1.05) | <.001 | 1.05 (1.03~1.07) | <.001 | 1.02 (1~1.05) | .041 | 1.03 (1~1.05) | .031 |
| Quintilesd | Quintilesd | Quintilesd | Quintilesd | Quintilesd | Quintilesd | Quintilesd | Quintilesd | Quintilesd |
| Q 1 | 1 (Ref) | | 1 (Ref) | | 1 (Ref) | | 1 (Ref) | |
| Q 2 | 1.3 (1.05~1.62) | .018 | 1.26 (1.01~1.58) | .043 | 1.06 (0.83~1.34) | .641 | 1.01 (0.79~1.28) | .957 |
| Q 3 | 2.39 (1.97~2.92) | <.001 | 1.96 (1.6~2.4) | <.001 | 1.57 (1.26~1.95) | <.001 | 1.55 (1.24~1.94) | <.001 |
| Q 4 | 1.98 (1.61~2.42) | <.001 | 2.04 (1.65~2.51) | <.001 | 1.29 (1.02~1.62) | .032 | 1.19 (0.93~1.51) | .16 |
| p for trend | <.001 | <.001 | <.001 | <.001 | .002 | .002 | .014 | .014 |
There was a linear relationship between VAI and the OR for HF in Model 3 (p for nonlinearity,.151), which used the restricted cubic spline model (Figure 2).
**Figure 2:** *Relationship between VAI and the odds ratio of HF. HF, heart failure; VAI, Visceral adiposity index.*
## Subgroup analysis
Sex, age, race, smoking status, hypertension, DM, CHD, liver disease, serum UA, and eGFR were used as stratification variables to observe the effect size trend, and a Forrest plot of data was generated (Figure 3). All associations were positive in the different subgroups, except for Mexican Americans and participants without hypertension. There were no significant interactions between VAI and sex, age, race, smoking status, hypertension, DM, CHD, liver disease, serum UA, eGFR, or albumin.
**Figure 3:** *Subgroup analysis of the association between VAI and HF. CI, confidence interval; eGFR, estimated glomerular filtration rate; HF, heart failure; OR, odds ratio; VAI, Visceral adiposity index.*
## DISCUSSION
Using data from a representative national sample of middle‐age and elderly populations in the US, we found that VAI was associated with HF after adjustment for other covariates, exhibiting a near linear dose–response relationship. Subgroup analysis makes it possible to better understand VAI and HF in different populations, suggesting that the direction of relationships between the VAI and HF in different subgroups was consistent with that in the total study population.
The overall prevalence of HF in our study was approximately $2.4\%$, which is consistent with previously published data. 3, 4 Previous studies have reported an association between obesity and HF, 26, 27, 28 and determined that obesity is the main risk factor for hypertension, CVD, and left ventricular hypertrophy (LVH), which are strong risk factors for the development of HF. 29, 30 In the Framingham Heart Study, which included 5881 participants, after adjusting for some risk factors, each unit increase in BMI increased the incidence of HF by $5\%$ in males and $7\%$ in females, and the risk for HF increased across the BMI range. 26 The Physicians’ Health Study, which included 21 094 males (mean age, 53 years) without known CHD at baseline, demonstrated that every 1 kg/m2 increase in BMI was associated with an $11\%$ increase in HF risk, and obese participants had a $180\%$ increase in HF risk. 31 A study involving 59 178 Finnish participants 25–74 years of age who had no HF at baseline reported that the multivariable adjusted risk ratio for HF was highest in the high BMI group (>30 kg/m2) among men and women, and abdominal obesity was associated with a greater risk for HF in men and women. 28 However, obesity includes both overall and abdominal obesity. Different obesity phenotypes may lead to different incidence, mortality, and treatment outcomes of HF. 32 BMI can be used as an indicator of overall obesity; however, for individuals with simple abdominal obesity, BMI may not be the best indicator, and there is even “normal weight obesity” in the population. Even for normal‐weight individuals, the risk for CVD may be higher among those with high WC. WC, an indicator of abdominal fat, is associated with cardiac metabolic diseases and CVD and can predict mortality. 33, 34 Therefore, other indicators are needed to evaluate abdominal obesity, among which VAT is one. Rao et al. demonstrated that VAT was independently associated with hospitalization for HFpEF in individuals without baseline CVD. 35 Selvaraj et al. suggested that patients with HFpEF had significantly increased pericardial and subcutaneous fat thicknesses compared to patients without HF. 12 Sorimachi et al. reported that female HFpEF patients had a higher VAT, and the accumulation of excess VAT played an important role in the pathophysiology of female HFpEF patients. 11 In these studies, the VAT was measured using abdominal CT or MRI. These methods are accurate but have high cost and low efficiency, and are rarely used in clinics. VAI can be calculated by measuring WC, height, weight, and TG and HDL‐c levels in the blood. The clinical operation is simple and the data are easy to obtain. Previous studies have concluded that VAI is associated with diabetes, hyperuricemia, metabolic syndrome, hypertension, atherosclerosis, and vascular calcification. 13, 14, 15, 16, 17, 18 However, to the best of our knowledge, results from previous studies investigating the correlation between VAI and HF are limited. A cohort study of 116 patients 35–80 years of age, who were hospitalized for aggravated HF between 2011 and 2013, demonstrated that VAI may be a good predictor of mortality in patients with ischemic heart failure, and that patients with higher VAI had a better survival prognosis. 36 However, this study only examined the relationship between VAI and mortality in patients with ischemic heart failure, and did not study the relationship between VAI and the prevalence of HF.
As expected, there was a linear relationship between VAI and the OR for HF. In terms of pathophysiological mechanism, VAT can induce cardiomyocyte hypertrophy, lead to myocardial fibrosis, and activate inflammatory pathways related to macrophage infiltration and cytokine gene expression. Excessive VAT accumulation may lead to higher circulating blood volume and more local and systemic atherogenic inflammatory factors. It may also increase the risk for stroke, increase heart wall pressure and myocardial injury, lead to left ventricular remodeling and, eventually, cause HF. 37, 38, 39 To the best of our knowledge, this is the first study to examine the association between VAI and HF in a large and representative national sample of adults in the US. Our study had the advantages of rigorous study protocols and quality controls, a large representative sample, and available data on many vital covariates by integrating the NHANES data. Nevertheless, this study had some limitations. First, the NHANES does not collect echocardiography and N‐terminal pro brain natriuretic peptide (NT‐proBNP) data from participants. Participants with HF were defined as those with self‐reported physician‐diagnosed HF. The same situation also occurs in hypertension, DM, anemia, liver disease, CHD, kidney disease, and a history of heart attack. Second, this was a cross‐sectional study that did not include follow‐up data. The changes in VAI and the risk for HF over time are unclear. Our study design did not permit identification of a causal association between VAI and HF during the study period. Third, it did not distinguish between the types of HF in participants and could not evaluate whether VAI has a different relationship with different types of HF.
## CONCLUSION
Results of the present study revealed that VAI was independently associated with the risk for HF. More simply stated, noninvasive scores of visceral adiposity permitted a simple noninvasive “one shot” assessment of HF risk(s). In view of the increasing prevalence and enormous health burden of HF, individuals with high VAI warrant greater attention to prevent HF. As such, its potential use as a novel marker of HF risk merits further investigation.
## CONFLICT OF INTEREST
The authors declare no conflict of interest.
## DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in the National Health and Nutrition Examination Survey (NHANES) at https://www.cdc.gov/nchs/nhanes/index.htm.
## References
1. McDonagh TA, Metra M, Adamo M. **2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure**. *Eur Heart J* (2021) **42** 3599-3726. PMID: 34447992
2. **Chinese guidelines for the diagnosis and treatment of heart failure 2018**. *Zhonghua Xin Xue Guan Bing Za Zhi* (2018) **46** 760-789. PMID: 30369168
3. Benjamin EJ, Muntner P, Alonso A. **Heart disease and stroke statistics‐2019 update: a report from the American Heart Association**. *Circulation* (2019) **139** e56-e528. PMID: 30700139
4. Benjamin EJ, Blaha MJ, Chiuve SE. **Heart disease and stroke statistics‐2017 update: a report from the American Heart Association**. *Circulation* (2017) **135** e146-e603. PMID: 28122885
5. Hao G, Wang X, Chen Z. **Prevalence of heart failure and left ventricular dysfunction in China: the China Hypertension Survey, 2012–2015**. *Eur J Heart Fail* (2019) **21** 1329-1337. PMID: 31746111
6. Lewis GA, Schelbert EB, Williams SG. **Biological phenotypes of heart failure with preserved ejection fraction**. *JACC* (2017) **70** 2186-2200. PMID: 29050567
7. Ather S, Chan W, Bozkurt B. **Impact of noncardiac comorbidities on morbidity and mortality in a predominantly male population with heart failure and preserved versus reduced ejection fraction**. *JACC* (2012) **59** 998-1005. PMID: 22402071
8. Dunlay SM, Roger VL, Redfield MM. **Epidemiology of heart failure with preserved ejection fraction**. *Nat Rev Cardiol* (2017) **14** 591-602. PMID: 28492288
9. Donataccio MP, Vanzo A, Bosello O. **Obesity paradox and heart failure**. *Eat Weight Disord* (2021) **26** 1697-1707. PMID: 32851592
10. Rao VN, Fudim M, Mentz RJ, Michos ED, Felker GM. **Regional adiposity and heart failure with preserved ejection fraction**. *Eur J Heart Fail* (2020) **22** 1540-1550. PMID: 32619081
11. Sorimachi H, Obokata M, Takahashi N. **Pathophysiologic importance of visceral adipose tissue in women with heart failure and preserved ejection fraction**. *Eur Heart J* (2021) **42** 1595-1605. PMID: 33227126
12. Selvaraj S, Kim J, Ansari BA. **Body composition, natriuretic peptides, and adverse outcomes in heart failure with preserved and reduced ejection fraction**. *JACC Cardiovasc Imaging* (2021) **14** 203-215. PMID: 32950445
13. Bagyura Z, Kiss L, Lux Á. **Association between coronary atherosclerosis and visceral adiposity index**. *Nutr Metab Cardiovasc Dis* (2020) **30** 796-803. PMID: 32127334
14. Nusrianto R, Tahapary DL, Soewondo P. **Visceral adiposity index as a predictor for type 2 diabetes mellitus in Asian population: a systematic review**. *Diabetes Metab Syndr* (2019) **13** 1231-1235. PMID: 31336469
15. Huang X, Jiang X, Wang L. **Visceral adipose accumulation increased the risk of hyperuricemia among middle‐aged and elderly adults: a population‐based study**. *J Transl Med* (2019) **17** 341. PMID: 31601236
16. Jung JY, Ryoo J‐H, Oh C‐M. **Visceral adiposity index and longitudinal risk of incident metabolic syndrome: Korean genome and epidemiology study (KoGES)**. *Endocr J* (2020) **67** 45-52. PMID: 31611471
17. Fan Y, He D, Liu S, Qiao Y, Gao H, Xin L. **Association between visceral adipose index and risk of hypertension in a middle‐aged and elderly Chinese population**. *Nutr Metab Cardiovasc Dis* (2021) **31** 2358-2365. PMID: 34090774
18. Son DH, Ha HS, Lee HS. **Association of the new visceral adiposity index with coronary artery calcification and arterial stiffness in Korean population**. *Nutr Metab Cardiovasc Dis* (2021) **31** 1774-1781. PMID: 33975738
19. Fryar CD, Ostchega Y, Hales CM, Zhang G, Kruszon‐Moran D. **Hypertension prevalence and control among adults: United States, 2015‐2016**. *NCHS Data Brief* (2017) 1-8
20. McClure ST, Schlechter H, Oh S. **Dietary intake of adults with and without diabetes: results from NHANES 2013‐2016**. *BMJ Open Diabetes Res Care* (2020) **8**
21. Levey AS, Stevens LA, Schmid CH. **A new equation to estimate glomerular filtration rate**. *Ann Intern Med* (2009) **150** 604-612. PMID: 19414839
22. Gojanovic M, Holloway‐Kew KL, Hyde NK. **The dietary inflammatory index is associated with low muscle mass and low muscle function in older Australians**. *Nutrients* (2021) **13** 1166. PMID: 33916033
23. Liu Z, Liu H, Deng Q. **Association between dietary inflammatory index and heart failure: results from NHANES (1999–2018)**. *Front Cardiovasc Med* (2021) **8**. PMID: 34307508
24. Martens P, Verluyten L, Van de Broek H. **Determinants of maximal dose titration of sacubitril/valsartan in clinical practice**. *Acta Cardiol* (2021) **76** 20-29. PMID: 31697901
25. Adamstein NH, MacFadyen JG, Rose LM. **The neutrophil‐lymphocyte ratio and incident atherosclerotic events: analyses from five contemporary randomized trials**. *Eur Heart J* (2021) **42** 896-903. PMID: 33417682
26. Kenchaiah S, Evans JC, Levy D. **Obesity and the risk of heart failure**. *N Engl J Med* (2002) **347** 305-313. PMID: 12151467
27. Bozkurt B, Aguilar D, Deswal A. **Contributory risk and management of comorbidities of hypertension, obesity, diabetes mellitus, hyperlipidemia, and metabolic syndrome in chronic heart failure: a scientific statement from the American Heart Association**. *Circulation* (2016) **134** e535-e578. PMID: 27799274
28. Hu G, Jousilahti P, Antikainen R, Katzmarzyk PT, Tuomilehto J. **Joint effects of physical activity, body mass index, waist circumference, and waist‐to‐hip ratio on the risk of heart failure**. *Circulation* (2010) **121** 237-244. PMID: 20048205
29. Lavie CJ, Laddu D, Arena R, Ortega FB, Alpert MA, Kushner RF. **Reprint of: healthy weight and obesity prevention**. *JACC* (2018) **72** 3027-3052. PMID: 30522635
30. Elagizi A, Kachur S, Lavie CJ. **An overview and update on obesity and the obesity paradox in cardiovascular diseases**. *Prog Cardiovasc Dis* (2018) **61** 142-150. PMID: 29981771
31. Kenchaiah S, Sesso HD, Gaziano JM. **Body mass index and vigorous physical activity and the risk of heart failure among men**. *Circulation* (2009) **119** 44-52. PMID: 19103991
32. Alagiakrishnan K, Banach M, Ahmed A, Aronow WS. **Complex relationship of obesity and obesity paradox in heart failure—higher risk of developing heart failure and better outcomes in established heart failure**. *Ann Med* (2016) **48** 603-613. PMID: 27427379
33. Piché M‐E, Poirier P, Lemieux I, Després JP. **Overview of epidemiology and contribution of obesity and body fat distribution to cardiovascular disease: an update**. *Prog Cardiovasc Dis* (2018) **61** 103-113. PMID: 29964067
34. Sahakyan KR, Somers VK, Rodriguez‐Escudero JP. **Normal‐weight central obesity: implications for total and cardiovascular mortality**. *Ann Intern Med* (2015) **163** 827-835. PMID: 26551006
35. Rao VN, Zhao D, Allison MA. **Adiposity and incident heart failure and its subtypes**. *JACC Heart Fail* (2018) **6** 999-1007. PMID: 30316935
36. Vogel P, Stein A, Marcadenti A. **Visceral adiposity index and prognosis among patients with ischemic heart failure**. *Sao Paulo Med J* (2016) **134** 211-218. PMID: 27191246
37. Murase T, Hattori T, Ohtake M. **Cardiac remodeling and diastolic dysfunction in DahlS.Z‐Lepr(fa)/Lepr(fa) rats: a new animal model of metabolic syndrome**. *Hypertension Res* (2012) **35** 186-193
38. Rodriguez Flores M, Aguilar Salinas C, Piché M‐E, Auclair A, Poirier P. **Effect of bariatric surgery on heart failure**. *Expert Rev Cardiovasc Ther* (2017) **15** 567-579. PMID: 28714796
39. Neeland IJ, Gupta S, Ayers CR. **Relation of regional fat distribution to left ventricular structure and function**. *Circ Cardiovasc Imaging* (2013) **6** 800-807. PMID: 23929898
|
---
title: 'Diabetic and stress‐induced hyperglycemia in spontaneous intracerebral hemorrhage:
A multicenter prospective cohort (CHEERY) study'
authors:
- Shaoli Chen
- Yan Wan
- Hongxiu Guo
- Jing Shen
- Man Li
- Yuanpeng Xia
- Lei Zhang
- Zhou Sun
- Xiaolu Chen
- Gang Li
- Quanwei He
- Bo Hu
journal: CNS Neuroscience & Therapeutics
year: 2022
pmcid: PMC10018104
doi: 10.1111/cns.14033
license: CC BY 4.0
---
# Diabetic and stress‐induced hyperglycemia in spontaneous intracerebral hemorrhage: A multicenter prospective cohort (CHEERY) study
## Abstract
In patients with hyperglycemia, stress‐induced hyperglycemia (SIH) was associated with a higher risk of pulmonary infection [risk ratios (RR): 1.477, $95\%$ confidence interval (CI): 1.004–2.172] and 30‐day (RR: 1.068, $95\%$ CI: 1.009–1.130) and 90‐day mortality after intracerebral hemorrhage (ICH) (RR: 1.060, $95\%$ CI: 1.000–1.124). Stress‐induced hyperglycemia is a sensitive predictor of the risk of pulmonary infection and all‐cause death after ICH.
### Introduction
Admission hyperglycemia is a common finding after spontaneous intracerebral hemorrhage (ICH) secondary to pre‐existing diabetes mellitus (DM) or stress‐induced hyperglycemia (SIH). Studies of the causal relationship between SIH and ICH outcomes are rare.
### Aim
We aimed to identify whether SIH or pre‐existing DM was the cause of admission hyperglycemia associated with ICH outcomes.
### Methods
Admission glycosylated hemoglobin (HbA1c), glucose levels, and comorbidity data from the prospective, multicenter cohort, Chinese Cerebral Hemorrhage: Mechanisms and Intervention Study (CHEERY), were collected and analyzed. According to different admission blood glucose and HbA1c levels, patients were divided into nondiabetic normoglycemia (NDN), diabetic normoglycemia (DN), diabetic hyperglycemia (DH), and SIH groups. Modified Poisson regression models were used to analyze ICH outcomes in the different groups.
### Results
In total, 1372 patients were included: 388 patients with admission hyperglycemia, 239 with DH, and 149 with SIH. In patients with hyperglycemia, SIH was associated with a higher risk of pulmonary infection [risk ratios (RR): 1.477, $95\%$ confidence interval (CI): 1.004–2.172], 30‐day (RR: 1.068, $95\%$ CI: 1.009–1.130) and 90‐day mortality after ICH (RR: 1.060, $95\%$ CI: 1.000–1.124).
### Conclusions
Admission hyperglycemia is a common finding after ICH, and SIH is a sensitive predictor of the risk of pulmonary infection and all‐cause death after ICH.
## INTRODUCTION
Spontaneous intracerebral hemorrhage (ICH) is the second most common and fatal type of stroke, 1 with a 30‐day mortality rate of up to $40\%$. 2 *Hyperglycemia is* always observed at admission when one suffers from ICH and is considered a predictor of poor outcomes in some studies, 3, 4, 5, 6, 7 which were challenged by other studies. 8, 9, 10 This discrepancy may be due to the causes of admission hyperglycemia, which could either be stress‐induced hyperglycemia (SIH) or pre‐existing diabetes mellitus (DM), which were not well differentiated in those studies. SIH is a transient hyperglycemic condition caused by acute diseases and is usually restricted to patients without DM, 11 and is defined as admission blood glucose ≥7.8 mmol/L, 12 which has been found to be directly responsive to the severity and predictable to the poor outcomes of ICH. 13, 14, 15, 16 However, there are still some shortcomings in these well‐designed reports: [1] up to $\frac{1}{3}$ of patients with occult diabetes may clutter the results 17 as some previous studies defined SIH in non‐DM patients only by disease history and not by measuring glycosylated hemoglobin A1c (HbA1c), 15, 16, 18 a reliable measure of the mean glucose concentration over the previous 3‐ to 4‐month time period 19, 20; [2] the few studies pertaining to SIH in patients with ICH were all relatively small populations (the largest included 328 patients) 9, 13, 14, 16; and [3] some were retrospective studies. 14 Therefore, we sought to determine whether SIH or pre‐existing DM was associated with poor outcomes of ICH in a large, multicentric, prospective cohort of patients with ICH having a long follow‐up.
## Participants and design
We analyzed data from the Chinese Cerebral Hemorrhage: Mechanisms and Intervention study (CHEERY) (registration number of China Clinical Trial Registration Center: ChiCTR1900020872, http://www.chictr.org.cn). The study protocol and data collection were conducted in strict accordance with the Declaration of Helsinki and approved by the Research Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China (ethical approval number: 2018‐S485). All patients provided informed consent before recruitment. Consecutive patients presenting with spontaneous ICH were admitted to 31 hospital centers between December 2018 and June 2021. Patients aged ≥18 years, with spontaneous ICH confirmed using computed tomography (CT), and within 24 h of onset were recruited for this study. Patients who met the following criteria were excluded: [1] hemorrhages secondary to trauma, primary subarachnoid hemorrhage, hemorrhagic conversion from ischemic stroke, and thrombolysis; [2] lack of data on admission glucose or HbA1c levels; and [3] unavailability of imaging and baseline information.
## Data collection and follow‐up
Relevant information was collected through the electronic medical record system: age, sex, disease history, admission systolic blood pressure (SBP), admission blood glucose, HbA1c level, time from symptom onset to admission, and surgical treatment. Admission blood glucose and HbA1c were measured after an overnight 8‐h fast. Baseline neurological deficits were assessed using the Glasgow Coma Scale (GCS). 21 Hematoma localization and intraventricular hemorrhage (IVH) were recorded according to the first head CT data after admission, and hematoma volume was calculated using the ABC/2 formula. 22, 23 Pulmonary infection was diagnosed by two well‐trained and experienced neurologists according to the modified Centers for Disease Control and Prevention criteria, in combination with the patient's clinical symptoms, laboratory, and radiological examinations within 1 week after ICH. 24, 25 Stroke neurologists conducted face‐to‐face or telephonic interviews with the enrolled patients on day 30 and 90 after the onset of ICH. The degree of functional recovery was evaluated according to a modified Rankin Scale (mRS) score, 23 and poor outcome was defined as mRS score of 3–6. 26, 27, 28
## Definition of subgroup
According to the latest consensus from the American Association of Clinical Endocrinologists and American Diabetes Association, 12 SIH was defined as having no DM history, HbA1c <$6.5\%$, and admission blood glucose ≥7.8 mmol/L. If the admission blood glucose <7.8 mmol/L, it was defined as nondiabetic normoglycemia (NDN). Diabetic hyperglycemia (DH) was defined as having a DM history or HbA1c ≥$6.5\%$, and admission blood glucose ≥7.8 mmol/L; and if admission blood glucose <7.8 mmol/L, it was defined as diabetic normoglycemia (DN).
## Statistical analysis
SPSS statistical software (version 26.0, SPSS Corporation, Chicago) was used to analyze the data, and the statistical significance was set at $p \leq 0.05.$ Categorical variables are expressed as percentages, and the χ2 test or Fisher's exact test was used to compare the differences between groups. The Kolmogorov–Smirnov test (KS test) for normality was used to assess the data distribution of continuous variables. Normally distributed variables are expressed as mean ± standard deviation (SD), and two groups were compared using Student's t‐test. Non‐normally distributed variables were expressed as median and interquartile ranges (first and third quartiles), and the Mann–Whitney U test or Kruskal–Wallis H test was used to compare differences between groups. In univariate analysis, variables reaching $p \leq 0.05$ were considered to have significant differences. Finally, age, sex, history of hypertension, time from symptom onset to admission, infratentorial hemorrhage, hematoma volume, GCS, IVH, and surgical treatment were included in the regression model in the multivariate analysis. Modified Poisson regression models were used to calculate risk ratios (RRs) 29 and the associated $95\%$ confidence intervals (CIs) for the association between different groups and the outcomes of interest, and figures were drawn using GraphPad Prism 9.0.
## Patient characteristics
From December 2018 to June 2021, 4248 patients with ICH were enrolled in the CHEERY study. After excluding 25 patients with non‐spontaneous ICH, 778 patients who presented with symptoms exceeding 24 h, 393 patients who lacked admission blood glucose data, and 1680 patients with missing HbA1c data, a total of 1372 patients with complete study data were finally included in the analysis. Among them, 984 ($71.7\%$) patients presented with normoglycemia, and 388 ($28.3\%$) had a hyperglycemic status on admission. According to the definitions for the different glucose classifications mentioned above, 826 patients were classified as having NDN, 158 patients were classified as having DN, 149 patients were classified as having SIH, and 239 patients were classified as having DH (Figure 1).
**FIGURE 1:** *Breakdown of the study population with identification of four groups of patients based on admission blood glucose, HbA1c, and history of DM. DH, diabetic hyperglycemia; DN, diabetic normoglycemia; NDN, nondiabetic normoglycemia; SIH, stress‐induced hyperglycemia.*
## Characteristics and outcomes of patients with different admission blood glucose levels
The baseline characteristics and outcomes of ICH patients with hyperglycemia or normoglycemia are compared in Table 1. The mean age of the included patients was 62.3 ± 11.6 years, 904 ($65.9\%$) patients were male, and 281 ($20.5\%$) patients had a history of DM. The average age, sex ratio, admission SBP, hematoma volume, and length of hospitalization were not significantly different between the hyperglycemic and normoglycemic groups. In the hyperglycemia group, the median of admission blood glucose was 9.5 (8.5–11.5) mmol/L, and the mean HbA1c was $6.4\%$, both of which were higher than those in the normoglycemic group ($p \leq 0.001$). Patients with hyperglycemia were more likely to have hypertension ($76.5\%$ vs. $67.1\%$, $$p \leq 0.002$$), DM ($47.2\%$ vs. $10.0\%$, $p \leq 0.001$), shorter time from symptom onset (3 vs. 3.5, $$p \leq 0.012$$), infratentorial hemorrhage ($18.3\%$ vs. $9.5\%$, $p \leq 0.001$), IVH ($22.9\%$ vs. $12.0\%$, $p \leq 0.001$), lower GCS score (13 vs. 14, $p \leq 0.001$), more frequent surgical treatment ($26\%$ vs. $13.1\%$, $p \leq 0.001$), and higher risk of pulmonary infection ($20.1\%$ vs. $10.3\%$, $p \leq 0.001$). In addition, patients with hyperglycemia were more likely to have poor outcomes, which were defined as mRS 3–6, on day 30 ($72.5\%$ vs. $57.4\%$, $p \leq 0.001$) and 90 ($63.2\%$ vs. $46.2\%$, $p \leq 0.001$), as well as higher risk of mortality during hospitalization ($6.4\%$ vs. $2.3\%$, $p \leq 0.001$) and 30 days($21.4\%$ vs. $7.8\%$, $p \leq 0.001$) and 90 days ($23.8\%$ vs. $11.1\%$, $p \leq 0.001$) after ICH onset compared with those with normoglycemia.
**TABLE 1**
| Unnamed: 0 | All subjects (N = 1372) | Admission blood glucose <7.8 (N = 984) | Admission blood glucose ≥7.8 (N = 388) | p‐value |
| --- | --- | --- | --- | --- |
| Age, y, Mean ± SD | 62.3 ± 11.6 | 62.2 ± 11.5 | 62.6 ± 11.9 | 0.158 |
| Male, (n, %) | 904 (65.9) | 651 (66.2) | 253 (65.2) | 0.738 |
| Hypertension, (n, %) | 957 (69.8) | 660 (67.1) | 297 (76.5) | 0.002 |
| DM, (n, %) | 281 (20.5) | 98 (10.0) | 183 (47.2) | <0.001 |
| Time from symptom onset to admission, h, median (IQR) | 3 (2, 7.5) | 3.5 (2, 8) | 3 (2, 6) | 0.012 |
| Admission SBP ≥140 mmHg, (n, %) | 1219 (88.8) | 869 (88.3) | 350 (90.2) | 0.316 |
| HBA1c ≥6.5%, (n, %) | 397 (28.9) | 158 (16.1) | 239 (61.6) | <0.001 |
| Admission blood glucose, mmol/L, median (IQR) | 6.5 (5.5, 8.1) | 5.9 (5.3, 6.6) | 9.5 (8.5, 11.5) | <0.001 |
| HBA1c, %, median (IQR) | 5.7 (5.3, 6.2) | 5.6 (5.3, 6.0) | 6.4 (5.6, 8.0) | <0.001 |
| Infratentorial hemorrhage, (n, %) | 159 (12.0) | 90 (9.5) | 69 (18.3) | <0.001 |
| Hematoma volume, ml, median (IQR) | 10.0 (4.6, 24.0) | 10.0 (4.5, 20.0) | 10.0 (4.3, 27.0) | 0.124 |
| IVH, (n, %) | 207 (15.1) | 118 (12.0) | 89 (22.9) | <0.001 |
| GCS, median (IQR) | 14 (12, 15) | 14 (13, 15) | 13 (9, 15) | <0.001 |
| Surgical treatment, (n, %) | 230 (16.8) | 129 (13.1) | 101 (26%) | <0.001 |
| Length of hospitalization, days, median (IQR) | 16.0 (11.0, 22.0) | 16.0 (12.0, 21.0) | 16.0 (9.0, 23.0) | 0.685 |
| Pulmonary infection, (n, %) | 179 (13.0) | 101 (10.3) | 78 (20.1) | <0.001 |
| 30‐day mRS (3–6), (n, %) | 821/1331 a (62.7) | 547/953 a (57.4) | 274/378 a (72.5) | <0.001 |
| 90‐day mRS (3–6), (n, %) | 679/1331 a (51.0) | 440/953 a (46.2) | 239/378 a (63.2) | <0.001 |
| In‐hospital death, (n, %) | 48 (3.5) | 23 (2.3) | 25 (6.4) | <0.001 |
| 30‐day death, (n, %) | 155/1331 a (11.6) | 74/953 a (7.8) | 81/378 a (21.4) | <0.001 |
| 90‐day death, (n, %) | 196/1331 a (14.7) | 106/953 a (11.1) | 90/378 a (23.8) | <0.001 |
## Multivariate analysis for association between hyperglycemia and patient outcomes
The multivariate analysis of the association between hyperglycemia and ICH outcomes is shown in Table 2. The following variables were used in the adjusted models: age, sex, history of hypertension, time from symptom onset to admission, infratentorial hemorrhage, IVH, GCS, and surgical treatment. The results revealed that hyperglycemia was independently associated with increased risk of pulmonary infection (RR: 1.403, $95\%$ CI: 1.066–1.846), poor 30‐day (RR: 1.038, $95\%$ CI: 1.003–1.074) and 90‐day outcomes (RR: 1.145, $95\%$ CI: 1.025–1.279), and 30‐day mortality due to ICH (RR: 1.041, $95\%$ CI 1.005–1.077), while hyperglycemia was not associated with mortality during hospitalization and 90 days after ICH (Table 2).
**TABLE 2**
| Unnamed: 0 | Admission blood glucose ≥7.8 | Admission blood glucose ≥7.8.1 | Admission blood glucose ≥7.8.2 |
| --- | --- | --- | --- |
| | RR | 95% CI | p value |
| Pulmonary infection | 1.403 | 1.066–1.846 | 0.016 |
| 30‐day mRS (3–6) | 1.038 | 1.003–1.074 | 0.034 |
| 90‐day mRS (3–6) | 1.145 | 1.025–1.279 | 0.016 |
| In‐hospital death | 1.138 | 0.577–2.242 | 0.710 |
| 30‐day death | 1.041 | 1.005–1.077 | 0.025 |
| 90‐day death | 1.029 | 0.991–1.069 | 0.132 |
## Characteristics and outcomes of ICH patients with different glucose classifications
Hyperglycemia at admission may be caused by SIH or DM. According to the HbA1c level, glucose level, and DM history, patients were classified into four groups: NDN, DN, DH, and SIH, as previously mentioned. The characteristics and outcomes of the four classifications are presented in Table 3. Overall, no significant differences in sex ($$p \leq 0.446$$) and age ($$p \leq 0.353$$) among the four groups were found. Compared with the other groups, patients with DH had a significantly larger hematoma volume ($p \leq 0.001$) and a higher incidence of IVH ($p \leq 0.001$). Patients with SIH had a significantly higher incidence of infratentorial ICH ($p \leq 0.001$), lower GCS score ($p \leq 0.001$), and a higher proportion of surgical treatment ($p \leq 0.001$) and higher risk of pulmonary infection ($p \leq 0.001$). In addition, patients with SIH were more likely to have poor outcomes (mRS 3–6) at 30 and 90 days after ICH ($p \leq 0.001$), and the risk of their death was higher during hospitalization, at 30 and 90 days after ICH, than the other groups ($p \leq 0.001$).
**TABLE 3**
| Unnamed: 0 | Admission blood glucose<7.8 | Admission blood glucose<7.8.1 | Admission blood glucose ≥7.8 | Admission blood glucose ≥7.8.1 | p value |
| --- | --- | --- | --- | --- | --- |
| | NDN (N = 826) | DN (N = 158) | DH (N = 239) | SIH (N = 149) | p value |
| Age, y, Mean ± SD | 62.0 ± 11.6 | 63.1 ± 10.8 | 63.2 ± 11.5 | 61.6 ± 12.4 | 0.353 |
| Male, (n, %) | 545 (66.0) | 106 (67.1) | 163 (68.2) | 90 (60.4) | 0.446 |
| Hypertension, (n, %) | 529 (64.0) | 131 (82.9) | 188 (78.7) | 109 (73.2) | <0.001 |
| DM, (n, %) | 0 (0) | 98 (62.0) | 183 (76.6) | 0 (0) | <0.001 |
| Time from symptom onset to admission (h), median (IQR) | 3.5 (2, 8) | 4 (2, 10.5) | 3 (2, 7) | 3 (2, 5) | 0.038 |
| Admission SBP ≥140 mmHg, (n, %) | 731 (88.5) | 138 (87.3) | 220 (92.1) | 130 (87.2) | 0.345 |
| Admission blood glucose, mmol/L, median (IQR) | 5.9 (5.2, 6.5) | 6.4 (5.4, 7.0) | 10.8 (9.1, 13.8) | 8.8 (8.3, 9.9) | <0.001 |
| HBA1c, %, median (IQR) | 5.6 (5.2, 5.9) | 6.7 (6.5, 7.4) | 7.8 (6.8, 9.1) | 5.6 (5.2, 6.0) | <0.001 |
| Infratentorial hemorrhage, (n, %) | 78 (9.8) | 12 (7.7) | 36 (15.4) | 33 (23.1) | <0.001 |
| Hematoma volume, (ml), median (IQR) | 10.0 (4.1, 20.0) | 10.0 (5.0, 28.9) | 12.8 (5.0, 35.0) | 11.8 (4.9, 30.5) | <0.001 |
| IVH, (n, %) | 98 (11.9) | 20 (12.7) | 58 (24.3) | 31 (20.8) | <0.001 |
| GCS, median (IQR) | 14 (13, 15) | 14 (11, 15) | 13 (9, 15) | 12 (7, 15) | <0.001 |
| Surgical treatment, (n, %) | 103 (12.5) | 26 (16.5) | 61 (25.5) | 40 (26.8) | <0.001 |
| Length of hospitalization, (days), median (IQR) | 17.0 (12.0, 22.0) | 15.0 (11.0, 20.0) | 17.0 (9.8, 26.0) | 16.0 (7.3, 21.0) | 0.051 |
| Pulmonary infection, (n, %) | 84 (10.2) | 17 (10.8) | 46 (19.2) | 32 (21.5) | <0.001 |
| 30‐day mRS (3–6), (n, %) | 457/799 (57.2) | 90/154 (58.4) | 169/235 (71.9) | 105/143 (73.4) | <0.001 |
| 90‐day mRS (3–6), (n, %) | 364/799 (45.2) | 76/154 (50.0) | 147/235 (63.4) | 92/143 (64.3) | <0.001 |
| In‐hospital death, (n, %) | 15 (1.8) | 8 (5.1) | 14 (5.9) | 11 (7.4) | <0.001 |
| 30‐day death, (n, %) | 51/799 (6.4) | 23/154 (14.9) | 43/235 (18.3) | 38/143 (26.6) | <0.001 |
| 90‐day death, (n, %) | 78/799 (9.8) | 28/154 (18.2) | 48/235 (20.4) | 42/143 (29.4) | <0.001 |
## Multivariate analysis for outcomes of patients with ICH and different glucose classifications
The results of the multivariate regression models (adjusted for age, sex, hypertension history, time from symptom onset to admission, infratentorial hemorrhage, hematoma volume, GCS, IVH, and surgical treatment) are shown in Table 4. Compared with patients with NDN, the risks of poor outcomes and death in patients with DN and DH did not increase.
**TABLE 4**
| RR (95% CI) | NDN | DN | p value | DH | p value.1 | SIH | p value.2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Pulmonary infection | Ref | 0.783 (0.474, 1.293) | 0.339 | 1.195 (0.853, 1.674) | 0.301 | 1.477 (1.004, 2.172) | 0.047 |
| 30‐day mRS (3–6) | Ref | 0.974 (0.924, 1.028) | 0.341 | 1.022 (0.981, 1.065) | 0.299 | 1.036 (0.987, 1.088) | 0.155 |
| 90‐day mRS (3–6) | Ref | 1.011 (0.846, 1.207) | 0.907 | 1.118 (0.979, 1.277) | 0.101 | 1.136 (0.975, 1.324) | 0.103 |
| In‐hospital death | Ref | 1.435 (0.593, 3.474) | 0.424 | 1.177 (0.513, 2.701) | 0.701 | 1.251 (0.506, 3.092) | 0.628 |
| 30‐day death | Ref | 1.021 (0.978, 1.066) | 0.343 | 1.020 (0.980, 1.062) | 0.34 | 1.068 (1.009, 1.130) | 0.022 |
| 90‐day death | Ref | 1.024 (0.977, 1.073) | 0.332 | 1.005 (0.961, 1.050) | 0.838 | 1.060 (1.000, 1.124) | 0.049 |
Patients with SIH had a higher risk of pulmonary infection (RR: 1.477, $95\%$ CI: 1.004–2.172), 30‐ (RR: 1.068, $95\%$ CI: 1.009–1.130) and 90‐day mortality (RR: 1.060, $95\%$ CI: 1.000–1.124) (Figure 2 and 3).
**FIGURE 2:** *Multivariate regression analysis of ICH poor outcomes of different subgroups. Adjusted: Adjusted for age, sex, time from symptom onset to admission, infratentorial hemorrhage, hematoma volume, GCS, IVH, hypertension history, and surgical treatment. DH, diabetic hyperglycemia; DN, diabetic normoglycemia; SIH, stress‐induced hyperglycemia.* **FIGURE 3:** *Multivariate regression analysis of ICH mortality of different subgroups. Adjusted: Adjusted for age, sex, time from symptom onset to admission, infratentorial hemorrhage, hematoma volume, GCS, IVH, hypertension history, and surgical treatment. DH, diabetic hyperglycemia; DN, diabetic normoglycemia; SIH, stress‐induced hyperglycemia.*
## DISCUSSION
In this study, we found that hyperglycemia was associated with poor prognosis and an increased risk of death after ICH onset. After multivariate regression analysis, admission hyperglycemia was an independent risk factor for pulmonary infection, poor 30‐ and 90‐day prognosis, and 30‐day mortality but did not increase the risk of 90‐day mortality. Compared with patients with NDN, DH did not increase the risk of poor outcome and mortality, whereas SIH was an independent risk factor for pulmonary infection and 30‐ and 90‐day death after ICH.
Previous studies have shown that ICH is often accompanied by hyperglycemia, and the association between admission hyperglycemia and the risk of death and adverse outcomes of ICH has been concerning. 3, 4, 5, 6, 7, 8, 9, 10 Hyperglycemia leads to peripheral nerve injury, 30 and hematoma perihematomal cell death 31 and decreased autophagy, 32 increasing the production of superoxide caused by tissue plasminogen activator, 33 and increasing the plasma kallikrein to promote the expansion of hematoma. 34 These processes may be the causes of hyperglycemia, which aggravates the poor outcomes of ICH. The results of INTERACT2 showed that hyperglycemia and DM were independent predictors of poor prognosis in patients with mild‐ to moderate–severe ICH, 7 and a considerable number of studies have shown that admission hyperglycemia is associated with poor outcomes and death risk after ICH. 3, 8, 10, 35, 36, 37, 38, 39, 40 In addition, our study revealed that admission hyperglycemia was associated with an increased risk of poor 30‐ and 90‐day poor outcomes and 30‐day mortality after ICH.
However, contradictions remain. Tetri et al. found that admission hyperglycemia was not an independent risk factor for the prognosis of ICH. 8, 9 Lee et al. found that in patients without diabetes, admission hyperglycemia was independently related to death and poor prognosis after ICH. 15, 18, 41, 42, 43 However, Passero et al. found that DM is associated with a higher risk of death and poor prognosis. 42, 44, 45 In most of the studies, hyperglycemia was usually measured as the admission blood glucose level without considering the pre‐onset blood glucose level or previous diabetes status. It must be noted that admission hyperglycemia could be caused by DM or SIH, which is a transient hyperglycemic condition caused by acute diseases, and some studies may not distinguish SIH from DM. In addition, patients with occult diabetes in some studies may be considered nondiabetic. All of the mentioned reasons may clutter the data and lead to misinterpretation of the results.
HbA1c is a reliable measure of the mean glucose concentration over the last 3‐ to 4‐months, 19, 20 which could recognize occult diabetes. Chu et al analyzed HbA1c values and found that SIH was related to the risk of death and adverse prognosis after ICH. 13, 46, 47 *In this* study, we combined the levels of HbA1c and admission glucose, along with DM history to distinguish between SIH and DM and classified patients into four groups: NDN, DN, DH, and SIH, as mentioned previously. We found that around $\frac{1}{5}$ of patients ($20.5\%$) had a history of DM, consistent with a previous report in which $15\%$–$30\%$ of patients with ICH had DM. 18, 48, 49, 50, 51 Combined with the HbA1c results, 397 ($28.9\%$) patients were d diagnosed with DM in this study, of whom 115 were diagnosed with occult DM ($29.0\%$). Multivariate regression analysis revealed that SIH was associated with a higher risk of mortality at 30‐day and 90‐day after ICH, as well as pulmonary infection after ICH. However, DH did not increase the risk of poor outcomes or mortality. This finding indicates that SIH was more likely to be a risk factor for mortality and poor outcome of ICH than DM.
A previous study recruited 2039 patients with acute stroke, of which 533 ($26.14\%$) were diagnosed with stroke‐associated pneumonia (SAP), and found that the stress hyperglycemia ratio (SHR) was significantly associated with the risk of SAP in patients without diabetes. 52 Additionally, hyperglycemia reportedly leads to the excessive release of inflammatory factors, such as tumor necrosis factor‐α (TNF‐α), interleukin‐1 (IL‐1) and interleukin‐6 (IL‐6), 53, 54 which were significant contributors to pulmonary infection. 55, 56, 57, 58 Simultaneously, increased proinflammatory factors and immunosuppression caused by stroke promote and accelerate the occurrence of pulmonary infection. 59, 60, 61 These findings shed light on the underlying mechanisms by which hyperglycemia increases post‐stroke pulmonary infections.
Evidence suggests that chronic hyperglycemia in patients with DM causes the body to form a self‐protection mechanism, preferentially down‐regulating glucose transporters (GLUT‐1 and GLUT‐3), allowing glucose to enter cells independently of insulin, thus reducing the acute fluctuation of glucose concentration and reducing endothelial cell apoptosis. 62 This phenomenon may be a potential reason for the better outcome of hyperglycemia in patients with DM than in patients without diabetes after ICH.
This study had several limitations. Although the sample size was large, there are few patients with DH and SIH, and only Chinese patients were included. Additionally, due to medical insurance policies and costs, the proportion of patients for whom HbA1c was measured was low, which was the most important reason restricting the inclusion of patients. Furthermore, some other variables known to be associated with poor outcomes of ICH, such as hematoma expansion, were not analyzed in this study due to unavailability of data, which should be further studied in the future.
In conclusion, admission hyperglycemia is common in ICH patients and is associated with poor outcomes, of which SIH may be prioritized over DH to predict the risk of pulmonary infection and 30‐ and 90‐day death due to ICH.
## AUTHOR CONTRIBUTIONS
Shaoli Chen, Yan Wan, and Hongxiu Guo conducted data analysis and wrote the manuscript. Gang Li, Quanwei He, and Bo Hu designed the study and wrote the manuscript. All authors helped with the data collection and literature searches. All the authors have approved this version of the manuscript for publication.
## CONFLICT OF INTEREST
Dr. Bo *Hu is* an editorial board member of CNS Neuroscience and Therapeutics and is a co‐author of this article. To minimize bias, they were excluded from all editorial decision‐making related to the acceptance of this article for publication.
## DATA AVAILABILITY STATEMENT
Data are available on request from the authors.
## References
1. Poon M, Fonville A, Al‐Shahi SR. **Long‐term prognosis after intracerebral haemorrhage: systematic review and meta‐analysis**. *J Neurol Neurosurg Psychiatry* (2014) **85** 660-667. PMID: 24262916
2. van Asch C, Luitse M, Rinkel G, van der Tweel I, Algra A, Klijn C. **Incidence, case fatality, and functional outcome of intracerebral haemorrhage over time, according to age, sex, and ethnic origin: a systematic review and meta‐analysis**. *Lancet Neurol* (2010) **9** 167-176. PMID: 20056489
3. Fogelholm R, Murros K, Rissanen A, Avikainen S. **Admission blood glucose and short term survival in primary intracerebral haemorrhage: a population based study**. *J Neurol Neurosurg Psychiatry* (2005) **76** 349-353. PMID: 15716524
4. Lehmann F, Schenk L, Schneider M. **Predictive relevance of baseline lactate and glucose levels in patients with spontaneous deep‐seated intracerebral hemorrhage**. *Brain Sci* (2021) **11** 633. PMID: 34069048
5. Guo X, Li H, Zhang Z. **Hyperglycemia and mortality risk in patients with primary intracerebral hemorrhage: a meta‐analysis**. *Mol Neurobiol* (2016) **53** 2269-2275. PMID: 25972238
6. Tan X, He J, Li L. **Early hyperglycaemia and the early‐term death in patients with spontaneous intracerebral haemorrhage: a meta‐analysis**. *Intern Med J* (2014) **44** 254-260. PMID: 24372661
7. Saxena A, Anderson C, Wang X. **Prognostic significance of hyperglycemia in acute intracerebral hemorrhage: the INTERACT2 study**. *Stroke* (2016) **47** 682-688. PMID: 26814235
8. Tetri S, Juvela S, Saloheimo P, Pyhtinen J, Hillbom M. **Hypertension and diabetes as predictors of early death after spontaneous intracerebral hemorrhage**. *J Neurosurg* (2009) **110** 411-417. PMID: 19249937
9. Zhao Y, Yang J, Zhao H, Ding Y, Zhou J, Zhang Y. **The association between hyperglycemia and the prognosis of acute spontaneous intracerebral hemorrhage**. *Neurol Res* (2017) **39** 152-157. PMID: 28019142
10. Zheng J, Yu Z, Ma L. **Association between blood glucose and functional outcome in intracerebral hemorrhage: a systematic review and meta‐analysis**. *World Neurosurg* (2018) **114** e756-e765. PMID: 29555604
11. Dungan K, Braithwaite S, Preiser J. **Stress hyperglycaemia**. *Lancet (London, England)* (2009) **373** 1798-1807. PMID: 19465235
12. Moghissi E, Korytkowski M, DiNardo M. **American Association of Clinical Endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control**. *Endocr Pract* (2009) **15** 353-369. PMID: 19454396
13. Chu H, Huang C, Tang Y, Dong Q, Guo Q. **The stress hyperglycemia ratio predicts early hematoma expansion and poor outcomes in patients with spontaneous intracerebral hemorrhage**. *Ther Adv Neurol Disord* (2022) **15** 17562864211070681. PMID: 35082921
14. Nouh C, Ray B, Xu C. **Quantitative analysis of stress‐induced hyperglycemia and intracranial blood volumes for predicting mortality after intracerebral hemorrhage**. *Transl Stroke Res* (2022) **13** 595-603. PMID: 35040036
15. Stead L, Jain A, Bellolio M. **Emergency Department hyperglycemia as a predictor of early mortality and worse functional outcome after intracerebral hemorrhage**. *Neurocrit Care* (2010) **13** 67-74. PMID: 20390379
16. Yoon J, Kim D, Sohn M. **Effect of stress hyperglycemia and intensive rehabilitation therapy in non‐diabetic hemorrhagic stroke: Korean Stroke Cohort for Functioning and Rehabilitation**. *Eur J Neurol* (2016) **23** 1658-1665. PMID: 27444813
17. Ma X, Pan J, Bao Y. **Combined assessment of glycated albumin and fasting plasma glucose improves the detection of diabetes in Chinese subjects**. *Clin Exp Pharmacol Physiol* (2010) **37** 974-979. PMID: 20557319
18. Stöllberger C, Exner I, Finsterer J, Slany J, Steger C. **Stroke in diabetic and non‐diabetic patients: course and prognostic value of admission serum glucose**. *Ann Med* (2005) **37** 357-364. PMID: 16179271
19. Sattar R, Kausar A, Siddiq M. **Advances in thermoplastic polyurethane composites reinforced with carbon nanotubes and carbon nanofibers: a review**. *J Plastic Film Sheeting* (2014) **31** 186-224
20. **Classification and diagnosis of diabetes**. *Diabetes Care* (2020) **43** S14-S31. PMID: 31862745
21. Teasdale G, Jennett B. **Assessment of coma and impaired consciousness. A practical scale**. *Lancet (London, England)* (1974) **2** 81-84. PMID: 4136544
22. Kothari R, Brott T, Broderick J. **The ABCs of measuring intracerebral hemorrhage volumes**. *Stroke* (1996) **27** 1304-1305. PMID: 8711791
23. van Swieten J, Koudstaal P, Visser M, Schouten H, van Gijn J. **Interobserver agreement for the assessment of handicap in stroke patients**. *Stroke* (1988) **19** 604-607. PMID: 3363593
24. Garner J, Jarvis W, Emori T, Horan T, Hughes J. **CDC definitions for nosocomial infections, 1988**. *Am J Infect Control* (1988) **16** 128-140. PMID: 2841893
25. Smith C, Kishore A, Vail A. **Diagnosis of stroke‐associated pneumonia: recommendations from the pneumonia in stroke consensus group**. *Stroke* (2015) **46** 2335-2340. PMID: 26111886
26. Dowlatshahi D, Demchuk A, Flaherty M, Ali M, Lyden P, Smith E. **Defining hematoma expansion in intracerebral hemorrhage: relationship with patient outcomes**. *Neurology* (2011) **76** 1238-1244. PMID: 21346218
27. Yogendrakumar V, Ramsay T, Fergusson D. **New and expanding ventricular hemorrhage predicts poor outcome in acute intracerebral hemorrhage**. *Neurology* (2019) **93** e879-e888. PMID: 31371565
28. Hofmeijer J, Kappelle L, Algra A, Amelink G, van Gijn J, van der Worp H. **Surgical decompression for space‐occupying cerebral infarction (the Hemicraniectomy After Middle Cerebral Artery infarction with Life‐threatening Edema Trial [HAMLET]): a multicentre, open, randomised trial**. *The Lancet Neurology* (2009) **8** 326-333. PMID: 19269254
29. Zou G. **A modified poisson regression approach to prospective studies with binary data**. *Am J Epidemiol* (2004) **159** 702-706. PMID: 15033648
30. Chiu C, Chen T, Chin L. **Investigation of the effect of hyperglycemia on intracerebral hemorrhage by proteomic approaches**. *Proteomics* (2012) **12** 113-123. PMID: 22065606
31. Song E, Chu K, Jeong S. **Hyperglycemia exacerbates brain edema and perihematomal cell death after intracerebral hemorrhage**. *Stroke* (2003) **34** 2215-2220. PMID: 12907821
32. Liu R, Wang J, Qiu X, Wu J. **Acute hyperglycemia together with hematoma of high‐glucose blood exacerbates neurological injury in a rat model of intracerebral hemorrhage**. *Neurosci Bull* (2014) **30** 90-98. PMID: 23884876
33. Won S, Tang X, Suh S, Yenari M, Swanson R. **Hyperglycemia promotes tissue plasminogen activator‐induced hemorrhage by Increasing superoxide production**. *Ann Neurol* (2011) **70** 583-590. PMID: 22002675
34. Liu J, Gao B, Clermont A. **Hyperglycemia‐induced cerebral hematoma expansion is mediated by plasma kallikrein**. *Nat Med* (2011) **17** 206-210. PMID: 21258336
35. Kimura K, Iguchi Y, Inoue T. **Hyperglycemia independently increases the risk of early death in acute spontaneous intracerebral hemorrhage**. *J Neurol Sci* (2007) **255** 90-94. PMID: 17350046
36. Kim Y, Han M, Kim C, Kim J, Cheong J, Ryu J. **Increased short‐term mortality in patients with spontaneous intracerebral hemorrhage and its association with admission glucose levels and leukocytosis**. *World Neurosurg* (2017) **98** 503-511. PMID: 27890760
37. Koga M, Yamagami H, Okuda S. **Blood glucose levels during the initial 72 h and 3‐month functional outcomes in acute intracerebral hemorrhage: the SAMURAI‐ICH study**. *J Neurol Sci* (2015) **350** 75-78. PMID: 25711829
38. Tapia‐Pérez J, Gehring S, Zilke R, Schneider T. **Effect of increased glucose levels on short‐term outcome in hypertensive spontaneous intracerebral hemorrhage**. *Clin Neurol Neurosurg* (2014) **118** 37-43. PMID: 24529227
39. Béjot Y, Aboa‐Eboulé C, Hervieu M. **The deleterious effect of admission hyperglycemia on survival and functional outcome in patients with intracerebral hemorrhage**. *Stroke* (2012) **43** 243-245. PMID: 21940959
40. Appelboom G, Piazza M, Hwang B. **Severity of intraventricular extension correlates with level of admission glucose after intracerebral hemorrhage**. *Stroke* (2011) **42** 1883-1888. PMID: 21636822
41. Lee S, Kim B, Bae H. **Effects of glucose level on early and long‐term mortality after intracerebral haemorrhage: the Acute Brain Bleeding Analysis Study**. *Diabetologia* (2010) **53** 429-434. PMID: 20091021
42. Passero S, Ciacci G, Ulivelli M. **The influence of diabetes and hyperglycemia on clinical course after intracerebral hemorrhage**. *Neurology* (2003) **61** 1351-1356. PMID: 14638954
43. Sun S, Pan Y, Zhao X. **Prognostic value of admission blood glucose in diabetic and non‐diabetic patients with intracerebral hemorrhage**. *Sci Rep* (2016) **6** 32342. PMID: 27562114
44. Arboix A, Massons J, García‐Eroles L, Oliveres M, Targa C. **Diabetes is an independent risk factor for in‐hospital mortality from acute spontaneous intracerebral hemorrhage**. *Diabetes Care* (2000) **23** 1527-1532. PMID: 11023147
45. Liebkind R, Gordin D, Strbian D. **Diabetes and intracerebral hemorrhage: baseline characteristics and mortality**. *Eur J Neurol* (2018) **25** 825-832. PMID: 29443444
46. Li S, Wang Y, Wang W, Zhang Q, Wang A, Zhao X. **Stress hyperglycemia is predictive of clinical outcomes in patients with spontaneous intracerebral hemorrhage**. *BMC Neurol* (2022) **22** 236. PMID: 35761206
47. Wang C, Wang W, Li G. **Prognostic value of glycemic gap in patients with spontaneous intracerebral hemorrhage**. *Eur J Neurol* (2022) **29** 2725-2733. PMID: 35652741
48. Tuttolomondo A, Pinto A, Di Raimondo D, Fernandez P, Licata G. **Stroke patterns, etiology, and prognosis in patients with diabetes mellitus**. *Neurology* (2005) **64** 581
49. Lithner F, Asplund K, Eriksson S, Hägg E, Strand T, Wester P. **Clinical characteristics in diabetic stroke patients**. *Diabete Metab* (1988) **14** 15-19. PMID: 3391327
50. Jørgensen H, Nakayama H, Raaschou H, Olsen T. **Stroke in patients with diabetes. The Copenhagen Stroke Study Stroke**. *Stroke* (1994) **25** 1977-1984. PMID: 8091441
51. Megherbi S, Milan C, Minier D. **Association between diabetes and stroke subtype on survival and functional outcome 3 months after stroke: data from the European BIOMED Stroke Project**. *Stroke* (2003) **34** 688-694. PMID: 12624292
52. Tao J, Hu Z, Lou F. **Higher stress hyperglycemia ratio is associated with a higher risk of stroke‐associated pneumonia**. *Front Nutr* (2022) **9**. PMID: 35273985
53. Esposito K, Nappo F, Marfella R. **Inflammatory cytokine concentrations are acutely increased by hyperglycemia in humans: role of oxidative stress**. *Circulation* (2002) **106** 2067-2072. PMID: 12379575
54. Morohoshi M, Fujisawa K, Uchimura I, Numano F. **Glucose‐dependent interleukin 6 and tumor necrosis factor production by human peripheral blood monocytes in vitro**. *Diabetes* (1996) **45** 954-959. PMID: 8666148
55. Kwan J, Horsfield G, Bryant T. **IL‐6 is a predictive biomarker for stroke associated infection and future mortality in the elderly after an ischemic stroke**. *Exp Gerontol* (2013) **48** 960-965. PMID: 23872300
56. Haeusler K, Schmidt W, Föhring F. **Cellular immunodepression preceding infectious complications after acute ischemic stroke in humans**. *Cerebrovasc Dis (Basel, Switzerland)* (2008) **25** 50-58
57. Prass K, Meisel C, Höflich C. **Stroke‐induced immunodeficiency promotes spontaneous bacterial infections and is mediated by sympathetic activation reversal by poststroke T helper cell type 1‐like immunostimulation**. *J Exp Med* (2003) **198** 725-736. PMID: 12939340
58. Santos Samary C, Pelosi P, Leme Silva P, Rieken Macedo Rocco P. **Immunomodulation after ischemic stroke: potential mechanisms and implications for therapy**. *Crit Care (London, England)* (2016) **20** 391
59. Meisel C, Schwab J, Prass K, Meisel A, Dirnagl U. **Central nervous system injury‐induced immune deficiency syndrome**. *Nat Rev Neurosci* (2005) **6** 775-786. PMID: 16163382
60. Emsley H, Hopkins S. **Acute ischaemic stroke and infection: recent and emerging concepts**. *Lancet Neurol* (2008) **7** 341-353. PMID: 18339349
61. Liu D, Chu S, Chen C, Yang P, Chen N, He X. **Research progress in stroke‐induced immunodepression syndrome (SIDS) and stroke‐associated pneumonia (SAP)**. *Neurochem Int* (2018) **114** 42-54. PMID: 29317279
62. Vanhorebeek I, Van den Berghe G. **Diabetes of injury: novel insights**. *Endocrinol Metab Clin North Am* (2006) **35** 859-872. PMID: 17127151
|
---
title: 'Risk factors based vessel‐specific prediction for stages of coronary artery
disease using Bayesian quantile regression machine learning method: Results from
the PARADIGM registry'
authors:
- Hyung‐Bok Park
- Jina Lee
- Yongtaek Hong
- So Byungchang
- Wonse Kim
- Byoung K. Lee
- Fay Y. Lin
- Martin Hadamitzky
- Yong‐Jin Kim
- Edoardo Conte
- Daniele Andreini
- Gianluca Pontone
- Matthew J. Budoff
- Ilan Gottlieb
- Eun Ju Chun
- Filippo Cademartiri
- Erica Maffei
- Hugo Marques
- Pedro de A. Gonçalves
- Jonathon A. Leipsic
- Sanghoon Shin
- Jung H. Choi
- Renu Virmani
- Habib Samady
- Kavitha Chinnaiyan
- Peter H. Stone
- Daniel S. Berman
- Jagat Narula
- Leslee J. Shaw
- Jeroen J. Bax
- James K. Min
- Woong Kook
- Hyuk‐Jae Chang
journal: Clinical Cardiology
year: 2023
pmcid: PMC10018106
doi: 10.1002/clc.23964
license: CC BY 4.0
---
# Risk factors based vessel‐specific prediction for stages of coronary artery disease using Bayesian quantile regression machine learning method: Results from the PARADIGM registry
## Abstract
### Background and Hypothesis
The recently introduced Bayesian quantile regression (BQR) machine‐learning method enables comprehensive analyzing the relationship among complex clinical variables. We analyzed the relationship between multiple cardiovascular (CV) risk factors and different stages of coronary artery disease (CAD) using the BQR model in a vessel‐specific manner.
### Methods
From the data of 1,463 patients obtained from the PARADIGM (NCT02803411) registry, we analyzed the lumen diameter stenosis (DS) of the three vessels: left anterior descending (LAD), left circumflex (LCx), and right coronary artery (RCA). Two models for predicting DS and DS changes were developed. Baseline CV risk factors, symptoms, and laboratory test results were used as the inputs. The conditional $10\%$, $25\%$, $50\%$, $75\%$, and $90\%$ quantile functions of the maximum DS and DS change of the three vessels were estimated using the BQR model.
### Results
The 90th percentiles of the DS of the three vessels and their maximum DS change were $41\%$–$50\%$ and $5.6\%$–$7.3\%$, respectively. Typical anginal symptoms were associated with the highest quantile ($90\%$) of DS in the LAD; diabetes with higher quantiles ($75\%$ and $90\%$) of DS in the LCx; dyslipidemia with the highest quantile ($90\%$) of DS in the RCA; and shortness of breath showed some association with the LCx and RCA. Interestingly, High‐density lipoprotein cholesterol showed a dynamic association along DS change in the per‐patient analysis.
### Conclusions
This study demonstrates the clinical utility of the BQR model for evaluating the comprehensive relationship between risk factors and baseline‐grade CAD and its progression.
## INTRODUCTION
Cardiovascular disease (CVD) is the primary cause of morbidity and mortality worldwide, with a global burden of 17 million deaths annually. 1 Among them, coronary artery disease (CAD) accounts for over $50\%$ of the total deaths and this number continues to increase. 2 Various physiological and behavioral cardiovascular (CV) risk factors have been found to be associated with the development of CAD. 3, 4, 5 Different symptoms can present themselves according to lesion severity or location and their interrelationships. 5 Almost $60\%$ of patients with stable chest pain exhibit non‐obstructive stenotic CAD with much less typical angina symptoms than obstructive CAD. 6, 7 In addition, various CV risk factors are associated with symptom presentation. 8, 9 *Coronary atherosclerosis* is a chronic and progressive process; thus, detecting subclinical atherosclerosis and intervening in its early phase has significant importance for clinical outcomes. 5, 9 Therefore, comprehensive studies are needed from the early to severe stages of CAD for optimized treatments. However, to date, most previous research has focused on obstructive CAD prediction via standard regression model analysis, overlooking the importance of the early stage of CAD as most deep and shallow machine learning models investigate only the average relationship between clinical outcome and risk factors. In contrast, the Bayesian quantile regression (BQR) model, a recently introduced machine learning method, is useful for analyzing the comprehensive association between clinical variables with various stages of CAD because BQR model yields multiple quantile regression curves. 10, 11, 12, 13 Particularly useful for revealing hidden independent dynamic associations of target clinical variables according to quantile stages of endpoint in a complex database such as clinical data; thus, it can be applied to specific patients for tailored therapy.
Therefore, we aimed to apply the BQR model to the association analysis between graded subclinical and clinical coronary atherosclerosis and CV risk factors to evaluate vessel‐specific dynamic interrelationships.
## Study design and population
We analyzed the data from Progression of AtheRosclerotic PlAque DetermIned by Computed TomoGraphic Angiography IMaging (PARADIGM, NCT02803411), a prospective, international, and multicenter observational registry designed to track coronary atherosclerosis in serially acquired coronary computed tomography angiography (CCTA). 14 Between 2003 and 2015, 2252 consecutive patients with suspected or known CAD who underwent serial CCTA at an interscan interval of ≥2 years were enrolled. The Institutional Review Boards of all participating hospitals approved this study protocol, which was conducted according to the Declaration of Helsinki revised in 2013. The need for informed consent was waived by the Severance Hospital Institutional Review Board because the study used anonymized data (approval number 2020‐3481‐001). After the exclusion of patients with non‐interpretable scans at baseline or follow‐up CCTA ($$n = 492$$), documented CAD before baseline CCTA ($$n = 227$$), and incomplete clinical information such as CV risk factors, symptom variables, and laboratory results at baseline or follow‐up CCTA ($$n = 70$$), 1463 patients who underwent per‐segment‐based quantitative CCTA plaque analysis including lumen diameter stenosis (DS) were included in this study.
## Data extraction and analysis
The baseline clinical characteristics and laboratory data were used as clinical variables, and the per‐segment‐based quantitative CCTA findings were used for a set of outcomes. We performed a vessel‐wise analysis with these data at all outcome‐level settings using the Bayesian truncated quantile regression model. For the vessel‐wise analysis, all 18 coronary segments were classified into the following three vessel groups: left anterior descending (LAD), left circumflex (LCx), and right coronary artery (RCA). The largest quantitative DS measurement in each vessel (LAD, LCx, or RCA) was regarded as the representative value for each vessel, and the largest DS among the vessels was regarded as the representative value for each patient. Most often, the LAD was included ($$n = 1264$$) followed by the RCA ($$n = 864$$) and the LCx ($$n = 718$$).
Figure S1 shows the histograms of DS values for the three vessels (LAD and LCx, and RCA) and each patient; the shapes of the histograms show that the data generating the distributions were not normally distributed and were truncated. Figure S2 shows the histograms of DS changes (defined as post‐DS minus pre‐DS divided by CCTA intervals) for the three vessels (the LAD, LCx, and RCA) and each patient.
We tested the following two models: the DS model (Model 1) and DS change model (Model 2). Multiple CV risk factors including the symptom variables were used to predict quantile DS values for the three vessels and each patient in Model 1 and also used to predict quantile DS changes in Model 2.
## Quantile regression modeling
The quantile regression model for DS prediction (Model 1) was defined as follows: [1] DSof(LAD,LCx,RCA,andper‐patient)=α+∑$i = 17$βi⋅Baselinesi+∑$i = 14$γi⋅SymptomTypesi+∑$i = 13$δi⋅LabExamsi+ϵθ, where Baselinesi were baseline CV risk factors including age, sex, body mass index (BMI), smoking, diabetes, hypertension, and dyslipidemia; Symptom Typesi were categorical risk factors denoting the types of patients’ symptoms comprised “typical angina, atypical angina, Noncardiac pain, and others” with “asymptotic” as the reference category; Lab Examsi were continuous variables from laboratory examinations including high‐density lipoprotein cholesterol (HDL‐C), low‐density lipoprotein cholesterol (LDL‐C), and triglycerides (TG); ϵθ was the error term with its θth quantile equal to zero (in our study, θ were $10\%$, $25\%$, $50\%$, $75\%$, and $90\%$).
Model 2 used the changes in DS values as the outcome variable, and the quantile regression model was specified as follows: [2] DSchangeof(LAD,LCx,RCA,andper‐patient)=α+∑$i = 17$βi⋅Baselinesi+∑$i = 14$γi⋅SymptomTypesi+∑$i = 13$δi⋅LabExams.i+ϵθ.
## Statistical analysis
All statistical analyses were performed using R software with package “ctqr” (version 4.1.0, R Foundation for Statistical Computing). 15 Continuous variables were presented as means and standard deviations. Categorical variables were presented as frequencies and percentages. Prediction performance was evaluated using the area under the curve (AUC) values of the receiver operating characteristic curves.
## Study population and AUC values for overall and the three major vessels
The baseline characteristics of the study population are presented in Table 1. The mean patient age was 62 years; $35.2\%$ were women, $59.4\%$ had hypertension, $46.3\%$ had dyslipidemia, and $24.1\%$ had diabetes mellitus. Most patients had atypical angina ($62.2\%$), and typical anginal symptoms were observed in only $6.5\%$ of the patients. AUC estimates for predicting obstructive stenosis (DS ≥ $50\%$) using a logistic regression model with risk factors are presented in Supplementary Figure 3. The AUC values were 0.67, 0.65, 0.78, and 0.73 for per‐patient, LAD, LCx, and RCA, respectively.
**Table 1**
| Unnamed: 0 | Patients (n = 1463) |
| --- | --- |
| Age, years | 61.8 ± 9.1 |
| Male | 1095 (64.8) |
| Body mass index, kg/m2 | 25.6 ± 3.4 |
| Current smoker | 320 (19.2) |
| Diabetes mellitus | 404 (24.1) |
| Hypertension | 993 (59.4) |
| Dyslipidemia | 772 (46.3) |
| Laboratory data | Laboratory data |
| HDL cholesterol, mg/dL | 49.6 ± 13.5 |
| LDL cholesterol, mg/dL | 112.5 ± 35.4 |
| Triglycerides | 145.5 ± 86.6 |
| Symptoms | Symptoms |
| Asymptomatic | 370 (22.2) |
| Typical angina | 109 (6.5) |
| Atypical angina | 1038 (62.2) |
| Noncardiac pain | 133 (8.0) |
| Others | 139 (8.3) |
## Intervessel correlation coefficients between stenosis measures
Table S1 shows the intervessel correlation coefficient estimates of the stenosis measures, revealing that the DSs of the three vessels were weakly correlated (<0.3). The low DS correlations between the vessels suggest the necessity of a per‐vessel analysis of DS for a more precise CAD diagnosis.
## BQR analysis for DS and DS change according to CV risk factors
The quantile estimates of $10\%$, $25\%$, $50\%$, $75\%$, and $90\%$ for the three vessels and their per‐patient values of the DS and DS changes are shown in Table S2. The mean measurements of the 90th percentiles were $41\%$–$50\%$ and $5.6\%$–$7.3\%$ in DS and DS change, respectively. Figures 1, 2, 3, 4 show the error bar charts of the coefficient estimates with $95\%$ confidence intervals for the selected risk factors for which at least one estimate was statistically significant among the five quantiles ($10\%$, $25\%$, $50\%$, $75\%$, and $90\%$), respectively for regression Models 1 and 2. The y‐axes were log‐scaled for clear visibility of the error bar charts.
In the per‐vessel analysis of DS, the typical anginal symptom was associated with the highest quantile ($90\%$) of DS in the LAD; diabetes was associated with higher quantiles ($75\%$ and $90\%$) of DS in the LCx; dyslipidemia was associated with the highest quantile ($90\%$) of DS in the RCA, whereas other symptoms showed some association with the LCx and RCA (Figure 1). Overall, the per‐patient analysis of DS, age, and hypertension was positively associated with all DS quantiles; in contrast, HDL‐C was negatively associated with most DS quantiles (Figure 2).
**Figure 1:** *Bayesian quantile regression analysis for DS in the three vessels (LAD, LCx, and RCA). Error bar charts of the coefficient estimates with 95% confidence intervals for the selected risk factors in the three vessels are presented. Risk factors with at least one statistically significant estimate among the five quantiles (10%, 25%, 50%, 75%, and 90%) of DS were chosen using the Bayesian quantile regression model (Model 1). The y‐axis of the error bar charts is log‐scaled. DS, diameter stenosis; HDL, high‐density lipoprotein cholesterol; LAD, left anterior descending coronary artery; LCx, left circumflex coronary artery; RCA, right coronary artery.* **Figure 2:** *Bayesian quantile regression analysis for DS in per‐patient. The error bar chart of the coefficient estimates with 95% confidence intervals for the selected risk factors per‐patient is presented. Risk factors with at least one statistically significant estimate among the five quantiles (10%, 25%, 50%, 75%, and 90%) of DS were chosen using the Bayesian quantile regression model (Model 1). The y‐axis of the error bar chart is log‐scaled. DS, diameter stenosis; HDL, high‐density lipoprotein cholesterol.*
In the per‐vessel analysis of DS change, HDL‐C showed a clear and dynamic relationship, a positive association with a low level of DS change and a negative association with a high level of DS change in the LAD and RCA; hypertension also showed a dynamic relationship with DS change in the LCx and DS change severity (Figure 3). In the overall per‐patient analysis of DS change, age, smoking, and hypertension showed a tendency to increase DS change, although no consistent associations were observed. However, unlike LDL‐C, which showed no significant association with DS change, HDL‐C showed a dynamic association with DS change which changed from positive to negative with DS severity (Figure 4).
**Figure 3:** *Bayesian quantile regression analysis for DS change in the three vessels (LAD, LCx, and RCA). Error bar charts of the coefficient estimates with 95% confidence intervals for the selected risk factors in the three vessels are presented. Risk factors with at least one statistically significant estimate among the five quantiles (10%, 25%, 50%, 75%, and 90%) of DS change were chosen using the Bayesian quantile regression model (Model 2). The y‐axis of the error bar charts is log‐scaled. BMI, body mass index; DS, diameter stenosis; HDL, high‐density lipoprotein cholesterol; LAD, left anterior descending coronary artery; LCx, left circumflex coronary artery; LDL‐C, low‐density lipoprotein cholesterol; RCA, right coronary artery.* **Figure 4:** *Bayesian quantile regression analysis for DS change in per‐patient. The error bar chart of the coefficient estimates with 95% confidence intervals for the selected risk factors per‐patient is presented. Risk factors with at least one statistically significant estimate among the five quantiles (10%, 25%, 50%, 75%, and 90%) of DS change were chosen using the Bayesian quantile regression model (Model 2). The y‐axis of the error bar chart is log‐scaled. DS, diameter stenosis; HDL, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol.*
## DISCUSSION
In the present study, we demonstrated the clinical utility of the Bayesian truncated quantile regression machine learning method to evaluate the comprehensive relationship between CV risk factors and baseline‐graded subclinical to clinical coronary artery stenosis and its progression.
First, while HDL‐C showed a consistent negative association with most DS levels, interestingly, the dynamic relationship was revealed for DS change severity, from positive relation to low‐level DS change and a negative association with high‐level DS change in our data set. The empirical results suggest that high HDL‐C has a preventive effect on CAD progression only for patients at a rapidly deteriorating stage. Hypertension is another CV risk factor exhibiting dynamic relation to DS change, from positive to negative, along with DS change severity. Typical angina symptoms were only associated with a high quantile of stenosis in the LAD and not in the LCx or RCA. Likewise, diabetes was strongly associated with LCx, and dyslipidemia was associated with RCA. Shortness of breath showed some relationship with a certain degree of stenosis in the LCx and RCA.
The empirical results from the BQR model provide clinical evidence supporting the implicit relationships among the risk factors. It has been known by clinical experience that LAD lesions are associated with typical anginal symptoms owing to their considerable accountability in the entire coronary perfusion 16, 17; similarly, it is known by experience that LCx or RCA lesions are more likely to be associated with vague symptoms than LAD lesions. 16, 18 However, to date, no scientific evidence has been provided.
In addition, HDL‐C showed a dynamic interrelationship with graded coronary stenosis and stenosis progression, which was the most distinctive utility of the BQR model that could not be achieved in any other standard regression models. Our empirical results from the BQR analysis might provide valuable clinical clues for enabling targeted management of CAD patients, especially since low HDL‐C levels could be an aggravating factor for rapid CAD progression.
Since Koenker and Bassett first introduced quantile regression models, they have been used in various research areas, such as investment, economics, and engineering, due to their multiple advantages over standard regression analysis. 19 Quantile regression has recently been regarded as an efficient analysis tool for income and wage studies in labor economics. The Bayesian Tobit quantile regression, an advanced version of the plain quantile regression model, has been utilized to estimate outage costs in the engineering field. 10, 11, 12 Although Wehby et al. 20 first introduced the utility of the BQR model in the medical field by presenting the different risk factors for low and high birth weight, it is not widely adopted probably because its interpretation seems somewhat unintuitive since the concept of quantile is less familiar than means. 21 However, with the increased interest in machine learning methods in medical research, quantile regression has recently attracted attention as a valuable data analysis tool in the medical research area. 13 Kuhudzai et al. 22 is the first study which indicated the impact of blood pressure risk factors in South Africa using BQR model. The study showed that the BQR model performs more accurate modeling for the hypertension estimate than classical approaches.
Although clinical models for estimating the pretest probability of CAD based on age, sex, and symptom typicality in patients with stable angina have been developed, 23, 24 recent studies raised the overestimation issue of these models, potentially due to the exclusion of other important CV risk factors such as diabetes, dyslipidemia, hypertension, smoking, and obesity. 25, 26 Novel imaging markers, including calcium score and multiple risk factor assessment using the machine learning method, have been evaluated to overcome this issue. 26, 27 However, most studies have shown modest performance for predicting obstructive CAD and are limited to a single outcome variable of $50\%$ DS. 25, 26, 27 To the best of our knowledge, the present study is the first to apply BQR analysis to the prediction of CAD and especially for CAD progression, by exploring the comprehensive association between CV risk factors and various stages of CAD. This pilot study can provide a framework for the cost‐efficient utilization of previously overlooked clinical information, thereby facilitating the development of a more accurate CAD pretest probability model. Furthermore, applying BQR analysis to complex clinical data will provide a hidden pattern of certain clinical risk factors for dynamically impacting certain targeted populations with specific stages of the disease and thus will be utilized in personalized therapy.
Recent studies have shown the possibility of deep learning‐based novel methods for detecting CAD in its early stage utilizing a conventional twelve‐lead electrocardiogram (ECG). 28 and the feasibility of convolutional neural networks for the prediction of calcium scores from traditional chest X‐ray radiography (CXR). 29 These innovative machine learning methods and their potential combined models could turn common clinical information from ECG and CXR into vital information thereby reducing unnecessary downstream tests.
This study has several limitations. First, we only included 1463 patients with complete clinical information; most had LAD lesions and the LCx and RCA lesions were only on 465 and 340 vessels, respectively. Thus, there were insufficient data for the evaluation of the LCx or RCA. Second, although we included major CV risk factors for CAD, further specified and various CV risk factors should be included to enhance the performance of this model. Lastly, this study could not present an elaborate CAD prediction model. To develop an advanced CAD prediction analysis, balanced vessel numbers and complete clinical data are needed.
In conclusion, we introduced the BQR machine learning method in the CV field to evaluate the complex interrelationship between CV risk factors and the different stages of CAD and its progression. Using this innovative method, we comprehensively determined the dominant association of each coronary vessel with symptoms or CV risk factors, which is clinically useful.
## CONFLICT OF INTEREST
Dr Chang receives funding from by the Korea Medical Device Development Fund grant funded by the Korea government (the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety) (Project Number: 1711139017); Dr Min receives funding from the National Institutes of Health (Grant Nos. R01 HL111141, R01 HL115150, R01 118019, and U01 HL 105907), the Qatar National Priorities Research Program (Grant No. 09‐370‐3‐089), and GE Healthcare. Dr Min served as a consultant to HeartFlow, serves on the scientific advisory board of Arineta, and has an equity interest in MDDX. Dr Bax receives unrestricted research grants from Biotronik, Medtronic, Boston Scientific, and Edwards Lifesciences. Dr Chun receives funding from National Research Foundation (NRF) grant funded by the Korea government (Ministry of Education, Science and Technology; NRF‐2015R1D1A1A01059717). Dr *Leipsic is* a consultant and holds stock options in HeartFlow and Circle CVI. He receives modest speaking fees from Philips and GE Healthcare. Dr Budoff receives grant support from the National Institutes of Health and GE Healthcare. Dr *Marques is* a Consultant and holds stock options for Cleerly Inc. Dr *Samady is* a cofounder and equity holder of Covanos, a consultant for Philips and Valo, and receives grant support from Phillips and St Jude Abbott/Medtronic. Dr *Andreini is* on the Speakers Bureau for GE Healthcare and receives grant support from GE Healthcare and Bracco. Dr Pontone receives institutional research grants from GE Healthcare, HeartFlow, Medtronic, Bracco, and Bayer. Dr Berman receives software royalties from Cedars‐Sinai. Dr Virmani has received institutional research support from 480 Biomedical, Abbott Vascular, Arterial Remodeling Technologies, BioSensors International, Biotronik, Boston Scientific, Celonova, Claret Medical, Cook Medical, Cordis, Edwards Lifesciences, Medtronic, MicroVention, OrbusNeich, ReCord, SINO Medical Technology, Spectranetics, Surmodics, Terumo Corporation, W.L. Gore and Xeltis. Dr Virmani also receives honoraria from 480 Biomedical, Abbott Vascular, Boston Scientific, Cook Medical, Lutonix, Medtronic, Terumo Corporation, and W.L. Gore, and is a consultant for 480 Biomedical, Abbott Vascular, Medtronic, and W.L. Gore. Dr *Min is* an employee and holds equity interest in Cleerly, Inc. He is also on the Medical Advisory Board at Arineta. The other authors report no conflicts.
## DATA AVAILABILITY STATEMENT
Due to privacy and ethical concerns, neither the data nor the source of the data can be made available.
## References
1. Roth GA, Abate D, Abate KH. **Global, regional, and national age‐sex‐specific mortality for 282 causes of death in 195 countries and territories, 1980–2017: a systematic analysis for the global burden of disease study 2017**. *The Lancet* (2018) **392** 1736-1788
2. Writing Group M, Lloyd‐Jones D, Adams RJ. **Heart disease and stroke statistics—2010 update: a report from the American Heart Association**. *Circulation* (2010) **121** 46
3. Khot UN, Khot MB, Bajzer CT. **Prevalence of conventional risk factors in patients with coronary heart disease**. *JAMA* (2003) **290** 898-904. PMID: 12928466
4. Greenland P, Knoll MD, Stamler J. **Major risk factors as antecedents of fatal and nonfatal coronary heart disease events**. *JAMA* (2003) **290** 891-897. PMID: 12928465
5. Shao C, Wang J, Tian J, Tang Y‐d. **Coronary artery disease: from mechanism to clinical practice**. *Coronary Artery Dis: Ther Drug Discov* (2020) 1-36
6. Patel MR, Peterson ED, Dai D. **Low diagnostic yield of elective coronary angiography**. *N Engl J Med* (2010) **362** 886-895. PMID: 20220183
7. Patel MR, Dai D, Hernandez AF. **Prevalence and predictors of nonobstructive coronary artery disease identified with coronary angiography in contemporary clinical practice**. *Am Heart J* (2014) **167** 846-852.e2. PMID: 24890534
8. Milner KA, Funk M, Richards S, Wilmes RM, Vaccarino V, Krumholz HM. **Gender differences in symptom presentation associated with coronary heart disease**. *Am J Cardiol* (1999) **84** 396-399. PMID: 10468075
9. Wang X, Yu D, Wang J, Huang J, Li W. **Analysis of coronary artery lesion degree and related risk factors in patients with coronary heart disease based on computer‐aided diagnosis of coronary angiography**. *Comput Math Methods Med* (2021) **2021** 1-10
10. Kim MS, Lee BS, Lee HS, Lee SH, Lee J, Kim W. **Robust estimation of outage costs in South Korea using a machine learning technique: Bayesian Tobit quantile regression**. *Appl Energy* (2020) **278**
11. Yu K, Lu Z, Stander J. **Quantile regression: applications and current research areas**. *J R Stat Soc: Ser D* (2003) **52** 331-350
12. Buchinsky M. **Quantile regression, Box‐Cox transformation model, and the US wage structure, 1963–1987**. *J Econom* (1995) **65** 109-154
13. Ton J, Cleophas AHZ. *Quantile Regression in Clinical Research* (2022)
14. Lee S‐E, Chang H‐J, Rizvi A. **Rationale and design of the progression of AtheRosclerotic PlAque DetermIned by Computed TomoGraphic Angiography IMaging (PARADIGM) registry: a comprehensive exploration of plaque progression and its impact on clinical outcomes from a multicenter serial coronary computed tomographic angiography study**. *Am Heart J* (2016) **182** 72-79. PMID: 27914502
15. Frumento P. **ctqr: Censored and Truncated Quantile Regression**. (2016)
16. Reeves TJ, Oberman A, Jones WB, Sheffield LT. **Natural history of angina pectoris**. *Am J Cardiol* (1974) **33** 423-430. PMID: 4273043
17. Kumpuris AG, Quinones MA, Kanon D, Miller RR. **Isolated stenosis of left anterior descending or right coronary artery: relation between site of stenosis and ventricular dysfunction and therapeutic implications**. *Am J Cardiol* (1980) **46** 13-20. PMID: 6966888
18. Lim HF, Dreifus LS, Kasparian H, Najmi M, Balis G. **Chest pain, coronary artery disease and coronary cine‐arteriography**. *Chest* (1970) **57** 41-46. PMID: 5410429
19. Koenker R, Bassett G. **Regression quantiles**. *Econometrica* (1978) **46** 33-50
20. Wehby GL, Murray JC, Castilla EE, Lopez‐Camelo JS, Ohsfeldt RL. **Prenatal care effectiveness and utilization in Brazil**. *Health Policy Plan* (2009) **24** 175-188. PMID: 19282483
21. Beyerlein A. **Quantile regression—opportunities and challenges from a user's perspective**. *Am J Epidemiol* (2014) **180** 330-331. PMID: 24989240
22. Kuhudzai AG, Van Hal G, Van Dongen S, Hoque M. **Modelling of South African hypertension: comparative analysis of the classical and Bayesian quantile regression approaches**. *Inquiry* (2022) **59**. PMID: 35373630
23. Diamond GA, Forrester JS. **Analysis of probability as an aid in the clinical diagnosis of coronary‐artery disease**. *N Engl J Med* (1979) **300** 1350-1358. PMID: 440357
24. Genders TSS, Steyerberg EW, Alkadhi H. **A clinical prediction rule for the diagnosis of coronary artery disease: validation, updating, and extension**. *Eur Heart J* (2011) **32** 1316-1330. PMID: 21367834
25. Rovai D, Neglia D, Lorenzoni V, Caselli C, Knuuti J, Underwood SR. **Limitations of chest pain categorization models to predict coronary artery disease**. *Am J Cardiol* (2015) **116** 504-507. PMID: 26081064
26. Genders T, Coles A, Hoffmann U. **The external validity of prediction models for the diagnosis of obstructive coronary artery disease in patients with stable chest pain: insights from the PROMISE trial**. *JACC. Cardiovasc Imaging* (2018) **11** 437-446. PMID: 28624401
27. Reeh J, Therming CB, Heitmann M. **Prediction of obstructive coronary artery disease and prognosis in patients with suspected stable angina**. *Eur Heart J* (2019) **40** 1426-1435. PMID: 30561616
28. Liu X, Wang H, Li Z, Qin L. **Deep learning in ECG diagnosis: a review**. *Knowledge‐Based Systems* (2021) **227**
29. Kamel PI, Yi PH, Sair HI, Lin CT. **Prediction of coronary artery calcium and cardiovascular risk on chest radiographs using deep learning**. *Radiol. Cardiothorac Imaging* (2021) **3**
|
---
title: Self‐reported sleep pattern and recurrence of atrial fibrillation after catheter
ablation
authors:
- Jiehui Cang
- Naiyang Shi
- Didi Zhu
- Yaowu Liu
- Qianxing Zhou
- Long Chen
journal: Clinical Cardiology
year: 2023
pmcid: PMC10018108
doi: 10.1002/clc.23975
license: CC BY 4.0
---
# Self‐reported sleep pattern and recurrence of atrial fibrillation after catheter ablation
## Abstract
### Background
Increasing evidence has shown the relationship between sleep and the recurrence of atrial fibrillation (AF). However, the association of different sleep patterns with AF recurrence after catheter ablation was rarely studied. We aimed to assess the role of different sleep behaviors in the risk of AF recurrence after catheter ablation.
### Methods and Results
A total of 416 consecutive participants from Zhongda hospital of Southeast University were finally analyzed. Sleep patterns were defined by chronotype, sleep duration, insomnia, snoring, and daytime sleepiness. A total of 208 patients ($50.0\%$) had a healthy sleep pattern within a mean follow‐up of 32.42 ± 18.18 months. The observed number of patients with AF recurrence was 10 ($50.0\%$), 80 ($42.6\%$), and 40 ($19.2\%$) in unhealthy, intermediate and healthy sleep groups, respectively ($p \leq .01$). After adjusting covariates, unhealthy sleep pattern was significantly associated with AF recurrence [hazard ratio = 3.47 ($95\%$ confidence interval CI: 1.726–6.979, $p \leq .001$)]. Sleep disorders such as inadequate sleep time (time <7 h or >8 h), insomnia and excessive sleepiness during daytime were associated with a higher risk of recurrence. Otherwise, improvement in sleep seemed to be associated with decreased risk of AF recurrence.
### Conclusion
This retrospective study indicates that adherence to a healthy sleep pattern is associated with a lower risk of AF recurrence. Also, improved sleep before ablation is associated with a lower risk of AF recurrence.
## INTRODUCTION
Atrial fibrillation (AF) is featured by absolutely irregular heart rhythm and can lead to a significantly increased risk of stroke and heart failure, which is associated with substantial morbidity, mortality, and economic cost. 1, 2, 3 Catheter ablation (CA) can effectively achieve rhythm control by ablating triggers and modifying atrial substrates with different forms of energy. 4 However, a significant decline in freedom from AF remains to be a challenge. A meta‐analysis showed that the 62‐month success rate of a single CA procedure was only $59\%$. 5 Previously identified lifestyle factors such as alcohol intake, 6 smoking, 7 and obesity 8 are associated with AF recurrence. Though sleep instability may be associated with the risk of recurrent AF, the specific association between sleep and the risk of AF recurrence is not unclear. 9 Emerging evidence has associated several sleep behaviors, such as excessive daytime sleepiness, 10 sleep quality, 11 sleep duration, 12 and insomnia with episodes of AF. Sleep disorders can alter the activation of sympathetic nerves and increase inflammation and oxidative stress, which would also increase the risk of AF recurrence. 9 Notably, different sleep behaviors are intrinsically linked, so different trials which focus on the same sleep behavior would show contradictory results. 13 Theoretically, *It is* more reasonable to pool different sleep behaviors together when exploring the influence of sleep on the recurrence of AF.
Li et al. previously put forward a new sleep pattern index score that integrates 5 different sleep behaviors: chronotype, sleep duration, excessive daytime sleepiness, snoring and insomnia. 14 Different from the previous Pittsburgh Sleep Quality Index (PSQI), the sleep pattern score is more succinct and easier for patients' follow‐up. What's more, it contains an evaluation of chronotype and daytime sleepiness, which is not included in PSQI.
Sleep pattern was proved to be significantly associated with episodes of cardiovascular diseases and incident arrhythmias. 14, 15 However, the association between sleep pattern and recurrent AF postprocedure is still unknown. In the current study, we aimed to retrospectively analyze the predictive role of newly developed sleep pattern index score in the recurrence of AF among patients who underwent AF ablation.
## Study population
This is a retrospective and single‐center study. Patients hospitalized in Zhongda Hospital of Southeast University and receiving successful CA (both radiofrequency and cyro‐balloon ablation) were all included in this study. Ablation was deemed successful in the absence of symptomatic or asymptomatic atrial tachyarrhythmias lasting >30 s identified by surface electrocardiogram (ECG) or Holter monitoring after the blanking period (3 months). We excluded patients: (i) New York Heart Association functional class IV; (ii) unstable angina or acute myocardial infarction within 3 months; (iii) chronic obstructive pulmonary disease; (iv) severe chronic renal or hepatic impairment; (v) thyroid dysfunction; (vi) rheumatic heart disease; (vii) noninitial procedure; (vii) self‐reporting obvious fluctuation in sleep quality postprocedure.
After excluding 234 patients, a total of 416 patients were included in this study (Figure S1). The baseline characteristics of included patients were presented in Table 1A. Considering the number of study participants, we finally divided patients into three groups according to their final score index: 0–1 (unhealthy sleep pattern), 2–3 (intermediate sleep pattern), 4–5 (healthy sleep pattern). A total of 208 patients ($50.0\%$) had a healthy sleep pattern. Table 1A shows that patients with a healthier sleep pattern appeared to be more likely to have a lower body mass index (BMI); have a smaller size of the left atrium (LA); be more likely to have lower blood pressure, glucose and left atrial diameter; be less likely to have chronic diseases such as hypertension, diabetes, coronary artery disease and chronic heart failure.
**Table 1A**
| Unnamed: 0 | Healthy sleep score | Healthy sleep score.1 | Healthy sleep score.2 |
| --- | --- | --- | --- |
| | 0–1 | 2–3 | 4–5 |
| Age, year | 63.50 ± 7.99 | 63.65 ± 9.55 | 63.15 ± 9.72 |
| Women (%) | 12 (60.0) | 74 (39.4) | 90 (43.3) |
| BMI, kg/m2 | 25.01 ± 3.47 | 25.31 ± 3.25 | 24.78 ± 2.96 |
| SBP, mm Hg | 130.75 ± 20.01 | 127.93 ± 18.37 | 127.99 ± 16.73 |
| DBP, mm Hg | 81.70 ± 12.14 | 77.82 ± 12.09 | 78.43 ± 12.34 |
| Glucose, mmol/L | 6.61 ± 2.72 | 5.91 ± 1.42 | 5.73 ± 1.28 |
| LA, cm | 4.23 ± 0.59 | 4.19 ± 0.52 | 4.07 ± 0.57 |
| Aspirin (%) | 4 (20.0) | 34 (18.1) | 34 (16.3) |
| β‐blocker (%) | 8 (40.0) | 79 (42.0) | 75 (36.0) |
| ACEI/ARNI (%) | 9 (45.0) | 67 (35.6) | 62 (29.8) |
| Spironolactone (%) | 3 (15.0) | 40 (21.3) | 35 (16.8) |
| Statins (%) | 6 (30.0) | 74 (39.3) | 67 (32.2) |
| Physical acticity, METs (min/week) | 819.60 ± 685.38 | 2088.40 ± 967.37 | 1714.05 ± 1124.67 |
| Current smoking (%) | 4 (20.0) | 43 (22.9) | 39 (18.8) |
| Current alcohol intake (%) | 1 (5.0) | 21 (11.2) | 22 (10.6) |
| Hypertension (%) | 13 (65.0) | 120 (63.8) | 118 (56.7) |
| Type 2 diabetes (%) | 4 (20.0) | 40 (21.3) | 15 (7.2) |
| Coronary artery disease (%) | 7 (35.0) | 53 (28.2) | 49 (23.6) |
| Heart failure (%) | 12 (60.0) | 77 (41.0) | 76 (36.5) |
| Nonpersistent AF (%) | 15 (75.0) | 122 (64.9) | 142 (68.3) |
| Substrate modification(%) | 3 (16.7) | 77 (45.3) | 71 (39.4) |
| Follow‐up (months) | 25.40 ± 16.62 | 28.64 ± 17.85 | 32.42 ± 18.18 |
| LR (%) | 10 (50.0) | 80 (42.6) | 40 (19.2) |
| ER (%) | 4 (20.0) | 26 (13.8) | 17 (8.2) |
| Total | 20 | 188 | 208 |
## Follow‐up and assessment of outcomes
All patients took uninterrupted oral anti‐coagulation and antiarrhythmia drugs for at least 3 months or long‐term with guidance from clinicians. Before the ablation procedure, antiarrhythmic drugs were usually discontinued ≥5 half‐lives before ablation, except for amiodarone. All patients were followed up in outpatient clinics or forms of telephone interviews. Patients who had a complaint about recurrence were asked to provide the first ECG or Holter monitoring which recorded the rhythm of AF, atrial flutter or atrial tachyarrhythmias. And for those who had no complaints about symptoms associated with recurrence, we performed a clinical assessment of recurrence, ECG and Holter monitoring to check AF recurrence. Follow‐up time was defined from the data of the procedure to the date of recurrence. While for those without recurrence, follow‐up time was defined from the date of the procedure to the date of the latest follow‐up.
We also collected information on patients who tried to improve sleep quality with longer‐use of hypnotics. “ Longer‐use” was defined as taking hypnotics at least 5 days a week. Improvement recorded by case history or self‐reported improvement was recorded.
## Assessment of sleep behaviors
Sleep behaviors were collected through a questionnaire. Chronotype was assessed by the question, “Do you consider yourself to be: (i) definitely a ‘morning’ person; (ii) indefinitely a ‘morning’ person.” Sleep duration was reported as the hours of sleep every 24 h (including naps). Insomnia symptoms were obtained by the question, “Do you usually have trouble falling asleep at night or do you wake up in the middle of the night?” with choices provided: (i) Yes; (ii) No. Snoring was asked by the question “Does your partner or a close relative or friend complain about your snoring?” with responses: (i) yes; or (ii) no. Daytime sleepiness was asked by the question, “Do you often doze off or fall asleep during the daytime when you don't mean to? ( eg, when working, reading or driving)” with the choices provided: (i) Yes; (ii) No. 14, 15 The final score for the sleep pattern was calculated by pooling 5 different sleep behaviors. Each sleep factor was coded 1 if meeting the healthy criteria and 0 if not. A higher score indicates a healthier sleep pattern.
## Assessment of other covariates
Demographic and lifestyle behaviors such as age, gender, preexisting conditions, drugs, smoking, alcohol intake, systolic blood pressure, diastolic blood pressure, glucose, and information associated with CA procedures were recorded according to electronic hospitalization systems. During follow‐up in an outpatient clinic, some lifestyle behaviors such as exercise, smoking, and alcohol intake were checked again by asking patients directly. Measurements such as the size of the left atrial, and left ventricular ejection fraction were recorded according to echocardiography results before the CA procedure. Besides, early recurrence was also recorded.
## Statistical analyses
Descriptive statistics were used to summarize patient characteristics. Baseline characteristics of the study participants were summarized across the healthy sleep score as mean ± SD or median (interquartile range) for continuous variables and n (%) for categorical variables, if appropriate. Wilcoxon's test or independent‐sample t‐test was performed to compare continuous variables in different groups, as appropriate. χ 2 test or Fisher's exact test was performed to compare categorical variables. Statistical tests were based on a two‐sided significance level of 0.05.
The analyses of time‐to‐clinical recurrence events were described by Kaplan–Meier curves and comparisons between the groups were performed by log‐rank test. Cox proportional hazard models were used to estimate the hazard ratios (HRs) and $95\%$ Poisson confidences (confidence interval [CIs]) for all initial predictors of the incidence of AF recurrence. Cox regression was imputed for univariable analyzes to assess potential predictors. Variables that were statistically significant in univariable analysis and those which were non‐statistically significant but had a clinical significance (including AF type, age, gender, hypertension, diabetes, coronary artery disease, heart failure, β‐blocker, and LAD) were all included in further multivariable analysis. A “Forward: likelihood” method was applied in multivariable analysis. When we performed a *Cox analysis* of different sleep behaviors, the same method was applied.
The IBM SPSS Statistics 25.0 software and R statistics were used to perform statistical analyses.
## Characteristics of patients with recurrence
Characteristics of patients with AF recurrence are concluded in Table 1B. Among 416 participants, we documented 130 patients ($31.3\%$) with an incidence of AF recurrence. The observed early recurrence of AF was 36 ($26.9\%$) and 12 ($4.2\%$) in patients without and with clinical recurrence, respectively ($p \leq .01$). According to Table 1B, women and patients with higher BMI are more likely to have a recurrence of atrial arrhythmia, though the difference between the two groups is not significant. Patients with recurrence have a lower healthy sleep pattern score of 2.93 ± 1.13, which is lower than that of patients without recurrence (3.67 ± 1.11, $p \leq .01$). As is shown in Table S1, patients with recurrence are more likely to need substrate modification such as ablation of the tricuspid‐valve isthmus, mitral‐valve isthmus and “BOX” ablation; are more likely to convert to sinus rhythm spontaneously during isolation of pulmonary veins; less likely to need electrical cardioversion during the ablation procedure.
**Table 1B**
| Unnamed: 0 | No‐recurrence (286) | Recurrence (130) | p‐value |
| --- | --- | --- | --- |
| Age <65 (%) | 134 (46.9) | 55 (42.3) | p = .388 |
| Women (%) | 130 (45.5) | 46 (35.4) | p = .540 |
| BMI, kg/m2 | 25.08 ± 3.16 | 24.87 ± 3.08 | p = .515 |
| SBP, mm Hg | 128.60 ± 18.40 | 126.98 ± 15.77 | p = .387 |
| DBP, mm Hg | 77.86 ± 12.13 | 79.29 ± 12.55 | p = .271 |
| Glucose, mmol/L | 5.88 ± 1.53 | 5.80 ± 1.24 | p = .577 |
| LA, mm | 4.09 ± 0.55 | 4.23 ± 0.55 | p = .013 |
| Healthy sleep pattern score | 3.67 ± 1.11 | 2.93 ± 1.13 | p < .01 |
| Aspirin (%) | 57 (19.9) | 15 (11.5) | p = .036 |
| β‐blocker (%) | 113 (39.5) | 49 (37.7) | p = .724 |
| ACEI/ARNI (%) | 99 (34.6) | 39 (30.0) | p = .354 |
| Spironolactone (%) | 47 (16.4) | 31 (23.8) | p = .073 |
| Statins (%) | 96 (33.6) | 51 (39.2) | p = .263 |
| Physical acticity, METs (min/week) | 944.86 ± 1467.10 | 1245.85 ± 2512.91 | p = .126 |
| Current smoking (%) | 54 (18.9) | 32 (24.6) | p = .181 |
| Current alcohol intake (%) | 32 (11.2) | 12 (9.2) | p = .547 |
| Hypertension (%) | 175 (61.2) | 76 (58.5) | p = .598 |
| Type 2 diabetes (%) | 41 (14.3) | 18 (13.8) | p = .894 |
| Coronary artery disease (%) | 80 (28.0) | 29 (22.3) | p = .223 |
| Heart failure (%) | 39 (13.6) | 26 (20.0) | p = .098 |
| Nonpersistent AF (%) | 203 (71.0) | 76 (58.5) | p = .012 |
| Substrate modification (%) | 100 (35.0) | 65 (50.0) | p = .04 |
| ER | 12 (4.2) | 36 (26.9) | p < .01 |
## Predictors of AF recurrence
The association between sleep patterns and the risk of AF recurrence is shown in Figure 1. Sleep pattern was significantly associated with the risk of recurrent AF ($p \leq .001$). After being adjusted by different factors, an unhealthy sleep pattern was still significantly associated with AF recurrence [HR = 3.47, $95\%$ CI (1.73–6.98), $p \leq .001$] when compared to a healthy sleep pattern. Intermediate sleep pattern was also observed to be associated with AF recurrence [HR = 2.20, $95\%$ CI (1.14–2.30), $p \leq .001$] (Table 2A).
**Figure 1:** *Cox regression analysis for prediction of AF recurrence. Cox regression analysis comparing the intermediate (red curve), unhealthy sleep pattern (green curve) to healthy sleep pattern (green curve), respectively. AF, atrial fibrillation* TABLE_PLACEHOLDER:Table 2A The relationship between the index score of the healthy sleep pattern and risk of each outcome was generally similar across subgroups by age (<65 or ≥65 years), AF type, gender, BMI categories (without obesity, and obesity), smoking status, drinking status (Figure S2).
## Different sleep behaviors and AF recurrence
Figure 2 showed the results of the Kaplan–Meier estimation of the time to AF recurrence postprocedure in patients with different sleep behaviors. After adjusted by different factors, parameters such as “adequate sleep duration” [HR = 0.53, $95\%$ CI (0.36–0.79)], “No insomnia” [HR = 0.47, $95\%$ CI (0.32–0.68)] and “No excessive daytime sleepiness” [HR = 0.61, $95\%$ CI (0.41–0.89)] were significantly associated with lower risk of AF recurrence, while “Morning chronotype” and “No snoring” was not associated with recurrent AF (Table 2B).
**Figure 2:** *Kaplan–Meier curve comparing the clinical outcomes of different sleep behaviors. Kaplan–Meier estimation of the time to AF recurrence after ablation in patients with different sleep behaviors. (A) Chronotype of sleep; (B) Sleep duration; (C) Insomnia; (D) Excessive sleepiness at daytime; (E) Snoring* TABLE_PLACEHOLDER:Table 2B
## Improved sleep preprocedure and AF recurrence
A total of 72 patients had a history of longer use of hypnotics and a total of 55 ($76.4\%$) successfully improved their sleep quality before the procedure. Demographic characteristics were concluded in Table S2. The average sleep pattern score of the “Effectiveness Group” was significantly higher (3.76 ± 1.16 vs. 2.18 ± 1.13, $p \leq .01$) preablation. A total of 34 ($61.8\%$) patients in the “Effectiveness Group” had a healthy sleep pattern preprocedure, while 16 patients ($94.1\%$) in the “Failure Group” had an intermediate or unhealthy sleep patterns before ablation (Table S3).
Kaplan–Meier curve presented patients in “Effectiveness Group” were less likely to come down with recurrent AF when compared to those with unhealthy sleep pattern (including 6 patients in the “Failure Group”) (log‐rank $$p \leq .008$$) (Figure S3). No significant difference between the “Effectiveness Group” and “Healthy Sleep Pattern Group” was observed.
## DISCUSSION
This retrospective study explored the association between sleep patterns and AF recurrence after CA. The results indicate that: [1] a healthy sleep pattern was associated with fewer episodes of recurrent AF; [2] three sleep behaviors (adequate sleep duration, no insomnia, and no excessive daytime sleepiness), rather than “Morning chronotype” and “No snoring,” were associated with lower risk of recurrence; [3] improved sleep before ablation was associated with lower risk of AF recurrence.
CA is an effective approach to achieving sinus rhythm for AF patients. However, many factors are associated with the failure of ablation. 16 Apart from previously identified parameters associated with fibrosis of LA, some risk factors resulting in the activation of the automatic nervous system draw increasing attention. Sleep disorder is receiving more focus as it is closely related to daily life. However, sleep itself is very complicated and different sleep parameters are intrinsically correlated. Sleep patterns including five different behaviors can be used to evaluate sleep quality easily and quickly. However, the sleep pattern index score is subjective, while the bands in this questionnaire are easier to recall and can lead to less recalling bias when compared to PSQI.
For the first time, our study explored the association of overall sleep patterns evaluated by “Sleep pattern index score” with the risk of AF recurrence after CA. This new metric system is not clinically used but has been validated to be associated with episodes of different cardiovascular diseases and arrhythmias. 14, 15, 17 Due to a limited number of patients, we divided patients into three groups according to their self‐reported score (0–1, 2–3, 4–5) when we explored the influence of sleep pattern on AF recurrence, which is different from previous research. 14 Kim et al. 9 found that improved sleep quality resulting from ablation was associated with a lower risk of recurrent AF in patients with non‐persistent AF, and sleep instability may be a predictor of recurrence. However, this study focused much more on the impact of alternation of sleep quality resulting from ablation, rather than whether sleep quality was improved before the procedure. Our study emphasized the association of sleep pattern pre‐procedure with AF recurrence. Considering the result of the subgroup analysis, we did not perform further analysis according to AF type. To weaken the impact of ablation to sleep patterns, we excluded patients reporting an obvious sleep instability after the procedure, which could decrease the bias in our research to some extent. Also, the exclusion of patients reporting sleep instability after ablation makes the interpretation of data easier based on the assessment of sleep patterns only once before CA.
Of different sleep disorders, obstructive sleep apnea (OSA) featuring snoring and excessive daytime sleepiness has been most studied. A meta‐analysis showed that untreated OSA can increase the risk of AF recurrence after CA. 18 However, one RCT which was performed to explore the influence of treatment of OSA on AF recurrence showed no difference in recurrence in the two groups. 19 The potential reason may be a small sample size ($$n = 25$$). Also, snoring is only a surrogate of OSA and the true impact of snoring on recurrence may not be equal to that of OSA. In our study, we explored the association between the parameter of 5 pooled behaviors and the risk of AF recurrence, and we found snoring and chronotype did not influence recurrence postprocedure. Previous studies showed that snoring was not associated with the incidence of AF, which may support our study results. 14, 20 Our study also explored the association between improvement in sleep preablation and the risk of AF recurrence. Considering that evaluating sleep pattern based on score twice a time could cause more recalling bias, we judged sleep pattern was improved based on recorded case history or self‐reported results. Results showed that patients who improved their sleep pattern successfully pre‐procedure had a lower risk of recurrence when compared to those with unhealthy sleep patterns. We hence set a hypothesis that improved sleep prior can reduce the risk of recurrent AF and sleep pattern score might be a useful tool to evaluate and guide sleep therapy before ablation. One potential mechanism can support this hypothesis: better sleep patterns can reduce the activation of the autonomic nerve, which will influence both trigger of AF and atrial substrate preprocedure. Remarkably, one patient with a healthy sleep pattern had a history of longer using hypnotics. We checked her medical history carefully and found she took hypnotics due to a complaint of short sleep duration (about 6 h/day). However, hypnotics did not improve this symptom well.
Though our study retrospectively explored the association of sleep and recurrence after ablation, we took different sleep behaviors as a whole, which has been proven to be critical for the research associated with sleep influence. 14, 15 Some potential mechanisms could explain the association between sleep patterns and AF recurrence observed in our study. It has been shown that sleep deprivation may disturb the autonomic nervous balance of sympathetic nervous and vagal outflows, which has been associated with induction and sustained arrhythmias. 14, 21 Also, sleep can affect a brand range of metabolic changes such as lipid, glucose levels, blood pressure and oxidative stress, which would also increase recurrence risk. 22, 23, 24 Additionally, various sleep behaviors may affect the development of cardiac arrhythmias via different and complementary pathways, so it is not surprising that their associations with recurrence exhibit an additive fashion when analyzed as a unit in the sleep pattern, as observed in our study.
## LIMITATIONS
This study has several limitations. Since this retrospective study was conducted in a single center, selection bias and recalling bias are inevitable. AF recurrence may be underestimated due to incomplete capture. Also, an assessment of sleep patterns was only conducted once before ablation. We did not dynamically assess sleep patterns postprocedure, though patients reporting sleep instability postprocedure were excluded. For those who have a history of longer‐use of hypnotics, the intervention was monitored in the outpatient clinic, and we did not assess their initial sleep pattern. Second, there are many other more objective ways of monitoring sleep quality. However, we did not use them in this research setting because most patients were followed up in the form of phone interviews. As we mentioned before, though evaluation of sleep patterns is subjective, bands of the questionnaire are easy to be recalled and can cause less recalling bias. Third, the number of participants was limited, which could weaken the power of evidence. Finally, since the effect of hypnotics was evaluated based on medical history documented in anamnesis or self‐reporting results rather than being monitored in the same way, evidence from this study was not powerful enough to prove sleep pattern was a specific causation. However, we set a hypothesis that the possibility that sleep pattern can predict AF recurrence and our findings could be the basis for further research to explore the causality of sleep patterns. A well‐designed prospective research including larger samples is demanded to confirm these findings.
## CONCLUSIONS
This retrospective study indicates that a healthy sleep pattern is associated with a lower risk of AF recurrence. Also, improvement in sleep before ablation is associated with a lower risk of recurrence. Our results support the hypothesis that better sleep patterns could reduce the recurrence of AF, which still requires more well‐designed studies to validate.
## AUTHOR CONTRIBUTIONS
Jiehui Cang and Long Chen: designed the study. Jiehui Cang and Didi Zhu: collected data. Jiehui Cang and Naiyang Shi: performed data analysis. Jiehui Cang: drafted article. Yaowu Liu: revised manuscript. All authors agree the submission of this manuscript.
## CONFLICTS OF INTEREST
The authors declare no conflicts of interest.
## DATA AVAILABILITY STATEMENT
Immediately following publication, the deidentified participant all‐calculated data that support the findings of this study will be shared upon request. In addition, the data can be applicable to any type of analyses, and they will be shared using methods such as Excel or CSV files via Email. Please contact the corresponding author directly to request data sharing.
## References
1. Turakhia MP, Ziegler PD, Schmitt SK. **Atrial fibrillation burden and short‐term risk of stroke: case‐crossover analysis of continuously recorded heart rhythm from cardiac electronic implanted devices**. *Circ Arrhythm Electrophysiol* (2015) **8** 1040-1047. PMID: 26175528
2. Freeman JV, Wang Y, Akar J, Desai N, Krumholz H. **National trends in atrial fibrillation hospitalization, readmission, and mortality for medicare beneficiaries, 1999–2013**. *Circulation* (2017) **135** 1227-1239. PMID: 28148599
3. Chugh SS, Havmoeller R, Narayanan K. **Worldwide epidemiology of atrial fibrillation**. *Circulation* (2014) **129** 837-847. PMID: 24345399
4. Hindricks G, Potpara T, Dagres N. **2020 ESC guidelines for the diagnosis and management of atrial fibrillation developed in collaboration with the European Association for Cardio‐Thoracic Surgery (EACTS): The Task Force for the diagnosis and management of atrial fibrillation of the European Society of Cardiology (ESC) developed with the special contribution of the European Heart Rhythm Association (EHRA) of the ESC**. *Eur Heart J* (2021) **42** 373-498. PMID: 32860505
5. Kis Z, Muka T, Franco OH. **The short and long‐term efficacy of pulmonary vein isolation as a sole treatment strategy for paroxysmal atrial fibrillation: a systematic review and meta‐analysis**. *Curr Cardiol Rev* (2017) **13** 199-208. PMID: 28124593
6. Bellandi F, Simonetti I, Leoncini M. **Long‐term efficacy and safety of propafenone and sotalol for the maintenance of sinus rhythm after conversion of recurrent symptomatic atrial fibrillation**. *Am J Cardiol* (2001) **88** 640-645. PMID: 11564387
7. Benjamin EJ, Al‐Khatib SM, Desvigne‐Nickens P. **Research priorities in the secondary prevention of atrial fibrillation: a national heart, lung, and blood institute virtual workshop report**. *J Am Heart Assoc* (2021) **10**. PMID: 34351783
8. Guglin M, Maradia K, Chen R, Curtis AB. **Relation of obesity to recurrence rate and burden of atrial fibrillation**. *Am J Cardiol* (2011) **107** 579-582. PMID: 21195377
9. Kim W, Na JO, Thomas RJ. **Impact of catheter ablation on sleep quality and relationship between sleep stability and recurrence of paroxysmal atrial fibrillation after successful ablation: 24‐hour holter‐based cardiopulmonary coupling analysis**. *J Am Heart Assoc* (2020) **9**. PMID: 33241769
10. Full KM, Lutsey PL, Norby FL. **Association between excessive daytime sleepiness and measures of supraventricular arrhythmia burden: evidence from the atherosclerosis risk in communities (ARIC) study**. *Sleep and Breathing* (2020) **24** 1223-1227. PMID: 32215831
11. Kwon Y, Gharib SA, Biggs ML. **Association of sleep characteristics with atrial fibrillation: the multi‐ethnic study of atherosclerosis**. *Thorax* (2015) **70** 873-879. PMID: 25986436
12. Zhuo C, Ji F, Lin X. **Depression and recurrence of atrial fibrillation after catheter ablation: a meta‐analysis of cohort studies**. *J Affect Disord* (2020) **271** 27-32. PMID: 32312694
13. Cappuccio FP, Cooper D, D'Elia L, Strazzullo P, Miller MA. **Sleep duration predicts cardiovascular outcomes: a systematic review and meta‐analysis of prospective studies**. *Eur Heart J* (2011) **32** 1484-1492. PMID: 21300732
14. Li X, Zhou T, Ma H. **Healthy sleep patterns and risk of incident arrhythmias**. *JACC* (2021) **78** 1197-1207. PMID: 34531019
15. Fan M, Sun D, Zhou T. **Sleep patterns, genetic susceptibility, and incident cardiovascular disease: a prospective study of 385 292 UK biobank participants**. *Eur Heart J* (2020) **41** 1182-1189. PMID: 31848595
16. Cai L, Yin Y, Ling Z. **Predictors of late recurrence of atrial fibrillation after catheter ablation**. *Int J Cardiol* (2013) **164** 82-87. PMID: 21737164
17. Li J, Yin J, Luo Y. **Association of healthy sleep pattern with the risk of cardiovascular disease and all‐cause mortality among people with diabetes: a prospective cohort study**. *Diabetes Res Clin Pract* (2022) **186**. PMID: 35271877
18. Neilan TG, Farhad H, Dodson JA. **Effect of sleep apnea and continuous positive airway pressure on cardiac structure and recurrence of atrial fibrillation**. *J Am Heart Assoc* (2013) **2**. PMID: 24275628
19. Caples SM, Mansukhani MP, Friedman PA, Somers VK. **The impact of continuous positive airway pressure treatment on the recurrence of atrial fibrillation post cardioversion: a randomized controlled trial**. *Int J Cardiol* (2019) **278** 133-136. PMID: 30522886
20. Lin GM, Colangelo LA, Lloyd‐Jones DM. **Association of sleep apnea and snoring with incident atrial fibrillation in the multi‐ethnic study of atherosclerosis**. *Am J Epidemiol* (2015) **182** 49-57. PMID: 25977516
21. Yin J, Jin X, Shan Z. **Relationship of sleep duration with all‐cause mortality and cardiovascular events: a systematic review and dose‐response meta‐analysis of prospective cohort studies**. *J Am Heart Assoc* (2017) **6**. DOI: 10.1161/JAHA.117.005947
22. Shamsuzzaman ASM, Winnicki M, Lanfranchi P. **Elevated C‐reactive protein in patients with obstructive sleep apnea**. *Circulation* (2002) **105** 2462-2464. PMID: 12034649
23. Jelic S, Padeletti M, Kawut SM. **Inflammation, oxidative stress, and repair capacity of the vascular endothelium in obstructive sleep apnea**. *Circulation* (2008) **117** 2270-2278. PMID: 18413499
24. Qi L. **MicroRNAs and other mechanisms underlying the relation between sleep patterns and cardiovascular disease**. *Eur Heart J* (2020) **41** 2502. PMID: 32380520
|
---
title: Associations of serum amino acids related to urea cycle with risk of chronic
kidney disease in Chinese with type 2 diabetes
authors:
- Wei Zhang
- Jun Zheng
- Jikun Zhang
- Ninghua Li
- Xilin Yang
- Zhong-Ze Fang
- Qiang Zhang
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC10018121
doi: 10.3389/fendo.2023.1117308
license: CC BY 4.0
---
# Associations of serum amino acids related to urea cycle with risk of chronic kidney disease in Chinese with type 2 diabetes
## Abstract
### Objective
Serum levels of amino acids related to urea cycle are associated with risk of type 2 diabetes mellitus (T2DM). Our study aimed to explore whether serum levels of amino acids related to urea cycle, i.e., arginine, citrulline, and ornithine, are also associated with increased risk of chronic kidney disease (CKD) in T2DM.
### Methods
We extracted medical records of 1032 consecutive patients with T2DM from the Electronic Administrative System of Liaoning Medical University First Affiliated Hospital (LMUFAH) system from May 2015 to August 2016. Of them, 855 patients with completed data available were used in the analysis. CKD was defined as estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m2. Serum amino acids were measured by mass spectrometry (MS) technology. Binary logistic regression was performed to obtain odds ratios (ORs) and their $95\%$ confidence intervals (CIs).
### Results
$52.3\%$ of the 855 T2DM patients were male, and 143 had CKD. In univariable analysis, high serum citrulline, high ratio of arginine to ornithine, and low ratio of ornithine to citrulline were associated with markedly increased risk of CKD (OR of top vs. bottom tertile: 2.87, $95\%$CI, 1.79-4.62 & 1.98, $95\%$CI,1.25-3.14 & 2.56, $95\%$CI, 1.61-4.07, respectively). In multivariable analysis, the ORs of citrulline and ornithine/citrulline ratio for CKD remained significant (OR of top vs. bottom tertile: 2.22, $95\%$CI, 1.29-3.82 & 2.24, 1.29-3.87, respectively).
### Conclusions
In Chinese patients with T2DM, high citrulline and low ornithine/citrulline ratio were associated with increased risk of CKD.
## Introduction
Chronic kidney disease (CKD) is one of the major microvascular complications of type 2 diabetes mellitus (T2DM) (1–3) and a leading cause of end-stage renal disease (ESRD) in patients with T2DM [4]. CKD is associated with an increasing risk of adverse outcomes including cardiovascular disease, infection and death, posing a significant threat to healthcare systems and having negative impacts on quality of life [5, 6]. Official figures showed that CKD affected between $8\%$ and $16\%$ of the population worldwide and was a leading cause of death [7]. A national survey conducted in China in 2012 revealed that the overall prevalence of CKD was $10.8\%$ (10.2–11.3) [8], which indicated that CKD has become an important public health problem in China.
Growing evidence from clinical research suggests that interventions such as glycemic control, blood pressure control, lipid-lowering measures, use of RAS blockers, weight control and lifestyle modifications can delay the progression of CKD, but the residual risk of CKD remains substantial. Therefore, it is essential to explore novel potential biomarkers that can be used to predict CKD in patients with T2DM.
Over recent decades, rapid advances in bioassay technology have led to the detection of a growing number of metabolites, which have provided new insights into pathways and biomarkers related to diabetes and its complications (9–12). It is worth noting that many studies have found that plasma amino acid levels are significantly altered in patients with T2DM or CKD. Amino acids involved in urea cycle might play an important role in inflammatory markers and oxidative stress [13, 14], many researchers took efforts to examine the relationship between such amino acids and T2DM or CKD. The complete urea cycle is mainly expressed in liver and the products and substrates in the cycle include arginine, citrulline and ornithine. An animal experiment showed that abnormal amino acid metabolism in the urea cycle might be associated with insulin resistance [15]. Nokhoijav et al. found that glutamine metabolism, urea cycle, and beta-oxidation make up crucial parts of the metabolic changes in T2DM. In addition, it has been demonstrated that CKD mice had high levels of arginine and citrulline [16]. Strong evidence supports the hypothesis that citrulline is a possible biomarker of kidney metabolism [17, 18]. However, the association between amino acids related to urea cycle and CKD in T2DM was largely unknown.
We conducted a cross-sectional survey in a Chinese population with T2DM to investigate any associations between serum levels of amino acids involved in urea cycle and the risk of CKD.
## Study design and population
The details of the study population and methods were described previously [19]. Briefly, from May 2015 to August 2016, we retrieved the electronic medical records of 1898 consecutive patients with a diagnosis of T2DM in Liaoning Medical University First Affiliated Hospital (LMUFAH), Jinzhou, China. T2DM was diagnosed by the 1999 World Health Organization’s criteria [20] or treated with antidiabetic drugs. A total of 1032 patients had metabolite data and complete data collection on age, gender, and body mass index (BMI). We used the data of 855 patients in the current analysis after excluding those who lacked creatinine or serum albumin information. The LMUFAH Clinical Research Ethics Committee approved the ethics of the study, and informed consent was waived due to the retrospective character of the cross-sectional study, which is consistent with the Helsinki Declaration.
## Data collection
Height, weight and blood pressure were measured by experienced physicians and nurses through standardized methods. To measure height and weight, study participants were asked to take off their shoes and heavy clothing. Body mass index (BMI) was calculated as weight in kilogram divided by height in squared meter height. Blood pressure was measured in a relaxed sitting position. Venous blood samples were drawn in the morning after at least 8 hours. High-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglyceride (TG), glycated hemoglobin (HbA1c), serum albumin (Alb), serum urea nitrogen (SUN) and serum creatinine (Scr) were assayed in the central laboratory of the hospital. Demographic, lifestyle information and clinical data were also documented, including age, gender, duration of diabetes, smoking status, alcohol consumption, diabetic retinopathy (DR), coronary heart disease (CHD), stroke, use of anti-diabetes drugs, lipid-lowering drugs and antihypertensive drugs.
## Clinical definitions
BMI was classified into four categories: underweight (<18.5 kg/m2), normal weight (18.5-23.9 kg/m2), overweight (24-27.9 kg/m2) and obesity (≥28 kg/m2) as recommended by the National Health Commission in China [21]. According to the criteria recommended by the American Diabetes Association [7], hyperglycemia was defined as HbA1c>$7.0\%$, and dyslipidemia was defined as TG≥1.7mmol/L, LDL-C ≥2.6mmol/L or HDL-C ≤1mmol/L in men and HDL-C ≤1.3mmol/L in women. Bilateral fundus photography was performed and DR was diagnosed by clinical manifestations of vascular abnormalities in the retina including microaneurysms, retinal hemorrhages, hard exudates or vitreous hemorrhage. Stroke was defined as subarachnoid hemorrhage, cerebral venous thrombosis, spinal cord stroke and ischemic stroke. CHD was defined as having a history of angina with abnormal electrocardiogram or on stress test, myocardial infarction, angina coronary artery bypass graft surgery or angioplasty.
## CKD definition
The new formula [22] for Chinese patients with type 2 diabetes was used to estimate glomerular filtration rate (eGFR) in milliliters per minute per 1.73m2, the formulas are: 313 × (Age)-0.494 × [Scr]-1.059 (mg/dl) × [Alb] 0.485 (g/dl) for men, and 783 × (Age)- 0.489 × [Scr]-0.877 (mg/dl) × [SUN]-0.150 (mg/dl) for women. In this analysis, CKD was defined as eGFR <60 mL/min/1.73m2 with or without kidney damage [23].
## Measurement of amino acids
The details of the quantification of the amino acids assessment method were mentioned in a previous article [24]. Briefly, mass spectrometry (MS) technology was applied to the metabolomics measurement. We collected capillary whole blood after a fast of at least 8 hours and stored it as dried blood spots (DBS) for metabolomic analysis. Metabolites in DBS were measured by direct infusion MS technology equipped with AB Sciex 4000 Qtrap system (AB Sciex, Framingham, MA, USA). High-purity water and Acetonitrile from Thermo Fisher (Waltham, MA, USA) were used as diluting agent and mobile phase.1-Butanol and acetyl chloride from Sigma-Aldrich (St Louis, MO, USA) was used to derive samples. Isotope-labeled internal standard samples of 12 amino acids (NSK-A) were purchased from Cambridge Isotope Laboratories (Tewksbury, MA, USA) while standard samples of the amino acids were purchased from Chrom systems (Grafelfing, Germany).
## Statistical analysis
For continuous variables, normally distributed data were presented as mean ± SD (standard deviation), and skewed variables were expressed as median (IQR). Normality was tested by checking the Q-Q plot. Qualitative variables were expressed as numbers (percentage). Non-paired Student t-test (or Mann-Whitney U test if appropriate) in continuous data and Chi-square test (or fisher test when appropriate) in categorical variables were used to compare the difference between CKD and non-CKD. Arginine, ornithine, citrulline and the ratios of any two of them were classified into three categories based on the 33rd and 66th percentiles, respectively. Binary logistic regression models were performed to obtain odds ratios (ORs) and their $95\%$ confidence intervals (CIs) of amino acids and the ratios of any two of them for the risk of CKD in univariate and multivariate analyses. A structured adjustment scheme was established to adjust for the effect of traditional risk factors on T2DM patients with CKD. We obtained the unadjusted OR values and the multivariable OR values adjusted for age, gender, BMI, duration of diabetes, systolic blood pressure (SBP), diastolic blood pressure (DBP), HDL-C, LDL-C, TG, HbA1c, smoking, drinking, anti-diabetes drugs, lipid-lowering drugs and antihypertensive drugs use.
Sensitivity analysis was performed to examine the consistency of the results in a 1032 population included patients with missing scream creatinine and serum albumin. A two-sided $P \leq 0.05$ in all analyses was considered statistically significant. SAS version 9.4 (SAS institute Inc., Cary, NC, USA) was used to conduct the statistical analysis.
## Characteristics of study subjects
The characteristics of the study patients are shown in Table 1. The patients had a mean age of 58.0 (SD: 13.8) years and a mean duration of T2DM of 5 [0-10] years. The mean BMI of the cohort was 25.3 (SD: 3.8) kg/m2, with $44.5\%$ of them being overweight and $21.0\%$ being obese. Of the 855 patients, $52.3\%$ were male, 143 were with CKD. The percentages of those with CHD, stroke and DR were 187 ($21.9\%$), 181 ($21.2\%$), and 115 ($13.5\%$), respectively. The subjects with CKD were older and had a longer duration of diabetes, higher SBP, lower DBP, worse clinical profile in HbA1c, HDL-C, LDL-C and TG. These patients were more likely to be female and less likely to smoke and drink. Also, these patients were more likely to use β-blockers use and less likely to use metformin. Nevertheless, patients with CKD were higher rates of CHD as compared with those without CKD. HDL-C, triglyceride, HbA1c and use of drugs other than metformin and β-blockers were similar in the two groups.
**Table 1**
| Variables | Total (n=855) | CKD (n=143) | Non-CKD (n=712) | P-value |
| --- | --- | --- | --- | --- |
| age | 58.0 ± 13.8 | 63.9 ± 14.4 | 56.9 ± 13.4 | <0.001 |
| Duration of diabetes, years | 5(0-10) | 8(2-13) | 5(0-10) | 0.001 |
| Male Gender | 447(52.3%) | 56(39.2%) | 391(54.92%) | <0.001 |
| BMI, kg/m2 | 25.3 ± 3.8 | 24.8 ± 4.1 | 25.4 ± 3.8 | 0.078 |
| BMI < 18.5 | | 8(5.6%) | 14(2.0%) | 0.025 |
| BMI ≥18.5 and < 24 | | 54(37.8%) | 230(32.3%) | |
| BMI ≥24 and < 28 | | 51(44.5%) | 317(44.5%) | |
| BMI ≥28 | | 30(21.0%) | 151(21.2%) | |
| Smoke, yes | | 35(24.5%) | 235(33.0%) | 0.045 |
| Drink, yes | | 24(16.8%) | 206(28.9%) | 0.003 |
| SBP, mmHg | | 144.7 ± 25.7 | 139.8 ± 24.0 | 0.030 |
| DBP, mmHg | | 78.7 ± 13.6 | 82.3 ± 13.3 | <0.001 |
| HDL-C, mmol/L | | 1.0(0.8-1.2) | 1.0(0.9-1.3) | 0.215 |
| <1.00 in men or <1.30 in women | | 76(53.2%) | 401(56.3%) | <0.001 |
| ≥1.00 in men or ≥1.30 in women | | 24(16.8%) | 209(29.4%) | |
| Unknown | | 43(14.3%) | 102(14.3%) | |
| LDL-C, mmol/L | | 2.6(2.1-3.1) | 2.8(2.3-3.4) | 0.027 |
| <2.60 | | 48(33.6%) | 243(34.1%) | <0.001 |
| ≥2.60 | | 52(36.4%) | 367(51.5%) | |
| Unknown | | 43(30.1%) | 102(14.3%) | |
| Triglyceride, mmol/L | | 1.7(1.1-2.2) | 1.7(1.1-2.4) | 0.718 |
| <1.70 | | 52(36.4%) | 316(44.4%) | <0.001 |
| ≥1.70 | | 48(33.6%) | 296(41.6%) | |
| Unknown | | 43(30.1%) | 100(14.0%) | |
| HbA1c, % | | 9.3 ± 2.5 | 9.6 ± 2.3 | 0.322 |
| <7 | | 18(12.6%) | 54(7.6%) | 0.014 |
| ≥7 and <8 | | 13(9.1%) | 84(11.8%) | |
| ≥8 | | 55(38.5%) | 355(49.9%) | |
| Unknown | | 57(39.9%) | 219(30.8%) | |
| Antidiabetic agents | | 122(85.3%) | 592(83.2%) | 0.524 |
| Insulin | | 110(76.9%) | 520(73.0%) | 0.335 |
| Metformin | | 32(22.4%) | 265(37.2%) | <0.001 |
| Other-OHD | | 61(42.7%) | 303(42.6%) | 0.982 |
| Hypotensive agents | | 69(48.3%) | 276(38.8%) | 0.035 |
| β-blockers | | 24(17.5%) | 60(8.4%) | 0.001 |
| ACEI | | 24(16.8%) | 89(12.5%) | 0.168 |
| ARB | | 20(14.0%) | 96(13.5%) | 0.873 |
| Lipid-lowering agents | | 50(35.0%) | 273(38.3%) | 0.447 |
| Statins | | 46(32.2%) | 266(37.4%) | 0.239 |
| Other lipid-lowering agents | | 4(2.8%) | 11(1.4%) | 0.489 |
| Coronary heart disease | 187(21.9%) | 47(32.9%) | 140(19.7%) | <0.001 |
| Stroke | 181(21.2%) | 38(26.6%) | 143(20.1%) | 0.083 |
| Diabetic retinopathy | 115(13.5%) | 22(15.4%) | 93(13.1%) | 0.458 |
The profile of serum amino acids related with urea cycle was shown in Table 2. Serum levels of citrulline, arginine and the ratios of arginine to ornithine and ornithine to citrulline were significantly different between the two groups. Compared with patients without CKD, the serum concentrations of citrulline and arginine were significantly higher in patients with CKD than those without. Ornithine was similar between the two groups. Moreover, the ratio of arginine to ornithine was markedly higher and the ratio of ornithine to citrulline was significantly lower in CKD than in those without CKD. Amino acids and their ratios were further stratified into tertiles, we observed that citrulline, the ratios of arginine to ornithine and ornithine to citrulline remained significant in the two comparison groups (all p values <0.05).
**Table 2**
| Variables | CKD b(n=143) | Non-CKD (n=712) | P-value |
| --- | --- | --- | --- |
| Arg, μmol/L | 12.6(7.0-17.9) | 9.6(5.4-17.1) | 0.030 |
| Arg <6.86, μmol/L | 35(24.5%) | 247(34.7%) | 0.059 |
| Arg≥6.86 and <14.69 μmol/L | 54(37.8%) | 227(31.9%) | |
| Arg≥14.69, μmol/L | 54(37.8%) | 238(33.4%) | |
| Cit, μmol/L | 22.9(18.1-32.5) | 19.2(15.3-24.9) | <0.001 |
| Cit<16.99, μmol/L | 28(19.6%) | 254(35.7%) | <0.001 |
| Cit≥16.99 and <23.19, μmol/L | 45(31.5%) | 237(33.3%) | |
| Cit≥23.19, μmol/L | 70(49.0%) | 221(31.0%) | |
| Orn, μmol/L | 17.4(12.8-23.6) | 17.4(13.0-23.7) | 0.767 |
| Orn<14.31, μmol/L | 46(32.2%) | 235(33.0%) | 0.762 |
| Orn≥14.31 and <20.98, μmol/L | 51(36.6%) | 232(32.6%) | |
| Orn≥20.98, μmol/L | 46(32.2%) | 245(34.4%) | |
| Arg:orn | 0.7(0.4-1.2) | 0.6(0.3-1.0) | 0.014 |
| Arg:orn <0.42 | 33(23.0%) | 252(35.4%) | 0.012 |
| Arg:orn≥0.42 and <0.85 | 50(35.0%) | 229(32.2%) | |
| Arg:orn≥0.85 | 60(42.0%) | 231(32.4%) | |
| Cit:arg | 2.0(1.3-3.2) | 1.9(1.2-3.3) | 0.372 |
| Cit:arg <1.38 | 44(30.8%) | 237(33.3%) | 0.705 |
| Cit:arg≥1.38 and <2.70 | 52(36.4%) | 234(32.9%) | |
| Cit:arg≥2.70 | 47(32.9%) | 241(33.9%) | |
| Orn:cit | 0.7(0.5-1.0) | 0.9(0.7-1.2) | <0.001 |
| Orn:cit <0.71 | 66(46.2%) | 217(30.5%) | <0.001 |
| Orn:cit≥0.71 and <1.04 | 46(32.2%) | 234(32.9%) | |
| Orn:cit≥1.04 | 31(21.7%) | 261(36.7%) | |
## Associations of amino acids and their ratios with CKD risk
Citrulline, the ratio of ornithine to citrulline and arginine to ornithine were associated with the risk of CKD in analysis. As shown in Table 3, the top tertile of serum of citrulline in univariate analysis was associated with markedly increased risk of CKD as compared with its bottom tertile (OR: 2.87 [$95\%$CI, 1.79-4.62]). After adjusting for possible confounders, i.e., age, gender, BMI, duration of T2DM, SBP, DBP, LDL-C, HDL-C, TG, HbA1c, drink, smoke, metformin, lipid-lowering drugs and β-blockers, in multivariable model 3, high levels of citrulline remained associated with significantly increased risk of CKD (OR: 2.38 [$95\%$CI, 1.42-4.00]). In both univariate and multivariable analysis, the ORs of bottom vs top tertiles of the ratio of ornithine to citrulline for CKD were 2.56 (1.61-4.07) and 2.24 (1.29-3.87), respectively, with all ORs remaining statistically significant (Table 4). For the ratio of arginine to ornithine, 3rd tertile increased the risk of CKD compared with 1st tertile in univariate analysis (OR:1.98 [$95\%$CI, 1.25-3.14]). Adjustmennt for traditional factors in multivariable model 3, the top tertile of the ratio of arginine to ornithine was also associated with increased risk of CKD as compared with its bottom tertile (OR: 2.14 [$95\%$CI, 1.28-3.57]). However, in multivariable model 4, further adjustment for the ornithine to citrulline ratio, the ratio of arginine to ornithine was no longer significant for the risk of CKD (Table 5).
## Sensitivity analysis
After inclusion the 177 study subjects with missing creatinine information and serum albumin information in the analysis, the effect sizes of citrulline and the ratio of ornithine to citrulline for CKD increased slightly and remained significant in univariable and multivariable analyses. ( Supplementary Tables 1, 2).
## Discussion
In this cross-sectional survey, we found that a higher concentration of citrulline was positively associated with increased risk of CKD in T2DM. The ratio of ornithine to citrulline was also associated with a markedly decreased risk of CKD in T2DM.
The progression of CKD increased the risk of all-cause mortality in T2DM patients [4, 25]. Unfortunately, CKD in T2DM develops silently until severe damage has occurred. It is therefore crucial to find biomarkers that can detect decreased kidney function early within T2DM. The pathogenesis of CKD is complex and multifactorial, age, gender and duration of T2DM were typically unmodifiable risk factors for CKD; glycemic control, hypertension, lipid abnormalities, smoking and physical activity were modifiable risk factors for CKD [26]. Insufficient insulin secretion was one of the risk factor of diabetic nephropathy [27]; 5-methoxytryptophan (5-MTP), acetylcarnitine, taurine and tiglylcarnitine, were strongly correlated with the development of CKD [28]. A prospective cohort study found that blood pressure levels, diabetes status, serum lipid status, obesity, smoking, and alcohol consumption affected the development of CKD [29]. However, these markers are not sufficient to fully understand the pathogenesis of CKD, especially in T2DM. In our current study, we found that citrulline significantly increased the risk of CKD in T2DM. Some previous studies observed that citrulline was associated with diabetes. Research performed by Zhou Yong et al. revealed that plasma citrulline levels elevated in diabetes [30]. A population-based study showed that plasma citrulline levels correlated with HbA1c levels [31]. An animal experiment concluded that plasma citrulline and ornithine levels were elevated in obese and insulin-deficient mice, and further suggested that citrulline could be an early indicator of obesity-dependent metabolic impairment [32]. Some encouraging data showed that citrulline could be a candidate predictor of renal injury, which was consistent with our findings. A GC-MS study showed that urinary metabolites differed between individuals with and without CKD in diabetes [33]. A previou experiment comprising twelve mice showed that citrulline and arginine levels were elevated in CKD mice [16]. Results derived from Framingham Heart Study (FHS) demonstrated that citrulline increased the risk of CKD (OR:1.48; $95\%$CI:1.19–1.83), indicating that the elevation of citrulline level in plasma might be a signal of underlying renal dysfunction [18]. The results of an eight-year follow-up study from Korea showed that high levels of citrulline were strongly associated with the development of CKD [34]. In addition, there were several studies documented that citrulline was associated with CKD progression or decreasing eGFR [17, 35, 36].
Interestingly, our study also found that the ornithine/citrulline ratio was reduced in CKD, which was in line with the results of some researchers. A study conducted in France in 2014 reported that citrulline/ornithine ratio increased in the late stages of CKD [36]. A population-based study from South Africa showed that citrulline/ornithine ratio was significantly higher in the diabetic group compared to the non-diabetic group in stage 1 of CKD, indicating that this ratio may predict early CKD [37]. It is worth noting that a small case-control study found that the ratio of citrulline to ornithine elevated in T2DM [38].
Citrulline is a non-essential amino acid, synthesized mainly in the intestine by the conversion of glutamine [39, 40]. In kidney, the enzyme dimethyl arginine dimethyl amino hydrolase (DDAH) metabolizes asymmetric dimethylarginine (ADMA) to generate dimethylamine and citrulline, and subsequently citrulline is converted to arginine, a process that requires catalysis by argininosuccinate synthase (ASS) and argininosuccinate lyase (ASL) [16, 41]. This could, to some extent, explain our findings that renal injury probably inhibits the activity of ASS or ASL, thereby affecting the conversion of citrulline to arginine (de novo synthesis of arginine) and resulting in abnormal arginine metabolism.
As we have described previously, citrulline and ornithine are two intermediates in arginine metabolism and an abnormal ratio of these two amino acids may reflect abnormal arginine metabolism. Abnormal ADMA metabolism is a manifestation of disorders of arginine metabolism. ADMA is produced by methylation of arginine residues in intracellular proteins by protein arginine N-methyltransferases (PRMTs), and over the past decades ADMA has been revealed to have biological properties that inhibit nitric oxide (NO) synthase, reduce NO bioavailability and lead to endothelial dysfunction [42, 43]. NO is a potent endothelial vasodilator that balances the constrictor and regulates vascular tone and blood pressure. Some findings showed that plasma ADMA levels increased in the early stage of CKD (44–47) and that ADMA was negatively associated with eGFR [48, 49]. Moreover, two cohort studies conducted in patients with CKD noted that ADMA favors CKD progression and renal function decline [50, 51]. ADMA, therefore, explains well many pathophysiological aspects of CKD. We speculate that elevated plasma citrulline levels or abnormal ratio of ornithine to citrulline may be a consequence of the massive accumulation of ADMA in the vivo, which inhibits the synthesis of NO, a deficiency of which is a major feature of CKD.
There are two plausible conjectures regarding the results of our study: [1] renal injury interferes with arginine metabolism, resulting in abnormal citrulline and the ratio of ornithine to citrulline; [2] impaired citrulline metabolism may limit the bioavailability of arginine for nitric oxide synthesis, leading to endothelial dysfunction affecting renal function. Either explanation, however, previous studies were consistent in suggesting that citrulline or the ratio of ornithine to citrulline was closely associated with renal function. Our study drew the same conclusion: in patients with CKD, citrulline levels were significantly elevated and the ratio of ornithine to citrulline decreased. Large cohort studies and animal experiments are needed to validate the actual link between these metabolites and CKD.
Our research had critical implications for both public health and clinical practice. We found that patients with CKD had higher levels of citrulline and lower levels of ornithine/citrulline ratio. Our current study discovered an association between amino acids related to the urea cycle and CKD, which dramatically widens our appreciation of pathological mechanisms of CKD. Furthermore, our findings shed light on the prevention of CKD and the delay of its progression.
There were several limitations in our current study. First, because our study was a retrospective cross-sectional investigation, the true causality could not be validated. However, our findings were consistent with some earlier studies, which might support our results. Second, given the nature of cross-sectional study, only statistical associations between citrulline, the ratio of ornithine to citrulline and CKD could be revealed, not causal relationships. Diabetes was not only accompanied by aberrations in amino acid metabolism, but also accelerated kidney damage, and we cannot determine whether urea cycle disturbances resulted from CKD in our study. These findings need to be confirmed in prospective cohort studies. Third, our study involved T2DM patients who were hospitalized and likely had more severe CKD and T2DM. Therefore, it is important to be cautious when extending our results to other populations. Forth, LDL-C, HDL-C, TG, and HbA1c had a large number of missing values. Taking into account that amino acids related to urea cycle rather than glucose and lipids were the main factors in this study, we regarded the missing values as a category. Finally, diet may affect citrulline levels [52], however, information on diet was not collected and not available to the analysis.
In conclusion, we found that high serum citrulline was significantly associated with increased risk of CKD and ornithine/citrulline ratio was inversely associated with risk of CKD in T2DM. Prospective cohort studies from different populations with T2DM are required to replicate our findings and mechanistic investigations are also warranted to understand molecular mechanisms underlying the biological link between serum amino acids involved in urea cycle and risk of CKD.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding author/s.
## Ethics statement
The studies involving human participants were reviewed and approved by First Affiliated Hospital of Liaoning Medical University. The ethics committee waived the requirement of written informed consent for participation. Written informed consent was not obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
QZ and Z-ZF conceived the project, designed experiments. WZ wrote the manuscript and analyzed data. JuZ collected the information and contributed to the writing of this manuscript. XY conceived the project, designed experiments and interpretated the data. NL contributed to the data interpretation and data analysis. JiZ collect the information and contributed to the data interpretation and wrote the manuscript and analyzed data. All authors edited the final version of the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1117308/full#supplementary-material
## References
1. Thomas MC, Cooper ME, Zimmet P. **Changing epidemiology of type 2 diabetes mellitus and associated chronic kidney disease**. *Nat Rev Nephrol* (2016) **12** 73-81. DOI: 10.1038/nrneph.2015.173
2. So WY, Kong AP, Ma RC, Ozaki R, Szeto CC, Chan NN. **Glomerular filtration rate, cardiorenal end points, and all-cause mortality in type 2 diabetic patients**. *Diabetes Care* (2006) **29**. DOI: 10.2337/dc06-0248
3. Thomas MC, Brownlee M, Susztak K, Sharma K, Jandeleit-Dahm KA, Zoungas S. **Diabetic kidney disease**. *Nat Rev Dis Primers* (2015) **1** 15018. DOI: 10.1038/nrdp.2015.18
4. Jiang G, Luk AOY, Tam CHT, Xie F, Carstensen B, Lau ESH. **Progression of diabetic kidney disease and trajectory of kidney function decline in Chinese patients with type 2 diabetes**. *Kidney Int* (2019) **95**. DOI: 10.1016/j.kint.2018.08.026
5. de Boer IH, Rue TC, Hall YN, Heagerty PJ, Weiss NS, Himmelfarb J. **Temporal trends in the prevalence of diabetic kidney disease in the united states**. *Jama* (2011) **305**. DOI: 10.1001/jama.2011.861
6. McCullough PA, Li S, Jurkovitz CT, Stevens L, Collins AJ, Chen SC. **Chronic kidney disease, prevalence of premature cardiovascular disease, and relationship to short-term mortality**. *Am Heart J* (2008) **156**. DOI: 10.1016/j.ahj.2008.02.024
7. Chen TK, Knicely DH, Grams ME. **Chronic kidney disease diagnosis and management: A review**. *Jama* (2019) **322**. DOI: 10.1001/jama.2019.14745
8. Zhang L, Wang F, Wang L, Wang W, Liu B, Liu J. **Prevalence of chronic kidney disease in China: a cross-sectional survey**. *Lancet* (2012) **379**. DOI: 10.1016/s0140-6736(12)60033-6
9. Bain JR. **Targeted metabolomics finds its mark in diabetes research**. *Diabetes* (2013) **62**. DOI: 10.2337/db12-1189
10. Floegel A, Stefan N, Yu Z, Mühlenbruch K, Drogan D, Joost HG. **Identification of serum metabolites associated with risk of type 2 diabetes using a targeted metabolomic approach**. *Diabetes* (2013) **62**. DOI: 10.2337/db12-0495
11. Wang TJ, Larson MG, Vasan RS, Cheng S, Rhee EP, McCabe E. **Metabolite profiles and the risk of developing diabetes**. *Nat Med* (2011) **17**. DOI: 10.1038/nm.2307
12. Zhou C, Zhang Q, Lu L, Wang J, Liu D, Liu Z. **Metabolomic profiling of amino acids in human plasma distinguishes diabetic kidney disease from type 2 diabetes mellitus**. *Front Med* (2021) **8**. DOI: 10.3389/fmed.2021.765873
13. Pietzner M, Kaul A, Henning AK, Kastenmüller G, Artati A, Lerch MM. **Comprehensive metabolic profiling of chronic low-grade inflammation among generally healthy individuals**. *BMC Med* (2017) **15** 210. DOI: 10.1186/s12916-017-0974-6
14. Carracedo J, Merino A, Briceño C, Soriano S, Buendía P, Calleros L. **Carbamylated low-density lipoprotein induces oxidative stress and accelerated senescence in human endothelial progenitor cells**. *FASEB J Off Publ Fed Am Societies Exp Biol* (2011) **25**. DOI: 10.1096/fj.10-173377
15. Li LO, Hu YF, Wang L, Mitchell M, Berger A, Coleman RA. **Early hepatic insulin resistance in mice: a metabolomics analysis**. *Mol Endocrinol* (2010) **24**. DOI: 10.1210/me.2009-0152
16. Mathew AV, Zeng L, Byun J, Pennathur S. **Metabolomic Profiling of Arginine Metabolome Links Altered Methylation to Chronic Kidney Disease Accelerated Atherosclerosis**. *J Proteom Bioinform* (2015) **Suppl 14**. DOI: 10.4172/jpb.S14-001
17. Cañadas-Garre M, Anderson K, McGoldrick J, Maxwell AP, McKnight AJ. **Proteomic and metabolomic approaches in the search for biomarkers in chronic kidney disease**. *J Proteomics* (2019) **193** 93-122. DOI: 10.1016/j.jprot.2018.09.020
18. Rhee EP, Clish CB, Ghorbani A, Larson MG, Elmariah S, McCabe E. **A combined epidemiologic and metabolomic approach improves CKD prediction**. *J Am Soc Nephrol: JASN* (2013) **24**. DOI: 10.1681/asn.2012101006
19. Li J, Cao YF, Sun XY, Han L, Li SN, Gu WQ. **Plasma tyrosine and its interaction with low high-density lipoprotein cholesterol and the risk of type 2 diabetes mellitus in Chinese**. *J Diabetes Invest* (2019) **10**. DOI: 10.1111/jdi.12898
20. Alberti KG, Zimmet PZ. **Definition, diagnosis and classification of diabetes mellitus and its complications. part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation**. *Diabetic Med J Br Diabetic Assoc* (1998) **15**. DOI: 10.1002/(sici)1096-9136(199807)15:7<539::Aid-dia668>3.0.Co;2-s
21. 21
National Health Commission of the People’s Republic of China. Criteria of weight for adults (2013). Available at: http://www.nhc.gov.cn/ewebeditor/uploadfile/2013/08/20130808135715967.pdf (Accessed October 3, 2022).. *Criteria of weight for adults* (2013)
22. Leung TK, Luk AO, So WY, Lo MK, Chan JC. **Development and validation of equations estimating glomerular filtration rates in Chinese patients with type 2 diabetes**. *Kidney Int* (2010) **77**. DOI: 10.1038/ki.2009.549
23. Levey AS, Eckardt KU, Tsukamoto Y, Levin A, Coresh J, Rossert J. **Definition and classification of chronic kidney disease: a position statement from kidney disease: Improving global outcomes (KDIGO)**. *Kidney Int* (2005) **67**. DOI: 10.1111/j.1523-1755.2005.00365.x
24. Wang Q, Sun T, Cao Y, Gao P, Dong J, Fang Y. **A dried blood spot mass spectrometry metabolomic approach for rapid breast cancer detection**. *OncoTarg Ther* (2016) **9**. DOI: 10.2147/ott.S95862
25. Coresh J, Turin TC, Matsushita K, Sang Y, Ballew SH, Appel LJ. **Decline in estimated glomerular filtration rate and subsequent risk of end-stage renal disease and mortality**. *Jama* (2014) **311**. DOI: 10.1001/jama.2014.6634
26. Harjutsalo V, Groop PH. **Epidemiology and risk factors for diabetic kidney disease**. *Adv chronic Kidney Dis* (2014) **21**. DOI: 10.1053/j.ackd.2014.03.009
27. Newgard CB. **Interplay between lipids and branched-chain amino acids in development of insulin resistance**. *Cell Metab* (2012) **15**. DOI: 10.1016/j.cmet.2012.01.024
28. Chen DQ, Cao G, Chen H, Argyopoulos CP, Yu H, Su W. **Identification of serum metabolites associating with chronic kidney disease progression and anti-fibrotic effect of 5-methoxytryptophan**. *Nat Commun* (2019) **10** 1476. DOI: 10.1038/s41467-019-09329-0
29. Yamagata K, Ishida K, Sairenchi T, Takahashi H, Ohba S, Shiigai T. **Risk factors for chronic kidney disease in a community-based population: a 10-year follow-up study**. *Kidney Int* (2007) **71**. DOI: 10.1038/sj.ki.5002017
30. Zhou Y, Qiu L, Xiao Q, Wang Y, Meng X, Xu R. **Obesity and diabetes related plasma amino acid alterations**. *Clin Biochem* (2013) **46**. DOI: 10.1016/j.clinbiochem.2013.05.045
31. Verdam FJ, Greve JW, Roosta S, van Eijk H, Bouvy N, Buurman WA. **Small intestinal alterations in severely obese hyperglycemic subjects**. *J Clin Endocrinol Metab* (2011) **96**. DOI: 10.1210/jc.2010-1333
32. Sailer M, Dahlhoff C, Giesbertz P, Eidens MK, de Wit N, Rubio-Aliaga I. **Increased plasma citrulline in mice marks diet-induced obesity and may predict the development of the metabolic syndrome**. *PloS One* (2013) **8**. DOI: 10.1371/journal.pone.0063950
33. Sharma K, Karl B, Mathew AV, Gangoiti JA, Wassel CL, Saito R. **Metabolomics reveals signature of mitochondrial dysfunction in diabetic kidney disease**. *J Am Soc Nephrol JASN* (2013) **24**. DOI: 10.1681/asn.2013020126
34. Lee H, Jang HB, Yoo MG, Park SI, Lee HJ. **mino Acid Metabolites Associated with Chronic Kidney Disease: An Eight-Year Follow-Up Korean Epidemiology Study**. *Biomedicines* (2020) **8**. DOI: 10.3390/biomedicines8070222
35. Shah VO, Townsend RR, Feldman HI, Pappan KL, Kensicki E, Vander Jagt DL. **Plasma metabolomic profiles in different stages of CKD**. *Clin J Am Soc Nephrol CJASN* (2013) **8**. DOI: 10.2215/cjn.05540512
36. Duranton F, Lundin U, Gayrard N, Mischak H, Aparicio M, Mourad G. **Plasma and urinary amino acid metabolomic profiling in patients with different levels of kidney function**. *Clin J Am Soc Nephrol CJASN* (2014) **9** 37-45. DOI: 10.2215/cjn.06000613
37. Mbhele T, Tanyanyiwa DM, Moepya RJ, Bhana S, Makatini MM. **Relationship between amino acid ratios and decline in estimated glomerular filtration rate in diabetic and non-diabetic patients in south Africa**. *Afr J Lab Med* (2021) **10**. DOI: 10.4102/ajlm.v10i1.1398
38. Kövamees O, Shemyakin A, Pernow J. **Amino acid metabolism reflecting arginase activity is increased in patients with type 2 diabetes and associated with endothelial dysfunction**. *Diabetes Vasc Dis Res* (2016) **13**. DOI: 10.1177/1479164116643916
39. Boelens PG, Melis GC, van Leeuwen PA, ten Have GA, Deutz NE. **Route of administration (enteral or parenteral) affects the contribution of l-glutamine to**. *Am J Physiol Endocrinol Metab* (2006) **291**. DOI: 10.1152/ajpendo.00252.2005
40. Boelens PG, van Leeuwen PA, Dejong CH, Deutz NE. **Intestinal renal metabolism of l-citrulline and l-arginine following enteral or parenteral infusion of l-alanyl-L-[2,15N]glutamine or l-[2,15N]glutamine in mice**. *Am J Physiol Gastrointest liver Physiol* (2005) **289**. DOI: 10.1152/ajpgi.00026.2005
41. Luiking YC, Ten Have GA, Wolfe RR, Deutz NE. **Arginine**. *Am J Physiol Endocrinol Metab* (2012) **303**. DOI: 10.1152/ajpendo.00284.2012
42. Kielstein JT, Zoccali C. **Asymmetric dimethylarginine: a novel marker of risk and a potential target for therapy in chronic kidney disease**. *Curr Opin Nephrol hypertension* (2008) **17**. DOI: 10.1097/MNH.0b013e328314b6ca
43. Schwedhelm E, Böger RH. **The role of asymmetric and symmetric dimethylarginines in renal disease**. *Nat Rev Nephrol* (2011) **7**. DOI: 10.1038/nrneph.2011.31
44. Mihout F, Shweke N, Bigé N, Jouanneau C, Dussaule JC, Ronco P. **Asymmetric dimethylarginine (ADMA) induces chronic kidney disease through a mechanism involving collagen and TGF-β1 synthesis**. *J Pathol* (2011) **223** 37-45. DOI: 10.1002/path.2769
45. Fleck C, Schweitzer F, Karge E, Busch M, Stein G. **Serum concentrations of asymmetric (ADMA) and symmetric (SDMA) dimethylarginine in patients with chronic kidney diseases**. *Clinica chimica acta; Int J Clin Chem* (2003) **336** 1-12. DOI: 10.1016/s0009-8981(03)00338-3
46. Kielstein JT, Böger RH, Bode-Böger SM, Frölich JC, Haller H, Ritz E. **Marked increase of asymmetric dimethylarginine in patients with incipient primary chronic renal disease**. *J Am Soc Nephrol JASN* (2002) **13**. DOI: 10.1681/asn.V131170
47. Uchida HA, Nakamura Y, Kaihara M, Norii H, Hanayama Y, Sugiyama H. **Steroid pulse therapy impaired endothelial function while increasing plasma high molecule adiponectin concentration in patients with IgA nephropathy**. *Nephrol dialysis Transplant Off Publ Eur Dialysis Transplant Assoc - Eur Renal Assoc* (2006) **21**. DOI: 10.1093/ndt/gfl423
48. Nkuipou-Kenfack E, Duranton F, Gayrard N, Argilés À, Lundin U, Weinberger KM. **Assessment of metabolomic and proteomic biomarkers in detection and prognosis of progression of renal function in chronic kidney disease**. *PLoS One* (2014) **9**. DOI: 10.1371/journal.pone.0096955
49. Stubbs JR, House JA, Ocque AJ, Zhang S, Johnson C, Kimber C. **Serum trimethylamine-N-Oxide is elevated in CKD and correlates with coronary atherosclerosis burden**. *J Am Soc Nephrol JASN* (2016) **27**. DOI: 10.1681/asn.2014111063
50. Ravani P, Tripepi G, Malberti F, Testa S, Mallamaci F, Zoccali C. **Asymmetrical dimethylarginine predicts progression to dialysis and death in patients with chronic kidney disease: a competing risks modeling approach**. *J Am Soc Nephrol JASN* (2005) **16**. DOI: 10.1681/asn.2005010076
51. Fliser D, Kronenberg F, Kielstein JT, Morath C, Bode-Böger SM, Haller H. **Asymmetric dimethylarginine and progression of chronic kidney disease: the mild to moderate kidney disease study**. *J Am Soc Nephrol JASN* (2005) **16**. DOI: 10.1681/asn.2005020179
52. Fragkos KC, Forbes A. **Citrulline as a marker of intestinal function and absorption in clinical settings: A systematic review and meta-analysis**. *United Eur Gastroenterol J* (2018) **6**. DOI: 10.1177/2050640617737632
|
---
title: Skeletal myotube-derived extracellular vesicles enhance itaconate production
and attenuate inflammatory responses of macrophages
authors:
- Atomu Yamaguchi
- Noriaki Maeshige
- Jiawei Yan
- Xiaoqi Ma
- Mikiko Uemura
- Mami Matsuda
- Yuya Nishimura
- Tomohisa Hasunuma
- Hiroyo Kondo
- Hidemi Fujino
- Zhi-Min Yuan
journal: Frontiers in Immunology
year: 2023
pmcid: PMC10018131
doi: 10.3389/fimmu.2023.1099799
license: CC BY 4.0
---
# Skeletal myotube-derived extracellular vesicles enhance itaconate production and attenuate inflammatory responses of macrophages
## Abstract
### Introduction
Macrophages play an important role in the innate immunity. While macrophage inflammation is necessary for biological defense, it must be appropriately controlled. Extracellular vesicles (EVs) are small vesicles released from all types of cells and play a central role in intercellular communication. Skeletal muscle has been suggested to release anti-inflammatory factors, but the effect of myotube-derived EVs on macrophages is unknown. As an anti-inflammatory mechanism of macrophages, the immune responsive gene 1 (IRG1)-itaconate pathway is essential. In this study, we show that skeletal muscle-derived EVs suppress macrophage inflammatory responses, upregulating the IRG1-itaconate pathway.
### Methods
C2C12 myoblasts were differentiated into myotubes and EVs were extracted by ultracentrifugation. Skeletal myotube-derived EVs were administered to mouse bone marrow-derived macrophages, then lipopolysaccharide (LPS) stimulation was performed and inflammatory cytokine expression was measured by RT-qPCR. Metabolite abundance in macrophages after addition of EVs was measured by CE/MS, and IRG1 expression was measured by RT-PCR. Furthermore, RNA-seq analysis was performed on macrophages after EV treatment.
### Results
EVs attenuated the expression of LPS-induced pro-inflammatory factors in macrophages. Itaconate abundance and IRG1 expression were significantly increased in the EV-treated group. RNA-seq analysis revealed activation of the PI3K-Akt and JAK-STAT pathways in macrophages after EV treatment. The most abundant miRNA in myotube EVs was miR-206-3p, followed by miR-378a-3p, miR-30d-5p, and miR-21a-5p.
### Discussion
Skeletal myotube EVs are supposed to increase the production of itaconate via upregulation of IRG1 expression and exhibited an anti-inflammatory effect in macrophages. This anti-inflammatory effect was suggested to involve the PI3K-Akt and JAK-STAT pathways. The miRNA profiles within EVs implied that miR-206-3p, miR-378a-3p, miR-30d-5p, and miR-21a-5p may be responsible for the anti-inflammatory effects of the EVs. In summary, in this study we showed that myotube-derived EVs prevent macrophage inflammatory responses by activating the IRG1-itaconate pathway.
## Graphical Abstract
## Introduction
Macrophages play an important role in the innate immune system and represent the front line of defense against bacterial infections [1]. They become activated in a pro-inflammatory way upon the detection of lipopolysaccharide (LPS), as characterized by the elevated expression of interleukin-1β (IL-1β), IL-6, and tumor necrosis factor-α (TNF-α) through the NF-κB pathway [2]. During several inflammatory diseases, suppressing excessive inflammatory responses of macrophages is crucial to avoid tissue damage [3].
Recently, mesenchymal stem cell-derived extracellular vesicles (EVs) have been reported to have an anti-inflammatory effect on macrophages [4]. EVs are lipid bilayer vesicles released from all types of cells and play a pivotal role in intercellular communication by encapsulating and delivering mRNAs, miRNAs, proteins, cytokines, and nucleic acids to distant organs and cells [5]. Also, cells change the release kinetics of EVs in response to various stimuli, and the contents in EVs also change in response to the cellular microenvironment [6].
Skeletal muscle is usually recognized as a locomotory organ, on the other hand, it is also known as the largest secretory organ in the human body and involved in as much as $75\%$ of the total metabolism in the body [7]. Furthermore, skeletal muscle is the only secretory organ that can be stimulated noninvasively and readily because it is widely distributed on the surface of the body and is a voluntarily controllable organ. In fact, it has been reported that high-intensity exercise with muscle contraction increases circulating EV amount [8] and that high-intensity ultrasound stimulation to cultured myotubes promotes EV secretion [9]. Thus, secretion of skeletal muscle-derived EVs can be more easily controlled than that of other organ-derived EVs, so the effect of skeletal muscle-derived EVs on macrophages is the key to controlling systemic inflammation. Furthermore, skeletal muscle is reported to secrete anti-inflammatory/immune modulatory factors [10]. However, the effect of skeletal myotube-derived EVs on macrophage inflammation has not been clarified yet.
As an important anti-inflammatory factor in macrophages, itaconate has been attracting attention. Itaconate, a tricarboxylic acid (TCA) cycle derivative, is known for its anti-inflammatory, antioxidant, anti-tumor, and anti-microbial properties [11]. Immune-responsive gene 1 (IRG1) has been regarded as a gene coding for immune-responsive gene 1 protein/cis-aconitic acid decarboxylase, an enzyme that catalyzes the production of itaconate by decarboxylating cis-aconitate [12]. During infection, macrophages reprogram immunometabolism by increasing itaconate production via upregulating IRG1 expression [13]. It is reported that for IRG1 activation, activities of the GR and JAK/STAT signaling pathways and the transcription factors C/ebpβ and Stat3 are required [14].
Here we show that skeletal myotube-derived EVs suppress macrophage inflammatory responses by inducing itaconate production in macrophages via upregulation of IRG1 expression.
## Cell culture
C2C12 myoblasts, mouse skeletal muscle cells, were purchased from ATCC. Myoblasts were seeded and cultured in 10 cm tissue culture dishes under $5\%$ CO2 in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with $10\%$ fetal bovine serum (FBS). When the cells reached $90\%$ confluence, the growth medium was changed to a differentiation medium (DMEM supplemented with $2\%$ horse serum) and differentiation into myotubes was started. After differentiation for 6 days refreshing the medium every 48 h, the efficiency of differentiation was confirmed by observing contraction by electrical stimulation (Supplemental Material 1). After differentiation, EVs were collected by incubating the myotubes in serum-free DMEM for 6 h.
To obtain bone marrow-derived macrophages (BMDMs), bone marrow cells were harvested from femurs and tibias of 7-week-old male C57BL/6J mice and cultured in a Petri dish under $5\%$ CO2 for eight days in RPMI 1640 with $10\%$ FBS, $25\%$ L929 cell supernatant, $1\%$ Penicillin/Streptomycin, and $1\%$ L-Glutamine. Differentiated BMDMs were plated in a 12-well tissue culture plate at a density of 3.0×105/well with macrophage culture media (RPMI 1640 supplemented with $10\%$ FBS, $10\%$ L929 cell supernatant, $1\%$ Penicillin/Streptomycin, and $1\%$ L-Glutamine).
The present study was approved by the Institutional Animal Care and Use Committee and all experiments were performed according to the Kobe University Animal Experimentation Regulations.
## EV extraction and addition to BMDMs
Myotube-derived EVs were isolated by ultracentrifugation following a previously described method [15]. Briefly, collecting medium was centrifuged at 1,000g for 10 min, followed by a second spin at 10,000g for 30 min to remove cell debris. Supernatant was collected and filtered through a 0.22 μm membrane, followed by a final centrifugation at 100,000g for 2 h to pellet EVs. The pellet was resuspended in macrophage culture media and filtered through a 0.22 μm membrane. Then, EVs were added to BMDMs at the concentration of 5.0×103 particles/cell. After 1.5 h-treatment by EVs, the cells were incubated with culture media overnight without myotube-derived EVs.
## EV characterization by tunable resistive pulse sensing, Western blotting and flow cytometry
The isolated EVs were characterized by their size and the presence of the EV marker CD63 [16]. The size distribution and concentration of EVs were measured using tunable resistive pulse sensing by qNano (Izon). The positive rate of CD63 in the collected EVs was analyzed using flow cytometry with a fluorescence-labeled CD63 antibody (Bio Legend Ltd., Japan) and magnetic beads coupled to a phosphatidylserine (PS)-binding protein (PS CaptureTM Exosome Flow Cytometry Kit, Fujifilm Wako Pure Chemical Co.) following the manufacturer’s instruction. EV-bound beads positive for CD63 antibody or isotype control were counted, and the positive rates were calculated using CytExpert (Beckman Coulter) software. Searching for the population of EV-bound beads using forward scatter (FSC) and side scatter (SSC) plots, the exosome population without aggregation was gated and the fluorescent signal in the corresponding histogram was evaluated. For Western blotting, EV proteins were extracted with $2\%$ SDS sample buffer for 10 min at 80°C and were migrated on $12.5\%$ sodium dodecyl sulfate-polyacrylamide gel. Following electrophoresis, proteins were transferred to a polyvinylidene difluoride membrane. The membrane was blocked with $5\%$ skim milk for 10 min at room temperature and then immunoblotted with anti-CD63 (1:200 dilution, sc-5275, Santa Cruz) at 4°C overnight. The membrane was then incubated with horseradish peroxidase-conjugated secondary antibodies (1:10,000, GE Healthcare, Waukesha, WI) for an hour. The membrane was detected using EzWestLumi One (ATTO) enhanced chemiluminescence solution. Finally, images were captured using the LAS-1000 imaging system (Fujifilm) with a chemiluminescent image analyzer.
## Cell viability assessment by Zombie Red™ immunofluorescence staining
The viability of BMDMs was analyzed 24 h after EV treatment or treatment with $1\%$ povidone-iodine (positive control) using Zombie Red™ as previously described [9]. Briefly, the cells were washed twice with PBS and stained with Zombie Red™ solution (1:1000) for 15 min. Then, BMDMs were fixed using $4\%$ paraformaldehyde for 30 min. After fixation, nuclei were stained with DAPI (1μg/mL) for 5 min. Stained images were observed using a BX50 fluorescence microscope at ×200 magnification (Olympus, Tokyo, Japan) and recorded with a digital camera (EOS Kiss X4, Canon, Tokyo, Japan). The numbers of total cells (blue) and dead cells (red) were counted and the percentage of live cells to total cells was calculated. Triplicate cell cultures and 5 random fields of each well were analyzed for each condition.
## RT-qPCR analysis
mRNAs from macrophages were isolated by TRIzol RNA Isolation protocol and used to make cDNA with iScriptTM cDNA Synthesis Kit (Bio-Rad). A StepOne™ Real-Time PCR thermal cycler was used to analyze the samples under the following conditions: 95° (3 min), 40 cycles of 95° (10 sec), and 60° (30 sec). The reaction mixture consisted of 8 μL cDNA, 1.5 μL 10× buffer, 0.3 μL 10 mM dNTPs, 1.5 μL 5 μM primers for each gene used in the study (F+R), 3.58 μL H2O, 0.075 μL Go Taq DNA polymerase, and 0.045 μL 2× SYBR green (Invitrogen). *Target* genes were Il-1β, Tnf-α, Il-6, Nf-kB p65, Nf-kB p50, and Irg1. For the analysis of pro-inflammatory genes, the cells were stimulated with 100 ng/mL LPS for 1.5 h at 24 h after EV treatment. Relative expression values for target genes were calculated by normalization to the expression of Glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Obtained data were analyzed by the delta/delta CT method [17]. The results are expressed as relative values with the control group or LPS-unstimulated group. The sequences for RT-qPCR primers are shown in Supplemental Material 2. Quadruplicate cell cultures and technical duplicates for each sample were analyzed.
## Metabolite analysis
At 24 h after EV treatment, BMDMs were washed with PBS twice and lysed in $80\%$ methanol containing 50 μM (+)-10-camphorsulfonic acid, 400 μM L-methionine sulfone, and 400 μM piperazine-1,4-bis(2-ethanesulfonic acid) (PIPES) as internal standards. The cells were incubated for 15 min at -80°, then scraped and centrifuged at 14,000 g for 5 min at 4°. The supernatant was collected and filtered using a Millipore 5 kDa cut-off membrane to remove solubilized proteins. The dried metabolites were dissolved in Milli-Q water after evaporation of the aqueous-layer extracts under vacuum using a FreeZone 2.5 Plus freeze dry system (Labconco, Kansas City, MO). The concentrations of intracellular metabolites were analyzed with a CE/MS (CE, Agilent G7100; MS, Agilent G6224AA LC/MSD TOF; Agilent Technologies, Palo Alto, CA) controlled by MassHunter Workstation Data Acquisition software (Agilent Technologies), as described previously [18]. The same validation was performed on BMDMs stimulated with LPS for 1.5 h at 24 h after EV treatment. For metabolite analysis on myotube-derived EVs, EVs extracted as above were used. The relative abundance of each metabolite to the control group or LPS group was calculated. Quadruplicate cell cultures were analyzed for each condition tested.
## RNA sequencing of BMDMs and myotube-derived EVs
Total RNA was extracted from BMDMs using TRIzol reagent (Takara Biotechnology, Japan) according to the manufacturer’s instructions. Raw RNA sequence data were obtained using an Illumina NovaSeq™ 6000 machine. After acquiring the raw data, the fold change (mean of each RNA in the EV group/mean of each RNA in the control group) and P-values were calculated for each RNA. These P-values were used to calculate the false discovery rate (FDR) for each RNA, which was further used as a filter to identify significant RNAs with a fold change ≥ 2 or ≤ 0.5 and an FDR < 0.05. The R 3.5.3 program was used to create the volcanic plots. The 20 most enriched pathways related to signaling transduction are presented and were used to reveal the associated pathways after a pathway analysis with the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database. For miRNA analysis in myotube-derived EVs, miRNA was extracted from isolated EVs and used to characterize miRNA profile in skeletal myotube-derived EVs using the above-mentioned method.
## Statistical analysis
All values are presented as mean ± SD. Statistical analysis was performed with Statistical 4 (OMS, Tokyo, Japan). Student’s t-test was used for two-group comparisons, and ANOVA (with Tukey’s multiple comparison test as a post-hoc analysis) was used for multiple comparisons. The sample size for each test needed to generate a power of at least 0.8 at a significance level of 0.05 (α = 0.05, β = 0.2) was calculated using power analysis using G Power software [19].
## Characterization of myotube-derived EVs
Isolated EVs were characterized using flow cytometry, Western blotting, and tunable resistive pulse sensing. As shown in Figure 1A, $84.3\%$ of the isolated particles were found to be positive for CD63 while less than $1\%$ of the particles treated with the isotype control showed the presence of CD63 (Figure 1B). In addition, presence of CD63 in the extracted EVs was also confirmed by Western blotting (Figure 1C). Regarding the size of the extracted EVs, most particles were within the size of 50-200 nm, which is EV size range (Figure 1D) [20].
**Figure 1:** *Extracellular vesicle (EV) characterization. (A) The presence of CD63 on isolated EVs was verified by flow cytometry. (B) CD63 was absent in particles treated with the isotype control. (C) Western blot analysis of CD63. (D) Size distribution of isolated EVs was measured by tunable resistive pulse sensing. n = 3.*
## Myotube-derived EVs attenuate LPS-induced inflammatory responses in BMDMs
Twenty-four hours after EV treatment, BMDMs were treated with 100 ng/mL LPS for 1.5 h as an inflammation model. To assess the effect of myotube-derived EVs on macrophage inflammation, mRNA expression levels of pro-inflammatory Il-1β, Il-6, and Tnf-α were measured. As shown in Figures 2A-E, LPS significantly upregulated the expression levels of Il-1β, Il-6, Tnf-α, Nf-kb p65, and Nf-kb p50 in BMDMs and myotube-derived EVs significantly prevented the upregulation of those factors. Furthermore, to investigate the effect of myotube-derived EVs on IRG1 expression in BMDMs, mRNA expression of Irg1 in BMDMs after EV treatment was measured. As shown in Figure 2F, EV treatment significantly increased the mRNA expression of Irg1 in BMDMs.
**Figure 2:** *mRNA expression levels in macrophages by qPCR. (A-E) Bone marrow-derived macrophages were stimulated with 100 ng/mL lipopolysaccharide (LPS) 24 h after extracellular vesicle (EV) treatment and the expression levels of pro-inflammatory genes were measured. (F)
Irg1 expression level in macrophages was measured after EV treatment. *p < 0.05, **p < 0.01 compared with LPS, **p < 0.01 compared with LPS, ††
p < 0.01 compared with CON (Student’s t-test). n = 4. Mean ± SD shown.*
## Myotube-derived EVs have no cytotoxicity on BMDMs
The cell viability of BMDMs was measured using Zombie Red™ immunofluorescence staining 24 h after EV treatment in order to examine the cytotoxicity of myotube-derived EVs on macrophages. As shown in Figure 3, while cells treated with $1\%$ povidone-iodine showed a significant decline in viability, EV treatment did not result in cell damage.
**Figure 3:** *Cell viability 24 h after extracellular vesicle (EV) treatment. (A) Bone marrow-derived macrophages were stained with Zombie Red™ immunofluorescence reagent. After fixation, the cells were counter-stained with DAPI. (B) The percentage of live cells to total cells was calculated. Triplicate cell cultures were analyzed and 5 random fields of each well were examined. Mean ± SD shown.*
## Myotube-derived EVs induce itaconate production in BMDMs
Intracellular levels of the metabolites were measured 24 h after EV treatment to assess the effect of myotube-derived EVs on the metabolite profile in BMDMs. As shown in Figure 4, itaconate was the most significantly elevated of the metabolites measured. Moreover, EVs induced an overall increase in metabolites in the TCA cycle. Subsequently, to assess the effect of EV treatment on the metabolite profile in BMDMs upon LPS stimulation, BMDMs were stimulated with LPS for 1.5 h at 24 h after EV treatment and the intracellular levels of the metabolites were quantified. As a result, EV+LPS group showed a significantly higher level of itaconate compared to LPS group (Figure 5).
**Figure 4:** *Metabolite profile in bone marrow-derived macrophages treated by myotube-derived extracellular vesicles. **p < 0.01 compared with CON (Student’s t-test). n = 4. Mean ± SD shown.* **Figure 5:** *Metabolite profile in bone marrow-derived macrophages treated by myotube-derived extracellular vesicles. *p < 0.05 compared with LPS (Student’s t-test). n = 4. Mean ± SD shown.*
## Metabolite analysis in myotube-derived EVs
To investigate the metabolite profile in C2C12 myotube-derived EVs, the abundance of metabolites in extracted EVs was quantified. As shown in Table 1, nine metabolites were detected and myotube-derived EVs were rich in lactate and pyruvate.
**Table 1**
| Metabolite | Abundance (nmol/1010 particles) |
| --- | --- |
| Lactate | 153.499 |
| Pyruvate | 5.042 |
| Citrate | 0.247 |
| Succinate | 0.156 |
| α-ketoglutarate | 0.055 |
| Malate | 0.052 |
| cis-Aconitate | 0.028 |
| Itaconate | 0.009 |
| D-Glucose 6-phosphate | 0.007 |
## RNA-seq analysis of EV-treated BMDMs
To investigate the mechanism by which myotube-derived EVs activated the IRG1-itaconate pathway in BMDMs, RNA sequencing analysis of BMDMs after EV treatment was performed. A total of 14,784 RNAs were identified by proteomic quantitative analysis, and according to the standard of a fold change of ≥ 2 or ≤ 0.5 as well as an FDR < 0.05, we screened 268 up-regulated RNAs and 95 down-regulated RNAs in the EV group versus the control group (Supplemental Material 3). Differentially expressed RNAs are displayed as a volcano plot (Figure 6A) and the top 10 of upregulated RNAs are shown in Table 2. Cxcl$\frac{1}{2}$ were the most upregulated RNAs in BMDMs by myotube-derived EV treatment. Enrichment pathway analysis was also conducted to identify the most activated pathways linked to signaling transduction after EV treatment. The 20 most enriched pathways are shown in Figure 6A, which included the PI3K-Akt, JAK-STAT, and adipocytokine signaling pathways. To compare the physiological action between myotube-derived EVs and pathological endotoxin, the same analyses were performed on BMDMs stimulated by LPS alone. Differentially expressed RNAs by LPS treatment are displayed as a volcano plot and the 20 most enriched pathways are shown in Figure 6B. The top 10 of upregulated RNAs by LPS stimulation are shown in Table 3. As a result, in LPS-stimulated group, Il12b was the most upregulated gene and the PI3K-Akt, JAK-STAT, and adipocytokine signaling pathways were not in the 20 most enriched pathways unlike the EV-treated group.
**Figure 6:** *RNA sequencing analysis of BMDMs after extracellular vesicle (EV) treatment. (A) Left: Volcano plot of differentially expressed RNAs in control group vs. EV group. Blue dots represent RNAs with statistically significant difference and red dots show RNAs with no statistically significant difference between EV group vs. control group. Right: Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis on differentially expressed RNAs; the 20 most enriched pathways linked to signaling transduction are presented. (B) Left: Volcano plot of differentially expressed RNAs in control group vs. LPS group. Blue dots represent RNAs with statistically significant difference and red dots show RNAs with no statistically significant difference between control group vs. LPS group. Right: Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis on differentially expressed RNAs; the 20 most enriched pathways linked to signaling transduction are presented. $$n = 3$.$* TABLE_PLACEHOLDER:Table 2 TABLE_PLACEHOLDER:Table 3
## miRNA profile in myotube-derived EVs
To investigate miRNAs contained in EVs, miRNA-seq analysis was performed. A total of 442 miRNAs were identified by proteomic quantitative analysis (Supplemental Material 4). The 20 most enriched miRNAs in the EVs are shown in Figure 7A and miR-206-3p was the most abundant miRNA, followed by miR-378a-3p, miR-30d-5p, and miR-21a-5p. Muscle-specific myomiRNAs accounted for 32.9 percent of total mapped miRNAs (Figure 7B).
**Figure 7:** *miRNA profile in skeletal myotube-derived extracellular vesicles (EVs). miRNA-sequencing analysis was performed on miRNAs extracted from the EVs. (A) The 20 most abundant miRNAs in myotube EVs. (B) Ratio of muscle-specific miRNAs to total miRNAs. n = 2.*
## Macrophage response to myotube-derived EV treatment in inflammation-related genes
To assess the response of macrophages to myotube EVs, pro-inflammatory mRNA expression levels in macrophages after EV treatment were measured. As shown in Figure 8A, EVs slightly upregulated the expression levels of pro-inflammatory factors in macrophages, but the response was very minor compared to the elevation caused by LPS. Furthermore, the expression levels of those pro-inflammatory factors were almost the same level as the control group 24 h after the treatment (Figure 8B).
**Figure 8:** *Macrophage responses to myotube-derived extracellular vesicle (EV) treatment in inflammation-related genes. (A) mRNA expression levels of pro-inflammatory factors were measured by qPCR after 1.5 h-treatment by skeletal muscle EVs or LPS. **p < 0.01 compared with CON, ††
p < 0.01 compared with LPS (Tukey’s multiple comparison test). (B) mRNA expression levels of pro-inflammatory factors were measured by qPCR 24 h after EV treatment. n = 4. Mean ± SD shown.*
## Discussion
This is the first report showing that skeletal myotube-derived EVs have an anti-inflammatory effect on macrophages. Skeletal myotube EVs prevented LPS-induced overexpression of inflammatory factors Il-1b, Il-6, and Tnf-a and suppressed the upregulation of the expression of Nf-kb in macrophages without causing a reduction in cell viability.
A variety of illnesses occur and develop as a result of the overexpression of IL-1β, IL-6, and TNF-α. Numerous inflammatory disorders have been found to improve with the suppression of these pro-inflammatory factors, and experiments evaluating their blockade have also been carried out [21]. Thus, the modulation of pro-inflammatory factors by myotube-derived EVs raises the possibility of a novel strategy for immune regulation utilizing skeletal muscle, which is the largest and most approachable secretory organ.
As the mechanism by which myotube-derived EVs exerted an anti-inflammatory effect in macrophages, Irg1 expression was upregulated in EV-treated macrophages, followed by an increase in the level of itaconate. Additionally, after LPS stimulation, EV-treated macrophages retained a higher level of itaconate compared to the untreated group. Given that IRG1 catalyzes the synthesis of itaconate [12], it is anticipated that myotube-derived EVs enhanced itaconate production in macrophages via IRG1 upregulation. Itaconate has been shown to reduce the severity of a variety of inflammatory disorders by reducing macrophage inflammatory responses (22–25). Furthermore, Lampropoulou et al. reported that itaconate alleviates inflammatory responses of macrophages in a concentration-dependent manner [26]. Therefore, enhancing itaconate synthesis in macrophages is crucial for the management of inflammatory disorders.
In this study, skeletal myotube-derived EVs most upregulated Cxcl1 and Cxcl2 expression levels in macrophages. Cxcl1 and Cxcl2 have been reported to activate the IRG1-inducing factor Protein Kinase C (PKC) (27–29). Whereas, during LPS-induced inflammation, other factors were preferentially upregulated. This suggests that skeletal myotube EVs trigger different responses in macrophages than the typical inflammatory responses induced by endotoxin. In addition, the result of pathway enrichment analysis shows that the PI3K-Akt signaling pathway, which induces PKC, the IL-17 signaling pathway, which induce C/ebpβ, the adipocytokine signaling pathway, which induces STAT3, and the JAK-STAT signaling pathways were within the top 20 most enriched pathways in EV-treated macrophages. Hall et al. reported that C/ebpβ, STAT3, and JAK-STAT pathway are involved in IRG1 activation [14]. On the other hand, in the top 20 most enriched pathways activated during LPS-induced inflammation, the PI3K-Akt, JAK-STAT, and adipocytokine signaling pathways, which were elevated after EV treatment, were not included. Thus, it is assumed that skeletal myotube-derived EVs activated the IRG1-itaconate pathway via multiple pathways, eliciting a response distinct from endotoxin-induced inflammatory responses.
miRNA analysis revealed that miRNA profile in skeletal muscle EVs is largely composed of miR-206-3p, a skeletal muscle-specific myomiRNA [30], and miR-378a-3p, a muscle-enriched miRNA [31]. Lin et al. reported that transfection of miR-mimic-206-3p into macrophages suppressed macrophage inflammation and transfection of miR-inhibitor-206-3p increased the level of inflammatory factors in macrophages [32]. Rückerl et al. identified miR-378a-3p as a factor contributing to the induction of anti-inflammatory macrophage reprogramming [33]. In addition, Kris et al. reported that miR-378a has anti-inflammatory effects on macrophages and its deficiency enhances severity of inflammation [34]. Taken together, miR-206 and miR-378a, which were abundant in skeletal myotube EVs, have been reported to exert anti-inflammatory effects in macrophages, but their detailed mechanisms and effects on the activation of the IRG1-itaconate pathway are still unclear and further studies are expected. Meanwhile, miR-30d, which was the third most abundant miRNA in skeletal myotube EVs, is reported to activate the JAK-STAT pathway by suppression of SOCS1 and SOCS3, negative regulators of the JAK-STAT pathway [35, 36]. Furthermore, the fourth most abundant miR-21a is reported to target PI3K-Akt inhibitor PTEN and downregulate its expression [37]. Based on these, it is suggested that these miRNAs may be involved in the activation of the JAK-STAT pathway and PI3K-Akt pathway in macrophages by skeletal myotube-derived EVs.
In this study, skeletal myotube EVs caused an elevation of inflammatory factors in macrophages to some extent. However, this response was dramatically small compared to the LPS-induced elevation of those factors. Moreover, after 24 h of EV treatment, the level of inflammatory factors recovered to the same level as non-treated group, suggesting that the inflammatory response caused by skeletal myotube-derived EVs is not a pathological hyperinflammatory response but a natural process of immunometabolism.
Additionally, myotube-derived EVs promoted the metabolism of the TCA cycle in macrophages as well as upregulation of lactate and pyruvate. Myotube EVs also increased the gene expression of phosphofructokinase, hexokinase, and pyruvate kinase, the rate-limiting enzymes of glycolysis, and isocitrate dehydrogenase and 2-oxoglutarate dehydrogenase, the rate-limiting enzymes of the TCA cycle (Supplemental Material 5), indicating that the EVs enhanced the metabolism of both glycolysis and the TCA cycle in macrophages. Metabolite analysis on myotube-derived EVs revealed that they are rich in lactate and pyruvate. Hui et al. reported that lactate and pyruvate can be a major carbon source, and thus energy source, for the TCA cycle [38]. Therefore, EVs possibly activated the TCA cycle by delivering myotube-derived lactate and pyruvate to macrophages. While the TCA cycle serves as an important regulatory function in driving energy production during macrophage activation, the accumulation of specific TCA cycle metabolites supports specific macrophage effector functions [39]. Thus, skeletal myotube-derived EVs, which can increase metabolites in the TCA cycle without causing excessive inflammatory responses, may be utilized to control macrophage dynamics.
While this study showed that skeletal myotube-derived EVs exert an anti-inflammatory effect by activating the IRG1-itaconate pathway via multiple pathways in macrophages, the details of how each pathway is involved in the activation of IRG1 are still unclear. Also, validation of the pathways activated by EVs and identification of contents of the skeletal myotube EVs which were responsible for these effects are expected as further studies.
In summary, this study found that skeletal myotube-derived EVs prevent macrophage inflammatory responses by activating the IRG1-itaconate pathway. These findings suggest a new immunoregulatory strategy utilizing skeletal muscle-derived EVs.
## Data availability statement
The original contributions presented in the study are publicly available. These data can be found here: https://ngdc.cncb.ac.cn/omix. ( OMIX repository, accession numbers OMIX003091 and OMIX003092).
## Ethics statement
The animal study was reviewed and approved by Kobe University Animal Care and Use Committee.
## Author contributions
AY: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing - original draft; NM: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing - original draft, Writing - review & editing; JY: Conceptualization, Supervision, Writing - review & editing; XM; Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft; MU: Formal analysis, Investigation, Methodology, Supervision, Validation, Visualization, Writing - original draft, Writing - review & editing; MM: Data curation, Formal analysis, Investigation, Software, Writing - review & editing; YN: Data curation, Funding acquisition, Resources, Software, Writing - review & editing; TH: Data curation, Formal analysis, Funding acquisition, Project administration, Resources, Software, Supervision, Writing - review & editing; HK: Conceptualization, Funding acquisition, Project administration, Supervision; HF: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing - review & editing; Z-MY: Conceptualization, Methodology, Project administration, Supervision, Writing - review & editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1099799/full#supplementary-material
## References
1. Weiss G, Schaible UE. **Macrophage defense mechanisms against intracellular bacteria**. *Immunol Rev* (2015) **264** 182-203. DOI: 10.1111/imr.12266
2. Orecchioni M, Ghosheh Y, Pramod AB, Ley K. **Corrigendum: Macrophage polarization: Different gene signatures in M1(LPS+) vs. classically and M2(LPS-) vs. alternatively activated macrophages**. *Front Immunol* (2020) **11**. DOI: 10.3389/fimmu.2020.00234
3. Funes SC, Rios M, Escobar-Vera J, Kalergis AM. **Implications of macrophage polarization in autoimmunity**. *Immunology* (2018) **154**. DOI: 10.1111/imm.12910
4. Harrell CR, Jovicic N, Djonov V, Arsenijevic N, Volarevic V. **Mesenchymal stem cell-derived exosomes and other extracellular vesicles as new remedies in the therapy of inflammatory diseases**. *Cells* (2019) **8** 1605. DOI: 10.3390/cells8121605
5. Andaloussi S EL, Mäger I, Breakefield XO, Wood MJ. **Extracellular vesicles: Biology and emerging therapeutic opportunities**. *Nat Rev Drug Discov* (2013) **12**. DOI: 10.1038/nrd3978
6. Kalluri R, LeBleu VS. **The biology, function, and biomedical applications of exosomes**. *Science* (2020) **367**. DOI: 10.1126/science.aau6977
7. Trovato E, Di Felice V, Barone R. **Extracellular vesicles: Delivery vehicles of myokines**. *Front Physiol* (2019) **10**. DOI: 10.3389/fphys.2019.00522
8. Whitham M, Parker BL, Friedrichsen M, Hingst JR, Hjorth M, Hughes WE. **Extracellular vesicles provide a means for tissue crosstalk during exercise**. *Cell Metab* (2018) **27** 237-251.e4. DOI: 10.1016/j.cmet.2017.12.001
9. Maeshige N, Langston PK, Yuan ZM, Kondo H, Fujino H. **High-intensity ultrasound irradiation promotes the release of extracellular vesicles from C2C12 myotubes**. *Ultrasonics* (2021) **110** 106243. DOI: 10.1016/j.ultras.2020.106243
10. Fiuza-Luces C, Garatachea N, Berger NA, Lucia A. **Exercise is the real polypill**. *Physiol (Bethesda)* (2013) **28**. DOI: 10.1152/physiol.00019.2013
11. Sano M, Tanaka T, Ohara H, Aso Y. **Itaconic acid derivatives: Structure, function, biosynthesis, and perspectives**. *Appl Microbiol Biotechnol* (2020) **104**. DOI: 10.1007/s00253-020-10908-1
12. Tallam A, Perumal TM, Antony PM, Jäger C, Fritz JV, Vallar L. **Gene regulatory network inference of immunoresponsive gene 1 (IRG1) identifies interferon regulatory factor 1 (IRF1) as its transcriptional regulator in mammalian macrophages**. *PloS One* (2016) **11**. DOI: 10.1371/journal.pone.0149050
13. O'Neill LAJ, Artyomov MN. **Itaconate: The poster child of metabolic reprogramming in macrophage function**. *Nat Rev Immunol* (2019) **19**. DOI: 10.1038/s41577-019-0128-5
14. Hall CJ, Boyle RH, Astin JW, Flores MV, Oehlers SH, Sanderson LE. **Immunoresponsive gene 1 augments bactericidal activity of macrophage-lineage cells by regulating β-oxidation-dependent mitochondrial ROS production**. *Cell Metab* (2013) **18**. DOI: 10.1016/j.cmet.2013.06.018
15. Monsel A, Zhu YG, Gennai S, Hao Q, Hu S, Rouby JJ. **Therapeutic effects of human mesenchymal stem cell-derived microvesicles in severe pneumonia in mice**. *Am J Respir Crit Care Med* (2015) **192**. DOI: 10.1164/rccm.201410-1765OC
16. Kowal J, Arras G, Colombo M, Jouve M, Morath JP, Primdal-Bengtson B. **Proteomic comparison defines novel markers to characterize heterogeneous populations of extracellular vesicle subtypes**. *Proc Natl Acad Sci USA* (2016) **113**. DOI: 10.1073/pnas.1521230113
17. Rao X, Huang X, Zhou Z, Lin X. **An improvement of the 2ˆ(-delta delta CT) method for quantitative real-time polymerase chain reaction data analysis**. *Biostat Bioinforma Biomath* (2013) **3** 71-85. PMID: 25558171
18. Kato H, Izumi Y, Hasunuma T, Matsuda F, Kondo A. **Widely targeted metabolic profiling analysis of yeast central metabolites**. *J Biosci Bioeng* (2012) **113**. DOI: 10.1016/j.jbiosc.2011.12.013
19. Kang H. **Sample size determination and power analysis using the G*Power software**. *J Educ Eval Health Prof* (2021) **18** 17. DOI: 10.3352/jeehp.2021.18.17
20. Biller SJ, Lundeen RA, Hmelo LR, Becker KW, Arellano AA, Dooley K. **Prochlorococcus extracellular vesicles: Molecular composition and adsorption to diverse microbes**. *Environ Microbiol* (2022) **24**. DOI: 10.1111/1462-2920.15834
21. Möller B, Villiger PM. **Inhibition of IL-1, IL-6, and TNF-alpha in immune-mediated inflammatory diseases**. *Springer Semin Immunopathol* (2006) **27** 391-408. DOI: 10.1007/s00281-006-0012-9
22. Ogger PP, Albers GJ, Hewitt RJ, O'Sullivan BJ, Powell JE, Calamita E. **Itaconate controls the severity of pulmonary fibrosis**. *Sci Immunol* (2020) **5**. DOI: 10.1126/sciimmunol.abc1884
23. Olagnier D, Farahani E, Thyrsted J, Blay-Cadanet J, Herengt A, Idorn M. **SARS-CoV2-mediated suppression of NRF2-signaling reveals potent antiviral and anti-inflammatory activity of 4-octyl-itaconate and dimethyl fumarate**. *Nat Commun* (2020) **11** 4938. DOI: 10.1038/s41467-020-18764-3
24. Song H, Xu T, Feng X, Lai Y, Yang Y, Zheng H. **Itaconate prevents abdominal aortic aneurysm formation through inhibiting inflammation**. *EBioMedicine* (2020) **57** 102832. DOI: 10.1016/j.ebiom.2020.102832
25. Zhang S, Jiao Y, Li C, Liang X, Jia H, Nie Z. **Dimethyl itaconate alleviates the inflammatory responses of macrophages in sepsis**. *Inflammation* (2021) **44**. DOI: 10.1007/s10753-020-01352-4
26. Lampropoulou V, Sergushichev A, Bambouskova M, Nair S, Vincent EE, Loginicheva E. **Itaconate links inhibition of succinate dehydrogenase with macrophage metabolic remodeling and regulation of inflammation**. *Cell Metab* (2016) **24**. DOI: 10.1016/j.cmet.2016.06.004
27. Chen B, Zhang D, Pollard JW. **Progesterone regulation of the mammalian ortholog of methylcitrate dehydratase (immune response gene 1) in the uterine epithelium during implantation through the protein kinase c pathway**. *Mol Endocrinol* (2003) **17**. DOI: 10.1210/me.2003-0207
28. Tsai YJ, Hao SP, Chen CL, Wu WB. **Thromboxane A2 regulates CXCL1 and CXCL8 chemokine expression in the nasal mucosa-derived fibroblasts of chronic rhinosinusitis patients**. *PloS One* (2016) **11**. DOI: 10.1371/journal.pone.0158438
29. Wang G, Huang W, Wang S, Wang J, Cui W, Zhang W. **Macrophagic extracellular vesicle CXCL2 recruits and activates the neutrophil CXCR2/PKC/NOX4 axis in sepsis**. *J Immunol* (2021) **207**. DOI: 10.4049/jimmunol.2100229
30. Horak M, Novak J, Bienertova-Vasku J. **Muscle-specific microRNAs in skeletal muscle development**. *Dev Biol* (2016) **410** 1-13. DOI: 10.1016/j.ydbio.2015.12.013
31. Xu T, Zhou Q, Che L, Das S, Wang L, Jiang J. **Circulating miR-21, miR-378, and miR-940 increase in response to an acute exhaustive exercise in chronic heart failure patients**. *Oncotarget* (2016) **7**. DOI: 10.18632/oncotarget.6966
32. Lin CC, Law BF, Hettick JM. **Acute 4,4'-methylene diphenyl diisocyanate exposure-mediated downregulation of miR-206-3p and miR-381-3p activates inducible nitric oxide synthase transcription by targeting Calcineurin/NFAT signaling in macrophages**. *Toxicol Sci* (2020) **173**. DOI: 10.1093/toxsci/kfz215
33. Rückerl D, Jenkins SJ, Laqtom NN, Gallagher IJ, Sutherland TE, Duncan S. **Induction of IL-4Rα-dependent microRNAs identifies PI3K/Akt signaling as essential for IL-4-driven murine macrophage proliferation**. *Blood* (2012) **120**. DOI: 10.1182/blood-2012-02-408252
34. Krist B, Florczyk U, Pietraszek-Gremplewicz K, Józkowicz A, Dulak J. **The role of miR-378a in metabolism, angiogenesis, and muscle biology**. *Int J Endocrinol* (2015) **2015** 281756. DOI: 10.1155/2015/281756
35. Wang S, Wen X, Han XR, Wang YJ, Shen M, Fan SH. **MicroRNA-30d preserves pancreatic islet β-cell function through negative regulation of the JNK signaling pathway**. *J Cell Physiol* (2018) **233**. DOI: 10.1002/jcp.26569
36. Lin X, Yu S, Ren P, Sun X, Jin M. **Human microRNA-30 inhibits influenza virus infection by suppressing the expression of SOCS1, SOCS3, and NEDD4**. *Cell Microbiol* (2020) **22**. DOI: 10.1111/cmi.13150
37. Li N, Qin JF, Han X, Jin FJ, Zhang JH, Lan L. **miR-21a negatively modulates tumor suppressor genes PTEN and miR-200c and further promotes the transformation of M2 macrophages**. *Immunol Cell Biol* (2018) **96** 68-80. DOI: 10.1111/imcb.1016
38. Hui S, Ghergurovich JM, Morscher RJ, Jang C, Teng X, Lu W. **Glucose feeds the TCA cycle**. *Nature* (2017) **551**. DOI: 10.1038/nature24057
39. Noe JT, Mitchell RA. **Tricarboxylic acid cycle metabolites in the control of macrophage activation and effector phenotypes**. *J Leukoc Biol* (2019) **106**. DOI: 10.1002/JLB.3RU1218-496R
|
---
title: 'Dietary B vitamins and glioma: A case–control study based on Chinese population'
authors:
- Weichunbai Zhang
- Jing Jiang
- Xun Kang
- Ce Wang
- Feng Chen
- Botao Zhang
- Shenglan Li
- Sijie Huang
- Wenbin Li
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC10018137
doi: 10.3389/fnut.2023.1122540
license: CC BY 4.0
---
# Dietary B vitamins and glioma: A case–control study based on Chinese population
## Abstract
### Background
Dietary antioxidants have long been thought to be likely to prevent the development of gliomas. Previous studies have reported vitamin A, C, and E protective effects against gliomas. B vitamins, one of the main vitamins in the diet, are closely related to human health, but the association with gliomas has rarely been reported.
### Objective
This study aimed to evaluate the relationship between five B vitamins and glioma.
### Methods
In this Chinese population-based case–control study, 506 glioma cases and 506 matched (age and sex) controls were included. The dietary intake of study participants was assessed using a valid 111-item food frequency questionnaire. The intake of five B vitamins was calculated based on participants’ dietary information from the food frequency questionnaire. The logistic regression model was used to examine the association between B vitamins and glioma, and the restriction cubic spline evaluated the dose–response relationship between the two.
### Results
After adjusting for confounding factors, thiamine (OR = 0.09, $95\%$CI: 0.05–0.20), riboflavin (OR = 0.12, $95\%$CI: 0.06–0.25), nicotinic acid (OR = 0.24, $95\%$CI: 0.12–0.47), folate (OR = 0.07, $95\%$CI: 0.03–0.15) and biotin (OR = 0.14, $95\%$CI: 0.07–0.30) in the highest tertile were associated with a significantly decreased risk of glioma compared with the lowest tertile. The results of thiamine and biotin in glioma with different pathological types and grades were different. The restricted cubic spline function showed significant dose–response relationships between the intake of five B vitamins and the risk of glioma. When B vitamins exceeded a specific intake, the risk of glioma did not change.
### Conclusion
Our study suggests that higher dietary intake of thiamine, riboflavin, nicotinic acid, and folate are associated with a decreased risk of glioma, but the results of biotin are not consistent among different populations. In the future, prospective studies should be conducted better to validate the effects of B vitamins on gliomas.
## Introduction
B vitamins are a group of water-soluble micronutrients required by all forms of cellular life, from bacteria to humans [1]. Unlike other nutrients, B vitamins are not classified based on chemical structural similarity but on their physiological functions in tissues and cells [2, 3]. As cofactors of hundreds of enzymes, B vitamins are mainly involved in energy metabolism, DNA and protein synthesis, and signal molecule synthesis [3, 4]. For B vitamins, the human body cannot synthesize itself, or the amount of synthesis is challenging to meet physiological needs, so it is still necessary to rely on various animal products and plants in the daily diet to obtain essential B vitamins [4]. In addition, exercise, alcohol consumption, certain drugs, and changes in body status can also affect the need for B vitamins [4, 5]. Thus, deficiencies in B vitamins remain a potential malnutrition problem worldwide [6].
Studies showed that the lack of B vitamins was closely related to cardiovascular diseases [7], neurodegenerative diseases [8, 9], kidney diseases [10], diabetes [11], and other chronic diseases. In recent years, the effects of B vitamins on cancer have also been discovered. Comin-Anduix et al. started subcutaneously injecting different doses of thiamine (vitamin B1) every day for 4 days after tumor implantation in mice. They found that low-dose thiamine can promote tumor growth, while high-dose thiamine can inhibit tumor growth. When thiamine supplementation was started on the 7th day before the tumor inoculation, the inhibitory effect was significantly enhanced, suggesting that thiamine has a preventive effect on cancer [12]. A meta-analysis of 6,184 colorectal cancer cases also found that higher intakes of thiamine could significantly reduce the risk of colorectal cancer (Odds ratio (OR) = 0.76, $95\%$confidence interval ($95\%$CI): 0.65–0.89) [13]. Similar results were also found for riboflavin (vitamin B2). Zschabitz et al. found that total riboflavin intakes were negatively associated with the risk of colorectal cancer in a prospective cohort of 88,045 postmenopausal women recruited from 1993 to 1998 (Hazard ratio (HR) = 0.81, $95\%$ CI: 0.66–0.99) [14]. Lu et al. also found that riboflavin had a protective effect against gastric cancer (OR = 0.56, $95\%$CI: 0.39–0.81) in the case–control study based on the Korean population, especially in the female population (OR = 0.52, $95\%$CI: 0.28–0.97) [15]. Although studies have also explored the relationship between nicotinic acid (vitamin B3) and digestive tract cancer, no significant results have been obtained [16]. Chen et al. found that niacinamide, a nicotinic acid derivative, can reduce the incidence of non-melanoma skin cancer by $23\%$ [17]. In contrast, folate (vitamin B9) had a broader range of effects against cancer. Lin et al. conducted a meta-analysis by including 10 studies on folate intake and pancreatic cancer and found that increasing dietary folate intake by 100 μg/day was associated with a $7\%$ decreased risk of pancreatic cancer (Relative risk (RR) = 0.93, $95\%$CI: 0.90–0.97) [18]. Some studies have found that folate was closely related to lung cancer [19], endometrial cancer [20], and prostate cancer [21].
Although these studies suggested that B vitamins were closely related to cancer, few studies reported the relationship between B vitamins and glioma. The pathogenesis of glioma was still unclear. It was currently believed that this mechanism may be related to genetic mutation of genes [22, 23], disorder of cell signal pathway [24], and defects in DNA damage repair [25]. Based on the available evidence, the physiological function of B vitamins also involved these aspects. Therefore, we could not ignore the impact of B vitamins on glioma. On the one hand, the general metabolic functions of B vitamins and their role in neurochemical synthesis may be considered to have specific effects on the brain [3], and the concentrations of B vitamins and their derivatives in the brain were significantly higher than in plasma (26–28). It seemed impossible to ignore the importance of B vitamins for the brain. On the other hand, the effects of other vitamins on glioma have been reported, especially vitamin A, vitamin C, and vitamin E. Epidemiological studies have shown that these vitamins with antioxidant effects have a certain preventive effect against glioma (29–31). In comparison, the evidence on B vitamins and glioma was minimal. Some studies have explored the association between B vitamins and brain tumors. Still, due to the variety of brain tumors, the results cannot represent the relationship between B vitamins and glioma, and these studies mainly focused on children (32–34). Therefore, we conducted a case–control study in a Chinese adult population to further explore the association between various B vitamins in the diet and glioma. This study evaluated the relationship between five B vitamins in the diet and glioma. It explored the dose–response relationship between the intake of B vitamins and the risk of glioma to provide the latest epidemiological evidence for vitamin prevention of glioma.
## Study population
This case–control study was initiated in 2021 and completed in 2022 at the Beijing Tiantan Hospital, Affiliated with Capital Medical University. Based on previous studies, we assumed that about $80\%$ of Chinese people took in B vitamins below the recommended level [35]. We further assumed that adequate B vitamins would reduce the risk of glioma by $43\%$ [31]. With $80\%$ power, and type I error of 0.05, the minimum sample size was calculated to be 256 cases and 256 healthy control subjects. Adult patients who were jointly diagnosed with glioma by neuro-oncology doctors and pathologists according to the 2021 neuro-oncology diagnostic criteria [36] about 3 months before the survey were included in the case group. On this basis, taking hormones and other drugs that interfere with diet, significant dietary behavior changes (such as weight loss, etc.), extreme energy intake (>5,000 or <400 kcal/day), pregnant women and nursing mothers, previous cancer (except glioma), and digestive, endocrine, and neurological conditions were excluded. The control group was recruited from the community’s healthy individuals who reported no glioma clinical manifestations and abnormalities in previous brain imaging studies. Each case was matched with the control by age (within 5 years) and sex during the study period. The corresponding controls were matched among the 506 eligible patients. In the end, a total of 506 pairs were included in the final statistical analysis. All participants provided informed consent, and the study protocol was approved by the Institutional Review Board of Beijing Tiantan Hospital, Capital Medical University (No. KY2022-203-02).
## Dietary assessment and calculation of B vitamins intake
The food frequency questionnaire was used to collect information on the type and amount of food intake of the research subjects in the past 12 months through face-to-face interviews. The food frequency questionnaire has been validated in previous studies, and its authenticity and reliability can meet the purpose of the study [37]. Based on this, referring to the existing articles on diet and glioma, we added and deleted several foods to make them more suitable for the study. The food frequency questionnaire in this study included refined grains ($$n = 9$$), whole grains ($$n = 2$$), tubers ($$n = 2$$), legumes and products ($$n = 5$$), vegetables ($$n = 23$$), fungi and algae ($$n = 4$$), fruits ($$n = 20$$), red meat ($$n = 4$$), poultry ($$n = 2$$), animal offal ($$n = 4$$), fish and seafood ($$n = 5$$), egg ($$n = 1$$), dairy products ($$n = 4$$), nut ($$n = 4$$), sweet food ($$n = 5$$), sugary drink ($$n = 3$$), tea and coffee ($$n = 3$$), condiment ($$n = 4$$), curing food ($$n = 3$$), processed products ($$n = 3$$), and alcohol ($$n = 4$$), a total of 114 items, basically covering the daily type of diet. The intake assessment for each food item consisted of three aspects: whether or not, the frequency of intake (daily/weekly/monthly), and single intake. In the questionnaire, g or ml was used as a unit to measure food intake, and pictures of different food volumes and qualities were provided as references to help subjects accurately assess intake. The daily intake of each food was calculated according to the frequency and single intake.
Five B vitamins were involved in the study, including thiamine, riboflavin, nicotinic acid, folate, and biotin (vitamin B7). The intakes of the five B vitamins were calculated using the China Food Composition Tables [38]. Chinese Food Composition Tables provided data on B vitamins and energy in each food. Combined with the daily intake of each food item, the total daily intake of five B vitamins and energy can be calculated.
## Other variables
In addition to the food frequency questionnaire, all participants were asked to complete other surveys, including basic information, disease history, and lifestyle habits. Basic information mainly included date of birth, sex, education levels (primary school and below, secondary school and university and above), occupation (manual workers, mental workers, or others), and household income (below 3,000 ¥/month, 3,000–10,000 ¥/month, or above 10,000 ¥/month). Disease history included allergies, head trauma, and family cancer, which have been reported as “yes” or “no.” Lifestyle habits included smoking status and physical activity. The subjects were classified as never smoking, former smoking, or current smoking according to their current smoking status. Physical activity was assessed using the International Physical Activity Questionnaire (IPAQ) [39], and metabolic equivalents were calculated to classify physical activity into low, moderate, or violent. In addition, according to previous studies, the proximity of residence to electromagnetic fields and or broadcast antennas may also be a risk factor for glioma, so defining this condition as living in a high-risk area was also listed as one of the potential confounding factors. For physical measurement, while the subjects were being examined, researchers measured their weight and height using calibrated instruments to calculate body mass index (BMI). BMI was measured by weight (kg)/height squared (m2).
To ensure survey quality, all surveys were conducted by one-on-one interviews with uniformly trained investigators with medical or epidemiological education.
## Statistical analysis
We used the t-test for continuous variables and the χ2 test for categorical variables to compare the general characteristics of the case and control groups. The Spearman correlation coefficient was used to evaluate the correlation between five B vitamins. We divided them into three groups based on the distribution of B vitamin intake. To explore the association between B vitamins and glioma, we used the lowest group as a reference and calculated the OR and $95\%$CI of each group by logistic regression. The univariate model (Model 1) was a crude model without adjusting for any confounding factors. The multivariate model (Model 2) adjusted for factors associated with the risk of glioma and B vitamin intake: age, BMI, occupation, education level, household income, high-risk residential areas, smoking status, history of allergies, history of head trauma, family history of cancer, physical activity, and energy intake.
We conducted a series of sensitivity analysis to test the robustness of our estimates by excluding participants with different ages, different sexes, different BMI, middle school and below, below 3,000 ¥/month, smoking, history of allergy or family history of cancer, and repeated regression analysis. In addition, to overcome the inherent limitations of B vitamins analysis as grade variables, the restricted cubic spline function was used to model the dose–response relationship in the multivariate adjustment model, with four nodes located at the 20th, 40th, 60th, and 80th percentiles of B-vitamin intake. The 10th percentile was used as the reference group (OR = 1) [40].
All analyses were performed using SPSS 26.0 and R 4.1.1. All reported p-values were 2-sided, and the significance level was set at $p \leq 0.05.$
## Characteristics of the study population and B vitamins
A total of 506 glioma patients were included in this study, including 7 patients with grade I, 98 cases with grade II, 73 cases with grade III, 255 cases with grade IV, and 73 cases that could not judge the pathological grade. The glioma population of each pathological grade and the corresponding control group had similar age distribution, and the sex composition was utterly consistent. Overall, glioma patients had higher BMI ($p \leq 0.001$), slightly fewer education levels ($p \leq 0.001$), more smoking ($$p \leq 0.039$$), more physical activity ($p \leq 0.001$), and were less likely to have allergies ($p \leq 0.001$) but more likely to have cancer in their families ($$p \leq 0.001$$). There were also differences in occupation ($$p \leq 0.024$$) and household income ($p \leq 0.001$; Table 1).
**Table 1**
| Unnamed: 0 | Grade I + II | Grade I + II.1 | Grade I + II.2 | Grade III | Grade III.1 | Grade III.2 | Grade IV | Grade IV.1 | Grade IV.2 | Others | Others.1 | Others.2 | Value of pa,b |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Case | Control | Value of pa | Case | Control | Value of pa | Case | Control | Value of pa | Case | Control | Value of pa | Value of pa,b |
| Age (years) | 38.50 ± 12.76 | 37.14 ± 12.73 | 0.443 | 41.40 ± 11.32 | 39.71 ± 11.09 | 0.365 | 44.38 ± 13.29 | 42.86 ± 13.02 | 0.193 | 43.62 ± 13.32 | 42.36 ± 12.86 | 0.562 | 0.072 |
| Sex, (%) | | | 1.000 | | | 1.000 | | | 1.000 | | | 1.000 | 1.000 |
| Male | 61.9 | 61.9 | | 58.9 | 58.9 | | 56.1 | 56.1 | | 45.2 | 45.2 | | |
| Female | 38.1 | 38.1 | | 41.1 | 41.1 | | 43.9 | 43.9 | | 54.8 | 54.8 | | |
| BMI (kg/m2) | 24.49 ± 3.06 | 23.14 ± 3.62 | 0.004 | 23.80 ± 3.05 | 23.05 ± 3.19 | 0.144 | 24.00 ± 3.35 | 23.12 ± 3.16 | 0.002 | 23.70 ± 3.35 | 22.69 ± 3.27 | 0.067 | <0.001 |
| High-risk residential area, (%) | | | 0.490 | | | 0.674 | | | 0.911 | | | 0.575 | 0.534 |
| Yes | 21.9 | 18.1 | | 20.5 | 17.8 | | 19.2 | 19.6 | | 28.8 | 24.7 | | |
| No | 78.1 | 81.9 | | 79.5 | 82.2 | | 80.8 | 80.4 | | 71.2 | 75.3 | | |
| Occupation, (%) | | | 0.190 | | | 0.597 | | | 0.095 | | | 0.009 | 0.024 |
| Manual workers | 29.5 | 21.9 | | 27.4 | 23.3 | | 22.0 | 20.4 | | 37.0 | 15.1 | | |
| Mental workers | 60.0 | 60.0 | | 53.4 | 61.6 | | 52.5 | 61.2 | | 39.7 | 57.5 | | |
| Others | 10.5 | 18.1 | | 19.2 | 15.1 | | 25.5 | 18.4 | | 23.3 | 27.4 | | |
| Education level, (%) | | | 0.009 | | | 0.005 | | | <0.001 | | | 0.001 | <0.001 |
| Primary school and below | 3.8 | 4.8 | | 8.2 | 1.4 | | 5.9 | 2.4 | | 13.7 | 1.4 | | |
| Middle school | 42.9 | 22.9 | | 41.1 | 23.3 | | 40.0 | 25.1 | | 45.2 | 30.1 | | |
| University and above | 53.3 | 72.4 | | 50.7 | 75.3 | | 54.1 | 72.5 | | 41.1 | 68.5 | | |
| Household income, (%) | | | <0.001 | | | <0.001 | | | <0.001 | | | 0.396 | <0.001 |
| <3,000 ¥/month | 11.4 | 17.1 | | 11.0 | 21.9 | | 6.7 | 18.4 | | 16.4 | 15.1 | | |
| 3,000–10,000 ¥/month | 76.2 | 49.5 | | 78.0 | 46.6 | | 78.4 | 47.8 | | 64.4 | 56.1 | | |
| >10,000 ¥/month | 12.4 | 33.3 | | 11.0 | 31.5 | | 14.9 | 33.7 | | 19.2 | 28.8 | | |
| Smoking status, (%) | | | 0.197 | | | 0.151 | | | 0.239 | | | 0.229 | 0.039 |
| Never smoking | 63.8 | 75.2 | | 74.0 | 71.2 | | 71.8 | 74.9 | | 68.5 | 80.8 | | |
| Former smoking | 13.3 | 9.6 | | 12.3 | 5.5 | | 13.7 | 9.0 | | 9.6 | 5.5 | | |
| Current smoking | 22.9 | 15.2 | | 13.7 | 23.3 | | 14.5 | 16.1 | | 21.9 | 13.7 | | |
| History of allergies, (%) | | | 0.092 | | | 0.149 | | | 0.116 | | | 0.007 | <0.001 |
| Yes | 5.7 | 12.4 | | 9.6 | 17.8 | | 8.6 | 12.9 | | 5.5 | 20.5 | | |
| No | 94.3 | 87.6 | | 90.4 | 82.2 | | 91.4 | 87.1 | | 94.5 | 79.5 | | |
| History of head trauma, (%) | | | 0.083 | | | 0.796 | | | 0.310 | | | 0.042 | 0.474 |
| Yes | 15.2 | 7.6 | | 11.0 | 12.3 | | 9.0 | 11.8 | | 13.7 | 4.1 | | |
| No | 84.8 | 92.4 | | 89.0 | 87.7 | | 91.0 | 88.2 | | 86.3 | 95.9 | | |
| Family history of cancer, (%) | | | 0.862 | | | 0.003 | | | 0.022 | | | 0.341 | 0.001 |
| Yes | 20.0 | 19.0 | | 39.7 | 17.8 | | 31.8 | 22.7 | | 28.8 | 21.9 | | |
| No | 80.0 | 81.0 | | 60.3 | 82.2 | | 68.2 | 77.3 | | 71.2 | 78.1 | | |
| Physical Activity, (%) | | | <0.001 | | | <0.001 | | | <0.001 | | | <0.001 | <0.001 |
| Low | 16.2 | 41.9 | | 12.3 | 39.7 | | 14.1 | 47.4 | | 9.6 | 52.1 | | |
| Moderate | 41.9 | 37.1 | | 42.5 | 42.5 | | 42.0 | 35.7 | | 37.0 | 31.5 | | |
| Violent | 41.9 | 21.0 | | 45.2 | 17.8 | | 43.9 | 16.9 | | 53.4 | 16.4 | | |
The education level ($p \leq 0.05$), household income ($p \leq 0.001$), and physical activity ($p \leq 0.001$) of glioma patients with different pathological grades were the same as the general population. In addition, compared with the corresponding control group, the group with grade I + II glioma had a higher BMI ($$p \leq 0.004$$), the group with grade III glioma had a higher percentage of family history of cancer ($$p \leq 0.003$$), the group with grade IV glioma had a higher BMI ($$p \leq 0.002$$) and a higher percentage of family history of cancer ($$p \leq 0.022$$), and the group with other glioma had more manual workers ($$p \leq 0.009$$), a lower history of allergies ($$p \leq 0.007$$) and a higher history of head trauma ($$p \leq 0.042$$). There were no significant differences in others (Table 1).
Regarding the intake of B vitamins, as shown in Table 2, the intakes of thiamine, riboflavin, nicotinic acid, folate, and biotin in the control group were significantly higher than those in the case group (Figure 1). In addition, there was a significant correlation between the intake of the individual B vitamins (Spearman coefficients ranged from 0.559 to 0.798; Supplementary Table S1).
## Association between B vitamins and glioma
The association results between the five B vitamins and glioma are shown in Table 2. In model 1, the intake of each B vitamin was significantly associated with the risk of glioma. After adjusting for age, BMI, and other variables (Model 2), the results for the categorical variable of B vitamin intake showed that, compared to the first tertile, the third tertile of thiamine was associated with a decreased risk of glioma (OR = 0.09, $95\%$CI: 0.05–0.20), the third tertile of riboflavin was associated with a decreased risk of glioma (OR = 0.12, $95\%$CI: 0.06–0.25), the third tertile of nicotinic acid was associated with a decreased risk of glioma (OR = 0.24, $95\%$CI: 0.12–0.47), the third tertile of folate was associated with a decreased risk of glioma (OR = 0.07, $95\%$CI: 0.03–0.15), and the third tertile of biotin was associated with a decreased risk of glioma (OR = 0.14, $95\%$CI: 0.07–0.30). The results of the analysis of the continuous variables showed that for each 0.1 mg/day increase in thiamine, the risk of glioma decreased by $9\%$ (OR = 0.91, $95\%$CI: 0.86–0.96), and for each 0.1 mg/day increase in riboflavin, the risk of glioma decreased by $30\%$ (OR = 0.70, $95\%$CI: 0.64–0.78), and for each 5 mg/day increase in nicotinic acid, the risk of glioma decreased by $38\%$ (OR = 0.62, $95\%$CI: 0.50–0.77), and for each 100 μg/day increase in folate, the risk of glioma decreased by $69\%$ (OR = 0.31, $95\%$CI: 0.23–0.42), and for each 10 μg/day increase in biotin, the risk of glioma decreased by $24\%$ (OR = 0.76, $95\%$CI: 0.66–0.88; Table 2).
## B vitamins and pathological classification and grading of glioma
Analysis of pathological subtypes of glioma showed that for astrocytoma, riboflavin (OR = 0.47, $95\%$CI: 0.28–0.78), nicotinic acid (OR = 0.56, $95\%$CI: 0.33–0.96), folate (OR = 0.11, $95\%$CI: 0.03–0.43) and biotin (OR = 0.55, $95\%$CI: 0.36–0.84) were associated with a significantly decreased risk, but the result of thiamine was not significant (OR = 0.91, $95\%$ CI: 0.81–1.01). For glioblastoma, thiamine (OR = 0.89, $95\%$CI: 0.79–0.99), riboflavin (OR = 0.79, $95\%$CI: 0.68–0.92), nicotinic acid (OR = 0.55, $95\%$CI: 0.36–0.83), and folate (OR = 0.24, $95\%$CI: 0.13–0.47) were associated with a significantly decreased risk, but the result of biotin was not significant (OR = 0.81, $95\%$ CI: 0.65–1.01). Due to the small sample size of oligodendroglioma, further analysis was not possible (Table 3).
**Table 3**
| Pathological classificationc | Model 1a | Value of p | Model 2b | Value of p.1 |
| --- | --- | --- | --- | --- |
| Astrocytoma | | | | |
| Thiamine | 0.97 (0.93–1.02) | 0.274 | 0.91 (0.81–1.01) | 0.084 |
| Riboflavin | 0.91 (0.86–0.97) | 0.004 | 0.47 (0.28–0.78) | 0.004 |
| Nicotinic acid | 0.92 (0.78–1.08) | 0.297 | 0.56 (0.33–0.96) | 0.035 |
| Folate | 0.58 (0.44–0.76) | <0.001 | 0.11 (0.03–0.43) | 0.001 |
| Biotin | 0.93 (0.83–1.04) | 0.212 | 0.55 (0.36–0.84) | 0.006 |
| Glioblastoma | | | | |
| Thiamine | 0.92 (0.88–0.97) | 0.001 | 0.89 (0.79–0.99) | 0.044 |
| Riboflavin | 0.93 (0.89–0.97) | <0.001 | 0.79 (0.68–0.92) | 0.002 |
| Nicotinic acid | 0.84 (0.73–0.96) | 0.011 | 0.55 (0.36–0.83) | 0.005 |
| Folate | 0.62 (0.53–0.74) | <0.001 | 0.24 (0.13–0.47) | <0.001 |
| Biotin | 0.95 (0.87–1.03) | 0.218 | 0.81 (0.65–1.01) | 0.065 |
The association between B vitamins and the pathological grade of glioma showed similar results. For low-grade gliomas, riboflavin (OR = 0.54, $95\%$CI: 0.37–0.79), nicotinic acid (OR = 0.50, $95\%$CI: 0.29–0.89), and folate (OR = 0.16, $95\%$CI: 0.06–0.44) were associated with a decreased risk, but the results of thiamine (OR = 0.90, $95\%$ CI: 0.80–1.01) and biotin (OR = 0.75, $95\%$ CI: 0.51–1.11) were not significant. For high-grade gliomas, thiamine (OR = 0.88, $95\%$CI: 0.82–0.96), riboflavin (OR = 0.73, $95\%$CI: 0.64–0.84), nicotinic acid (OR = 0.63, $95\%$CI: 0.46–0.86), folate (OR = 0.26, $95\%$CI: 0.16–0.43) and biotin (OR = 0.79, $95\%$CI: 0.65–0.96) were associated with a decreased risk (Table 4).
**Table 4**
| Glioma gradingc | Model 1a | Value of p | Model 2b | Value of p.1 |
| --- | --- | --- | --- | --- |
| Low grade | | | | |
| Thiamine | 0.96 (0.91–1.02) | 0.231 | 0.90 (0.80–1.01) | 0.055 |
| Riboflavin | 0.89 (0.82–0.96) | 0.003 | 0.54 (0.37–0.79) | 0.001 |
| Nicotinic acid | 0.85 (0.69–1.04) | 0.113 | 0.50 (0.29–0.89) | 0.018 |
| Folate | 0.60 (0.46–0.78) | <0.001 | 0.16 (0.06–0.44) | <0.001 |
| Biotin | 0.97 (0.87–1.08) | 0.572 | 0.75 (0.51–1.11) | 0.151 |
| High grade | | | | |
| Thiamine | 0.95 (0.91–0.98) | 0.002 | 0.88 (0.82–0.96) | 0.003 |
| Riboflavin | 0.94 (0.90–0.97) | <0.001 | 0.73 (0.64–0.84) | <0.001 |
| Nicotinic acid | 0.89 (0.80–0.99) | 0.026 | 0.63 (0.46–0.86) | 0.003 |
| Folate | 0.65 (0.57–0.74) | <0.001 | 0.26 (0.16–0.43) | <0.001 |
| Biotin | 0.97 (0.91–1.04) | 0.407 | 0.79 (0.65–0.96) | 0.018 |
## Sensitivity analysis
The results of sensitivity analysis showed that after excluding participants with different ages, different sexes, different BMI, middle school and below, below 3,000 ¥/month, smoking, history of allergy or family history of cancer, we observed that most results of thiamine, riboflavin, nicotinic acid and folate were consistent with the overall results. However, biotin was not significant in people with low BMI (Supplementary Table S2).
## Dose–response relationship
In Figure 2, we used restricted cubic splines to describe the relationships between B vitamins and glioma. There were significant dose–response relationships between the intake of five B vitamins and the risk of glioma. There was a nonlinear dose–response relationship between thiamine and the risk of glioma, and when intake exceeded 0.67 mg/day, the risk of glioma decreased significantly with increasing intake. After the intake exceeded 0.91 mg/day, the risk of glioma tended to stabilize (P-nonlinearity < 0.0001). There was a linear dose–response relationship between riboflavin and the risk of glioma, and the risk of glioma decreased significantly with the increase in intake. After the intake exceeded 1.06 mg/day, the risk of glioma tended to stabilize (P-nonlinearity = 0.4040). There was a nonlinear dose–response relationship between nicotinic acid and the risk of glioma, and when intake exceeded 12.83 mg/day, the risk of glioma decreased significantly with increasing intake. After the intake exceeded 18.03 mg/day, the risk of glioma tended to stabilize (P-nonlinearity = 0.0211). There was a linear dose–response relationship between folate and the risk of glioma. When the intake exceeded 224.27 μg/day, the risk of glioma decreased significantly with the increase in intake. After the intake exceeded 404.93 μg/day, the risk of glioma tended to stabilize (P-nonlinearity = 0.2168). There was a nonlinear dose–response relationship between biotin and the risk of glioma, and when intake exceeded 28.81 μg/day, the risk of glioma decreased significantly with increasing intake. After the intake exceeded 40.62 μg/day, the risk of glioma tended to stabilize (P-nonlinearity < 0.0001).
**Figure 2:** *The restricted cubic spline for the associations between B vitamins and glioma. The lines represent adjusted odds ratios based on restricted cubic splines for the intake in the regression model. Knots were placed at the 20th, 40th, 60th, and 80th percentiles of the B vitamins intake, and the reference value was set at the 10th percentile. The adjusted factors were the same as in Model 2. (A) Thiamine, (B) Riboflavin, (C) Nicotinic acid, (D) Folate, (E) Biotin.*
## Discussion
Previous observational studies on vitamins and glioma focused on vitamins A, C, and E, while studies on B vitamins were relatively rare. Our study evaluated the association between the intake of five B vitamins and glioma in the Chinese population. The results showed that thiamine, riboflavin, nicotinic acid, folate, and biotin were significantly negatively associated with the risk of glioma. The results of different pathological classifications were different. Thiamine only had a significant impact on glioblastoma. Similarly, it also existed between biotin and astrocytoma, and riboflavin and folate seemed to have a greater effect against astrocytoma. The results of different pathological grades were similar to those of pathological types. Thiamine and biotin only had significant effects against high-grade gliomas and had no significant correlation with low-grade gliomas, but the effects of riboflavin, nicotinic acid, and folate had a greater effect against low-grade glioma. The results of sensitivity analysis suggested that thiamine, riboflavin, nicotinic acid, and folate had relatively robust associations with glioma. Still, the results of biotin were not consistent with those of the general population, suggesting that its effects on glioma were different. The restricted cubic spline model further confirmed significant dose–response relationships between the five B vitamins and the risk of glioma, in which the dose–response relationships between thiamine, nicotinic acid, biotin, and glioma were nonlinear. In contrast, the dose–response relationships between riboflavin, folate, and glioma were linear.
Based on this, epidemiological studies have mainly focused on thiamine and nervous system diseases. Jung et al. found that thiamine deficiency can lead to Wernicke’s encephalopathy, which was often manifested as headache, inattention, irritability, confusion, apathy, impaired consciousness of the immediate situation, coma, and other obvious neurological symptoms [41]. But most studies on thiamine and cancer have focused on the body. Liu et al. based on a meta-analysis of seven articles and 6,184 colorectal cancer patients, found that high-dose thiamine intake significantly reduced the risk of colorectal cancer (OR = 0.76, $95\%$CI: 0.65–0.89) [13]. Cancarini et al. found that high intake of thiamine was associated with a lower risk of breast cancer (RR = 0.61, $95\%$CI: 0.38–0.97) in a prospective cohort of 10,786 women followed for an average of 16.5 years, especially in estrogen receptor-negative and progesterone receptor-negative breast cancer [42]. However, there have been few reported studies on thiamine and glioma. Our study might be the first to find that higher dietary thiamine intake can significantly reduce the risk of glioma (OR = 0.09, $95\%$CI: 0.05–0.20), but this effect varied among different pathological types and grades of gliomas. It was protective against high-grade gliomas, such as glioblastoma, but similar results were not found in astrocytomas or other low-grade gliomas. The results of the dose–response relationship suggested that the relationship between thiamine intake and the risk of glioma was non-linear, with no change in the risk of glioma beyond 0.91 mg/day (P-nonlinearity < 0.0001). But the mechanism of the effect of thiamine on glioma was still unclear, which was related to the antioxidation of thiamine in vivo [43, 44]. Oxidative stress and the production of free radicals may promote the occurrence and development of cancer [45], and gliomas were no exception [46]. Cell experiments showed that thiamine significantly decreased the malondialdehyde level and increased the levels of superoxide dismutase and catalase in C6 rat glioma cells, inhibiting oxidative stress to some extent [47].
Compared with thiamine, riboflavin was more closely related to cancer, especially in digestive tract cancer, riboflavin showed a certain preventive effect. In a case–control study involving 756 controls and 377 cases of gastric cancer, Lu et al. assessed dietary riboflavin intake through a food frequency questionnaire. They found that riboflavin was significantly negatively correlated with the risk of glioma (OR = 0.56, $95\%$CI: 0.39–0.81) and interacted with methionine synthetase reductase [15]. Based on the Women’s Health Initiative Observational Study cohort, Zschabitz et al. conducted a colorectal cancer study. They found that higher riboflavin intake significantly reduced the risk of colorectal cancer (RR = 0.81, $95\%$ CI: 0.66–0.99), but this association was statistically significant only when the intake exceeded 3.97 mg/day [14]. Similar results were also verified in a meta-analysis that included 8 articles [48]. However, the effect of riboflavin on cancer varied by location, and its protective effect has not been consistent in studies on other cancer sites [49, 49]. There were still few reports on riboflavin and glioma, and only studies have been conducted in the Middle East. Heydari et al. found no significant association between riboflavin and glioma by comparing riboflavin intake in 128 gliomas and 256 healthy individuals in a hospital case–control study in Iran (OR = 0.57, $95\%$CI: 0.18–1.78) [31]. This was different from our results. In this study, there was a significant negative correlation between riboflavin and the risk of glioma (OR = 0.12, $95\%$CI: 0.06–0.25), and consistent results were obtained in different pathological subtypes and grades of glioma. This was not contradictory to the Iranian study. On the one hand, there were dietary differences between the two regions, and dietary riboflavin intake was significantly different. There was no difference in riboflavin intake in the Iranian population between glioma patients and healthy people (case: 2.50 ± 0.58 mg/day, control: 2.60 ± 1.29 mg/day), and was much higher than that in our study population [51]. On the other hand, the dose–response relationship showed that when riboflavin intake exceeds 1.06 mg/day, the risk of glioma no longer changed, so it was necessary to repeat this study in different populations. Riboflavin can participate in the glutathione redox cycle, maintain the oxidative/antioxidant balance, resist oxidative stress, and inhibit cancer development [52]. Riboflavin also had a synergistic effect with folate in DNA synthesis and repair [53, 54], which provided a possible explanation for riboflavin to prevent the occurrence and development of glioma.
Based on the beneficial effects of nicotinic acid on the skin, most cancer studies related to nicotinic acid have also focused on areas rich in epithelial tissue. Several clinical trials and animal studies have shown that nicotinic acid can prevent skin cancer, especially non-melanoma skin cancer, by promoting DNA repair, inhibiting pro-inflammatory mediators, and reducing light damage to the skin [43, 55, 56]. Oral nicotinamide was found to be safe and effective in reducing the incidence of new non-melanoma skin cancer in high-risk patients in phase 3, double-blind, randomized, controlled trials [17]. In addition to skin cancer, nicotinic acid had a similar effect in other epithelial-rich areas, such as esophageal cancer [57] and endometrial cancer [58]. We also found the preventive effect of nicotinic acid in glioma populations. Higher nicotinic acid intake reduced the risk of glioma by $76\%$ (OR = 0.24, $95\%$CI: 0.12–0.47) and had roughly the same effect on astrocytoma and glioblastoma. The dose–response relationship suggested that the effect of nicotinic acid against the risk of glioma was slightly different before and after the intake of 18.03 mg/day. At present, the mechanism of nicotinic acid inhibiting the occurrence and development of glioma was not clear. Yang et al. observed that nicotinic acid selectively targeted glioblastoma cells and retained most normal glial cells and neurons through overnight treatment of U251 glioblastoma cells with different concentrations of nicotinic acid. It was also observed that nicotinic acid decomposes F-actin stress fibers, which in turn affects cell-matrix adhesion, suggesting that nicotinic acid can inhibit the invasion of glioma cells [59]. Li et al. found that nicotinic acid can cause the loss of mesenchymal phenotype in U251 glioblastoma cells. It has been found to inhibit glioma invasion in vitro and in vivo. The primary mechanism was that nicotinic acid promoted snail1 degradation and enhanced intercellular adhesion, indicating that epithelial-mesenchymal transition in glioma cells was inhibited [60].
Of all the B vitamins, folate had the most extensive effect on cancer. Increasing folate intake within the physiological dose had a protective effect against cancer [61]. Epidemiological studies have shown that higher folate intake can significantly reduce the risk of colorectal cancer [62], pancreatic cancer [18], and endometrial cancer [20]. But the reverse results have been obtained in the study of lung cancer [19] and prostate cancer [21]. Compared with other B vitamins, there were more studies on folate and brain tumors. However, most of them focused on assessing the relationship between the folate intake of pregnant women and the risk of brain tumors in their offspring. Milne et al. collected 327 brain tumor cases and 867 healthy controls based on 10 Australian pediatric oncology centers and collected maternal folate intake through a questionnaire and found that higher folate intake during pregnancy significantly reduced the risk of brain tumors in offspring (OR = 0.60, $95\%$CI: 0.38–0.98), and this association was also found in low-grade gliomas (OR = 0.44, $95\%$CI: 0.22–0.89) [33]. Another study based on this population also reported that folate in childhood (3–14 years old) was associated with an overall reduced risk of brain tumors (OR = 0.63, $95\%$CI: 0.41–0.97), especially in low-grade gliomas (OR = 0.52, $95\%$CI: 0.29–0.92) [32]. The study on folate intake and glioma in adults was still insufficient. Our study found the epidemiological evidence for the protective effect of folate intake against gliomas in adults (OR = 0.31, $95\%$CI: 0.23–0.42), and the results for low-grade gliomas, including astrocytomas, were similar to previous studies, that was, folate intake had a more significant protective effect against low-grade gliomas. The study of serum folate concentration and glioma also found that the proportion of patients with glioma who were lower than the serum folate biological reference value was higher ($p \leq 0.001$) [63]. However, Chen et al. conducted a population-based case–control study in eastern Nebraska and found that dietary folate intake was not associated with the risk of glioma (OR = 0.90, $95\%$CI: 0.50–1.50) and was considered to be related to the small sample size of the study [64]. From the results of the dose–response relationship, the range of folate intake was also consistent with previous studies. It was generally believed that there was a U-shaped relationship between folate and cancer and that when folate intake was between 150 μg/day and 1,000 μg/day, the risk of cancer was significantly reduced [61]. Folate is essential for normal DNA synthesis and repair. Its deficiency can activate proto-oncogenes by reducing intracellular S-adenosyl methionine and altering cytosine methylation in DNA, which can also lead to an imbalance of DNA precursors, misincorporation of uracil into DNA and chromosome breakage, which may be the potential mechanism of folate in preventing gliomas [65]. In addition, folate can also be used as an adjuvant to assist in treating gliomas. It limited the proliferation of glioma cells and increased the sensitivity to temozolomide-induced apoptosis by inducing DNA methylation [66].
The study of biotin and cancer has mainly focused on biotin as a ligand in the treatment of new drugs [67, 68]. Few studies have reported the effect of biotin on cancer risk. Although our study found the protective effect of biotin against glioma, the results were inconsistent in sensitivity analysis, such as different BMI. In different pathological types, biotin was only significantly associated with astrocytoma (OR = 0.55, $95\%$CI: 0.36–0.84) but not in glioblastoma (OR = 0.81, $95\%$CI: 0.65–1.01). In different pathological grades, the results of low-grade gliomas and high-grade gliomas were also the opposite. This had to make us carefully consider the relationship between biotin and glioma.
This study had several advantages. First, we explored the relationship between five dietary B vitamins and glioma. As previous studies have paid more attention to the relationship between vitamin A, vitamin C, vitamin E, and glioma, the results of this study supplement the evidence of the preventive effect of B vitamins against glioma. Secondly, the studies on B vitamins and glioma often stayed in cell and animal experiments. The results of this study were mutually confirmed with the existing experimental evidence, especially for thiamine, nicotinic acid, and folate, which were lack of clinical study. Third, we thoroughly explored the relationship between gliomas with different pathological subtypes, pathological grades, and B vitamins. The dose–response relationships between B vitamins and the risk of glioma were described for the first time, which were consistent with the existing dose–response relationship between B vitamins and cancer. These provided further population evidence for the prevention and treatment of glioma by B vitamins. However, there were still some limitations in this study. On the one hand, B vitamins participated in one-carbon metabolism in different forms, such as the synthesis of nucleotides and the metabolism of amino acids, which was of great significance for maintaining normal cell growth. However, because the food composition table failed to provide a broader content of B vitamins, such as vitamin B12, this study only evaluated the five common B vitamins, and could not explore the association between more B vitamins and glioma, especially the interaction between B vitamins. But this has been far the most comprehensive result about B vitamins and gliomas. On the other hand, the association between the concentration of B vitamins in the body and glioma has not been evaluated. Although the food frequency questionnaire had been verified and had good repeatability and validity. However, due to the limitation of methods, there was still a certain bias in the assessment of the intake of B vitamins through the food frequency questionnaire. Assessing the concentration of B vitamins in the body can reduce this uncertainty. Since the concentration of B vitamins in the subjects was not detected in this study, the association between internal exposure and glioma should be evaluated in future studies in combination with the circulating concentration of B vitamins in or biomarkers. In addition, since this study was a case–control study, their causality cannot be verified.
## Conclusion
In summary, in this study of B vitamins and gliomas, we observe that thiamine, riboflavin, nicotinic acid, and folate were associated with a significantly decreased risk of gliomas. Still, the roles of biotin are different in different populations. The relationship between them should be further verified by prospective cohort studies in the future.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by the Institutional Review Board of Beijing Tiantan Hospital, Capital Medical University (no. KY2022-203-02). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
WL and WZ contributed to the conception or design of the work and wrote the manuscript. WZ, JJ, and SH contributed to data collection and analysis. XK, CW, FC, BZ, and SL contributed to data collection and management. All authors contributed to the article and approved the submitted version.
## Funding
This study was supported by the Talent Introduction Foundation of Tiantan Hospital (no. RCYJ-2020-2025-LWB) and Advanced Research and Training Program of Beijing Double Leading Scholars from China academy of Chinese Medical Science (no. 2-759-02-DR).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1122540/full#supplementary-material
## References
1. Liu Z, Farkas P, Wang K, Kohli MO, Fitzpatrick TB. **B vitamin supply in plants and humans: the importance of vitamer homeostasis**. *Plant J* (2022) **111** 662-82. DOI: 10.1111/tpj.15859
2. Said HM. **Intestinal absorption of water-soluble vitamins in health and disease**. *Biochem J* (2011) **437** 357-72. DOI: 10.1042/BJ20110326
3. Kennedy DO. **B vitamins and the brain: mechanisms, dose and efficacy–a review**. *Nutrients* (2016) **8** 68. DOI: 10.3390/nu8020068
4. Peterson CT, Rodionov DA, Osterman AL, Peterson SN. **B vitamins and their role in immune regulation and cancer**. *Nutrients* (2020) **12** 12. DOI: 10.3390/nu12113380
5. Revuelta JL, Buey RM, Ledesma-Amaro R, Vandamme EJ. **Microbial biotechnology for the synthesis of (pro) vitamins, biopigments and antioxidants: challenges and opportunities**. *Microb Biotechnol* (2016) **9** 564-7. DOI: 10.1111/1751-7915.12379
6. Whatham A, Bartlett H, Eperjesi F, Blumenthal C, Allen J, Suttle C. **Vitamin and mineral deficiencies in the developed world and their effect on the eye and vision**. *Ophthalmic Physiol Opt* (2008) **28** 1-12. DOI: 10.1111/j.1475-1313.2007.00531.x
7. Jeon J, Park K. **Dietary vitamin b6 intake associated with a decreased risk of cardiovascular disease: a prospective cohort study**. *Nutrients* (2019) **11** 11. DOI: 10.3390/nu11071484
8. Shrestha L, Shrestha B, Gautam K, Khadka S, Mahara RN. **Plasma vitamin b-12 levels and risk of alzheimer's disease: a case-control study**. *Gerontol Geriatr Med* (2022) **8** 1692866509. DOI: 10.1177/23337214211057715
9. Qiu F, Wu Y, Cao H, Liu B, Du M, Jiang H. **Changes of peripheral nerve function and vitamin b12 level in people with parkinson's disease**. *Front Neurol* (2020) **11** 549159. DOI: 10.3389/fneur.2020.549159
10. Hsueh YM, Huang YL, Lin YF, Shiue HS, Lin YC, Chen HH. **Plasma vitamin b12 and folate alter the association of blood lead and cadmium and total urinary arsenic levels with chronic kidney disease in a Taiwanese population**. *Nutrients* (2021) **13** 13. DOI: 10.3390/nu13113841
11. Saravanan P, Sukumar N, Adaikalakoteswari A, Goljan I, Venkataraman H, Gopinath A. **Association of maternal vitamin b12 and folate levels in early pregnancy with gestational diabetes: a prospective UK cohort study (pride study)**. *Diabetologia* (2021) **64** 2170-82. DOI: 10.1007/s00125-021-05510-7
12. Comin-Anduix B, Boren J, Martinez S, Moro C, Centelles JJ, Trebukhina R. **The effect of thiamine supplementation on tumour proliferation. A metabolic control analysis study**. *Eur J Biochem* (2001) **268** 4177-82. DOI: 10.1046/j.1432-1327.2001.02329.x
13. Liu Y, Xiong WJ, Wang L, Rang WQ, Yu C. **Vitamin b1 intake and the risk of colorectal cancer: a systematic review of observational studies**. *J Nutr Sci Vitaminol (Tokyo)* (2021) **67** 391-6. DOI: 10.3177/jnsv.67.391
14. Zschabitz S, Cheng TY, Neuhouser ML, Zheng Y, Ray RM, Miller JW. **B vitamin intakes and incidence of colorectal cancer: results from the women's health initiative observational study cohort**. *Am J Clin Nutr* (2013) **97** 332-43. DOI: 10.3945/ajcn.112.034736
15. Lu YT, Gunathilake M, Lee J, Choi IJ, Kim YI, Kim J. **Riboflavin intake, mtrr genetic polymorphism (rs1532268) and gastric cancer risk in a Korean population: a case-control study**. *Br J Nutr* (2022) **127** 1026-33. DOI: 10.1017/S0007114521001811
16. Liu Y, Yu Q, Zhu Z, Zhang J, Chen M, Tang P. **Vitamin and multiple-vitamin supplement intake and incidence of colorectal cancer: a meta-analysis of cohort studies**. *Med Oncol* (2015) **32** 434. DOI: 10.1007/s12032-014-0434-5
17. Chen AC, Martin AJ, Choy B, Fernandez-Penas P, Dalziell RA, McKenzie CA. **A phase 3 randomized trial of nicotinamide for skin-cancer chemoprevention**. *N Engl J Med* (2015) **373** 1618-26. DOI: 10.1056/NEJMoa1506197
18. Lin HL, An QZ, Wang QZ, Liu CX. **Folate intake and pancreatic cancer risk: an overall and dose-response meta-analysis**. *Public Health* (2013) **127** 607-13. DOI: 10.1016/j.puhe.2013.04.008
19. Takata Y, Shu XO, Buchowski MS, Munro HM, Wen WQ, Steinwandel MD. **Food intake of folate, folic acid and other b vitamins with lung cancer risk in a low-income population in the southeastern United States**. *Eur J Nutr* (2020) **59** 671-83. DOI: 10.1007/s00394-019-01934-5
20. Du L, Wang Y, Zhang H, Zhang H, Gao Y. **Folate intake and the risk of endometrial cancer: a meta-analysis**. *Oncotarget* (2016) **7** 85176-84. DOI: 10.18632/oncotarget.13211
21. Collin SM. **Folate and b12 in prostate cancer**. *Adv Clin Chem* (2013) **60** 1-63. DOI: 10.1016/b978-0-12-407681-5.00001-5
22. Gao S, Jin L, Liu G, Wang P, Sun Z, Cao Y. **Overexpression of RASD1 inhibits glioma cell migration/invasion and inactivates the AKT/mTOR signaling pathway**. *Sci Rep* (2017) **7** 3202. DOI: 10.1038/s41598-017-03612-0
23. Sim HW, Nejad R, Zhang W, Nassiri F, Mason W, Aldape KD. **Tissue 2-hydroxyglutarate as a biomarker for isocitrate dehydrogenase mutations in gliomas**. *Clin Cancer Res* (2019) **25** 3366-73. DOI: 10.1158/1078-0432.CCR-18-3205
24. Hu Y, Jiao B, Chen L, Wang M, Han X. **Long non-coding RNA GASL1 may inhibit the proliferation of glioma cells by inactivating the TGF-beta signaling pathway**. *Oncol Lett* (2019) **17** 5754-60. DOI: 10.3892/ol.2019.10273
25. Ulgen E, Can O, Bilguvar K, Oktay Y, Akyerli CB, Danyeli AE. **Whole exome sequencing-based analysis to identify DNA damage repair deficiency as a major contributor to gliomagenesis in adult diffuse gliomas**. *J Neurosurg* (2019) **132** 1435-46. DOI: 10.3171/2019.1.JNS182938
26. Reynold S. **Vitamin transport diseases of brain: focus on folates, thiamine and riboflavin**. *Brain Disorders Therapy* (2014) 3. DOI: 10.4172/2168-975X.1000120
27. Spector R, Johanson CE. **Vitamin transport and homeostasis in mammalian brain: focus on vitamins b and e**. *J Neurochem* (2007) **103** 425-38. DOI: 10.1111/j.1471-4159.2007.04773.x
28. Uchida Y, Ito K, Ohtsuki S, Kubo Y, Suzuki T, Terasaki T. **Major involvement of na(+) -dependent multivitamin transporter (SLC5A6/SMVT) in uptake of biotin and pantothenic acid by human brain capillary endothelial cells**. *J Neurochem* (2015) **134** 97-112. DOI: 10.1111/jnc.13092
29. Zhang W, Jiang J, He Y, Li X, Yin S, Chen F. **Association between vitamins and risk of brain tumors: a systematic review and dose-response meta-analysis of observational studies**. *Front Nutr* (2022) **9** 935706. DOI: 10.3389/fnut.2022.935706
30. Lv W, Zhong X, Xu L, Han W. **Association between dietary vitamin a intake and the risk of glioma: evidence from a meta-analysis**. *Nutrients* (2015) **7** 8897-904. DOI: 10.3390/nu7115438
31. Heydari M, Shayanfar M, Sharifi G, Saneei P, Sadeghi O, Esmaillzadeh A. **The association between dietary total antioxidant capacity and glioma in adults**. *Nutr Cancer* (2021) **73** 1947-56. DOI: 10.1080/01635581.2020.1817954
32. Greenop KR, Miller M, Bailey HD, de Klerk NH, Attia J, Kellie SJ. **Childhood folate, b6, b12, and food group intake and the risk of childhood brain tumors: results from an Australian case-control study**. *Cancer Causes Control* (2015) **26** 871-9. DOI: 10.1007/s10552-015-0562-z
33. Milne E, Greenop KR, Bower C, Miller M, van Bockxmeer FM, Scott RJ. **Maternal use of folic acid and other supplements and risk of childhood brain tumors**. *Cancer Epidemiol Biomark Prev* (2012) **21** 1933-41. DOI: 10.1158/1055-9965.EPI-12-0803
34. Stalberg K, Haglund B, Stromberg B, Kieler H. **Prenatal exposure to medicines and the risk of childhood brain tumor**. *Cancer Epidemiol* (2010) **34** 400-4. DOI: 10.1016/j.canep.2010.04.018
35. Fan Y, Liu A, He Y, Yang X, Xu G, Ma G. **Assessment of nutrient adequacy of adult residents in China**. *Ying Yang Xue Bao* (2012) **34** 15-19
36. Louis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D. **The 2021 who classification of tumors of the central nervous system: a summary**. *Neuro-Oncology* (2021) **23** 1231-51. DOI: 10.1093/neuonc/noab106
37. Zhao WH, Huang ZP, Zhang X, He L, Willett W, Wang JL. **Reproducibility and validity of a Chinese food frequency questionnaire**. *Biomed Environ Sci* (2010) **23** 1-38. DOI: 10.1016/S0895-3988(11)60014-7
38. Yang YX. *China food composition tables* (2018)
39. Craig CL, Marshall AL, Sjostrom M, Bauman AE, Booth ML, Ainsworth BE. **International physical activity questionnaire: 12-country reliability and validity**. *Med Sci Sports Exerc* (2003) **35** 1381-95. DOI: 10.1249/01.MSS.0000078924.61453.FB
40. Zhang W, Du J, Li H, Yang Y, Cai C, Gao Q. **Multiple-element exposure and metabolic syndrome in Chinese adults: a case-control study based on the Beijing population health cohort**. *Environ Int* (2020) **143** 105959. DOI: 10.1016/j.envint.2020.105959
41. Jung YC, Chanraud S, Sullivan EV. **Neuroimaging of wernicke's encephalopathy and korsakoff's syndrome**. *Neuropsychol Rev* (2012) **22** 170-80. DOI: 10.1007/s11065-012-9203-4
42. Cancarini I, Krogh V, Agnoli C, Grioni S, Matullo G, Pala V. **Micronutrients involved in one-carbon metabolism and risk of breast cancer subtypes**. *PLoS One* (2015) **10** e138318. DOI: 10.1371/journal.pone.0138318
43. Hrubsa M, Siatka T, Nejmanova I, Voprsalova M, Kujovska KL, Matousova K. **Biological properties of vitamins of the b-complex, part 1: vitamins b1, b2, b3, and b5**. *Nutrients* (2022) **14** 14. DOI: 10.3390/nu14030484
44. Lonsdale D. **A review of the biochemistry, metabolism and clinical benefits of thiamin(e) and its derivatives**. *Evid Based Complement Alternat Med* (2006) **3** 49-59. DOI: 10.1093/ecam/nek009
45. Klaunig JE. **Oxidative stress and cancer**. *Curr Pharm Des* (2018) **24** 4771-8. DOI: 10.2174/1381612825666190215121712
46. Olivier C, Oliver L, Lalier L, Vallette FM. **Drug resistance in glioblastoma: the two faces of oxidative stress**. *Front Mol Biosci* (2020) **7** 620677. DOI: 10.3389/fmolb.2020.620677
47. Ergul M, Taskiran AS. **Thiamine protects glioblastoma cells against glutamate toxicity by suppressing oxidative/endoplasmic reticulum stress**. *Chem Pharm Bull (Tokyo)* (2021) **69** 832-9. DOI: 10.1248/cpb.c21-00169
48. Liu Y, Yu QY, Zhu ZL, Tang PY, Li K. **Vitamin b2 intake and the risk of colorectal cancer: a meta-analysis of observational studies**. *Asian Pac J Cancer Prev* (2015) **16** 909-13. DOI: 10.7314/apjcp.2015.16.3.909
49. Ma E, Iwasaki M, Kobayashi M, Kasuga Y, Yokoyama S, Onuma H. **Dietary intake of folate, vitamin b2, vitamin b6, vitamin b12, genetic polymorphism of related enzymes, and risk of breast cancer: a case-control study in Japan**. *Nutr Cancer* (2009) **61** 447-56. DOI: 10.1080/01635580802610123
50. Siassi F, Ghadirian P. **Riboflavin deficiency and esophageal cancer: a case control-household study in the Caspian littoral of Iran**. *Cancer Detect Prev* (2005) **29** 464-9. DOI: 10.1016/j.cdp.2005.08.001
51. Aminianfar A, Vahid F, Shayanfar M, Davoodi SH, Mohammad-Shirazi M, Shivappa N. **The association between the dietary inflammatory index and glioma: a case-control study**. *Clin Nutr* (2020) **39** 433-9. DOI: 10.1016/j.clnu.2019.02.013
52. Saedisomeolia A, Ashoori M. **Riboflavin in human health: a review of current evidences**. *Adv Food Nutr Res* (2018) **83** 57-81. DOI: 10.1016/bs.afnr.2017.11.002
53. Thakur K, Tomar SK, Singh AK, Mandal S, Arora S. **Riboflavin and health: a review of recent human research**. *Crit Rev Food Sci Nutr* (2017) **57** 3650-60. DOI: 10.1080/10408398.2016.1145104
54. Powers HJ. **Interaction among folate, riboflavin, genotype, and cancer, with reference to colorectal and cervical cancer**. *J Nutr* (2005) **135** S2960-6. DOI: 10.1093/jn/135.12.2960S
55. Giacalone S, Spigariolo CB, Bortoluzzi P, Nazzaro G. **Oral nicotinamide: the role in skin cancer chemoprevention**. *Dermatol Ther* (2021) **34** e14892. DOI: 10.1111/dth.14892
56. Nikas IP, Paschou SA, Ryu HS. **The role of nicotinamide in cancer chemoprevention and therapy**. *Biomol Ther* (2020) **10** 10. DOI: 10.3390/biom10030477
57. Ma JL, Zhao Y, Guo CY, Hu HT, Zheng L, Zhao EJ. **Dietary vitamin b intake and the risk of esophageal cancer: a meta-analysis**. *Cancer Manag Res* (2018) **10** 5395-410. DOI: 10.2147/CMAR.S168413
58. Petridou E, Kedikoglou S, Koukoulomatis P, Dessypris N, Trichopoulos D. **Diet in relation to endometrial cancer risk: a case-control study in Greece**. *Nutr Cancer* (2002) **44** 16-22. DOI: 10.1207/S15327914NC441_3
59. Yang X, Mei S, Niu H, Li J. **Nicotinic acid impairs assembly of leading edge in glioma cells**. *Oncol Rep* (2017) **38** 829-36. DOI: 10.3892/or.2017.5757
60. Li J, Qu J, Shi Y, Perfetto M, Ping Z, Christian L. **Nicotinic acid inhibits glioma invasion by facilitating snail1 degradation**. *Sci Rep* (2017) **7** 43173. DOI: 10.1038/srep43173
61. Jagerstad M. **Folic acid fortification prevents neural tube defects and may also reduce cancer risks**. *Acta Paediatr* (2012) **101** 1007-12. DOI: 10.1111/j.1651-2227.2012.02781.x
62. Moazzen S, Dolatkhah R, Tabrizi JS, Shaarbafi J, Alizadeh BZ, de Bock GH. **Folic acid intake and folate status and colorectal cancer risk: a systematic review and meta-analysis**. *Clin Nutr* (2018) **37** 1926-34. DOI: 10.1016/j.clnu.2017.10.010
63. Liu N, Jiang J, Song YJ, Zhao SG, Tong ZG, Song HS. **Impact of MTHFR polymorphisms on methylation of MGMT in glioma patients from Northeast China with different folate levels**. *Genet Mol Res* (2013) **12** 5160-71. DOI: 10.4238/2013.October.29.10
64. Chen H, Ward MH, Tucker KL, Graubard BI, McComb RD, Potischman NA. **Diet and risk of adult glioma in eastern Nebraska, United States**. *Cancer Causes Control* (2002) **13** 647-55. DOI: 10.1023/a:1019527225197
65. Duthie SJ. **Folic acid deficiency and cancer: mechanisms of DNA instability**. *Br Med Bull* (1999) **55** 578-92. DOI: 10.1258/0007142991902646
66. Hervouet E, Debien E, Campion L, Charbord J, Menanteau J, Vallette FM. **Folate supplementation limits the aggressiveness of glioma via the remethylation of DNA repeats element and genes governing apoptosis and proliferation**. *Clin Cancer Res* (2009) **15** 3519-29. DOI: 10.1158/1078-0432.CCR-08-2062
67. Liu Q, Zhou L, Lu R, Yang C, Wang S, Hai L. **Biotin and glucose co-modified multi-targeting liposomes for efficient delivery of chemotherapeutics for the treatment of glioma**. *Bioorg Med Chem* (2021) **29** 115852. DOI: 10.1016/j.bmc.2020.115852
68. Paganelli G, Bartolomei M, Ferrari M, Cremonesi M, Broggi G, Maira G. **Pre-targeted locoregional radioimmunotherapy with 90y-biotin in glioma patients: phase I study and preliminary therapeutic results**. *Cancer Biother Radiopharm* (2001) **16** 227-35. DOI: 10.1089/10849780152389410
|
---
title: Subcutaneous transplantation of human embryonic stem cells-derived pituitary
organoids
authors:
- Hiroo Sasaki
- Hidetaka Suga
- Kazuhito Takeuchi
- Yuichi Nagata
- Hideyuki Harada
- Tatsuma Kondo
- Eiji Ito
- Sachi Maeda
- Mayu Sakakibara
- Mika Soen
- Tsutomu Miwata
- Tomoyoshi Asano
- Hajime Ozaki
- Shiori Taga
- Atsushi Kuwahara
- Tokushige Nakano
- Hiroshi Arima
- Ryuta Saito
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC10018142
doi: 10.3389/fendo.2023.1130465
license: CC BY 4.0
---
# Subcutaneous transplantation of human embryonic stem cells-derived pituitary organoids
## Abstract
### Introduction
The pituitary gland, regulating various hormones, is central in the endocrine system. As spontaneous recovery from hypopituitarism is rare, and exogenous-hormone substitution is clumsy, pituitary replacement via regenerative medicine, using pluripotent stem cells, is desirable. We have developed a differentiation method that in mice yields pituitary organoids (POs) derived from human embryonic stem cells (hESC). Efficacy of these POs, transplanted subcutaneously into hypopituitary mice, in reversing hypopituitarism was studied.
### Methods
hESC-derived POs were transplanted into inguinal subcutaneous white adipose tissue (ISWAT) and beneath dorsal skin, a relatively avascular region (AR), of hypophysectomized severe combined immunodeficient (SCID) mice. Pituitary function was evaluated thereafter for ¾ 6mo, assaying basal plasma ACTH and ACTH response to corticotropin-releasing hormone (CRH) stimulation. Histopathologic examination of organoids 150d after transplantation assessed engraftment. Some mice received an inhibitor of vascular endothelial growth factor (VEGF) to permit assessment of how angiogenesis contributed to subcutaneous engraftment.
### Results
During follow-up, both basal and CRH-stimulated plasma ACTH levels were significantly higher in the ISWAT group ($p \leq 0.001$ – 0.05 and 0.001 – 0.005, respectively) than in a sham-operated group. ACTH secretion also was higher in the ISWAT group than in the AR group. Histopathologic study found ACTH-producing human pituitary-cell clusters in both groups of allografts, which had acquired a microvasculature. POs qPCR showed expression of angiogenetic factors. Plasma ACTH levels decreased with VEGF-inhibitor administration.
### Conclusions
Subcutaneous transplantation of hESC-derived POs into hypopituitary SCID mice efficaciously renders recipients ACTH-sufficient.
## Introduction
The pituitary gland, an important endocrine center, regulates homeostasis via various hormones. The anterior pituitary lobe secretes adrenocorticotropic hormone (ACTH), growth hormone, thyroid-stimulating hormone, luteinizing hormone, follicle-stimulating hormone, and prolactin. The posterior pituitary lobe secretes oxytocin and vasopressin. These hormones support a wide variety of physiological functions. Deficiency of pituitary hormones thus can cause severe systemic disease, variably manifest [1]. For example, ACTH deficiency can cause adrenocortical insufficiency, resulting in impaired consciousness, electrolyte imbalance, hypotension, and compromised immunity, which can at worst be fatal [2]. The most common cause of hypopituitarism is pituitary adenoma [3, 4]. Other non-pituitary tumors, such as craniopharyngioma or meningioma, also can cause hypopituitarism. Non-neoplastic causes include Sheehan syndrome, elevated intracranial pressure with empty sella syndrome, traumatic brain injury, aneurysmal subarachnoid hemorrhage, and hypophysitis; some instances are idiopathic [3, 4]. Present treatment for hypopituitarism is, with rare exceptions [5], limited to administration of hormones identified as deficient. Dosage adjustment is difficult. Irregular administration – patients forget! – also poses problems, lessening utility [2]. If pituitary-gland tissue derived from pluripotent stem cells (PSCs) could be deployed clinically, however, these complications of hormone replacement therapy might be eliminated.
We have successfully established efficient differentiation of human embryonic stem cells (hESC) and human induced pluripotent stem cells (hiPSC) into pituitary organoids (POs) in vitro [6, 7], working from a three-dimensional differentiation method using mouse ESC [7]. Transplanting hESC-derived POs under the renal capsule of hypopituitary mice improves activity levels and mortality [6, 8]. Differentiation of hESC-derived POs under feeder-free conditions is under way [9]; clinical application is in the offing. However, several issues remain. Among them is determination of the site and method of transplantation. We hitherto have transplanted PSCs-derived POs into mice using a renal subcapsular site, but the trans-retroperitoneal approach required is substantially invasive. Graft removal from this site, should tumor improbably develop [10, 11], and renal injury with insufficiency also are concerns. Site and method of transplantation thus require refinement. This study accordingly sought to identify an easy, relatively non-invasive, and – in case of tumorigenesis – extirpation-accessible approach for PO graft placement.
## Maintenance and differentiation culture of hESCs
hESCs (KhES-1) were provided by RIKEN BioResource Center and were used in accordance with the hESC research guidelines of the Japanese government. All experimental protocols and procedures were approved by the Ethics Committee of Nagoya University Graduate School of Medicine (approval ES-001). Maintenance and differentiation culture of hESCs was performed as described (6–8, 12) with modifications [9] (Supplementary Figure 1). POs harvested 100-200d after differentiation were used.
## Mice and hypophysectomy
All animal experiments were approved by the Animal Experimentation Committee of the Nagoya University Graduate School of Medicine and were performed in accordance with institutional guidelines for animal care and use. Severe combined immunodeficient (SCID) male mice aged 8-9wk (C.B-17/Icr-Hsd-Prkdcscid, Japan SLC, Shizuoka, Japan) underwent transaural hypophyseal ablation [13]. Mice were anaesthetized with intraperitoneal (i.p.) injection of a mixture of 3 agents (medetomidine 0.75mg/kg, midazolam 4mg/kg, butorphanol 5mg/kg) [14] and pituitary tissue was aspirated from the sella turcica via the auditory meatus using a needle (KN-390, Natsume Seisakusyo, Tokyo, Japan) fitted to a 1ml syringe containing 0.2 ml saline. After the procedure, mice were injected with the medetomidine antagonist atipamezole (0.75mg/kg i.p.). Since hypopituitary SCID mice are sickly, they were bred in a clean environment in the P2A laboratory, a newly constructed animal facility at our university. Their cages were changed twice a week. Complete hypophysectomy was confirmed post-mortem in all mice used (Supplementary Figure 2).
Blood collection and ACTH determination ACTH levels were assessed as a biomarker of pituitary function. Blood samples were collected by tail transection between 1300 and 1700, with sampling before and 1h after administration of human CRH (2μg/kg, i.p.; Tanabe, Osaka, Japan). Plasma was separated from blood samples by centrifugation at 1,000 x g for 15min, 4°C. Plasma ACTH assay used an ACTH ELISA kit (MD Bioproducts, Oakdale, MN) reactive against human and mouse ACTH, with solution absorbance (ACTH concentration) read using Cytation 5 (Biotek, Winooski, VT). CRH loading tests were conducted 1wk after hypophysectomy and 1/~4wk after PO transplantation, including in sham-operated mice, until 6mo later (Figure 1A). Blood samples were collected repeatedly from the same mice. To prevent adrenal crisis, all mice received intramuscular dexamethasone, 0.2mg/0.61ml, after each blood collection. Mice with plasma ACTH levels < 10pg/ml after CRH stimulation were classed as hypopituitary and used as subjects.
**Figure 1:** *Inguinal subcutaneous white adipose tissue (ISWAT) transplantation of hESC-derived pituitary organoids (POs) (A) Schema, mouse handling protocol (hypophyseal ablation, confirmation of hypopituitarism, POs transplantation, and allograft-function testing). (B) Shaved left inguinal area, restrained supine mouse under inhalation anesthesia. (C) Vessel in filmy adipose tissue (arrowhead) overlying femoral vein and artery (arrowhead) viewed through 4mm vertical skin incision. (D) Pocket created in ISWAT. (E) POs emplacement into pocket via syringe fitted with wide-bore tip. (F) Emplaced POs. (G) Nylon suture closure of adipose tissue over POs. Scale intervals, 1mm.*
## Determination of PO spontaneous ACTH secretion in vitro
Five POs were incubated at 37°C for 72h in 2.5ml cell culture medium (Iscove’s modified Dulbecco’s medium, Sigma-Aldrich, St. Louis, MO; Ham’s F12, Thermo Fisher Scientific, Waltham, MA (1:1); $1\%$ GlutaMAX, Thermo Fisher Scientific; $1\%$ Chemically Defined Lipid Concentrate, Thermo Fisher Scientific; 450 µM 1-thioglycerol, Sigma-Aldrich; and $20\%$ KnockOut Serum Replacement, Thermo Fisher Scientific). Culture supernatants were collected. ACTH concentrations in supernatants were determined using an electrochemiluminescence immunoassay (ECLIA) kit (SRL, Tokyo, Japan) employed clinically in Japan.
## Transplantation methods
Mice were anesthetized with isoflurane and placed supine. In inguinal subcutaneous white adipose tissue (ISWAT) transplantation, the left inguinal area was shaved. A 4mm vertical skin incision was made and a pocket in subcutaneous adipose tissue was created, with placement of 5 POs harvested from cell culture medium into the pocket using a wide-bore tip under microscopy. Nylon-suture closure over the POs was followed by skin closure (Figures 1B–G). In the sham-operated group, a vertical skin incision was made in the left inguinal region, a small pocket was created in the subcutaneous fat, and surgical wound closure was performed without PO transplantation. In the avascular region (AR) transplantation group, a vertical skin incision was made in the left dorsal skin, with placement of 5 POs. After these manipulations, mice received intramuscular dexamethasone (0.2mg/0.61ml).
## Evaluation of mice transplanted with POs
Weight was followed and rate of weight loss was evaluated. Activity testing used a running wheel device (ENV-044; Med Associates, Georgia, VT).
## Histological assessment
Transplanted cell aggregates, skin, and fat, from SCID mice were fixed in $10\%$ formalin, dehydrated for paraffin infiltration, and sectioned by sliding microtome. Sections at 5µm were stained with hematoxylin and eosin (H&E) or subjected to immunofluorescence microscopy for various antigens with nuclear 4′,6-diamidino-2-phenylindole counterstaining. Antigen targets included ACTH (mouse, 1:200, 10C-CR1096M1; Fitzgerald), LHX3 (LIM homeobox protein 3, rabbit, 1:3000, AS4002S; RIKEN custom), human nuclei (mouse, 1:1000, MAB4383; Millipore), E-cadherin (rat, 1:50, M108; Takara) and SMA (smooth-muscle actin, mouse, 1:200, M0851; DAKO).
RNA extraction and cDNA synthesis from POs and undifferentiated hESCs RNA was extracted from POs and undifferentiated hESCs using the RNeasy Mini Kit (Qiagen, Hilden, Germany) following manufacturer’s instructions. RNA quality was evaluated using TapeStation 4150 (Agilent Technologies, Santa Clara, CA). cDNA was synthesized using ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo, Osaka, Japan).
## Quantitative PCR
Quantitative PCR (qPCR) of 5 POs and 5 transplanted POs (differentiated from 10,000 ESCs/sample) used a LightCycler 480 system (Roche Diagnostics, Rotkreuz, Switzerland). Data were normalized to those for GAPDH as an endogenous control and determined using standard curve-based relative quantitation. Primers used were: VEGFA, forward 5’-CTGTCAGGGCTGCTTCTTC-3’, reverse 5’-TTGCTGTGCTTTGGGGATTC -3’; VEGFB, forward 5’-TTGACTGTGGAGCTCATGGG-3’, reverse 5’-TGTGTTCTTCCAGGGACATCT-3’; VEFGC, forward 5’-TGTTTTCCTCGGATGCTGGA-3’,reverse 5’-ACATTGGCTGGGGAAGAGTT-3’; FGF2, forward 5’-AGGAGTGTGTGCTAACCGTT-3’, reverse 5’-CAGTTCGTTTCAGTGCCACA-3’; ANGPT2, forward 5’-TGACTGCCACGGTGAATAAT-3’, reverse 5’-CGTGTAGATGCCATTCGTGG-3’; GAPDH, forward 5’-CATCACTGCCACCCAGAAGACTG-3’, reverse 5’-ATGCCAGTGAGCTTCCCGTTCAG -3’. These primer sequences do not cross-react between human and mouse genes. qPCR was performed in POs and contralateral ISWAT. As thus assayed, expression of these genes was clearly lower in mouse ISWAT than in POs (Supplementary Figure 3).
## Statistical analysis
All data were analyzed using IBM SPSS statistics software (version 28.0.0.0, IBM, Armonk, NY). Data are expressed as means ± standard error. Comparisons between groups were performed using Student’s t-test. Comparisons among groups were performed by one-way ANOVA with post hoc Tukey’s test. P values of < 0.05 (*), < 0.01 (**), and < 0.001 (***) were considered significant.
## Assessment of subcutaneous transplantation methods
Four methods were examined: Pre-vascularization of subcutaneous tissue using temporary placement of either a gelatin hydrogel sustained-release device containing basic fibroblast growth factor (FGF) with heparin (GEL) [15] or of a medically approved vascular access catheter (deviceless, DL) [16]; graft siting in ISWAT [17]; and graft siting in a relatively AR beneath dorsal skin. We hypothesized that blood supply would determine graft fate. AR siting served as a control for the 3 other options. Adipose tissue is inherently vessel-rich, while the GEL and DL pre-vascularization methods increase blood supply to the graft site. However, those methods require invasive manipulation 2wk or 1mo, respectively, before transplantation, potentially stressing hypopituitary mice. Plasma ACTH levels 1mo after transplantation of hESC-derived POs were highest in ISWAT-cohort mice (ACTH levels in POs culture medium 30100 pg/ml; all transplanted POs from the same lot). ISWAT transplantation thus appeared best (Figure 1, Supplementary Figure 4).
## Comparisons among ISWAT, sham-operated, and AR groups
ISWAT group ($$n = 6$$), sham-operated group ($$n = 6$$), and AR group ($$n = 5$$) mice, as stated above, served as controls for higher-vascularity ISWAT work, with follow-up of plasma ACTH levels for 6mo. Pre-transplant basal plasma ACTH levels and CRH-stimulated plasma ACTH levels did not significantly differ among the 3 groups (ISWAT vs. sham vs. AR, basal; 4.6 ± 1.8 pg/ml vs. 2.3 ± 1.6 pg/ml vs. 2.5 ± 1.1pg/ml, $$p \leq 0.27$$ – 0.92, stimulated; 1.7 ± 0.8 pg/ml vs. 1.7 ± 2.3 pg/ml vs. 7.2 ± 2.1 pg/ml, $$p \leq 0.06$$ – 0.72).
After transplantation, basal plasma ACTH levels (“basal”, Figure 2A) consistently were higher in the ISWAT group than in the sham-operated group. At 2, 4, 8, and 17wk after transplantation, ACTH values differed significantly between the groups ($p \leq 0.001$ – 0.05, Figure 2A). CRH-stimulated plasma ACTH levels (“stimulated”, Figure 2A) also were higher in the ISWAT group than in the sham-operated group, with statistically significant differences between the groups at 2, 4, 8, 21, and 26wk ($p \leq 0.001$ – 0.005, Figure 2A).
**Figure 2:** *Evaluation of grafted mice All data presented as mean ± SEM. *p<0.05, **p<0.01, ***p<0.001. (A) Basal and CRH-stimulated serum ACTH levels in mice subjected to POs transplantation; sham-operated (Sham), avascular region (AR), and ISWAT cohorts. Sham, n=6. AR, n=5. ISWAT, n=6. Statistical assessment, one-way ANOVA with post hoc Tukey’s test. (B) CRH loading test in intact SCID mice. n=6. (C) Running-wheel activity test. Sham, n=3. ISWAT, n=3. Intact, n=3. Student’s t-test. (D) Percentage change in body weight, ISWAT and Sham mice. Percentage change in body weight, abscissa; time course, ordinate. Student’s t-test.*
Pre-transplant ACTH secretion in vitro, as assessed by ACTH levels in POs culture medium, did not differ significantly between the ISWAT and AR groups (ISWAT vs. AR, 44416 ± 8435 pg/ml vs. 45800 ± 17291 pg/ml, $$p \leq 0.876$$); i.e., ACTH secretory capacity of transplanted POs was similar between groups. Basal plasma ACTH levels in intact mice were 182 ± 40.7 pg/ml and CRH-stimulated plasma ACTH levels were 278.6 ± 43.2 pg/ml ($$n = 6$$, Figure 2B). After transplantation, plasma basal ACTH levels and CRH-stimulated ACTH levels were higher in the AR group than in the sham-operated group and lower than in the ISWAT group, with ISWAT plasma basal ACTH levels significantly higher than AR levels at 2wk ($$p \leq 0.009$$) and ISWAT plasma CRH-stimulated ACTH levels significantly higher than AR levels at 2, 8, and 21wk ($p \leq 0.001$ – 0.013). These results indicated that POs in subcutaneous tissues released ACTH more efficiently when implanted in well-vascularized sites such as adipose tissue than in non-vascularized sites.
Whilst running wheel testing found greater activity in the ISWAT group than in the sham-operated group, the ISWAT group was slightly less active than the intact group (Figure 2C). The rate of weight loss in the ISWAT group was modest by comparison with that in the sham-operated group, but still > $10\%$ (Figure 2D).
## Macroscopic and histological findings after subcutaneous transplantation
On macroscopy 4wk after ISWAT transplantation, graft neovascularization was apparent (Figure 3A). At 21wk, no adipose tissue was observed macroscopically, but the grafts appeared intact (Figure 3B). On microscopy of the skin and subjacent grafts harvested en bloc at 21wk, transplanted cell aggregates lay within subcutaneous tissue (Figures 3C, G). Fluorescence immunomicroscopy revealed that the grafts expressed ACTH; E-cadherin, a marker of oral ectoderm; and LHX3, a pituitary-progenitor marker (Figures 3D, E, H, I). Simultaneous expression of human nuclei indicated that the cells in question were transplanted hESC-derived pituitary cells (Figure 3I). Furthermore, SMA, a vessel-wall marker, was expressed as clusters around and within the grafts (Figures 3F, J), indicating neovascularization. These observations indicate that the hESC-derived POs engraft and function in vivo. To confirm by microscopy that POs were present was not possible in AR-group mice that secreted ACTH poorly, whereas in an AR-group mouse with relatively good ACTH secretion tissue with engrafted POs was found on H&E staining. Fluorescence immunomicroscopy also demonstrated reactivity, although fewer cells marked than in ISWAT material (Supplementary Figure 5).
**Figure 3:** *Macroscopic and histological findings after ISWAT transplantation (A) Macroscopic appearance, graft site, 4wk after transplantation. Dotted line, grafted POs. Arrowhead, vessel associated with graft. Scale intervals, 1mm. (B) Macroscopic appearance, 21wk after transplantation. Arrow, engrafted POs. (C, G) Skin and subjacent tissue including graft. Several POs are included in the section. Hematoxylin/eosin (H, E). Yellow box, (D–J), (D–F, H–J) Immunofluorescence photomicrographs, various antigens targeted. Counterstaining with 4,6-diamidine-2 -phenylindole dihydrochloride (DAPI). (D, H) Adrenocorticotropic hormone (ACTH, red) and LIM-homeobox protein (LHX3, green), a pituitary progenitor marker. (E, I) Human nuclei (hunuclei, green) and E-cadherin (E-cad, white), an oral ectoderm marker. (F, J) Smooth muscle actin (SMA, red), a vessel-wall marker. Scale bars uniformly 100 μm.*
## Promoting vascularization of hESC-derived POs
Acknowledging that hESC-derived POs function after subcutaneous transplantation, the question arose of how vascularization affects engraftment of POs transplanted subcutaneously. Vascular endothelial growth factor (VEGF), FGF2, and angiopoietin 2 (ANGPT2) are related to early-stage angiogenesis (18–20). We evaluated by qPCR whether the POs expressed these genes. Pre-transplant POs, POs harvested 12h after transplantation, and undifferentiated hESCs as controls were analyzed. Expression levels of VEGFA, VEGFB, VEGFC, and ANGPT2 in the POs were significantly higher than those in undifferentiated hESCs. Expression levels of VEGFC and ANGPT2 in particular rose after transplantation (Figures 4A–E). These results suggested that the POs themselves express angiogenic factors, which might promote engraftment into subcutaneous tissue.
**Figure 4:** *Expression and effects of angiogenic factors in hESC-derived POs All data presented as mean ± SEM. *p<0.05, **p<0.01, ***p<0.001. Quantitative PCR results, expression of VEGFA
(A), VEGFB
(B), VEGFC
(C), FGF2
(D), and ANGPT2
(E) in POs without transplantation, transplanted POs (“with transplantation”), respectively PO Tx (-) and PO Tx (+), and undifferentiated ESC. Expression was normalized to that of GAPDH. PO Tx (-), n=3. PO Tx (+), n=3. ESC, n=3. Statistical assessment, one-way ANOVA with post hoc Tukey’s test. (F) Comparison of basal and CRH-stimulated serum ACTH levels with bevacizumab (BEV) or vehicle administration. BEV, n=9. Vehicle, n=9. Student’s t-test.*
Finally, to assess further the importance of angiogenesis in survival of subcutaneous POs grafts, we examined POs function in vivo by administering bevacizumab, a VEGF inhibitor (2mg/kg, i.p., 2x/wk for 2wk [21]; Selleck, Houston, TX). Hypopituitary SCID mice were divided after ISWAT transplantation into a bevacizumab administration group ($$n = 9$$) and a vehicle administration group ($$n = 9$$). Whilst ACTH levels in the culture medium of POs used in grafting did not differ significantly between the groups (bevacizumab vs. vehicle, 19445 ± 6774 pg/ml vs. 20295 ± 6610 pg/ml, $$p \leq 0.93$$), plasma ACTH levels after CRH stimulation were significantly lower in the bevacizumab group than in the vehicle group (4.6 ± 3.0 pg/ml vs. 34.5 ± 11.7 pg/ml, $$p \leq 0.035$$). We infer that angiogenesis is important for engrafted-POs functionality in ISWAT (Figure 4F).
## Discussion
This study indicated that hESC-derived POs could be engrafted into and function in the subcutaneous tissue of hypopituitary SCID mice. The transplanted mice responded to CRH with release of ACTH into the circulation, indicating that injected CRH stimulated the transplanted ACTH-producing cells. Comparisons between ISWAT and AR transplantation indicated that vessel-rich subcutaneous adipose tissue is better than the relatively vessel-poor potential space beneath back skin, yielding more persistent ACTH secretion. PSC-derived POs transplantation improves physical activity levels and body weight [6, 8]; our work confirmed this. Moreover, hESC-derived POs express angiogenic factors such as VEGF and ANGPT2, suggesting that POs autonomously promote vascularization and engraftment. Levels of VEGFC and ANGPT2 expression rose in POs after transplantation; perhaps adiponectin, a cytokine released from adipocytes, contributed to this [22]. Finally, ACTH secretion was reduced by bevacizumab administration, indicating that angiogenesis is important, at least in subcutaneous transplantation. Peri-implant vascularity could be important for engraftment. Perhaps relatively high ACTH secretion observed in some AR-group mice reflected serendipitous proximity to blood vessels, resulting in successful PO engraftment. Of weight here is that the subcutaneous vascular plexus is abundant in the adipose tissue layer [23]. Humoral factors from adipose tissue such as adiponectin may contribute by supporting angiogenesis. These results and considerations support intra-adipose tissue implantation of hESC-derived POs if a subcutaneous site is selected.
Demonstration of function in subcutaneously transplanted PSC-derived POs is an important step in regenerative-medicine technology. Kidney subcapsular transplantation and subcutaneous transplantation differ importantly. POs recipient patients suffer from hypopituitarism and thus are sensitive to stresses such as invasive procedures. Both kidney subcapsular and subcutaneous transplantation require general anesthesia, but in humans, subcutaneous transplantation can be performed under local anesthesia, substantially less invasive than the alternative. Kidney subcapsular transplantation could damage a normal kidney. Risks of collateral damage during subcutaneous transplantation are by contrast low; the surgery itself is easy and can be done quickly, in mouse or in human, perhaps permitting outpatient work. Important in transplantation of cells derived from PSC is graft removal if tumor develops. Subcutaneous transplantation allows relatively simple and non-invasive removal.
That transplantation into adipose tissue is more effective than transplantation into an avascular site is important for clinical application. PSC-derived pancreatic-endoderm cells engrafted successfully in the deep subcutaneous tissues of the abdominal wall of type 1 diabetes human patients [24, 25]. Our success is consonant with theirs, supporting the merits of adipose tissue endografts.
Mouse ACTH secretion varied. The reasons for the variation include: 1) ACTH secretion in reaction to stressors (a little stress during sample collection can affect ACTH values), 2) hemolysis, and 3) large gaps in results caused by small measurement errors due to minute sample volume. Although at no blood collection point did differences achieve statistical significance, the transplant group tended to secrete higher levels of ACTH than the sham group. This has the potential to prevent adrenal crisis, which is an important goal of pituitary regenerative medicine.
To evaluate adrenal function, levels of ACTH and corticosterone, a hormone regulated by ACTH, must be determined simultaneously. In this study, however, we selected blood collection from the tail to permit repeated sampling. That sample volumes were very small precluded measurement of both ACTH and corticosterone at the same time. However, corticosterone values increased on CRH loading after transplantation of POs, although these were placed in a renal subcapsular site and the mice were decapitated [6, 8].
On histologic study, POs after subcutaneous transplantation differ from POs in vitro; they are soft and they collapse, losing their original appearance. More pituitary cells are seen in Figure 3, showing POs transplanted nearly 200d after differentiation, than in Supplementary Figure 1: The more time in vitro, the more differentiated [7].
Subcutaneous transplantation of POs increased ACTH secretion and improved physical activity in hypopituitary mice but without equaling normal mice and with some weight loss. How many POs are required to normalize plasma ACTH levels awaits study. Our methods of inducing differentiation of POs from PSCs generate ACTH-producing cells efficiently, with fewer cells dedicated to production of other adenohypophyseal hormones (data not shown). We speculate that growth hormone deficiency [26] and central hypogonadotropic hypogonadism [27] contributed to decreased activity and to weight loss.
Since we studied SCID mice, we did not investigate immune responses. Our follow-up work will focus on treatment of “wild-type” mice with hypopituitarism, addressing the utility of immunosuppression or of HLA-editing iPSC-derived POs [28]. Solving this issue will bring us closer to clinical application in humans.
## Conclusion
We indicated that hESC-derived POs function following subcutaneous transplantation in mice. An appropriate site for subcutaneous transplantation is adipose tissue, which is richly vascularized. Angiogenesis is important for subcutaneous engraftment of hESC-derived POs.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding author.
## Ethics statement
The animal study was reviewed and approved by Animal Experimentation Committee of the Nagoya University Graduate School of Medicine.
## Author contributions
Authorship: Participation included writing of the article, HSa and HSu. Research design, HSa, HSu, KT, YN, HH, TK, EI, TM, ST, AK, TN, HA, and RS. Performance of the research, HSa, HSu, SM, MSa, MSo, TM, TA, and HO. Data analysis, HSa and ST. All authors contributed to the article and approved the submitted version.
## Conflict of interest
ST and AK are employed by Sumitomo Pharma. TN is employed by Sumitomo Chemical. HSu has received research funding from Sumitomo Pharma and Sumitomo Chemical. Authors are co-inventors on patent applications related to the study presented in this article.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be constructed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1130465/full#supplementary-material
## References
1. Schneider HJ, Aimaretti G, Kreitschmann-Andermahr I, Stalla GK, Ghigo E. **Hypopituitarism**. *Lancet* (2007.0) **369**. DOI: 10.1016/s0140-6736(07)60673-4
2. Hahner S, Spinnler C, Fassnacht M, Burger-Stritt S, Lang K, Milovanovic D. **High incidence of adrenal crisis in educated patients with chronic adrenal insufficiency: a prospective study**. *J Clin Endocrinol Metab* (2015.0) **100**. DOI: 10.1210/jc.2014-3191
3. Prodam F, Caputo M, Mele C, Marzullo P, Aimaretti G. **Insights into non-classic and emerging causes of hypopituitarism**. *Nat Rev Endocrinol* (2021.0) **17**. DOI: 10.1038/s41574-020-00437-2
4. Tanriverdi F, Dokmetas HS, Kebapcı N, Kilicli F, Atmaca H, Yarman S. **Etiology of hypopituitarism in tertiary care institutions in Turkish population: analysis of 773 patients from pituitary study group database**. *Endocrine* (2014.0) **47** 198-205. DOI: 10.1007/s12020-013-0127-4
5. Joshi MN, Whitelaw BC, Carroll PV. **MECHANISMS IN ENDOCRINOLOGY: Hypophysitis: diagnosis and treatment**. *Eur J Endocrinol* (2018.0) **179** R151-r163. DOI: 10.1530/eje-17-0009
6. Ozone C, Suga H, Eiraku M, Kadoshima T, Yonemura S, Takata N. **Functional anterior pituitary generated in self-organizing culture of human embryonic stem cells**. *Nat Commun* (2016.0) **7**. DOI: 10.1038/ncomms10351
7. Kasai T, Suga H, Sakakibara M, Ozone C, Matsumoto R, Kano M. **Hypothalamic contribution to pituitary functions is recapitulated**. *Cell Rep* (2020.0) **30** 18-24.e5. DOI: 10.1016/j.celrep.2019.12.009
8. Suga H, Kadoshima T, Minaguchi M, Ohgushi M, Soen M, Nakano T. **Self-formation of functional adenohypophysis in three-dimensional culture**. *Nature* (2011.0) **480** 57-62. DOI: 10.1038/nature10637
9. Nakano T. **Method for producing cell mass including pituitary tissue, and cell mass thereof. JP Patent application WO2019103129A1**
10. Hong SG, Winkler T, Wu C, Guo V, Pittaluga S, Nicolae A. **Path to the clinic: assessment of iPSC-based cell therapies**. *Cell Rep* (2014.0) **7**. DOI: 10.1016/j.celrep.2014.04.019
11. Nori S, Okada Y, Nishimura S, Sasaki T, Itakura G, Kobayashi Y. **Long-term safety issues of iPSC-based cell therapy in a spinal cord injury model: oncogenic transformation with epithelial-mesenchymal transition**. *Stem Cell Rep* (2015.0) **4**. DOI: 10.1016/j.stemcr.2015.01.006
12. Eiraku M, Watanabe K, Matsuo-Takasaki M, Kawada M, Yonemura S, Matsumura M. **Self-organized formation of polarized cortical tissues from ESCs and its active manipulation by extrinsic signals**. *Cell Stem Cell* (2008.0) **3**. DOI: 10.1016/j.stem.2008.09.002
13. Falconi G, Rossi GL. **Transauricular hypophysectomy in rats and mice**. *Endocrinology* (1964.0) **74**. DOI: 10.1210/endo-74-2-301
14. Kawai S, Takagi Y, Kaneko S, Kurosawa T. **Effect of three types of mixed anesthetic agents alternate to ketamine in mice**. *Exp Anim* (2011.0) **60**. DOI: 10.1538/expanim.60.481
15. Uematsu SS, Inagaki A, Nakamura Y, Imura T, Igarashi Y, Fathi I. **The optimization of the prevascularization procedures for improving subcutaneous islet engraftment**. *Transplantation* (2018.0) **102**. DOI: 10.1097/tp.0000000000001970
16. Pepper AR, Gala-Lopez B, Pawlick R, Merani S, Kin T, Shapiro AM. **A prevascularized subcutaneous device-less site for islet and cellular transplantation**. *Nat Biotechnol* (2015.0) **33**. DOI: 10.1038/nbt.3211
17. Yasunami Y, Nakafusa Y, Nitta N, Nakamura M, Goto M, Ono J. **A novel subcutaneous site of islet transplantation superior to the liver**. *Transplantation* (2018.0) **102**. DOI: 10.1097/tp.0000000000002162
18. Carmeliet P, Jain RK. **Molecular mechanisms and clinical applications of angiogenesis**. *Nature* (2011.0) **473** 298-307. DOI: 10.1038/nature10144
19. Ortega S, Ittmann M, Tsang SH, Ehrlich M, Basilico C. **Neuronal defects and delayed wound healing in mice lacking fibroblast growth factor 2**. *Proc Natl Acad Sci USA* (1998.0) **95**. DOI: 10.1073/pnas.95.10.5672
20. Nag S, Nourhaghighi N, Venugopalan R, Asa SL, Stewart DJ. **Angiopoietins are expressed in the normal rat pituitary gland**. *Endocr Pathol* (2005.0) **16** 67-73. DOI: 10.1385/ep:16:1:067
21. Lin Y, Dong MQ, Liu ZM, Xu M, Huang ZH, Liu HJ. **A strategy of vascular-targeted therapy for liver fibrosis**. *Hepatology* (2022.0) **76**. DOI: 10.1002/hep.32299
22. Sakata N, Yoshimatsu G, Tanaka T, Yamada T, Kawakami R, Kodama S. **Mechanism of transplanted islet engraftment in visceral white adipose tissue**. *Transplantation* (2020.0) **104**. DOI: 10.1097/tp.0000000000003400
23. Yousef H, Alhajj M, Sharma S. *Anatomy, skin (Integument), epidermis* (2022.0)
24. Shapiro AMJ, Thompson D, Donner TW, Bellin MD, Hsueh W, Pettus J. **Insulin expression and c-peptide in type 1 diabetes subjects implanted with stem cell-derived pancreatic endoderm cells in an encapsulation device**. *Cell Rep Med* (2021.0) **2**. DOI: 10.1016/j.xcrm.2021.100466
25. Ramzy A, Thompson DM, Ward-Hartstonge KA, Ivison S, Cook L, Garcia RV. **Implanted pluripotent stem-cell-derived pancreatic endoderm cells secrete glucose-responsive c-peptide in patients with type 1 diabetes**. *Cell Stem Cell* (2021.0) **28**. DOI: 10.1016/j.stem.2021.10.003
26. Salomon F, Cuneo RC, Hesp R, Sönksen PH. **The effects of treatment with recombinant human growth hormone on body composition and metabolism in adults with growth hormone deficiency**. *N Engl J Med* (1989.0) **321**. DOI: 10.1056/nejm198912283212605
27. Katznelson L, Finkelstein JS, Schoenfeld DA, Rosenthal DI, Anderson EJ, Klibanski A. **Increase in bone density and lean body mass during testosterone administration in men with acquired hypogonadism**. *J Clin Endocrinol Metab* (1996.0) **81**. DOI: 10.1210/jcem.81.12.8954042
28. Xu H, Wang B, Ono M, Kagita A, Fujii K, Sasakawa N. **Targeted disruption of HLA genes**. *Cell Stem Cell* (2019.0) **24** 566-578.e7. DOI: 10.1016/j.stem.2019.02.005
|
---
title: 'Obesity, malnutrition, and the prevalence and outcome of hypertension: Evidence
from the National Health and Nutrition Examination Survey'
authors:
- Heng-Zhi Zhang
- Yi-Han Wang
- Ying-Lin Ge
- Shu-Yu Wang
- Jin-Yu Sun
- Lu-Lu Chen
- Shuang Su
- Ying Sun
journal: Frontiers in Cardiovascular Medicine
year: 2023
pmcid: PMC10018144
doi: 10.3389/fcvm.2023.1043491
license: CC BY 4.0
---
# Obesity, malnutrition, and the prevalence and outcome of hypertension: Evidence from the National Health and Nutrition Examination Survey
## Abstract
### Background
Nutritionally unhealthy obesity is a newly introduced phenotype characterized by a combined condition of malnutrition and obesity. This study aims to explore the combined influence of obesity and nutritional status on the prevalence and outcome of hypertension.
### Methods
Participants collected from the National Health and Nutrition Examination Survey (NHANES) database were divided into four subgroups according to their obesity and nutritional conditions, as defined by waist circumference and serum albumin concentration. The lean-well-nourished was set as the reference group. Logistic regression models were applied to evaluate the hypertension risk. Kaplan–*Meier analysis* and Cox proportional hazard regression models were used to assess the survival curve and outcome risk of participants with hypertension.
### Results
A total of 28,554 participants with 10,625 hypertension patients were included in the analysis. The lean-malnourished group showed a lower hypertension risk (odds ratio [OR] 0.85, $95\%$ confidence interval [CI]: 0.77–0.94), while the obese-well-nourished condition elevated the risk (OR 1.47, $95\%$ CI: 1.3–1.67). Two malnourished groups had higher mortality risks (HR 1.42, $95\%$ CI: 1.12–1.80 and HR 1.31, $95\%$ CI: 1.03–1.69 for the lean and obese, respectively) than the reference group. The outcome risk of the obese-well-nourished group (HR 1.02, $95\%$ CI: 0.76–1.36) was similar to the lean-well-nourished.
### Conclusion
Malnutrition was associated with a lower risk of developing hypertension in both lean and obese participants, but it was associated with a worse outcome once the hypertension is present. The lean-malnourished hypertension patients had the highest all-cause mortality risk followed by the obese-malnourished. The obese-well-nourished hypertension patients showed a similar mortality risk to the lean-well-nourished hypertension patients.
## 1. Introduction
Arterial hypertension is a leading risk factor for multiple cardiovascular diseases (CVDs) and renal disability [1]. About 10 million deaths globally can be attributed to hypertension each year [2, 3]. Despite the relatively stable global average blood pressure in these decades, the prevalence of hypertension is continuously increasing in various low- and middle-income regions [1, 4].
Due to the unhealthy diet and behavior patterns, obesity, a condition strongly associated with type 2 diabetes and CVDs, has become a growing worldwide health problem [5]. However, obesity is a phenomenon of high heterogeneity and can occur under a broad spectrum of metabolic situations [6]. Interestingly, malnutrition and obesity can be observed simultaneously in individuals as part of the so-called “double burden of malnutrition” [7, 8]. This hybrid condition of malnutrition and obesity is recognized to be highly implicated with the inflammatory state and the risk of non-communicable diseases [7].
Recently, an intriguing research on heart failure patients revealed that the obese-malnourished participants had significantly higher comorbidity burden and less favorable cardiac outcomes compared with other nutrition and obesity statuses [9]. This result led to our speculation about the potential influence of this phenotype on hypertension. Although obesity has long been identified as an important risk factor for hypertension [10], the relationship between nutrition status and blood pressure remains vague [11, 12]. To our knowledge, no previous research has investigated the combined effect of nutrition and obesity status on hypertension. In this study, we aim to explore the association of nutrition status defined by serum albumin (SA) levels and abdominal obesity with the prevalence and outcomes of clinical hypertension.
## 2.1. Data source and study population
The National Health and Nutrition Examination Survey (NHANES) is a multistage health survey based on interviews and physical/laboratory examinations of the civilian US population. The National Death Index (NDI) is a centralized death record information database collecting the follow-up information from the date of medical examination to either death or censoring (December 31, 2015). This study was based on the publicly available data of 7 consecutive NHANES cycles (2001–2002, 2003–2004, 2005–2006, 2007–2008, 2009–2010, 2011–2012, 2013–2014) and the NDI database. The demographic data, body measurements, blood pressure, CVD, smoking/drinking status, medical conditions, standard biochemistry profiles, income, and education levels of participants were extracted. The race of participants was categorized as non-Hispanic white, non-Hispanic black, Mexican American, other Hispanic, and Other. The exclusion criteria were as follows: [1] participants aged <18 or > 80 years, [2] without body mass index (BMI) or waist circumference records, [3] without blood pressure records, [4] pregnant individuals, [5] diagnosed with cancer, [6] deceased within 3 months. After selecting patients with hypertension, 10,625 participants were finally enrolled. National Center for Health Statistics Research Ethics Review Board approved the analysis, and informed consent was acquired from all individuals.
## 2.2. Study definitions
Abdominal obesity was defined by waist circumference, with a cut-off value of $\frac{102}{88}$ cm for males and females, respectively, as proposed by the National Cholesterol Education Program-Adult Treatment Panel III (NCEP-ATP III) [13]. Nutrition status was defined by the serum albumin concentration as in the previous study [9]. Participants with SA < 45 g/L or ≥ 45 g/L were recognized as malnourished or well-nourished.
Each participant’s blood pressure was measured following the American Heart Association standardized protocol three times after resting 5 minutes in a seated position. Hypertension was defined as previously described [14, 15]: [1] average systolic blood pressure ≥ 130 mmHg or diastolic blood pressure ≥ 80 mmHg, [2] self-reported hypertension, or [3] self-reported administration of anti-hypertensive medications.
The income levels of participants were evaluated by family income-to-poverty ratios (PIRs), which was calculated as the ratio of family income to the federal poverty level. PIR was categorized as <1.33, 1.33–<3.50, and ≥ 3.50, following the qualification criterion for the US federal Supplemental Nutrition Assistance Program [16].
## 2.3. Covariates
Multiple covariates related to hypertension were assessed in this study to minimize bias. To be specific, age, sex, body mass index (BMI), race (non-Hispanic white, non-Hispanic Black, Mexican American, other Hispanic, and other races), diabetes, taking prescriptions for hypertension, smoking, alcohol drinking, and education level of the participants were collected from the NHANES and were adjusted in the statistical analysis.
## 2.4. Statistical analysis
The statistical reporting recommendations by the American Heart Association were followed in the study [17]. Participants were categorized into four groups according to their waist circumference and nutrition status. The group with low waist circumference and fine nutrition status was selected as the reference group (the lean-well-nourished group). Kolmogorov–Smirnov test was used to assess the normality. Continuous variables with normal distribution were provided as mean ± standard deviation, while skewed distributed variables were provided as median with interquartile range. Categorical variables were reported as percentages. ANOVA test, Kruskal-Wallis test, and chi-square test were adapted for comparing the baseline characteristics of continuous variables with normal and skewed distribution, and categorical variables, as appropriate.
Logistic regression models were applied to evaluate the association of obesity and malnutrition with the prevalence of hypertension, and Cox proportional hazard regression models were used to assess the association of obesity and malnutrition with all-cause death in participants with hypertension. Odds ratios (ORs) and hazard ratios (HRs) with $95\%$ confidence intervals (CIs) were calculated in the logistic and Cox regression analyses, respectively. In either logistic or Cox regression analysis, unadjusted analysis was performed at first. Then two different adjusted models were performed to minimize the bias caused by covariates. Model 1 was adjusted for age, sex, race, diabetes, having CVDs, taking prescriptions for hypertension, smoking/drinking status, and education level. Model 2 was adjusted for model 1 covariates plus BMI. An additional age-stratified analysis for the young (below 45 years old) and the elder (above 45 years old) were then conducted with the same method.
Moreover, we further assessed the association between obesity and nutrition status with the prognosis of hypertension by Kaplan–Meier survival analysis. Pairwise comparisons of the survival rates in different groups were performed by the log-rank test. A two-tail value of $p \leq 0.05$ was considered statically significant. All statistical analyses were performed using the R software (version 3.6.1; R Foundation for Statistical Computing, Vienna). An additional age-stratified analysis for the young (below 45 years old) and the elder (above 45 years old) were then conducted with the same method.
## 3.1. Baseline characteristics
The demographics, cardiovascular health status, cardiovascular risk factors, and behavioral factors of the participants were provided in Table 1. A total of 28,554 participants were included in the study, including 10,625 ($37.2\%$) participants diagnosed with hypertension. During a median follow-up of 6.8 years, 2,162 ($7.6\%$) deaths were observed, including 376 ($17.3\%$) CVD deaths.
**Table 1**
| Unnamed: 0 | Lean-well-nourished* (n = 5,133) | Lean-malnourished (n = 8,104) | Obese-well-nourished (n = 2,679) | Obese-malnourished (n = 12,638) | p value |
| --- | --- | --- | --- | --- | --- |
| Age | 35.0 [26.0;48.0] | 45.0 [32.0;60.0]# | 47.0 [35.0;61.0]#,† | 51.0 [39.0;64.0]#,†,& | <0.001 |
| Sex (Male/Female), n (%) | 3,806/1,327 (74.1/25.9) | 4,634/3,470 (57.2/42.8)# | 1,587/1,092 (59.2/40.8)#,† | 4,389/8,249 (34.7/65.3)#,†,& | <0.001 |
| Hypertension (No/Yes), n (%) | 4,113/1,020 (80.1/19.9) | 5,883/2,221 (72.6/27.4)# | 1,464/1,215 (56.4/45.4)#,† | 6,469/6,169 (51.2/48.8)#,†,& | <0.001 |
| Race: | | | | | <0.001 |
| Non-Hispanic White | 2,405 (46.9%) | 3,310 (40.8%) | 1,457 (54.4%) | 5,492 (43.5%) | |
| Non-Hispanic Black | 711 (13.9%) | 1,821 (22.5%) | 338 (12.6%) | 3,117 (24.7%) | |
| Mexican American | 935 (18.2%) | 1,389 (17.1%) | 509 (19.0%) | 2,423 (19.2%) | |
| Other Hispanic | 397 (7.7%) | 678 (8.4%) | 225 (8.4%) | 1,030 (8.2%) | |
| Other races | 685 (13.3%) | 906 (11.2%) | 150 (5.6%) | 576 (4.6%) | |
| PIR level: | | | | | <0.001 |
| <1.33 | 1,449 (28.2%) | 2,309 (28.5%) | 723 (27.0%) | 4,004 (31.7%) | |
| 1.33–3.5 | 1,623 (31.6%) | 2,710 (33.4%) | 895 (33.4%) | 4,335 (34.3%) | |
| ≥3.5 | 2,061 (40.2%) | 3,085 (38.1%) | 1,061 (39.6%) | 4,299 (34.0%) | |
| Education level: | | | | | <0.001 |
| Below high school | 1,148 (22.4%) | 2,210 (27.3%) | 646 (24.1%) | 3,693 (29.2%) | |
| High School | 1,149 (22.4%) | 1,756 (21.7%) | 660 (24.6%) | 3,080 (24.4%) | |
| Above high school | 2,836 (55.3%) | 4,138 (51.1%) | 1,373 (51.3%) | 5,865 (46.4%) | |
| CVD (No/Yes), n (%) | 4,940/193 (96.2/3.8) | 7,514/590 (92.7/7.3) | 2,461/218 (91.9/8.1) | 11,131/1,507 (88.1/11.9) | <0.001 |
| Mortality status (No/Yes), n (%) | 4,938/195 (96.2/3.8) | 7,427/677 (91.6/8.4) | 2,515/164 (93.9/6.1) | 11,512/1,126 (91.1/8.9) | <0.001 |
| CVD death (No/Yes), n (%) | 5,105/28 (99.5/0.5) | 7,984/120 (98.5/1.5) | 2,648/31 (98.8/1.2) | 12,441/197 (98.4/1.6) | <0.001 |
| SBP | 117.3 [109.3;127.3] | 118.0 [108.7;130.0]# | 124.0 [114.0;135.3]#,† | 122.7 [112.7;136.0]#,† | <0.001 |
| DBP | 70.7 [64.0;78.0] | 70.0 [62.7;77.3]# | 74.0 [65.3;81.7]#,† | 72.0 [64.0;80.0]#,†,& | <0.001 |
| WC | 86.1 [79.1;93.6] | 86.0 [80.0;94.3]# | 105.8 [99.4;112.0]#,† | 107.0 [99.0;116.2]#,†,& | <0.001 |
| BMI | 24.0 [21.7;26.3] | 24.2 [22.0;26.5]# | 30.4 [28.1;33.4]#,† | 31.9 [28.8;36.2]#,†,& | <0.001 |
| Total-to-HDL ratio | 3.5 [2.8;4.5] | 3.4 [2.8;4.4]# | 4.3 [3.4;5.3]#,† | 4.0 [3.2;4.9]#,†,& | <0.001 |
| Triglycerides | 101.0 [68.0;158.0] | 98.0 [68.0;150.0]# | 149.0 [99.0;232.0]#,† | 133.0 [90.0;199.0]#,†,& | <0.001 |
| Diabetes (No/Yes), n (%) | 4,818/315 (93.9/6.1) | 7,315/789 (90.3/9.7) | 2,274/405 (84.9/15.1) | 9,653/2,985 (76.4/23.6) | <0.001 |
| HbA1c | 5.3 [5.1;5.5] | 5.4 [5.1;5.6]# | 5.5 [5.2;5.8]#,† | 5.6 [5.3;6.0]#,†,& | <0.001 |
| FPG | 89.0 [83.0;96.0] | 90.0 [83.0;98.0]# | 93.0 [86.0;103.0]#,† | 95.0 [87.0;109.0]#,†,& | <0.001 |
| eGFR | 109.5 [89.8;130.8] | 99.3 [77.7;122.4]# | 125.2 [96.3;159.1]#,† | 123.3 [91.4;161.0]#,† | <0.001 |
| Smoking (No/Yes), n (%) | 2,816/231 7 (54.9/45.1) | 4,245/3,859 (52.4/47.6) | 1,435/1,244 (53.6/46.4) | 6,932/5,706 (54.9/45.1) | 0.003 |
| Drinking (No/Yes), n (%) | 4,694/439 (91.4/8.6) | 7,155/949 (88.3/11.7) | 2,387/292 (89.1/10.9) | 2,387/292 (89.1/10.9) | <0.001 |
Among four groups classified by abdominal obesity and nutrition status, the obese-malnourished group had the most participants (12,638, $44.3\%$), much higher than the obese-well-nourished group [2,679]. The obese-malnourished group was generally the oldest (median 51 years) and had the highest BMI (median 31.9), drinking percentage ($16.2\%$), and crude prevalence of diabetes ($23.6\%$). The obese-malnourished group also presented the highest crude prevalence of hypertension ($48.8\%$). The obese-malnourished group had the largest absolute number [8,249] and proportion ($65.3\%$) of females. Being the opposite of the obese-malnourished, the lean-well-nourished group presented the youngest age (median 35 years), lowest alcohol drinking proportion ($8.6\%$), and lowest crude prevalence of diabetes ($6.1\%$) and hypertension ($19.9\%$).
## 3.2. The association of obesity and malnutrition with the prevalence of hypertension
Table 2 shows the results of the logistic regression models assessing the association of obesity and malnutrition with the prevalence of hypertension. In the multivariable-adjusted model 1, compared with the lean-well-nourished group, both two obese groups showed a significantly higher risk of hypertension, with ORs of 2.29 ($95\%$ CI: 2.04–2.57) and 1.94 ($95\%$ CI, 1.77–2.13) respectively for the well−/malnourished. Interestingly, the old lean-malnourished group had a significantly lower risk for hypertension (OR 0.83, $95\%$ CI: 0.73–0.95) compared with the reference group, but the significance was not observed in the younger participants.
**Table 2**
| Unnamed: 0 | Unadjusted | Unadjusted.1 | Model 1* | Model 1*.1 | Model 2** | Model 2**.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p |
| Total | Total | Total | Total | Total | Total | Total |
| Lean-well-nourished | Reference | Reference | Reference | Reference | Reference | Reference |
| Lean-malnourished | 1.52 (1.4–1.66) | <0.001 | 0.89 (0.81–0.98) | 0.017 | 0.85 (0.77–0.94) | 0.001 |
| Obese-well-nourished | 3.35 (3.02–3.71) | <0.001 | 2.29 (2.04–2.57) | <0.001 | 1.47 (1.3–1.67) | <0.001 |
| Obese-malnourished | 3.85 (3.56–4.15) | <0.001 | 1.94 (1.77–2.13) | <0.001 | 1.07 (0.96–1.19) | 0.252 |
| Age below 45 | Age below 45 | Age below 45 | Age below 45 | Age below 45 | Age below 45 | Age below 45 |
| Lean-well-nourished | Reference | Reference | Reference | Reference | Reference | Reference |
| Lean-malnourished | 1.07 (0.92–1.24) | 0.380 | 0.99 (0.85–1.16) | 0.921 | 0.95 (0.81–1.11) | 0.537 |
| Obese-well-nourished | 3.16 (2.66–3.74) | <0.001 | 2.87 (2.4–3.43) | <0.001 | 1.76 (1.45–2.13) | <0.001 |
| Obese-malnourished | 2.98 (2.63–3.4) | <0.001 | 2.75 (2.37–3.19) | <0.001 | 1.38 (1.15–1.66) | <0.001 |
| Age above 45 | Age above 45 | Age above 45 | Age above 45 | Age above 45 | Age above 45 | Age above 45 |
| Lean-well-nourished | Reference | Reference | Reference | Reference | Reference | Reference |
| Lean-malnourished | 1.03 (0.92–1.16) | 0.568 | 0.83 (0.73–0.95) | 0.005 | 0.82 (0.72–0.93) | 0.002 |
| Obese-well-nourished | 2.09 (1.81–2.41) | <0.001 | 1.89 (1.62–2.21) | <0.001 | 1.29 (1.1–1.53) | 0.002 |
| Obese-malnourished | 2.22 (1.99–2.48) | <0.001 | 1.62 (1.43–1.83) | <0.001 | 0.97 (0.84–1.11) | 0.662 |
In model 2, which adjusted covariates in model 1 plus BMI, a significant elevation of hypertension risk was still observed in the obese-well-nourished group (OR 1.47, $95\%$ CI: 1.3–1.67). The older lean-malnourished group still showed a lower risk (OR 0.82, $95\%$ CI: 0.72–0.93) comparing with participants with normal waist circumference and nutrition status. However, the significant rise in hypertension risk in the obese-malnourished group was only observed in younger participants (OR 1.38, $95\%$ CI: 1.15–1.66) in model 2.
## 3.3. The association of obesity and malnutrition with the outcome of hypertension patients
Among 10,625 participants with hypertension, 1,370 ($12.9\%$) deaths were observed during the follow-up. Figure 1 and Table 3 show the results of the Kaplan–*Meier analysis* for evaluating the association of obesity and malnutrition with the risk of all-cause death in hypertension patients. Compared with the lean-well-nourished group, both malnourished groups showed significantly elevated death risk. Interestingly, the difference between the two malnourished groups is also statistically significant, with the lean-malnourished having a higher mortality risk. However, no significant difference was observed between the lean-well-nourished and the obese-well-nourished groups.
**Figure 1:** *Kaplan–Meier survival curve of the four groups divided by obesity and nutrition status.* TABLE_PLACEHOLDER:Table 3 We then examined the survival of hypertension participants in each group by multi-adjusted Cox models to eliminate the bias caused by cardiovascular covariates (Table 4). In model 1, compared with the lean-well-nourished, only the elevation of mortality risk in the lean-malnourished group reached significance (HR 1.42, $95\%$ CI: 1.12–1.79). After further adjustment of BMI in model 2, the unstratified analysis found a significantly higher risk in the lean-malnourished group (HR 1.42, $95\%$ CI: 1.12–1.80) and obese-malnourished group (HR 1.31, $95\%$ CI: 1.03–1.69). In the age-stratified model 2, the lean-malnourished group had higher outcome risks (HR 2.61, $95\%$ CI 1.10–6.18 and HR 1.36, $95\%$ CI 1.06–1.74 for the young and the elder, respectively), and the risk elevation of the obese-malnourished group also nearly reached statistical significance (HR 1.29, $95\%$ CI 1.00–1.67).
**Table 4**
| Unnamed: 0 | Unadjusted | Unadjusted.1 | Model 1* | Model 1*.1 | Model 2** | Model 2**.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | HR (95% CI) | p | HR (95% CI) | p | HR (95% CI) | p |
| Total | Total | Total | Total | Total | Total | Total |
| Lean-well-nourished | Reference | Reference | Reference | Reference | Reference | Reference |
| Lean-malnourished | 2.06 (1.63, 2.61) | <0.001 | 1.42 (1.12, 1.79) | 0.004 | 1.42 (1.12, 1.80) | 0.004 |
| Obese-well-nourished | 1.16 (0.87, 1.53) | 0.307 | 0.92 (0.69, 1.22) | 0.568 | 1.02 (0.76, 1.36) | 0.908 |
| Obese-malnourished | 1.61 (1.29, 2.01) | <0.001 | 1.15 (0.92, 1.45) | 0.223 | 1.31 (1.03, 1.69) | 0.031 |
| Age below 45 | Age below 45 | Age below 45 | Age below 45 | Age below 45 | Age below 45 | Age below 45 |
| Lean-well-nourished | Reference | Reference | Reference | Reference | Reference | Reference |
| Lean-malnourished | 2.64 (1.13, 6.18) | 0.025 | 2.61 (1.10, 6.18) | 0.029 | 2.61 (1.10, 6.18) | 0.029 |
| Obese-well-nourished | 1.85 (0.70, 4.86) | 0.212 | 1.66 (0.63, 4.37) | 0.309 | 1.61 (0.58, 4.45) | 0.361 |
| Obese-malnourished | 2.16 (0.97, 4.81) | 0.058 | 1.63 (0.70, 3.79) | 0.252 | 1.57 (0.60, 4.06) | 0.356 |
| Age above 45 | Age above 45 | Age above 45 | Age above 45 | Age above 45 | Age above 45 | Age above 45 |
| Lean-well-nourished | Reference | Reference | Reference | Reference | Reference | Reference |
| Lean-malnourished | 1.67 (1.31, 2.13) | <0.001 | 1.35 (1.06, 1.73) | 0.017 | 1.36 (1.06, 1.74) | 0.015 |
| Obese-well-nourished | 0.96 (0.71, 1.28) | 0.77 | 0.88 (0.66, 1.18) | 0.401 | 0.98 (0.72, 1.33) | 0.901 |
| Obese-malnourished | 1.27 (1.00, 1.60) | 0.048 | 1.12 (0.88, 1.42) | 0.344 | 1.29 (1.00, 1.67) | 0.052 |
## 4. Discussion
Nutritionally unhealthy obesity, or the individual-level double burden of malnutrition, is a newly introduced phenotype characterized by a combined condition of malnutrition and obesity, caused mainly by poor food quality and inadequate micronutrient consumption [9, 18]. Traditionally, the double burden of malnutrition is thought to affect primarily low- to middle-income countries [7, 8]. However, our study observed an individual-level double burden of malnutrition (low SA levels plus high waist circumference) in a considerable proportion of US participants, suggesting the long-standing importance of promoting favorable diet and health behaviors not only in developing countries but still in developed ones. We found that the demographic distribution of the obese-malnourished was typically older and often women, which is generally consistent with a recent study on Asian participants [9]. This demographic distribution pattern is similar to that of malnutrition alone, the underlying mechanisms of which may include socioeconomic burden and metabolic features of aging [19].
Our results suggested that the association between obesity and the all-cause mortality risk of hypertension patients is modified by nutrition status. The obese-malnourished hypertension patients have a significantly higher mortality risk than the lean-well-nourished, whereas the risk of the obese-well-nourished is comparable with that of the reference group. This result implies that malnutrition may be an outcome indicator of hypertension which is independent of obesity. Similarly, a newly published study on heart failure patients reported that compared with the obese-well-nourished, the obese-malnourished status has a higher outcome risk, higher likelihood of comorbidities, and notable cardiac remodeling [9]. Malnutrition in obese individuals is often overlooked in clinical practice, but recent evidence, including our study, suggested the significant association between this phenotype and the prognosis of CVDs and other diseases [20, 21].
This study further enhanced the understanding of the value of the SA test in clinical practice. Our results suggested that SA test can be an easy and inexpensive test in clinical practice to assist in evaluating the prognosis of hypertension patients. Beyond our findings, low SA levels have also been found to be an indicator of the emergence and worsening of some CVDs [22], and an independent predictor of ischemic heart disease and ischemic stroke even after adjustments for BMI, liver functions, and kidney functions [23]. The underlying mechanisms of the association of low SA levels and cardiovascular dysfunctions are ambiguous. Speculations are that it is related to the inflammatory condition in malnutritional status with low SA. As human albumin is the most copious antioxidant in the whole blood, the insufficiency of SA will initialize oxidative stress and inflammation, which play a key role in the pathogenesis of hypertension and multiple CVDs [22, 24]. At the same time, low SA levels can be the result of various causes including kidney diseases and general organ dysfunction due to CVDs [25], which may directly affect or reflect the cardiovascular health condition.
Among the malnourished hypertension participants, we found that obese individuals had a better outcome than the lean. This phenomenon is related to the so-called obesity paradox, i.e., obesity is associated with a higher risk for developing CVDs, but a better prognosis if the disease is present [26]. Recent evidence argued that this paradox might stem more from the poorer catabolic reserve and more cachexia of the lean group rather than the potential benefit of the obesity itself, as studies have confirmed that low body fat percentage and low BMI are independent predictors of worse outcomes of CVDs [27, 28]. Other possible causes for the better outcome of obese individuals may include younger age at presentation, lower prevalence of smoking, lower levels of atria natriuretic peptide, and more usage of cardiac medications [26].
Our study suggested that malnutrition may reduce the prevalence of hypertension. After being stratified by obesity status, the two malnourished groups presented a lower risk for developing hypertension than their well-nourished counterparts. This result is partly in line with earlier observations that SA level is positively associated with blood pressure [12, 29]. However, this result does not imply that weakening nutritional condition is a desirable way to prevent hypertension since malnutrition is a harmful state associated with various comorbidities and remarkably higher mortality risk [30, 31]. We speculate that this result is related to the SA’s function of maintaining colloid osmotic pressure, but precise underlying causes are worthy of further research.
Body mass index remains hitherto the most widely used measure to report obesity and related cardiovascular risk. However, since only weight and height are considered, BMI cannot comprehensively describe the heterogeneous composition of body weight and fat distribution within the obese population [32]. Compared with BMI, waist circumference can better indicate visceral adiposity and provide additive information [33, 34], and was therefore recommended as a vital sign in clinical practice according to a recent consensus [35]. Previous studies have reported that when analyzed in the same model, waist circumference was a risk factor for CVDs, but BMI was usually found to be a neutral or even protective factor (36–38). It is a strength that we used waist circumference to define obesity and adjusted for BMI in the analysis to fully investigate the potential value of obesity and nutrition status in evaluating the risk of hypertension.
## 4.1. Limitations
First, on the association between obesity and nutrition status with the prognosis of hypertension, this study only explored the risk of all-cause mortality without confirmation of the causal relationship. There remains the possible existence of multiple confounding factors that should not be ignored. Low SA levels may result from different causes such as chronic diseases and inflammatory conditions, many of which are directly related to worse cardiac outcomes, so the precise causal chain still needs to be further identified. Second, the baseline waist circumference and SA level of participants may change during the follow-up, but we cannot analyze the possible influence of these changes on the outcome. Third, the nutritional status was only evaluated by SA levels in this study. However, SA levels were affected by concurrent health issues apart from nutrition status. Therefore, further validation based on other nutritional indicators should be performed in the following study. Fourth, hypertension defined by both examination and self-report, may affect the robustness of the results. Lastly, due to the limitation of the data source, this study could not distinguish between primary and secondary hypertension.
## 5. Conclusion
This study based on NHANES observed the existence of an individual-level double burden of malnutrition in US citizens. We found that the association between obesity and the prognosis of hypertension is altered by nutrition status. Malnutrition defined by SA was associated with a lower risk of developing hypertension in both lean and obese participants, but it was associated with a worse outcome if the hypertension is present. The lean-malnourished hypertension patients had the highest all-cause death risk followed by the obese-malnourished. The obese- and lean-well-nourished hypertension patients showed a similar mortality risk lower than the mal-nourished.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding authors.
## Author contributions
H-ZZ, J-YS, SS, and YS conceived and designed the study. Y-LG, S-YW, J-YS, and L-LC analyzed the data. H-ZZ, Y-HW, and Y-LG wrote the paper. All authors provided critical revisions of the manuscript and approved the final manuscript.
## Funding
This study was supported in part by the Natural Science Foundation of Jiangsu Province (No. 21KJB320006) and the College Students Innovation and Entrepreneurship Training Program of Jiangsu Province (No. 202110312006Z and No. 202210312010Z).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Zhou B, Perel P, Mensah GA, Ezzati M. **Global epidemiology, health burden and effective interventions for elevated blood pressure and hypertension**. *Nat Rev Cardiol* (2021) **18** 785-802. DOI: 10.1038/s41569-021-00559-8
2. Stanaway JD, Afshin A, Gakidou E, Lim SS, Abate D, Abate KH. **Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990–2017: a systematic analysis for the global burden of disease study 2017**. *Lancet* (2018) **392** 1923-94. DOI: 10.1016/s0140-6736(18)32225-6
3. Olsen MH, Angell SY, Asma S, Boutouyrie P, Burger D, Chirinos JA. **A call to action and a lifecourse strategy to address the global burden of raised blood pressure on current and future generations: the lancet commission on hypertension**. *Lancet* (2016) **388** 2665-712. DOI: 10.1016/S0140-6736(16)31134-5
4. **Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants**. *Lancet* (2021) **398** 957-80. DOI: 10.1016/S0140-6736(21)01330-1
5. **Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128.9 million children, adolescents, and adults**. *Lancet* (2017) **390** 2627-42. DOI: 10.1016/S0140-6736(17)32129-3
6. Samocha-Bonet D, Dixit VD, Kahn CR, Leibel RL, Lin X, Nieuwdorp M. **Metabolically healthy and unhealthy obese--the 2013 stock conference report**. *Obes Rev* (2014) **15** 697-708. DOI: 10.1111/obr.12199
7. Wells JC, Sawaya AL, Wibaek R, Mwangome M, Poullas MS, Yajnik CS. **The double burden of malnutrition: aetiological pathways and consequences for health**. *Lancet* (2020) **395** 75-88. DOI: 10.1016/S0140-6736(19)32472-9
8. Popkin BM, Corvalan C, Grummer-Strawn LM. **Dynamics of the double burden of malnutrition and the changing nutrition reality**. *Lancet* (2020) **395** 65-74. DOI: 10.1016/S0140-6736(19)32497-3
9. Chien SC, Chandramouli C, Lo CI, Lin CF, Sung KT, Huang WH. **Associations of obesity and malnutrition with cardiac remodeling and cardiovascular outcomes in Asian adults: a cohort study**. *PLoS Med* (2021) **18** e1003661. DOI: 10.1371/journal.pmed.1003661
10. Stamler R, Stamler J, Riedlinger WF, Algera G, Roberts RH. **Weight and blood pressure. Findings in hypertension screening of 1 million Americans**. *JAMA* (1978) **240** 1607-10. DOI: 10.1001/jama.1978.03290150053024
11. Oda E. **Decreased serum albumin predicts hypertension in a Japanese health screening population**. *Intern Med* (2014) **53** 655-60. DOI: 10.2169/internalmedicine.53.1894
12. Vargas CM, Obisesan T, Gillum RF. **Association of serum albumin concentration, serum ionized calcium concentration, and blood pressure in the third National Health and nutrition examination survey**. *J Clin Epidemiol* (1998) **51** 739-46. DOI: 10.1016/S0895-4356(98)00047-X
13. **Executive summary of the third report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (adult treatment panel III)**. *JAMA* (2001) **285** 2486-97. DOI: 10.1001/jama.285.19.2486
14. Bakris G, Ali W, Parati G. **ACC/AHA versus ESC/ESH on hypertension guidelines: JACC guideline comparison**. *J Am Coll Cardiol* (2019) **73** 3018-26. DOI: 10.1016/j.jacc.2019.03.507
15. Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and Management of High Blood Pressure in adults: executive summary: a report of the American College of Cardiology/American Heart Association task force on clinical practice guidelines**. *J Am Coll Cardiol* (2018) **71** 2199-269. DOI: 10.1016/j.jacc.2017.11.005
16. Saydah SH, Siegel KR, Imperatore G, Mercado C, Gregg EW. **The Cardiometabolic risk profile of young adults with diabetes in the U.S**. *Diabetes Care* (2019) **42** 1895-902. DOI: 10.2337/dc19-0707
17. Althouse AD, Below JE, Claggett BL, Cox NJ, de Lemos JA, Deo RC. **Recommendations for statistical reporting in cardiovascular medicine: a special report from the American Heart Association**. *Circulation* (2021) **144** e70-91. DOI: 10.1161/CIRCULATIONAHA.121.055393
18. Freeman AM, Aggarwal M. **Malnutrition in the obese: commonly overlooked but with serious consequences**. *J Am Coll Cardiol* (2020) **76** 841-3. DOI: 10.1016/j.jacc.2020.06.059
19. O'Keeffe M, Kelly M, O'Herlihy E, O'Toole PW, Kearney PM, Timmons S. **Potentially modifiable determinants of malnutrition in older adults: a systematic review**. *Clin Nutr* (2019) **38** 2477-98. DOI: 10.1016/j.clnu.2018.12.007
20. Chien SC, Chen CY, Lin CF, Yeh HI. **Critical appraisal of the role of serum albumin in cardiovascular disease**. *Biomark Res* (2017) **5** 31. DOI: 10.1186/s40364-017-0111-x
21. Robinson MK, Mogensen KM, Casey JD, McKane CK, Moromizato T, Rawn JD. **The relationship among obesity, nutritional status, and mortality in the critically ill**. *Crit Care Med* (2015) **43** 87-100. DOI: 10.1097/CCM.0000000000000602
22. Arques S. **Human serum albumin in cardiovascular diseases**. *Eur J Intern Med* (2018) **52** 8-12. DOI: 10.1016/j.ejim.2018.04.014
23. Ronit A, Kirkegaard-Klitbo DM, Dohlmann TL, Lundgren J, Sabin CA, Phillips AN. **Plasma albumin and incident cardiovascular disease: results from the CGPS and an updated meta-analysis**. *Arterioscler Thromb Vasc Biol* (2020) **40** 473-82. DOI: 10.1161/ATVBAHA.119.313681
24. Rodrigo R, Gonzalez J, Paoletto F. **The role of oxidative stress in the pathophysiology of hypertension**. *Hypertens Res* (2011) **34** 431-40. DOI: 10.1038/hr.2010.264
25. Levitt DG, Levitt MD. **Human serum albumin homeostasis: a new look at the roles of synthesis, catabolism, renal and gastrointestinal excretion, and the clinical value of serum albumin measurements**. *Int J Gen Med* (2016) **9** 229-55. DOI: 10.2147/IJGM.S102819
26. Ortega FB, Lavie CJ, Blair SN. **Obesity and cardiovascular disease**. *Circ Res* (2016) **118** 1752-70. DOI: 10.1161/CIRCRESAHA.115.306883
27. Lavie CJ, De Schutter A, Patel D, Artham SM, Milani RV. **Body composition and coronary heart disease mortality--an obesity or a lean paradox?**. *Mayo Clin Proc* (2011) **86** 857-64. DOI: 10.4065/mcp.2011.0092
28. Lavie CJ, McAuley PA, Church TS, Milani RV, Blair SN. **Obesity and cardiovascular diseases: implications regarding fitness, fatness, and severity in the obesity paradox**. *J Am Coll Cardiol* (2014) **63** 1345-54. DOI: 10.1016/j.jacc.2014.01.022
29. Hostmark AT, Tomten SE, Berg JE. **Serum albumin and blood pressure: a population-based, cross-sectional study**. *J Hypertens* (2005) **23** 725-30. DOI: 10.1097/01.hjh.0000163139.44094.1d
30. Correia MI, Waitzberg DL. **The impact of malnutrition on morbidity, mortality, length of hospital stay and costs evaluated through a multivariate model analysis**. *Clin Nutr* (2003) **22** 235-9. DOI: 10.1016/S0261-5614(02)00215-7
31. Dziedzic T, Slowik A, Szczudlik A. **Serum albumin level as a predictor of ischemic stroke outcome**. *Stroke* (2004) **35** e156-8. DOI: 10.1161/01.STR.0000126609.18735.be
32. Neeland IJ, Poirier P, Despres JP. **Cardiovascular and metabolic heterogeneity of obesity: clinical challenges and implications for management**. *Circulation* (2018) **137** 1391-406. DOI: 10.1161/CIRCULATIONAHA.117.029617
33. Pischon T, Boeing H, Hoffmann K, Bergmann M, Schulze MB, Overvad K. **General and abdominal adiposity and risk of death in Europe**. *N Engl J Med* (2008) **359** 2105-20. DOI: 10.1056/NEJMoa0801891
34. Poirier P. **Adiposity and cardiovascular disease: are we using the right definition of obesity?**. *Eur Heart J* (2007) **28** 2047-8. DOI: 10.1093/eurheartj/ehm321
35. Ross R, Neeland IJ, Yamashita S, Shai I, Seidell J, Magni P. **Waist circumference as a vital sign in clinical practice: a consensus statement from the IAS and ICCR working group on visceral obesity**. *Nat Rev Endocrinol* (2020) **16** 177-89. DOI: 10.1038/s41574-019-0310-7
36. Garvey WT, Mechanick JI. **Proposal for a scientifically correct and medically actionable disease classification system (ICD) for obesity**. *Obesity (Silver Spring)* (2020) **28** 484-92. DOI: 10.1002/oby.22727
37. de Hollander EL, Bemelmans WJ, Boshuizen HC, Friedrich N, Wallaschofski H, Guallar-Castillon P. **The association between waist circumference and risk of mortality considering body mass index in 65- to 74-year-olds: a meta-analysis of 29 cohorts involving more than 58 000 elderly persons**. *Int J Epidemiol* (2012) **41** 805-17. DOI: 10.1093/ije/dys008
38. Coutinho T, Goel K, Correa de Sa D, Kragelund C, Kanaya AM, Zeller M. **Central obesity and survival in subjects with coronary artery disease: a systematic review of the literature and collaborative analysis with individual subject data**. *J Am Coll Cardiol* (2011) **57** 1877-86. DOI: 10.1016/j.jacc.2010.11.058
|
---
title: Untargeted metabolomics unravel serum metabolic alterations in smokers with
hypertension
authors:
- Yang Shen
- Pan Wang
- Xinchun Yang
- Mulei Chen
- Ying Dong
- Jing Li
journal: Frontiers in Physiology
year: 2023
pmcid: PMC10018148
doi: 10.3389/fphys.2023.1127294
license: CC BY 4.0
---
# Untargeted metabolomics unravel serum metabolic alterations in smokers with hypertension
## Abstract
Background: Cigarette smoking is an important environmental risk factor for cardiovascular events of hypertension (HTN). Existing studies have provided evidence supporting altered gut microbiota by cigarette smoking, especially in hypertensive patients. Metabolic biomarkers play a central role in the functional potentials of the gut microbiome but are poorly characterized in hypertensive smokers. To explore whether serum metabolomics signatures and compositions of HTN patients were varied in smokers, and investigate their connecting relationship to gut microbiota, the serum metabolites were examined in untreated hypertensive patients using untargeted liquid chromatography-mass spectrometry (LC/MS) analysis.
Results: A dramatic difference and clear separation in community features of circulating metabolomics members were seen in smoking HTN patients compared with the non-smoking controls, according to partial least squares discrimination analysis (PLS-DA) and orthogonal partial least squares discrimination analysis (OPLS-DA). Serum metabolic profiles and compositions of smoking patients with HTN were significantly distinct from the controls, and were characterized by enrichment of 12-HETE, 7-Ketodeoxycholic acid, Serotonin, N-Stearoyl tyrosine and Deoxycholic acid glycine conjugate, and the depletion of Tetradecanedioic acid, Hippuric acid, Glyceric acid, 20-Hydroxyeicosatetraenoic acid, Phenylpyruvic acid and Capric acid. Additionally, the metabolome displayed prominent functional signatures, with a majority proportion of the metabolites identified to be discriminating between groups distributed in Starch and sucrose metabolism, Caffeine metabolism, Pyruvate metabolism, Glycine, serine and threonine metabolism, and Phenylalanine metabolic pathways. Furthermore, the observation of alterations in metabolites associated with intestinal microbial taxonomy indicated that these metabolic members might mediate the effects of gut microbiome on the smoking host. Indeed, the metabolites specific to smoking HTNs were strongly organized into co-abundance networks, interacting with an array of clinical parameters, including uric acid (UA), low-denstiy lipoprotein cholesterol (LDLC) and smoking index.
Conclusions: In conclusion, we demonstrated disparate circulating blood metabolome composition and functional potentials in hypertensive smokers, showing a linkage between specific metabolites in blood and the gut microbiome.
## 1 Introduction
Overwhelming evidences regarding the consequences of smoking have shown that cigarette smoking powerfully enhanced the risks of all-cause mortality, cardiovascular mortality and major adverse cardiovascular events (Chi et al., 2022; Thiravetyan and Vathesatogkit, 2022). Tobacco smoking and even second-hand exposure has been demonstrated to be associated with cardiovascular risk as well as the development of hypertension (HTN) (Dikalov et al., 2019; Bernabe-Ortiz and Carrillo-Larco, 2021). In fact, there has been a growing interest on investigating the role of tobacco consumption on HTN. Hypertensive smokers have been suggested to be more likely to develop malignant or renovascular HTN than non-smokers (Virdis et al., 2010) Cigarette smoking has been shown to raise the daytime and average 24-h blood pressure (BP) and heart rate in treated hypertensive patients (Ohta et al., 2016). Moreover, smokers with HTN were observed to exhibit higher proportion of left ventricular hypertrophy and worse BP control than non-smokers (Journath et al., 2005).
Gut microbiome has emerged as a research hotspot in cardiovascular diseases during the recent years. *Bacterial* genera and species were reported to be altered in smokers and HTN patients, respectively (Aguilar, 2017; Lee et al., 2018; Nakai et al., 2021). Aberrant microbial community and imbalanced composition and function of gut microbes were indicated to be a consequence of cigarette smoking (Shanahan et al., 2018; Bai et al., 2022), but also as a crucial contributor to disrupting metabolic processes and leading to hypertensive disorders (Li et al., 2017; Yan et al., 2020; Avery et al., 2021). Notably, the metabolites transporting into bloodstream acted as an important bridge for the linkage between gut microbiota and host pathology and physiology (Guest et al., 2016). Past studies have attempted potential alterations in gut microbial functionality in HTN subjects related to smoking status. One important aspect was to examine the varied metabolic functions of gut microbiome in hypertensive cigarette smokers as compared with non-smokers (Wang et al., 2021). It was found that the gut microbiota was disordered among smoking HTN patients, with lower microbial α-diversity and significant difference of β-diversity on axes. In addition, dramatic shifts in the intestinal composition at genus and species levels were found among smokers with HTN, including reduced enrichment of Phycisphaera and *Clostridium asparagiforme* (Wang et al., 2021). For another, studies have shown that the smoke-induced gut microbiota dysbiosis impaired gut metabolites directly (Bai et al., 2022). They showed increased bile acid metabolites, especially taurodeoxycholic acid in the colon of mice after smoke exposure (Bai et al., 2022). In addition, the serum metabolome of smoking patients has been identified to differ from that of non-smoking individuals (Xu et al., 2013; Zhang et al., 2022). For instance, Xu et al. indicated that compared with former and non-smokers, in male current smokers, the concentrations of four unsaturated diacyl-phosphatidylcholines (PCs) and five amino acids (arginine, aspartate, glutamate, ornithine and serine) were increased, while three saturated diacyl-PCs, one lysoPC and four acyl-alkyl-PCs, as well as kynurenine showed lower content. Furthermore, higher levels of carnitine and PC aa C32:1, and a lower level of hydroxysphingomyeline [SM (OH)] C22:2 were found in female current smokers (Xu et al., 2013). However, due to the limited number of studies, more investigations are essential to uncover the interrelationship between altered metabolic features and gut microbiota within smokers with HTN.
In an attempt to fully explore the metabolite profiles, and investigate the specific interactive links between metabolites in circulation and the gut microbiome in hypertensive smokers, we employed untargeted liquid chromatography-mass spectrometry (LC/MS) analysis on un-medicated smoking or non-smoking individuals within the clinical context of HTN.
## 2.1 Study participants recruitment
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of Beijing Chaoyang Hospital. All applicable institutional regulations regarding the ethical use of information and samples from human participants were abided, and signed informed consent for the survey have been received from each individual prior to data collection.
The participants were enrolled from our previous study cohort in China (Li et al., 2017) All the individuals wre ethnic Han from employees of the Kailuan Group Corportion of Tangshan city, with simial and stable life-time environment of residence as well as dietary habit. Individuals were excluded if they had suffered from serious cancer, heart or renal failure, stroke, peripheral artery disease or immunodeficiency disorders. All the patients were newly diagnosed hypertensive patients prior to antihypertensive treatment, and none of them have been exposed to antibiotics, probiotics, statin, aspirin, insulin, metformin or nifedipine, etc. Including medicinal herbs during the last 2 months prior to sample collection.
HTN patients with complete information for smoking, including duration and amount for tobacco consumption, as well as smoking coefficient were included in the current research. HTN was diagnosed as systolic BP ≥ 130 mm Hg and/or diastolic BP ≥ 80 mm Hg according to the 2017 ACC/AHA guidelines (Whelton et al., 2017). The measurement of BP was executed by professional nurses or physicians with a random-zero mercury column sphygmomanometer, and subjects were in a sitting position. At every 5 minutes intervals, the BP readings were recorded repeatedly three times, the average of which was regarded as the formal data. Patients having consumed >1 cigarette/day for more than half a year were considered to be smokers as we described previously (Wang et al., 2021), and non-smokers were individuals without a history of tobacco use.
There were 32 non-smokers with HTN (HTN-NS), and 30 cigarette smokers with HTN (HTN-S) in this study. Relevant demographic and clinical profiles such as age, sex, height, weight, body mass index (BMI), fasting blood glucose, total cholesterol, triglyceride, etc. of participants were collected.
## 2.2 Serum sample collection and preparations
Peripheral vein blood was collected from all recruited participants under a fasting state with vacuous tubes, and then separated into serum through centrifugation at 3,000 rpm, 4°C for 10 min. Each aliquot of the obtained serum samples was placed at −80°C and stored until further procedure. For sample preparations before metabolomics determination, serum samples were thawed under room temperature, mixed with $80\%$ methanol and 2.9 mg/mL DL-o-Chlorophenylalanine, vortexed for 30 s and centrifuged at 12,000 rpm and 4°C for 15 min. The supernatant of the solution was proceeded for ultrahigh-performance liquid chromatography with LC/MS detection.
## 2.3 Untargeted metabolome profiling
LC/MS determination was conducted on the platform (Thermo, Ultimate 3000LC, Orbitrap Elite) with Hypergod C18 Column. The conditions for chromatographic separation was at 40°C, and 0.3 mL/min for the flow rate, with water+$0.1\%$ formic acid, and acetonitrile+$0.1\%$ formic acid, respectively. The temperature for automatic injector was at 4°C. For ES + mode, the total ion chromatograms of samples were obtained under 300°C for heater temperature, 45 arb for sheath gas flow rate, 15 arb for aux gas flow rate, one arb for sweep gas flow rate, 3.8 kV spray voltage, 350°C for capillary temperature, and $30\%$ S-Lens RF level. While ES- mode was performed with spray voltage at 3.2 kV and S-Lens RF level at $60\%$. Peaks were extracted from the raw data and analysis was preprocessed with SIEVE software (Thermo). The data of total ion current was normalized and information for features, including retention time, compound molecular weight, and peak intensity were obtained (Dunn et al., 2011).
## 2.4 Multivariate analysis
Multivariate statistical analyses were conducted based on the serum metabolome composition with SIMCA software (V14.1, Sartorius Stedim Data Analytics AB, Umea, Sweden) to discriminate HTN-S patients from HTN-NS individuals. Firstly, principal component analysis (PCA) as an un-supervised analysis, was carried out to produce new characteristic variables from metabolite variables through linear combination according to weights, and further classify distinct group of samples with the obtained variables (Wiklund et al., 2008). Besides, partial least squares discrimination analysis (PLS-DA) as a supervised analysis has been the most frequently used method for classification in metabonomics data analysis. Regression model was combined with dimension reduction in PLS-DA and discriminant analysis was performed based on regression results with discriminant thresholds (Aggio et al., 2010). Orthogonal partial least squares discriminant analysis (OPLS-DA) was performed to exclude the metabolic orthogonal variables which are not related to classification variables, and analyze the non-orthogonal variables and orthogonal variables respectively (Trygg and Wold, 2002). For annotation strategies, the m/z values and mass of compounds were matched to the featured peaks in the METLIN database and the metabolites were identified. METLIN database enhances accurate quantification and facilitates it to more effectively use the data in metabolite databases (Tautenhahn et al., 2012; Zhu et al., 2013; Alseekh et al., 2021).
## 2.5 Metabolite pathway identification
The differentially expressed compounds between groups were annotated to be involved in metabolic pathways according to Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway database (http://www.kegg.jp/kegg/pathway.html) (Kanehisa et al., 2016). By both enrichment analysis and topological analysis of the pathways matched with metabolites, key KEGG pathways mostly correlated with the metabolites were detected.
## 2.6 Gut microbial genera and species identification
The metagenomic sequencing, gene catalog construction, taxonomic annotation and abundance profiling of gut microbes at genus and species levels were performed as described in our previous studies (Li et al., 2017; Wang et al., 2021). The whole metagenome shotgun sequencing data of the specimens evaluated in present study have been previously deposited in a public dataset at the EMBL European Nucleotide Archive underthe BioProject accession code PRJEB13870.
## 2.7 Statistical analysis
Subject characteristics were quantitatively described with mean and SD. For continuous variables, range was shown and count and percent prevalence were summarized for categorical variables. The relative abundances of metabolic members from smoking individuals with HTN were compared to non-smoking controls. Z-score was calculated based on the mean and standard deviation of the data set. Z-score=(x−μ)/σ, where x was a specific score, µ was the mean, and s was the standard deviation (Sreekumar et al., 2009). For univariate analysis, $p \leq 0.05$ was defined to reach statistical significance in two-tailed Student’s t-test. Metabolic compounds with $p \leq 0.05$, and variable importance in the projection (VIP) > 1 for the first principal component of OPLS-DA model, were regarded as statistically different between groups. The VIP and the t-tests are two popular strategies for metabolic biomarker selection. The VIP criterion is to infer biomarkers from the multivariate models and the t-test aims at selecting them in a univariate mode. We used the multivariate VIP (VIP >1) in conjunction with univariate t-test ($p \leq 0.05$) to identify discriminatory metabolite in the current work, as other investigators performed previously (Chen et al., 2022; Han et al., 2022). A significance threshold of one for the VIP was suggested to lead to much improved enrichment of true positives in the selection (Franceschi et al., 2012; Saccenti et al., 2014). A cutoff at $p \leq 0.05$ in univariate t-test has been frequently adopted by researchers (Chen et al., 2022; Han et al., 2022; Wang et al., 2022). Spearman’s correlation analysis was performed to evaluate the interactions among clinical measures, smoking-related serum metabolites, and HTN-S-related gut microbial composition. The cutoff for correlation coefficient was ≥0.3 or ≤ −0.3, and p values were <0.05. The visualization of multiomics correlations was performed using the OmicStudio tools (https://www.omicstudio.cn/tool) with igraph package (version 1.2.6) in R (version 3.6.3). Partial least squares structural equation modeling (PLS-SEM) (Nitzl et al., 2016) was performed with the Smart-PLS three software. The ratio of indirect-to-total effect, variance accounted for (VAF) score, which determines the proportion of the variance explained by the mediation process, was used to examine the significance of mediation effect. Random forest analysis was performed using the random forest package in R to predict the individuals as HTN-NS or HTN-S based on their profiles of genera, species and metabolites. Variable importance by mean decrease in Gini index was calculated for the random forest models. Furthermore, the receiver operation characteristic (ROC) curves for genera, species and metabolites were applied to distinguish the individuals with HTN-S from HTN-NS, statistical significance determined by applying the method of DeLong et al. using MedCalc version 11.4.2.0 (DeLong et al., 1988).
## 3.1 Biochemical characterization of patients
The measurements of the anthropometric and biochemical data of the study participants were presented in Table 1. The average age of hypertensive smokers was 50.9, and non-smokers mainly aged at 53.7 years. As expected, the duration, amount and coefficient of smoking were significantly higher in the HTN-S group as compared with the non-smoking HTN group. Other baseline clinical characteristics between HTN-NS and HTN-S in systolic blood pressure (SBP), diastolic blood pressure (DBP), Fasting blood glucose and blood lipid index, etc. were similar.
**TABLE 1**
| Characteristics | HTN-NS | HTN-S | p-value |
| --- | --- | --- | --- |
| Number | 32 | 30 | — |
| Age, years | 53.7 ± 6.2 | 50.9 ± 4.8 | 0.053 |
| Male/female sex | 29/3 | 30/0 | 0.239 |
| Smoking duration (year) | 0 | 25.0 (10.0–32.0) | <0.001 |
| Smoking amount (cigarette/day) | 0 | 10.0 (7.0–20.0) | <0.001 |
| Smoking coefficient (year cigarette/day) | 0 | 270.0 (100.0–500.0) | <0.001 |
| Systolic BP, mmHg | 140.0 ± 15.7 | 138.3 ± 18.3 | 0.698 |
| Diastolic BP, mmHg | 89.3 ± 9.4 | 89.6 ± 9.4 | 0.908 |
| HR, bmp | 72.4 ± 6.1 | 71.6 ± 9.5 | 0.691 |
| Body mass index, kg/m2 | 25.4 ± 3.4 | 25.6 ± 2.6 | 0.756 |
| Uric acid, μmol/L | 368.00 (275.50–420.00) | 360.50 (303.00–400.00) | 0.688 |
| Creatinine, μmol/L | 71.50 (61.00–91.10) | 70.00 (63.00–76.00) | 0.356 |
| Fasting blood glucose, mmol/L | 5.57 ± 0.60 | 5.64 ± 0.96 | 0.765 |
| Total cholesterol, mmol/L | 5.39 ± 0.96 | 5.35 ± 0.74 | 0.880 |
| Triglyceride, mmol/L | 1.32 (0.95–2.06) | 1.76 (1.17–2.35) | 0.120 |
| HDLC, mmol/L | 1.35 ± 0.27 | 1.27 ± 0.27 | 0.248 |
| LDLC, mmol/L | 2.58 ± 0.79 | 2.47 ± 0.58 | 0.562 |
| TBIL, μmol/L | 13.95 (10.95–19.05) | 13.95 (11.00–18.50) | 0.933 |
| Hemoglobin, g/L | 157.50 (150.00–161.00) | 161.00 (156.00–164.00) | 0.058 |
| Blood platelet, *10^9/L | 222.00 (190.00–251.50) | 220.00 (185.00–255.00) | 0.882 |
| White blood cell, *10^9/L | 6.20 (5.35–6.85) | 6.30 (6.00–7.20) | 0.244 |
## 3.2 Characterization of the serum metabolomic profiling
Serum metabolite assessment was performed with LC/MS, and 3,436 and 4,079 metabolic peaks were detected within ES+ and ES- mode, respectively. Across over 7,000 distinct metabolic features obtained, we identified 561 endogenous small-molecule compounds. The overall discrepancy in the serum metabolome profiles between HTN-S and non-smokers were assessed through multivariate statistical analysis including PCA, PLS-DA and OPLS-DA in Figure 1. In both positive and negative modes, the PCA score plots showed that samples in HTN-S were mixed with those in HTN-NS group (Figures 1A,B). It was noted that, marginal different distributions of samples from HTN-S and HTN-NS were detected in PC1 and PC2 axis under ES + mode. Discriminant analysis with PLS-DA under positive and negative ionic mode revealed significantly separated clustering of hypertensive smokers and non-smoking controls, respectively (Figures 1C,D). And plots obtained in the OPLS-DA models further validated the two distinct clusters of subjects from disparate groups in ES− and ES+ (Figures 1E,F).
**FIGURE 1:** *Smoking or not conduced to dissimilarity of serum metabolic community in hypertensive patients. (A, B), Changes of overall metabolic signatures in hypertensive smokers as compared with non-smokers were identified with PCA score plots based on negative (ES−) and positive (ES+) mode, respectively. Hotelling’s T-squared ellipse indicated 95% confidence interval. The distributions of samples in PC1 and PC2 coordinate axis were further shown with box plots. The first and third quartile (25th and 75th percentile) was expressed with boxes, and median was represented with the inside line. Whiskers extend 1.5 times the inter quartile range from the outer bounds. p values were derived from two-tailed Student’s t-test. (C, D), PLS-DA score plots of both ES- and ES + mode serum metabolomic data from HTN-S patients and HTN-NS controls. Subjects from each group were completely separated. (E, F), Score plot of the OPLS-DA showing disparate metabolic profiling in HTN patients smoking cigarette or not. OPLS-DA method is a supervised multiple regression analysis for identifying discernible patterns between different groups. HTN-S, smokers with HTN; HTN-NS, non-smokers with HTN.*
## 3.3 Metabolite markers for discriminating HTN-S from HTN-NS
To evaluate the detailed differences in metabolome between groups, relative abundance of each metabolite was analyzed. A Supplementary Table S1 reporting both nominal p-value and FDR has been provided (Supplementary Table S1). The detailed information for annotation levels of the annotated features has been described in Supplementary Table S2. As shown in Figures 2A,B, all the metabolic features detected were depicted with fold change of abundance between HTN-S and HTN-NS, VIP of OPLS-DA model in the multivariate statistical analysis, and p values in the univariate analysis. We observed 184 apparently increased and 421 reduced metabolicfeatures based on univariate analysis in HTN-S as compared with non-smoking patients, with 186 and 235 decreased under ES+ and ES- mode, 57 and 127 elevated under ES+ and ES- mode, respectively (Figures 2A,B).
**FIGURE 2:** *Identification of the differential metabolites associated with smoking and non-smoking HTN. (A, B), Volcano plots showing the important metabolites concluded from the OPLS-DA model using a threshold of variable importance for the projection (VIP) > 1. A is based on metabolomics data in ES- mode, and B is in ES- mode. Comparison of the relative abundance of each metabolite in groups was further validated using the p values from two-tailed Student’s t-test. Each dot represented a metabolite, blue denoted downregulated ones, and pink represented those upregulated in smoking HTN. The value of VIP was expressed as the dot size. Among the varied metabolites between groups, those successfully identified were further labeled with names. (C, D), Hierarchical cluster analysis heat-maps of identified metabolites with significant disparate levels between smoking HTN patients and non-smoking controls. The relative abundance of each metabolite in each individual is depicted.*
Among these compounds, features with VIP scores >1 were considered as significant different in HTN-S, and those successfully identified metabolites were labeled in the volcano plots. These serum metabolites with large discrepancy between HTN-S patients and non-smoking controls were further visualized in heat-map (Figures 2C,D). Z-score comparison for these differentially abundant metabolites was performed in each individual (Supplementary Figures S1A, B). Of note, compared with non-smoking subjects, we found that most serum metabolites under ES- were depleted in HTN-S, such as 20-Hydroxyeicosatetraenoic acid, 1-Stearoyl lysophosphatidic acid, Hippuric acid, Glyceric acid and Glutaric acid, and most serum metabolites in ES+ were less abundant in hypertensive patients with tobacco consumption, e.g., N-Acetyl-L-aspartic acid, L-Pipecolic acid, Phenylacetic acid and Capric acid, etc.
Additionally, dramatic co-abundance correlations were revealed among these discriminative metabolites (Figures 3A,B). For instance, Glutaric acid was positively related with Pyridoxine 5′-phosphate, Glyceric acid, Orotic acid and 2-Hydroxyadipic acid. Especially, more profound association between metabolites was observed for those under negative mode. Thus altogether, these findings suggest that HTN patients with smoking behavior exhibit significantly different metabolic profiles compared with those of non-smoking subjects.
**FIGURE 3:** *Association analysis of differential metabolites between HTN-S patients and HTN-NS controls by Spearman correlation. (A), Co-abundance correlation of the distinct metabolites identified in HTN-S as compared with HTN-NS based on ES- mode. (B), Heat map depicting the potential relationship between different metabolic compositions in ES+. Negative correlation is described in orange and positive correlation is labeled in blue. *, p < 0.05; **, p < 0.01; ***, p < 0.001; Spearman’s correlation.*
## 3.4 Pathway enrichment analysis of discrepant metabolites
Concerning the main metabolic pathways and signaling pathways that the differential metabolites participated in, KEGG enrichment analysis was performed, and potential functions wer determined. In order to more comprehensively describe the functional capacities of metabolites and the pathways they participate in, we conducted analysis to explore the crucial pathways these metabolites involved in. The distinct metabolites between groups are labeled and visualized in KEGG pathway map, where enhanced metabolites were in red and depressed ones in blue (Supplementary Figure S2). Each metabolite was assigned to the corresponding KEGG pathway it acts in. The detailed information for pathways those discrepant metabolites matched to, including pathway name, the total number of all metabolites within each pathway, the number and name of differential metabolites matched in each pathway, p values and impact was shown in Supplementary Table S3. Within KEGG database, the distinct metabolites between groups were identified to be involved in multiple pathways regarding Glycine, serine and threonine metabolism, Neomycin, kanamycin and gentamicin biosynthesis, Pyruvate metabolism, Phenylalanine metabolism and Alanine, aspartate and glutamate metabolism, etc. ( Figures 4A,B). There were 10 differential metabolites enriched in Phenylalanine metabolism, and 33 metabolites in Glycine, serine and threonine metabolism, which matched with more differential metabolites than the other pathways. Moreover, the pathways of Starch and sucrose metabolism and Phenylalanine metabolism exhibited higher impact in the analysis under ES- and ES+, respectively. Phenylalanine metabolism and Glycine, serine and threonine metabolism were the most significant metabolic pathways these metabolites functioned on.
**FIGURE 4:** *Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the differentially expressed compounds for group HTN-S vs. HTN-NS. (A, B), Bubble plots in ES− and ES + showing the enriched metabolic pathways of varied metabolic compounds between groups, respectively. The color and Y-axis of dots are based on the -lnP-value, and the enrichment degree is more significant when the color is darker. The size and X-axis of dots represent the impact factor of the pathway in the analysis.*
## 3.5 Associations among serum metabolites and intestinal microbiota and clinical indicators
In order to assess the relationships and explore the potential interaction of altered serum metabolites, gut microbiota profiles and clinical parameters in participants, Spearman’s correlation analysis was conducted subsequently. *Discriminative* genera and species between hypertensive smokers and HTN-NS (Wang et al., 2021) were reanalyzed and evaluated in the present study. The association of top 40 differential with the top 40 distinct intestinal genera and species, respectively were shown in heat-maps (Figure 5).
**FIGURE 5:** *Smoking HTN-associated serum metabolites correlated with gut microbial genera and species differentiating HTN-S vs HTN-NS individuals. (A, C, E, G), The association of gut microbial genera and species with top 40 serum metabolites in the study cohort was described with heatmap. The genera, species and metabolites included were those identified as significantly disparate between HTN-S vs. HTN-NS samples. A is for correlation of genera and metabolites, and C is association of species with metabolites, respectively. Positive associations are in red, and negative associations are in blue. *p < 0.05, **p < 0.01, and ***p < 0.001, Spearman’s rank correlation. (B, D, F, H), Correlation network was produced based on integration of microbiome and metabolome datasets. Differential microbial variances were highly linked with differential metabolites. The correlation coeffcient is ≥ 0.4 or ≤ −0.4, and p < 0.05, calculated from Spearman correlation. Color of the nodes has been updated to represent log2FC, and different data type for metabolite/clinical/microbes is denoted with the node shape. Log10 p-values is described with node size to highlight the major correlations, and spearman Rho was displayed using continuous scaling of coloring of the edges. The datatype was used for attribute circular layout to group each category.*
These metabolites were detected to be strongly related to the abundance of gut microbiota. For instance, Pyridoxine 5′-phosphate and Hippuric acid exhibited a significantly positive correlation with Natronorubrum, Microcoleus, Thermaerobacter, Halorhabdus and Oscillatoria, etc. Serotonin and Palmitoylcarnitine were prominently negatively associated with Ruminiclostridium, Chloroflexus, and Oscillatoria, etc. It was interesting that, Pyridoxine 5′-phosphate, Hippuric acid and T etradecanedioic acid showed a significant correlation with most discriminative genera and species. Co-abundance network in Figures 5B,D,F,H further showed that Pyridoxine 5′-phosphate and Hippuric acid were positively linked to a large cluster of fecal bacteria, such as Thermococcus sp. ES1, Clostridium sp. CAG:557, Bacillus clausii, Caldatribacterium saccharofermentans, Lactobacillus fuchuensi and so on, indicating Pyridoxine 5′-phosphate and Hippuric acid might be potential gut microbiome relevant small-molecule products.
Analogously, as shown in Figure 6, most of clinical features were significantly correlated with altered serum metabolome in HTN-S. Particularly, smoking index were dramatically positively associated with Ubiquinone 6, 8,9−DiHETrE and Serotonin, but negatively related with Glyceric acid, Orotic acid, Pyridoxine 5′-phosphate, Capric acid, Octadecanamide and N-Acetyl-L-aspartic acid (Figures 6A,C). Co-variation between the altered serum metabolites and clinical parameters exhibited significant and complicated association (Figures 6B,D). LDLC was positively linked to Pyridoxine 5'−phosphate, Dihydrouracil and L−Serine, Fasting blood glucose with L−Lactic acid and Glutaric acid, and conversely smoke showed negative association with Pyridoxine 5′-phosphate, 1-Methylxanthine and Octadecanamide, implying a possible contribution of metabolites to the host.
**FIGURE 6:** *Correlation between clinical indexes of subjects and the important serum metabolites altered specifically in smoking HTNs. (A, C), Heat map of the Spearman’s rank correlation coefficient of differential serum metabolic compounds and clinical indexes. Red squares indicate positive associations between metabolites and clinical indexes; blue squares indicate negative associations. The statistical significance was labeled with *p < 0.05, **p < 0.01, respectively. (B, D), Correlation network describing the significant linkage between metabolites and clinical parameters. The correlation coefficient is ≥ 0.3 or ≤ −0.3, and p < 0.05, tested by Spearman correlation. Color of the nodes has been updated to represent log2FC, and different data type for metabolite/clinical/microbes is denoted with the node shape. Log10 p-values is described with node size to highlight the major correlations, and spearman Rho was displayed using continuous scaling of coloring of the edges. The datatype was used for attribute circular layout to group each category. SmokeN: smoking amount (cigarette/day); smokey: smoking duration (year); smokey*N: smoking coefficient (year cigarette/day).*
## 3.6 Random forest classifier identifying HTN-S with metabolic and microbial biomarkers
To further explore the potential biomarker signatures of microbiome and metabolome for distinguishing smoking hypertensive patients, we conducted random forest disease classifier using the relative genera, species and metabolites abundances as variables, respectively. On the basis of the feature importance for the random forest model, as measured with the mean decrease Gini, we obtained ranked lists of metabolic and microbial features crucial for HTN-S. The top 30 most discriminatory microbial biomarkers were primarily from genus Oscillatoria, Pseudobutyrivibrio, Anaeroarcus, Kyrpidia, Parvimonas, and species Pseudomonas stutzeri, Actinomyces sp. ph3, Nocardioides insulae, Lachnospiraceae bacterium.28.4, Clostridium. sp. etc. ( Figures 7A,B). And serum metabolites such as Serotonin, Ubiquinone 6, Octadecanamide, N-Acetyl-L-aspartic acid, Pyridoxine-5-phosphate and L-Pipecolic acid contributed the most to discriminate HTN-S from HTN-NS (Figure 7C). We implemented receiver operating characteristic (ROC) curves to evaluate the discriminative values of variables including top30 gut genera, species and serum metabolites. It showed an area under the curve (AUC) of species = 0.70 ($$p \leq 0.006$$), genera = 0.79 ($p \leq 0.001$), metabolites = 0.65 ($$p \leq 0.046$$), species + metabolites = 0.71 ($$p \leq 0.004$$), and genera + metabolites revealed an AUC of 0.82 ($p \leq 0.001$) (Figure 7D). Comparing to the other variables, the variable of genera + metabolites was more effective to classify HTN-S samples from HTN-NS. For metabolite and microbial biomarkers identified in the present study, we have performed a validation in the form of out-of-bag error for the random forest predictive model. Out-of-bag error rate of the random forest model with variables of the top 30 most distinct genera, species and metabolites to distinguish smoking HTN patients from non-smoking individuals is 0.258, 0.194 and 0.113, respectively (Supplementary Figures S3A–C).
**FIGURE 7:** *Random forest classification based on gut microbiome and metabolites to identify HTN-S patients from HTN-NS controls. (A–C), Random forest models were conducted respectively to evaluate the potential of fecal genera/species and serum metabolites to discriminate between HTN-S and HTN-NS. The top 30 most distinguishing genera/species and metabolites between HTN-S and HTN-NS in the random forest analysis. Mean decrease Gini shown on the x-axis indicates the importance of each variable (genera, species or metabolites) for the classification. (D), Receiver operating characteristic (ROC) curves showed the sensitivity, specificity and area under the curve (AUC) with explanatory variables of the top 30 most distinguishing genera, species or metabolites to distinguish smoking HTN patients from non-smoking individuals. (E), The mediation effects of top 30 most distinguishing gut genera/species (indirect effect) on the total effect of smoking on top 30 metabolites. Path coefficients (beta) are labeled beside each path and indirect effect (VAF score) are denoted below each mediator variables. TE, total effect; ADE, average direct effect; ACME, average causal mediation effect.*
Based on the crosstalk among smoking, gut microbiome and metabolites in HTN, we employed PLS-SEM to test the mediation effects of gut genera and species (indirect effect) on the total effect of smoking on metabolites (Figure 7E). The VAF score was used to estimate the proportion of indirect effect to total effects, and a VAF score at $20\%$–$80\%$ suggests a partial indirect effect. The mediation model evaluating the strength of the indirect effects, identified that the direct relationship between smoking and metabolites was statistically significant, and for indirect effect, the effect of species was statistically significant. The VAF for genera and species between smoking and metabolites was $20.3\%$ while that of the species was $40.1\%$ (Figure 7E). Thus, the contribution of smoking to serum metabolome was partially mediated by influencing the gut microbiome composition.
## 4 Discussion
In the current study, we identified profound association between cigarette smoking status and serum metabolomic profiles among hypertensive patients. Both PLS-DA and OPLS-DA models derived from untargeted metabolomics analysis showed significant discriminations in metabolic profiles and characteristics between smoking HTN patients and the non-smoking controls. Interestingly, it was noted that the majority of discrepant serum metabolites obtained, such as L-serine, Pipecolic acid, L-Lactic acid, were significantly deficient in cigarette smokers subjected to HTN. Furthermore, a classifier based on intestinal microbiota at genus level and core metabolites was established to accurately distinguish smokers from non-smokers among hypertensive patients.
It has been widely recognized that smoking and HTN are both crucial risk factors for cardiovascular diseases (Hopkins and Williams, 1986). Previous studies demonstrated that the combination of HTN and current smoking would extert an additive effect on the risk of developing cardiovascular and cerebrovascular diseases (Huangfu et al., 2017; Hara et al., 2019). Specifically, Huangfu et al. have shown that the cumulative incidence rate of ischemic stroke was $0.85\%$, $2.05\%$, $3.19\%$ and $8.14\%$ among non-HTN/non-smokers, non-HTN/smokers, HTN/non-smokers, and HTN/smokers, respectively. In addition, participants with coexistence of cigarette smoking and HTN were suggested to be at the highest risk for ischemic stroke disorders (Huangfu et al., 2017). Investigators reported that HTN and current smoking had a synergistic effect on the risk of progressing from moderate to severe cerebral small vessel diseases with OR: 10.59, $95\%$ CI: 3.97–28.3, and synergy index: 4.03, $95\%$ CI: 1.17–28.3) (Hara et al., 2019).
The gut microbiome is known to exert a vital impact on host physiology, and emerging evidence have successively confirmed that gut microbiota dysregulation is closely implicated in the occurrence and development of cardiovascular diseases and metabolic disorders (Tilg and Kaser, 2011; Tang et al., 2017). Previously, researchers have described the intestinal microbial dysibosis in smoking hypertensive patients, which is mainly manifested with decreased α-diversity and inclined to Provotella-dominant type (Wang et al., 2021). Simultaneously, in cigarette smokers suffering HTN, dramatic alterations of intestinal genus and species composition were detected, such as reduced enrichment of physphaera and Clostridium asparagiformme, etc. The fecal microbial metabolites are believed to induce host responses in the intestine and even at the distant organs (Liu et al., 2022). The metabolite profiles facilitates us to explain a high proportion of the varied functions for gut microbiome and thus have been regarded as intermediates that mediate the host-microbiota crosstalk (Zierer et al., 2018; Visconti et al., 2019). To the best of our knowledge, here we for the first time examined the serum metabolome of smokers with HTN, and further explored the correlation with gut microbiome.
L-pipecolic acid is known as a cyclic amino acid derived from L-lysine (Pérez-García et al., 2019). Previous studies indicated that the excessive accumulation of pipecolic acid is associated with host disorders of Zellweger syndrome, chronic liver diseases and pyridoxine-dependent epilepsy, etc. ( Yu et al., 2020). Moreover, Yuan X and colleagues showed that, when compared to ulcerative colitis patients without depression/anxiety, those with ulcerative colitis and depression/anxiety subjects exhibited much lower abundance of L-pipecolic acid. It was most attractive that, prophylactic administration of L-pipecolic acid was identified to significantly reduce depressive-like behaviors in mice with colitis and prominently alleviate inflammatory cytokine levels in their colon, blood and brain (Yuan et al., 2021). Similarly, in the present work, we also observed apparent decrease in the abundance of L-pipecolic acid among smoking HTN patients. Further we revealed that L-pipecolic acid was positively interacted with butyrate-producing bacteria such as Clostridim spp. ( Clostridium sp. CAG:470; Clostridium sp. CAG:557), Ruminococcus spp. ( Ruminococcus sp CAG:382) and Bacillus clausii. Butyrate-producing bacteria are considered as a group of beneficial bacteria that produce butytric acids through fermenting dietary fiber (Nylund et al., 2015). For example, the Gram-positive gut bacteria Ruminococcus, has been demonstrated to be quite enriched under health status but markedly depleted in numerous diseases including human motor neuron disease (Rowin et al., 2017; Saad et al., 2021). Besides, *Bacillus clausii* was verified to possess immunomodulatory activity, and play important role in regulating cell growth and differentiation, cell adhesion, signal transcription and transduction, vitamin production and protection of the intestine from genotoxic agents. Therefore, in recent years, preparations containing *Bacillus clausii* have been frequently applied in the treatment or prevention of gut barrier impairment (Lopetuso et al., 2016).
L-serine is generally regarded as non-essential amino acid, but the term “conditionally non-essential amino acid” might be more appropriate for it, since under some circumstances, vertebrates are unable to produce it with sufficient quantities to meet the necessary cellular requirements (Metcalf et al., 2018). It was reported that compared with the mice that compromised from *Klebsiella pneumonia* lung infection, enrichment of L-serine was detected in mice that survived during the infection and L-serine was indicated to be associated with the host surviving. Furthermore, L-serine was able to facilitate macrophage phagocytosis, and participate in a natural way to promote host clearance of lung pathogens (Liu et al., 2018). In addition, it was shown that the level of L-serine was also significantly reduced in murine lungs infected with Pasteurella multocida. Exogenous supplementation of L-serine would significantly enhance the survival rate of infected mice and suppress the colonization of *Pasteurella multocida* in mice lungs (He et al., 2019). Our findings that the abundance of serum metabolite L-serine, was depleted in smokers with HTN, is in agreement with the changes previously observed in other lung diseases (Liu et al., 2018; He et al., 2019).
Actually, several pervious studies have reported the serum metabolomic profiles between essential hypertension and healthy controls (Dołegowska et al., 2009; Shi et al., 2022; Sun et al., 2022). Metabolites such as 2-methylbutyrylcarnitine (Shi et al., 2022) and 12-HETE (Dołegowska et al., 2009) have been demonstrated to be dramatically enriched, while L-Serine (Shi et al., 2022; Sun et al., 2022) was depleted in hypertensive patients in comparison with normotensive controls. It is worth noting that our findings in the current work confirm that the abundance of these serum metabolites discrepant between normal individuals and hypertensive patients, including enhanced 2-methylbutyrylcarnitine and Tetranor 12-HETE, and suppressed L-Serine were further more severely altered in hypertensive smokers.
Cotinine is the utmost important nicotine metabolite (Rolle-Kampczyk et al., 2016). Benowitz and his colleagues indicated that cotinine has been proved to be a suitable marker to differentiate smoke burdened from unburdened persons (Benowitz, 1996; Benowitz et al., 2009). Furthermore, Rolle-Kampczyk UE et al. found that urine cotinine levels were significantly higher among mothers who smoked during pregnancy in comparison with non-smokers (Rolle-Kampczyk et al., 2016). The metabolite cotinine was not detected in the present study, which might be due to the different study population, sample collection site and metabolome detection methods. In addition, Gu F et al. reported that both cotinine and serotonin were positively correlated with current smoking status in a cohort from the Environment And Genetics in Lung Cancer Etiology study (Gu et al., 2016). Although metabolite cotinine was not detected to be discrepant in the present study, we identified higher level of serotonin in smoking HTN patients, which confirmed previous study to a certain extent (Gu et al., 2016).
Previous studies have elucidated the potential capacity of gut microbiome combined with metabolites as biomarkers for distinguishing various diseases, such as thyroid carcinoma, spontaneous preterm birth, Alzheimer’s disease, etc (Feng et al., 2019; Flaviani et al., 2021; Xi et al., 2021). Previously, a discriminate predictive model based on eight metabolites as well as five genera displayed excellent distinguishing effect between thyroid carcinoma patients and healthy controls (Feng et al., 2019). In the present study, we constructed a random forest classifier with the combination of 30 gut bacterial genera and 30 metabolites, which was able to discriminate HTN-S from HTN-NS with an AUC of 0.82. These results illustrated that biomarker signatures according to the gut microbiome and metabolome exert strong reliability in identifying smokers with HTN from non-smokers, which emphasized the significance of gut microbiota and metabolome. On the basis of the previous results, we also conducted mediation analysis, showing that the contribution of smoking status to the serum metabolome was partially mediated by affecting the composition of the intestinal microbiome.
Nevertheless, several limitations have to be acknowledged in the present study. Firstly, the number of participants was relatively small which might restrict the generalization of our results. Since population used to identify the biomarkers did not show any clinical, anthropometrical, or biochemical difference, further studies are still needed to validate these biomarkers in other hypertensive population with cardiovascular diseases, and a direct causal relationship among cigarette smoking, fecal microbiota, serum metabolites during HTN development is warranted to be elucidated in a cell/animal model thoroughly. Secondly, untargeted LC-MS was conducted when analyzing metabolite compositions in smoking and non-smoking HTN patients. Target mass spectrometry, which is more sensitive and more quantitative would also be necessary to further confirm the present findings. Lastly, we admit that the information regarding mode of birth and alcoholic drinking is lacking for the participants.
## 5 Conclusion
In summary, the findings based on this study demonstrated significant discrepancy in circulating blood metabolome profiles in HTN-S when compared with HTN-NS. A combination of gut bacterial genera and serum metabolites was capable to discriminate HTN-S from HTN-NS with good performance. In addition, the contribution of smoking to host metabolome was identified to be mediated at least partially by affecting the gut microbiome. Taken together, it is proposed that smoking cessation is extremely essential for hypertensive patients, which might be helpful to improve metabolic homeostasis and avoid future cardiovascular events by regulating gut microbiome and metabolites.
## Data availability statement
The original contribution presented in the study are included in the article/Supplementary Material, the data presented in the study are deposited in the metabolights repository, accession number MTBLS 7057.
## Ethics statement
The studies involving human participants were reviewed and approved by the ethics committee of Beijing Chaoyang Hospital, Capital Medical University (China). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
YS, PW, YD, and JL conceived and designed the study. YS, PW, XY, MC, YD, and JL conducted the experiment. YS and PW analyzed the data. XY and MC advised and helped analyses. YS and PW wrote the manuscript, YD and JL supervised all processes for this manuscript. All authors have read and agreed to the published version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys.2023.1127294/full#supplementary-material
## References
1. Aggio R. B., Ruggiero K., Villas-Boas S. G.. **Pathway activity profiling (PAPi): From the metabolite profile to the metabolic pathway activity**. *Bioinformatics* (2010) **26** 2969-2976. DOI: 10.1093/bioinformatics/btq567
2. Aguilar A.. **Hypertension: Microbiota under pressure**. *Nat. Rev. Nephrol.* (2017) **13** 3. DOI: 10.1038/nrneph.2016.173
3. Alseekh S., Aharoni A., Brotman Y., Contrepois K., D'Auria J., Ewald J.. **Mass spectrometry-based metabolomics: A guide for annotation, quantification and best reporting practices**. *Nat. Methods* (2021) **18** 747-756. DOI: 10.1038/s41592-021-01197-1
4. Avery E. G., Bartolomaeus H., Maifeld A., Marko L., Wiig H., Wilck N.. **The gut microbiome in hypertension: Recent advances and future perspectives**. *Circ. Res.* (2021) **128** 934-950. DOI: 10.1161/CIRCRESAHA.121.318065
5. Bai X., Wei H., Liu W., Coker O. O., Gou H., Liu C.. **Cigarette smoke promotes colorectal cancer through modulation of gut microbiota and related metabolites**. *Gut* (2022) **71** 2439-2450. DOI: 10.1136/gutjnl-2021-325021
6. Benowitz N. L.. **Cotinine as a biomarker of environmental tobacco smoke exposure**. *Epidemiol. Rev.* (1996) **18** 188-204. DOI: 10.1093/oxfordjournals.epirev.a017925
7. Benowitz N. L., Dains K. M., Dempsey D., Herrera B., Yu L., Jacob P.. **Urine nicotine metabolite concentrations in relation to plasma cotinine during low-level nicotine exposure**. *Nicotine Tob. Res.* (2009) **11** 954-960. DOI: 10.1093/ntr/ntp092
8. Bernabe-Ortiz A., Carrillo-Larco R. M.. **Second-hand smoking, hypertension and cardiovascular risk: Findings from Peru**. *BMC Cardiovasc Disord.* (2021) **21** 576. DOI: 10.1186/s12872-021-02410-x
9. Chen J., Zheng Q., Zheng Z., Li Y., Liao H., Zhao H.. **Analysis of the differences in the chemical composition of monascus rice and highland barley monascus**. *Food Funct.* (2022) **13** 7000-7019. DOI: 10.1039/d2fo00402j
10. Chi Y., Wang X., Jia J., Huang T.. **Smoking status and type 2 diabetes, and cardiovascular disease: A comprehensive analysis of shared genetic Etiology and causal relationship**. *Front. Endocrinol. (Lausanne).* (2022) **13** 809445. DOI: 10.3389/fendo.2022.809445
11. DeLong E. R., DeLong D. M., Clarke-Pearson D. L.. **Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach**. *Biometrics* (1988) **44** 837-845. DOI: 10.2307/2531595
12. Dikalov S., Itani H., Richmond B., Vergeade A., Rahman S. M. J., Boutaud O.. **Tobacco smoking induces cardiovascular mitochondrial oxidative stress, promotes endothelial dysfunction, and enhances hypertension**. *Am. J. Physiol. Heart Circ. Physiol.* (2019) **316** H639-H646. DOI: 10.1152/ajpheart.00595.2018
13. Dołegowska B., Błogowski W., Kedzierska K., Safranow K., Jakubowska K., Olszewska M.. **Platelets arachidonic acid metabolism in patients with essential hypertension**. *Platelets* (2009) **20** 242-249. DOI: 10.1080/09537100902849836
14. Dunn W. B., Broadhurst D., Begley P., Zelena E., Francis-McIntyre S., Anderson N.. **Procedures for large-scale metabolic profiling of serum and plasma using gas chromatography and liquid chromatography coupled to mass spectrometry**. *Nat. Protoc.* (2011) **6** 1060-1083. DOI: 10.1038/nprot.2011.335
15. Feng J., Zhao F., Sun J., Lin B., Zhao L., Liu Y.. **Alterations in the gut microbiota and metabolite profiles of thyroid carcinoma patients**. *Int. J. Cancer* (2019) **144** 2728-2745. DOI: 10.1002/ijc.32007
16. Flaviani F., Hezelgrave N. L., Kanno T., Prosdocimi E. M., Chin-Smith E., Ridout A. E.. **Cervicovaginal microbiota and metabolome predict preterm birth risk in an ethnically diverse cohort**. *JCI Insight* (2021) **6** e149257. DOI: 10.1172/jci.insight.149257
17. Franceschi P., Masuero D., Vrhovsek U., Mattivi F., Wehrens R.. **A benchmark spike‐in data set for biomarker identification in metabolomics**. *J. Chemom.* (2012) **26** 16-24. DOI: 10.1002/cem.1420
18. Gu F., Derkach A., Freedman N. D., Landi M. T., Albanes D., Weinstein S. J.. **Cigarette smoking behaviour and blood metabolomics**. *Int. J. Epidemiol.* (2016) **45** 1421-1432. DOI: 10.1093/ije/dyv330
19. Guest P. C., Guest F. L., Martins-de Souza D.. **Making sense of blood-based proteomics and metabolomics in psychiatric research**. *Int. J. Neuropsychopharmacol.* (2016) **19** pyv138. DOI: 10.1093/ijnp/pyv138
20. Han B., Meng F., Niu Y., Liu H., Li J., Peng X.. **Effect of delphinidin on metabolomic profile in breast carcinogenesis**. *Anticancer Agents Med. Chem.* (2022) **22**. DOI: 10.2174/1871520622666220616101659
21. Hara M., Yakushiji Y., Suzuyama K., Nishihara M., Eriguchi M., Noguchi T.. **Synergistic effect of hypertension and smoking on the total small vessel disease score in healthy individuals: The kashima scan study**. *Hypertens. Res.* (2019) **42** 1738-1744. DOI: 10.1038/s41440-019-0282-y
22. He F., Yin Z., Wu C., Xia Y., Wu M., Li P.. **l-Serine lowers the inflammatory responses during Pasteurella multocida infection**. *Infect. Immun.* (2019) **87** 006777-e719. DOI: 10.1128/IAI.00677-19
23. Hopkins P. N., Williams R. R.. **Identification and relative weight of cardiovascular risk factors**. *Cardiol. Clin.* (1986) **4** 3-31. DOI: 10.1016/s0733-8651(18)30632-5
24. Huangfu X., Zhu Z., Zhong C., Bu X., Zhou Y., Tian Y.. **Smoking, hypertension, and their combined effect on ischemic stroke incidence: A prospective study among inner Mongolians in China**. *J. Stroke Cerebrovasc. Dis.* (2017) **26** 2749-2754. DOI: 10.1016/j.jstrokecerebrovasdis.2017.06.048
25. Journath G., Nilsson P. M., Petersson U., Paradis B. A., Theobald H., Erhardt L.. **Hypertensive smokers have a worse cardiovascular risk profile than non-smokers in spite of treatment--a national study in Sweden**. *Blood Press* (2005) **14** 144-150. DOI: 10.1080/08037050510034220
26. Kanehisa M., Sato Y., Kawashima M., Furumichi M., Tanabe M.. **KEGG as a reference resource for gene and protein annotation**. *Nucleic Acids Res.* (2016) **44** D457-D462. DOI: 10.1093/nar/gkv1070
27. Lee S. H., Yun Y., Kim S. J., Lee E. J., Chang Y., Ryu S.. **Association between cigarette smoking status and composition of gut microbiota: Population-based cross-sectional study**. *J. Clin. Med.* (2018) **7** 282. DOI: 10.3390/jcm7090282
28. Li J., Zhao F., Wang Y., Chen J., Tao J., Tian G.. **Gut microbiota dysbiosis contributes to the development of hypertension**. *Microbiome* (2017) **5** 14. DOI: 10.1186/s40168-016-0222-x
29. Liu Q., Li B., Li Y., Wei Y., Huang B., Liang J.. **Altered faecal microbiome and metabolome in IgG4-related sclerosing cholangitis and primary sclerosing cholangitis**. *Gut* (2022) **71** 899-909. DOI: 10.1136/gutjnl-2020-323565
30. Liu S., Zhang P., Liu Y., Gao X., Hua J., Li W.. **Metabolic regulation protects mice against**. *Exp. Lung Res.* (2018) **44** 302-311. DOI: 10.1080/01902148.2018.1538396
31. Lopetuso L. R., Scaldaferri F., Franceschi F., Gasbarrini A.. **Bacillus clausii and gut homeostasis: State of the art and future perspectives**. *Expert Rev. Gastroenterol. Hepatol.* (2016) **10** 943-948. DOI: 10.1080/17474124.2016.1200465
32. Metcalf J. S., Dunlop R. A., Powell J. T., Banack S. A., Cox P.. **L-serine: A naturally-occurring amino acid with therapeutic potential**. *Neurotox. Res.* (2018) **33** 213-221. DOI: 10.1007/s12640-017-9814-x
33. Nakai M., Ribeiro R. V., Stevens B. R., Gill P., Muralitharan R. R., Yiallourou S.. **Essential hypertension is associated with changes in gut microbial metabolic pathways: A multisite analysis of ambulatory blood pressure**. *Hypertension* (2021) **78** 804-815. DOI: 10.1161/HYPERTENSIONAHA.121.17288
34. Nitzl C., Roldán J., Carrión G. C.. **Mediation analysis in partial least squares path modeling: Helping researchers discuss more sophisticated models**. *Industrial Manag. Data Syst.* (2016) **116** 1849-1864. DOI: 10.1108/IMDS-07-2015-0302
35. Nylund L., Nermes M., Isolauri E., Salminen S., de Vos W. M., Satokari R.. **Severity of atopic disease inversely correlates with intestinal microbiota diversity and butyrate-producing bacteria**. *Allergy* (2015) **70** 241-244. DOI: 10.1111/all.12549
36. Ohta Y., Kawano Y., Hayashi S., Iwashima Y., Yoshihara F., Nakamura S.. **Effects of cigarette smoking on ambulatory blood pressure, heart rate, and heart rate variability in treated hypertensive patients**. *Clin. Exp. Hypertens.* (2016) **38** 510-513. DOI: 10.3109/10641963.2016.1148161
37. Pérez-García F., Brito L. F., Wendisch V. F.. **Function of L-pipecolic acid as compatible solute in corynebacterium glutamicum as basis for its production under hyperosmolar conditions**. *Front. Microbiol.* (2019) **10** 340. DOI: 10.3389/fmicb.2019.00340
38. Rolle-Kampczyk U. E., Krumsiek J., Otto W., Roder S. W., Kohajda T., Borte M.. **Metabolomics reveals effects of maternal smoking on endogenous metabolites from lipid metabolism in cord blood of newborns**. *Metabolomics* (2016) **12** 76. DOI: 10.1007/s11306-016-0983-z
39. Rowin J., Xia Y., Jung B., Sun J.. **Gut inflammation and dysbiosis in human motor neuron disease**. *Physiol. Rep.* (2017) **5** e13443. DOI: 10.14814/phy2.13443
40. Saad A. H., Ahmed M. S., Aboubakr M., Ghoneim H. A., Abdel-Daim M. M., Albadrani G. M.. **Impact of dietary or drinking water Ruminococcus sp. supplementation and/or heat stress on growth, histopathology, and bursal gene expression of broilers**. *Front. Vet. Sci.* (2021) **8** 663577. DOI: 10.3389/fvets.2021.663577
41. Saccenti E., Hoefsloot H., Smilde A. K., Westerhuis J. A., Hendriks M.. **Reflections on univariate and multivariate analysis of metabolomics data**. *Metabolomics* (2014) **10** 361-374. DOI: 10.1007/s11306-013-0598-6
42. Shanahan E. R., Shah A., Koloski N., Walker M. M., Talley N. J., Morrison M.. **Influence of cigarette smoking on the human duodenal mucosa-associated microbiota**. *Microbiome* (2018) **6** 150. DOI: 10.1186/s40168-018-0531-3
43. Shi M., He J., Li C., Lu X., He W. J., Cao J.. **Metabolomics study of blood pressure salt-sensitivity and hypertension**. *Nutr. Metab. Cardiovasc Dis.* (2022) **32** 1681-1692. DOI: 10.1016/j.numecd.2022.04.002
44. Sreekumar A., Poisson L. M., Rajendiran T. M., Khan A. P., Cao Q., Yu J.. **Metabolomic profiles delineate potential role for sarcosine in prostate cancer progression**. *Nature* (2009) **457** 910-914. DOI: 10.1038/nature07762
45. Sun J., Ding W., Liu X., Zhao M., Xi B.. **Serum metabolites of hypertension among Chinese adolescents aged 12-17 years**. *J. Hum. Hypertens.* (2022) **36** 925-932. DOI: 10.1038/s41371-021-00602-8
46. Tang W. H., Kitai T., Hazen S. L.. **Gut microbiota in cardiovascular health and disease**. *Circ. Res.* (2017) **120** 1183-1196. DOI: 10.1161/CIRCRESAHA.117.309715
47. Tautenhahn R., Cho K., Uritboonthai W., Zhu Z., Patti G. J., Siuzdak G.. **An accelerated workflow for untargeted metabolomics using the METLIN database**. *Nat. Biotechnol.* (2012) **30** 826-828. DOI: 10.1038/nbt.2348
48. Thiravetyan B., Vathesatogkit P.. **Long-term effects of cigarette smoking on all-cause mortality and cardiovascular outcomes in Thai population: Results from a 30-year cohort study**. *Asia Pac J. Public Health* (2022) **34** 761-769. DOI: 10.1177/10105395221106860
49. Tilg H., Kaser A.. **Gut microbiome, obesity, and metabolic dysfunction**. *J. Clin. Invest.* (2011) **121** 2126-2132. DOI: 10.1172/JCI58109
50. Trygg J., Wold S.. **Orthogonal projections to latent structures (O-PLS)**. *J. Chemom.* (2002) **16** 119-128. DOI: 10.1002/cem.695
51. Virdis A., Giannarelli C., Neves M. F., Taddei S., Ghiadoni L.. **Cigarette smoking and hypertension**. *Curr. Pharm. Des.* (2010) **16** 2518-2525. DOI: 10.2174/138161210792062920
52. Visconti A., Le Roy C. I., Rosa F., Rossi N., Martin T. C., Mohney R. P.. **Interplay between the human gut microbiome and host metabolism**. *Nat. Commun.* (2019) **10** 4505. DOI: 10.1038/s41467-019-12476-z
53. Wang J., Wang Y., Zeng Y., Huang D.. **Feature selection approaches identify potential plasma metabolites in postmenopausal osteoporosis patients**. *Metabolomics* (2022) **18** 86. DOI: 10.1007/s11306-022-01937-0
54. Wang P., Dong Y., Jiao J., Zuo K., Han C., Zhao L.. **Cigarette smoking status alters dysbiotic gut microbes in hypertensive patients**. *J. Clin. Hypertens. (Greenwich).* (2021) **23** 1431-1446. DOI: 10.1111/jch.14298
55. Whelton P. K., Carey R. M., Aronow W. S., Casey D. E., Collins K. J., Dennison Himmelfarb C.. **ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: A report of the American College of cardiology/American heart association task force on clinical practice guidelines**. *J. Am. Coll. Cardiol.* (2017) **71** e127-e248. DOI: 10.1016/j.jacc.2017.11.006
56. Wiklund S., Johansson E., Sjostrom L., Mellerowicz E. J., Edlund U., Shockcor J. P.. **Visualization of GC/TOF-MS-based metabolomics data for identification of biochemically interesting compounds using OPLS class models**. *Anal. Chem.* (2008) **80** 115-122. DOI: 10.1021/ac0713510
57. Xi J., Ding D., Zhu H., Wang R., Su F., Wu W.. **Disturbed microbial ecology in alzheimer's disease: Evidence from the gut microbiota and fecal metabolome**. *BMC Microbiol.* (2021) **21** 226. DOI: 10.1186/s12866-021-02286-z
58. Xu T., Holzapfel C., Dong X., Bader E., Yu Z., Prehn C.. **Effects of smoking and smoking cessation on human serum metabolite profile: Results from the KORA cohort study**. *BMC Med.* (2013) **11** 60. DOI: 10.1186/1741-7015-11-60
59. Yan X., Jin J., Su X., Yin X., Gao J., Wang X.. **Intestinal flora modulates blood pressure by regulating the synthesis of intestinal-derived corticosterone in high salt-induced hypertension**. *Circ. Res.* (2020) **126** 839-853. DOI: 10.1161/CIRCRESAHA.119.316394
60. Yu K., Liu H., Kachroo P.. **Pipecolic acid quantification using gas chromatography-coupled mass spectrometry**. *Bio Protoc.* (2020) **10** e3841. DOI: 10.21769/BioProtoc.3841
61. Yuan X., Chen B., Duan Z., Xia Z., Ding Y., Chen T.. **Depression and anxiety in patients with active ulcerative colitis: Crosstalk of gut microbiota, metabolomics and proteomics**. *Gut Microbes* (2021) **13** 1987779. DOI: 10.1080/19490976.2021.1987779
62. Zhang R., Sun X., Huang Z., Pan Y., Westbrook A., Li S.. **Examination of serum metabolome altered by cigarette smoking identifies novel metabolites mediating smoking-BMI association**. *Obes. (Silver Spring)* (2022) **30** 943-952. DOI: 10.1002/oby.23386
63. Zhu Z. J., Schultz A. W., Wang J., Johnson C. H., Yannone S. M., Patti G. J.. **Liquid chromatography quadrupole time-of-flight mass spectrometry characterization of metabolites guided by the METLIN database**. *Nat. Protoc.* (2013) **8** 451-460. DOI: 10.1038/nprot.2013.004
64. Zierer J., Jackson M. A., Kastenmüller G., Mangino M., Long T., Telenti A.. **The fecal metabolome as a functional readout of the gut microbiome**. *Nat. Genet.* (2018) **50** 790-795. DOI: 10.1038/s41588-018-0135-7
|
---
title: Identification and validation of ferroptosis-related hub genes in obstructive
sleep apnea syndrome
authors:
- Peijun Liu
- Dong Zhao
- Zhou Pan
- Weihua Tang
- Hao Chen
- Ke Hu
journal: Frontiers in Neurology
year: 2023
pmcid: PMC10018165
doi: 10.3389/fneur.2023.1130378
license: CC BY 4.0
---
# Identification and validation of ferroptosis-related hub genes in obstructive sleep apnea syndrome
## Abstract
### Background
By 2020, the prevalence of Obstructive Sleep Apnea Syndrome (OSAS) in the US has reached 26. 6–$43.2\%$ in men and 8.7–$27.8\%$ in women. OSAS promotes hypertension, diabetes, and tumor growth through unknown means. Chronic intermittent hypoxia (CIH), sleep fragmentation, and increased pleural pressure are central mechanisms of OSAS complications. CIH exacerbates ferroptosis, which is closely related to malignancies. The mechanism of ferroptosis in OSAS disease progression remains unknown.
### Methods
OSAS-related datasets (GSE135917 and GSE38792) were obtained from the GEO. Differentially expressed genes (DEGs) were screened using the R software and intersected with the ferroptosis database (FerrDb V2) to get ferroptosis-related DEGs (f-DEGs). GO, DO, KEGG, and GSEA enrichment were performed, a PPI network was constructed and hub genes were screened. The TCGA database was used to obtain the thyroid cancer (THCA) gene expression profile, and hub genes were analyzed for differential and survival analysis. The mechanism was investigated using GSEA and immune infiltration. The hub genes were validated with RT-qPCR, IHC, and other datasets. Sprague-Dawley rats were randomly separated into normoxia and CIH groups. ROS, MDA, and GSH methods were used to detect CIH-induced ferroptosis and oxidative stress.
### Results
GSEA revealed a statistically significant difference in ferroptosis in OSAS (FDR < 0.05). HIF1A, ATM, HSPA5, MAPK8, MAPK14, TLR4, and CREB1 were identified as hub genes among 3,144 DEGs and 74 f-DEGs. HIF1A and ATM were the only two validated genes. F-DEGs were mainly enriched in THCA. HIF1A overexpression in THCA promotes its development. HIF1A is associated with CD8 T cells and macrophages, which may affect the immunological milieu. The result found CIH increased ROS and MDA while lowering GSH indicating that it could cause ferroptosis. In OSAS patients, non-invasive ventilation did not affect HIF1A and ATM expression. Carvedilol, hydralazine, and caffeine may be important in the treatment of OSAS since they suppress HIF1A and ATM.
### Conclusions
Our findings revealed that the genes HIF1A and ATM are highly expressed in OSAS, and can serve as biomarkers and targets for OSAS.
## Introduction
Obstructive sleep apnea syndrome (OSAS) is one of the most frequent chronic respiratory disorders, affecting almost a billion people globally and having devastating effects on both individuals and society [1]. OSAS can increase the prevalence of neurological disorders such as malignant tumors, coronary heart disease, pulmonary heart disease, diabetes, vestibular abnormalities, and depression (2–4). OSAS is characterized by recurrent full or incomplete pharyngeal collapse during sleep, leading to chronic intermittent hypoxia (CIH) and sleep fragment, with a recent trend toward younger onset, particularly in infants with congenital developmental abnormalities [5, 6]. OSAS-induced intermittent hypoxia and sleep fragment increase cancer or its aggressiveness, as well as the occurrence of antitumor therapy resistance [7]. Furthermore, OSAS-related repeated upper airway obstruction can cause physical discomfort due to prolonged CIH [8, 9]. OSAS is an individually variable condition with diverse symptoms and endotypes, the focus of this study is the pathophysiology and mechanisms of sleep breathing problems [10].
Ferroptosis is a new kind of iron-dependent cell death described by Dixon in 2012 that is morphologically distinct from apoptosis and autophagy [11]. The morphological features of ferroptosis are mainly damaged cells with intact cell membranes, increased mitochondrial membrane density, reduced or absent mitochondrial cristae, mitochondrial membrane shrinkage, and outer membrane rupture. The chromatin of cells is not condensed, and their nuclei are of normal size [12]. *The* genes associated with ferroptosis can be categorized based on six modules: Drivers, Suppressors, Markers, Inducers, Inhibitors, and Diseases [13].
Ferroptosis is essential for the occurrence and development of pathological processes and diseases that include cerebral hemorrhage, ischemic stroke, sepsis, cancer, and myocardial infarction [14]. Ferroptosis has demonstrated tremendous promise as a cancer therapy, and OSAS not only relates to metabolic and cardiovascular illnesses but also to the progression of cancer [15]. OSAS is a prevalent form of respiratory illness [16]. There are few findings on the association between ferroptosis and OSAS, but there are numerous correlation studies between ferroptosis and other respiratory illnesses. Several recent studies have shown that ferroptosis is a potential therapeutic target for lung diseases including acute lung injury, chronic obstructive pulmonary disease (COPD), pulmonary fibrosis (PF), lung infection, and asthma [17]. Numerous animal and cellular models of acute lung injury (ALI) have established the role of ferroptosis in the course of the disease [18].
Through the promotion of autophagy, reactive oxygen species (ROS) are able to upregulate both ferroptosis and intracellular ferritin [19]. CIH can cause an increase in ROS, which can lead to cellular ferroptosis as a possible consequence [20]. ROS/HIF1A leads to increased oxidative stress and increases systemic inflammation, whereas inflammation can also enhance HIF1A expression and aggravate the oxidative stress reaction; these two phenomena are closely connected [21]. HIF1A is both a regulatory protein and a transcription factor in the molecular physiology of oxygen homeostasis [19]. HIF1A regulates multiple glycolysis, proliferation, invasion, and survival genes in response to hypoxia [22, 23]. There is a significant correlation between HIF1A and lung cancer, and the expression of HIF1A in non-small cell lung cancer is mediated by the AKT and ERK signaling pathways [24].
Because the lungs are in a hyperoxic physiological state when compared to other human organs, the pathophysiology of obstructive sleep apnea is mostly manifested as CIH, which is unique from other lung diseases that are in a permanent hypoxic state [25]. Oxidative stress has the potential to produce a significant number of ROS, the accumulation of which is one of the primary processes that contribute to the promotion of ferroptosis. This may be very useful for understanding the function of ferroptosis in OSAS [26]. As a consequence of this, we have developed the research hypothesis that ferroptosis plays an important role in OSAS. This prompted us to perform bioinformatics and CIH animal experiments to investigate the relationship between OSAS, tumors, and ferroptosis.
## Acquisition of datasets and RNA degradation
The RNAseq data (GSE135917 and GSE38792) related to OSAS were downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) [27, 28]. Both of their sequencing platforms are GPL6244, total RNA was isolated from human subcutaneous fat. The first group of GSE135917 consisted of 10 OSAS patients and 8 normal people, whereas the second group consisted of 24 OSAS patients who were sampled individually after treatment with a continuous positive airway pressure (CPAP) ventilator. In brief, CPAP was started following the biopsy and after 2 weeks of self-reported CPAP use for more than 4 h per night. The follow-up biopsy was conducted on the other side of the abdomen.
The ReadAffy function from the affy package (version 1.72.0) was utilized to read the raw data included in the cell format files. The data were imported into SIMCA 14.1 (Sartorius, Malmö, Sweden) for analysis, and four components were chosen to construct the PLS-DA model [29, 30].
## F-DEGs, GSEA, and GSVA
The raw data are normalized using the Affymetrix platform in R software (version 4.0.5). The RMA functions were then used to process data using the affy package. The Limma package (version 3.50.3) was used to perform DEGs between OSAS and normal groups with the empirical bayesian t-test, and |log2-Fold change (FC)| >0.5 and $P \leq 0.05$ were utilized as DEG screening criteria. The heatmap package was used to create a heat map that displayed the top 50 genes with the most significant genes. The ferroptosis-related genes were downloaded from the FerrDb V2 database (http://www.zhounan.org/ferrdb/), which mainly contained gene sets such as “Driver,” “Suppressor,” and “Marker.” A Venn diagram was utilized to illustrate the intersection genes of FerrDb V2 and DEGs (f-DEGs) [31].
*The* genes in the dataset were ranked by OSAS phenotype and logFC value, and Gene Set Enrichment Analysis (GSEA) was performed on 664 gene sets from WikiPathways. The statistical differences were determined by the normalized enrichment score (|NES| >1), and FDR < 0.25. The dataset was read and subjected to Gene set variation analysis(GSVA) analysis using the GSVA package (version 1.42.0) to acquire a GSVA score for each sample. The Limma package was used to compare the GSVA scores between the OSAS and the normal group with the Bayesian t-test, adj. $P \leq 0.05$ and |logFC| > 0.1 represent significant differences.
## Screening F-DEGs biomarkers
The f-DEGs were transformed with the org.Hs.eg.db package (version 3.10.0) and then enriched with the cluster profiler package (version 4.2.2) for gene ontology, disease ontology, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis [32, 33]. The q value ≤ 0.05 and gene counts ≥3 were considered significant.
## Construction of PPI network and identification of hub genes
The f-DEGs were analyzed using the STRING database (https://cn.string-db.org/) and a combined interaction score>0.4 was considered statistically significant. Subsequently, the string interactions were visualized using Cytoscape software (version 3.9.0). CytoNCA (2.1.6), a Cytoscape plug-in, was used to filter the density and importance of modules in the PPI networks. Then, the R software screened out the hub genes in the network, based on the following principles: betweenness, closeness, and degree are greater than average.
## The hub gene of F-DEGs analysis in THCA
The THCA gene expression profiles were downloaded from the TCGA database. There were 568 cases of TCGA (510 cases in the tumor group and 58 cases in the normal group). The online database ULCAN (http://ualcan.path.uab.edu) analyzed the gene expression level of HIF1A in cancer tissues and normal tissues of TCGA database samples. Validation of the effect of HIF1A expression on THCA survival through the Kaplan-Meier plotter (http://kmplot.com/analysis/). Transcriptome data for THCA were grouped according to pathology type or gene expression. CIBERSORT was used to compute immune cell composition based on gene expression profiles. R software (including vioplot, ggpubr, and ggExtra packages) was used to analyze the correlation between HIF1A and THCA immune infiltration.
## Immune infiltration in OSAS
In the GSE135917 dataset, single sample gene set enrichment analysis was used to calculate the per sample infiltration levels of immune cell types. The ggboxplot function in the ggpubr package (version 0.4.0) was used to plot box line plots of immune cells in the two groups of samples. The heatmaps were plotted using the pheatmap package (version 1.0.12), showing the relationship between hub genes and various immune cells.
We performed immune cell analysis on the GSE135917 dataset to understand the role of OSAS on the immune environment of adipocytes. OSAS was associated with an increase in activated CD4 T cells, gamma delta T cells, and regulatory T cells (Supplementary Figure S2A). A correlation heat map was generated based on the immune cell contact relationship (Supplementary Figure S2B). Analysis of the relationship between hub genes and immune cells demonstrated that HIF1A was the most important gene in OSAS immune cells (Supplementary Figures S2C, D). According to Figure 5 and Supplementary Figure S2, we found that HIF1A is critical for the development of THCA and OSAS, which may increase the risk of disease by affecting the immune microenvironment.
## F-DEGs validation in GSE38792 and animal model
GSE38792 was utilized to validate the differential expression of hub genes between OSAS patients and healthy volunteers. A variance analysis was performed using the stat_compare_means function in the ggpubr package (version 0.4.0), and the results were presented in box plots [34].
Eight male Sprague-Dawley rats (SPF grade, weight of 230–250 g) were obtained from the central laboratory of the Animal Experimental Center at Renmin Hospital of Wuhan University. The rats were fed in a 12-h alternating day and night environment with corresponding humidity and temperature control. The rats were randomly divided into two groups: [1] Normal control group, in which the rats were kept in a normoxic environment for 8 weeks; [2] Chronic intermittent hypoxia group (CIH), in which the mice were kept in a chronic intermittent hypoxic environment 8 h per day for 8 weeks. The oxygen concentration in the modeling chamber alternated between 30 s of hypoxia (FiO2, $10\%$) and 60 s of reoxygenation (FiO2, $21\%$). After 8 weeks, we administered general anesthesia to rats, isolated the abdominal aorta, and obtained arterial blood for arterial blood gas analysis using a blood gas needle.
## Reverse transcription and real-time PCR analysis
Total RNA was isolated and purified using RNAiso Plus reagent (9108, Takara, Japan) according to the manufacturer's protocol. Based on the TaqMan probe method, The mRNA expressions of hub genes (HIF1A, ATM, MAPK8, and MAPK14) were detected according to the instructions of probe One-Step qRT-PCR Kit (D7277, Beyotime, China). The reactions were carried out in the cycler under the following conditions: 50°C for 20 min, 95°C for 2 min, 95°C for 15 s and 60°C for 20 s (40 cycles in total). GAPDH was used as a housekeeping gene, and the RT-qPCR primer sequences are listed in Table 1. The relative gene expression level was calculated using the 2– ΔΔCT method [35].
**Table 1**
| Target | Forward primer (5'-3') | Reverse primer (3'-5') | bp |
| --- | --- | --- | --- |
| Hif1a (Rat) | AAGCAGCAGGAATTGGAACG | CTCGTTTCCAAGAAAGCGACA | 75 |
| Atm (Rat) | CAGCTTTAGAGAGGTGTGTAATGA | AAGTCTCTGCCAGCCAGTTG | 89 |
| Mapk8 (Rat) | ACAGCTCGGAACACCTTGTC | TCGCCTGACTGGCTTTAAGT | 167 |
| Mapk14 (Rat) | GCACTACAACCAGACAGTGGA | GTCCCCGTCAGACGCATTAT | 129 |
| GAPDH (Rat) | CCGCATCTTCTTGTGCAGTG | CGATACGGCCAAATCCGTTC | 79 |
## ROS measurements and ELISA assay
ROS production was detected by in situ staining of fresh lung tissue with a ROS fluorescent probe-dihydroethidium (DHE, D7008, Sigma, 1:500). In this process, sections incubated with stain sections were observed under a fluorescence microscope (IX53, Olympus, Japan). Fluorescein-labeled ROS-positive sections emitted red fluorescence (excitation wavelength 490 nm, emission wavelength 560 nm). At least three sites were selected for each sample. The fluorescence intensity of the cell sections was measured using Image J software.
The lungs were quickly removed after the rats were anesthetized and sacrificed. The saline was rinsed and blotted dry with filter paper. The lung tissue was ground with a cryogenic grinder at 4°C, and the supernatant was centrifuged and assayed according to the manufacturer's instructions for the MDA and GSH kits (A001-3-1/A006-2-1, Nanjing Jiancheng Bioengineering Institute, China). The contents of MDA and GSH were measured in the normal and CIH groups for intra-assay and inter-assay repeated experiments, respectively. The intra- and inter-coefficients of variation were both < $10\%$.
## Immunohistochemistry
After the inguinal fat was removed from the rat, the adipose tissue was pressed with oil-absorbing paper [36]. The expressions of HIF1A and Atm in rat inguinal adipose tissue were detected using immunohistochemistry. The sections (6 μm) were processed with deparaffinized and rehydration using xylene and different concentration gradients of ethanol. The sections were added to citric acid antigen repair solution (pH 6.0) and then heated in a 95°C water bath for 20 min for antigen repair [37]. The primary antibodies used were HIF1A (1:500, K000487P, Solarbio, China), ATM (1:200, K009314P, Solarbio, China). The sections were incubated with HRP-labeled secondary antibodies for 1 h at room temperature. The relative IOD of the immune sections was measured with ImageJ software.
## ROC curve, hub genes relationships, and drug therapy OSAS prediction
We estimated the area under the ROC curve (AUC) using version 1.8 of the standard pROC tool for the R software. The Pearson correlation coefficient was used to conduct correlation analysis on the hub genes. The CPAP-treated population belongs to the second group in the dataset GSE135917. The paired t-test was performed to compare the variation in hub genes expression before and after the application of a non-invasive ventilator. The DrugBank database was used to search for probe targets (Version 4.2) [38].
## Statistics analysis
GraphPad Prism 8 statistical software (La Jolla, CA, USA) was utilized for statistical analyses. All results were expressed as a mean ± standard deviation (SD) from three independent experiments. The data were tested with the Shapiro-Wilk normality test and the variance homogeneity test. Comparisons were tested using paired or unpaired t-tests and the level of confidence was set at $95\%$ ($P \leq 0.05$).
## The flowchart of the research and animal model of OSAS
Figure 1A depicts the whole study methodology, including bioinformatics analysis and rat validation model. The animal model of OSAS consists of three components: an intermittent hypoxic chamber, a nitrogen tank, and an oxygen compressor that varies FiO2 between 10 and $21\%$ (Figure 1B).
**Figure 1:** *The research flowcharts and animal models. (A) Flowchart of bioinformatics analysis and experimental validation. (B) The experimental model of chronic intermittent hypoxia in rats.*
## RNA degradation and PLS-DA model
To evaluate the accuracy of the gene sequencing in the GSE135917 dataset, we employed RNA degradation curve analysis and PLS-DA. The modest slope of each curve shows that mRNA is not degraded explicitly (Supplementary Figure S1A). The 2D and 3D structure of the PLS-DA model is depicted in Supplementary Figures S1B, C. Four components were used to construct the PLS-DA model, which had a cumulative explanatory power of $99.4\%$ and a cumulative predictive power for the dependent variable of $73\%$. To determine if the PLS-DA model was overfitted, the model was evaluated by holding the X matrix constant and randomly rearranging the Y matrix variables 200 times to generate permutation test results (Supplementary Figure S1D). The results indicate that the normal and OSA groups can be discriminated against without difficulty and that the model is not overfitting.
## Identification of DEGs and GSEA, GSVA enrichment analysis
After GSE135917 Validation, we filtered the DEGs and performed GSEA and GSVA. A total of 23,281 gene expression values were gathered. The OSAS samples contained 1,470 down-regulated genes and 1,674 up-regulated genes, as compared to the normal controls.
The heat map depicts the 50 genes with the greatest differences (positive and negative values) according to logFC values in the dataset (Figure 2A). Ferroptosis, Alzheimer's disease, insulin signaling, mapk signaling pathway, and oxidative stress response were all significantly different between the OSAS and normal groups. The ferroptosis pathway was significantly different between the two groups (NES = 1.8234, $$P \leq 0.0005$$, and FDR = 0.005) (Figure 2B). The WikiPathways were evaluated using gene set variation analysis, and it was determined that there was a significant difference in ferroptosis between the two groups, with elevated sample scores in the OSAS group (Figure 2C). The ferroptosis gene set included 64 genes, and the circular heat map revealed that GCLM, HMGCR, SLC38A1, and CHMP5 were elevated in OSAS (Figure 2D). This grants the theory that ferroptosis may play a significant role in the pathophysiology of OSAS.
**Figure 2:** *Identification of the DEGs in the GSE135917 dataset, GSEA analysis, and GSVA heatmap analysis. (A) Heatmap plot of the DEGs for OSAS vs. controls. (B, C) GSEA and GSVA analysis in wikipathways for OSAS vs. controls. (|NES| > 1, FDR < 0.1, P-value < 0.05). (D) Ferroptosis genes ring heatmap.*
## F-DEGs enrichment analysis
To determine whether DEGs are involved in ferroptosis, we retrieved f-DEGs from the FerrDb V2 database and performed functional enrichment on them. The intersection of 543 ferroptosis-related genes from the FerrDb V2 database with DEGs using Venn diagrams generated 74 f-DEGs (Figure 3A). The KEGG pathway database was used to identify 102 enriched pathways, including ferroptosis, endocrine resistance, mitophagy-animal, autophagy-animal, and Kaposi's sarcoma-associated herpesvirus infection (Figure 3B, Supplementary Table S1).
**Figure 3:** *Identification of f-DEGs,f-DEGs enrichment analysis. (A) 74 f-DEGs were obtained from the intersection of the FerrDb V2 database with DEGs. (B) KEGG pathway analysis of f-DEGs. (C) GO enrichment analysis f-DEGs. (D) GO term analysis on f-DEGs using the Metascape website. (E) Disease Ontology analysis of f-DEGs.*
The most enriched pathways in terms of GO terms were those associated with neuron death, regulation of neuron death, cellular response to metal ions, cellular response to chemical stress, and reaction to metallic ions (Figure 3C). On the Metascape website (https://metascape.org/), GO analysis of f-DEG revealed that ferroptosis-related pathways are involved in the regulation of neuronal death (Figure 3D). The f-DEGs were evaluated in the Disease Ontology database, and thyroid carcinoma, papillary thyroid carcinoma, thyroid cancer, and colon cancer were discovered to be enriched disorders (Figure 3E). We discovered that f-DEGs are closely associated with THCA and may play an important role in its onset.
## Identification of hub genes
By establishing a PPI network, we were able to identify the hub genes in f-DEGs and establish their relationship. A 71-node, 142-edge PPI network was constructed based on the biological interactions of 74 f-DEGs. The PPI enrichment P-value is 9.9910-16, and the average local clustering coefficient is 0.419. CytoNCA (2.1.6) was applied to analyze f-DEGs, resulting in a network of 53 nodes and 284 edges (Figures 4A, B). Furthermore, the hub genes were discovered using the same screening method: HIF1A, ATM, HSPA5, MAPK14, KRAS, MAPK8, TLR4, and CREB1 (Figure 4C). In the network, HIF1A is the most significant gene. Since f-DEGs are primarily enriched in THCA (Figure 3), we hypothesize that HIF1A may be the principal gene of THCA.
**Figure 4:** *PPI network and cytoscape to obtain hub genes. (A, B) F-DEGs were analyzed with a PPI network and visualized with Cytoscape. (C) Hub network construction of HIF1A, ATM, MAPK14, TLR4, HSPA5, MAPK8, and CREB1.*
## HIF1A over expression promotes THCA progress
We investigated the association between HIF1A and THCA, which was obtained from the TCGA database. The effect of high HIF1A expression in OSAS on thyroid cancer survival discovered that HIF1A expression in thyroid tissue was also increased, and patients with high HIF1A expression had a lower survival time. As a result, HIF1A overexpression in OSAS may accelerate the occurrence and progression of thyroid cancer (Figures 5A, B). This research examines the relationship between HIF1A, TNM, and thyroid cancer stage to determine the impact of HIF1A on thyroid cancer. HIF1A expression is lowest in normal people and increases with the T and N stages. In the M stage, the 0 stage was highest. HIF1A promotes thyroid carcinoma in patients (Figure 5C). To investigate whether HIF1A promotes THCA, the enrichment pathway was examined. HIF1A is enriched in the ferroptosis, ROS, cancer, and apoptosis pathways of THCA (Figure 5D). The immune milieu plays a crucial role in the development of tumors. HIF1A expression is intimately associated with dendritic cells, macrophages, mast cells, and CD8 T cells in the immunological milieu of THCA (Figures 5E, G). HIF1A causes alterations in the immunological microenvironment of thyroid cancer, which may be associated with ferroptosis of thyroid cells produced by an increase in ROS. OSAS potentially promote the onset and progression of THCA by increasing HIF1A expression.
**Figure 5:** *Increased expression of HIF1A promotes the progression of THCA. (A) The HIF1A expression difference in the tumor group vs. the normal group. (B) Kaplan-Meier plot illustrating the influence of HIF1A expression level on THCA survival. (C) HIF1A and TNM staging of thyroid cancer. (D) GSEA analysis of THCA pathway differences (FDR < 0.25 is deemed significant). (E–G) Correlation of THCA and HIF1A with immunological infiltration. *P < 0.05, **P < 0.01, ***P < 0.001.*
## Hub genes validation and CIH experimental validation
We validated the expression of hub genes in another dataset (GSE38792). OSAS had elevated the expression of HIF1A, ATM, HSPA5, MAPK8, MAPK14, TLR4, and CREB1. HIF1A and ATM exhibited statistically significant differences (Figure 6A). HIF1A and Atm mRNA expressions were elevated in rats exposed to CIH, with P-values of 0.037 and 0.002, respectively (Figure 6B). Rearrangement of the OSAS phenotype based on HIF1A expression demonstrated that the ferroptosis pathway remained significantly distinct, NES = 1.4508007, $P \leq 0.05$ (Figure 6C). PO2 (CIH group): 51.86 ± 2.97 mmHg, PO2 (Normal group): 69.87 ± 2.45 mmHg. There is a statistical difference between the two groups ($P \leq 0.05$).
**Figure 6:** *The hub genes (HIF1A and ATM) were validated in another OSAS dataset and the rats qRT-PCR experiment. (A) The hub genes were imported into the GSE38792 dataset for validation. (B) The mRNA levels of hub genes in the rats of CIH and Normal groups. (C) GSEA analysis of HIF1A expression level, NES = 1.4508007, P = 0.024. *P < 0.05, **P < 0.01.*
Figure 7 depicts the values of the oxidative stress markers (ROS, MDA, and GSH) and the hub genes (HIF1A, Atm). CIH exposure increased the expression of ROS in rat lung tissue compared to the control group (Figures 7A, B). MDA was greatly increased in the CIH group compared to the control group, although GSH was dramatically decreased, and the P-values were all significant ($P \leq 0.01$) (Figures 7C, D). The expression of HIF1A and Atm proteins in inguinal fat was examined using immunohistochemical techniques in CIH-exposed rats and was significantly elevated in the CIH group with a significant P-value of 0.0049 (Figures 7 E, F). Both Figures 6, 7 demonstrate that HIF1A and ATM play a significant role in the process of OSAS-induced ferroptosis, which is intimately connected to the elevation in ROS levels and the decline in GSH levels.
**Figure 7:** *Establishment of chronic intermittent hypoxia model (CIH) in rats, validated by ROS, MDA, GSH, and IHC. (A, B) The expression of ROS in rat lung tissue in different groups. Original magnification is × 400, Scare bar, 25 μm. (C, D) The expression of MDA and GSH in different groups. (E, F) HIF1A and ATM protein expression variations in different groups were detected by immunohistochemistry. Magnification × 200, Scare bar, 100 μm. *P < 0.05.*
## ROC diagnostic curve and hub genes relationships
We performed a diagnostic curve analysis to assess the significance of hub genes (HIF1A and ATM) in the diagnosis of OSAS. The respective areas of HIF1A under the curves (AUC) were 0.838 and 0.80 in both datasets, corresponding to an AUC of 0.838, 0.812 for ATM, respectively (Figures 8A, B). The AUC of the combined HIF1A and ATM model for the diagnosis of OSAS was 0.875 in both datasets, suggesting that the model is beneficial for the identification of OSAS (Figures 8C, D). To assess the interaction analysis between hub genes, Pearson correlation analysis was performed. HIFA and ATM were found to be strongly correlated in both datasets, with correlation coefficients (r) of 0.56, 0.70 and P-values of 0.02, 1.2e-3, respectively (Figures 8E, F).
**Figure 8:** *HIF1A and ATM diagnosis curves, expression correlation analysis, and non-invasive ventilator pairing therapy modification. (A, B) In GSE135917 and GSE38792 datasets, HIF1A, and ATM expression were examined for ROC. (C, D) HIF1A and ATM combined models in GSE135917 and GSE38792 datasets. (E, F) The person correlation analysis of HIF1A and ATM. (G) Differential expression of HIF1A and ATM before and after non-invasive ventilator (CPAP) treatment in GSE135917.*
It was found that there was no significant change in HIF1A and ATM in OSAS patients treated with ventilators, with P-values of 0.74, and 0.31, respectively (Figure 8G). There was no significant change in HIF1A and ATM after short-term non-invasive ventilator (CPAP) therapy, and we investigated pharmacological treatment of OSAS in the following phase.
## Drugs from the drugbank
DrugBank includes the most comprehensive information on medications and their targets, so it provides us with the most important drug targets. Eight drugs targeting these two hub genes were identified from the Drugbank (Table 2). Among these drugs, the modulators are mainly 2-methoxy estradiol, carvedilol, ENMD-1198, and FG-2216, the inducers are hydralazine, and the inhibitors are PX-478 and vadarestat, The stabilizer has caffeine. The approved pharmaceuticals include carvedilol, hydralazine, and caffeine, which are potential OSAS treatments.
**Table 2**
| Drugbank ID | Name | Group | Target | Actions |
| --- | --- | --- | --- | --- |
| DB02342 | 2-Methoxyestradiol | Investigational | HIF1A | Modulator |
| DB01136 | Carvedilol | Approved, investigational | HIF1A | Modulator |
| DB05959 | ENMD-1198 | Investigational | HIF1A | Modulator |
| DB08687 | FG-2216 | Investigational | HIF1A | Modulator |
| DB01275 | Hydralazine | Approved | HIF1A | Inducer |
| DB06082 | PX-478 | Investigational | HIF1A | Inhibitor |
| DB12255 | Vadadustat | Investigational | HIF1A | Stabilization |
| DB00201 | Caffeine | Approved | ATM | Inhibitor |
## Discussion
Chronic intermittent hypoxia is the main pathophysiological mechanism of OSAS. CIH can lead to an increase in ROS, which may result in cellular ferroptosis [20]. Ferroptosis was important in CIH-induced liver and cardiac injury [39, 40], however, there are limited investigations on CIH for lung, adipose tissue, and tumor. We analyzed OSAS and ferroptosis-related f-DEGs for functional pathways. The HIF1A and ATM are hub genes in f-DEGs utilizing the PPI network and Cytoscape software.
The KEGG pathway showed that ferroptosis is closely related to autophagy, which is essential for driving cells to undergo ferroptosis. Regulatory mechanisms and signaling pathways for autophagy-dependent ferroptosis may enhance the study of chemo modulators of ferroptosis and may be developed for therapeutic interventions in human diseases [41]. Ferroptosis is connected to lysosomes and autophagosomes, according to CC enrichment analysis. The molecular function of ferroptosis may involve DNA transport and activation of the mitogen-activated protein kinase pathway. An anesthetic drug (lidocaine) attenuates pulmonary epithelial cell ferroptosis in a hypoxia/reoxygenation-induced model by modulating the p38 MAPK pathway, suggesting that it is closely related to ferroptosis [42]. This is consistent with the enrichment results of f-DEGs in the Disease Ontology database, indicating that CIH may affect the progress of cancers, particularly THCA.
The THCA data was gathered from the TCGA database, the hub genes (HIF1A, ATM) in f-DEGs were investigated, and HIF1A was overexpressed in THCA. There is a lack of research on the relationship between OSAS and thyroid cancer at the present, however, HIF1A is closely associated with both diseases. According to our findings, OSAS can cause CIH, which increases ROS, promotes ferroptosis, and contributes to the onset and progression of THCA. Hypoxia causes cervical lymph node metastases and thyroid cancer recurrence, however, its mechanism is unknown [43]. FGF11 interacted with HIF1A to increase thyroid cancer growth and metastasis [44]. HIF1A expression correlated positively with Medullary thyroid carcinoma prognosis (MTC) [45]. Most likely, the presence of fewer T cells in OSAS patients is due to higher HIF1A levels [46]. The analysis of THCA immune cell infiltration revealed a more significant reduction of CD8 T cells [47]. OSAS is regarded as a low-grade systemic inflammation triggered by CIH, which can increase the inflammatory factors NF-kB and HIF1A [48]. In our research, active CD4 T cells were increased in OSAS, although activated B cells were lower, indicating that OSAS may be associated with immune dysregulation. CIH and chronic inflammation may inhibit the activation of immune cells, especially lymphocytes and monocytes. We believe that the activation of immune cells helps OSAS patients to fight chronic inflammation. However, this may affect their metabolism, such as fat metabolism, and could exacerbate their disease progression. Additionally, it was discovered that an overexpression of HIF1A was associated with decreased immunological activity as well as a decreased survival rate in malignancies [49]. Consequently, our findings may demonstrate that HIF1A is probable to cause the progression of OSAS and THCA diseases via the ferroptosis mechanism and immunological microenvironment modifies.
The validation in the GSE38792 dataset revealed that the expression of HIF1A and ATM was also elevated. Subsequent immunohistochemistry assays confirmed that HIF1A and Atm proteins were also raised in the adipose tissue of rats. CIH causes a rise in cellular HIF1A, which results in increased NADPH oxidase 4 (NOX4) activity, which can produce more ROS [50]. NOX4 promotes ferroptosis-dependent cytotoxicity via increasing oxidative stress-induced lipid peroxidation [51]. Ataxia-telangiectasia mutated (ATM) is a protein kinase that is necessary for cellular inflammatory toxicity along with oxidative stress-induced cell death [52]. Chen et al. identified ATM as the primary ferroptosis kinase using siRNA knockdown [53]. Previous research has demonstrated that repeated hypoxic reoxygenation of CIH results in an excessive generation of ROS [54]. Our investigation revealed that CIH can not only increase ROS and MDA but also decrease GSH. ROS can cause ferroptosis by activating autophagy and increasing intracellular iron levels through increasing ferritin and transferrin receptors [19].
Our results reveal that HIF1A is significantly positively correlated to ATM in the course of OSAS. Hypoxia can inhibit mTORC signaling, leading to ATM-dependent HIF1A phosphorylation at serine 696 and mediating the downregulation of mTORC1 signaling [55]. The ROC curve indicates that the model is advantageous for identifying OSAS. Compared to polysomnography, however, this technique is invasive. Polysomnography is a complicated procedure that is easily influenced by the patient's mental state. The combination of the two may thus aid in the identification of OSAS. It is still debatable whether a ventilator or surgery should be used to treat OSAS [56]. Figure 8G shows that 2 weeks of ventilator treatment did not reduce the expression of HIF1A and ATM, implying that CPAP treatment is a long-term procedure. In the DrugBank database, we anticipate discovering drugs that could be utilized to treat OSAS. Hydralazine can reduce the protein levels of HIF1A and its downstream target genes to increase cellular antioxidant capacity [57]. CIH can cause raised HIFA and ATM, which can generate high ROS, that can accelerate ferroptosis, resulting in diminished or non-existent mitochondrial cristae, ruptured and contracted outer mitochondrial membranes, and darkened mitochondria (Figure 9).
**Figure 9:** *Graphical summary illustrating elevated HIF1A and ATM expression, mediates ROS elevation by chronic intermittent hypoxia promotes ferroptosis.*
Our research also has some limitations. Our raw data comes from online databases. *The* genes that cause ferroptosis are still inadequately known. In vitro research aids in elucidating the molecular process underlying ferroptosis.
## Conclusion
Our research discovered that HIF1A and ATM are crucial genes in the process of CIH that leads to ferroptosis and that changes in the immunological microenvironment promote the progression of tumor disorders such as THCA. At present, the main method for diagnosing OSAS is polysomnography, but there is a lack of specific markers. The hub genes (HIF1A and ATM) can serve as biomarkers and therapeutic targets for OSAS. Therefore, this study may provide new insights into the role of ferroptosis in the pathogenesis of OSAS.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
This research was carried out in accordance with the Regulations of Experimental Animal Administration issued by the State Committee of Science and Technology of the People's Republic of China, with the approval of the Ethics Committee in Renmin Hospital of Wuhan University(IACUC Issue No: 20220501A).
## Author contributions
All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur.2023.1130378/full#supplementary-material
## References
1. Benjafield AV, Ayas NT, Eastwood PR, Heinzer R, Ip MS, Morrell MJ. **Estimation of the global prevalence and burden of obstructive sleep apnoea: a literature-based analysis**. *The Lancet Respiratory Medicine.* (2019) **7** 687-98. DOI: 10.1016/s2213-2600(19)30198-5
2. I Almendros I, Gileles-Hillel A, Khalyfa A, Wang Y, Zhang SX, Carreras A. **Adipose tissue macrophage polarization by intermittent hypoxia in a mouse model of OSA: effect of the tumor microenvironment**. *Cancer Lett* (2015) **361** 233-39. DOI: 10.1016/j.canlet.2015.03.010
3. Gao T, Zhang Q, Hou J, Zhu K, Sun B, Chen J. **Vestibular-evoked myogenic potentials in patients with severe obstructive sleep apnea**. *J Int Med Res.* (2020) **48** 300060520909717. DOI: 10.1177/0300060520909717
4. Chen L, Ma W, Tang W, Zha P, Wang C, Chen D. **Prevalence of obstructive sleep apnea in patients with diabetic foot ulcers**. *Front Endocrinol.* (2020) **11** 416. DOI: 10.3389/fendo.2020.00416
5. M Zaffanello M, Antoniazzi F, Tenero L, Nosetti L, Piazza M, Piacentini G. **Sleep-disordered breathing in paediatric setting: existing and upcoming of the genetic disorders**. *Ann Transl Med* (2018) **6** 343. DOI: 10.21037/atm.2018.07.13
6. Shi Y, Luo H, Liu H, Hou J, Feng Y, Chen J. **Related biomarkers of neurocognitive impairment in children with obstructive sleep apnea**. *Int J Pediatr Otorhinolaryngol* (2019) **116** 38-42. DOI: 10.1016/j.ijporl.2018.10.015
7. MMartínez-García MÁ, Campos-Rodriguez F, Barbé F. **Cancer and OSA: current evidence from human studies**. *Chest.* (2016) **150** 451-63. DOI: 10.1016/j.chest.2016.04.029
8. Hou J, Zhao L, Yan J, Ren X, Zhu K, Gao T. **MicroRNA expression profile is altered in the upper airway skeletal muscle tissue of patients with obstructive sleep apnea-hypopnea syndrome**. *J Int Med Res.* (2019) **47** 4163-82. DOI: 10.1177/0300060519858900
9. Liu X, Ma Y, Ouyang R, Zeng Z, Zhan Z, Lu H. **The relationship between inflammation and neurocognitive dysfunction in obstructive sleep apnea syndrome**. *J Neuroinflammation.* (2020) **17** 229. DOI: 10.1186/s12974-020-01905-2
10. S Schütz SG, Dunn A, Braley TJ, Pitt B, Shelgikar AV. **New frontiers in pharmacologic obstructive sleep apnea treatment: a narrative review**. *Sleep Med Rev* (2021) **57** 101473. DOI: 10.1016/j.smrv.2021.101473
11. Dixon SJ, Lemberg KM, Lamprecht MR, Skouta R, Zaitsev EM, Gleason CE. **Ferroptosis: an iron-dependent form of nonapoptotic cell death**. *Cell.* (2012) **149** 1060-72. DOI: 10.1016/j.cell.2012.03.042
12. H Yan HF, Zou T, Tuo QZ, Xu S, Li H, Belaidi AA. **Ferroptosis: mechanisms and links with diseases**. *Signal Transduct Target Ther* (2021) **6** 49. DOI: 10.1038/s41392-020-00428-9
13. Ye LF, Chaudhary KR, Zandkarimi F, Harken AD, Kinslow CJ, Upadhyayula PS. **Radiation-Induced Lipid Peroxidation Triggers Ferroptosis and Synergizes with Ferroptosis Inducers**. *ACS Chem Biol.* (2020) **15** 469-84. DOI: 10.1021/acschembio.9b00939
14. Zhang GY, Liu MZ, Liu C. **Mechanisms and pharmacological applications of ferroptosis: a narrative review**. *Ann Transl Med.* (2021) **9** 1503. DOI: 10.21037/atm-21-1595
15. 15.Disturbed Sleep OSAS and Metabolic Diseases. J Diabetes Res. (2019) 2019:1463045. 10.1155/2019/146304531641672. *J Diabetes Res.* (2019) **2019** 1463045. DOI: 10.1155/2019/1463045
16. Schmickl CN, Landry SA, Orr JE, Chin K, Murase K, Verbraecken J. **Acetazolamide for OSA and central sleep apnea: a comprehensive systematic review and meta-analysis**. *Chest.* (2020) **158** 2632-45. DOI: 10.1016/j.chest.2020.06.078
17. Xu W, Deng H, Hu S, Zhang Y, Zheng L, Liu M. **Role of ferroptosis in lung diseases**. *J Inflamm Res.* (2021) **14** 2079-90. DOI: 10.2147/JIR.S307081
18. J Li J, Lu K, Sun F, Tan S, Zhang X, Sheng W. **Panaxydol attenuates ferroptosis against LPS-induced acute lung injury in mice by Keap1-Nrf2/HO-1 pathway**. *J Transl Med* (2021) **19** 96. DOI: 10.1186/s12967-021-02745-1
19. E Park E, Chung SW. **ROS-mediated autophagy increases intracellular iron levels and ferroptosis by ferritin and transferrin receptor regulation**. *Cell Death Dis* (2019) **10** 822. DOI: 10.1038/s41419-019-2064-5
20. G Yuan G, Nanduri J, Khan S, Semenza GL, Prabhakar NR. **Induction of HIF-1alpha expression by intermittent hypoxia: involvement of NADPH oxidase, Ca2+ signaling, prolyl hydroxylases, and mTOR**. *J Cell Physiol* (2008) **217** 674-85. DOI: 10.1002/jcp.21537
21. Palazon A, Goldrath AW, Nizet V, Johnson RS. **transcription factors, inflammation, and immunity**. *Immunity.* (2014) **41** 518-28. DOI: 10.1016/j.immuni.2014.09.008
22. Dewhirst MW, Cao Y, Moeller B. **Cycling hypoxia and free radicals regulate angiogenesis and radiotherapy response**. *Nat Rev Cancer.* (2008) **8** 425-37. DOI: 10.1038/nrc2397
23. E LaGory EL, Giaccia AJ. **The ever-expanding role of HIF in tumour and stromal biology**. *Nat Cell Biol* (2016) **18** 356-65. DOI: 10.1038/ncb3330
24. Wan J, Wu W. **Hyperthermia induced HIF-1a expression of lung cancer through AKT and ERK signaling pathways**. *J Exp Clin Cancer Res.* (2016) **35** 119. DOI: 10.1186/s13046-016-0399-7
25. Mateika JH, Syed Z. **Intermittent hypoxia, respiratory plasticity and sleep apnea in humans: present knowledge and future investigations**. *Respir Physiol Neurobiol* (2013) **188** 289-300. DOI: 10.1016/j.resp.2013.04.010
26. Wu X, Gong L, Xie L, Gu W, Wang X, Liu Z. **NLRP3 deficiency protects against intermittent hypoxia-induced neuroinflammation and mitochondrial ROS by promoting the PINK1-Parkin pathway of mitophagy in a murine model of sleep apnea**. *Front Immunol.* (2021) **12** 628168. DOI: 10.3389/fimmu.2021.628168
27. Gharib SA, Hayes AL, Rosen MJ, Patel SR. **A pathway-based analysis on the effects of obstructive sleep apnea in modulating visceral fat transcriptome**. *Sleep.* (2013) **36** 23-30. DOI: 10.5665/sleep.2294
28. Gharib SA, Hurley AL, Rosen MJ, Spilsbury JC, Schell AE, Mehra R. **Obstructive sleep apnea and CPAP therapy alter distinct transcriptional programs in subcutaneous fat tissue**. *Sleep* (2020) **43** zsz314. DOI: 10.1093/sleep/zsz314
29. J Westerhuis JA, Kourti T, MacGregor JF. **Analysis of multiblock and hierarchical PCA and PLS models**. *J Chemom* (1998) **12** 301-21. DOI: 10.1002/(SICI)1099-128X(199809/10)12:53.0.CO;2-S
30. Robotti E, Marengo E. **Chemometric Multivariate Tools for Candidate Biomarker Identification: LDA, PLS-DA, SIMCA, Ranking-PCA**. *Methods Mol Biol.* (2016) **1384** 237-67. DOI: 10.1007/978-1-4939-3255-9_14
31. N Zhou N, Bao J. **FerrDb: a manually curated resource for regulators and markers of ferroptosis and ferroptosis-disease associations**. *Database* (2020) **2020** baaa021. DOI: 10.1093/database/baaa021
32. **The gene ontology (GO) database and informatics resource**. *Nucleic Acids Res* (2004) **32** 258-61. DOI: 10.1093/nar/gkh036
33. Kanehisa M, Goto S, Kawashima S, Okuno Y, Hattori M. **The KEGG resource for deciphering the genome**. *Nucleic Acids Res* (2004) **32** D277-80. DOI: 10.1093/nar/gkh063
34. Bangdiwala SI. **Why analyse variances in order to compare means**. *Int J Inj Contr Saf Promot.* (2015) **22** 89-91. DOI: 10.1080/17457300.2014.996984
35. K Livak KJ, Schmittgen TD. **Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) method**. *Methods* (2001) **25** 402-08. DOI: 10.1006/meth.2001.1262
36. S. Khan SN, Sawaki D, Arnaud C, Pépin JL, Gaucher J, Derumeaux G. **Intermittent hypoxia induces premature adipose tissue senescence leading to cardiac remodeling**. *Arch Cardiovasc Dis Supp* (2020) **12** 247-48. DOI: 10.1016/j.acvdsp.2020.03.116
37. W Liu W, Zhao D, Wu X, Yue F, Yang H, Hu K. **Rapamycin ameliorates chronic intermittent hypoxia and sleep deprivation-induced renal damage via the mammalian target of rapamycin (mTOR)/NOD-like receptor protein 3 (NLRP3) signaling pathway**. *Bioengineered* (2022) **13** 5537-50. DOI: 10.1080/21655979.2022.2037872
38. Wishart DS, Feunang YD, Guo AC, Lo EJ, Marcu A, Grant JR. **DrugBank 5.0: a major update to the DrugBank database for 2018.**. *Nucleic Acids Res.* (2018) **46** D1074-82. DOI: 10.1093/nar/gkx1037
39. Huang J, Xie H, Yang Y, Chen L, Lin T, Wang B. **The role of ferroptosis and endoplasmic reticulum stress in intermittent hypoxia-induced myocardial injury**. *Sleep Breath* (2022) **3** 1. DOI: 10.1007/s11325-022-02692-1
40. Chen LD, Wu RH, Huang YZ, Chen MX, Zeng AM, Zhuo GF. **The role of ferroptosis in chronic intermittent hypoxia-induced liver injury in rats**. *Sleep Breath.* (2020) **24** 1767-73. DOI: 10.1007/s11325-020-02091-4
41. Wu Y, Zhang S, Gong X, Tam S, Xiao D, Liu S. **The epigenetic regulators and metabolic changes in ferroptosis-associated cancer progression**. *Mol Cancer.* (2020) **19** 39. DOI: 10.1186/s12943-020-01157-x
42. Ma X, Yan W, He N. **Lidocaine attenuates hypoxia/reoxygenation-induced inflammation, apoptosis, and ferroptosis in lung epithelial cells by regulating the p38 MAPK pathway**. *Mol Med Rep.* (2022) **25** 150. DOI: 10.3892/mmr.2022.12666
43. Zheng F, Chen J, Zhang X, Wang Z, Chen J, Lin X. **The HIF-1α antisense long non-coding RNA drives a positive feedback loop of HIF-1α mediated transactivation and glycolysis**. *Nat Commun.* (2021) **12** 1341. DOI: 10.1038/s41467-021-21535-3
44. B Chen B, Feng M, Yao Z, Zhang Z, Zhang K, Zhou L. **Hypoxia promotes thyroid cancer progression through HIF1α/FGF11 feedback loop**. *Exp Cell Res* (2022) **416** 113159. DOI: 10.1016/j.yexcr.2022.113159
45. Lodewijk L, van Diest P, van der Groep P, Ter Hoeve N, Schepers A, Morreau J. **Expression of HIF-1α in medullary thyroid cancer identifies a subgroup with poor prognosis**. *Oncotarget.* (2017) **8** 28650-59. DOI: 10.18632/oncotarget.15622
46. C Thorn CE, Knight B, Pastel E, McCulloch LJ, Patel B, Shore AC. **Adipose tissue is influenced by hypoxia of obstructive sleep apnea syndrome independent of obesity**. *Diabetes Metab* (2017) **43** 240-47. DOI: 10.1016/j.diabet.2016.12.002
47. Shi M, Cui F, Yang CY, Zhang H, Wang YP, Wei L. **Effects of chronic intermittent hypobaric hypoxia on immune function in rat**. *Chin J Appl Physiol.* (2009) **25** 433-38. DOI: 10.13459/j.cnki.cjap.2009.04.009
48. Ryan S, Taylor CT, McNicholas WT. **Systemic inflammation: a key factor in the pathogenesis of cardiovascular complications in obstructive sleep apnoea syndrome?**. *Postgrad Med J.* (2009) **85** 693-98. DOI: 10.1136/thx.2008.105577
49. B. Chen B, Li L, Li M, Wang X. **HIF1A expression correlates with increased tumor immune and stromal signatures and aggressive phenotypes in human cancers**. *Cell Oncol* (2020) **43** 877-88. DOI: 10.1007/s13402-020-00534-4
50. Xiao R, Wang S, Guo J, Liu S, Ding A, Wang G. **Ferroptosis-related gene NOX4, CHAC1, and HIF1A are valid biomarkers for stomach adenocarcinoma**. *J Cell Mol Med.* (2022) **26** 1183-93. DOI: 10.1111/jcmm.17171
51. Park MW, Cha HW, Kim J, Kim JH, Yang H, Yoon S. **NOX4 promotes ferroptosis of astrocytes by oxidative stress-induced lipid peroxidation via the impairment of mitochondrial metabolism in Alzheimer's diseases**. *Redox Biol.* (2021) **41** 101947. DOI: 10.1016/j.redox.2021.101947
52. T Aki T, Uemura K. **Cell death and survival pathways involving ATM protein kinase**. *Genes* (2021) **12** 1581. DOI: 10.3390/genes12101581
53. Chen PH, Wu J, Ding CK, Lin CC, Pan S, Bossa N. **Kinome screen of ferroptosis reveals a novel role of ATM in regulating iron metabolism**. *Cell Death Differ.* (2020) **27** 1008-22. DOI: 10.1038/s41418-019-0393-7
54. Semenza GL, Prabhakar NR. **HIF-1-dependent respiratory, cardiovascular, and redox responses to chronic intermittent hypoxia**. *Antioxid Redox Signal.* (2007) **9** 1391-96. DOI: 10.1089/ars.2007.1691
55. H Cam H, Easton JB, High A, Houghton PJ. **Houghton.mTORC1 signaling under hypoxic conditions is controlled by ATM-dependent phosphorylation of HIF-1α.**. *Mol Cell* (2010) **40** 509-20. DOI: 10.1016/j.molcel.2010.10.030
56. Tagaya M, Otake H, Suzuki K, Yasuma F, Yamamoto H, Noda A. **The comparison of nasal surgery and CPAP on daytime sleepiness in patients with OSAS**. *Rhinology.* (2017) **55** 269-73. DOI: 10.4193/Rhin17.026
57. Mehrabani M, Nematollahi MH, Tarzi ME, Juybari KB, Abolhassani M, Sharifi AM. **Protective effect of hydralazine on a cellular model of Parkinson's disease: a possible role of hypoxia-inducible factor (HIF)-1α**. *Biochem Cell Biol.* (2020) **98** 405-14. DOI: 10.1139/bcb-2019-0117
|
---
title: 'Mortality and survival in nonagenarians during the COVID-19 pandemic: Unstable
equilibrium of aging'
authors:
- Daria A. Kashtanova
- Veronika V. Erema
- Maria S. Gusakova
- Ekaterina R. Sutulova
- Anna Yu. Yakovchik
- Mikhail V. Ivanov
- Anastasiia N. Taraskina
- Mikhail V. Terekhov
- Lorena R. Matkava
- Antonina M. Rumyantseva
- Vladimir S. Yudin
- Anna A. Akopyan
- Irina D. Strazhesko
- Irina S. Kordiukova
- Alexandra I. Akinshina
- Valentin V. Makarov
- Olga N. Tkacheva
- Sergey A. Kraevoy
- Sergey M. Yudin
journal: Frontiers in Medicine
year: 2023
pmcid: PMC10018166
doi: 10.3389/fmed.2023.1132476
license: CC BY 4.0
---
# Mortality and survival in nonagenarians during the COVID-19 pandemic: Unstable equilibrium of aging
## Abstract
### Introduction
Aging puts the human body under an immense stress and makes it extremely susceptible to many diseases, often leading to poor outcomes and even death. Long-living individuals represent a unique group of people who withstood the stress of time and offer an abundance of information on the body’s ability to endure the pressure of aging. In this study, we sought to identify predictors of overall one-year mortality in 1641 long-living individuals. Additionally, we analyzed risk factors for COVID-19-related morality, since statistics demonstrated an extreme vulnerability of older adults.
### Methods
We conducted a two-stage evaluation, including a comprehensive geriatric assessment for major aging-associated: frailty, cognitive impairment, frontal lobe dysfunction, chronic pain, anxiety, risk of falls, sensory deficit, depression, sarcopenia, risk of malnutrition, fecal and urinary incontinence, dependence in Activities of Daily Living, dependence in Instrumental Activities of Daily Living, polypragmasia, and orthostatic hypotension; extensive blood testing, a survey, and a one-year follow-up interview.
### Results
The most reliable predictors of overall mortality were cognitive impairment, malnutrition, frailty, aging-associated diseases and blood markers indicating malnutrition-induced metabolic dysfunctions (decreased levels of protein fractions, iron, 25-hydroxyvitamin D, and HDL), and aging biomarkers, such as IGF-1 and N-terminal pro b-type natriuretic peptide. In post-COVID 19 participants, the most significant mortality predictors among geriatric syndromes were depression, frontal lobe dysfunction and frailty, and similar to overall mortality blood biomarkers - 25-hydroxyvitamin D, IGF-1, HDL as well as high white blood cell, neutrophils counts and proinflammatory markers. Based on the results, we built a predictive model of overall mortality in the long-living individuals with f-score=0.76.
### Conclusion
The most sensitive and reliable predictors of mortality were modifiable. This is another evidence of the critical importance of proper geriatric care and support for individuals in their “golden years”. These results could facilitate geriatric institutions in their pursuit for providing improved care and could aid physicians in detecting early signs of potentially deadly outcomes. Additionally, our findings could be used in developing day-to-day care guidelines, which would greatly improve prevention statistics.
## Introduction
Globally, people are living longer, and the number of older adults is increasing. Aging causes significant changes in the human body and results in numerous chronic conditions, which make the older population progressively more susceptible to diseases and poor outcomes [1, 2]. This has been unequivocally demonstrated by the COVID-19 pandemic. Counterintuitively, aging also causes some of the risk factors to gain – protective properties. Identifying predictors of poor outcomes in older adults would contribute to timely disease prevention and provide a better understanding of the physiology of aging. Sadly, older adults are too often excluded from relevant studies, the findings of which could directly affect their wellbeing. Here, we focused on long-living adults who represent a unique aging phenotype. A number of studies have addressed mortality in long-living adults has been addressed in (3–5). The Danish 1905-Cohort Survey investigated factors associated with mortality in over 2000 participants aged 85 years and above. The authors found that the 15-month mortality was associated with the degree of disability; a low level of mental and physical activity; and, in women, low self-esteem [6]. In the PLAD study of mortality in a Chinese cohort of the oldest-old, age and aging-associated diseases were predictors of mortality, whereas high MMSE scores and high level of physical activity contributed to survival; moreover, the survival rate was higher in women [7]. The Mugello Study found that cognitive disorders, ADL, polypragmasy, and renal dysfunction were predictors of the 12-month mortality in nonagenarians [8].
During the COVID-19 pandemic, several studies analyzed the COVID-19-related mortality in people older than 90 years of age. The mortality rate was higher in the 90-year-old hospitalized patients with functional dependencies [1]. Analysis of mortality rates in people aged 60 years and above showed that dementia was associated with an increased COVID-19-related mortality rate [9].
These used data from cohorts from different countries and projects and have constantly revealed common patterns, which significantly contribute to mortality. They were mostly focused on non-modifiable markers reflective of current health status. However, prevention of undesirable outcomes largely relies on predictive and modifiable markers.
In the present study, we analyzed data from 1,641 long-living Russian adults aged 90 years and above. We assessed their health status and examined associations between geriatric syndromes (GSs), clinical and biochemical parameters and one-year mortality. Presently, this study is the largest of this design type in Russia. Coincidentally, it was carried out during the COVID-19 pandemic; therefore, we investigated the association between the studied factors and COVID-19-related mortality.
## Study design
Initially, we recruited 2020 long-living adults aged 90 years and above. To provide validity, we based our results on full sets of data, which were collected in a two-stage procedure. Data sets for 379 participants were incomplete at the end of the second stage; therefore, we excluded them from the study, which brought the number of participants down to 1,641.
All participants provided informed consent.
The first stage was conducted in 2020: participants were visited by a physician and a nurse for a comprehensive geriatric assessment, biomaterial sampling (whole blood and blood serum), and survey completion. The survey aimed at collecting information about health status, assessing lifelong risks of chronic diseases, and analyzing the lifestyle and socioeconomic background. The comprehensive geriatric assessment focused on the following GSs (Table 1) and followed the Clinical guidelines on frailty, approved by the Ministry of Health of the Russian Federation [10]. Supplementary Table S1 provides more information about GS assessment:
**Table 1**
| Geriatric syndrome | Assessment method |
| --- | --- |
| Frailty | The short physical performance battery (SPPB) |
| Cognitive impairment | The mini-mental state examination (MMSE) |
| Frontal lobe dysfunction | The frontal assessment battery (FAB) |
| Chronic pain | Questionnaire 1 (provided in Supplementary Table S1: Questionnaires) |
| Anxiety | Questionnaire 2 (provided in Supplementary Table S1: Questionnaires) |
| Risk of falls | Questionnaire 3 (provided in Supplementary Table S1: Questionnaires) |
| Sensory deficit | Questionnaire 4 (provided in Supplementary Table S1: Questionnaires) |
| Depression | The five-item geriatric depression scale (GDS 5) |
| Sarcopenia | The Simple questionnaire to rapidly diagnose sarcopenia (SARC-F) |
| Risk of malnutrition | The mini nutritional assessment (MNA) |
| Fecal and urinary incontinence | The Barthel Index |
| Activities of daily living (ADLs) | The Barthel Index |
| Instrumental activities of daily living (IADLs) | The Lawton scale |
| Polypragmasia | Interpreted as a simultaneous administration of five or more medications |
| Orthostatic hypotension | Blood pressure test (sitting vs. standing) |
The blood and serum samples were tested for:Complete blood count and white blood cell differential (with Sysmex hematology analyzers).Glucose metabolic panel (glycosylated hemoglobin (colorimetric analysis), glucose (hexokinase analysis) and insulin (immunoassay) levels).Lipid panel (cholesterol (enzymatic analysis), triglycerides (homogeneous enzymatic colorimetric assay), low-density lipoprotein [a direct measurement method for colorimetric determination of cholesterol oxidase and cholesterol esterase) and high-density lipoprotein levels (homogeneous enzymatic colorimetric assay)]; 4. Hepatic cytolysis markers (ALT and AST), Bilirubin and gamma-GT.Markers of hepatocyte cytolysis (ALT and AST) (kinetic UV method), bilirubin [colorimetric assay with a diazo reagent (Endrashik method)], GGT (kinetic colorimetric assay).Ferritin (enzyme-immunoassay), homocysteine (enzyme-immunoassay), fibrinogen (Clauss Method: by thrombin clotting time in diluted plasma).Uric acid [enzymatic (uricase).Renal function test: creatinine (enzymatic), urea (kinetic UV method (urease)) and cystatin C levels (immunoturbidimetry).Prostate-Specific Antigen (PSA) (chemiluminescent immunoassay analysis).Protein fractions: albumin and globulin (capillary electrophoresis).
Hormonal screening for sex hormones: testosterone (solid phase chemiluminescent immunoassay analysis), estrogen (solid phase chemiluminescent immunoassay analysis) and dehydroepiandrosterone sulfate (enzyme-immunoassay); thyroid hormones: thyroid-stimulating hormone (chemiluminescent immunoassay analysis), free T3 (enzyme-immunoassay), and adipokines: adiponectin and leptin (solid phase enzyme-immunoassay)).
Insulin-like growth factor (enzyme-immunoassay).
N-terminal pro–brain natriuretic peptide as a marker of age-related cardiovascular diseases (electrochemiluminescence immunoassay).
25-hydroxyvitamin D (chemiluminescent microparticle immunoassay).
The second stage was conducted a year later: the participanrs were interviewed by phone to collect up-to-date data on their overall health status, past diseases, Covid-19 status (vaccination status, date of contraction, duration, and severity), and vital status (including, when appropriate, the date and cause of death).
Complete data sets were obtained for 1,641 participants, 347 of whom had recovered and 113 had died from COVID-19.
The present longitudinal observational cohort study is a joint effort by the Centre for Strategic Planning and Management of Biomedical Health Risks and Pirogov’s Russian Clinical and Research Center of Gerontology of the Federal Medical Biological Agency. The study is approved by the Local ethics Committee of the Pirogov’s Russian Clinical and Research Center of Gerontology (Protocol 30 from 24 December, 2019).
## Statistical analysis
For statistical analysis, we used Statsmodels, a Python (v.3.6.9) module. The Shapiro–Wilk test showed that most data were distributed non-normally; therefore, we applied a Box-*Cox data* transformation. For data description, we used the median and interquartile range (IQR). The categorical variables are expressed as numbers and percentages.
To establish statistically significant associations between complete blood count and blood chemistry and mortality, we performed logistic regression analysis. We used the least squares method (Statsmodels, Python 3.8.) to estimate the model parameters. To measure the importance of the independent variables and calculate the value of ps, we used an F-test. Age and sex were used as covariates.
We used the following function: y = β_1 * x_1 + β_2 * x_2 + β_3 * x_3 + β_0: y—mortality (1—the number of participants who died within a year after the examination; 0—the number of surviving participants); x1, x2 and x3—sex, age, and the factors, respectively. To establish statistically significant associations between the GSs and mortality, we performed logistic regression analysis (Statsmodels v0.12.2, Python 3.8.). The results are presented as odds ratio (OR), the logistic regression coefficient, and the Pearson’s correlation coefficient.
To decide whether to accept or reject the null hypothesis, we applied the Bonferroni correction.
To build a prognostic model, we used the Random forests algorithm in Scikit-learn, a software machine learning library for the Python programming language, and the data on the tests results, aging-associated diseases, and mortality. We split the data into the training set ($80\%$) and the tests set ($20\%$). The training set was first standardized using the StandardScaler (Scikit-learn). To evaluate the model, we used the ROC-curve and K-fold cross-validation. To calculate the confidence interval for the ROC curve and AUC, we used bootstrap percentile re-sampling with 100 re-samples per model. CI for the ROC curve and AUC was $95\%$.
As a result, we identified variables significant associated with mortality.
## Results
Out of 1,641 participants, 603 ($36.7\%$) were home-based; 538 ($32.8\%$) resided in elderly care facilities; 500 ($30.5\%$) were inpatients. Vital status (dead or alive) was available for 1,641 participants (out of 2020 initially enrolled); $75\%$ of 1,641 them were women, $32.1\%$ of whom lived alone. The excluded participants ($$n = 379$$) did not exhibit any significant differences from other long-living individuals in the given parameters. i.e., gender, age, inclusion criteria, and clinical profiles.
The age median was 92 years (Q1-Q3 91–94 years). By late 2021, 552 ($33.62\%$) participants died. Figure 1 shows the participant inclusion algorithm and key examination results.
**Figure 1:** *Study design and key results.*
By 2021, $33.6\%$ ($$n = 552$$) of the participants ($$n = 1$$,641) had died. Table 2 shows characteristics of the participants.
**Table 2**
| Unnamed: 0 | N† | Results: median [Q1; Q3] or n (%) |
| --- | --- | --- |
| Age in years, median [Q1; Q3] | 1641 | 92 (91; 94) |
| Women, n (%) | 1641 | 1,234 (75.2%) |
| BMI, kg/m2, median [Q1; Q3] | 1503 | 25.5 (22.9; 28.6) |
| Living alone, n (%) | 1633 | 525 (32.1%) |
| Current smoking status, n (%) | 1579 | 6 (0.4%) |
| SPPB score, median [Q1; Q3] | 1537 | 3 (1; 6) |
| Frailty (SPPB≤7), n (%) | 1639 | 1,438 (87.7%) |
| MMSE score, median [Q1; Q3] | 1542 | 23 (17; 26) |
| Cognitive impairment (diagnosed by a neurologist; other than mild cognitive impairment), n (%) | 1542 | 826 (53.6%) |
| FAB score, median [Q1; Q3] | 1523 | 11 (7; 16) |
| Frontal lobe dysfunction, n (%) | 1564 | 1,198 (76.6%) |
| Dependence in ADL, n (%) | 1593 | 1,463 (91.8%) |
| Dependence in IADL, n (%) | 1599 | 1,511 (94.5%) |
| Chronic pain, n (%) | 1580 | 996 (63%) |
| Anxiety, n (%) | 1018 | 362 (35.6%) |
| Risk of falls, n (%) | 1576 | 917 (58.2%) |
| Sensory deficit, n (%) | 1585 | 1,483 (93.6%) |
| GDS-5 score, median [Q1; Q3] | 1493 | 1 (0; 3) |
| Depression, n (%) | 1584 | 764 (48.2%) |
| MNA score, median [Q1; Q3] | 327 | 20 (17; 22.5) |
| No malnutrition/risk of malnutrition/malnutrition, n (%) | 1425 | 199 (14%) /895 (62.8%) /331 (23.2%) |
| Urinary incontinence, n (%) | 493 | 335 (73.5%) |
| Fecal incontinence, n (%) | 456 | 151 (33.1%) |
| Polypragmasia, n (%) | 1464 | 719 (49.1%) |
| Orthostatic hypotension, n (%) | 1053 | 294 (28%) |
| Aging-associated diseases, n (%) | 1638 | 1,570 (95.8%) |
| Cancer, n (%) | 1606 | 113 (7%) |
| Cardiovascular diseases, n (%) | 1611 | 694 (43.1%) |
| Diabetes mellitus, n (%) | 1612 | 231 (14.3%) |
| COPD, n (%) | 1610 | 234 (14.5%) |
| Sarcopenia, n (%) | 1450 | 1,321 (91.1%) |
| Participants who died in the course of the study, n (%) | 1641 | 552 (33.6%) |
| Participants infected by COVID-19, n (%) | 347 | 21.1% |
| Participants who died from COVID-19, n (% of those who were infected by COVID-19) | 113 | 32.6% |
Most participants were affected by at least one GS at the time of enrollment; $90\%$ of them suffered from frailty. The design of this study did not require random sampling; therefore, we cannot provide a specific estimate of the prevalence of GSs in the older adults in Moscow. Nonetheless, we can safely say that there was a positive correlation between age and the number of GSs in this region [11].
Figure 1 shows that the main causes of death in the older adults were cerebrovascular accidents ($28.7\%$) and cardiovascular diseases ($19.6\%$). Another major cause of mortality was COVID-19. This number, however, could be even higher due to COVID-19 complications, but which might have been classified as resulting from cardiovascular diseases.
## All-cause mortality
We investigated the associations between all-cause mortality and the geriatric assessment and blood test results.
## Associations between the geriatric syndromes and all-cause mortality
We found statistically significant associations between mortality and GSs, including malnutrition, ADL and IADL, frailty, frontal lobe dysfunction, a high risk of falls, depression, cognitive impairment, and aging-related diseases.
Table 3 presents the significant associations between the GSs and mortality, adjusted for sex, age, and multiple testing. See Supplementary Table S2 presents all associations between the GSs and mortality.
**Table 3**
| Age, sex and GS | N† (alive/dead) | In alive (n, % from the alive) | In deceased (n, % from the dead) | OR | CC | P-value | ROC AUC (95% CI) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Age | 1,641 (1,089/552) | 92 [91; 94] (mean: 92.51) | 92 [91;94] (mean: 92.81) | 1.13 [1.02; 1.25] | 0.12 | 0.02 | 0.50 [0.41, 0.56] |
| Sex | 1,641 (1,089/552) | m: 1089 (257, 23,6%)w: 1089 (832, 76,4%) | m: 552 (150, 27,2%)w: 552 (402, 72,8%) | 0.92 [0.83; 1.02] | −0.08 | 0.11 | 0.52 [0.48, 0.56] |
| Depression | 1,584 (1,068/516) | 484 (45.3%) | 280 (54.3%) | 1.04 [1.02; 1.06] | 0.17 | 0.001 | 0.55 [0.49, 0.62] |
| Malnutrition /risk of malnutrition | 1,425 (975/450) | 193 (19.8%) 626 (64.2%) | 138 (30.7%) 269 (59.8%) | 1.07 [1.04; 1.09] | 0.3 | <0.0001 | 0.59 [0.53, 0.66] |
| Cognitive impairment | 1,542 (1,031/511) | 488 (47.3%) | 338 (66.1%) | 1.09 [1.06; 1.11] | 0.4 | <0.0001 | 0.61 [0.54, 0.65] |
| Frontal lobe dysfunction | 1,564 (1,049/515) | 752 (71.7%) | 446 (86.6%) | 1.08 [1.07; 1.13] | 0.4 | <0.0001 | 0.60 [0.53, 0.66] |
| Dependence in ADL | 1,593 (1,061/532) | 954 (89.9%) | 509 (95.7%) | 1.05 [1.03; 1.09] | 0.2 | <0.0001 | 0.57 [0.50, 0.63] |
| Dependence in IADL | 1,599 (1,065/534) | 988 (92.8%) | 523 (97.9%) | 1.05 [1.03; 1.10] | 0.3 | <0.0001 | 0.56 [0.51, 0.61] |
| Frailty | 1,639 (1,088/551) | 922 (84.7%) | 516 (93.6%) | 1.06 [1.04; 1.10] | 0.3 | <0.0001 | 0.57 [0.51, 0.63] |
| AADs | 1,638 (1,087/551) | 1,026 (94.4%) | 544 (98.7%) | 1.05 [1.03; 1.10] | 0.3 | <0.0001 | 0.55 [0.49, 0.61] |
Cognitive impairment was the main contributor to mortality in the older adults. It even surpassed frailty, which is conventionally viewed as an unsuccessful aging phenotype, and aging-associated diseases.
## Association between the test results and mortality
The multivariate regression analysis, the results of which were adjusted for sex, age, and multiple testing revealed significant associations between the test results and mortality (Table 4). All associations are presented in Supplementary Table S3.
**Table 4**
| Test | In alive (n = 1,089)median [Q1; Q3] | In deceased (n = 552)median [Q1; Q3] | OR | CC (normalized) | p-value | ROC AUC (95% CI) |
| --- | --- | --- | --- | --- | --- | --- |
| Hemoglobin, g/dL | 12.6 [11.3; 13.7] | 12.2 [11; 13.5] | 0.90 [0.85; 0.96] per unit of measure | −0.20 | 0.0004 | 0.56 [0.50, 0.62] |
| МСHС, g/dL | 33.2 [32.4; 34] | 32.9 [31.9; 33.8] | 0.86 [0.80; 0.93] per unit of measure | −0.22 | <0.0001 | 0.58 [0.52, 0.65] |
| RDW, % | 13.9 [13.2; 14.8] | 14.25 [13.3; 15.5] | 1.13 [1.07; 1.18] per unit of measure | 0.27 | <0.0001 | 0.58 [0.51, 0.63] |
| Total protein, g/L | 70 [66; 75] | 69 [64; 73] | 0.96 [0.95; 0.98] per unit of measure | −0.26 | <0.0001 | 0.57 [0.50, 0.63] |
| Albumin, g/L | 39.2 [35.8; 41.925] | 36.8 [33.3; 40] | 0.90 [0.88; 0.92] per unit of measure | −0.51 | <0.0001 | 0.64 [0.59, 0.68] |
| α1-globulin, g/L | 3.1 [2.8; 3.4] | 3.3 [3; 3.7] | 1.83 [1.54; 2.17] per unit of measure | 0.39 | <0.0001 | 0.60 [0.54, 0.66] |
| Total cholesterol, mmol/L | 4.93 [4.11; 5.72] | 4.72 [3.9; 5.6] | 0.86 [0.79; 0.94] per unit of measure | −0.19 | 0.0007 | 0.55 [0.50, 0.60] |
| HDL, mmol/L | 1.3 [1.1; 1.6] | 1.18 [0.96; 1.4] | 0.33 [0.24; 0.45] per unit of measure | −0.40 | <0.0001 | 0.60 [0.56, 0.65] |
| AC | 2.7 [2.1; 3.4] | 2.9 [2.2; 3.8] | 1.18 [1.08; 1.29] per unit of measure | 0.20 | 0.0001 | 0.56 [0.51, 0.62] |
| ALT, U/L | 12 [9; 15] | 11 [8; 15] | 0.98 [0.97; 0.99] per unit of measure | −0.20 | 0.0001 | 0.53 [0.47, 0.58] |
| GGT, U/L | 18 [14; 29] | 17 [12; 27] | 0.97 [0.94; 0.98] per 10 units of measure | −0.18 | 0.0008 | 0.54 [0.49, 0.59] |
| hsCRP, mg/L | 2.62 [1.31; 6.77] | 3.99 [1.83; 10.71] | 1.02 [1.01; 1.02] per unit of measure | 0.34 | <0.0001 | 0.58 [0.52, 0.64] |
| Free T3, pmol/L | 3.7 [3.3; 4] | 3.5 [2.98; 3.9] | 0.57 [0.49; 0.67] per unit of measure | −0.40 | <0.0001 | 0.61 [0.56, 0.66] |
| 25(ОН)D, ng/mL | 8 [6; 13] | 6 [5; 9] | 0.95 [0.92; 0.95] per unit of measure | −0.47 | <0.0001 | 0.62 [0.57, 0.68] |
| Insulin, μU/mL | 6.9 [4.7; 11.35] | 5.9 [3.7; 9.4] | 0.98 [0.92; 0.95] per unit of measure | −0.28 | <0.0001 | 0.56 [0.48, 0.62] |
| Leptin, ng/mL | 12.85 [4.8; 29.4] | 6.9 [2.82; 16.6] | 0.98 [0.98; 0.99] per unit of measure | −0.41 | <0.0001 | 0.60 [0.53, 0.69] |
| IGF-1, ng/mL | 104.2 [82.6; 135.6] | 90.55 [69.15; 121.1] | 0.91 [0.89; 0.94] per 10 units of measure | −0.37 | <0.0001 | 0.60 [0.54, 0.66] |
| Cystatin C, mg/L | 1.74 [1.51; 2.06] | 1.82 [1.56;2.15] | 1.44 [1.19; 1.78] per unit of measure | 0.20 | 0.0003 | 0.55 [0.49, 0.61] |
| NT-proBNP, pg./mL | 566.5 [282; 1149.5] | 785.5 [381; 1878.75] | 1.27 [1.17; 1.37] (by 100 times) | 0.32 | <0.0001 | 0.57 [0.50, 0.63] |
| BMI, kg/m2 | 25.7 [23.37; 28.9] | 24.8 [22.2; 27.9] | 0.96 [0.93; 0.98] per unit of measure | −0.19 | 0.0007 | 0.55 [0.50, 0.61] |
The results show that the most significant predictor of mortality was the turnover of protein and iron. Interestingly, mortality was significantly associated with the association between mortality and HDL and AC but not with LDL. The levels of hsCRP and α1-globulin suggested an association between mortality and inflammation. Based on the hsCRP levels, the participants had low-grade, not acute inflammation. Table 4 presents the median values and quartiles 1 and 3 for all significant associations.
It is worth mentioning that α1-globulin was highly predictive of mortality: its increase by 1 g/L resulted in a mortality OR of 1.83.
Thus, the most significant biochemical mortality predictors were the markers of inflammation, nutritional deficiency, and lower levels of 25(ОН)D.
## Predictive model of 1-year mortality
Based on the logistic regression analysis, we selected the variables with significant associations with mortality and used them as input. Figure 2 show the importance of each variable. The most important variables were albumin (0.09), C-reactive protein (0.07), HDL (0.06), leptin (0.06), IGF-1 (0.05), NT-proBNT (0.05), cystatin C (0.05), free Т3 (0.05), α1-globulin (0.05), and MCHC. Supplementary Table S4 provides a complete list of significant associations.
**Figure 2:** *Assessment of the relative importance of the input variables, based on the Gini coefficient. The Gini coefficients are absolute values. The relative importance is indicated by the deviation from point zero (the Gini coefficient); the direction and color of the bars show the effect on the risk of all-cause mortality: the red indicates a lowering risk with an increasing coefficient; the blue—an increasing risk with an increasing coefficient. AC, atherogenic coefficient; ADL, activity of daily-living; ALT, alanine transaminase; free T3, free triiodothyronine; GGT, gamma-glutamyl transferase; HDL, high-density lipoprotein cholesterol; IADL, instrumental activity of daily-living; IGF-1, insulin-like growth factor 1; MCHC, mean corpuscular hemoglobin concentration; NT-proBNP, N-terminal pro b-type natriuretic peptide; RDW, red cell distribution width.*
The highest prognostic accuracy was achieved at 100 trees; a random subset of three features, with a maximum tree depth of 5; a minimum of one sample; and weight values of {0: 1, 1: 3,7}.
Figure 3 shows the ROC-curve of the random forest model. ROC AUC was 0.68 ($95\%$CI 54.7–76.0). To find the point of maximum accuracy, we built an F-measure diagram for a threshold and obtained an F-score of 0.76.
**Figure 3:** *ROC-curve of the predictive model based on the randomized forest algorithm.*
Thus, the most important features were the markers of protein metabolism, inflammation, cardiac failure, glucose and lipid metabolism, and thyroid function. Despite a relatively low AUC, the model has a good f-score and, after additional validation, could be used in medical decision-making, such as evaluating drug load and other preventive strategies for the oldest-old. Notably, age was not an important feature; the most important features were modifiable parameters.
## COVID-19-related mortality
The present study coincided with the COVID_19 pandemic. It was an epidemiologically difficult period for Russia, with a sharp rise in COVID-19 cases in the spring and summer of 2020 [12].
We separately analyzed the COVID-19-related mortality in the long-living adults using the above statistical methods. The results showed that COVID-19 contributed to an increased rate of mortality rate observed in this study. We found that 347 participants had been infected with COVID-19; 113 of them had died from it. Hence, the COVID-19-related mortality rate was $32.6\%$. It should be mentioned that many participants could have had mild or asymptomatic COVID-19: many studies have reported that from 23 to $54\%$ of people have no disease symptoms (13–15). Therefore, an increased all-cause mortality rate in our study could be due to mild/asymptomatic infection and its complications, including thrombosis, or other factors contributing to increased mortality in older individuals [16].
Figure 4 shows the COVID-19-related mortality dynamics in the entire Moscow population and in the study participants from Moscow from April 2020 to October 2021.
**Figure 4:** *COVID-19-related mortality dynamics in the entire Moscow population and in the study participants from Moscow from April 2020 to October 2021. The blue line: officially reported COVID-19-related mortality in Moscow (17); the red line: COVID-19-related mortality in the long-living individuals.*
Mortality in the long-living individuals peaked in the spring–summer of 2020 and the fall–winter of 2020–2021, despite the fact that this portion of the population lived in a relative isolation. However, this trend was consistent with an overall increase in all-cause mortality and COVID-19 waves. Notably, a marked difference between the mortality rates. However, we should take into consideration the imposed restrictions and large sample size.
## Associations between geriatric syndromes and COVID-19-related mortality
Detailed distribution of GSs and levels of biochemical markers are presented in the Supplementary Material.
Table 5 shows the GSs with a significantly different distribution of the recovered and deceased participants. The mortality rate was higher in those who had suffered from cognitive impairment and frailty. These results partially match the results on all-cause mortality (see paragraph 1.1., Results). However, malnutrition was not a significant predictor of COVID-19-related mortality, despite its significant association with all-cause mortality. All associations between the GS and COVID-19-related mortality are presented in Supplementary Table S5.
**Table 5**
| Age, sex, GS | Recovered from COVID-19 (n = 234)N (% from all recovered) | Infected with COVID-19, died (n = 113)N (% from all deceased) | OR | CC | p-value | ROC AUC (95% CI) |
| --- | --- | --- | --- | --- | --- | --- |
| Age | 92 [91; 94] (mean: 92.6) | 92 [91; 94] (mean: 92.76) | 1.07 [0.86; 1.34] | 0.07 | 0.53 | 0.47 [0.36, 0.58] |
| Sex | m: 70 (29,9%)w: 164 (70,1%) | m: 42 (37,2%)w: 71 (62,8%)w: 1089 (832, 76,4%) | 0.86 [0.69; 1.07] | −0.15 | 0.18 | 0.53 [0.43, 0.64] |
| Depression | 96 (41%) | 60 (53.1%) | 1.007 [1.002; 1.013] | 0.3 | 0.01* | 0.56 [0.45, 0.68] |
| Frontal lobe dysfunction | 170 (72.6%) | 89 (78.8%) | 1.007 [1.001; 1.013] | 0.3 | 0.02* | 0.57 [0.45, 0.69] |
| Frailty | 223 (95.3%) | 111 (98.2%) | 1.006 [1.0; 1.013] | 0.3 | 0.04* | 0.54 [0.44, 0.65] |
## Associations between the test results and COVID-19-related mortality
The results of the laboratory analysis also varied.
We did not find associations between COVID-19-related mortality rate and levels of cystatin C, ferritin, and neutrophiles, which were associated with all-cause mortality in the entire cohort. However, we found significant associations with the following markers: GGT, insulin, glycohemoglobin, ALT, and lymphocytes (Table 6).
**Table 6**
| Test | Recovered (n = 234)Median [Q1; Q3] | Deceased (n = 113)Median [Q1; Q3] | OR | CC | P-value | ROC AUC (95% CI) |
| --- | --- | --- | --- | --- | --- | --- |
| WBC, cells × 109/L | 5.65 [4.74; 7.05] | 6.54 [5.08; 7.69] | 1.05 [1.0; 1.14] per unit of measure | 0.3 | 0.03 | 0.49 [0.34, 0.63] |
| RDW, % | 13.8 [13.1; 14.72] | 14.2 [13.45; 15.05] | 1.12 [1.01; 1.25] per unit of measure | 0.3 | 0.03 | 0.54 [0.45, 0.67] |
| Neutrophils, cells × 109/L | 3.2 [2.52; 4.15] | 3.69 [2.96; 4.76] | 1.17 [1.06; 1.4] per unit of measure | 0.3 | 0.006 | 0.58 [0.47, 0.70] |
| Cystatin C, mg/L | 1.75 [1.55; 2.07] | 1.85 [1.61; 2.2] | 1.77 [1.13; 2.85] per unit of measure | 0.3 | 0.01 | 0.57 [0.45, 0.70] |
| NT-proBNP, pg./mL | 505 [237; 1,090] | 775 [376.5; 1532.5] | 1.3 [1.1; 1.54] by 100 times | 0.4 | 0.002 | 0.59 [0.48, 0.69] |
| IGF-1, ng/mL | 112.5 [83.38; 138.93] | 98.6 [77.1; 134] | 0.99 [0.99; 0.1] per unit of measure | −0.3 | 0.03 | 0.55 [0.44, 0.65] |
| α1-globulin, g/L | 3.1 [2.8; 3.4] | 3.3 [2.9; 3.6] | 1.82 [1.25; 2.66] per unit of measure | 0.4 | 0.002 | 0.61 [0.48, 0.75] |
| HDL, mmol/L | 1.26 [1.08; 1.55] | 1.16 [0.96; 1.43] | 0.42 [0.22; 0.81] per unit of measure | −0.3 | 0.01 | 0.57 [0.44, 0.70] |
| Free T3 | 3.7 [3.3; 4] | 3.5 [3.1; 4] | 0.67 [0.49; 0.94] per unit of measure | −0.3 | 0.02 | 0.57 [0.44, 0.67] |
| 25(OH)D, ng/mL | 8 [6; 13] | 7 [5.5; 10] | 0.95 [0.92; 0.98] per unit of measure | −0.4 | 0.001 | 0.59 [0.46, 0.71] |
All associations are presented in Supplementary Table S6.
## Discussion
With a sharp decline in the rate of non-senescent mortality, the overwhelming majority of deaths are now caused by aging. This trend could be partially attributed to the quality of life, access to health care and other socioeconomic factors. Today, $15.8\%$ of Russians, or one out of seven, is an older than 65 years of age [18], compared with $15.5\%$ in early 2020. Population aging entails economic, budgetary, and health care implications. If this trend continues, the number of people of working age might decrease drastically. Therefore, for timely screening and disease prevention as part of ambulatory care, it is crucial to identify the causes of mortality, health risks, and protective factors in older adults and long-living individuals. Presently, the study of aging has emerged as a new and promising trend. However, long-living individuals—the most abundant source of biological data—are under-examined. The present study focused on this growing population group in Russia.
The geriatric assessment and analysis showed that mortality in the long-living cohort was associated with cognitive dysfunction of any cause, frailty, malnutrition, depression, functional disability, and comorbidity; COVID-19-related mortality was associated with depression, frontal lobe dysfunction, and frailty. Many authors have reported associations between mortality and various cognitive dysfunctions (including dementia) in the youngest-and middle-old [19, 20]; however, reports on the associations in the oldest-old, or long-living adults, are scarce. In the German longitudinal six-year-long study, the authors used the Cognitive Telephone Screening Instrument (COGTEL) to assess cognitive functioning in people over 70 years of age. The results showed that the mortality rate in the subjects with low COGTEL scores were $60\%$ higher than in those with higher COGTEL scores, especially in men [20].
Interestingly, some studies consider cognitive functioning dynamics to be another predictor of mortality, along with cognitive impairment [21]. A Chinese study of older adults (mean = 82 years) showed a $75\%$ higher mortality rate in the subjects who had demonstrated a more rapid decline in cognitive functioning measured by the MMSE. However, this association was more marked in the youngest old (under 80) and those who had initially scored higher [22].
Cognitive decline and its association with a high risk of mortality could sometimes result from cerebrovascular diseases [19]. Reduced cerebral perfusion inevitably leads to chronic cerebrovascular ischemia that often affects cognitive functions, such as memory. Hence, cognitive disorders have proven to be a sensitive marker of clinical outcome in the oldest-old. Wang et al. [ 23] found associations between COVID-19-related mortality and dementia: the results showed that the patients with AD, not only vascular dementia, were at a significantly higher risk of death. Depression often accompanies dementia. In our study, depression in the participants who had recovered from COVID-19 was associated with a higher risk of mortality; pre- COVID-19 depression was a risk factor for COVID-19-related mortality. Our results are consistent with the results of a study conducted in a smaller cohort of younger older adults [24]. These results substantiate the need for early prevention and screening for cognitive dysfunctions and monitoring cognitive functioning and mood in older adults.
We found that malnutrition was another significant, yet modifiable, risk factor for mortality, along with the laboratory markers of inadequate nutrition. The prognostic model demonstrated that its contribution was comparable to that of the age of the participants, unlike the above-described syndromes. We feel obligated to once again stress the high prevalence of this syndrome in older adults, even though malnutrition has been mentioned in many other studies as a prevalent condition associated with poor outcome (25–28). Long-living individuals, despite their exceptional characteristics (and probably, due to these characteristics), are often one of the most disadvantaged groups lacking access to some of the basic things, such as information and social and economic support. Risk of malnutrition and the effect of diagnosed malnutrition on poor outcome were clearly shown by routine examination/tests results. BMI, insulin and leptin levels were negatively correlated with the risk of mortality. Previous studies have shown that a slightly higher BMI could be a protective factor [29, 30], in older adults, probably, because aging is often accompanied by emaciation. Statistically significant were the correlations between mortality and such biochemical indicators as protein and iron turnovers (total protein, albumin, hemoglobin levels and mean cell hemoglobin concentration, and the size of red blood cells). Unfortunately, these results demonstrate inadequate testing of older individuals for these basic parameters, despite the long-standing discussions on the increased need for protein-rich food in older people [31, 32]. Hence, proper screening for malnutrition and the risk of malnutrition with a follow-up nutritional therapy should be put in place. It is safe to assume that there is no need to abide by a strict range of glucose metabolism and BMI, since they could be protective at slightly increased levels. However, further research is needed to establish the target levels of these markers in the advanced age.
We also found significant associations between mortality and ADL, IADL, and frailty. These syndromes have been associated with COVID-19-related mortality in a number of studies (33–35). Many GSs could not be treated in people over 90; however, the data on their association with mortality can aid in identifying those at risk of COVID-19-related death, performing detailed diagnosis and developing a more effective treatment strategy.
We found that the level of 25(ОН)D was a vital factor in the long-living individuals. This association has been described in younger older adults. In February 2022, DeJaeger et al. [ 36] published their study of 1915 men aged 49 to 74 with a follow-up of about 12 years. Their results showed that 25(ОН)D deficiency doubled the risk of mortality. In our study, most participants were very deficient in 25(ОН)D; however, in those who died from COVID-19 and other causes, a critical deficiency in 25(ОН)D had been a risk factor for mortality. This indicates another potential therapeutic target and serves as the evidence of inadequate geriatric care. The therapeutic benefit of 25(ОН)D was tested in many studies during the pandemic. Oristrell et al. [ 37] reported that patients supplemented with 25(ОН)D until achieving 25OHD levels ≥30 ng/ml were at a lower risk of lower risk of SARS-CoV2 infection and severe COVID-19. Our results confirm that 25(ОН)D supplementation can be beneficial even in people over 90.
The long-living individuals who died within a year since the beginning of the study had been affected by many aging-associated diseases, had elevated levels of cystatin C, GGT, N-terminal proBNP, and total cholesterol. Those who died from COVID-19 had higher while blood cell and neutrophile counts, while those who died from other causes—higher apha-1-globulin and hsCRP levels. For the most part, all these parameters were within the normal range; however, their increase in the cohort of more vulnerable individuals indicated raised inflammatory markers. Therefore, comorbidity and inflammaging are directly associated with not only all-cause mortality but also with COVID-19 -related mortality, including in long-living individuals [38, 39]. Increased levels of free T3 free in our study contributed to a higher survival rate. There is evidence to suggest that hyperthyroidism in an advanced age is more dangerous than hypothyroidism [40]. Its levels in most participants were within the normal range which could be the reason we did not observe this association in our study.
Low IGF-1 was associated with a higher risk of mortality. The contribution of IGF-1 to longevity is still unclear. On the one hand, long-living individuals demonstrated low IGF-1, which could be genetic [41]. On the other hand, other studies also demonstrated the association between low IGF-1 and mortality [42]. This is yet another evidence of the delicate “equilibrium” observed in long-living individuals. Promising strategies to expand life expectancy in the oldest-old are shown in the Figure 5.
**Figure 5:** *Promising strategies to expand life expectancy in the oldest-old. BMI, body mass index; GSs, geriatric syndromes; hsCRP, high-sensitivity C-reactive protein; NT-proBNP, N-terminal pro b-type natriuretic peptide.*
The present study has its limitations. First, all participants were from Moscow and the Moscow region; therefore, the results could not be extrapolated to the entire country. Second, COVID-19 status could not be established for a number of participants, which might have rendered the COVID-19 group (of those who contracted the disease or died from it) incomplete.
Despite the above-mentioned limitations, our study clearly demonstrated insufficiencies in geriatric care even in the regions with high-quality healthcare. Improved geriatric care could aid in expanding the active period of life, particularly in the oldest-old. Moreover, our results showed that biological, not chronological, age takes the lead in determining health, even in an advanced age. Therefore, prevention of aging as a complex phenomenon can facilitate a solution of socioeconomic and health care problems.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by the Local Ethics Committee of the Russian Gerontological Research and Clinical Center (Protocol No 30, December 24, 2019). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
DK, VE, MG, AAA, and IS: conceptualization ideas. AR, ES, and MI: data curation. ES, MI, and AT: formal analysis and software. AAA, IS, IK, AR, and DK: investigation. DK, VE, MG, and VY: methodology. VY, VM, OT, SK, and SY: project administration. AAA, IS, IK, and AIA: resources. DK, OT, VY, SK, and SY: supervision. DK, VE, MG, and AY: validation. ES, MI, MT, and VE: visualization. DK, VE, MG, ES, AY, MI, LM, and AR: writing—original draft. DK, VE, MG, ES, MI, MT, AR, VY, and LM: writing—review and editing. All authors have read and agreed to the published version of the manuscript.
## Funding
This study was funded by the Centre for Strategic Planning and Management of Biomedical Health Risks.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmed.2023.1132476/full#supplementary-material
## References
1. Ortonobes Roig S, Soler-Blanco N, Torrente Jiménez I, Van den Eynde OE, Moreno-Ariño M, Gómez-Valent M. **Clinical and pharmacological data in COVID-19 hospitalized nonagenarian patients**. *Rev Esp Quimioter* (2021) **34** 145-50. DOI: 10.37201/req/130.2020
2. Marcon G, Tettamanti M, Capacci G, Fontanel G, Spanò M, Nobili A. **COVID-19 mortality in Lombardy: the vulnerability of the oldest old and the resilience of male centenarians**. *Aging* (2020) **12** 15186-95. DOI: 10.18632/aging.103872
3. Lee SJ, Go AS, Lindquist K, Bertenthal D, Covinsky KE. **Chronic conditions and mortality among the oldest old**. *Am J Public Health* (2008) **98** 1209-14. DOI: 10.2105/AJPH.2007.130955
4. Formiga F, Ferrer A, Chivite D, Pujol R. **Survival after 7 years of follow-up at ninety. The NonaSantfeliu study**. *Eur J Intern Med* (2011) **22** e164-5. DOI: 10.1016/j.ejim.2011.07.001
5. Tiainen K, Luukkaala T, Hervonen A, Jylhä M. **Predictors of mortality in men and women aged 90 and older: a nine-year follow-up study in the vitality 90+ study**. *Age Ageing* (2013) **42** 468-75. DOI: 10.1093/ageing/aft030
6. Nybo H, Petersen HC, Gaist D, Jeune B, Andersen K, McGue M. **Predictors of mortality in 2,249 nonagenarians—the Danish 1905-cohort survey**. *J Am Geriatr Soc* (2003) **51** 1365-73. DOI: 10.1046/j.1532-5415.2003.51453.x
7. Zou C, Zhou Y, Dong B, Hao Q, Chen S, Zhou J. **Predictors of 49-month mortality in Chinese nonagenarians and centenarians in PLAD study**. *Aging Clin Exp Res* (2015) **27** 821-7. DOI: 10.1007/s40520-015-0355-y
8. Pancani S, Lombardi G, Sofi F, Gori AM, Boni R, Castagnoli C. **12-month survival in nonagenarians inside the Mugello study: on the way to live a century**. *BMC Geriatr* (2022) **22** 194. DOI: 10.1186/s12877-022-02908-9
9. Tahira AC, Verjovski-Almeida S, Ferreira ST. **Dementia is an age-independent risk factor for severity and death in COVID-19 inpatients**. *Alzheimers Dement* (2021) **17** 1818-31. DOI: 10.1002/alz.12352
10. 10.Consultant Plus. Clinical guidelines on frailty. The Ministry of Health of the Russian Federation. (2023). Available: https://www.consultant.ru/document/cons_doc_LAW_369879/?ysclid=ldlsvzsm78101522039. *Clinical guidelines on frailty. The Ministry of Health of the Russian Federation* (2023)
11. Ostapenko VS, Runikhina NK, Sharashkina NV. **Prevalence of frailty and its correlation with chronic non-infectious diseases among outpatients in Moscow**. *Rus J Ger Med* (2020) **10** 131-7. DOI: 10.37586/2686-8636-2-2020-131-137
12. 12.Coronavirus COVID-19: Official information about coronavirus in Russia. The Government of the Russian Federation. (2022) Available online at: https://xn--80aesfpebagmfblc0a.xn--p1ai/. *Official information about coronavirus in Russia. The Government of the Russian Federation* (2022)
13. Alshami A, Alattas R, Anan H, Alhalimi A, Alfaraj A, Al QH. **Silent disease and loss of taste and smell are common manifestations of SARS-COV-2 infection in a quarantine facility: Saudi Arabia**. *PLoS One* (2020) **15** e0241258. DOI: 10.1371/journal.pone.0241258
14. Al-Qahtani M, AlAli S, AbdulRahman A, Salman AA, Otoom S, Atkin SL. **The prevalence of asymptomatic and symptomatic COVID-19 in a cohort of quarantined subjects**. *Int J Infect Dis* (2021) **102** 285-8. DOI: 10.1016/j.ijid.2020.10.091
15. Li Y, Shi J, Xia J, Duan J, Chen L, Yu X. **Asymptomatic and symptomatic patients with non-severe coronavirus disease (COVID-19) have similar clinical features and Virological courses: a retrospective single center study**. *Front Microbiol* (2020) **11** 11. DOI: 10.3389/fmicb.2020.01570
16. Perrotta F, Corbi G, Mazzeo G, Boccia M, Aronne L, D’Agnano V. **COVID-19 and the elderly: insights into pathogenesis and clinical decision-making**. *Aging Clin Exp Res* (2020) **32** 1599-608. DOI: 10.1007/s40520-020-01631-y
17. 17.Territory of the Russian Federation—The Russian Government. Operational headquarters to prevent the importation and spread of a new coronavirus infection on the territory of the Russian Federation—The Government of the Russian Federation. (2022). Available online at: http://government.ru/department/556/events. *Operational headquarters to prevent the importation and spread of a new coronavirus infection on the territory of the Russian Federation—The Government of the Russian Federation* (2022)
18. 18.SiteSoft. The population of the Russian Federation by sex and age. Federal State Statistics Service. (2022). Available online at: https://rosstat.gov.ru/compendium/document/13284. *The population of the Russian Federation by sex and age. Federal State Statistics Service* (2022)
19. Hassing LB, Johansson B, Berg S, Nilsson SE, Pedersen NL, Hofer SM. **Terminal decline and markers of cerebro- and cardiovascular disease: findings from a longitudinal study of the oldest old**. *J Gerontol B Psychol Sci Soc Sci* (2002) **57** P268-76. DOI: 10.1093/geronb/57.3.P268
20. Perna L, Wahl H-W, Mons U, Saum K-U, Holleczek B, Brenner H. **Cognitive impairment, all-cause and cause-specific mortality among non-demented older adults**. *Age Ageing* (2015) **44** 445-51. DOI: 10.1093/ageing/afu188
21. Bortone I, Zupo R, Castellana F, Aresta S, Lampignano L, Sciarra S. **Motoric cognitive risk syndrome, subtypes and 8-year all-cause mortality in aging phenotypes: the Salus in Apulia study**. *Brain Sci* (2022) **12** 861. DOI: 10.3390/brainsci12070861
22. Lv X, Li W, Ma Y, Chen H, Zeng Y, Yu X. **Cognitive decline and mortality among community-dwelling Chinese older people**. *BMC Med* (2019) **17** 63. DOI: 10.1186/s12916-019-1295-8
23. Wang Y, Li M, Kazis LE, Xia W. **Clinical outcomes of COVID-19 infection among patients with Alzheimer’s disease or mild cognitive impairment**. *Alzheimers Dement* (2022) **18** 911-23. DOI: 10.1002/alz.12665
24. Bayrak M, Çadirci K. **The associations of life quality, depression, and cognitive impairment with mortality in older adults with COVID-19: a prospective, observational study**. *Acta Clin Belg* (2022) **77** 588-95. DOI: 10.1080/17843286.2021.1916687
25. Feng L, Chu Z, Quan X, Zhang Y, Yuan W, Yao Y. **Malnutrition is positively associated with cognitive decline in centenarians and oldest-old adults: a cross-sectional study**. *eClinicalMedicine* (2022) 101336. DOI: 10.1016/j.eclinm.2022.101336
26. Song Y, Liu M, Jia W-P, Han K, Wang S-S, He Y. **The association between nutritional status and functional limitations among centenarians: a cross-sectional study**. *BMC Geriatr* (2021) **21** 1-8. DOI: 10.1186/s12877-021-02312-9
27. Zupo R, Castellana F, Bortone I, Griseta C, Sardone R, Lampignano L. **Nutritional domains in frailty tools: working towards an operational definition of nutritional frailty**. *Ageing Res Rev* (2020) **64** 101148. DOI: 10.1016/j.arr.2020.101148
28. Zupo R, Castellana F, Guerra V, Donghia R, Bortone I, Griseta C. **Associations between nutritional frailty and 8-year all-cause mortality in older adults: the Salus in Apulia study**. *J Intern Med* (2021) **290** 1071-82. DOI: 10.1111/joim.13384
29. Pes GM, Licheri G, Soro S, Longo NP, Salis R, Tomassini G. **Overweight: a protective factor against comorbidity in the elderly**. *Int J Environ Res Public Health* (2019) **16** 3656. DOI: 10.3390/ijerph16193656
30. Donini LM, Pinto A, Giusti AM, Lenzi A, Poggiogalle E. **Obesity or BMI paradox? Beneath the tip of the iceberg**. *Front Nutr* (2020) **7** 53. DOI: 10.3389/fnut.2020.00053
31. Coelho-Júnior HJ, Rodrigues B, Uchida M, Marzetti E. **Low protein intake is associated with frailty in older adults: a systematic review and meta-analysis of observational studies**. *Nutrients* (2018) **10** 1334. DOI: 10.3390/nu10091334
32. Krok-Schoen JL, Archdeacon Price A, Luo M, Kelly OJ, Taylor CA. **Low dietary protein intakes and associated dietary patterns and functional limitations in an aging population: a NHANES analysis**. *J Nutr Health Aging* (2019) **23** 338-47. DOI: 10.1007/s12603-019-1174-1
33. Fagard K, Gielen E, Deschodt M, Devriendt E, Flamaing J. **Risk factors for severe COVID-19 disease and death in patients aged 70 and over: a retrospective observational cohort study**. *Acta Clin Belg* (2022) **77** 487-94. DOI: 10.1080/17843286.2021.1890452
34. Sablerolles RSG, Lafeber M, van Kempen JAL, van de Loo BPA, Boersma E, Rietdijk WJR. **Association between clinical frailty scale score and hospital mortality in adult patients with COVID-19 (COMET): an international, multicentre, retrospective, observational cohort study**. *Lancet Healthy Longev* (2021) **2** e163-70. DOI: 10.1016/S2666-7568(21)00006-4
35. Ramos-Rincon J-M, Moreno-Perez O, Pinargote-Celorio H, Leon-Ramirez J-M, Andres M, Reus S. **Clinical frailty score vs hospital frailty risk score for predicting mortality and other adverse outcome in hospitalized patients with COVID-19: Spanish case series**. *Int J Clin Pract* (2021) **75** e14599. DOI: 10.1111/ijcp.14599
36. Dejaeger M, Antonio L, Bouillon R, Moors H, Wu FCW, O’Neill TW. **Aging men with insufficient vitamin D have a higher mortality risk: no added value of its free fractions or active form**. *J Clin Endocrinol Metab* (2022) **107** e1212-20. DOI: 10.1210/clinem/dgab743
37. Oristrell J, Oliva JC, Casado E, Subirana I, Domínguez D, Toloba A. **Vitamin D supplementation and COVID-19 risk: a population-based, cohort study**. *J Endocrinol Investig* (2022) **45** 167-79. DOI: 10.1007/s40618-021-01639-9
38. Sabbatinelli J, Matacchione G, Giuliani A, Ramini D, Rippo MR, Procopio AD. **Circulating biomarkers of inflammaging as potential predictors of COVID-19 severe outcomes**. *Mech Ageing Dev* (2022) **204** 111667. DOI: 10.1016/j.mad.2022.111667
39. Müller L, Di Benedetto S. **How Immunosenescence and Inflammaging may contribute to Hyperinflammatory syndrome in COVID-19**. *Int J Mol Sci* (2021) **22** 2539. DOI: 10.3390/ijms222212539
40. Barbesino G. **Thyroid function changes in the elderly and their relationship to cardiovascular health: a mini-review**. *GER* (2019) **65** 1-8. DOI: 10.1159/000490911
41. Vitale G, Pellegrino G, Vollery M, Hofland LJ. **ROLE of IGF-1 system in the modulation of longevity: controversies and new insights from a centenarians’ perspective**. *Front Endocrinol* (2019) **10** 27. DOI: 10.3389/fendo.2019.00027
42. Sanders JL, Guo W, O’Meara ES, Kaplan RC, Pollak MN, Bartz TM. **Trajectories of IGF-I predict mortality in older adults: the cardiovascular health study**. *J Gerontol A Biol Sci Med Sci* (2018) **73** 953-9. DOI: 10.1093/gerona/glx143
|
---
title: 'Intravenous dexamethasone administration during anesthesia induction can improve
postoperative nutritional tolerance of patients following elective gastrointestinal
surgery: A post-hoc analysis'
authors:
- Feng Tian
- Xinxiu Zhou
- Junke Wang
- Mingfei Wang
- Zhou Shang
- Leping Li
- Changqing Jing
- Yuezhi Chen
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC10018170
doi: 10.3389/fnut.2023.1093662
license: CC BY 4.0
---
# Intravenous dexamethasone administration during anesthesia induction can improve postoperative nutritional tolerance of patients following elective gastrointestinal surgery: A post-hoc analysis
## Abstract
### Aim
To investigate the effect of intravenous dexamethasone administration on postoperative enteral nutrition tolerance in patients following gastrointestinal surgery.
### Methods
Based on the previous results of a randomized controlled study to explore whether intravenous administration of dexamethasone recovered gastrointestinal function after gastrointestinal surgery, we used the existing research data from 1 to 5 days post operation in patients with enteral nutrition tolerance and nutrition-related analyses of the changes in serum indices, and further analyzed the factors affecting resistance to enteral nutrition.
### Result
The average daily enteral caloric intake was significantly higher in patients receiving intravenous administration of dexamethasone during anesthesia induction than in controls (8.80 ± 0.92 kcal/kg/d vs. 8.23 ± 1.13 kcal/kg/d, $$P \leq 0.002$$). Additionally, intravenous administration of 8 mg dexamethasone during anesthesia induction can reduce the changes in postoperative day (POD) 3, POD5, and preoperative values of serological indices, including ΔPA, ΔALB, and ΔRBP ($P \leq 0.05$). In the subgroup analysis, dexamethasone significantly increased the average daily enteral nutrition caloric intake in patients undergoing enterotomy (8.98 ± 0.87 vs. 8.37 ± 1.17 kcal/kg/d, $$P \leq 0.010$$) or in female patients (8.94 ± 0.98 vs. 8.10 ± 1.24 kcal/kg/d, $$P \leq 0.019$$). The changes of serological indexes (ΔPA, ΔALB, and ΔRBP) in the dexamethasone group were also significantly different on POD3 and POD5 ($P \leq 0.05$). In addition, multivariate analysis showed that dexamethasone use, surgical site, and age might influence enteral nutrition caloric tolerance.
### Conclusion
Postoperative enteral nutrition tolerance was significantly improved in patients receiving intravenous administration of dexamethasone during anesthesia induction, especially in patients following enterotomy surgery, with significant improvements in average daily enteral caloric intake, PA levels, ALB levels, and RBP levels.
### Clinical trial registration
http://www.chictr.org.cn, identifier: ChiCTR1900024000.
## 1. Introduction
Trauma and stress caused by surgery can lead to a catabolic state. Studies have shown that patients can lose ~2 kg of body weight during recovery, even after uncomplicated elective surgery [1, 2]. Postoperative malnutrition is more common in approximately $40\%$ of patients undergoing gastrointestinal surgery due to inflammatory reactions, gastrointestinal dysfunction, and loss of gastrointestinal reserve function [3, 4].
Nutritional deficiency after gastrointestinal surgery is considered to be one of the important risk factors for postoperative complications and morbidity [5, 6], which may not only increase the length of hospital stay (LOS) and treatment cost but also affect the survival of cancer patients due to delayed adjuvant therapy after operation (7–10).
For decades, clinicians have been trying to improve the prognosis of surgical patients by reducing complications caused by nutritional deficiencies. Although enhanced recovery after surgery (ERAS) protocols and preoperative administration of oral nutritional supplements (ONSs) can improve the nutritional status of patients, some patients undergoing abdominal surgery suffer from malnutrition (11–14). Therefore, improving nutritional status as soon as possible after gastrointestinal surgery is particularly important. Postoperative stress in some patients leads to gastrointestinal motility dysfunction and intolerance to enteral nutrition, which limits the recovery of early gastrointestinal function. Data from one of our previous studies, the effect of dexamethasone on postoperative gastrointestinal motility (DOPGM) trial [15] concluded that a single intravenous dose of 8 mg dexamethasone at anesthesia induction significantly decreased the time to return of flatus improved abdominal distension at 72 h, and promoted tolerance of a liquid diet. However, our study did not calculate the average daily enteral nutritional energy tolerance during the intervention period. In addition, it also raised an important new problem that was not adequately addressed in the preplanned analysis: since there is no difference in LOS and quality of life (QoL), will this secondary outcome affect its application in real life? Different indicators reflect the clinical significance of various aspects. LOS and QoL might not be the only indicators for evaluating the applicability of dexamethasone in real-life clinical settings. Moreover, although the LOS and QoL did not improve significantly, the patient's time to first flatus and tolerance of a liquid diet was shortened. Abdominal distension was reduced at 72 h after surgery, which may improve the postoperative nutritional intake and nutritional status such as pre-albumin (PA), albumin (ALB), and retinol-binding protein (RBP), among others.
In this post-hoc analysis of the DOPGM trial, we analyzed the changes in postoperative indicators related to nutritional status between the two groups with random intervention to verify the hypothesis of whether a single intravenous dose of 8 mg dexamethasone at induction of anesthesia can improve postoperative enteral nutrition tolerance and nutritional status in the patients undergoing elective gastrointestinal surgery.
## 2.1. Patients and methods
This is a post hoc analysis of DOPGM, a prospective, double-blind, single-center, and randomized controlled trial carried out in the Department of Gastroenterology, Shandong Provincial Hospital, China. The study design, ethical approval, inclusion criteria, and procedures have been previously reported [15].
After obtaining informed consent, the 126 patients were randomized into two groups. One group received 8 mg of intravenous dexamethasone during the induction of anesthesia, and the other group received normal saline. All patients underwent standardized general anesthesia and elective gastrointestinal surgery. Our main aim was to assess the effects of preoperative dexamethasone administration on patient outcomes in terms of postoperative enteral nutrition tolerance. All 126 patients were included, whose PA, ALB, hemoglobin (Hb), lymphocyte count (LC), RBP, and fasting blood glucose (FBG) levels were measured preoperatively and on postoperative days (PODs) 1, 3, and 5 were recorded as part of the clinical routine. Postoperative energy and protein requirements were estimated according to the European Society of Clinical Nutrition and Metabolism (ESPAN) guidelines [16]. The energy requirement was calculated according to 30 kcal/kg of body weight, while the protein requirement was 1.5 g/kg of body weight after the operation. On the first postoperative day, all patients started consuming clear liquids via oral or tube feeding. We considered the patient to be tolerant of the liquid diet if there were no reports of nausea, vomiting, or significant abdominal distention after an intake of 200 ml of clear liquid. The clear liquid diet was gradually adjusted to enteral nutrition (Abbott, Ensure, 1.06 kcal/ml) on the second postoperative day. According to our department's routine management process for enteral nutrition supplements after gastrointestinal surgery, we set that enteral nutrition provided $20\%$ of the total target caloric intake from the second postoperative day and increased it by $10\%$ daily. The rest of the caloric intake was supplied by parenteral nutrition. When enteral nutrition met $60\%$ of the total caloric requirement, parenteral nutrition was stopped. Researchers have previously recorded the actual amount of daily enteral nutrition. The patients recorded the type and amount of the diet. The caloric and protein contents of the food were recorded according to the China Food Composition Tables [Yang [17]] so that the actual caloric and protein intakes on PODs 1–5 were recorded.
## 2.2. Outcome measures
The average daily enteral nutrition caloric intake and serum indices PA, ALB, RBP, LC, FBG, and the changes between preoperative and postoperative serum index values (including ΔPA, ΔALB, ΔRBP, ΔLC, ΔFBG, etc.) were compared between the two groups to evaluate the enteral nutrition tolerance and nutritional status of the patients.
## 2.3. Statistical analysis
Statistical analyses were performed using IBM SPSS Statistics 26.0. Normally distributed continuous variables are reported as mean and standard deviation, and an independent sample t-test was used to compare the differences between the treatment and control groups. Categorical variables are presented as numbers and analyzed using the χ2 or Fisher's exact test, as appropriate. Linear regression analysis was used for univariate and multivariate analyses. Two-sided P-values were reported where necessary, with the significance level set at $P \leq 0.05.$ A $95\%$ confidence interval was used for all statistical analyses. Bar graphs and forest graphs were generated using GraphPad Prism 7.0.4.
## 3.1. There was no difference in preoperative baseline among the 126 patients
In total, 126 participants completed the initial intervention. There were no significant demographic differences between the two groups. Compared with the control group, the preoperative RBP value was slightly lower in the dexamethasone group. There were no significant differences in the other indices. The baseline characteristics of the 126 participants included in the analysis are shown in Table 1.
**Table 1**
| Unnamed: 0 | Dexamethasone (n = 64) | Control (n = 62) | P-value |
| --- | --- | --- | --- |
| Age (mean ± SD) | 60.77 ± 12.63 | 61.06 ± 11.13 | 0.888 |
| Gender | Gender | Gender | 0.489 |
| Female | 20 (31.3%) | 23 (37.1%) | |
| Male | 44 (68.8%) | 39 (62.9%) | |
| BMI | 24.74 ± 3.45 | 24.12 ± 3.34 | 0.341 |
| Site of surgery | Site of surgery | Site of surgery | 0.808 |
| Enterotomy | 41 (64.1%) | 41 (66.1%) | |
| Gastrectomy | 23 (35.9%) | 21 (33.9%) | |
| Serum indices | Serum indices | Serum indices | Serum indices |
| Hb (g/L) | 126.44 ± 20.93 | 129.08 ± 20.02 | 0.475 |
| RBP (μg/L) | 32.20 ± 10.13 | 37.02 ± 12.96 | 0.022 |
| FBG (mmol/L) | 5.48 ± 1.02 | 5.28 ± 1.048 | 0.310 |
| ALB (g/L) | 38.85 ± 4.02 | 40.19 ± 4.60 | 0.084 |
| PA (mg/L) | 212.41 ± 63.32 | 231.28 ± 65.49 | 0.103 |
| LC (109/L) | 1.79 ± 0.67 | 1.83 ± 0.74 | 0.798 |
| nrs 2002 score | nrs 2002 score | nrs 2002 score | 0.827 |
| < 3 | 55 (85.9%) | 55 (88.7%) | |
| ≥3 | 9 (14.1%) | 7 (11.3%) | |
## 3.2. Patients in the dexamethasone group had better tolerance to enteral nutrition after surgery
Postoperative average daily caloric intake through enteral nutrition was significantly higher in the dexamethasone group than in the control group (8.80 ± 0.92 vs. 8.23 ± 1.13 kcal/kg/d, $$P \leq 0.002$$; Table 2 and Figure 1). With regard to caloric intake through enteral nutrition for each postoperative day, the dexamethasone group was higher than the control group on POD 2–4 ($P \leq 0.05$), and there was no difference in POD 5 ($$P \leq 0.086$$). The results of subgroup analysis showed that dexamethasone significantly increased the average daily enteral nutrition caloric intake in patients undergoing enterotomy (8.98 ± 0.87 vs. 8.37 ± 1.17 kcal/kg/d, $$P \leq 0.010$$; Table 3) or in female patients (8.94 ± 0.98 vs. 8.10 ± 1.24 kcal/kg/d, $$P \leq 0.019$$; Table 4). However, no significant differences were found between the subgroup of gastrostomy surgery patients (8.48 ± 0.94 vs. 7.95 ± 1.02 kcal/kg/d, $$P \leq 0.083$$; Figure 2). Enteral nutrition intake with dexamethasone was significantly higher in male patients than in controls, but this did not reach statistical significance (8.73 ± 0.90 vs. 8.31 ± 1.07 kcal/kg/d, $$P \leq 0.052$$; Figure 2). Subgroup analysis revealed that dexamethasone significantly improved tolerance to enteral nutrition in female patients and undergoing enterotomy (Figure 3).
## 3.3. The decline in nutrition-related indices after surgery was smaller in the dexamethasone group
Compared with the control group, the dexamethasone group showed fewer changes in nutrition-related indices, such as ΔPA, ΔALB, and ΔRBP, on POD 3 and POD 5 [Figure 4; ΔPA: POD 3, 60.36 mg/L vs. 86.01, $95\%$ CI (−45.28, −6.02), $$P \leq 0.11$$; POD 5, 48.64 vs. 74.42 mg/L, $95\%$ CI (−47.72, −3.84), $$P \leq 0.022$$; ΔALB:POD 3, 3.00 vs. 4.52 g/L, $95\%$ CI (−2.99, −0.05), $$P \leq 0.043$$; POD 5, 1.57 vs. 3.43 g/L, $95\%$ CI (−3.51, −0.22), $$P \leq 0.027$$; ΔRBP: POD 3, 9.78 vs. 13.58 μg/L, $95\%$ CI (−7.18, −0.41), $$P \leq 0.028$$; POD 5, 6.02 vs. 11.02 μg/L, $95\%$ CI (−8.85, −1.15), $$P \leq 0.011$$]. Moreover, the results of the subgroup analysis showed that in patients undergoing enterotomy surgery, dexamethasone can reduce the declining level of PA and ALB values on the POD 3 [ΔPA: 61.08 vs. 85.01 mg/L, $95\%$ CI (−47.00, −0.86), $$P \leq 0.042$$; ΔALB:2.65 vs. 4.33g/L, $95\%$ CI (−3.31, −0.05), $$P \leq 0.044$$] and the declining level of RBP values both on the POD3 and POD 5 [ΔRBP: POD3, 9.50 vs. 14.29 μg/L, $95\%$ CI (−9.08, −0.51), $$P \leq 0.029$$; POD 5, 6.54 vs. 12.02 μg/L, $95\%$ CI (−10.25, −0.72), $$P \leq 0.025$$; Figure 5]. Similarly, in a subgroup analysis of female patients, the dexamethasone group had reduced changes in PA value on the POD 3 [ΔPA: 42.78 vs. 74.99 mg/L, $95\%$ CI (−63.78, −0.65), $$P \leq 0.046$$; Figure 6]. However, such changes did not reach statistical significance in male patients or patients undergoing gastrectomy surgery.
**Figure 4:** *The decrease of PA, ALB, RBP in POD3 and POD5. Δ represents the difference between preoperative value and postoperative value. *P < 0.05.* **Figure 5:** *The decrease of PA, ALB, RBP in POD3 and POD5 in enterotomy surgery group. Δ represents the difference between preoperative value and postoperative value. ns P > 0.05, *P < 0.05.* **Figure 6:** *The decrease of PA in POD3 and POD5 in female patients. Δrepresents the difference between preoperative value and postoperative value. ns P > 0.05, *P < 0.05.*
## 3.4. Influencing factors of average daily enteral nutrition caloric intake
Univariate linear regression analysis of factors affecting the average daily caloric intake through enteral nutrition showed that surgical site, age, and intravenous dexamethasone might affect the average daily enteral nutrition caloric intake (Table 5). Multivariate linear regression analysis showed that surgical site, age, and intravenous dexamethasone use might be significant predictors of daily average enteral nutrition energy intake. With increasing age, the degree of enteral nutrition tolerance decreased (Table 6).
## 4. Discussion
In our study, PA, ALB, RBP, FBG, and LC were used to evaluate nutrition-related indicators, which was consistent with previous studies. These indicators have been used to evaluate the nutritional status of patients in previous studies (18–21). In this post-hoc analysis of prospectively collected data from the DOPGM trial, we observed that the average daily caloric intake through enteral nutrition was significantly higher in the dexamethasone group than in the control group. This may be related to the faster recovery of intestinal function and faster tolerance of a liquid diet in the dexamethasone group. However, differences in daily enteral nutrient caloric intake are shown on POD 2–4. As time goes on, this difference between the two groups will no longer be statistically significant on POD 5. In terms of nutrition-related serological indices, there was no statistical difference between the two groups. Still, we found that the decline in nutrition-related indices after surgery, such as ΔPA, ΔALB, and ΔRBP, reached statistical significance on POD 3 and POD 5. These results suggest that 8 mg single-dose intravenous dexamethasone can improve postoperative nutritional status in patients with short-term nutritional status. Subgroup analysis showed that the average daily caloric intake through enteral nutrition was higher in patients undergoing enterotomy surgery than in the control group. However, in patients undergoing gastrectomy, the average daily caloric intake through enteral nutrition in the dexamethasone group did not show obvious advantages, which may be related to the longer duration of gastric surgery, greater surgical traumatic stress, longer time to a liquid diet, longer time to gastrointestinal function and motility recovery, and poor enteral nutrition tolerance. This may also be the reason for the lower decline in nutrition-related indices after enterotomy. Interestingly, female patients in the dexamethasone group also showed a similar change. However, we have not found in previous studies that after being given dexamethasone, females haves a better recovery of intestinal function after surgery than males. This may be due to intestinal flora differences between male and female patients with enteral nutrition absorption. Thus, further research is still needed to determine the reasons for the difference in caloric absorption caused by sex differences.
With the change in treatment mode and the popularization of the ERAS concept, the perioperative fasting time and surgical stress have been reduced in recent years. Although these measures improve the nutritional status of patients after major surgery, there are some still suffer from postoperative malnutrition, which is associated with poor postoperative outcomes. These include an increased incidence of infections, depression of the immune system, impaired wound healing, and increased mortality [22]. Gastrointestinal dysfunction is an important factor that affects nutritional absorption after surgery. Early enteral feeding is particularly important to reduce surgical stress and the risk of postoperative complications caused by malnutrition and insufficient feeding, especially for patients who have nutritional risks before surgery or require gastrointestinal surgery [23, 24]. Our previous studies have shown that preoperative intravenous dexamethasone can promote faster recovery of gastrointestinal function and better tolerance to a liquid diet. Meanwhile, this post-hoc analysis study showed that treatment with dexamethasone could improve short-term postoperative nutritional status. These findings strongly support the idea that preoperative dexamethasone administration can improve patients' postoperative recovery.
Meanwhile, inflammation could be another key factor in explaining these outcomes [25]. Surgery is a type of trauma that can cause a series of reactions, including releasing stress hormones and inflammatory mediators. In severe cases, it can even cause the so-called “systemic inflammatory response syndrome,” which significantly impacts metabolism [26, 27]. In addition, previous studies have shown that inflammation can affect the nutritional support of patients in different ways [28, 29], such as affecting appetite and gastrointestinal function, reducing food intake, and increasing insulin resistance [30]. At the cellular level, cytokines such as IL-6 interfere with the satiety center, leading to anorexia, delayed gastric emptying, and skeletal muscle protein catabolism [31]. In contrast, previous studies have shown that dexamethasone significantly reduces IL-6 levels [32]. Prevention of nausea and vomiting and reduction of pain may have been another reason for the increased food intake in the dexamethasone group [33]. Whether additional administration could promote the recovery of gastrointestinal function to improve nutritional status after gastrectomy still requires further prospective trials.
Correlation analysis showed that dexamethasone administration was an important predictor of the average daily enteral nutrition intake. This may be related to the reduction of intestinal stress and the promotion of gastrointestinal peristalsis. In addition, it was reported that a patient with esophageal cancer cachexia was treated with dexamethasone combined with nutritional drugs, and his nutritional status was significantly improved and he could tolerate chemotherapy [34]. This may be due to the fact that corticosteroids such as dexamethasone can inhibit brain edema and improve appetite on the one hand, and stimulate the expression of neuropeptide y and prevent the synthesis of promelanocortin on the other hand, leading to increased appetite and hunger, thereby reducing the application of parenteral nutrition and improving the tolerance of enteral nutrition [35]. This finding is consistent with our previous findings. The increase in age, the increase in basic diseases, the decline in various body functions, and the use of anesthetics and antibiotics significantly impact the recovery of gastrointestinal peristalsis in the elderly, and the tolerance of enteral nutrition in the elderly decreases. Jang and Jeong [36] concluded in an analysis of early nutritional tolerance after gastrectomy: age (≥70 years), gender, tumor obstruction and operation time are related to poor tolerance of enteral nutrition, and male and tumor obstruction are independent influencing factors of poor tolerance. Therefore, age negatively correlates with the average daily tolerance to enteral nutrition.
## 5. Strengths and limitations
This post hoc analysis was based on the random nature of previous clinical trials, which ensured the balance of data between the two groups. However, this study has some limitations. First, we did not monitor cytokines such as IL-6, which may provide more detailed information. Second, the sample size of this experiment may be too small to find significant interactions in some research results. Finally, because this is a post-hoc analysis, our results are based on the study hypothesis of the first trial; therefore, further randomized controlled trials with independent samples are needed to verify the tolerance of enteral nutrition.
## 6. Conclusion
In a post hoc analysis of a previous clinical trial involving dexamethasone, we found that dexamethasone improved postoperative enteral nutrition tolerance, particularly in a subgroup of patients following enterotomy surgery, as well as significantly improved postoperative average daily enteral nutritional caloric intake and changes in nutrition-related serological indicators.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by Shandong Provincial Hospital Ethics Committee. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
FT, XZ, and JW: analysis and interpretation, literature search, and writing manuscript. MW and ZS: materials, data collection, and processing. LL: design. YC and CJ: supervision, critical review, and funding acquisition. All authors have read and agreed to the published version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Schricker T, Meterissian S, Donatelli F, Carvalho G, Mazza L, Eberhart L. **Parenteral nutrition and protein sparing after surgery: do we need glucose?**. *Metabolism.* (2007) **56** 1044-50. DOI: 10.1016/j.metabol.2007.03.013
2. Phillips BE, Smith K, Liptrot S, Atherton PJ, Varadhan K, Rennie MJ. **Effect of colon cancer and surgical resection on skeletal muscle mitochondrial enzyme activity in colon cancer patients: a pilot study**. *J Cachexia Sarcopenia Muscle.* (2013) **4** 71-7. DOI: 10.1007/s13539-012-0073-7
3. Bistrian BR, Blackburn GL, Vitale J, Cochran D, Naylor J. **Prevalence of malnutrition in general medical patients**. *JAMA.* (1976) **235** 1567-70. DOI: 10.1001/jama.235.15.1567
4. Zhang H, Wang Y, Mu SY, Jiang ZM, Shi D, Yu K. **Prevalence of nutritional risks, malnutrition and application of nutritional support rates at one Chongqing teaching hospital**. *Zhonghua Yi Xue Za Zhi.* (2012) **92** 3417-9. PMID: 23327702
5. Choi WJ, Kim J. **Nutritional care of gastric cancer patients with clinical outcomes and complications: a review**. *Clin Nutr Res.* (2016) **5** 65-78. DOI: 10.7762/cnr.2016.5.2.65
6. Yu W, Seo BY, Chung HY. **Postoperative body-weight loss and survival after curative resection for gastric cancer**. *Br J Surg.* (2002) **89** 467-70. DOI: 10.1046/j.0007-1323.2001.02046.x
7. Weimann A, Braga M, Harsanyi L, Laviano A, Ljungqvist O, Soeters P. **ESPEN guidelines on enteral nutrition: surgery including organ transplantation**. *Clin Nutr.* (2006) **25** 224-44. DOI: 10.1016/j.clnu.2006.01.015
8. Plauth M, Cabre E, Campillo B, Kondrup J, Marchesini G, Schutz T. **ESPEN guidelines on parenteral nutrition: hepatology**. *Clin Nutr.* (2009) **28** 436-44. DOI: 10.1016/j.clnu.2009.04.019
9. Tokunaga M, Tanizawa Y, Bando E, Kawamura T, Terashima M. **Poor survival rate in patients with postoperative intra-abdominal infectious complications following curative gastrectomy for gastric cancer**. *Ann Surg Oncol.* (2013) **20** 1575-83. DOI: 10.1245/s10434-012-2720-9
10. Jiang N, Deng JY, Ding XW, Zhang L, Liu HG, Liang YX. **Effect of complication grade on survival following curative gastrectomy for carcinoma**. *World J Gastroenterol.* (2014) **20** 8244-52. DOI: 10.3748/wjg.v20.i25.8244
11. Bozzetti F, Gianotti L, Braga M, Carlo VDi, Mariani L. **Postoperative complications in gastrointestinal cancer patients: the joint role of the nutritional status and the nutritional support**. *Clin Nutr.* (2007) **26** 698-709. DOI: 10.1016/j.clnu.2007.06.009
12. Fujitani K, Tsujinaka T, Fujita J, Miyashiro I, Imamura H, Kimura Y. **Prospective randomized trial of preoperative enteral immunonutrition followed by elective total gastrectomy for gastric cancer**. *Br J Surg.* (2012) **99** 621-9. DOI: 10.1002/bjs.8706
13. Klek S, Sierzega M, Szybinski P, Szczepanek K, Scislo L, Walewska E. **Perioperative nutrition in malnourished surgical cancer patients - a prospective, randomized, controlled clinical trial**. *Clin Nutr.* (2011) **30** 708-13. DOI: 10.1016/j.clnu.2011.07.007
14. Gustafsson UO, Scott MJ, Schwenk W, Demartines N, Roulin D, Francis N. **Guidelines for perioperative care in elective colonic surgery: enhanced recovery after surgery (ERAS(R)) Society recommendations**. *Clin Nutr.* (2012) **31** 783-800. DOI: 10.1016/j.clnu.2012.08.013
15. Chen Y, Dong C, Lian G, Li D, Yin Y, Yu W. **Dexamethasone on postoperative gastrointestinal motility: a placebo-controlled, double-blinded, randomized controlled trial**. *J Gastroenterol Hepatol.* (2020) **35** 1549-54. DOI: 10.1111/jgh.15020
16. Weimann A, Braga M, Carli F, Higashiguchi T, Hubner M, Klek S. **ESPEN guideline: clinical nutrition in surgery**. *Clin Nutr.* (2017) **36** 623-50. DOI: 10.1016/j.clnu.2017.02.013
17. Yang YX. *China Food Composition Tables: Standard Edition* (2018)
18. **Modalities for assessing the nutritional status in patients with diabetes and cancer**. *Diabetes Res Clin Pract.* (2018) **142** 162-72. DOI: 10.1016/j.diabres.2018.05.039
19. Smith SH. **Using albumin and prealbumin to assess nutritional status**. *Nursing.* (2017) **47** 65-6. DOI: 10.1097/01.NURSE.0000511805.83334.df
20. Winkler MF, Gerrior SA, Pomp A, Albina JE. **Use of retinol-binding protein and prealbumin as indicators of the response to nutrition therapy**. *J Am Diet Assoc.* (1989) **89** 684-7. DOI: 10.1016/S0002-8223(21)02227-6
21. Zhang Y, Zhang M, Yang X, Wang H. **Significance of retinol binding protein and prealbumin in neonatal nutritional evaluation**. *Pak J Pharm Sci.* (2018) **31** 1613-6. PMID: 30203747
22. Yeh DD, Fuentes E, Quraishi SA, Cropano C, Kaafarani H, Lee J. **Adequate nutrition may get you home: effect of caloric/protein deficits on the discharge destination of critically ill surgical patients**. *JPEN J Parenter Enteral Nutr.* (2016) **40** 37-44. DOI: 10.1177/0148607115585142
23. Horowitz M, Neeman E, Sharon E, Ben-Eliyahu S. **Exploiting the critical perioperative period to improve long-term cancer outcomes**. *Nat Rev Clin Oncol.* (2015) **12** 213-26. DOI: 10.1038/nrclinonc.2014.224
24. Gustafsson UO, Oppelstrup H, Thorell A, Nygren J, Ljungqvist O. **Adherence to the ERAS protocol is associated with 5-year survival after colorectal cancer surgery: a retrospective cohort study**. *World J Surg.* (2016) **40** 1741-7. DOI: 10.1007/s00268-016-3460-y
25. Merker M, Gomes F, Stanga Z, Schuetz P. **Evidence-based nutrition for the malnourished, hospitalised patient: one bite at a time**. *Swiss Med Wkly.* (2019) **149** w20112. DOI: 10.4414/smw.2019.20112
26. Gillis C, Carli F. **Promoting perioperative metabolic and nutritional care**. *Anesthesiology.* (2015) **123** 1455-72. DOI: 10.1097/ALN.0000000000000795
27. Alazawi W, Pirmadjid N, Lahiri R, Bhattacharya S. **Inflammatory and immune responses to surgery and their clinical impact**. *Ann Surg.* (2016) **264** 73-80. DOI: 10.1097/SLA.0000000000001691
28. Schuetz P. **Food for thought: why does the medical community struggle with research about nutritional therapy in the acute care setting?**. *BMC Med.* (2017) **15** 38. DOI: 10.1186/s12916-017-0812-x
29. Schuetz P. **“Eat your lunch!” - controversies in the nutrition of the acutely, non-critically ill medical inpatient**. *Swiss Med Wkly.* (2015) **145** w14132. DOI: 10.4414/smw.2015.14132
30. Felder S, Braun N, Stanga Z, Kulkarni P, Faessler L, Kutz A. **Unraveling the link between malnutrition and adverse clinical outcomes: association of acute and chronic malnutrition measures with blood biomarkers from different pathophysiological states**. *Ann Nutr Metab.* (2016) **68** 164-72. DOI: 10.1159/000444096
31. Kuhlmann MK, Levin NW. **Potential interplay between nutrition and inflammation in dialysis patients**. *Contrib Nephrol.* (2008) **161** 76-82. DOI: 10.1159/000129759
32. Zargar-Shoshtari K, Sammour T, Kahokehr A, Connolly AB, Hill AG. **Randomized clinical trial of the effect of glucocorticoids on peritoneal inflammation and postoperative recovery after colectomy**. *Br J Surg.* (2009) **96** 1253-61. DOI: 10.1002/bjs.6744
33. **Dexamethasone versus standard treatment for postoperative nausea and vomiting in gastrointestinal surgery: randomised controlled trial (DREAMS Trial)**. *BMJ.* (2017) **357** j1455. DOI: 10.1136/bmj.j1455
34. Stefani S, Andayani DE. **Nutritional medical therapy in cachexia patient with oesophageal adenocarcinoma metastases on dexamethasone therapy: A case report**. *J Pak Med Assoc.* (2021) **71** S143-5. PMID: 33785960
35. Liu L, Xu S, Wang X, Jiao H, Zhao J, Lin H. **Effect of dexamethasone on hypothalamic expression of appetite-related genes in chickens under different diet and feeding conditions**. *J Anim Sci Biotechnol.* (2016) **7** 23. DOI: 10.1186/s40104-016-0084-x
36. Jang A, Jeong O. **Tolerability of early oral nutrition and factors predicting early oral nutrition failure after gastrectomy**. *Clin Nutr.* (2020) **39** 3331-6. DOI: 10.1016/j.clnu.2020.02.019
|
---
title: Excessive consumption of mucin by over-colonized Akkermansia muciniphila promotes
intestinal barrier damage during malignant intestinal environment
authors:
- Shuang Qu
- Yinghui Zheng
- Yichun Huang
- Yicheng Feng
- Kunyao Xu
- Wei Zhang
- Yawen Wang
- Kaili Nie
- Meng Qin
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC10018180
doi: 10.3389/fmicb.2023.1111911
license: CC BY 4.0
---
# Excessive consumption of mucin by over-colonized Akkermansia muciniphila promotes intestinal barrier damage during malignant intestinal environment
## Abstract
Gut microbiota disorders damage the intestinal barrier, which causes intestinal disease. Thus, we screened the microbiota with significant changes using an in situ malignant colorectal cancer (CRC) model. Among the colonies with increased abundance, *Akkermansia muciniphila* (A. muciniphila) is known for its characteristic of breaking down mucin, which is an essential component of the intestinal barrier. The role of A. muciniphila remains controversial. To investigate the effect of excess A. muciniphila on the intestinal barrier, we established an over-colonized A. muciniphila mouse model by administering a live bacterial suspension after disrupting the original gut microbiome with antibiotics. The results showed that over-colonization of A. muciniphila decreased intestinal mucin content. The mRNA and protein expression levels of tight junction proteins also decreased significantly in the over-colonized A. muciniphila mouse model. Our findings reveal that excess colonization by A. muciniphila breaks the dynamic balance between mucin secretion and degradation, reduces the thickness of the intestinal mucus layer, and damages the intestinal barrier, which would eventually aggravate the development of colitis and CRC. These results will raise awareness about the safety of A. muciniphila serving as a probiotic.
## Introduction
A strong association has been reported between the gut microbiota and various diseases, particularly intestinal diseases, including inflammatory bowel diseases (IBD), such as Crohn’s disease; (Palm et al., 2014), ulcerative enteritis (Kummen et al., 2017), and irritable bowel syndrome (Liu et al., 2017), as well as colorectal cancer (CRC; Coker et al., 2022). Among them, CRC has high incidence and mortality rates worldwide (Sung et al., 2021). Since metagenomics next-generation sequencing technology has been developed, researchers are discovering the relationships between the gut microbiota and intestinal diseases (Rajagopala et al., 2017). The complex and extensive microbiota that colonizes the gut is an essential part of the intestinal contents and plays a crucial role in many physiological processes, including immunological control and the regulation of digestion and metabolism (Schultz et al., 2017; Zheng et al., 2020). Accordingly, dysbiosis of the gut microbiota is a major contributor to intestinal diseases (Mahalhal et al., 2018; Wong and Yu, 2019; Lee, 2021). The gut microbiota change in quantity and composition with changes in food intake, lifestyle, use of medications, and other factors, which raises the risk of intestinal disease. The intestinal barrier, which stops invasion by pathogenic microorganisms, includes the gut microbiota as the key component. The intestinal barrier system is damaged by dysbiosis of the gut microbiota, excess colonization of pathogenic microbiota, excess consumption of mucus, and thinning of the intestinal mucus layer, which expose the intestinal epithelial cells to the microbial environment (Van der Sluis et al., 2006; Johansson et al., 2008; Kim and Ho, 2010). Inflammation develops from bacteria directly contacting epithelial cells and is exacerbated by the toxic byproducts produced by pathogenic bacteria. Inflammation is a major contributor to intestinal disease and CRC.
The intestinal mucus layer is mainly composed of mucins. The degradation of mucins is an adverse factor that may lead to disturbances in the host intestinal environment (Paone and Cani, 2020). The intestinal goblet cells continuously secrete mucin, which provides colonic bacteria with a plentiful energy supply to survive (Ouwehand et al., 2005). Some intestinal microbiota have a slow degradation effect on mucin, such as *Akkermansia muciniphila* (A. muciniphila; Sicard et al., 2017). Akkermansia muciniphila is a Gram-negative, strictly anaerobic bacterium that inhabits the human intestine, and is a representative genus in Verrucomicrobia (Derrien et al., 2004). Akkermansia muciniphila uses mucin as its sole source of carbon and nitrogen, and produces 61 enzymes that destroy mucin (Derrien et al., 2004). A decrease in the abundance of A. muciniphila was thought to be associated with obesity, diabetes mellitus type 2, nonalcoholic fatty liver disease, and cardiovascular diseases (Everard et al., 2013; Zhang et al., 2013; Everard et al., 2014; Li et al., 2017). However, studies on A. muciniphila in colitis and CRC are controversial, as A. muciniphila promotes mucin secretion and reduces intestinal permeability through extracellular vesicles (Chelakkot et al., 2018). It has also been shown that A. muciniphila has beneficial effects on intestinal barrier function during intestinal inflammation. For example, A. muciniphila helps to restore intestinal barrier function in DSS-induced colitis, improve the symptoms of colitis, and repair intestinal barrier damage (Kang et al., 2013; Zhai et al., 2019; Wang et al., 2020). However, depleting the mucus layer may be exacerbated by the consumption of mucin by A. muciniphila, leading to thinning of the mucus layer and increased inflammation. Akkermansia muciniphila interferes with reconstruction of the intestinal mucosa and aggravates the symptoms of colitis caused by *Salmonella typhimurium* (Ganesh et al., 2013). Other studies have demonstrated that A. muciniphila acts as a pathobiont to promote colitis in a genetically susceptible host. Our previous study reported an overgrowth of A. muciniphila in DSS-induced acute ulcerative colitis (Huang et al., 2022). However, it remains unclear whether this is compensatory regulation of the intestinal microbiota or a pathogenic effect of A. muciniphila. In addition, when intestinal inflammation progresses to CRC, the intestinal barrier is completely disrupted, and colonization of A. muciniphila is rarely reported at that time.
In this study, we explored changes in the intestinal microbiota of an in situ malignant CRC model and investigated the effect of over-colonized A. muciniphila on the intestinal barrier.
## Reagents
CT26 cells were purchased from Procell Life Science & Technology Co., Ltd. (Wuhan, China). Roswell Park Memorial Institute (RPMI) 1640 medium and fetal bovine serum (FBS) were purchased from Gibco (Shanghai, China). Ampicillin sodium salt, neomycin sulfate, metronidazole, and vancomycin HCl were purchased from Macklin Biochemical Co., Ltd. (Shanghai, China). Akkermansia muciniphila was obtained from the BeNa Culture Collection (Henan Province, China). The thioglycollate medium was purchased from Qingdao Hope Bio-Technology Co., Ltd. (Qingdao, China).
## Animals
The animal experiments used 20–22 g male BALB/C mice provided by the Beijing HFK Bioscience Co., Ltd. (Beijing, China). All mice were raised under standard laboratory conditions with a $\frac{12}{12}$ h dark/light cycle and free access to water and food. All procedures were carried out following the Care and Use of Laboratory Animals (license number: 2022D019).
## Cell culture and the in situ malignant CRC mouse model
CT26 cells (CVCL_7254) were cultured in RPMI 1640 medium supplemented with $10\%$ FBS and $1\%$ penicillin and streptomycin (PS) and were incubated at 37°C in $5\%$ CO2. The cells were subcultured in trypsin-EDTA ($0.25\%$), and a single-cell suspension was obtained at a concentration of 107 cells/mL. The suspension was injected into the left axilla of four mice in a volume of 100 μL per mouse. Subcutaneous tumors of about 1 cm in diameter grew after 10 days. The subcutaneous tumor-bearing mice were euthanized, their skin was disinfected, and the subcutaneous tumor was peeled off, and immediately immersed in saline containing 100 U/mL PS. The flesh-like tissues of the actively growing tumors were cut into 1 mm3 pieces.
The mice were randomly divided into the control group (CTL group, $$n = 5$$), the colonic tumor model group (CRC group, $$n = 7$$), and the sham surgery group (SHAM group, $$n = 5$$). Mice in the CRC group were anesthetized with inhaled isoflurane and placed in the supine position. An incision of about 1 cm was made in the center of the abdomen, and the cecum was carefully removed. The local plasma membrane at the junction between the cecum and colon was scraped off with a needle, a prepared tumor block was glued into the scratch with 3 M medical tissue glue, the external intestinal segment was backfilled into the abdominal cavity, and the wound was sutured. The entire process was performed aseptically. The mice were routinely housed in an SPF-grade animal laboratory for 15 days after awakening.
## Akkermansia muciniphila culture and the over-colonized Akkermansia muciniphila mouse model
Liquid cultures of A. muciniphila for oral gavage were grown in fluid thioglycollate medium in an anaerobic chamber. The 5–7 days cultures were harvested, centrifuged at 4,000 × g for 10 min, and resuspended in sterile saline.
Fifteen mice were randomly divided into three groups ($$n = 5$$) of the control group (CTL group), the antibiotic cocktail group (1 g/L ampicillin, 1 g/L metronidazole, 0.5 g/L vancomycin, and 0.5 g/L neomycin, AMVN group), and the A. muciniphila + antibiotic cocktail group (AKK group). Ten mice were orally treated once daily for 2 days with 200 μL of an antibiotic cocktail before transplant. They were randomly assigned to the AMVN group or the AKK group. Mice in the AKK group were administered 400 μL of A. muciniphila (1 × 109 CFU in sterile saline) once daily for 7 days, while mice in the CTL and AMVN groups were given equal doses of sterile saline. The weights of the mice were recorded daily. Fecal samples, blood, and serum were collected on the day after the last gavage. Then, all mice were humanly euthanized, and the colonic and fecal samples were collected. The stool, partial colonic tissue, and serum were stored at −80°C and the blood was stored at 4°C until analysis. Partial colonic segments were stored in tissue fixative (Wuhan Servicebio Technology Co., Ltd.) at 4°C for further procedures.
## Sectioning and staining of the colonic tissue
The intact colon was removed from the euthanized mice, and a 1 cm section of colonic tissue was excised about 2 cm from the anus and placed in $4\%$ paraformaldehyde for fixation. After the tissue was dehydrated in alcohol, it was hyalinized with xylene and embedded in paraffin. The embedded paraffin block was fixed and cut into 5–8 μm thick slices on a microtome and placed in a 45°C thermostat for drying.
The paraffin was removed from the sections with xylene, the hematoxylin and eosin (H&E) sections were washed with high to low concentrations of alcohol, and finally with distilled water before staining. The sections were placed in an aqueous hematoxylin solution for 3 min to stain the nuclei, and then in an eosin solution for 3 min to stain the cytoplasm. The sections were dehydrated with anhydrous ethanol and xylene, dried, and sealed with gum.
The colonic sections were stained with Alcian blue to detect the mucins. The colonic sections were dewaxed, dehydrated in gradient alcohol, and rehydrated in distilled water. The sections were soaked in an Alcian blue acidified solution for 3 min, stained with an Alcian blue staining solution for 30 min, and rinsed in running water for 5 min. The samples were re-stained in nuclear solid staining solution for 5 min and rinsed in running water for 1 min. The stained slides were scanned using Pannoramic SCAN (3DHISTECH Kft). Image analysis was performed with ImageJ.
## Immunofluorescence analysis
The colonic sections were dewaxed and dehydrated in gradient alcohol. The sections were blocked with $2\%$ BSA at 37°C for 30 min. Fluorescently labeled anti-ZO1 tight junction protein rabbit pAb (GB111402, Servicebio, Beijing, China) and anti-MUC2 rabbit pAb (GB11344, Servicebio) was added dropwise to the sections and incubated at 37°C for 30 min. The sections were rinsed three times with 0.01 mol/L PBS (pH = 7.4) for 5 min each time. Cell contours were formed using DAPI negative staining and blocked in buffered glycerol (analytically pure non-fluorescent glycerol mixed with pH = 9.2, 0.2 M carbonate buffer at 9:1). The stained slides were scanned as indicated above. Image analysis was performed with ImageJ.
## Fecal DNA extraction and 16S ribosomal DNA gene sequencing
Feces were collected under sterile conditions before the mice were euthanized. The DNA was extracted from frozen fresh feces using the Stool Genomic DNA Extraction Kit (D2700, Solarbio® Life Science) according to the manufacturer’s instructions. The 16S ribosomal DNA (16S rDNA) gene sequencing method was described in our previous study (Huang et al., 2022). The V3–V4 16S rDNA target region of the ribosomal RNA gene was amplified by PCR (95°C for 5 min, followed by 30 cycles at 95°C for 1 min, 60°C for 1 min, and 72°C for 1 min and a final extension at 72°C for 7 min) using the forward primer 341F 5′-CCTACGGGNGGCWGCAG-3′ and the reverse primer 806R 5′-GGACTACHVGGGTATCTAAT-3′ (amplicon size was 466). PCR reagents were from New England Biolabs, United States.
The amplicons were extracted and purified from $2\%$ agarose gels using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States) according to the manufacturer’s instructions. Then, those amplicons were quantified using ABI StepOnePlus Real-Time PCR System (Life Technologies, Foster City, United States). The purified amplicons were equimolar pooled and paired-end sequenced (PE250) on the Illumina NovaSeq 6000 according to the standard protocol ($$n = 5$$/group).
## Quantitative real-time PCR analysis
The steps for fecal DNA extraction are described in Section 2.7. The DNA concentrations were determined by a spectrophotometer (NanoDrop, Thermo Fisher, Waltham, MA, United States). Quantitative real-time PCR was used to quantify the bacteria. Quantitative real-time PCR was performed with the 2X SG Fast qPCR Master Mix (High Rox; B639273, Beyotime, Beijing, China), and relative DNA expression was measured and analyzed with the ABI QuantStudio 6 Flex (Thermo Fisher). Each reaction was performed in triplicate. The 2−ΔΔCt method was used to calculate the relative quantity of DNA compared to the internal control. Eubacteria was used as the internal reference gene. The final results of the AMVN and AKK groups were calculated relative to the CTL group. The relative primer sequences are available in Supplementary Table 1 ($$n = 3$$/group).
## RNA extraction and quantitative reverse transcription PCR analysis
RNA was extracted from frozen colonic tissue using the RNAeasy™ Animal RNA Isolation Kit and a Spin Column (R0027, Beyotime) according to the manufacturer’s instructions. The contents of occludin, claudin-4, and ZO-1 mRNA were measured by quantitative reverse transcription PCR. Quantitative reverse transcription PCR was performed using the BeyoFast™ SYBR Green One-Step qRT-PCR Kit (D7268S, Beyotime), and relative RNA expression was measured and analyzed using the ABI QuantStudio 6 Flex (Thermo Fisher). Each reaction was performed in triplicate. The relative amount of RNA compared to the internal control was calculated by the 2−ΔΔCt method. GAPDH was used as the internal reference gene. The relative primer sequences are available in Supplementary Table 1 ($$n = 4$$ for the AMVN group and $$n = 5$$ for the AKK group).
## Data preprocessing and bioinformatics analysis
The 16S rDNA gene sequencing data were processed and analyzed by referencing our previous study (Huang et al., 2022). First, we used FASTP for filtering to obtain high-quality reads from the raw sequencing data. Then, the paired-end clean reads were merged as raw tags using FLASH (version 1.2.11) with a minimum overlap of 10 bp and a mismatch error rate of $2\%$.
The clean reads retained after quality control were clustered into operational taxonomic units (OTUs) of ≥$97\%$ similarity using the UPARSE (version 9.2.64) pipeline. All chimeric tags were removed using the UCHIME algorithm, and effective tags were obtained for further analysis. The tag sequence with the highest abundance was selected as the representative sequence within each cluster. The representative OTU sequences were classified into organisms using a naive Bayesian model and RDP classifier (version 2.2) based on the SILVA database (version 132) with a confidence threshold value of 0.8.
Alpha diversity indices were calculated using Usearch (version 10.0.240) and visualized using GraphPad Prism 9 (version 0.1.1; GraphPad Software Inc., La Jolla, CA, United States). The Vegan package (version 2.5–7) in R (version 4.0.5; The R Foundation for Statistical Computing, Vienna, Austria) was used to depict the Bray-Curtis distances for non-metric multidimensional scaling (NMDS). The graph of the NMDS plot and the stacked bar plots of the microbiota composition were visualized in R using the ggplot2 package (version 3.5.5). The unweighted pair group method with the arithmetic means (UPGMA) function in the R Vegan package was used to obtain the hierarchical clusters among the samples. The biomarker features were screened in each group using linear discriminant analysis effect size (LEfSe) analysis and LEfSe software (version 1.0).
## Statistical analysis
All data were analyzed and graphs were prepared using GraphPad Prism 8.0. Two groups of data were analyzed using Student’s t-test and Welch’s t-test. Two-way ANOVA followed by Tukey’s multiple comparison test was used to analyze the data from more than two groups. Finally, the data were presented as mean ± SEM, and values of $p \leq 0.05$ were considered significant.
## Malignant damage of the colonic tissue in the in situ malignant CRC model
An in situ malignant CRC model was used to study the changes in the intestinal microenvironment in the cancer state. To assess the development of CRC, we examined the body weight and survival rate of the mice for 3 weeks. The mice were humanely euthanized in W3, and the colorectal tissue was collected for H&E staining to demonstrate intestinal damage. The body weight of the mice in the CRC group was significantly lower than that of the other groups (Figure 1A). Some mice in the CRC group died from tumors (Figure 1B). H&E staining displayed immune cell infiltration and colonic tissue damage. As shown in Figure 1C, the in situ tumors severely damaged the colonic tissue. A decrease in the number of intestinal glands in the intrinsic layer, a disorganized and distorted arrangement of the crypts, an increase in the number of basal plasma cells, and inflammatory infiltration were observed in the CRC group compared with the CTL and SHAM groups.
**Figure 1:** *Malignant damage to the colonic tissue in the in situ malignant colorectal cancer (CRC) model. (A) Changes in the body weights of mice over 3 weeks. (B) Survival rate. (C) Hematoxylin and eosin (H&E) staining at W3. Scale bar, 100 μm. Data are mean ± SEM; n ≥ 3, *p < 0.05, **p < 0.005.*
## Decreased gut microenvironment abundance in the in situ malignant CRC model
To investigate variations in the intestinal microenvironment of the mice, feces were collected during W3 to perform 16S rDNA sequencing. Intestinal microbiota diversity indicates the stability of the micro-ecology of the microbiota and its resistance to invasion by external pathogens (He et al., 2021). Alpha diversity was analyzed in multiple dimensions (dilution curves), perspectives (different indices), and forms, including the Ace index, the Chao1 index, the Simpson index, the Shannon index, and richness (Figures 2A–E). Higher alpha diversity indices indicate higher community diversity. The Ace index (Figure 2A), Chao1 index (Figure 2B), Simpson index (Figure 2C), Shannon index (Figure 2D), and richness (Figure 2E) were significantly lower in the CRC group than those in the CTL and SHAM groups. The gradual flattening of the dilution curves in all groups indicated that the number of species in the gut did not increase with the number of sequences. Therefore, the number of samples measured in this experiment was sufficient to characterize the microbiota. The results indicate that the abundance of the intestinal microbiota decreased significantly in mice with CRC and the stability of the intestinal microenvironment was disrupted.
**Figure 2:** *Decreased abundance of intestinal microbiota in CRC mice. Alpha diversity analysis of the three groups: Ace index (A), Chao1 index (B), Simpson index (C), Shannon index (D), and the richness dilution curve (E). (F) Non-metric multidimensional scaling (NMDS) based on the operational taxonomic units (OTUs) in the three groups. Each point on the graph represents a sample. (G,H) Structural changes in the intestinal microbiota of the CRC model mice. The relative abundance of the top 10 microbial taxa was assessed at the (G) phylum and (H) generic levels. Data are mean ± SEM; n = 5, “ns,” not significant, **p < 0.005, ***p < 0.0005, and ****p < 0.0001.*
We used NMDS and UPGMA clustering analysis to examine the structural differences among the three groups of samples. NMDS analysis reflects samples as points in a multidimensional space based on species information. The NMDS scatter plot (Figure 2F) reflected the discrepancies between the three groups of samples based on the distance between points. The 2D stress value was 0.06, which was well represented. NMDS analysis showed that the composition of the intestinal microenvironment in the CRC group was significantly distinct from that of the CTL and SHAM groups, suggesting that the gut microbial composition changed in mice due to CRC. The samples were classified in the UPGMA classification tree based on the beta diversity distance matrix information. Similar samples had fewer common branches. UPGMA clustering analysis (Supplementary Figure 1) showed similar findings as the NMDS analysis.
## Structural changes in the intestinal microenvironment of the in situ malignant CRC model
The changes in the composition of the intestinal microbiota in CRC mice were further investigated. We analyzed the community structure of the entire sample at the phylum level. The relative abundances of the top 10 microbial taxa are reported as stacked histograms (Figure 2G). The relative abundances of Bacteroidetes, Patescibacteria, and Actinobacteria decreased in the CRC group, while the relative abundances of Firmicutes, Proteobacteria, Verrucomicrobia, and Epsilonbacteraeota increased compared with those in the CTL and SHAM groups. The relative abundance results at the phylum level demonstrated that the intestinal microbiota composition was altered in the CRC mice.
To further reveal the changes at the genus level, the relative abundance results of the top 10 genera are shown in Figure 2H. The structural variations in the CRC group were significant compared to those in the CTL and SHAM groups. CRC decreased the abundances of Candidatus saccharimonas, Ruminococcaceae UCG-014, Alistipes, and Bacteroides, while it significantly increased the abundances of Enterococcus, Escherichia shigella, Romboutsia, and Akkermansia. Similarly, the structure of the intestinal microbiota was significantly altered in the CRC group and homeostasis of the intestinal microenvironment was disrupted.
## Identification of the signature bacteria in CRC and excess colonization of Akkermansia muciniphila
We used LEfSe software to count significantly different biomarkers in the CRC and SHAM groups and further identify the key microbiota in the guts of the CRC mice. Linear discriminant analysis (LDA) effects were used to identify characteristic biomarkers (Figure 3A). The logarithmic LDA score for significant differences was set to 3, and the corresponding cladogram is provided in Supplementary Figure 2. Bacteroides acidifaciens, Lactobacillus gasseri, Ruminococcaceae, Prevotellaceae, and C. saccharimonas were enriched in the SHAM group. However, E. Shigella from Enterobacteriaceae, Enterococcus, Romboutsia, Lactococcus, and Akkermansia were enriched in the CRC group.
**Figure 3:** *Identification of signature bacteria in CRC and excess colonization of Akkermansia muciniphila. (A) Linear discriminant analysis effect size (LEfSe) analysis between the CRC and SHAM groups with the highest linear discriminant analysis (LDA) score [log (LDA score) ≥ 3]. Relative abundances of Bacteroides acidifaciens
(B), Lactobacillus gasseri
(C), Candidatus saccharimonas
(D), Escherichia Shigella
(E), Enterococcus
(F), and Akkermansia
(G). Data are mean ± SEM; n ≥ 3, “ns,” not significant, *p < 0.05, and **p < 0.005.*
The relative abundance statistics of the characteristic microbiota were further analyzed to determine the environmental differences among the three groups. The relative abundances of B. acidifaciens (Figure 3B), L. gasseri (Figure 3C), and C. saccharimonas (Figure 3D) decreased significantly in the CRC group. In contrast, the relative abundances of pathogenic bacteria, such as E. Shigella (Figure 3E) and Enterococcus (Figure 3F), increased. Notably, the relative abundance of Akkermansia increased compared to that in the CTL and SHAM groups (Figure 3G).
We repeated the same experiments to examine the changes in the gut microbiota in the CRC model. The gut microbiota of the CRC group was significantly different from that of the control group, as shown in Supplementary Figures 3A–E. Significant differences were observed at the phylum and genus levels, and the relative abundances of Verrucomicrobia and Akkermansia increased significantly in the CRC group (Supplementary Figures 3F,G). The LEfSe analysis also showed that Akkermansia was an extremely significantly enriched bacterial species in the model group (Supplementary Figure 3H). These results suggest that the relative abundance of Akkermansia in the intestines of CRC mice increased significantly. These results reveal the reproducibility of the aforementioned experiments.
## Excess colonization of Akkermansia muciniphila leads to minor inflammation
The over-colonized A. muciniphila mouse model was established by gastric gavage with live A. muciniphila after the antibiotic treatment to assess the condition of the intestinal barrier during A. muciniphila over-colonization (Figure 4A). The feces of the mice were collected on the final day, and colonization by A. muciniphila was detected by qRT-PCR. As shown in Figure 4B, the A. muciniphila mRNA relative expression level in the AKK group was significantly higher than that in the CTL and AMVN groups. To show the role of antibiotic in A. muciniphila over-colonization, we added a group (AKK−) with the same concentration of A. muciniphila gavage but no antibiotic pretreatment. The results indicated no significant difference in the relative expression of A. muciniphila between the AKK− and CTL groups (Supplementary Figure 4). However, the relative expression of A. muciniphila was significantly higher in the AKK group than in the AKK-group. These results show that the antibiotic cocktail treatment caused dysregulation of the intestinal microbiota, which aided A. muciniphila colonization and the over-colonized A. muciniphila mouse model was successfully established. Figure 4C shows the body weights of the mice before and after the A. muciniphila gavage. The body weights of the AKK and AMVN groups were lower on day 0 due to the antibiotic treatment compared to the CTL group. Mice in the AKK group had significantly lower body weights.
**Figure 4:** *Excess colonization of the Akkermansia muciniphila mouse model. (A) Schematic diagram of the experiment. (B) Relative expression of A. muciniphila in mice feces tested by qRT-PCR. (C) Body weights before and after administering A. muciniphila. (D) The quantity of white blood cells (WBC). The percentages of neutrophils (NE) (E), lymphocytes (Lymph) (F), and monocytes (Mon) (G). Data are mean ± SEM; n ≥ 3, “ns,” not significant, *p < 0.05, **p < 0.005, and ***p < 0.0005.*
We determined whether the over-colonization of A. muciniphila caused inflammation using hematological analysis. As shown in Figures 4D–G, the number of white blood cells (WBCs) in the AKK group increased significantly compared with that in the CTL and AMVN groups. After the mice were treated with A. muciniphila, the percentages of neutrophils (NE) and monocytes (Mon) increased, whereas the percentages of lymphocytes (Lymph) decreased. These results indicate that the excess colonization of A. muciniphila caused mild inflammation. H&E staining of the colonic tissues (Figure 5A) revealed that excess colonization of A. muciniphila destroyed crypts, caused edema, and resulted in massive inflammatory cell infiltration. This result was consistent with the hematological analysis. The excess colonization of A. muciniphila inflamed and destroyed the colonic tissue.
**Figure 5:** *Disruption of the intestinal barrier by excess colonization of Akkermansia muciniphila.
(A) H&E staining of colonic tissues. Scale bar, 100 μm. (B) Alcian blue staining of colonic tissues. Scale bar, 100 μm. (C) Quantitative analysis of Alcian blue staining; relative mean. (D) Immunofluorescence staining of colon tissues for MUC2 (red) and ZO-1 (green). DAPI (blue) was used to stain nuclei. Scale bar, 100 μm. Quantitative analysis of MUC2 (E) and ZO-1 (F) relative fluorescence intensity. The relative mRNA expression of occludin (G), claudin-4 (H), and ZO-1 (I) in the intestinal tissue in the AKK and AMVN groups. Data are mean ± SEM; n ≥ 3, “ns,” not significant, *p < 0.05, **p < 0.005, ***p < 0.0005, and ****p < 0.0001.*
## Disruption of the intestinal barrier by excess Akkermansia muciniphila colonization
The mucus layer is the first defense barrier against pathogens, and goblet cells secrete mucins to maintain the thickness of the mucus layer (Gustafsson and Johansson, 2022). To determine the effect of A. muciniphila over-colonization on the mucus layer, colonic tissues were stained with Alcian blue (goblet cells stain blue). As shown in Figure 5B, the CTL group exhibited regular goblet cell morphology and a normal mucus layer, while the AMVN group developed slight thinning of the mucus layer. Notably, the mucus layer in the colonic tissue of the AKK group was severely damaged and the number of goblet cells decreased significantly. The area of blue in the images was determined quantitatively to assess the degree of disruption of the mucus layer (Figure 5C). The relative mean of the Alcian blue-stained area was significantly lower in the AKK group than that in the CTL and AMVN groups.
Mucin 2 (MUC2) is a secreted mucin, and the main mucin in the colonic barrier, which plays a role in lubrication and protects the colonic mucosa against bacteria and toxins. The MUC2 level has been used as an important indicator of colonic mucosal integrity. ZO-1 is an indicator of the function of the tight junction barrier between intestinal epithelial cells. Thus, the expression levels of MUC2 and ZO-1 were detected in colonic tissues by immunofluorescence (Figure 5D). The MUC2 protein was red, ZO-1 was green, and DAPI was used to stain the cell nuclei. The relative fluorescence intensities of MUC2 and ZO-1 were statistically quantified among the three groups. As shown in Figures 5E,F, relative fluorescence intensity was significantly lower in the AKK group than that in the CTL and AMVN groups, indicating that the MUC2 and ZO-1 levels were low in the colonic tissue of the AKK group. In addition, the relative mRNA expression levels of occludin, claudin-4, and ZO-1 also decreased in the intestinal tissue (Figures 5G–I). These results suggest that excess colonization of A. muciniphila led to substantial catabolism of mucin and tight junction proteins in colonic tissues and an impaired intestinal barrier.
## Discussion
In this study, we prepared an in situ malignant CRC model by transplanting tumor tissues from subcutaneous tumors created with CT26 cells. Then, we examined the changes in intestinal microbial ecology and composition in CRC mice using 16S rDNA sequencing. Remarkably, the relative abundance of A. muciniphila increased at the genus level in the LDA analysis. In our recent study, we noticed a similar over-colonization in mice with DSS-induced ulcerative colitis (Huang et al., 2022). As mentioned before, the characteristics of A. muciniphila are related to the mucus layer of the intestinal barrier. Therefore, we established an over-colonized A. muciniphila mouse model to confirm whether the role of A. muciniphila in intestinal barrier damage is compensatory regulation or aggravation. The level of intestinal barrier-related mucins and tight junction proteins decreased in intestinal tissues, confirming that the over-colonized A. muciniphila caused intestinal barrier damage.
In our study, we established the over-colonized A. muciniphila mouse model by gastric gavage with live A. muciniphila (1 × 109 CFU) after the antibiotic treatment. Some studies (Everard et al., 2013; Plovier et al., 2017) have indicated that A. muciniphila (2 × 108 CFU) gavage alone has no detrimental effect. Notably, the antibiotic cocktail was an important factor contributing to colonization by A. muciniphila. As the same time, we used the Eubacteria primers as an internal reference for total bacteria (or as the housekeeping gene). These primers are universal eubacteria community primers (UniF340 and UniR514), which represent all bacteria and are widely used in qRT-PCR and 16S rDNA studies to determine bacterial abundance or capture the 16S rDNA of all bacteria (Wang et al., 1996; Lawley et al., 2017; Kuhbandner et al., 2019; Chen et al., 2022). The target threshold cycle (Ct) for the relative qRT-PCR expression results was divided by the internal reference Ct using the delta–delta Ct method. Thus, our qRT-PCR results reflect the relative amount of A. muciniphila to the overall microbiota. The abundance of A. muciniphila in the AKK group was significantly higher than that in the CTL and AMVN groups, so A. muciniphila was over-colonized.
The intestinal mucus barrier is the first natural barrier against the invasion of harmful substances. It plays an active role defending against the invasion of disease-causing microorganisms and assisting in the colonization of probiotics (Johansson et al., 2008). MUC2 is a heavily glycosylated secretory mucin secreted by intestinal goblet cells and is the main component of the intestinal mucus barrier (Yao et al., 2021). Akkermansia muciniphila normally colonizes the outer layer of loose mucus and upregulates the synthesis of MUC2 by goblet cells through metabolites (Meng et al., 2020). Although A. muciniphila degrades the mucus layer, it does not change its thickness, which reaches a dynamic equilibrium (Derrien et al., 2017). In our study, over-colonization of A. muciniphila resulted in a significantly lower MUC2 level than normal, suggesting that over-colonized A. muciniphila consumes more mucin than goblet cells secrete. As a result, both the intestinal mucus barrier and the balance of the intestinal barrier were compromised.
The primary component of the intestinal barrier is intestinal mucosal epithelial cells, which are held together by tight junction proteins. The mechanical integrity and proper operation of the intestinal mucosal barrier are maintained by these tight junctions (Suzuki, 2020). Tight junctions consist of occludin, claudins, junction adhesion molecules, and peripheral cytoplasmic proteins, such as ZO-1, ZO-2, and ZO-3. Our study revealed that ZO-1 content decreased significantly in the colonic tissue of the over-colonized A. muciniphila mouse model, and the relative mRNA expression levels of occludin, claudin-4, and ZO-1 also decreased in the intestinal tissue. As shown in our results, an overabundance of A. muciniphila led to a thinning of the intestinal mucus layer, which allowed harmful bacteria to directly contact the intestinal epithelial cells, further causing epithelial cell damage and disrupting the tight junctions between the cells. At the same time, the persistent deficiency of mucins and the difficulties in repairing the intestinal mucus layer were accompanied by further weakening of the repair capacity of the intestinal epithelial cells. These two factors contributed to disrupt the intestinal barrier and aggravate the development of colitis and CRC. These factors also contributed to bacterial translocation, providing opportunities for pathogenic bacteria and potentially carcinogenic metabolites. Harmful substances can access the peripheral circulation, causing systemic tissue damage and possibly leading to serious autoimmune diseases, such as type 1 diabetes (Costa et al., 2016) or systemic lupus erythematosus (Ogunrinde et al., 2019).
Akkermansia muciniphila is widely considered to be a next-generation probiotic. Clinical trial results have shown that pasteurized A. muciniphila has a positive effect on overweight and obese human volunteers rather than live A. muciniphila (Depommier et al., 2019). Many conflicting results about the effects of A. muciniphila have been reported by IBD-related studies (Zhang et al., 2021). It has also been shown that A. muciniphila exerts anti-inflammatory effects through extracellular vesicles (Kang et al., 2013; Chelakkot et al., 2018). However, one study revealed that A. muciniphila interferes with mucosal reconstitution and exacerbates inflammation in *Salmonella enterica* typhimurium-induced gut inflammation (Ganesh et al., 2013). This may be due to A. muciniphila-enhanced catabolism of mucin products, which provides other gut pathogenic bacteria the substances they needed to grow. Another study found that the abundance of A. muciniphila increased significantly in the mouse intestines after worsening of spontaneous colitis caused by NLRP6 gene knockout in IL-10−/− mice (Seregin et al., 2017), and that A. muciniphila administration orally worsened the colitis. It has been demonstrated that the abundance of A. muciniphila is strongly associated with Parkinson’s disease and is a disease signature microbiota (Hill-Burns et al., 2017; Heintz-Buschart et al., 2018). A higher abundance of A. muciniphila has also been observed in multiple sclerosis and Alzheimer’s disease (Berer et al., 2017; Vogt et al., 2017). Although the causality and mechanisms associated with these observations remain unclear, these findings suggest that A. muciniphila may be a double-edged sword.
Our findings support the idea that intestinal A. muciniphila, which also be considered as probiotics, may have negative effects in certain pathological situations and aggravate intestinal disease symptoms. Akkermansia muciniphila becomes a potential pathogenic agent under some conditions. Thus, it is crucial to analyze the function of A. muciniphila in many contexts before use in clinical treatment. Our findings raise awareness of the possible pathogenic role of live A. muciniphila as a probiotic. Live A. muciniphila aggravates damage to the intestinal barrier and promotes the development of inflammation and cancer. Thus, it is not possible to use A. muciniphila to treat patients with diseases, such as colitis and CRC.
However, A. muciniphila also affects the ability of intestinal goblet cells to secrete mucin; the mechanism underlying this effect remains unknown and requires further research. Furthermore, we did not test the effect of A. muciniphila on other gut microorganisms, and the interactions between A. muciniphila and other bacteria were not determined. These limitations will be answered in our future studies. The findings of this study have important implications for understanding how A. muciniphila interferes with the reconstruction of a damaged intestinal mucus layer. They suggest that we must consider the patient’s condition when applying A. muciniphila to the clinic.
## Conclusion
Our findings indicate that the abundance and structure of the gut microbiota are disturbed in CRC mice with substantial damage to the intestinal barrier. Notably, the number of A. muciniphila increased significantly in the context of intestinal disease. Excess A. muciniphila degraded mucins rather than increasing mucin synthesis in the intestinal tissues. Disruption of the tight junctions of intestinal epithelial cells and the ability to produce mucin to repair the mucosal barrier worsened the intestinal mucosal environment, which promotes the displacement of microbiota and exacerbates inflammation and cancer. Our study suggests that over-colonized A. muciniphila contributes to the genesis and progression of intestinal diseases by disrupting the intestinal barrier through excessive consumption of mucin.
## Data availability statement
The data presented in the study are deposited in the Genome Sequence Archive (Genomics, Proteomics and Bioinformatics 2021) in National Genomics Data Center (Nucleic Acids Res 2022), China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA009251) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa/browse/CRA009251.
## Ethics statement
The animal study was reviewed and approved by Animal Ethics Committee of the Institute of Chinese Materia Medica.
## Author contributions
SQ, YZ, and YH contributed to conception and design of the study under the guidance of MQ, and KN. SQ, and YZ carried out most experiments. YH was responsible for the bioinformatic analysis. SQ and YZ wrote the first draft of the manuscript. YF, KX, WZ, and YW carried out article correction and provided guidance on the overall study. All authors contributed to the article and approved the submitted version.
## Funding
This work was partially or fully sponsored by the National Key Research and Development Program of China with grant no. ZK20200085.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2023.1111911/full#supplementary-material
## References
1. Berer K., Gerdes L. A., Cekanaviciute E., Jia X., Xiao L., Xia Z.. **Gut microbiota from multiple sclerosis patients enables spontaneous autoimmune encephalomyelitis in mice**. *Proc. Natl. Acad. Sci. U. S. A.* (2017) **114** 10719-10724. DOI: 10.1073/pnas.1711233114
2. Chelakkot C., Choi Y., Kim D. K., Park H. T., Ghim J., Kwon Y.. **Akkermansia muciniphila-derived extracellular vesicles influence gut permeability through the regulation of tight junctions**. *Exp. Mol. Med.* (2018) **50** e450. DOI: 10.1038/emm.2017.282
3. Chen Y. L., Bai L., Dilimulati D., Shao S., Qiu C., Liu T.. **Periodontitis salivary microbiota aggravates ischemic stroke through IL-17A**. *Front. Neurosci.* (2022) **16** 876582. DOI: 10.3389/fnins.2022.876582
4. Coker O. O., Liu C., Wu W. K. K., Wong S. H., Jia W., Sung J. J. Y.. **Altered gut metabolites and microbiota interactions are implicated in colorectal carcinogenesis and can be non-invasive diagnostic biomarkers**. *Microbiome* (2022) **10** 35. DOI: 10.1186/s40168-021-01208-5
5. Costa F. R., Françozo M. C., de Oliveira G. G., Ignacio A., Castoldi A., Zamboni D. S.. **Gut microbiota translocation to the pancreatic lymph nodes triggers NOD2 activation and contributes to T1D onset**. *J. Exp. Med.* (2016) **213** 1223-1239. DOI: 10.1084/jem.20150744
6. Depommier C., Everard A., Druart C., Plovier H., Van Hul M., Vieira-Silva S.. **Supplementation with Akkermansia muciniphila in overweight and obese human volunteers: a proof-of-concept exploratory study**. *Nat. Med.* (2019) **25** 1096-1103. DOI: 10.1038/s41591-019-0495-2
7. Derrien M., Belzer C., de Vos W. M.. **Akkermansia muciniphila and its role in regulating host functions**. *Microb. Pathog.* (2017) **106** 171-181. DOI: 10.1016/j.micpath.2016.02.005
8. Derrien M., Vaughan E. E., Plugge C. M., de Vos W. M.. **Akkermansia muciniphila gen. Nov., sp. nov., a human intestinal mucin-degrading bacterium**. *Int. J. Syst. Evol. Microbiol.* (2004) **54** 1469-1476. DOI: 10.1099/ijs.0.02873-0
9. Everard A., Belzer C., Geurts L., Ouwerkerk J. P., Druart C., Bindels L. B.. **Cross-talk between Akkermansia muciniphila and intestinal epithelium controls diet-induced obesity**. *Proc. Natl. Acad. Sci. U. S. A.* (2013) **110** 9066-9071. DOI: 10.1073/pnas.1219451110
10. Everard A., Lazarevic V., Gaïa N., Johansson M., Ståhlman M., Backhed F.. **Microbiome of prebiotic-treated mice reveals novel targets involved in host response during obesity**. *ISME J.* (2014) **8** 2116-2130. DOI: 10.1038/ismej.2014.45
11. Ganesh B. P., Klopfleisch R., Loh G., Blaut M.. **Commensal Akkermansia muciniphila exacerbates gut inflammation in salmonella typhimurium-infected gnotobiotic mice**. *PLoS One* (2013) **8** e74963. DOI: 10.1371/journal.pone.0074963
12. Gustafsson J. K., Johansson M. E. V.. **The role of goblet cells and mucus in intestinal homeostasis**. *Nat. Rev. Gastroenterol. Hepatol.* (2022) **19** 785-803. DOI: 10.1038/s41575-022-00675-x
13. He T., Cheng X., Xing C.. **The gut microbial diversity of colon cancer patients and the clinical significance**. *Bioengineered* (2021) **12** 7046-7060. DOI: 10.1080/21655979.2021.1972077
14. Heintz-Buschart A., Pandey U., Wicke T., Sixel-Döring F., Janzen A., Sittig-Wiegand E.. **The nasal and gut microbiome in Parkinson's disease and idiopathic rapid eye movement sleep behavior disorder**. *Mov. Disord.* (2018) **33** 88-98. DOI: 10.1002/mds.27105
15. Hill-Burns E. M., Debelius J. W., Morton J. T., Wissemann W. T., Lewis M. R., Wallen Z. D.. **Parkinson's disease and Parkinson's disease medications have distinct signatures of the gut microbiome**. *Mov. Disord.* (2017) **32** 739-749. DOI: 10.1002/mds.26942
16. Huang Y., Zheng Y., Yang F., Feng Y., Xu K., Wu J.. **Lycium barbarum Glycopeptide prevents the development and progression of acute colitis by regulating the composition and diversity of the gut microbiota in mice**. *Front. Cell. Infect. Microbiol.* (2022) **12** 921075. DOI: 10.3389/fcimb.2022.921075
17. Johansson M. E., Phillipson M., Petersson J., Velcich A., Holm L., Hansson G. C.. **The inner of the two Muc2 mucin-dependent mucus layers in colon is devoid of bacteria**. *Proc. Natl. Acad. Sci. U. S. A.* (2008) **105** 15064-15069. DOI: 10.1073/pnas.0803124105
18. Kang C. S., Ban M., Choi E. J., Moon H. G., Jeon J. S., Kim D. K.. **Extracellular vesicles derived from gut microbiota, especially Akkermansia muciniphila, protect the progression of dextran sulfate sodium-induced colitis**. *PLoS One* (2013) **8** e76520. DOI: 10.1371/journal.pone.0076520
19. Kim Y. S., Ho S. B.. **Intestinal goblet cells and mucins in health and disease: recent insights and progress**. *Curr. Gastroenterol. Rep.* (2010) **12** 319-330. DOI: 10.1007/s11894-010-0131-2
20. Kuhbandner K., Hammer A., Haase S., Terbrack E., Hoffmann A., Schippers A.. **MAdCAM-1-mediated intestinal lymphocyte homing is critical for the development of active experimental autoimmune encephalomyelitis**. *Front. Immunol.* (2019) **10** 903. DOI: 10.3389/fimmu.2019.00903
21. Kummen M., Holm K., Anmarkrud J. A., Nygård S., Vesterhus M., Høivik M. L.. **The gut microbial profile in patients with primary sclerosing cholangitis is distinct from patients with ulcerative colitis without biliary disease and healthy controls**. *Gut* (2017) **66** 611-619. DOI: 10.1136/gutjnl-2015-310500
22. Lawley B., Munro K., Hughes A., Hodgkinson A. J., Prosser C. G., Lowry D.. **Differentiation of Bifidobacterium longum subspecies longum and infantis by quantitative PCR using functional gene targets**. *Peer J.* (2017) **5** e3375. DOI: 10.7717/peerj.3375
23. Lee M. H.. **Harness the functions of gut microbiome in tumorigenesis for cancer treatment**. *Cancer Commun.* (2021) **41** 937-967. DOI: 10.1002/cac2.12200
24. Li J., Zhao F., Wang Y., Chen J., Tao J., Tian G.. **Gut microbiota dysbiosis contributes to the development of hypertension**. *Microbiome* (2017) **5** 14. DOI: 10.1186/s40168-016-0222-x
25. Liu H. N., Wu H., Chen Y. Z., Chen Y. J., Shen X. Z., Liu T. T.. **Altered molecular signature of intestinal microbiota in irritable bowel syndrome patients compared with healthy controls: a systematic review and meta-analysis**. *Dig. Liver Dis.* (2017) **49** 331-337. DOI: 10.1016/j.dld.2017.01.142
26. Mahalhal A., Williams J. M., Johnson S., Ellaby N., Duckworth C. A., Burkitt M. D.. **Oral iron exacerbates colitis and influences the intestinal microbiome**. *PLoS One* (2018) **13** e0202460. DOI: 10.1371/journal.pone.0202460
27. Meng X., Wang W., Lan T., Yang W., Dahai Y., Fang X.. **A purified aspartic protease from Akkermansia Muciniphila plays an important role in degrading Muc2**. *Int. J. Mol. Sci.* (2020) **21** 72. DOI: 10.3390/ijms21010072
28. Ogunrinde E., Zhou Z., Luo Z., Alekseyenko A., Li Q. Z., Macedo D.. **A link between plasma microbial translocation, microbiome, and autoantibody development in first-degree relatives of systemic lupus erythematosus patients**. *Arthritis Rheum.* (2019) **71** 1858-1868. DOI: 10.1002/art.40935
29. Ouwehand A. C., Derrien M., de Vos W., Tiihonen K., Rautonen N.. **Prebiotics and other microbial substrates for gut functionality**. *Curr. Opin. Biotechnol.* (2005) **16** 212-217. DOI: 10.1016/j.copbio.2005.01.007
30. Palm N. W., de Zoete M. R., Cullen T. W., Barry N. A., Stefanowski J., Hao L.. **Immunoglobulin a coating identifies colitogenic bacteria in inflammatory bowel disease**. *Cells* (2014) **158** 1000-1010. DOI: 10.1016/j.cell.2014.08.006
31. Paone P., Cani P. D.. **Mucus barrier, mucins and gut microbiota: the expected slimy partners?**. *Gut* (2020) **69** 2232-2243. DOI: 10.1136/gutjnl-2020-322260
32. Plovier H., Everard A., Druart C., Depommier C., Van Hul M., Geurts L.. **A purified membrane protein from Akkermansia muciniphila or the pasteurized bacterium improves metabolism in obese and diabetic mice**. *Nat. Med.* (2017) **23** 107-113. DOI: 10.1038/nm.4236
33. Rajagopala S. V., Vashee S., Oldfield L. M., Suzuki Y., Venter J. C., Telenti A.. **The human microbiome and cancer**. *Cancer Prev. Res. (Phila.)* (2017) **10** 226-234. DOI: 10.1158/1940-6207.Capr-16-0249
34. Schultz B. M., Paduro C. A., Salazar G. A., Salazar-Echegarai F. J., Sebastián V. P., Riedel C. A.. **A potential role of salmonella infection in the onset of inflammatory bowel diseases**. *Front. Immunol.* (2017) **8** 191. DOI: 10.3389/fimmu.2017.00191
35. Seregin S. S., Golovchenko N., Schaf B., Chen J., Pudlo N. A., Mitchell J.. **NLRP6 protects Il10(−/−) mice from colitis by limiting colonization of Akkermansia muciniphila**. *Cell Rep.* (2017) **19** 733-745. DOI: 10.1016/j.celrep.2017.03.080
36. Sicard J. F., Le Bihan G., Vogeleer P., Jacques M., Harel J.. **Interactions of intestinal bacteria with components of the intestinal mucus**. *Front. Cell. Infect. Microbiol.* (2017) **7** 387. DOI: 10.3389/fcimb.2017.00387
37. Sung H., Ferlay J., Siegel R. L., Laversanne M., Soerjomataram I., Jemal A.. **Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA Cancer J. Clin.* (2021) **71** 209-249. DOI: 10.3322/caac.21660
38. Suzuki T.. **Regulation of the intestinal barrier by nutrients: the role of tight junctions**. *Anim. Sci. J.* (2020) **91** e13357. DOI: 10.1111/asj.13357
39. Van der Sluis M., De Koning B. A., De Bruijn A. C., Velcich A., Meijerink J. P., Van Goudoever J. B.. **Muc2-deficient mice spontaneously develop colitis, indicating that MUC2 is critical for colonic protection**. *Gastroenterology* (2006) **131** 117-129. DOI: 10.1053/j.gastro.2006.04.020
40. Vogt N. M., Kerby R. L., Dill-McFarland K. A., Harding S. J., Merluzzi A. P., Johnson S. C.. **Gut microbiome alterations in Alzheimer's disease**. *Sci. Rep.* (2017) **7** 13537. DOI: 10.1038/s41598-017-13601-y
41. Wang R. F., Cao W. W., Cerniglia C. E.. **PCR detection and quantitation of predominant anaerobic bacteria in human and animal fecal samples**. *Appl. Environ. Microbiol.* (1996) **62** 1242-1247. DOI: 10.1128/aem.62.4.1242-1247.1996
42. Wang L., Tang L., Feng Y., Zhao S., Han M., Zhang C.. **A purified membrane protein from Akkermansia muciniphila or the pasteurised bacterium blunts colitis associated tumourigenesis by modulation of CD8(+) T cells in mice**. *Gut* (2020) **69** 1988-1997. DOI: 10.1136/gutjnl-2019-320105
43. Wong S. H., Yu J.. **Gut microbiota in colorectal cancer: mechanisms of action and clinical applications**. *Nat. Rev. Gastroenterol. Hepatol.* (2019) **16** 690-704. DOI: 10.1038/s41575-019-0209-8
44. Yao D., Dai W., Dong M., Dai C., Wu S.. **MUC2 and related bacterial factors: therapeutic targets for ulcerative colitis**. *EBioMedicine* (2021) **74** 103751. DOI: 10.1016/j.ebiom.2021.103751
45. Zhai R., Xue X., Zhang L., Yang X., Zhao L., Zhang C.. **Strain-specific anti-inflammatory properties of two Akkermansia muciniphila strains on chronic colitis in mice**. *Front. Cell. Infect. Microbiol.* (2019) **9** 239. DOI: 10.3389/fcimb.2019.00239
46. Zhang T., Ji X., Lu G., Zhang F.. **The potential of Akkermansia muciniphila in inflammatory bowel disease**. *Appl. Microbiol. Biotechnol.* (2021) **105** 5785-5794. DOI: 10.1007/s00253-021-11453-1
47. Zhang X., Shen D., Fang Z., Jie Z., Qiu X., Zhang C.. **Human gut microbiota changes reveal the progression of glucose intolerance**. *PLoS One* (2013) **8** e71108. DOI: 10.1371/journal.pone.0071108
48. Zheng D., Liwinski T., Elinav E.. **Interaction between microbiota and immunity in health and disease**. *Cell Res.* (2020) **30** 492-506. DOI: 10.1038/s41422-020-0332-7
|
---
title: Regulation by Nrf2 of IL-1β-induced inflammatory and oxidative response in
VSMC and its relationship with TLR4
authors:
- Zoe González-Carnicero
- Raquel Hernanz
- Marta Martínez-Casales
- María Teresa Barrús
- Ángela Martín
- María Jesús Alonso
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC10018188
doi: 10.3389/fphar.2023.1058488
license: CC BY 4.0
---
# Regulation by Nrf2 of IL-1β-induced inflammatory and oxidative response in VSMC and its relationship with TLR4
## Abstract
Introduction: Vascular oxidative stress and inflammation play an important role in the pathogenesis of cardiovascular diseases (CVDs). The proinflammatory cytokine Interleukin-1β (IL-1β) participates in the vascular inflammatory and oxidative responses and influences vascular smooth muscle cells (VSMC) phenotype and function, as well as vascular remodelling in cardiovascular diseases. The Toll-like receptor 4 (TLR4) is also involved in the inflammatory response in cardiovascular diseases. A relationship between Interleukin-1β and Toll-like receptor 4 pathway has been described, although the exact mechanism of this interaction remains still unknown. Moreover, the oxidative stress sensitive transcription factor nuclear factor-erythroid 2-related factor 2 (Nrf2) promotes the transcription of several antioxidant and anti-inflammatory genes. Nuclear factor-erythroid 2-related factor 2 activators have shown to possess beneficial effects in cardiovascular diseases in which oxidative stress and inflammation are involved, such as hypertension and atherosclerosis; however, the molecular mechanisms are not fully understood. Here, we analysed the role of Toll-like receptor 4 in the oxidative and inflammatory effects of Interleukin-1β as well as whether nuclear factor-erythroid 2-related factor 2 activation contributes to vascular alterations by modulating these effects.
Materials: *For this* purpose, vascular smooth muscle cells and mice aortic segments stimulated with Interleukin-1β were used.
Results: Interleukin-1β induces MyD88 expression while the Toll-like receptor 4 inhibitor CLI-095 reduces the Interleukin-1β-elicited COX-2 protein expression, reactive oxygen species (ROS) production, vascular smooth muscle cells migration and endothelial dysfunction. Additionally, Interleukin-1β increases nuclear factor-erythroid 2-related factor 2 nuclear translocation and expression of its downstream proteins heme oxygenase-1, NAD(P)H:quinone oxidoreductase 1 and superoxide dismutase-2, by an oxidative stress-dependent mechanism; moreover, Interleukin-1β reduces the expression of the nuclear factor-erythroid 2-related factor 2 inhibitor Keap1. The nuclear factor-erythroid 2-related factor 2 activator tert-butylhydroquinone (tBHQ) reduces the effects of Interleukin-1β on the increased reactive oxygen species production and the expression of the proinflammatory markers (p-p38, p-JNK, p-c-Jun, COX-2), the increased cell proliferation and migration and prevents the Interleukin-1β-induced endothelial dysfunction in mice aortas. Additionally, tert-butylhydroquinone also reduces the increased MyD88 expression, NADPHoxidase activity and cell migration induced by lipopolysaccharide.
Conclusions: In summary, this study reveals that Toll-like receptor 4 pathway contributes to the prooxidant and proinflammatory Interleukin-1β-induced effects. Moreover, activation of nuclear factor-erythroid 2-related factor 2 prevents the deleterious effects of Interleukin-1β, likely by reducing Toll-like receptor 4-dependent pathway. Although further research is needed, the results are promising as they suggest that nuclear factor-erythroid 2-related factor 2 activators might protect against the oxidative stress and inflammation characteristic of cardiovascular diseases.
## Introduction
Cardiovascular diseases (CVDs) are one of the leading causes of morbidity and mortality worldwide (Şahin and İlgün, 2022). Vascular oxidative stress and inflammation play a significant role in the pathogenesis of CVDs, including hypertension and atherosclerosis (Senoner and Dichtl, 2019; Abbate et al., 2020; Griendling et al., 2021; Melton and Qiu, 2021). Among others, endothelial dysfunction and vascular smooth muscle cell (VSMC) alterations, such as abnormal migration and proliferation, play pivotal roles in the development of these pathologies (Wang et al., 2012; Eun et al., 2015; Badran et al., 2020; Griendling et al., 2021; Melton and Qiu, 2021).
Interleukin-1β (IL-1β) is a prototypical proinflammatory cytokine belonging to the IL-1 family (Abbate et al., 2020). IL-1β binds to its corresponding receptor and mostly signals through the myeloid differentiation factor 88 (MyD88) to facilitate downstream upregulation of inflammatory genes (Abbate et al., 2020; Melton and Qiu, 2021). Several reports suggest that IL-1β levels are increased in patients with CVDs such as hypertension and atherosclerosis, where this cytokine participates in the vascular proinflammatory and oxidative responses (Krishnan et al., 2014; Pfeiler et al., 2019; Mai and Liao, 2020; Melton and Qiu, 2021; Silveira Rossi et al., 2022). IL-1β also influences the phenotype and functions of VSMC, and induces cell migration and proliferation, participating in vascular remodelling in CVDs (Eun et al., 2015; Melton and Qiu, 2021).
On the other hand, the Toll-like receptor 4 (TLR4), a member of the family of pattern recognition receptors, participates in the inflammatory response in vascular diseases. TLR4 is expressed on the surface of several cell types, including VSMC (Nunes et al., 2019). Its activation occurs in response to both exogenous pathogen-associated molecular patterns (PAMPs) and endogenous molecules released by cells following tissue damage, called damage-associated molecular patterns (DAMPs) (Nunes et al., 2019). Although the classic TLR4 signalling pathway involves both MyD88-dependent and independent mechanisms (Liu et al., 2014), the TLR4-induced inflammatory response occurs mainly through activation of MyD88-dependent pathways (Liu et al., 2014; Biancardi et al., 2017). Signalling through TLR4 contributes to vascular inflammatory pathologies such as atherosclerosis and hypertension (De Batista et al., 2014; Yang et al., 2014; Hernanz et al., 2015). Furthermore, a relationship between IL-1β and TLR4 has recently been described, since treatment with this cytokine increases the activity of the TLR4/NF-κB pathway (Liu et al., 2014; Xu et al., 2020); however, the exact mechanism by which IL-1β interacts with this receptor is still unknown.
Reactive oxygen species (ROS) have been recognized as mediators of cellular signalling. At moderate levels, ROS are involved in physiological processes, but an excessive ROS production leads to disruption of redox signalling and to molecular damage (Xia et al., 2017; Griendling et al., 2021). To prevent ROS-induced alterations there are several antioxidant systems, among others, the nuclear factor-erythroid 2-related factor 2 (Nrf2). Nrf2 is a redox transcription factor (Yamamoto et al., 2018; Robledinos-Antón et al., 2019) which, under basal conditions, is restricted to the cytoplasm where it interacts with Kelch-like ECH-associated protein 1 (Keap1), leading to its ubiquitination (Yamamoto et al., 2018; Robledinos-Antón et al., 2019). Under stress conditions (ROS, electrophiles), Keap1 cysteines are oxidized, allowing Nrf2 to be released from its repressor Keap1 and translocate into the nucleus, wherein it heterodimerizes with small Maf proteins (sMAF) and binds to antioxidant response elements (AREs) on DNA, leading to transcription of ARE-driven genes. This set of genes regulates the expression of phase II detoxifying enzymes, including NAD(P)H:quinone oxidoreductase 1 (NQO1) or glutathione peroxidase (GPx), and antioxidant proteins such as heme oxygenase-1 (HO-1) (Yamamoto et al., 2018; Robledinos-Antón et al., 2019). Nrf2 has been related to different CVDs such as hypertension and atherosclerosis (Alonso-Piñeiro et al., 2021; Tanase et al., 2022). Thus, Nrf2 downregulation in stroke-prone spontaneously hypertensive rats (SHRSP) contributes to the increased oxidative stress and vascular dysfunction observed in this model (Lopes et al., 2015); however, patients who suffer from atherosclerosis have higher levels of Nrf2/HO-1 than healthy subjects (Fiorelli et al., 2019), suggesting that Nrf2 activation is an adaptive mechanism against oxidative stress characteristic of this CVDs.
Tert-butylhydroquinone (tBHQ) is a Nrf2 activator with antioxidant properties that enhances Nrf2-mediated transcription by promoting dissociation of Nrf2-Keap1 (Zhang and Hannink, 2003). Activators of Nrf2 have shown to have a beneficial effect in pathologies involving oxidative stress and inflammation (Xia et al., 2017; Robledinos-Antón et al., 2019). Thus, sulforaphane reduces ROS production both in vessels and VSMC from SHRSP (Lopes et al., 2015), and resveratrol, which also activates Nrf2, reduces ROS generation and lipid peroxidation, thus contributing to prevent from atherosclerosis (Parsamanesh et al., 2021). On the other hand, it has been described that tBHQ prevents microvascular endothelial dysfunction and remodelling and reduces blood pressure in angiotensin II-induced hypertension (Wang et al., 2018). Based on the mentioned data, Nrf2 signalling pathway is currently considered an important defence mechanism against several CVDs; however, the mechanisms underlying the preventive effects of Nrf2 are barely known.
The aim of this study was to analyse whether TLR4 pathway contributes to the proinflammatory and prooxidant IL-1β-induced effects; in addition, the result of Nrf2 activation by tBHQ on these IL-1β-mediated effects and the possible mechanisms involved were also evaluated.
## Ethic statements
All experimental procedures were approved by the Ethical Committee of Research of the Universidad Autónoma de Madrid and Dirección General de Medio Ambiente, Comunidad de Madrid, Spain (PROEX $\frac{183.2}{20}$). Animal care and experimental procedures conformed to the current Spanish laws (RD $\frac{53}{2013}$) and are also conformed to the Directive $\frac{2010}{63}$/EU of the European Parliament for animal experiments. The studies also comply with the ARRIVE guidelines for reporting experiments involving animals. C57BL/6 mice were obtained from colonies maintained at the Animal Quarters of the Facultad de Medicina of the Universidad Autónoma de Madrid. During treatment, mice were housed with constant room temperature, humidity and light cycle (12-h light/dark) and they had free access to tap water and were fed with standard mice chow ad libitum.
## Cell culture
Experiments were performed using the mouse aortic VSMC line MOVAS (ATCC® CRL-2797™, Manassas, VA, United States). Cells were maintained in High Glucose Dulbecco’s Modified Eagle Medium [High Glucose-DMEM (Sigma Chemical Co., St. Louis, MO, United States)] supplemented with $10\%$ (v/v) fetal bovine serum (FBS) and 0.2 mg/mL of the G-418 solution (Roche Diagnostics, Manheim, Germany). VSMC were incubated at 37°C in a humidified $5\%$ CO2 atmosphere. Cells from passages 3-12 were used for the experiments. Cells were starved in High Glucose-DMEM, $0\%$ FBS and 0.2 mg/mL of the G-418 solution. Afterwards, cells were stimulated with 10 ng/mL IL-1β for 1 h with or without pretreatment for 24 h with the Nrf2 activator tBHQ (20 μM) or for 1 h with apocynin (30 μM), catalase (1,000 U/mL) or the general intracellular TLR4 inhibitor CLI-095 (1 μM). In other set of experiments, cells were stimulated with hydrogen peroxide (H2O2, 100 μM, 1 h) or lipopolysaccharide (LPS, 10 μg/mL, 3 h) in the absence or the presence of tBHQ (20 μM, 24 h).
## Western blot for protein expression analysis
Protein expression was determined by Western blot in whole-cell lysates (15–30 µg) or nuclear and cytosolic fractions (15–20 µg) from VSMC. Nuclear and cytosolic extracts were obtained as previously described (Martín et al., 2012). Proteins were separated by $7.5\%$ or $12\%$ SDS-PAGE and transferred to polyvinyl difluoride membranes that were incubated with antibodies for Nrf2, HO-1, NQO1, SOD1, SOD2, Keap1, TLR4, MyD88, COX-2, p-JNK, p-p38 and p-c-Jun (Table 1). After washing, membranes were then incubated with secondary anti-rabbit (1:2,000; Bio-Rad Laboratories, Hercules, CA, United States) or anti-mouse antibodies (1:4,000; Bio-Rad Laboratories) conjugated to horseradish peroxidase. Proteins were detected using a horseradish peroxidase-luminol/enhancer chemiluminescence system (Bio-Rad Laboratories). The same membrane was used to determine the expression of α-tubulin, p38 and JNK in total extracts or GAPDH and TBP in cytosolic and nuclear extracts, respectively, as loading control by using mouse monoclonal or rabbit polyclonal antibodies (Table 1). Immunoblot signals were quantified using the Image Lab Software version 6.0 (Bio-Rad Laboratories). For protein expression, the ratio between signals on the immunoblot corresponding to the protein studied and that of tubulin, p-38, JNK or TBP was calculated. The protein expression in control cells was assigned the value of 1 to compare with treated cells.
**TABLE 1**
| Antibody | Dilution | Source | Selling company | Catalog number |
| --- | --- | --- | --- | --- |
| Nrf2 | 1:1,000 | Rabbit monoclonal | Cell Signaling Technology (Danvers, MA, United States) | 12721 |
| HO-1 | 1:1,000 | Rabbit polyclonal | ENZO Life Sciences (Lausen, Switzerland) | ADI-SPA-895 |
| NQO1 | 1:500 | Mouse monoclonal | Santa Cruz Biotechnology (Santa Cruz, CA, United States) | sc-32793 |
| SOD1 | 0.05 μg/mL | Rabbit polyclonal | ENZO Life Sciences | ADI-SOD-101 |
| SOD2 | 0.005 μg/mL | Rabbit polyclonal | ENZO Life Sciences | ADI-SOD-111 |
| Keap1 | 1:1,000 | Rabbit monoclonal | Cell Signaling Technology | 8047 |
| TLR4 | 1:500 | Mouse monoclonal | Santa Cruz Biotechnology | sc-293072 |
| MyD88 | 1:1,000 | Rabbit polyclonal | Abcam (Cambridge, United Kingdom) | ab2064 |
| COX-2 | 1:150 | Rabbit polyclonal | Cayman Chemical (Ann Arbor, MI, United States) | 160126 |
| p-JNK | 1:500 | Rabbit polyclonal | Cell Signaling Technology | 9251 |
| JNK | 1:1,000 | Rabbit monoclonal | Cell Signaling Technology | 9258 |
| p-p38 | 1:1,000 | Mouse monoclonal | Cell Signaling Technology | 9216 |
| p38 | 1:2,000 | Mouse monoclonal | Cell Signaling Technology | 9217 |
| p-c-Jun | 1:500 | Rabbit polyclonal | Santa Cruz Biotechnology | sc-16312 |
| Tubulin | 1:20,000 | Mouse monoclonal | Sigma Chemical Co. | T5168 |
| TBP | 1:1,000 | Rabbit polyclonal | Santa Cruz Biotechnology | sc-204 |
| GAPDH | 1:1,000 | Mouse monoclonal | Calbiochem (Temecula, CA, United States) | CB1001 |
## Detection of ROS production: Fluorescence microscopy—Flow cytometry
The oxidative fluorescent dye dihydroethidium (DHE) and the colorant 2′,7′-dichlorofluorescein diacetate (DCFH-DA) were used to evaluate superoxide anion (O2. −) and H2O2 production, respectively. Hydroethidine is able of freely cross cell membranes and oxidize in the presence of O2. − to ethidium bromide, which is trapped inside the cell due to its ability to intercalate into DNA. Ethidium bromide is excited at a wavelength of 546 nm and has an emission spectrum at 600–700 nm. DCFH-DA is a permeable dye able of diffusing through the cell membrane. DCFH-DA is degraded by intracellular esterases to 2′-7′ dichlorofluorescein (DCF), which binds to intracellular H2O2 and emits fluorescence at a wavelength of 535 nm when excited at a wavelength of 485 nm.
## Fluorescence microscopy
Briefly, VSMC were plated onto glass coverslips placed in 12-well plates and cultured as described above. Cells at a confluence of $40\%$–$60\%$ were stimulated with 10 ng/mL IL-1β or 100 μM H2O2 for 1 h in the absence or the presence of 20 μM tBHQ for 24 h. After that, cells were then incubated with 10 μM DHE in cell culture medium for 30 min at 37°C. The images were then acquired using a fluorescence microscope (Zeiss Axioplan 2, Carl Zeiss Microscopy, LLC, Thornwood, NY, United States) and a cool fluorescence colour camera (Leica DFC7000T). The fluorescence intensity values of 10–12 nuclei per experiment were measured using ImageJ software (http://rsb.info.nih.gov/ij). Data were expressed as an increase in the fluorescence intensity related to the control value.
## Flow cytometry
Briefly, VSMC were plated in 12-well plates, cultured as described above, and stimulated with 10 ng/mL IL-1β or 100 μM H2O2 for 1 h in the absence or the presence of 20 μM tBHQ for 24 h or 1 μM CLI-095 for 1 h. Cells were then incubated with 10 μM DHE or 20 μM DCFH-DA in cell culture medium for 30 min at 37°C. Next, cells were trypsinized and collected in tubes for centrifugation. After two washes with PBS, cells in the pellet were resuspended in 500 μl PBS and analysed by flow cytometry using a Beckman Coulter Cytomics FC500 MPL cytometer (Beckman Coulter, Miami, FL, United States). Data were expressed as an increase in the fluorescence intensity related to the control value.
## Lucigenin assay
The superoxide anion generated by nicotinamide adenine dinucleotide phosphate (NADPH) oxidase activity was determined using a chemiluminescence assay using lucigenin and NADPH. For this, VSMC were treated with 10 ng/mL IL-1β or 100 μM H2O2 for 1 h, or 10 μg/mL LPS for 3 h, with or without pretreatment with tBHQ 20 μM for 24 h, and then they were homogenized in a lysis buffer (50 mM KH2PO4, 1 mM ethylene glycol tetra acetic acid—EGTA, 150 mM sucrose, pH = 7.4). The resulting homogenate was then transferred to a 96 wells plate together with the lysis buffer and lucigenin (5 µM). The assay was performed by duplicate. Basal luminescence was measured every 1.8 s during 3 min in a plate luminometer (GloMax®-Multi Detection System: Promega, Madison, WI, United States). The reaction was initiated by adding NADPH (100 µM) to the samples in a final volume of 300 μl. Luminescence was determined every 1.8 s for 3 min. Baseline activity in the absence of NADPH was subtracted. Activity was expressed as relative light units per μg of protein. Protein concentration was determined by using the Micro BCA™ protein assay kit (Thermo Fisher Scientific, Rockford, IL, United States). Data were expressed as an increase in the luminescence related to the control cells.
## In vitro wound healing assay
To evaluate the effect of tBHQ and CLI-095 on IL-1β- and LPS-induced migration, wound healing assay was used. For this, cells were seeded and cultured in 12-well plates. Once the cells reached a confluence of $90\%$, the medium was removed, a wound was made with a P10 pipette tip, and a line was drawn through the centre of the wells, perpendicular to the wound. After two washes with PBS (to wash away any cell debris remaining in the wound area), serum-free medium was added. A picture was taken at time zero at the site of intersection of the line and the wound. Then, VSMC were stimulated with 10 ng/mL IL-1β or 100 μM H2O2 for 24 h in the absence or presence of 20 μM tBHQ or 1 μM CLI-095. After 24 h, we took a picture in the same location. ImageJ software was used to determine the area of wound closure compared to time 0 for the stimulus and with respect to the control situation.
## Cell proliferation assay
3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyltetrazolium bromide (MTT) was used to evaluate VSMC proliferation. For this, cells were seeded in 12-well plates. Once the cells reached a confluence of $50\%$–$60\%$, VSMC were starved in serum-free medium and stimulated with 10 ng/mL IL-1β or 100 μM H2O2 for 24 h in the absence or presence of 20 μM tBHQ or 1 μM CLI-095. Then, the medium was removed and MTT solution was added for 3 h. MTT is transformed into formazan crystals (insoluble in water) by the action of mitochondrial dehydrogenases in living cells. The blue formazan crystals are solubilized with DMSO, resulting in a colorimetric reaction leading to the appearance of a purple coloration. The absorbance was measured at 570 nm using a FLUOstar omega spectrophotometer (BMG Labtech, Germany).
## Vascular function
Some experiments were performed in aortic segments from 3-month C57BL/6 male mice. Thoracic aorta was divided into four segments 2 mm in length, each one incubated for 24 h without or with 10 ng/mL IL-1β, 20 μM tBHQ and IL-1β+tBHQ or IL-1β, 1 μM CLI-095 and IL-1β+CLI-095 in incubation medium (DMEM low glucose supplemented with $1\%$ (v/v) FBS, $1\%$ Penicillin-Streptomycin and $1\%$ Glutamine). Afterwards, segments were transferred to a wire myograph to measure vascular reactivity. After a 30-min equilibration period in oxygenated Krebs-Henseleit solution (KHS, in mM: 115 NaCl, 25 NaHCO3, 4.7 KCl, 1.2 MgSO4.7H2O, 2.5 CaCl2, 1.2 KH2PO4, 11.1 glucose, and 0.01 Na2EDTA) bubbled with a $95\%$ O2-$5\%$ CO2 mixture (pH = 7.4), arterial segments were stretched to their optimal lumen diameter for active tension development. Segment´s contractility was tested by an initial exposure to a high K+ solution (120 mM K+-KHS, which was identical to KHS except that NaCl was replaced by KCl on an equimolar basis). Then, endothelium-dependent vasodilation was analysed by performing a single cumulative concentration-response curve to ACh (1 nM—10 μM) in arteries precontracted with phenylephrine (Phe) at a concentration that produce approximately $50\%$ of the contraction induced by K+-KHS.
## Statistical analysis
All values are expressed as mean ± standard error (S.E.M.); n denotes the number of animals or the number of different cultures (each one obtained from one different passage) used in each experiment. For cell culture experiments, data are expressed as n-fold increase relative to the average value of controls in each plate or each blot. For vascular reactivity experiments, vasodilator responses induced by ACh were expressed as the % of the previous tone in each case.
Results were analysed by using paired Student’s t-test, one-way or two-way ANOVA followed by Bonferroni post-test by using GraphPad Prism Software (San Diego, CA, United States). Values were considered to be significant when p-value is less than 0.05.
## Drugs/chemicals and antibodies
FBS, Penicillin-Streptomycin, Glutamine, IL-1β, tBHQ, LPS, H2O2, apocynin, catalase, DHE, DCFH-DA, MTT, Phe, ACh, Lucigenin and NADPH were obtained from Sigma Chemical Co. CLI-095 was obtained from InvivoGen (San Diego, CA, United States). All drugs were dissolved in distilled water except for tBHQ, CLI-095, DHE, DCFH-DA, which were dissolved in DMSO, and Lucigenin, which was dissolved in acetic acid. Neither DMSO nor acetic acid have any effect on VSMC.
## TLR4 pathway is involved in Interleukin-1β effects
As mentioned, IL-1β is associated to increased ROS production; the primary enzymatic sources of cardiovascular ROS are NADPH oxidases. Accordingly, IL-1β (10 ng/mL, 1 h) increased the NADPH oxidase activity (Supplementary Figure S1A) and the subsequent production of hydrogen peroxide and superoxide anion in VSMC (Supplementary Figures S1B–D). In addition, treatment of VSMC with IL-1β induced the nuclear translocation of the AP-1 subunit phospho-c-Jun and the COX-2 protein expression (Supplementary Figure S2). In this study we also prove that oxidative stress is involved in the proinflammatory effects mediated by this cytokine. Thus, IL-1β-elicited nuclear translocation of p-c-Jun was reduced by the antioxidant apocynin (30 µM) and the H2O2 scavenger catalase (1,000 U/mL) (Supplementary Figures S2A, B). Similarly, both drugs reduced the IL-1β-induced COX-2 expression (Supplementary Figures S2C, D). Neither apocynin nor catalase modified the nuclear p-p-Jun or COX-2 expressions (Supplementary Figure S2). These results confirm that ROS participate in IL-1β-induced activation of AP-1 and COX-2 in VSMC.
Next, we analysed the effect of IL-1β on the TLR4 signalling pathway. Treatment of mouse aortic VSMC with IL-1β did not modify TLR4 expression (Figure 1A); however, it increased the expression of its adapter protein MyD88 (Figure 1B). The effect of IL-1β on MyD88 protein expression was prevented by using the general intracellular TLR4 inhibitor CLI-095 (1 μM) (Figure 1B), suggesting that the TLR4/MyD88-dependent signalling pathway was activated, at least in part, by IL-1β in VSMC. Next, we analysed the contribution of TLR4 to the observed effects of IL-1β. The inhibition of TLR4 by CLI-095 reduced the increase in COX-2 expression induced by IL-1β (Figure 1C). Besides, CLI-095 also reduced the IL-1β-increased O2.- and H2O2 production (Figures 1D, E). CLI-095 alone did affect neither COX-2 and MyD88 expressions nor O2 − and H2O2 production (Figures 1B, C). Moreover, we found that IL-1β (10 ng/mL, 24 h) increased VSMC migration, and this effect was reduced by blocking the TLR4 pathway; CLI-095 did not affect cell migration (Figure 1F). The IL-1β-induced VSMC proliferation was also prevented by CLI-095 (results not shown).
**FIGURE 1:** *Role of TLR4 pathway on the IL-1β-induced effects. Effect of Interleukin-1β (IL-1β, 10 ng/mL) on TLR4 expression (A). Effect of CLI-095 (1 µM), IL-1β and CLI-095 + IL-1β on myeloid differentiation factor 88 (MyD88) and cyclooxygenase-2 (COX-2) protein expression (B,C). Effect of CLI-095 on IL-1β-induced superoxide anion (O2
.-) and hydrogen peroxide (H2O2) production evaluated by flow cytometry (D,E) and cell migration (F) in vascular smooth muscle cells. Representative blots are shown in upper panels. Images of cell migration by wound healing assay are included; bar scale represents 50 μm. *p < 0.05 vs. control; #p < 0.05 vs. IL-1β by Student’s t-test (n = 5–12). Effect of CLI-095, IL-1β and CLI-095 + IL-1β on acetylcholine (ACh)-induced relaxation in mice aortic segments precontracted with phenylephrine (G). *p < 0.05 vs. control; #p < 0.05 vs. IL-1β by two-way ANOVA followed by Bonferroni’s post test (n = 5). See methods for the incubation times.*
Thereafter, whether TLR4 inhibition interferes with the IL-1β effect on vasodilator responses were also analysed. Incubation of mouse aortas with IL-1β (10 ng/mL, 24 h) induced endothelial dysfunction, while cotreatment with the TLR4 antagonist improved ACh-induced endothelium-dependent vasodilatation (Figure 1G); however, IL-1β did not affect the response to K-KHS, regardless of the presence or absence of CLI-095 (data not shown). Furthermore, CLI-095 alone did not modify ACh-induced relaxation (Figure 1G). Taken together, all these results indicate that the TLR4 pathway contributes to the IL-1β-mediated effects.
## Interleukin-1β-induced Nrf2 activation is dependent on reactive oxygen species production
Later, we analysed the effect of IL-1β on the Nrf2 pathway. Treatment of VSMC with IL-1β (10 ng/mL, 1 h) increased the Nrf2 nuclear translocation (Figure 2A); in addition, IL-1β also enhanced the expression of its downstream proteins, HO-1, NQO1, SOD1, and SOD2 (Figures 2B–E). Importantly, IL-1β reduced the expression of the Nrf2 inhibitory protein Keap1 (Figure 2F).
**FIGURE 2:** *Effect of IL-1β on the Nrf2 pathway. Effect of Interleukin-1β (IL-1β, 10 ng/mL, 1 h) on the nuclear Nrf2 (nuclear factor-erythroid 2-related factor 2) protein expression (A) in vascular smooth muscle cells; a representative blot of the cytosolic (Cy) and nuclear (Nu) expression is also shown; nuclear TATA-binding protein (TBP) and cytosolic GAPDH expressions are also shown to guarantee the successful cellular fractioning. Effect of IL-1β on heme oxygenase-1 [HO-1, (B)], NAD(P)H:quinone oxidoreductase 1 [NQO1, (C)], superoxide dismutase 1 [SOD1, (D)], superoxide dismutase 2 [SOD2, (E)] and Keap1 (F) protein expression. Representative blots are shown in upper panels; the same loading control for SOD1 and SOD2 protein expression was used. Tert-butylhydroquinone (tBHQ, 20 μM, 24 h) was used as positive control. *p < 0.05 vs. control by Student’s t-test (n = 6–30).*
As described above, IL-1β increased the NADPH oxidase activity and the subsequent ROS production in VSMC. Therefore, the involvement of oxidative stress in the effects of IL-1β on Nrf2 pathway was studied. IL-1β-induced Nrf2 nuclear translocation was reduced by apocynin (30 µM) and catalase (1,000 U/mL) (Figures 3A, B); similarly, apocynin reduced the effect of IL-1β on HO-1 and SOD2 protein expression (Figures 3C, E). However, catalase did reduce the IL-1β-induced expression neither of HO-1 nor of SOD2 (Figures 3D, F). Neither apocynin nor catalase modified the proteins studied (Figure 3). In addition, treatment of VSMC with H2O2 (100 μM, 1 h) increased the Nrf2 nuclear translocation (Supplementary Figure S3A) as well as the protein expression of HO-1 and NQO1 (Supplementary Figures S3B, C). All together, these results show that ROS participate in Nrf2 activation by IL-1β.
**FIGURE 3:** *Role of oxidative stress in the IL-1β-induced effects on the Nrf2 pathway. Effect of apocynin (Apo, 30 μM, 2 h) and catalase (Cat, 1,000 U/mL, 2 h) on the nuclear Nrf2 (nuclear factor-erythroid 2-related factor 2) protein expression (A,B) induced by Interleukin-1β (IL-1β, 10 ng/mL, 1 h) in vascular smooth muscle cells; a representative blot of the cytosolic (Cy) and nuclear (Nu) expression is also shown; nuclear TATA-binding protein (TBP) and cytosolic GAPDH expressions are also shown to guarantee the successful cellular fractioning. Effect of Apo and Cat on heme oxygenase-1 [HO-1, (C,D)] and superoxide dismutase 2 [SOD2, (E,F)] protein expressions induced by IL-1β. Representative blots are shown in upper panels. Tert-butylhydroquinone (tBHQ, 20 μM, 24 h) was used as positive control. *p < 0.05 vs. control; #p < 0.05 vs. IL-1β by Student’s t-test (n = 6–10).*
## Nrf2 activation downregulates Interleukin-1β prooxidant and proinflammatory effects
We next analysed the effect of Nrf2 activation on IL-1β-induced effects. For this, the Nrf2 activator tBHQ was used. As expected, tBHQ (20 µM) increased the nuclear translocation of Nrf2 as well as the protein levels of HO-1 and NQO1, while it reduced the expression of Keap1 (Figure 4). Furthermore, pretreatment of VSMC for 24 h with the Nrf2 activator tBHQ increased the IL-1β-induced Nrf2 nuclear translocation as well as the HO-1 and NQO1 protein expression, although the levels reached were similar to those of tBHQ alone (Figures 4A–C). Additionally, tBHQ plus IL-1β reduced Keap1 protein expression, being this effect similar to that induced by either IL-1β or tBHQ alone (Figure 4D).
**FIGURE 4:** *Modulation by Nrf2 activation of the IL-1β-induced effects on the Nrf2 pathway. Effect of tert-butylhydroquinone (tBHQ, 20 μM, 24 h), Interleukin-1β (IL-1β, 10 ng/mL, 1 h) and tBHQ + IL-1β on the nuclear Nrf2 (nuclear factor-erythroid 2-related factor 2) protein expression (A) in vascular smooth muscle cells; a representative blot of the cytosolic (Cy) and nuclear (Nu) expression is also shown; nuclear TATA-binding protein (TBP) and cytosolic GAPDH expressions are also shown to guarantee the successful cellular fractioning. Effect of tBHQ, IL-1β and tBHQ + IL-1β on heme oxygenase-1 [HO-1, (B)], NAD(P)H:quinone oxidoreductase 1 [NQO1, (C)] and Keap1 (D) protein expressions. Representative blots are shown in upper panels. *p < 0.05 vs. control; #p < 0.05 vs. IL-1β by Student’s t-test (n = 6–10).*
Next, we investigated the effect of Nrf2 activation on the IL-1β-elicited prooxidant effects. tBHQ did modify neither the basal NADPH oxidase activity nor the production of both hydrogen peroxide and superoxide anion (Figure 5); however, it reduced the increased IL-1β-induced NADPH oxidase activity (Figure 5A) as well as the IL-1β-elicited increase of both H2O2 and O2. − production (Figures 5B–D). In addition, we evaluated whether Nrf2 activation affects the H2O2-elicited effects. Pretreatment of VSMC for 24 h with 20 µM tBHQ increased the H2O2-induced Nrf2 nuclear translocation (Supplementary Figure S4A) and the HO-1 and NQO1 protein expression (Supplementary Figures S4B, C), although these effects were similar to those caused by tBHQ alone. Furthermore, tBHQ reduced the H2O2-elicited NADPH oxidase activity and O2. − production (Supplementary Figures S4D–F).
**FIGURE 5:** *Modulation by Nrf2 activation of the IL-1β-induced effects on oxidative stress. Effect of tert-butylhydroquinone (tBHQ, 20 μM, 24 h), Interleukin-1β (IL-1β, 10 ng/mL, 1 h) and tBHQ + IL-1β on NAPDH oxidase activity (A), hydrogen peroxide [H2O2, (B)] and superoxide anion (O2
.-) production evaluated by flow cytometry (C) and by fluorescence microscopy (D) in vascular smooth muscle cells; representative fluorescent photomicrographs are also shown; images were captured with a fluorescence microscope; bar scale represents 100 μm. *p < 0.05 vs. control; #p < 0.05 vs. IL-1β by Student’s t-test (n = 6–10).*
Thereafter, we explored the effect of Nrf2 activation on IL-1β-elicited proinflammatory effects. Pretreatment of VSMC for 24 h with the Nrf2 activator tBHQ did not modify the protein expression of MyD88, p-p38, p-JNK2, nuclear p-c-Jun or COX-2, but it reduced the IL-1β-induced increase of these proteins, as shown in Figure 6. On the other hand, we found that bacterial LPS, the main TLR4 ligand, increased MyD88 protein expression, and this increase was reduced by tBHQ (Supplementary Figure S5A); in addition, tBHQ reduced the LPS-elicited increase of NADPH oxidase activity and cell migration (Supplementary Figures S5B, C). Our results, therefore, suggest that Nrf2 activation protects against the detrimental prooxidant and proinflammatory effects of IL-1β, probably through TLR4 pathway modulation.
**FIGURE 6:** *Modulation by Nrf2 activation of IL-1β-induced-effects on proinflammatory markers. Effect of tert-butylhydroquinone (tBHQ, 20 μM, 24 h), Interleukin-1β (IL-1β, 10 ng/mL, 1 h) and tBHQ + IL-1β on myeloid differentiation factor 88 [MyD88, (A)], p-p38 (B), p-JNK2 (C), nuclear p-c-Jun (D) and cyclooxygenase-2 [COX-2, (E)] protein expression in vascular smooth muscle cells. Representative blots are shown in upper panels. A representative blot of the cytosolic (Cy) and nuclear (Nu) expression of p-c-Jun is also shown (D); nuclear TATA-binding protein (TBP) and cytosolic GAPDH expressions are also shown to guarantee the successful cellular fractioning *p < 0.05 vs. control; #p < 0.05 vs. IL-1β by Student’s t-test (n = 5–12).*
## Nrf2 activation protects against Interleukin-1β induced VSMC migration and proliferation, and endothelial dysfunction
The activator of Nrf2 alone did modify neither cell migration nor proliferation; however, it reduced the VSMC migration and proliferation induced by incubation for 24 h with 10 ng/mL IL-1β (Figures 7A, B). In addition, tBHQ (20 μM) also reduced the cell migration elicited by H2O2 (100 μM, 24 h), as shown in Figure 7C. In contrast, H2O2 reduced cell proliferation, which was further decreased by the Nrf2 activator tBHQ (Figure 7D).
**FIGURE 7:** *Modulation by Nrf2 activation of the IL-1β- and H2O2-induced effects on cell migration and proliferation as well as on IL-1β-induced endothelial dysfunction. Effect of tert-butylhydroquinone (tBHQ, 20 μM, 24 h), Interleukin-1β (IL-1β, 10 ng/mL, 1 h) and tBHQ + IL-1β (A,B) and tBHQ, hydrogen peroxide (H2O2, 100 μM, 1 h) and tBHQ + H2O2
(C,D) on cell migration and proliferation in vascular smooth muscle cells. Images of cell migration by wound healing assay are included; bar scale represents 50 μm. *p < 0.05 vs. control; #p < 0.05 vs. IL-1β by Student’s t-test (n = 6–10). Effect of tBHQ, IL-1β (10 ng/mL, 24 h) and tBHQ + IL-1β on acetylcholine (ACh)-induced relaxation in mice aortic segments precontracted with phenylephrine (E). *p < 0.05 vs. control; #p < 0.05 vs. IL-1β by two-way ANOVA followed by Bonferroni’s post test (n = 4).*
Finally, the effect of Nrf2 activation on IL-1β-induced endothelial dysfunction was also analysed. As shown in Figure 7E, tBHQ did not modify the vasodilator response induced by ACh, but it partially prevented the IL-1β-induced endothelial dysfunction. Neither tBHQ alone nor its combination with IL-1β did modify the contraction of K-KHS (results not shown).
In summary, these results indicate a beneficial role for Nrf2 activation on vascular alterations such as VSMC migration and proliferation and the endothelial dysfunction induced by IL-1β.
## Discussion
In the present study we found that TLR4 pathway contributes to the prooxidant and proinflammatory effects induced by IL-1β in VSMC, being these effects reduced by activation of the redox-sensitive transcription factor Nrf2. Thus, Nrf2 activation prevents the detrimental effects of this cytokine on VSMC migration, proliferation and vascular endothelial dysfunction, likely by interfering with the IL-1β-induced TLR4 pathway.
IL-1β is one of the most relevant proinflammatory cytokines belonging to the IL-1 family (Abbate et al., 2020). The binding of IL-1β to its receptor, mainly through MyD88, activates signalling pathways involving MAPKs as well as transcription factors such as NF-κB and AP-1 (Krishnan et al., 2014). IL-1β levels are increased in several pathologies that involve oxidative stress and inflammation, including some CVDs such as atherosclerosis and hypertension (Krishnan et al., 2014; Mai and Liao, 2020; Melton and Qiu, 2021). For instance, it has been described that IL-1β contributes to atherogenesis by initiation, formation, and growth of the atheroma plaques (Abbate et al., 2020). Accordingly, Bhaskar et al. [ 2011] have shown that blocking IL-1β-mediated activity reduces plaque formation and subsequent atherosclerosis progression. Additionally, IL-1β also participates in the pathogenesis of hypertension (Melton and Qiu, 2021). Thus, inhibition of IL-1β activity reduces blood pressure in models of DOCA/salt- (Ling et al., 2017) and angiotensin II-induced hypertension in mice (Akita et al., 2021). It is widely known that IL-1β increases ROS production, thus contributing to the progression of vascular damage related to those CVDs (Shen et al., 2016). In agreement with previous studies (Martín et al., 2012), we observed that IL-1β increased the NADPH oxidase activity and the subsequent production of O2. − and H2O2 in VSMC. Furthermore, IL-1β contributes to the development of CVDs not only by inducing oxidative stress, but also due to its proinflammatory effects by promoting the expression of a variety of inflammatory mediators. Thus, IL-1β induces the activation of the vascular cell adhesion molecule 1 (VCAM-1), which recruits inflammatory cells to the VSMC at the atherosclerotic lesion (Wang et al., 1995) and promotes vascular calcification (Sánchez-Duffhues et al., 2019). In a previous report, we have shown that IL-1β also increases COX-2 expression in VSMC from SHR (Martín et al., 2012). COX-2 is the inducible isoform of COX enzyme that catalyses the formation of prostaglandins, whose expression and activation is induced by oxidative stress and inflammatory stimuli (Zhao et al., 2021). COX-2 acts as an inflammatory mediator by releasing vasoconstrictors which may induce cardiovascular dysfunction (Félétou et al., 2011; Bomfim et al., 2012; Zhao et al., 2021), playing a critical role in the endothelium-dependent contraction related to CVDs progression (Zhao et al., 2021). It has been described that AP-1, which can be activated by IL-1β (Shen et al., 2016), contributes to COX-2 expression in VSMC (Palacios-Ramírez et al., 2019). In our study, we found that IL-1β induces the nuclear translocation of the AP-1 subunit phospho-c-Jun as well as COX-2 protein expression in VSMC. Additionally, our results using the antioxidant apocynin and the hydrogen peroxide scavenger catalase confirm the role of oxidative stress in the IL-1β-induced proinflammatory effects.
TLR4 contribution to the vascular damage by activating inflammatory signalling pathways involving oxidative stress mechanisms, and its closely association with the pathogenesis of CVDs such as atherosclerosis and hypertension, have been described (De Batista et al., 2014; Yang et al., 2014; Hernanz et al., 2015; Qi et al., 2021). In addition, a possible connection between IL-1β and TLR4 has been suggested; thus, treatment with this cytokine activates the TLR4/NF-κB signalling via MyD88-dependent and -independent pathways (Liu et al., 2014; Gu et al., 2017), and this effect was abolished after TLR4 inhibition (Xu et al., 2020). We found that treatment of VSMC with IL-1β did not modify TLR4 expression, in agreement to that found by Chen et al. [ 2013]. However, IL-1β increased the expression of its adapter protein MyD88. This effect of IL-1β on MyD88 protein expression was prevented by using the specific intracellular TLR4 inhibitor CLI-095, which disrupts the interactions of TLR4 with its adaptors (Matsunaga et al., 2011), suggesting that the TLR4/MyD88-dependent signalling pathway was activated by this cytokine in VSMC. As mentioned above, IL-1β increases COX-2 expression in VSMC. Activation of the TLR4 pathway mediates the transcription of COX-2, a well-known TLR4 target protein in VSMC (Bomfim et al., 2012; Biancardi et al., 2017). Accordingly, we found that the increase in COX-2 expression induced by IL-1β was reduced by CLI-095. In addition, the IL-1β-elicited increase of both O2. − and H2O2 production was also reduced by the TLR4 inhibitor.
Under normal conditions, VSMC, the major component of the vascular wall, have a characteristic contractile phenotype; however, under oxidative stress conditions and excessive inflammation these cells turn to a dedifferentiated phenotype which makes them more prone to migrate and proliferate (Badran et al., 2020; Qi et al., 2021). Several studies have reported that inflammatory mediators such as IL-1β induces VSMC migration and proliferation (Martín et al., 2012; Wang et al., 2012; Eun et al., 2015), then promoting vascular damage development (Badran et al., 2020). Downregulation or inhibition of TLR4 has been shown to reduce angiotensin II-induced VSMC migration and proliferation (De Batista et al., 2014; Qi et al., 2021), and this would contribute to explain the beneficial effects of TLR4 blockade in vascular remodelling and mechanical alterations observed in hypertension (Hernanz et al., 2015). Here, we found that inhibition of TLR4 reduces IL-1β-induced VSMC migration and proliferation. Cytokines also induce cardiovascular damage by impairment of endothelial-dependent relaxations (Briones et al., 2005; Jimenez-Altayó et al., 2006), and its inhibition in many cases ameliorates vascular endothelial dysfunction, as described by Vallejo et al. [ 2014] for IL-1β. Here, we confirm that IL-1β impairs endothelial-dependent relaxation. More importantly, the general TLR4 inhibitor CLI-095 prevented the deleterious effect of IL-1β on endothelial function. The beneficial effect of inhibiting TLR4 on endothelium-dependent vasodilation has been described by our group and others in several models of hypertension (De Batista et al., 2014; Hernanz et al., 2015) and atherosclerosis (Chen et al., 2020). Although we have not analysed the mechanisms by which TLR4 blockade improves the IL-1β-induced endothelial dysfunction, the reduction of oxidative stress might contribute to this effect, as has been reported in the hypertension-associated endothelial dysfunction (Hernanz et al., 2015). All these results reveal an interaction between IL-1β and TLR4, although the exact mechanism for this needs further investigation.
As mentioned above, IL-1β increases ROS production. An excessive ROS production leads to a disruption of redox homeostasis and molecular damage (Xia et al., 2017; Griendling et al., 2021). To counteract high ROS levels, cells may activate antioxidant mechanisms, being the redox-sensitive transcription factor Nrf2 one of these mechanisms. Nrf2 protects against oxidative stress by inducing the expression of antioxidant proteins and phase II detoxification enzymes, including HO-1, NQO1 and SOD, in response to changes in the intracellular redox balance (Yamamoto et al., 2018). HO-1 overexpression in atherosclerotic lesions is considered to be protective (Kishimoto et al., 2019), whereas inhibition of this enzyme resulted in the progression of atherosclerosis (Ishikawa et al., 2012). Furthermore, Nrf2 activation may have a protective role in hypertension (Lopes et al., 2015; Guzik and Touyz, 2017; Xia et al., 2017; Wang et al., 2018). Therefore, activation of Nrf2 is a potential target against CVDs. In our study, we found that treatment of VSMC with IL-1β increased Nrf2 nuclear translocation and enhanced the expression of its downstream proteins, simultaneously reducing the expression of the Nrf2 repressor protein Keap1. This Nrf2 activation could be explained as a compensatory mechanism of the cell against the deleterious effect of this cytokine. Of note, the use of apocynin suggests that oxidative stress is involved in Nrf2 activation by IL-1β. Consistent with this, exogenous hydrogen peroxide also increased the nuclear translocation of Nrf2 as well as the protein expression of HO-1 and NQO1. It is known that oxidative modification of Keap1 cysteines results in a conformational change that leads to the detachment of Nrf2 from Keap1 and the inhibition of its ubiquitination; however, oxidative stress, by increasing the downstream Nrf2-driven gene p62/SQSTM1, activates specific autophagy of Keap1, and thus can modify its expression (Katsuragi et al., 2016).
tBHQ is a synthetic phenolic antioxidant widely used as selective Nrf2 activator. tBHQ acts modifying thiol groups of cysteines on the Keap1 protein, which induces a conformational change in Keap1; this leads to the release of Nrf2 from its repressor, therefore allowing induction of antioxidant and anti-inflammatory responses by this transcription factor (Zhang and Hannink, 2003). As expected, after VSMC incubation with tBHQ, increased nuclear Nrf2 translocation was found; in addition, tBHQ enhanced the protein expression of HO-1 and NQO1 while reduced Nrf2 inhibitor Keap1 expression. Pretreatment of IL-1β-stimulated VSMC with tBHQ induced similar effects on Nrf2 pathway to that induced by tBHQ alone; in addition, the effects of coincubation of VSMC with H2O2 and tBHQ on Nrf2 translocation and the expression of its downstream antioxidant enzymes HO-1 and NQO1 were also similar to incubation with tBHQ alone, indicating that the observed effects are those induced by the Nrf2 activator. On the other hand, we found reduced IL-1β-induced NADPH oxidase activity as well as IL-1β-mediated O2. − and H2O2 production in cells pretreated with tBHQ, which denotes a protective role of Nrf2 activation against the prooxidant effects of IL-1β. It is known that H2O2 activates NADPH oxidase leading to further O2. − production (Li et al., 2001; Hernanz et al., 2015). Herein we confirm these results and, consistent with the antioxidant effect of tBHQ, we observed that this activator also reduces the increased NADPH oxidase activity and the subsequent O2. − production induced by the exogenous addition of H2O2 in VSMC.
In addition to the protective effect against oxidative stress, Nrf2 also has anti-inflammatory properties (Alonso-Piñeiro et al., 2021). Thus, activators of Nrf2 inhibit the transcription of proinflammatory cytokines and they are used as anti-inflammatory drugs (Muri et al., 2020). Therefore, the result of Nrf2 activation on the observed IL-1β-induced proinflammatory effects was analysed. Although tBHQ alone did not affect the expression of the proinflammatory proteins, it reduced the IL-1β-induced MyD88 expression as well as the increased phosphorylation of MAPKs (p-p38 and p-JNK$\frac{1}{2}$), which likely reduced p-c-Jun and COX-2 expression. This is consistent with a previous report showing that activation of HO-1 reduced AP-1/COX-2 expression (Subedi et al., 2019). Our results indicate that Nrf2 activation keeps against the detrimental proinflammatory effects of IL-1β. Previously we have shown that oxidative stress plays a role in the proinflammatory effects induced by IL-1β in VSMC; therefore, the antioxidant properties of tBHQ would contribute to explain the beneficial effect of Nrf2 activation on the proinflammatory pathways induced by the cytokine.
On the other hand, we found that Nrf2 activation by tBHQ reduces VSMC migration and proliferation induced by IL-1β, suggesting a protective role of this transcription factor on the cytokine-induced vascular remodelling. The fact that tBHQ also reduces the H2O2-induced cell migration allows us to propose that the antioxidant properties derived from Nrf2 activation would contribute to this effect. By contrast, H2O2 reduces VSMC proliferation, in agreement to that found by Nickenig et al. [ 2002], and this effect was further reduced by Nrf2 activation. Differences in the source of ROS, either endogenous or exogenous, as well as in the concentration might explain the different effects of ROS in cell proliferation.
Besides the previously mentioned action on the detrimental prooxidant and proinflammatory effects of IL-1β, tBHQ partially prevents the IL-1β-induced endothelial dysfunction. This is consistent with a previous study by Wang et al. [ 2018], which describes that tBHQ prevented angiotensin II-induced endothelial dysfunction by a Nrf2 dependent mechanism. Similarly, Lopes et al. [ 2015] found that sulforaphane corrected the impaired endothelial function in SHRSP.
Finally, we found that activation of Nrf2 with tBHQ reduces the effects of the TLR4 ligand LPS on MyD88 expression, NADPH oxidase activity and cell migration. These results, together to the previously mentioned reduction of IL-1β MyD88 expression by tBHQ, allow us to propose that protection of Nrf2 activation against the detrimental prooxidant and proinflammatory effects of IL-1β, are due, at least in part, to the interference with the TLR4 pathway activated by the cytokine, although further experiments are needed to confirm whether the signalling pathway involves MyD88-dependent or -independent mechanisms.
In summary, the present study demonstrates that IL-1β elicits oxidative stress, inflammation, cell migration and proliferation, as well as endothelial dysfunction, by mechanisms involving its relationship with TLR4 pathway, although the exact mechanism by which IL-1β interacts with this receptor needs further investigation; however, we cannot rule out that some observed effects of IL-1β might also be produced by acting on its canonical IL-1R receptor (Figure 8). Moreover, our results reveal that activation of the redox-sensitive transcription factor Nrf2 induces vascular protection against IL-1β deleterious effects, at least in part, through interference with TLR4 pathway. Thus, by reducing TLR4 signalling pathway, Nrf2 activation reduces IL-1β-induced ROS generation and the increased expression of proinflammatory markers, in which oxidative stress participates. Furthermore, the activation of this transcription factor also prevents IL-1β-induced VSMC migration and proliferation, as well as the IL-1β-induced endothelial dysfunction (Figure 8). Therefore, our observations allow as to propose that activation of antioxidant and anti-inflammatory mechanisms such as Nrf2 might be an important therapeutic target in pathological conditions with oxidative and inflammatory components. However, the clinical application of Nrf2 activators for treatment of CVDs should be deeply analysed.
**FIGURE 8:** *Proposed mechanism by which Nrf2 activation protects against vascular IL-1β effects. IL-1β increases oxidative stress and induces inflammation, cell migration and proliferation, as well as endothelial dysfunction, partially by a mechanism involving its interaction with TLR4 pathway. Nrf2 activation by tBHQ protects against these IL-1β vascular deleterious effects. The dashed lines indicate additional mechanisms unexplored in our study. IL-1β, Interleukin-1β; TLR4, toll-like receptor 4; MyD88, myeloid differentiation factor 88; ROS, reactive oxygen species; O2
.
−, superoxide anion; H2O2, hydrogen peroxide; tBHQ, tert-butylhydroquinone; Nrf2, nuclear factor-erythroid 2-related factor 2; HO-1, heme oxygenase-1; NQO1; NAD(P)H:quinone oxidoreductase 1; MAPK, mitogen-activated protein kinase; AP-1, activator protein-1; COX-2, cyclooxygenase-2.*
## Data availability statement
The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.
## Ethics statement
The animal study was reviewed and approved by Ethical Committee of Research of the Universidad Autónoma de Madrid.
## Author contributions
AM, RH, MA designed the research study; ZG-C, AM, RH, MM-C, MB performed the experimental work; ZG-C, AM, RH, MA wrote the paper; all authors approved the MS.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2023.1058488/full#supplementary-material
## References
1. Abbate A., Toldo S., Marchetti C., Kron J., Van Tassell B. W., Dinarello C. A.. **Interleukin-1 and the inflammasome as therapeutic targets in cardiovascular disease**. *Circ. Res.* (2020) **126** 1260-1280. DOI: 10.1161/CIRCRESAHA.120.315937
2. Akita K., Isoda K., Ohtomo F., Isobe S., Niida T., Sato-Okabayashi Y.. **Blocking of interleukin-1 suppresses angiotensin II-induced renal injury**. *Clin. Sci. (Lond)* (2021) **135** 2035-2048. DOI: 10.1042/CS20201406
3. Alonso-Piñeiro J. A., Gonzalez-Rovira A., Sánchez-Gomar I., Moreno J. A., Durán-Ruiz M. C.. **Nrf2 and heme oxygenase-1 involvement in atherosclerosis related oxidative stress**. *Antioxidants* (2021) **10** 1463. DOI: 10.3390/antiox10091463
4. Badran A., Nasser S. A., Mesmar J., El-Yazbi A. F., Bitto A., Fardoun M. M.. **Reactive oxygen species: Modulators of phenotypic switch of vascular smooth muscle cells**. *Int. J. Mol. Sci.* (2020) **21** 8764. DOI: 10.3390/ijms21228764
5. Bhaskar V., Yin J., Mirza A. M., Phan D., Vanegas S., Issafras H.. **Monoclonal antibodies targeting IL-1 beta reduce biomarkers of atherosclerosis**. *Atherosclerosis* (2011) **216** 313-320. DOI: 10.1016/j.atherosclerosis.2011.02.026
6. Biancardi V. C., Bomfim G. F., Reis W. L., Al-Gassimi S., Nunes K. P.. **The interplay between Angiotensin II, TLR4 and hypertension**. *Pharmacol. Res.* (2017) **120** 88-96. DOI: 10.1016/j.phrs.2017.03.017
7. Bomfim G. F., Dos Santos R. A., Oliveira M. A., Giachini F. R., Akamine E. H., Tostes R. C.. **Toll-like receptor 4 contributes to blood pressure regulation and vascular contraction in spontaneously hypertensive rats**. *Clin. Sci. (Lond)* (2012) **122** 535-543. DOI: 10.1042/CS20110523
8. Briones A. M., Salaices M., Vila E.. **Ageing alters the production of nitric oxide and prostanoids after IL-1beta exposure in mesenteric resistance arteries**. *Mech. Ageing Dev.* (2005) **126** 710-721. DOI: 10.1016/j.mad.2005.01.006
9. Chen G., Chen X. L., Xu C. B., Lin J., Luo H. L., Xie X.. **Toll-like receptor protein 4 monoclonal antibody inhibits mmLDL-induced endothelium-dependent vasodilation dysfunction of mouse mesenteric arteries**. *Microvasc. Res.* (2020) **127** 103923. DOI: 10.1016/j.mvr.2019.103923
10. Chen T. M., Li J., Liu L., Fan L., Li X. Y., Wang Y. T.. **Effects of heme oxygenase-1 upregulation on blood pressure and cardiac function in an animal model of hypertensive myocardial infarction**. *Int. J. Mol. Sci.* (2013) **14** 2684-2706. DOI: 10.3390/ijms14022684
11. De Batista P. R., Palacios R., Martín A., Hernanz R., Médici C. T., Silva M. A. S. C.. **Toll-like receptor 4 upregulation by angiotensin II contributes to hypertension and vascular dysfunction through reactive oxygen species production**. *PLoS ONE* (2014) **9** e104020. DOI: 10.1371/journal.pone.0104020
12. Eun S. Y., Ko Y. S., Park S. W., Chang K. C., Kim H. J.. **IL-1β enhances vascular smooth muscle cell proliferation and migration via P2Y2 receptor-mediated RAGE expression and HMGB1 release**. *Vasc. Pharmacol.* (2015) **72** 108-117. DOI: 10.1016/j.vph.2015.04.013
13. Félétou M., Huang Y., Vanhoutte P. M.. **Endothelium-mediated control of vascular tone: COX-1 and COX-2 products**. *Br. J. Pharmacol.* (2011) **164** 894-912. DOI: 10.1111/j.1476-5381.2011.01276.x
14. Fiorelli S., Porro B., Cosentino N., Di Minno A., Manega C. M., Fabbiocchi F.. **Activation of Nrf2/HO-1 pathway and human atherosclerotic plaque vulnerability: An**. *Cells* (2019) **8** 356. DOI: 10.3390/cells8040356
15. Griendling K. K., Camargo L. L., Rios F. J., Alves-Lopes R., Montezano A. C., Touyz R. M.. **Oxidative stress and hypertension**. *Circ. Res.* (2021) **128** 993-1020. DOI: 10.1161/CIRCRESAHA.121.318063
16. Gu H., Jiao Y., Yu X., Li X., Wang W., Ding L.. **Resveratrol inhibits the IL-1β-induced expression of MMP-13 and IL-6 in human articular chondrocytes viaTLR4/MyD88-dependent and-independent signaling cascades**. *Int. J. Mol. Med.* (2017) **39** 734-740. DOI: 10.3892/ijmm.2017.2885
17. Guzik T. J., Touyz R. M.. **Oxidative stress, inflammation, and vascular aging in hypertension**. *Hypertension* (2017) **70** 660-667. DOI: 10.1161/HYPERTENSIONAHA.117.07802
18. Hernanz R., Martínez-Revelles S., Palacios R., Martín A., Cachofeiro V., Aguado A.. **Toll-like receptor 4 contributes to vascular remodelling and endothelial dysfunction in angiotensin II-induced hypertension**. *Br. J. Pharmacol.* (2015) **172** 3159-3176. DOI: 10.1111/bph.13117
19. Ishikawa K., Navab M., Lusis A. J.. **Vasculitis, atherosclerosis, and altered HDL composition in heme-oxygenase-1-knockout mice**. *Int. J. Hypertens.* (2012) **2012** 948203. DOI: 10.1155/2012/948203
20. Jiménez-Altayó F., Briones A. M., Giraldo J., Planas A. M., Salaices M., Vila E.. **Increased superoxide anion production by interleukin-1beta impairs nitric oxide-mediated relaxation in resistance arteries**. (2006) **316** 42-52. DOI: 10.1124/jpet.105.088435
21. Katsuragi Y., Ichimura Y., Komatsu M.. **Regulation of the Keap1-Nrf2 pathway by p62/SQSTM1**. *Curr. Opin. Toxicol.* (2016) **1** 54-61. DOI: 10.1016/j.cotox.2016.09.005
22. Kishimoto Y., Kondo K., Momiyama Y.. **The protective role of heme oxygenase-1 in atherosclerotic diseases**. *Int. J. Mol. Sci.* (2019) **20** 3628. DOI: 10.3390/ijms20153628
23. Krishnan S. M., Sobey C. G., Latz E., Mansell A., Drummond G. R.. **IL-1β and IL-18: Inflammatory markers or mediators of hypertension?**. *Br. J. Pharmacol.* (2014) **171** 5589-5602. DOI: 10.1111/bph.12876
24. Li W. G., Miller F. J., Zhang H. J., Spitz D. R., Oberley L. W., Weintraub N. L.. **H(2)O(2)-induced O(2) production by a non-phagocytic NAD(P)H oxidase causes oxidant injury**. *J. Biol. Chem.* (2001) **276** 29251-29256. DOI: 10.1074/jbc.M102124200
25. Ling Y. H., Krishnan S. M., Chan C. T., Diep H., Ferens D., Chin-Dusting J.. **Anakinra reduces blood pressure and renal fibrosis in one kidney/DOCA/salt-induced hypertension**. *Pharmacol. Res.* (2017) **116** 77-86. DOI: 10.1016/j.phrs.2016.12.015
26. Liu L., Gu H., Liu H., Jiao Y., Li K., Zhao Y.. **Protective effect of resveratrol against IL-1β-induced inflammatory response on human osteoarthritic chondrocytes partly via the TLR4/MyD88/NF-κB signaling pathway: An “**. *Int. J. Mol. Sci.* (2014) **15** 6925-6940. DOI: 10.3390/ijms15046925
27. Lopes R. A., Neves K. B., Tostes R. C., Montezano A. C., Touyz R. M.. **Downregulation of nuclear factor erythroid 2-related factor and associated antioxidant genes contributes to redox-sensitive vascular dysfunction in hypertension**. *Hypertension* (2015) **66** 1240-1250. DOI: 10.1161/HYPERTENSIONAHA.115.06163
28. Mai W., Liao Y.. **Targeting IL-1β in the treatment of atherosclerosis**. *Front. Immunol.* (2020) **11** 589-654. DOI: 10.3389/fimmu.2020.589654
29. Martín A., Pérez-Girón J. V., Hernanz R., Palacios R., Briones A. M., Fortuño A.. **Peroxisome proliferator-activated receptor-γ activation reduces cyclooxygenase-2 expression in vascular smooth muscle cells from hypertensive rats by interfering with oxidative stress**. *J. Hypertens.* (2012) **30** 315-326. DOI: 10.1097/HJH.0b013e32834f043b
30. Matsunaga N., Tsuchimori N., Matsumoto T., Li M.. **TAK-242 (resatorvid), a small-molecule inhibitor of toll-like receptor (TLR) 4 signaling, binds selectively to TLR4 and interferes with interactions between TLR4 and its adaptor molecules**. *Mol. Pharmacol.* (2011) **79** 34-41. DOI: 10.1124/mol.110.068064
31. Melton E., Qiu H.. **Interleukin-1β in multifactorial hypertension: Inflammation, vascular smooth muscle cell and extracellular matrix remodeling, and non-coding RNA regulation**. *Int. J. Mol. Sci.* (2021) **22** 8639. DOI: 10.3390/ijms22168639
32. Muri J., Wolleb H., Broz P., Carreira E. M., Kopf M.. **Electrophilic Nrf2 activators and itaconate inhibit inflammation at low dose and promote IL-1β production and inflammatory apoptosis at high dose**. *Redox Biol.* (2020) **36** 101647. DOI: 10.1016/j.redox.2020.101647
33. Nickenig G., Baudler S., Müller C., Werner C., Werner N., Welzel H.. **Redox-sensitive vascular smooth muscle cell proliferation is mediated by GKLF and Id3**. *FASEB J.* (2002) **16** 1077-1086. DOI: 10.1096/fj.01-0570com
34. Nunes K. P., de Oliveira A. A., Mowry F. E., Biancardi V. C.. **Targeting toll-like receptor 4 signalling pathways: Can therapeutics pay the toll for hypertension?**. *Br. J. Pharmacol.* (2019) **176** 1864-1879. DOI: 10.1111/bph.14438
35. Palacios-Ramírez R., Hernanz R., Martín A., Pérez-Girón J. V., Barrús M. T., González-Carnicero Z.. **Pioglitazone modulates the vascular contractility in hypertension by interference with ET-1 pathway**. *Sci. Rep.* (2019) **9** 16461. DOI: 10.1038/s41598-019-52839-6
36. Parsamanesh N., Asghari A., Sardari S., Tasbandi A., Jamialahmadi T., Xu S.. **Resveratrol and endothelial function: A literature review**. *Pharmacol. Res.* (2021) **170** 105725. DOI: 10.1016/j.phrs.2021.105725
37. Pfeiler S., Winkels H., Kelm M., Gerdes N.. **IL-1 family cytokines in cardiovascular disease**. *Cytokine* (2019) **122** 154215. DOI: 10.1016/j.cyto.2017.11.009
38. Qi H. M., Cao Q., Liu Q.. **TLR4 regulates vascular smooth muscle cell proliferation in hypertension via modulation of the NLRP3 inflammasome**. *Am. J. Transl. Res.* (2021) **13** 314-325. PMID: 33527026
39. Robledinos-Antón N., Fernández-Ginés R., Manda G., Cuadrado A.. **Activators and inhibitors of NRF2: A review of their potential for clinical development**. *Oxid. Med. Cell Longev.* (2019) **2019** 9372182. DOI: 10.1155/2019/9372182
40. Şahin B., İlgün G.. **Risk factors of deaths related to cardiovascular diseases in World Health Organization (WHO) member countries**. *Health Soc. Care Community* (2022) **30** 73-80. DOI: 10.1111/hsc.13156
41. Sánchez-Duffhues G., García de Vinuesa A., van de Pol V., Geerts M. E., de Vries M. R., Janson S. G. T.. **Inflammation induces endothelial-to-mesenchymal transition and promotes vascular calcification through downregulation of BMPR2**. *J. Pathol.* (2019) **247** 333-346. DOI: 10.1002/path.5193
42. Senoner T., Dichtl W.. **Oxidative stress in cardiovascular diseases: Still a therapeutic target?**. *Nutrients* (2019) **11** 2090. DOI: 10.3390/nu11092090
43. Shen W. C., Liang C. J., Huang T. M., Liu C. W., Wang S. H., Young G. H.. **Indoxyl sulfate enhances IL-1β-induced E-selectin expression in endothelial cells in acute kidney injury by the ROS/MAPKs/NFκB/AP-1 pathway**. *Arch. Toxicol.* (2016) **90** 2779-2792. DOI: 10.1007/s00204-015-1652-0
44. Silveira Rossi J. L., Barbalho S. M., Reverete de Araujo R., Bechara M. D., Sloan K. P., Sloan L. A.. **Metabolic syndrome and cardiovascular diseases: Going beyond traditional risk factors**. *Diabetes Metab. Res. Rev.* (2022) **38** e3502. DOI: 10.1002/dmrr.3502
45. Subedi L., Lee J., Yumnam S., Ji E., Kim S.. **Anti-inflammatory effect of sulforaphane on LPS-activated microglia potentially through JNK/AP-1/NF-κB inhibition and Nrf2/HO-1 activation**. *Cells* (2019) **8** 194. DOI: 10.3390/cells8020194
46. Tanase D. M., Apostol A. G., Costea C. F., Tarniceriu C. C., Tudorancea I., Maranduca M. A.. **Oxidative stress in arterial hypertension (HTN): The nuclear factor erythroid factor 2-related factor 2 (Nrf2) pathway, implications and future perspectives**. *Pharmaceutics* (2022) **14** 534. DOI: 10.3390/pharmaceutics14030534
47. Vallejo S., Palacios E., Romacho T., Villalobos L., Peiró C., Sánchez-Ferrer C. F.. **The interleukin-1 receptor antagonist anakinra improves endothelial dysfunction in streptozotocin-induced diabetic rats**. *Cardiovasc Diabetol.* (2014) **13** 158. DOI: 10.1186/s12933-014-0158-z
48. Wang C., Luo Z., Carter G., Wellstein A., Jose P. A., Tomlinson J.. **NRF2 prevents hypertension, increased ADMA, microvascular oxidative stress, and dysfunction in mice with two weeks of ANG II infusion**. *Am. J. Physiol. Regul. Integr. Comp. Physiol.* (2018) **314** R399-R406. DOI: 10.1152/ajpregu.00122.2017
49. Wang L., Zhang J., Fu W., Guo D., Jiang J., Wang Y.. **Association of smooth muscle cell phenotypes with extracellular matrix disorders in thoracic aortic dissection**. *J. Vasc. Surg.* (2012) **56** 1698-1709. DOI: 10.1016/j.jvs.2012.05.084
50. Wang X., Feuerstein G. Z., Gu J.-L., Lysko P. G., Yue T. L.. **Interleukin-1β induces expression of adhesion molecules in human vascular smooth muscle cells and enhances adhesion of leukocytes to smooth muscle cells**. *Atherosclerosis* (1995) **115** 89-98. DOI: 10.1016/0021-9150(94)05503-b
51. Xia N., Li H., Daiber A., Förstermann U.. **Antioxidant effects of resveratrol in the cardiovascular system**. *Br. J. Pharmacol.* (2017) **174** 1633-1646. DOI: 10.1111/bph.13492
52. Xu X., Liu X., Yang Y., He J., Jiang M., Huang Y.. **Resveratrol exerts anti-osteoarthritic effect by inhibiting TLR4/NF-κB signaling pathway via the TLR4/Akt/FoxO1 axis in IL-1β-stimulated SW1353 cells**. (2020) **14** 2079-2090. DOI: 10.2147/DDDT.S244059
53. Yamamoto M., Kensler T. W., Motohashi H.. **The KEAP1-NRF2 system: A thiol-based sensor-effector apparatus for maintaining redox homeostasis**. *Physiol. Rev.* (2018) **98** 1169-1203. DOI: 10.1152/physrev.00023.2017
54. Yang K., Zhang X. J., Cao L. J., Liu X. H., Liu Z. H., Wang X. Q.. **Toll-like receptor 4 mediates inflammatory cytokine secretion in smooth muscle cells induced by oxidized low-density lipoprotein**. *PLoS ONE* (2014) **9** e95935. DOI: 10.1371/journal.pone.0095935
55. Zhang D. D., Hannink M.. **Distinct cysteine residues in Keap1 are required for Keap1-dependent ubiquitination of Nrf2 and for stabilization of Nrf2 by chemopreventive agents and oxidative stress**. *Mol. Cell Biol.* (2003) **23** 8137-8151. DOI: 10.1128/MCB.23.22.8137-8151.2003
56. Zhao S., Cheng C. K., Zhang C. L., Huang Y.. **Interplay between oxidative stress, cyclooxygenases, and prostanoids in cardiovascular diseases**. *Antioxid. Redox Signal* (2021) **34** 784-799. DOI: 10.1089/ars.2020.8105
|
---
title: Financial precarity, food insecurity, and psychological distress prospectively
linked with use of potentially dangerous dietary supplements during the pandemic
in the US
authors:
- S. Bryn Austin
- Ariel L. Beccia
- Amanda Raffoul
- Destiny A. Jackson
- Vishnudas Sarda
- Jaime E. Hart
- Jorge E. Chavarro
- Janet Rich-Edwards
journal: Frontiers in Public Health
year: 2023
pmcid: PMC10018192
doi: 10.3389/fpubh.2023.1120942
license: CC BY 4.0
---
# Financial precarity, food insecurity, and psychological distress prospectively linked with use of potentially dangerous dietary supplements during the pandemic in the US
## Abstract
### Introduction
Supplements sold with claims to promote weight loss, cleansing/detoxing, increased energy, or boosted immunity can be dangerous, and consumers experiencing extreme stressors may be especially vulnerable to deceptive claims. The purpose of our study was to investigate associations of financial strain and psychological distress during the COVID-19 pandemic with use of supplements sold for weight loss, cleanse/detox, energy, or immunity.
### Methods
We used repeated-measures data gathered over five survey waves from April/May 2020–April 2021 from the COVID-19 Substudy ($$n = 54$$,951), within three prospective US national cohorts (Nurses' Health Study 2, Nurses' Health Study 3, and Growing Up Today Study), to investigate longitudinal associations between financial strain and psychological distress and risk of use of potentially dangerous types of supplements. Surveys assessed use of supplements prior to and during the first year of the pandemic, as well as financial precarity, food insecurity, depressive and anxiety symptoms, perceived stress, and daily hassles. We fit sociodemographic-adjusted modified Poisson GEE models to estimate risk ratios (RRs) and $95\%$ confidence intervals (CIs) for associations between baseline or lagged time-varying predictors and prevalent or incident (i.e., new-onset) use of each supplement type.
### Results
At baseline in April/May 2020, soon after pandemic onset, current use of supplement types was: weight loss $2.7\%$; cleanse/detox $3.2\%$; energy $4.4\%$; immune $22.6\%$. By the end of the study period, cumulative incidence was: weight loss $3.5\%$; cleanse/detox $3.7\%$; energy $4.5\%$; immune $21.3\%$. In prevalent-use analyses, financial precarity, food insecurity, and psychological distress were associated with up to 2.4 times the risk of use of these types of supplements across the study period. Similarly, in incident-use analyses, financial precarity and psychological distress were associated with up to 2.1 times the risk of initiating use; whereas, high food insecurity was associated with nearly 1.8 times higher risk of onset of weight-loss supplements use but was not associated with onset of use of other types of supplements.
### Discussion
We found consistent evidence that during the first year of the pandemic, participants experiencing elevated financial strain and psychological distress were at heightened risk of initiating use of potentially dangerous types of supplements. Our findings raise concerns about deceptive claims about the safety and product effectiveness by manufacturers of these supplements to profit from vulnerable consumers during the pandemic.
## Introduction
The US market in dietary supplements is massive and fast growing, with over half of adults [1] and nearly a third of children [2] having taken supplements in the past month [3]. With the onset of the COVID-19 pandemic, consumer anxiety about infection risk increased appreciably, and the industry saw rising sales for dietary supplements often marketed with deceptive claims of boosting “wellness” [4, 5]. These types of potentially dangerous supplements include those sold with claims to lead to weight loss, cleanse/detox, increased energy, or boosted immunity. These types of supplements are not medically recommended, not effective in producing healthful outcomes, and due to toxic ingredients, many have been linked with serious physical harms. Due to widely documented adulteration of supplements on the US market with illicit and toxic ingredients, these products have been linked with numerous serious physical harms, including cardiovascular events, liver injury, and death (6–10). A recent national study found Latin individuals of all genders make up $36\%$, and women of all race/ethnicity groups make of $74\%$, of serious liver injury cases, some requiring transplant, caused by green tea extract, a common liver toxin in weight-loss and energy supplements [9]. In addition, commonly used dietary supplements can render ineffective prescription medications, such as those for hormonal contraception or to treat blood clots, HIV/AIDS, or organ transplants, with dire effects (6–13). Yet companies selling these types of supplements persist in deceptive promotion of their products [14], misleading consumers about their safety and effectiveness.
Widespread psychological distress during the pandemic has been documented [15], with an estimated threefold increase in depressive symptoms in US adults within the first few months of the pandemic [16]. Marketing of dietary supplements often uses positive framing of consumer agency and self-care [17] and, both explicitly and implicitly [18], often includes deceptive claims of health protection [19, 20], such as claims to be a healthful and effective way to reduce risk of illness, weight gain, flagging energy, or the effects of aging. It is plausible that the combination of marketing emphasizing consumer agency and self-care and deceptive claims of safety and effectiveness may lead consumers experiencing elevated psychological distress to use these products during the pandemic, especially if they are seeking to bolster a sense of control and protection in the face of intense and unpredictable pandemic-related stressors. Little is known, however, regarding how psychological distress during the pandemic may be associated with risk of using potentially dangerous types of supplements.
Economic shocks during the pandemic both exacerbated psychological distress and reduced access to food and other basic necessities, with households in communities of color disproportionately impacted by pandemic-related financial precarity and food insecurity [21, 22]. By April and May of 2020, food insecurity rate nationwide doubled due to the pandemic [23], peaking at an estimated $23\%$ of households [22], and by 2021 food insecurity was estimated to affect 42 million people in the United States [23]. A study of US Census *Bureau data* collected from April–June 2020 found that household income shocks, such as suddenly becoming unemployed or experiencing reduced work hours, were associated with higher risk of depression and anxiety [24]. A recent US nationally representative study found job loss to be associated with substantially elevated psychological distress [25]. In research conducted before the pandemic with the National Health and Nutrition Examination Survey, prevalence of past-month use of supplements of any type, which could include vitamins or other types of supplements, was lower in food-insecure ($36\%$) compared to food-secure ($55\%$) households, but prevalence of use still exceeded a third of food-insecure households [26]. Potentially dangerous types of supplements are purchased in households at all income levels, but the burden of purchasing proportional to household income is two to four times higher for lower-income than higher-income households, as shown in a national study of weight-loss supplements [27]. As a result, while it is plausible that lower-income households may reduce purchasing of these products temporarily during times of heightened financial precarity and food insecurity, prior research suggests that many low-income households continue to purchase potentially dangerous types of supplements, perhaps with the commonly held false belief the products are safe and effective in protecting their own and their family's health [28], thus compounding the financial hardship these households face. Yet no studies we are aware of have assessed how financial precarity and food insecurity during the pandemic may be associated with risk of using potentially dangerous types of supplements.
Revenue during the pandemic has grown precipitously for the US dietary supplements industry. In 2020, US revenue increased over prior years by more than $7 billion, representing $14.5\%$ growth in a single year, and amounting to $56 billion in total revenue by year's end [29]. By comparison, in the 5 years preceding the pandemic, the industry's growth in the US averaged $2–$2.5 billion/year [29]. Industry analysts estimate that growth in the US supplements market is ~$1.4 billion/year higher than it would have been without the COVID-19 pandemic-related surge in sales, and by 2024 analysts predict US supplements revenue to exceed $66 billion [29].
Prior to the pandemic, widespread dietary supplement fraud, defined as fraud perpetrated for economic gain through selling dietary supplements, was already considered an important public health threat in certain sectors of the supplements market [30, 31]. For instance, one content analysis of marketing claims of weight-loss and muscle-building supplements found packages promoted on average 6.5 claims, largely relaying deceptive information claiming to reduce weight, body mass index, or fat; cleanse or detox organs; and boost immunity; furthermore, nearly half of the product packages falsely claimed scientific evidence verified these effects [32].
To address gaps in the literature related to use of potentially dangerous dietary supplements during the pandemic, we gathered longitudinal survey data on financial precarity, food insecurity, psychological indicators of distress, and use of supplements sold for weight loss, cleanse/detox, energy, or immunity in the COVID-19 Substudy, embedded within three existing US national prospective cohorts. The purpose of our study was to investigate associations of financial strain and psychological distress during the COVID-19 pandemic with use of these types of supplements. We hypothesized that financial precarity, food insecurity, and psychological distress would be prospectively, positively associated with both persistent use and new-onset use of potentially dangerous types of supplements.
## Study design
The present study uses data provided by participants in the COVID-19 Substudy, which is a closed cohort embedded within three US national, ongoing prospective cohorts Nurses' Health Study 2 (NHS2) [33], Nurses' Health Study 3 (NHS3) [34], and Growing Up Today Study (GUTS) [35] cohorts. NHS2 participants were ages 25–42 years old at enrollment in 1989; NHS3 is an open cohort launched in 2010 where participants were aged 19–49 years at enrollment; GUTS participants were aged 9–16 years at enrollment, which was conducted in two phases with the first in 1996 and the second in 2004. All NHS2 and NHS3 were nursing professionals at the time of enrollment, and all GUTS participants are the children of NHS2 participants [34]. Eligible participants were those currently enrolled in NHS$\frac{2}{3}$ and GUTS with an email on record with cohort administrators and who had returned any pending prior questionnaires (see Supplementary Figure 1 for additional exclusions). By combining the three cohorts, the COVID-19 Substudy included 58,612 participants living in all US states and a very large age range from young to middle adulthood through elder years (GUTS $$n = 6$$,725; NHS3 $$n = 12$$,323; NHS2 $$n = 39$$,566). Participant ages in 2021 were: GUTS 26–40 years; NHS3 30–70 years; NHS2 57–74 years.
For the COVID-19 Substudy, the baseline questionnaire was administered in April-May 2020. During the first 4 months, we conducted monthly follow-ups from baseline among all participants. After the first 4 months, we conducted follow-up questionnaires every 3 months through April 2021 for a total of seven study waves administered to all participants (see Supplementary Figure 2, which also details as to when during the study period our variables of interest were assessed). The overall study was approved by the Brigham and Women's Hospital Institutional Review Board (IRB) and the current analyses by the Boston Children's Hospital IRB.
## Analytic sample
For the current study, we excluded participants who reported living outside the United States at any point during the study period ($$n = 820$$), because individuals living outside the country may have had different pandemic-related experiences given the substantial heterogeneity in COVID-19 response across countries. We then excluded those who had missing data on key covariates (i.e., age, cohort, race/ethnicity, gender identity, and geographic region of residence; $$n = 1$$,649) and/or the auxiliary variables used in the construction of inverse-probability-of-censoring weights (IPW) (i.e., COVID-19 infection and related symptomology, occupational status, and mental health status; $$n = 1$$,208). We also excluded those who never provided information on their use of weight-loss, cleanse/detox, energy, or immune supplements ($$n = 417$$), resulting in a total of 54,951 eligible participants. For analyses of incident use of these types of supplements, eligibility was further restricted to the 43,469 participants who provided outcome information at baseline and each subsequent study wave and reported no use prior to pandemic onset. List-wise deletion was used to handle missingness on each predictor separately, resulting in final analytic sample size ranges (which varied for each supplement type) of 40,906–54,951 for prevalent-use analyses and 32,814–43,469 for incident-use analyses.
## Predictors
Financial strain was assessed in two ways: The first was with an original single-item measure of general financial strain administered at baseline only that asked, “Since the pandemic began, how much of a concern is having enough money for essentials like food and clothing or for paying rent or mortgage?” with response options including “extremely,” “moderately,” “somewhat,” or “not at all concerning.” We used this assessment at baseline as a proxy for experiences with financial precarity across survey waves. The second was a two-item measure of food insecurity derived from the validated Household Food Security Survey Module (HFSSM) [36]. The items asked participants to rate the frequency with which they worried about and/or experienced food running out before having money to buy more. Following federal guidelines [37], we categorized those who responded “sometimes” or “often” to both items as having high food insecurity, those who responded “sometimes” or “often” to one item as having moderate food insecurity, and those who responded “never” to both items as having low food insecurity. These items were administered once during the follow-up survey administered in July 2020; given research showing that food insecurity tended to track over time within individuals during the first year of the COVID-19 pandemic [38, 39], we used this single assessment as a proxy for experiences with food insecurity across survey waves.
Indicators of psychological distress included depressive symptoms, anxiety symptoms, perceived stress, and daily hassles. Depressive and anxiety symptoms were assessed at each study wave using the Patient Health Questionnaire (PHQ-2) [40] and the Generalized Anxiety Disorder Scale (GAD-2) [41], respectively, both of which were operationalized using the validated binary cut-off score of 3 to represent probable presence of symptoms [40, 41]. Perceived stress was assessed at most study waves (baseline, May 2020, June 2020, July 2020, and October 2020) using the abbreviated Perceived Stress Scale (PSS-4) [42] and operationalized using a binary indicator contrasting high vs. low levels of perceived stress, where high perceived stress was defined as having a PSS-4 score in the top tertile of the score distribution and low perceived stress was defined as having a score in the lower two tertiles [42]. Last-observation-carried-forward imputation was used to fill in values for the two study waves that did not administer the PSS-4 (i.e., January 2021 and April 2021). Finally, daily hassles, defined as experiences in everyday life that an individual finds salient and harmful or threatening to one's own wellbeing, was assessed at baseline only using the Daily Hassles Scale (DHS) [43] and was operationalized in the same manner as the PSS-4 (i.e., using a binary indicator contrasting high vs. low levels of daily hassles, defined using tertiles) [44]. We used this assessment at baseline as a proxy for daily hassles experiences across survey waves.
## Outcomes
We considered use of any of four potentially dangerous types of supplements as outcomes, which were assessed at five of the seven survey waves. Participants were asked at baseline in April/May 2020, “Are you currently using any of the following types of dietary supplements?” with subsequent items asking specifically about use of weight loss, cleanse/detox, energy, and immune supplements. Response options for each supplement type included “No;” “Yes, started before the outbreak;” “Yes; started after the outbreak began.” On follow-up surveys administered in June 2020, October 2020, January 2021, and April 2021, participants were asked if they were currently using each type of supplement. We considered both prevalent use (i.e., current use at a given study wave, regardless of prior use status) and incident use (i.e., new use at a given study wave among those who reported no pre-pandemic use and no prior use at any waves). All outcomes were operationalized as binary and time-varying (i.e., prevalent/incident use vs. no use at a given study wave).
## Covariates
We identified a parsimonious set of covariates that represented potential confounders of the relationship between the previously listed predictors and use of potentially dangerous types of supplements, including age in years, cohort (NHS2, NHS3, and GUTS), gender identity (cisgender woman, cisgender man, and transgender/gender diverse), race/ethnicity (Asian, Black, Hispanic/Latin, White, and other/unlisted), US geographic region of residence (Midwest, Northeast, South, and West), and current healthcare worker status (yes and no).
## Data analysis
We fit a series of modified Poisson models to estimate risk ratios (RRs) and $95\%$ confidence intervals (CIs) for associations between each predictor and use of each type of supplement. The first set of models included only baseline data to estimate cross-sectional associations between predictors and prevalent supplement use. The second set of models included repeated-measures data from all relevant study waves to estimate longitudinal associations between predictors and prevalent supplement use across the study period. For predictors that were time-varying, we lagged their values by one study wave to ensure correct temporal ordering (i.e., associations between predictor at wave w−1 and outcomes at wave w), using their baseline values for the unobserved pre-baseline period [45]. Finally, the third set of models estimated longitudinal associations between predictors and incident supplement use across the study period among the subset of participants who reported no use at any wave prior to outcome measurement; here, participants were excluded from the analysis after their first report of the outcome (i.e., use of a type of supplement). To account for intra-cluster correlation due to repeated measures, the second and third set of models were fit using generalized estimating equations with robust standard errors and an exchangeable working correlation matrix. All models adjusted for the aforementioned covariates.
Attrition was substantial across the study period (see Supplementary Figure 2: May 2020 $19.3\%$, June 2020 $23.8\%$, July 2020 $25.4\%$, October 2020 $27.8\%$, January 2021 $25.0\%$, and April 2021 $17.2\%$); therefore, we used IPW to re-weight the data such that they reflected the sample composition at baseline. Following the methods outlined by Fewell et al. [ 45] and expanded upon by VanderWeele et al. [ 46] for implementing IPW in the context of repeated-measures data, we built weights that predicted censoring at each study wave (for prevalent-use models) or the cumulative probability of remaining uncensored over time (for incident-use models), conditional on demographic characteristics (age, cohort, gender identity, race/ethnicity, geographic region of residence, and healthcare worker status) and a set of auxiliary variables that were associated with loss-to-follow-up and/or item non-response. The auxiliary variables included COVID-19 infection and related symptomology, unemployment, and mental health status. We then incorporated these weights into the longitudinal models described previously such that the resulting estimates could be interpreted as the associations that would have been observed had there been no attrition (assuming no omitted-variable bias). Analyses were conducted in R version 4.1.0 using the geepack package [47].
## Results
Table 1 presents baseline characteristics of the study sample of 54,951 participants living throughout the United States. The sample was largely made up of white cisgender women ranging in age from young adulthood through elder years. At baseline, just under $3\%$ reported experiencing moderate or high food insecurity and slightly more than $6\%$ reported feeling moderately or extremely concerned about having the financial resources to provide household essentials. Depressive ($12.9\%$) and anxiety ($21.2\%$) symptoms were common at baseline, and the prevalence of current use of supplement types was as follows: weight loss $2.7\%$; cleanse/detox $3.2\%$; energy $4.4\%$; and immune $22.6\%$. Figure 1 depicts the fluctuation in current use of each type of supplement across the five waves, which ranged from lowest to highest as follows: weight-loss 2.7–$3.5\%$; cleanse/detox 3.0–$3.5\%$; energy 4.1–$4.5\%$; immune 18.2–$25.0\%$.
The results of multivariable models estimating cross-sectional associations between financial strain and psychological distress and supplement use at baseline, controlling for confounders, are presented in Table 2. Participants reporting financial strain or psychological distress at baseline were at higher risk of using all four types of supplements at baseline, ranging from $8\%$ to over $180\%$ higher risk.
**Table 2**
| Unnamed: 0 | Supplement type | Supplement type.1 | Supplement type.2 | Supplement type.3 |
| --- | --- | --- | --- | --- |
| | Weight-loss | Cleanse/detox | Energy | Immune |
| | RR (95% CI) | RR (95% CI) | RR (95% CI) | RR (95% CI) |
| Financial stressor | Financial stressor | Financial stressor | Financial stressor | Financial stressor |
| Concern over essentials | | | | |
| Extremely concerning | 1.88 (1.37, 2.60) | 2.00 (1.51, 2.63) | 2.18 (1.75, 2.72) | 1.41 (1.27, 1.56) |
| Moderately concerning | 1.96 (1.57, 2.46) | 1.72 (1.39, 2.13) | 2.05 (1.73, 2.43) | 1.28 (1.19, 1.39) |
| Somewhat concerning | 1.64 (1.43, 1.87) | 1.47 (1.29, 1.66) | 1.53 (1.37, 1.71) | 1.23 (1.18, 1.28) |
| Not at all concerning | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) |
| Food insecurity | | | | |
| High | 2.84 (1.79, 4.51) | 2.44 (1.59, 3.75) | 2.65 (1.88, 3.73) | 1.23 (1.01, 1.50) |
| Moderate | 2.32 (1.75, 3.08) | 1.68 (1.25, 2.25) | 2.21 (1.76, 2.77) | 1.20 (1.07, 1.34) |
| Low | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) |
| Psychological distress | Psychological distress | Psychological distress | Psychological distress | Psychological distress |
| Depressive symptoms | 1.19 (1.04, 1.37) | 1.21 (1.06, 1.37) | 1.50 (1.36, 1.66) | 1.08 (1.04, 1.13) |
| Anxiety symptoms | 1.08 (0.96, 1.22) | 1.17 (1.05, 1.31) | 1.36 (1.25, 1.49) | 1.13 (1.09, 1.17) |
| High perceived stress | 1.39 (1.12, 1.73) | 1.27 (1.02, 1.57) | 1.50 (1.29, 1.74) | 1.17 (1.09, 1.26) |
| High daily hassles | 1.37 (1.12, 1.69) | 1.43 (1.19, 1.71) | 1.53 (1.31, 1.78) | 1.27 (1.19, 1.36) |
Table 3 presents results of multivariable models using repeated measures (N of observations = 184,148–222,388, which varied for each supplement type) gathered across the five waves of data collection in the first year of the pandemic, where baseline or lagged indicators of financial strain and psychological distress predicted prevalent use of potentially dangerous types of supplements over the study period and/or at the subsequent wave, respectively. Similar to the cross-sectional models, these longitudinal models indicate that participants reporting financial strain or psychological distress experienced a higher risk of using most types of supplements assessed, ranging from $5\%$ to $140\%$ higher risk. Of note, food insecurity and concern about not having enough money for essentials like food and clothing or for paying rent or mortgage were consistently prospectively associated with elevated risk of using these products.
**Table 3**
| Unnamed: 0 | Supplement type | Supplement type.1 | Supplement type.2 | Supplement type.3 |
| --- | --- | --- | --- | --- |
| | Weight-loss | Cleanse/detox | Energy | Immune |
| | RR (95% CI) | RR (95% CI) | RR (95% CI) | RR (95% CI) |
| Financial stressor | Financial stressor | Financial stressor | Financial stressor | Financial stressor |
| Concern over essentials | | | | |
| Extremely concerning | 1.84 (1.48, 2.29) | 1.60 (1.29, 1.97) | 1.86 (1.60, 2.17) | 1.39 (1.28, 1.51) |
| Moderately concerning | 1.91 (1.64, 2.23) | 1.56 (1.34, 1.82) | 1.76 (1.55, 1.99) | 1.32 (1.24, 1.40) |
| Somewhat concerning | 1.56 (1.42, 1.72) | 1.37 (1.26, 1.50) | 1.49 (1.38, 1.61) | 1.22 (1.18, 1.26) |
| Not at all concerning | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) |
| Food insecurity | | | | |
| High | 2.35 (1.64, 3.39) | 1.62 (1.09, 2.41) | 2.22 (1.70, 2.91) | 1.23 (1.03, 1.46) |
| Moderate | 1.70 (1.37, 2.11) | 1.33 (1.05, 1.67) | 1.63 (1.36, 1.95) | 1.25 (1.14, 1.36) |
| Low | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) |
| Psychological distress b | Psychological distress b | Psychological distress b | Psychological distress b | Psychological distress b |
| Depressive symptoms | 1.07 (0.99, 1.16) | 1.04 (0.97, 1.12) | 1.23 (1.17, 1.30) | 1.03 (1.00, 1.05) |
| Anxiety symptoms | 1.02 (0.96, 1.09) | 1.04 (0.98, 1.10) | 1.12 (1.07, 1.17) | 1.05 (1.03, 1.07) |
| High perceived stress | 1.18 (1.03, 1.35) | 1.14 (1.00, 1.30) | 1.27 (1.15, 1.40) | 1.08 (1.03, 1.13) |
| High daily hassles | 1.37 (1.19, 1.58) | 1.32 (1.16, 1.52) | 1.34 (1.20, 1.49) | 1.27 (1.21, 1.34) |
Finally, Table 4 presents results of multivariable models estimating the associations between financial strain and psychological distress with incident use of each type of supplements, restricted to those not previously reporting use of a supplement type. By the end of the study period, cumulative incidence of use post-pandemic onset was: weight loss $3.5\%$; cleanse/detox $3.7\%$; energy $4.5\%$; and immune $21.3\%$. Participants experiencing financial precarity and psychological distress were at increased risk of beginning to use some types of supplements during the first year of the pandemic. In contrast, high food insecurity was associated with nearly 1.8 times higher risk of onset of weight-loss supplements use but was not associated with onset of use of other types of supplements. The most consistent pattern of risk was observed for participants experiencing concern about not having enough money for essentials, ranging from $21\%$ to $111\%$ elevated risk. Similarly, participants experiencing psychological distress, such as depressive and anxiety symptoms, high perceived stress, and high daily hassles were at increased risk of new use of supplements sold with claims to lead to weight loss, energy, or immune boosting, ranging from $16\%$ to $47\%$ elevated risk.
**Table 4**
| Unnamed: 0 | Supplement type | Supplement type.1 | Supplement type.2 | Supplement type.3 |
| --- | --- | --- | --- | --- |
| | Weight-loss | Cleanse/detox | Energy | Immune |
| | RR (95% CI) | RR (95% CI) | RR (95% CI) | RR (95% CI) |
| Cumulative incidence, % b | 3.5 | 3.7 | 4.5 | 21.3 |
| Financial stressor | Financial stressor | Financial stressor | Financial stressor | Financial stressor |
| Concern over essentials | | | | |
| Extremely concerning | 1.74 (1.26, 2.41) | 1.47 (1.04, 2.06) | 2.11 (1.63, 2.73) | 1.42 (1.22, 1.65) |
| Moderately concerning | 1.49 (1.17, 1.89) | 1.54 (1.23, 1.93) | 1.46 (1.18, 1.80) | 1.25 (1.12, 1.38) |
| Somewhat concerning | 1.35 (1.18, 1.54) | 1.25 (1.09, 1.42) | 1.36 (1.21, 1.53) | 1.21 (1.14, 1.27) |
| Not at all concerning | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) |
| Food insecurity | | | | |
| High | 1.77 (1.03, 3.06) | 0.92 (0.44, 1.96) | 1.64 (0.99, 2.70) | 0.92 (0.67, 1.28) |
| Moderate | 1.17 (0.80, 1.71) | 0.91 (0.60, 1.38) | 1.26 (0.91, 1.73) | 1.17 (1.00, 1.37) |
| Low | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) |
| Psychological distress c | Psychological distress c | Psychological distress c | Psychological distress c | |
| Depressive symptoms | 1.31 (1.14, 1.50) | 1.11 (0.96, 1.29) | 1.46 (1.29, 1.64) | 1.17 (1.10, 1.24) |
| Anxiety symptoms | 1.16 (1.03, 1.31) | 1.10 (0.98, 1.24) | 1.32 (1.19, 1.46) | 1.20 (1.15, 1.26) |
| High perceived stress | 1.22 (0.94, 1.58) | 0.96 (0.72, 1.28) | 1.47 (1.20, 1.80) | 1.26 (1.14, 1.41) |
| High daily hassles | 1.39 (1.13, 1.70) | 1.15 (0.93, 1.42) | 1.47 (1.23, 1.74) | 1.34 (1.23, 1.45) |
## Discussion
As the COVID-19 pandemic caused widespread psychological distress and financial hardship to millions of Americans [15, 21, 22] the dietary supplements industry enjoyed rapid market growth [29] selling myriad products promising to consumers health protection broadly and specifically claiming to promote weight loss, cleansing/detoxing, energy, and immunity. Given concerns about inadequate consumer protections by government to prevent deceptive claims, we undertook our study to identify predictors of use of potentially dangerous supplement types during the first year of the pandemic. In a large cohort study of nearly 55,000 US adults, we found that participants experiencing heightened financial precarity, food insecurity, and psychological distress were at up to 2.4 times the risk of use of these products throughout the first year of the pandemic. Furthermore, among those who did not previously report use, these same adverse experiences were associated with up to over two times the risk of beginning to use these potentially dangerous products during the first year of the pandemic. To our knowledge, this study is the first to investigate patterns of dietary supplements use associated with financial strain and psychological distress during the pandemic, and our findings raise concerns about industry practices that may exploit vulnerabilities of consumers hardest hit by the pandemic to sell them products that are not only ineffective, but also may put their own and their family's health in greater peril.
Since the passage of the US federal Dietary Supplements Health Education Act (DSHEA) in 1994, the US supplements industry has grown appreciably, earning $56 billion in total revenue from American consumers by the end of 2020 [29]. As a direct result of DSHEA, however, the Food and Drug Administration (FDA) is prohibited from requiring rigorous prescreening of supplements for safety or efficacy before they enter the market [48]. Growth for the industry has only accelerated since the start of the pandemic [29]. Continued industry growth without robust government oversight should raise alarms among medical and public health professionals, given the substantial evidence linking certain types of supplements with serious injury in consumers, including cardiovascular events, liver injury, and death (6–10).
As is now well-documented, the pandemic provoked widespread psychological distress, with a threefold increase in depressive symptoms within just a few months after pandemic onset in the United States [15, 16]. Also well-documented were the profound financial shocks that many experienced in the first year of the pandemic, including reduced income and job loss, leading some to extreme financial precarity, food insecurity, and exacerbated psychological distress [24, 25]. In fact, within months of the start of the pandemic in the United States, food insecurity had doubled nationally [23], peaking at $23\%$ of households by April/May 2020 [22], and by 2021 an estimated 42 million people across the country were struggling with food insecurity [23]. However, no prior studies we are aware of have investigated how these types of widespread adverse experiences during the pandemic may have increased the risk of use of potentially dangerous types of dietary supplements commonly marketed with positive framing of consumer agency and self-care [17, 18] and with deceptive claims to support consumer health [19, 20]. Our study findings support our hypotheses that during the first year of the pandemic, consumers experiencing financial precarity, food insecurity, and psychological distress would be at elevated risk of both prevalent use and new onset of use of potentially dangerous dietary supplements sold for weight loss, cleanse/detox, energy, and immunity. It is plausible that for these types of supplements, consumers facing both financial and psychological adversity may have been particularly vulnerable to product marketing that emphasized consumer agency and self-care along with deceptive claims of safety and effectiveness. Perhaps vulnerable consumers sought to bolster a sense of control and protection as a way to cope with intense pandemic-related stressors. Future research will be needed to explore how deceptive marketing may influence consumer motivations to use potentially dangerous types of dietary supplements without evidence of safety or effectiveness, particularly in times of financial precarity and mental distress. In addition, as prior research has documented disproportionate financial burden on lower-income households attributable to purchases of deceptive weight-loss supplements [27], new research is needed to estimate the added financial burden attributable to supplements on households experiencing financial precarity and other pandemic-related adversities.
Our study has several limitations. Our study cohorts are not representative of the US population, as all participants in NHS2 and NHS3 were professional nurses or nursing trainees at the time of enrollment, and GUTS participants are all children of NHS2 participants. As a result, communities of color, men, and low-income communities are underrepresented in the cohorts. In addition, while both high or moderate food insecurity were positively associated with prevalent use of potentially dangerous types of supplements in the first year of the pandemic, in incident-use analyses, only high food insecurity was associated with new-onset use and only for weight-loss supplements. As only 1,075 participants reported food insecurity at baseline, it is possible that food insecurity is not associated with new-onset use of other types of supplements or that our analyses were not sufficiently powered to detect an association. Our assessment of use of potentially dangerous types of dietary supplements is self-reported, and we do not have information on dose or frequency of use or brand of supplement used, reducing precision in our outcome measurement. Unlike other primary predictors, general financial strain, food insecurity, and daily hassles were assessed at only one time point (so had to be treated as time invariant) and with brief measures that cannot fully capture the complexity of each construct, increasing the likelihood of misclassification. Sample attrition after April/May 2020 was substantial, and it is plausible that bias was introduced if participants most severely affected by the pandemic stopped completing surveys at a higher rate than other participants. In response, we implemented IPW, which is considered the most rigorous analytic approach to mitigate the effects of attrition bias on study findings.
The US dietary supplements industry achieved an unprecedented $14.5\%$ revenue growth during the first year of the pandemic, reaching $56 billion in total revenue by the end of 2020 [29]. Our finding that people experiencing heightened financial precarity, food insecurity, and psychological distress during the first year of the pandemic were at greatest risk of using potentially dangerous supplements provides disturbing evidence that companies selling these products profited directly from vulnerable members of our communities. Our findings also amplify ongoing concerns about consumer safety, given the persistent problems of deceptive advertising [4, 5, 32] and widespread consumer misconception that these products are safe and effective [28].
Our findings have important implications for clinicians, public health nutrition professionals, and government to protect consumers from the growing problem of predatory and potentially dangerous supplements. Clinicians should routinely query patients about their use of these types of supplements and counsel them as to the risks. Public health nutrition professionals working with individuals and households experiencing food insecurity and financial precarity should be aware that, despite the expense, these clients may use these products, perhaps with the mistaken belief promoted by industry that the products will protect themselves or their families from illness. Large-scale public health surveillance surveys designed to assess nutrition or risk behaviors should also assess dietary supplement use, particularly the types of potentially dangerous supplements as addressed in our study. Furthermore, descriptive and analytic epidemiologic studies are needed to build on our findings to help illuminate other determinants of use of potentially dangerous supplements and possible leverage points for preventive interventions. Finally, the FDA and Federal Trade Commission have a clear responsibility to take aggressive action against companies employing deceptive advertising and tainting their products with dangerous and illegal ingredients.
Given the myriad serious health risks (6–10) presented by supplements sold with claims to promote weight loss, cleanse/detox, energy, and immunity, our study underscores the compelling need for effective intervention. As the industry continues to capitalize on the COVID-19 pandemic to accelerate the pace and scale of its dangerous products, the need is urgent for a robust and evidence-based public health response to mitigate consumer harm linked with these deceptive products.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by Brigham and Women's Hospital. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
SA, AB, and AR conceived of the study. AB and VS carried out data analyses. SA and AB produced the initial draft of the manuscript. AR, DJ, VS, JR-E, JC, and JH provided feedback on data analyses and interpretation and critical revisions to the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpubh.2023.1120942/full#supplementary-material
## References
1. Kantor ED, Rehm CD, Du M, White E, Giovannucci EL. **Trends in dietary supplement use among US adults from 1999–2012**. *J Am Med Assoc.* (2016) **316** 1464-74. DOI: 10.1001/jama.2016.14403
2. Panjwani AA, Cowan AE, Jun S, Bailey RL. **Trends in nutrient- and non-nutrient-containing dietary supplement use among US children from 1999 to 2016**. *J Pediatr.* (2021) **231** 131-140.e132. DOI: 10.1016/j.jpeds.2020.12.021
3. 3.Food Drug Administration. Statement from FDA Commissioner Scott Gottlieb M.D on the Agency's New Efforts to Strengthen Regulation of Dietary Supplements by Modernizing and Reforming FDA's Oversight. (2019). Available online at: https://www.fda.gov/news-events/press-announcements/statement-fda-commissioner-scott-gottlieb-md-agencys-new-efforts-strengthen-regulation-dietary (accessed September 1, 2021).. *Statement from FDA Commissioner Scott Gottlieb M.D on the Agency's New Efforts to Strengthen Regulation of Dietary Supplements by Modernizing and Reforming FDA's Oversight.* (2019)
4. Rachul C, Marcon AR, Collins B, Caulfied T. **COVID-19 and ‘immune boosting' on the internet: a content analysis of Google search results**. *BMJ Open.* (2020) **10** e040989. DOI: 10.1136/bmjopen-2020-040989
5. Bramstedt KA. **Unicorn poo and blessed waters: COVID-19 quackery and FDA warning letters**. *Therap Innov Regul Sci.* (2021) **55** 239-44. DOI: 10.1007/s43441-020-00224-1
6. Rao N, Spiller HA, Hodges NL, Chounthirath T, Casavant MJ, Kamboj AK. **An increase in dietary supplement exposures reported to US poison control centers**. *J Med Toxicol.* (2017) **13** 227-37. DOI: 10.1007/s13181-017-0623-7
7. Geller AI, Shehab N, Weidle NJ, Lovegrove MC, Wolpert BJ, Timbo BB. **Emergency department visits for adverse events related to dietary supplements**. *N Engl J Med.* (2015) **373** 1531-40. DOI: 10.1056/NEJMsa1504267
8. Or F, Kim Y, Simms J, Austin SB. **Taking stock of dietary supplements' harmful effects on children, adolescents, and young adults**. *J Adolesc Health.* (2019) **65** 455-61. DOI: 10.1016/j.jadohealth.2019.03.005
9. Hoofnagle JH, Bonkovsky HL, Phillips EJ, Li YJ, Ahmad J, Barnhart H. **HLA-B**. *Hepatology.* (2021) **73** 2484-93. DOI: 10.1002/hep.31538
10. 10.National Institute of Diabetes Digestive Kidney Diseases. LiverTox: Clinical and Research Information on Drug-Induced Liver Injury: Green Tea. (2020). Available online at: https://pubmed.ncbi.nlm.nih.gov/31643176/ (accessed May 15, 2022).. *LiverTox: Clinical and Research Information on Drug-Induced Liver Injury: Green Tea.* (2020)
11. Asher GN, Corbett AH, Hawke RL. **Common herbal dietary supplement—drug interactions**. *Am Fam Phys* (2017) **96** 101-07
12. Fan CSS, Wen S, Lee H. **Warfarin and food, herbal or dietary supplement interactions: a systematic review**. *Br J Clin Pharmacol.* (2021) **87** 352-74. DOI: 10.1111/bcp.14404
13. Corey RL, Rakela J. **Complementary and alternative medicine: risks and special considerations in pretransplant and posttransplant patients**. *Nutr Clin Pract.* (2014) **9** 322-31. DOI: 10.1177/0884533614528007
14. Wharton S, Bonder R, Jeffery A, Christensen RAG. **The safety and effectiveness of commonly-marketed natural supplements for weight loss in populations with obesity: a critical review of the literature from 2006 to 2016**. *Crit Rev Food Sci Nutr.* (2020) **60** 1614-30. DOI: 10.1080/10408398.2019.1584873
15. Holingue C, Kalb LG, Riehm KE, Bennett D, Kapteyn A, Veldhuis CB. **Mental distress in the United States at the beginning of the COVID-19 pandemic**. *Am J Public Health.* (2020) **110** 1628-34. DOI: 10.2105/AJPH.2020.305857
16. Ettman CK, Abdalla SM, Cohen GH, Sampson L, Vivier PM, Galea S. **Prevalence of depression symptoms in US adults before and during the COVID-19 pandemic**. *JAMA Netw Open.* (2020) **3** e2019686. DOI: 10.1001/jamanetworkopen.2020.19686
17. Derkatch C. **“Wellness” as incipient illness: dietary supplements in a biomedical culture**. *Present Tense.* (2012) **2** 1-10
18. Delivett CP, Klepacz NA, Farrow CV, Thomas JM, Raats MM, Nash RA. **Front-of-pack images can boost the perceived health benefits of dietary products**. *Appetite.* (2020) **155** 104831. DOI: 10.1016/j.appet.2020.104831
19. Nichter M, Thompson J. **For my wellness, not just my illness: North Americans' use of dietary supplements culture, medicine, and psychiatry**. *Cult Med Psychiatry.* (2006) **30** 175-222. DOI: 10.1007/s11013-006-9016-0
20. Wu W-Y, Linn CT, Fu C-S, Sukoco BM. **The role of endorsers, framing, and rewards on the effectiveness of dietary supplement advertisements**. *J Health Commun.* (2012) **17** 54-75. DOI: 10.1080/10810730.2011.585689
21. Morales DX, Morales SA, Beltran TF. **Racial/ethnic disparities in household food insecurity during the COVID-19 pandemic: a nationally representative study**. *J Racial Ethnic Health Dispar.* (2021) **8** 1300-14. DOI: 10.1007/s40615-020-00892-7
22. Schanzenbach DW, Pitts A. *How Much Has Food Insecurity Risen? Evidence From the Census Household Pulse Survey.* (2020)
23. Gross R, Edmunds M, Schwartz P. *Food Security: A Community Driver of Health American Public Health Association, AcademyHealth, and Kaiser Permanente.* (2021)
24. Donnelly R, Farina MP. **How do state policies shape experiences of household income shocks and mental health during the COVID-19 pandemic?**. *Soc Sci Med.* (2021) **269** 113557. DOI: 10.1016/j.socscimed.2020.113557
25. Garcia MA, Homan PA, García C, Brown TH. **The color of COVID-19: structural racism and the disproportionate impact of the pandemic on older Black and Latinx adults**. *J Gerontol B Psychol Sci Soc Sci.* (2021) **76** e75-e80. DOI: 10.1093/geronb/gbaa114
26. Cowan AE, Jun S, Gahche JJ, Tooze JA, Dwyer JT, Eicher-Miller HA. **Dietary supplement use differs by socioeconomic and health-related characteristics among U.S. adults, NHANES 2011–2014**. *Nutrients.* (2018) **10** 1-12. DOI: 10.3390/nu10081114
27. Austin SB Yu K, Liu SH, Dong F, Tefft N. **Household expenditures on dietary supplements sold for weight loss, muscle building, and sexual function: disproportionate burden by gender and income**. *Prev Med Rep.* (2017) **6** 236-41. DOI: 10.1016/j.pmedr.2017.03.016
28. Pillitteri JL, Shiffman S, Rohay JM, Harkins AM, Burton SL, Wadden TA. **Use of dietary supplements for weight loss in the United States: results of a national survey**. *Obesity (Silver Spring).* (2008) **16** 790-6. DOI: 10.1038/oby.2007.136
29. Grebow J. **Dietary supplement sales success post-COVID: How can industry keep the momentum going after the pandemic?**. *Nutr Outlook* (2021)
30. Wheatley VM, Spink J. **Defining the public health threat of dietary supplement fraud**. *Comprehens Rev Food Sci Food Saf.* (2013) **12** 599-613. DOI: 10.1111/1541-4337.12033
31. 31.Congressional Research Service. Regulation of Dietary Supplements: Background and Issues for Congress (R43062). Washington, DC: Congressional Research Service (2021).. *Regulation of Dietary Supplements: Background and Issues for Congress (R43062)* (2021)
32. Hua SV, Granger B, Bauer K, Roberto CA. **A content analysis of marketing on the packages of dietary supplements for weight loss and muscle building**. *Prev Med Rep.* (2021) **23** 1-5. DOI: 10.1016/j.pmedr.2021.101504
33. Solomon CG, Willett WC, Carey VJ, Rich-Edwards J, Hunter DJ, Colditz GA. **A prospective study of pregravid determinants of gestational diabetes mellitus**. *J Am Med Assoc.* (1997) **278** 1078-83. DOI: 10.1001/jama.1997.03550130052036
34. Bao Y, Bertoia ML, Lenart EB, Stampfer JM, Willett WC, Speizer FE. **Origin, methods, and evolution of the three Nurses' Health Studies**. *Am J Public Health.* (2016) **106** 1573-81. DOI: 10.2105/AJPH.2016.303338
35. Field AE, Camargo CA Jr, Taylor CB, Berkey CS, Roberts SB, Colditz GA. **Peer, parent and media influences on the development of weight concerns and frequent dieting among preadolescent and adolescent girls and boys**. *Pediatrics.* (2001) **107** 54-60. DOI: 10.1542/peds.107.1.54
36. 36.US Department of Agriculture. Guide to Measuring Household Food Insecurity: Revised 2000. US Department of Agriculture (2000). Available online at: https://www.fns.usda.gov/guide-measuring-household-food-security-revised-2000 (accessed July 31, 2022).. *Guide to Measuring Household Food Insecurity: Revised 2000. US Department of Agriculture* (2000)
37. 37.US Department of Agriculture. Food Insecurity in the US: Measurement. (2022). Available online at: https://www.ers.usda.gov/topics/food-nutrition-assistance/food-security-in-the-u-s/measurement/ (accessed July 31, 2022).. *Food Insecurity in the US: Measurement* (2022)
38. Rogers AM, Lauren BN, Woo Baidal JA, Ozanne EM, Hur C. **Persistent effects of the COVID-19 pandemic on diet, exercise, risk for food insecurity, and quality of life: a longitudinal study among US adults**. *Appetite.* (2021) **167** 1-7. DOI: 10.1016/j.appet.2021.105639
39. Adams EL, Caccavale LJ, Smith D, Bean MK. **Longitudinal patterns of food insecurity, the home food environment, and parent feeding practices during COVID-19**. *Obes Sci Pract.* (2021) **7** 415-24. DOI: 10.1002/osp4.499
40. Kroenke K, Spitzer RL, Williams JB. **The Patient Health Questionnaire-2: validity of a two-item depression screener**. *Med Care* (2003) **2003** 1284-92. DOI: 10.1097/01.MLR.0000093487.78664.3C
41. Kroenke K, Spitzer RL, Williams JB, Monahan PO, Löwe B. **Anxiety disorders in primary care: prevalence, impairment, comorbidity, and detection**. *Ann Intern Med.* (2007) **146** 317-25. DOI: 10.7326/0003-4819-146-5-200703060-00004
42. Cohen S, Kamarck T, Mermelstein R. **A global measure of perceived stress**. *J Health Soc Behav.* (1983) **24** 385-96. DOI: 10.2307/2136404
43. Lazarus RS, Folkman S. *Manual for the Hassles and Uplift Scales, Research Edition.* (1989). DOI: 10.1037/t06473-000
44. Kanner AD, Coyne JC, Schaefer C, Lazarus RS. **Comparison of two modes of stress measurement: daily hassles and uplifts versus major life events**. *J Behav Med.* (1981) **4** 1-39. DOI: 10.1007/BF00844845
45. Fewell Z, Hernán MA, Tilling K, Choi H, Sterne JAC. **Controlling for time-dependent confounding using marginal structural models**. *Stata J.* (2004) **4** 402-20. DOI: 10.1177/1536867X0400400403
46. VanderWeele TJ, Hawkley LC, Cacioppo JT. **On the reciprocal association between loneliness and subjective well-being**. *Am J Epidemiol.* (2012) **176** 777-84. DOI: 10.1093/aje/kws173
47. Halekoh U, Højsgaard S, Yan J. **The R package geepack for generalized estimating equations**. *J Stat Softw.* (2006) **15** 1-11. DOI: 10.18637/jss.v015.i02
48. Pomeranz JL, Barbosa G, Killian C, Austin SB. **The dangerous mix of adolescents and dietary supplements for weight loss and muscle building: legal strategies for state action**. *J Public Health Manag Pract.* (2015) **21** 496-503. DOI: 10.1097/PHH.0000000000000142
|
---
title: Gradual deterioration of fatty liver disease to liver cancer via inhibition
of AMPK signaling pathways involved in energy-dependent disorders, cellular aging,
and chronic inflammation
authors:
- Sha-Sha Meng
- Hong-Wei Gu
- Ting Zhang
- Yu-Sang Li
- He-Bin Tang
journal: Frontiers in Oncology
year: 2023
pmcid: PMC10018212
doi: 10.3389/fonc.2023.1099624
license: CC BY 4.0
---
# Gradual deterioration of fatty liver disease to liver cancer via inhibition of AMPK signaling pathways involved in energy-dependent disorders, cellular aging, and chronic inflammation
## Abstract
### Introduction
Hepatocellular carcinoma (HCC) is the most prevalent primary liver cancer kind. According to recent research, a fatty liver increases the risk of hepatocellular cancer. Nevertheless, the AMPK signaling pathway is crucial. In addition, 5’-AMP-activated protein kinase (AMPK) is strongly linked to alterations in the tumor microenvironment, such as inflammation, hypoxia, and aging. The objective of this study is to evaluate the impact of the AMPK signaling pathway on the progression of fatty liver to HCC.
### Methods
In this study, we established a mouse liver cancer model using high-fat diets and nano-nitrosamines (nano-DEN). In addition, we employed a transcriptomic technique to identify all mRNAs detected in liver samples at the 25th weekexpression of proteins linked with the LKB1-AMPK-mTOR signaling pathway, inflammation, aging, and hypoxia was studied in microarrays of liver cancer tissues from mice and humans. These proteins included p-AMPK, LKB1, mTOR, COX-2, β-catenin, HMGB1, p16, and HIF-1α.
### Results
Data were collected at different times in the liver as well as in cancerous and paracancerous regions and analyzed by a multispectral imaging system. The results showed that most of the genes in the AMPK signaling pathway were downregulated. Prakk1 expression was upregulated compared to control group but downregulated in the cancerous regions compared to the paracancerous regions. Stk11 expression was downregulated in the cancerous regions. Mtor expression was upregulated in the cancerous regions. During liver cancer formation, deletion of LKB1 in the LKB1-AMPK-mTOR signaling pathway reduces phosphorylation of AMPK. It contributed to the upregulation of mTOR, which further led to the upregulation of HIF1α. In addition, the expression of β-catenin, COX-2, and HMGB1 were upregulated, as well as the expression of p16 was downregulated.
### Discussion
These findings suggest that changes in the AMPK signaling pathway exacerbate the deterioration of disrupted energy metabolism, chronic inflammation, hypoxia, and cellular aging in the tumor microenvironment, promoting the development of fatty liver into liver cancer.
## Introduction
Primary liver cancer will be the third biggest cause of cancer deaths worldwide in 2020. There will be approximately 906,000 new cases of liver cancer and 830,000 deaths worldwide in 2020, with the number of deaths close to the number of new cases [1]. Nonalcoholic fatty liver disease (NAFLD) has a global prevalence of $25\%$, is one of the primary causes of cirrhosis and HCC, and is a very common chronic disease believed to be linked to metabolic dysfunction [2]. NAFLD is an increasingly significant risk factor for liver cancer development [3]. It is estimated that about $\frac{1}{4}$ of the world’s population has NAFLD, and the incidence of NAFLD-related HCC is increasing at an annual rate of $9\%$ [4].
NAFLD is a prevalent cause of cryptogenic cirrhosis and HCC of unknown origin, and the development of NAFLD-HCC is linked to alterations in inflammatory and tumor-related signaling pathways [5]. However, the specific mechanism by which NAFLD develops into HCC remains unclear. Therefore, it is vital to identify new treatment targets to monitor the evolution of liver cancer that originates from a fatty liver. It has been shown that p-AMPK inhibits NAFLD and disrupts fat formation in HCC [6, 7]. AMPK is an energy receptor that senses glucose availability independently of changes in adenine nucleotides [8]. Activation of the AMPK signaling pathway inhibits the glycolysis pathway, which in turn inhibits the progression of liver cancer [9]. AMPK is a key molecule in the regulation of biological energy metabolism and plays an important role in this process.
Chronic liver inflammation is associated with tumorigenesis [10]. In most cases, liver cancer occurs in the context of chronic liver inflammation caused by viral hepatitis and alcoholic or nonalcoholic steatohepatitis [11]. Deregulated Wnt/β-catenin signaling is one of the main genetic alterations in human liver cancer [12]. In addition, there is growing evidence that abnormalities in Wnt/β-catenin signaling promote the development and progression of liver cancer [13].
In the majority of instances, liver cancer develops in the context of chronic liver inflammation resulting from viral hepatitis and alcoholic or nonalcoholic steatohepatitis. HCC exists in an immunosuppressed milieu that favors tumor evasion, and hypoxia can influence intercellular communication in the microenvironment of the tumor [14, 15]. Hypoxia is a key factor in the induction of transcription of HIF1A, HIF-1α protein accumulates to promote angiogenesis [16]. It is currently believed that HIF-1 and AMPK are important regulators of the switch between reprogramming and oxidative metabolism [17].
Cellular senescence is a program that prevents the malignant transformation of senescent cells following oncogenic pathway activation and DNA damage. Senescent cells are metabolically active and secrete cytokines and chemokines that shape the function and composition of their microenvironment [18]. Increased expression of the p16 oncogene associated with cellular senescence leads to senescence. In contrast, cancer cell overcome this effect by inactivating the gene through homologous deletion or hypermethylation [19].
The specific mechanism by which nonalcoholic fatty liver develops into HCC remains unclear. AMPK, inflammation-related β-catenin, COX-2, HMGB1, hypoxia-related HIF-1α, and cellular senescence-related p16 are all important in the development of liver cancer. Therefore, this paper explores the influence and mechanism of the AMPK signaling pathway on the tumor microenvironment during the progression from fatty liver to liver cancer. It can provide a reference for the treatment of liver cancer that develops from fatty liver.
## Reagents
The following drug was used in this experiment: diethylnitrosamine (DEN; Tokyo Chemical Industry Co., Ltd., Tokyo, Japan). The antibodies used in this study included antibodies against β-catenin, COX-2, HMGB1, p16, HIF-1α (Abcam Inc., Cambridge, MA, UK), anti-phospho-AMPKα (pThr172; Sigma Chemical Co., St. Louis, MO, USA), LKB1 and mTOR (ABclonal Technology, Wuhan, China). The instruments used in the experiment were as follows: a multispectral quantitative instrument (Cambridge Research & Instrumentation, Inc., Woburn, MA, USA, model Nuance), a fusion instrument (Servicebio, Wuhan, China, Model F-16), and a sampling handle (Servicebio, Wuhan, China, model SH-20).
## Animal care
Before the initiation of experiments, 25 male Kunming mice (18-22 g) from the Hubei Experimental Animal Center were acclimatized for 7 days under SPF conditions. The animals were reared in a temperature-controlled animal feeding center (20-25°C) with a 12-hour light-dark cycle. Kunming mice were randomly divided into two groups: control ($$n = 5$$), model ($$n = 20$$). The control group was fed a normal diet, while the model group was fed a high-fat diet. During the experiment, the animals in the model group were injected intraperitoneally with the chemical poison nano-DEN (16.5 mg/kg) weekly [20].
The control group did not have any treatment, and the mice were sacrificed on the first day of the experiment. When the mice were sacrificed, the livers were collected immediately and collected for histological and protein analysis. Twenty mice from the model group were sacrificed, and liver tissues were collected from mice at the 10th, 15th, 20th, and 30th week, with five mice sacrificed each week. Then, protein levels were assessed by immunohistochemical staining. The livers of the control, HF (high-fat), ND (nano-DEN), and HFND (co-exposure of high-fat diet and nano-DEN) group mice were transcriptomic analyzed at the 25th week. The livers of mice in the HFND group were divided into HFNDC (high fat-diet and nano-DEN’s co-exposure to cancerous regions) and HFNDP (high fat-diet and nano-DEN’s co-exposure to paracancerous regions). The data that support the findings of this study have been deposited into CNGB Sequence Archive [21] of China National GeneBank DataBase [22] with accession number CNP0003645. For hepatocellular steatosis, specimens were scored 0 – 3 points. 0 points: no fat; 1 point: steatosis occupying < $33\%$ of the hepatic parenchyma; 2 points: 34 – $66\%$ of the hepatic parenchyma; and 3 points: more than $66\%$ of the hepatic parenchyma [23].
The protocols for the management and usage of animals and experiments for this study were performed by the Laboratory Animal Society of the Research Facilities Committee. All research on mice was approved by the Animal Experiment Ethics Committee of South-Central Minzu University, Wuhan, China (permit number: 2018-SCUEC-AEC-010).
## Human samples and their tissue micro-array (TMA) analysis
Liver tissue samples from 30 liver cancer patients were obtained from the Key Laboratory of Chinese Internal Medicine of MOE of the Beijing Dongzhimen Hospital of Beijing University of Chinese Medicine. The cancerous regions were obtained from carcinoma tissues and reviewed by professional pathologists. The paracancerous regions were obtained from normal tissues over 3 cm from the borderline of the cancerous regions in the same fixed slice. The collection and follow-up manipulation of all pathological tissue samples were approved by the Committee on the Ethics of Experiments of South-Central Minzu University in China (Permit Number: 2017-SCUEC-MEC-007).
The tissues were extracted using a tissue chip handle (2 mm) and then put into a 96-well paraffin mold (2 mm). After that, the mold was heated to keep the tissue flat. The remaining paraffin was added to the mold to fill the gaps between the holes. The TMA was made after the tissue had cooled [24].
## Immunohistochemical analysis
The sliced sections were immersed in xylene followed by absolute alcohol for dewaxing and dehydration. After incubation in $3\%$ H2O2 hydrogen peroxide, for antigen retrieval, sections were heated in a microwave oven for 10 min in 0.01 M citric saline (pH = 6.0). After blocking in an oven at 37°C for 1 hour with the appropriate serum albumin, the sections were incubated with anti-phospho-AMPKα (pThr172; 1:100 dilution), LKB1 (1:100 dilution), mTOR (1:100 dilution), β-catenin (1:200 dilution), COX-2 (1:200 dilution), HMGB1 (1:200 dilution), p16 (1:200 dilution) and HIF-1α (1:100 dilution) antibodies overnight. After the corresponding secondary antibodies were directly applied to the tissue sections and incubated at 37°C for 1 hour, they were confirmed to be microscopic positive by staining with diaminobenzidine (DAB) chromophore followed by counterstaining with hematoxylin. Finally, the slices were sequentially dehydrated and sealed by gradient alcohol and xylene. As a negative control, $1\%$ bovine serum albumin (BSA) was used to replace the primary antibody on sections that were proven to be positive for p-AMPKα (pThr172), LKB1, mTOR, β-catenin, COX-2, HMGB1, p16 and HIF-1α in the present experiments. Then, imaging analysis of sections was performed by using a Nikon 50i light microscope imaging system (Nikon).
## Statistical analysis
Prism 9 (GraphPad Software) was used to process the data. Bar charts were constructed, and statistical analyses were performed. The quantitative immunohistochemical staining results showed positive expression intensity. All data are presented as the mean ± SEM. A paired t-test was used to compare the differences between the two groups in the cancerous and paracancerous regions, and an unpaired t-test was used to compare the differences between the groups at different times and the control group. *, $p \leq 0.05$, **, $p \leq 0.01$, and ***, $p \leq 0.001$ were considered statistically significant, very significant, and extremely significant, respectively, and $p \leq 0.05$ was considered not significant.
## Analysis of differential gene expression of the AMPK signaling pathway in the livers of mice with liver cancer
In Figure 1A, we were able to observe the pathological structural changes in the livers of different groups of mice from several representative images. In the control group, the liver structure was intact and clear, while the HF group showed a large number of fat vacuoles (2.60 ± 0.24, $p \leq 0.001$). Multiple inflammatory sites and abnormal aggregation of cells with partial fatty degeneration (1.60 ± 0.40, $p \leq 0.001$) could be seen in the ND group. In the HFND group, there was an accumulation of abnormal cells with enlarged nuclei and multiple points of inflammation in the cancerous regions, which were distinctly different from normal liver tissue and showed more steatosis (2.20 ± 0.20, $p \leq 0.001$). In our earlier research, it was found that only mice treated with nano-DEN showed significant tumor nodules at the 25th to 35th week [25]. However, very obvious tumor nodules already appeared at the 25th week after co-exposure of high-fat diet and nano-DEN, which significantly accelerated the carcinogenesis process.
**Figure 1:** *General view of mouse livers and KEGG pathways: AMPK signaling pathway and mRNA expression of Prakk1, Stk11, and Mtor. (A) General view of the liver samples and Hepatocellular steatosis score. (B) KEGG pathways: Changes in AMPK signaling pathway genes in different groups. (C) Statistical quantification of the mRNA levels of Prakk1, Stk11, and Mtor in the livers of mice. Upregulated genes are marked in red, and downregulated genes are marked in green. The experimental data are expressed as the mean ± SEM. Compared to the control, *p < 0.05; **p < 0.01; ***p < 0.001. Compared to HFNDP, #
p < 0.05. Scale bar, 50 µm and 200 µm.*
We also analyzed genes with differential expression in the AMPK signaling pathway in the livers of mice with liver cancer. We found that $60.0\%$ (HF vs. control; $\frac{21}{35}$), $72.7\%$ (ND vs. control; $\frac{8}{11}$), $76.9\%$ (HFND vs. control; $\frac{20}{26}$), and $83.3\%$ (HFNDC vs. HFNDP; $\frac{5}{6}$) of the genes in the AMPK signaling pathway were downregulated (Figure 1B).
According to Figure 1C, the expression of Prakk1 was upregulated in the HF group (1.98 ± 0.07, $p \leq 0.001$), ND group (1.35 ± 0.20, $p \leq 0.05$), and the HFNDC group (1.73 ± 0.18, $p \leq 0.05$) compared to the control group (1.27 ± 0.07), while the expression was downregulated in the HFNDC compared to the HFNDP (2.40 ± 0.09, $p \leq 0.05$). Stk11 expression was upregulated in the HF group (18.32 ± 0.66, $p \leq 0.05$) and ND group (17.63 ± 1.38, $p \leq 0.05$) compared to the control group (16.81 ± 0.41). Conversely, Stk11 expression was downregulated in the HFNDC (13.01 ± 0.40, $p \leq 0.001$), while cancerous regions was downregulated compared to the HFNDP (18.81 ± 1.56, $p \leq 0.05$). The expression of Mtor was upregulated in the HF group (4.14 ± 0.08, $p \leq 0.05$), ND group (3.43 ± 0.11, $p \leq 0.05$), and the HFNDC (5.29 ± 0.35, $p \leq 0.01$) compared to the control group (2.90 ± 0.47), while HFNDC was also significantly upregulated compared to the HFNDP (3.71 ± 0.20, $p \leq 0.05$). So we can see that the AMPK signaling pathway and related genes show more severe changes in the HFND group than in the ND group during the same period.
## Downregulation of LKB1 and p-AMPK expression and upregulation of mTOR expression in liver tissues of mice with liver cancer
The expression levels of LKB1, p-AMPK, and mTOR proteins in the liver tissues are summarized in Figure 2. The experimental results were as follows: in Figures 2A, B, compared with the control group, p-AMPK expression was continuously increased from the 10th to the 20th week, and the expression in the cancerous regions was decreased at the 30th week. Compared with the paracancerous regions (12316 ± 1115), the expression in the cancerous regions (8715 ± 1395) was downregulated (downregulated to $71\%$ of the paracancerous regions, $p \leq 0.05$). At the 20th week, a large number of fat vacuoles appeared in the liver tissues of mice, the expression of p-AMPKα was abnormal with time, and liver metabolism was disturbed (Figure 2A).
**Figure 2:** *Immunohistochemical staining of p-AMPK, mTOR, and LKB1 proteins in mouse livers at different time points and their quantitative representative images. (A) Representative images of immunohistochemical staining of p-AMPK in mouse livers during liver carcinogenesis. (B) Quantitative multispectral image of the p-AMPK protein. (C) Representative images of immunohistochemical staining of mTOR in mouse livers during liver carcinogenesis. (D) Quantitative multispectral image of the mTOR protein. (E) Representative images of immunohistochemical staining of LKB1 in mouse livers during liver carcinogenesis. (F) Quantitative multispectral image of the LKB1 protein. The experimental data are expressed as the mean ± SEM. Compared to the control, *p < 0.05; **p < 0.01; ***p < 0.001. Compared to paracancerous regions, *p < 0.05. Scale bar, 50 µm. ns, not statistically.*
As shown in Figures 2C, D, compared with the control group, the expression levels of mTOR continuously increased from the 10th to the 30th week. Compared with the paracancerous regions (10868 ± 1867), the expression in the cancerous regions (28626 ± 4396) was upregulated (upregulated to $263\%$ of that in the paracancerous regions, $p \leq 0.05$).
As shown in Figures 2E, F, compared with the control group, the total expression of LKB1 was continuously increased from the 10th to the 15th week, and decreased at the 20th week. The expression in the cancerous regions was significantly decreased at the 30th week. Compared with the paracancerous regions (25991 ± 4322), the expression in the cancerous regions (11925 ± 3046) was downregulated (downregulated to $46\%$ of that in the paracancerous regions, $p \leq 0.05$). Compared with the nuclear expression of LKB1 in the paracancerous regions (4048 ± 649), cancerous regions (2331 ± 331) was downregulated (downregulated to $58\%$ of that in the paracancerous regions, $p \leq 0.05$). Compared with the paracancerous regions (3.50 ± 0.48), the nucleation rate in the cancerous regions (1.82 ± 0.32) was downregulated (downregulated to $52\%$ of that in the paracancerous regions, $p \leq 0.01$). LKB1 may be decompensated and increased in the process of liver cancer, and its expression in cancer tissues is absent after carcinogenesis. LKB1 is an upstream kinase of AMPK and regulates AMPK phosphorylation. Therefore, p-AMPK increases first and decreases after carcinogenesis, leading to increased expression of mTOR.
## Expression of the inflammation-related proteins β-catenin, COX-2, and HMGB1 was upregulated in liver tissues of mice with liver cancer
The expression levels of β-catenin, COX-2, and HMGB1 in the liver tissues are illustrated in Figure 3. We found that compared with the control group, the expression of β-catenin increased from the 10th to the 20th week, and decreased at the 30th week in the cancerous regions (Figures 3A, B). In addition, compared with the paracancerous regions (25852 ± 6759), the expression of the cancerous regions (31316 ± 4932) was upregulated (upregulated to $121\%$ of that in the paracancerous regions, $p \leq 0.05$). Compared with the nuclear expression of β-catenin in the paracancerous regions (5343 ± 831), cancerous regions (10748 ± 2127) was upregulated (upregulated to $201\%$ of that in the paracancerous regions, $p \leq 0.05$). Compared with the paracancerous regions (4.92 ± 0.67), the nucleation rate of the cancerous regions (11.48 ± 1.60) was upregulated (upregulated to $233\%$ of that in the paracancerous regions, $p \leq 0.05$).
**Figure 3:** *Immunohistochemical staining of β-catenin, COX-2, and HMGB1 proteins in mouse livers at different time points and their quantitative representative images. (A) Representative images of immunohistochemical staining of β-catenin in mouse livers during liver carcinogenesis. (B) Quantitative multispectral image of the β-catenin protein. (C) Representative images of immunohistochemical staining of COX-2 in mouse livers during liver carcinogenesis. (D) Quantitative multispectral image of the COX-2 protein. (E) Representative images of immunohistochemical staining of HMGB1 in mouse livers during liver carcinogenesis. (F) Quantitative multispectral image of the HMGB1 protein. The experimental data are expressed as the mean ± SEM. Compared to the control, *p < 0.05; **p < 0.01; ***p < 0.001. Compared to paracancerous regions, *p < 0.05; **p < 0.01. Scale bar, 50 µm. ns, not statistically.*
As shown in Figures 3C, D, compared with the control group, COX-2 expression continuously increased from the 10th to the 30th week, and the highest expression was observed at the 20th week. Compared with the paracancerous regions (26686 ± 3473), the expression in the cancerous regions (39816 ± 3673) was upregulated (upregulated to $149\%$ of that in the paracancerous regions, $p \leq 0.05$). As shown in Figures 3E, F, compared with the control group, HMGB1 expression continuously increased from the 10th week to the 30th week. Compared with the paracancerous regions (20295 ± 5462), the expression in the cancerous regions (37954 ± 6362) was upregulated (upregulated to $187\%$ of that in the paracancerous regions, $p \leq 0.05$).
## Downregulation of aging factor p16 expression and upregulation of hypoxic factor HIF-1α expression in liver tissues of liver cancer mice
Figure 4 shows the expression levels of p16 and HIF-1α in the liver tissues. As shown in Figures 4A, B, compared with the control group, the total expression of p16 continuously increased from the 10th to the 20th week, and the expression in the cancerous regions decreased at the 30th week. Compared with the paracancerous regions (33703 ± 3711), the expression in the cancerous regions (20421 ± 3377) was downregulated (downregulated to $61\%$ of that in the paracancerous regions, $p \leq 0.01$). Compared with the nuclear expression of p16 in the paracancerous regions (9668 ± 1272), cancerous regions (5828 ± 971) was downregulated (downregulated to $60\%$ of that in the paracancerous regions, $p \leq 0.01$). Compared with the paracancerous regions (6.90 ± 0.86), the nucleation rate of the cancerous regions (5.14 ± 0.73) was downregulated (downregulated to $75\%$ of that in the paracancerous regions, $p \leq 0.05$).
**Figure 4:** *Immunohistochemical staining of p16 and HIF-1α proteins in mouse livers at different time points and their quantitative representative images. (A) Representative images of immunohistochemical staining of p16 in mouse livers during liver carcinogenesis. (B) Quantitative multispectral image of the p16 protein. (C) Representative images of immunohistochemical staining of HIF-1α in mouse livers during liver carcinogenesis. (D) Quantitative multispectral image of the HIF-1α protein. The experimental data are expressed as the mean ± SEM. Compared to the control, *p < 0.05; **p < 0.01; ***p < 0.001. Compared to paracancerous regions, *p < 0.05. Scale bar, 50 µm. ns, not statistically.*
According to Figures 4C, D, compared with the control group, the total expression of HIF-1α was continuously increased from the 10th to the 30th week, and compared with the expression in the paracancerous regions (43064 ± 5037), cancerous regions (67346 ± 9082) was upregulated (upregulated to $156\%$ of that in the paracancerous regions, $p \leq 0.05$). Compared with the nuclear expression of HIF-1α in the paracancerous regions (3551 ± 763), cancerous regions (5400 ± 1104) was upregulated (upregulated to $152\%$ of that in the paracancerous regions, $p \leq 0.05$). Compared with the paracancerous regions (3.24 ± 0.98), the nucleation rate of the cancerous regions (5.20 ± 0.95) was upregulated (upregulated to $161\%$ of that in the paracancerous regions, $p \leq 0.05$).
## Downregulation of LKB1 and p-AMPK expression and upregulation of mTOR expression in cancerous regions of liver cancer patientsns
As shown in Figures 5A, B, compared with the paracancerous regions (18103 ± 2668), the expression of p-AMPK in the cancerous regions (10687 ± 1062) was downregulated (downregulated to $59\%$ of that in the paracancerous regions, $p \leq 0.01$). Moreover, compared with the well-differentiated group, the expression of p-AMPK in the cancerous regions of the moderately and poorly differentiated group was downregulated more significantly ($p \leq 0.05$).
**Figure 5:** *Immunohistochemical staining of p-AMPK, mTOR, and LKB1 proteins in liver cancer tissues and their quantitative images. (A) Representative image of immunohistochemical staining of the p-AMPK protein in liver tissues. (B) Quantitative multispectral image of the p-AMPK protein. (C) Representative image of immunohistochemical staining of the mTOR protein in liver tissues. (D) Quantitative multispectral image of the mTOR protein. (E) Representative image of immunohistochemical staining of the LKB1 protein in liver tissues. (F) Quantitative multispectral image of the LKB1 protein. The experimental data are expressed as the mean ± SEM. Compared to paracancerous regions, *p < 0.05; **p < 0.01; ***p < 0.001. Scale bar, 50 µm. ns, not statistically.*
According to Figures 5C, D, the expression of mTOR was significantly upregulated in the cancerous regions (14624 ± 1461) compared with the paracancerous regions (7194 ± 920; upregulated to $203\%$ of that in the paracancerous regions, $p \leq 0.001$). However, there was no significant difference in tissues expression between the cancerous and paracancerous regions in the well-differentiated group. However, the expression of mTOR was more significantly upregulated in the cancerous regions of the intermediate and moderately and poorly differentiated group ($p \leq 0.001$) compared to the well-differentiated group.
However, compared with the paracancerous regions (14802 ± 1524), the total expression of LKB1 in the cancerous regions (10348 ± 1007) was downregulated (downregulated to $70\%$ of that in the paracancerous regions, $p \leq 0.01$). Compared with the well-differentiated group, the total expression in the moderately and poorly differentiated group was downregulated more significantly ($p \leq 0.01$). Compared with the paracancerous regions (5595 ± 594), the nuclear expression of LKB1 in the cancerous regions (3989 ± 474) was downregulated (downregulated to $71\%$ of that in the paracancerous regions, $p \leq 0.05$). Compared with the paracancerous regions (5.96 ± 0.60), the nucleation rate in the cancerous regions (3.91 ± 0.49) was downregulated (downregulated to $66\%$ of that in the paracancerous regions, $p \leq 0.05$; Figures 5E, F).
## Expression of the inflammation-related proteins β-catenin, HMGB1, and COX-2 was upregulated in the cancerous regions of liver cancer patients
As illustrated in Figures 6A, B, we found that the total expression of β-catenin in the cancerous regions, the nuclear expression, and the nucleation rate were significantly upregulated compared with those in the paracancerous regions. Compared with the paracancerous regions (61827 ± 4985), the total expression of β-catenin in the cancerous regions (82403 ± 7283) was upregulated (upregulated to $133\%$ of that in the paracancerous regions, $p \leq 0.01$). Compared with the well-differentiated group, the total expression in the moderately and poorly differentiated group increased more significantly ($p \leq 0.01$). Compared with the paracancerous regions (23083 ± 1824), the nuclear expression of β-catenin in the cancerous regions (37680 ± 3402) was upregulated (upregulated to $163\%$ of that in the paracancerous regions, $p \leq 0.001$), and the cancerous regions of the moderately and poorly differentiated group was upregulated more obviously than that in the well-differentiated group ($p \leq 0.001$). Compared with the paracancerous regions (16.55 ± 1.04), the nucleation rate in the cancerous regions (21.35 ± 1.51) was significantly increased (upregulated to $129\%$ of that in the paracancerous regions, $p \leq 0.01$), and the cancerous regions of the moderately and poorly differentiated group was increased more significantly than that of the well-differentiated group ($p \leq 0.001$).
**Figure 6:** *Immunohistochemical staining of β-catenin, HMGB1, and COX-2 proteins in liver cancer tissues and their quantitative images. (A) Representative image of immunohistochemical staining of the β-catenin protein in liver tissues. (B) Quantitative multispectral image of the β-catenin protein. (C) Representative image of immunohistochemical staining of the HMGB1 protein in liver tissues. (D) Quantitative multispectral image of the HMGB1 protein. (E) Representative image of immunohistochemical staining of the COX-2 protein in liver tissues. (F) Quantitative multispectral image of the COX-2 protein. The experimental data are expressed as the mean ± SEM. Compared to paracancerous regions, *p < 0.05; **p < 0.01; ***p < 0.001. Scale bar, 50 µm. ns, not statistically.*
We found that compared with the paracancerous regions (78445 ± 7836), the total expression of HMGB1 in the cancerous regions (97202 ± 8054) was upregulated (upregulated to $124\%$ of that in the paracancerous regions ($p \leq 0.01$). Compared with the well-differentiated group, the total expression in the moderately and poorly differentiated group increased more significantly ($p \leq 0.01$). Compared with the paracancerous regions (8068 ± 1038), the nuclear expression of HMGB1 in the cancerous regions (16655 ± 1983) was upregulated (upregulated to $206\%$ of that in the paracancerous regions, $p \leq 0.001$), and the cancerous regions of the moderately and poorly differentiated group was upregulated more obviously than that in the well-differentiated group ($p \leq 0.001$). Compared with the paracancerous regions (1.76 ± 0.23), the nucleation rate in the cancerous regions (3.71 ± 0.45) was significantly increased (upregulated to $211\%$ of that in the paracancerous regions, $p \leq 0.001$), and the cancerous regions of the moderately and poorly differentiated group was increased more significantly than that of the well-differentiated group ($p \leq 0.001$; Figures 6C, D).
As shown in Figures 6E, F, compared with that in the paracancerous regions (79676 ± 8100), the expression of COX-2 in the cancerous regions (110194 ± 8022) was upregulated (upregulated to $138\%$ of the paracancerous region, $p \leq 0.01$). However, COX-2 was upregulated more significantly in the moderately and poorly differentiated group than in the well-differentiated group ($p \leq 0.05$).
## Downregulation of aging factor p16 expression and upregulation of hypoxic factor HIF-1α expression in the cancerous regions of liver cancer patients
Figure 7 shows the expression levels of aging factor p16 and hypoxia factor HIF-1α in human liver cancer tissues. We found that the total expression, nuclear expression, and nucleation rate of p16 were downregulated in the cancerous regions compared to the paracancerous regions (Figures 7A, B). The total expression of p16 was downregulated in the cancerous regions (60204 ± 3320) compared to the paracancerous regions (70364 ± 4891; downregulated to $86\%$ of the paracancerous regions, $p \leq 0.05$). However, total expression was more significantly downregulated in the moderately and poorly differentiated groups than in the well-differentiated group ($p \leq 0.05$). Nuclear expression of p16 was downregulated (downregulated to $71\%$ of that in the paracancerous regions, $p \leq 0.05$) in the cancerous regions (26525 ± 1291) compared to the paracancerous regions (37352 ± 3442). The downregulation was more pronounced in the moderately and poorly differentiated group than in the well-differentiated group in the cancerous regions ($p \leq 0.05$). In addition, the nucleation rate of p16 was downregulated (downregulated to $75\%$ of that in the paracancerous regions, $p \leq 0.01$) in the cancerous regions (8.06 ± 0.41) compared to the paracancerous regions (10.77 ± 0.91), and it was more significantly downregulated ($p \leq 0.05$) in the moderately and poorly differentiated group compared to the well-differentiated group in the cancerous regions.
**Figure 7:** *Immunohistochemical staining of p16 and HIF-1α proteins in liver cancer tissues and their quantitative images. (A) Representative image of immunohistochemical staining of the p16 protein in liver tissues. (B) Quantitative multispectral image of the p16 protein. (C) Representative image of immunohistochemical staining of the HIF-1α protein in liver tissues. (D) Quantitative multispectral image of the HIF-1α protein. The experimental data are expressed as the mean ± SEM. Compared to paracancerous regions, *p < 0.05; **p < 0.01. Scale bar, 50 µm. ns, not statistically.*
As shown in Figures 7C, D, the total expression, nuclear expression, and nucleation rate of HIF-1α were upregulated in the cancerous regions compared to the paracancerous regions. The total expression of HIF-1α was upregulated in the cancerous regions (89486 ± 8137; upregulated to $155\%$ of that in paracancerous regions, $p \leq 0.001$) compared to the paracancerous regions (57765 ± 5427), and more significantly upregulated in the moderately and poorly differentiated group compared to the well-differentiated group ($p \leq 0.001$). The nuclear expression of HIF-1α was not significantly different (upregulated to $114\%$ of that in the paracancerous regions, $p \leq 0.05$) from that in the paracancerous regions (19158 ± 1766) compared to that in the cancerous regions (21932 ± 3184). The nucleation rate was significantly upregulated (upregulated to $150\%$ of that in the paracancerous regions, $p \leq 0.01$) from that in the cancerous regions (11.43 ± 1.21) compared to that of the paracancerous regions (7.64 ± 0.60). We found that the nucleation rate was more significantly downregulated ($p \leq 0.01$) in the moderately and poorly differentiated group than in the well-differentiated group in the cancerous regions.
## Correlation of p-AMPK with the expression of proteins related to LKB1, mTOR, inflammatory factors (β-catenin, COX-2, HMGB1), aging factor (p16), and hypoxia factor (HIF-1α) at the same site
Based on the quantitative results, we used the statistical methods of Pearson and significant difference to calculate and analyze the correlations between the expression of groups of two proteins in liver tissues (p-AMPK&LKB1, p-AMPK& mTOR, p-AMPK&β-catenin, p-AMPK&COX-2, p-AMPK&HMGB1, p-AMPK&HIF-1α, p-AMPK&p16). The results are shown in Figure 8. The results showed that p-AMPK was positively correlated with the expression of LKB1 ($r = 0.61$) and p16 ($r = 0.62$) and negatively correlated with the expression of mTOR (r = - 0.44), β-catenin (r = - 0.43), COX-2 (r = - 0.41), HMGB1 ($r = 0.13$) and HIF1α (r = - 0.25). The correlations between p-AMPK and the expression of these seven proteins in liver tissue were greater in the moderately and poorly differentiated group than in the well-differentiated group. We found a relatively high correlation between p-AMPK & LKB1, p-AMPK & p16.
**Figure 8:** *Correlation of p-AMPK with LKB1, mTOR, inflammatory factors (COX-2, β-catenin, HMGB1), aging factor (p16), and hypoxia factor (HIF-1α) proteins expressed at the same site in the same region of tumor tissue from patients with liver cancer. (A) Correlation between p-AMPK and LKB1 expression in the same region. (B) Correlation for p-AMPK and p16 expression in the same region. (C) Correlation of p-AMPK and HMGB1 expression in the same region. (D) Correlation of p-AMPK and mTOR expression in the same region. (E) Correlation of p-AMPK and β-catenin expression in the same region. (F) Correlation of p-AMPK and COX-2 expression in the same region. (G) Correlation of p-AMPK and HIF-1α expression in the same region. When r > 0, the two variables are positively correlated. When r < 0, the two variables are negatively correlated. When |r| ≥ 0.8, the two variables can be considered highly correlated. When 0.8 > |r| ≥ 0.5, the two variables can be considered moderately correlated. When 0.5 > |r| ≥ 0.3, the two variables can be considered slightly correlated, and |r| < 0.3 means the correlation is weak and uncorrelated.*
## Discussion
In recent years, increasing attention has been given to the changes in the tumor microenvironment of liver cancer. It is well accepted that a fatty liver is a risk factor for liver cancer [26]. In our study, we also found that the co-exposed group had more severe liver damage at the 25th week than mice given a high-fat diet or nano-DEN alone. In addition, by observing the livers of mice at various periods during the experiment, we determined that severe liver steatosis is also present during the development of liver cancer. AMPK is a protein kinase that inhibits lipid production via phosphorylation and deactivates important adipogenic genes such as SREBP-1 [27]. Further development of fatty liver leads to the formation of steatohepatitis, which is characterized by inflammation and hypoxia in the microenvironment [28]. AMPK is a key master switch that regulates lipid metabolism by directly phosphorylating proteins or regulating gene transcription in specific tissues such as the liver [29].
AMPK exhibits duality in tumors, where it is rapidly activated under conditions of hypoxia and a lack of nutrients to maintain metabolic homeostasis and support the survival of cancer cells [30, 31]. Therefore, activation of AMPK in the hypoxic tumor microenvironment may be the mechanism by which tumor cells enter a protective quiescent state [30]. There may be a mechanism through which AMPK is increased in a compensatory manner during tumor formation and assists cancer cells in adapting to their environment. The deterioration of the tumor microenvironment, such as in inflammation and hypoxia, mediated by AMPK promoted the formation of liver cancer. We found that severe steatosis of the liver occurs during the development of liver cancer. NAFLD also becomes a metabolic dysfunction-related fatty liver disease, while it is a risk factor for HCC [32]. AMPK activation reduces the formation of NAFLD [33]. We found that p-AMPKα expression first increases and then decreases during liver injury, and finally, liver metabolism may be disturbed to form liver cancer due to the poor microenvironment.
Significantly more genes in the AMPK signaling pathway were discovered to be downregulated than upregulated in this study. The expression of Prakk1 was downregulated compared to the paracancerous regions, but the expression was higher than that of the control group. Interestingly, its upstream gene Stk11 was less expressed in cancerous regions than in the control, while the downstream gene Mtor was upregulated in cancerous regions. Similarly, the expression of p-AMPK and LKB1 in mouse livers first increased and then decreased during the development of liver cancer and was ultimately downregulated in cancerous regions. The expression of the mTOR was continuously increased and upregulated in cancerous regions. Research shows that, AMPK negatively regulates aerobic glycolysis in tumor cells and inhibits tumor growth in vivo [34]. LKB1 is a tumor suppressor that is deficiently expressed in lung and breast cancers and has been identified as a key upstream kinase required for AMPK activation [35]. LKB1 expression is elevated in chronic liver disease and early-stage liver cancer, and LKB1 becomes a more likely target for other posttranslational modifications in early-stage liver cancer [36].
mTOR is typically increased in malignancies, such as HCC, and is linked to a poor prognosis, poor tumor differentiation, and early recurrence [37]. LKB1 has been reported to directly phosphorylate Thr172 of AMPKα in vitro and activate its kinase activity [38]. In liver cancer, the deletion of LKB1 downregulates AMPK and promotes the expression of mTOR [34]. These are consistent with our study’s findings. As a key upstream kinase that activates AMPK, LKB1 was downregulated in liver cancer, preventing AMPK activation and upregulating mTOR expression. The LKB1/AMPK/mTOR signaling pathway is aberrantly activated.
Our study showed that the expression of β-catenin, COX-2, and HMGB1 increased during the progression of liver cancer and that the expression was higher in cancerous regions than in paracancerous regions. The same was seen in human liver cancer samples. It has been found that activation of the Wnt/β-catenin signaling drives liver cancer formation and that β-catenin expression is abnormal in the majority of human liver cancer samples [34, 39]. In previous studies, we confirmed the relationship of the β-catenin/COX-2 ring [40]. Elevated β-catenin expression and nucleation promote COX-2 expression, and elevated COX-2 expression, in turn, promotes β-catenin expression, HMGB1 is located downstream of the β-catenin/COX-2 loop, and elevated expression of β-catenin and COX-2 promotes HMGB1 expression [40, 41]. It has been found that AMPK is an upstream signal of β-catenin, and the activation of AMPK can inhibit the activity of the Wnt signaling pathway and prevent the accumulation of β-catenin in the nucleus [42, 43]. Thus, the downregulation of AMPK exacerbates the involvement of the β-catenin/COX-2 loop and HMGB1 signaling in the inflammatory response. But more specific mechanisms still need to be further studied.
Protein phosphorylation is also tightly correlated with the level of oxidative stress in the cell; therefore, oxidative stress can promote phosphorylation of p16, resulting in the cessation of cell division and early senescence, thereby preventing the formation of tumors [44]. The expression of p16 increases significantly with age in a variety of rodent and human tissues in both healthy and diseased states. The p16 gene has been reported to be silenced in various human tumors [45, 46]. These are consistent with the results of the present study, in which we also found that p16 expression was first elevated during the formation of liver cancer and downregulated in the cancerous regions after tumor formation. Exogenous activation of LKB1/AMPK signaling increased p16 expression, and a decrease in p16 expression was observed in LKB1 knockdown cells [47, 48]. These could explain the change in p-AMPK upregulation and then downregulation over time in our results. The expression of p16 was similarly upregulated and then downregulated. Inhibition of the AMPK signaling pathway resulted in the downregulation of p16 expression.
Enhanced activation of AMPK in ovarian cancer inhibits aerobic glycolysis and promotes the degradation of HIF-1α ubiquitination [49]. It has been shown that in order to adapt to a hypoxic environment, tumor cells express a high level of HIF-1α, which stimulates cell metabolism and proliferation [50]. It has been found that the inactivation of AMPKα promotes a metabolic shift to aerobic glycolysis and that these metabolic effects require HIF-1α to maintain stability, as silencing HIF-1α reverses the shift to aerobic glycolysis and the biosynthetic and proliferative advantages of reduced AMPKα signaling in tumor cells [51]. Therefore, the high expression of HIF-1α in the development of liver cancer is influenced by AMPK.
There are many studies that have demonstrated the close relationship between AMPK and β-catenin, p16 and HIF-1α, respectively. However, there may be a mutual relationship between them. We will conduct a deeper study next.
## Conclusion
In summary, we can conclude that fatty liver deteriorates into liver cancer due to AMPK-mediated metabolic disorders, cellular aging and chronic inflammation in the liver. Malignant alterations in the tumor microenvironment inhibit the AMPK signaling pathway, resulting in a disruption of energy metabolism, chronic inflammation, fat deposition, and cellular aging. Promotes the progression of fatty liver to cancer of the liver.
## Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://db.cngb.org/cnsa/, CNP0003645.
## Ethics statement
The studies involving human participants were reviewed and approved by the Committee on the Ethics of Experiments of South-Central Minzu University in China (Permit Number: 2017-SCUEC-MEC-007). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements. The animal study was reviewed and approved by the Laboratory Animal Society of the Research Facilities Committee. All research on mice was approved by the Animal Experiment Ethics Committee of South-Central Minzu University, Wuhan, China (permit number: 2018-SCUEC-AEC-010).
## Authors contributions
S-SM: Investigation, Methodology, Data analysis, Writing - Original Draft. H-WG: Investigation, Methodology, Data analysis, Writing - Original Draft. TZ: Investigation, Methodology, Data analysis. Y-SL: Conceptualization, Data analysis, Writing - review & editing, Resource, Supervision, Funding acquisition. H-BT: Conceptualization, Data analysis, Writing - Original Draft, Writing - review & editing, Resource, Supervision, Funding acquisition, Project administration. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc.2023.1099624/full#supplementary-material
## References
1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A. **Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA Cancer J Clin* (2021) **71**. DOI: 10.3322/caac.21660
2. Powell EE, Wong VW, Rinella M. **Non-alcoholic fatty liver disease**. *Lancet* (2021) **397**. DOI: 10.1016/S0140-6736(20)32511-3
3. Yang H, Deng Q, Ni T, Liu Y, Lu L, Dai H. **Targeted inhibition of LPL/FABP4/CPT1 fatty acid metabolic axis can effectively prevent the progression of nonalcoholic steatohepatitis to liver cancer**. *Int J Biol Sci* (2021) **17**. DOI: 10.7150/ijbs.64714
4. Huang DQ, El-Serag HB, Loomba R. **Global epidemiology of NAFLD-related HCC: trends, predictions, risk factors and prevention**. *Nat Rev Gastroenterol Hepatol* (2021) **18**. DOI: 10.1038/s41575-020-00381-6
5. Younossi ZM, Koenig AB, Abdelatif D, Fazel Y, Henry L, Wymer M. **Global epidemiology of nonalcoholic fatty liver disease-meta-analytic assessment of prevalence, incidence, and outcomes**. *Hepatology* (2016) **64** 73-84. DOI: 10.1002/hep.28431
6. Liu G, Kuang S, Cao R, Wang J, Peng Q, Sun C. **Sorafenib kills liver cancer cells by disrupting SCD1-mediated synthesis of monounsaturated fatty acids**. *FASEB J* (2019) **33**. DOI: 10.1096/fj.201802619RR
7. Chyau CC, Wang HF, Zhang WJ, Chen CC, Huang SH, Chang CC. **Antrodan alleviates high-fat and high-fructose diet-induced fatty liver disease in C57BL/6 mice model**. *Int J Mol Sci* (2020) **21** 360. DOI: 10.3390/ijms21010360
8. Lin SC, Hardie DG. **AMPK: Sensing glucose as well as cellular energy status**. *Cell Metab* (2018) **27** 299-313. DOI: 10.1016/j.cmet.2017.10.009
9. Bi L, Ren Y, Feng M, Meng P, Wang Q, Chen W. **HDAC11 regulates glycolysis through the LKB1/AMPK signaling pathway to maintain hepatocellular carcinoma stemness**. *Cancer Res* (2021) **81**. DOI: 10.1158/0008-5472.CAN-20-3044
10. Wang G, Wang Q, Liang N, Xue H, Yang T, Chen X. **Oncogenic driver genes and tumor microenvironment determine the type of liver cancer**. *Cell Death Dis* (2020) **11** 313. DOI: 10.1038/s41419-020-2509-x
11. Matter MS, Marquardt JU, Andersen JB, Quintavalle C, Korokhov N, Stauffer JK. **Oncogenic driver genes and the inflammatory microenvironment dictate liver tumor phenotype**. *Hepatology* (2016) **63**. DOI: 10.1002/hep.28487
12. Xu C, Xu Z, Zhang Y, Evert M, Calvisi DF, Chen X. **β-catenin signaling in hepatocellular carcinoma**. *J Clin Invest* (2022) **132** e154515. DOI: 10.1172/JCI154515
13. He S, Tang S. **WNT/β-catenin signaling in the development of liver cancers**. *BioMed Pharmacother* (2020) **132**. DOI: 10.1016/j.biopha.2020.110851
14. Zhang B, Tang B, Gao J, Li J, Kong L, Qin L. **A hypoxia-related signature for clinically predicting diagnosis, prognosis and immune microenvironment of hepatocellular carcinoma patients**. *J Transl Med* (2020) **18** 342. DOI: 10.1186/s12967-020-02492-9
15. Wang ZH, Peng WB, Zhang P, Yang XP, Zhou Q. **Lactate in the tumour microenvironment: From immune modulation to therapy**. *EBioMedicine* (2021) **73**. DOI: 10.1016/j.ebiom.2021.103627
16. Wen Y, Zhou X, Lu M, He M, Tian Y, Liu L. **Bclaf1 promotes angiogenesis by regulating HIF-1α transcription in hepatocellular carcinoma**. *Oncogene* (2019) **38**. DOI: 10.1038/s41388-018-0552-1
17. Moldogazieva NT, Mokhosoev IM, Terentiev AA. **Metabolic heterogeneity of cancer cells: An interplay between HIF-1, GLUTs, and AMPK**. *Cancers (Basel)* (2020) **12** 862. DOI: 10.3390/cancers12040862
18. Greten TF, Eggert T. **Cellular senescence associated immune responses in liver cancer**. *Hepat Oncol* (2017) **4**. DOI: 10.2217/hep-2017-0011
19. Rayess H, Wang MB, Srivatsan ES. **Cellular senescence and tumor suppressor gene p16**. *Int J Cancer* (2012) **130**. DOI: 10.1002/ijc.27316
20. Zhang T, Gu HW, Gao JX, Li YS, Tang HB. **Ethanol supernatant extracts of gynura procumbens could treat nanodiethylnitrosamine-induced mouse liver cancer by interfering with inflammatory factors for the tumor microenvironment**. *J Ethnopharmacol* (2022) **285**. DOI: 10.1016/j.jep.2021.114917
21. Guo X, Chen F, Gao F, Li L, Liu K, You L. **CNSA: a data repository for archiving omics data**. *Database (oxford)* (2020) **2020** baaa055. DOI: 10.1093/database/baaa055
22. Chen FZ, You LJ, Yang F, Wang LN, Guo XQ, Gao F. **CNGBdb: China national GeneBank DataBase**. *Yi Chuan* (2020) **42** 799-809. DOI: 10.16288/j.yczz.20-080
23. Matsumoto M, Hada N, Sakamaki Y, Uno A, Shiga T, Tanaka C. **An improved mouse model that rapidly develops fibrosis in non-alcoholic steatohepatitis**. *Int J Exp Pathol* (2013) **94** 93-103. DOI: 10.1111/iep.12008
24. Qin QF, Li XJ, Li YS, Zhang WK, Tian GH, Shang HC. **AMPK-ERK/CARM1 signaling pathways affect autophagy of hepatic cells in samples of liver cancer patients**. *Front Oncol* (2019) **9**. DOI: 10.3389/fonc.2019.01247
25. Li YS, Leng CL, Chen MT, Zhang WK, Li XJ, Tang HB. **Mouse hepatic neoplasm formation induced by trace level and low frequency exposure to diethylnitrosamine through β-catenin signaling pathway**. *Toxicol Res (Camb)* (2016) **5**. DOI: 10.1039/c5tx00317b
26. Marengo A, Rosso C, Bugianesi E. **Liver cancer: Connections with obesity, fatty liver, and cirrhosis**. *Annu Rev Med* (2016) **67**. DOI: 10.1146/annurev-med-090514-013832
27. Fang C, Pan J, Qu N, Lei Y, Han J, Zhang J. **The AMPK pathway in fatty liver disease**. *Front Physiol* (2022) **13**. DOI: 10.3389/fphys.2022.970292
28. Wang H, Mehal W, Nagy LE, Rotman Y. **Immunological mechanisms and therapeutic targets of fatty liver diseases**. *Cell Mol Immunol* (2021) **18** 73-91. DOI: 10.1038/s41423-020-00579-3
29. Wang Q, Liu S, Zhai A, Zhang B, Tian G. **AMPK-mediated regulation of lipid metabolism by phosphorylation**. *Biol Pharm Bull* (2018) **41**. DOI: 10.1248/bpb.b17-00724
30. Laderoute KR, Amin K, Calaoagan JM, Knapp M, Le T, Orduna J. **5'-AMP-activated protein kinase (AMPK) is induced by low-oxygen and glucose deprivation conditions found in solid-tumor microenvironments**. *Mol Cell Biol* (2006) **26**. DOI: 10.1128/MCB.00166-06
31. Kottakis F, Bardeesy N. **LKB1-AMPK axis revisited**. *Cell Res* (2012) **22**. DOI: 10.1038/cr.2012.108
32. Cataldo I, Sarcognato S, Sacchi D, Cacciatore M, Baciorri F, Mangia A. **Pathology of non-alcoholic fatty liver disease**. *Pathologica* (2021) **113** 194-202. DOI: 10.32074/1591-951X-242
33. Smith BK, Marcinko K, Desjardins EM, Lally JS, Ford RJ, Steinberg GR. **Treatment of nonalcoholic fatty liver disease: role of AMPK**. *Am J Physiol Endocrinol Metab* (2016) **311**. DOI: 10.1152/ajpendo.00225.2016
34. Jung TY, Ryu JE, Jang MM, Lee SY, Jin GR, Kim CW. **Naa20, the catalytic subunit of NatB complex, contributes to hepatocellular carcinoma by regulating the LKB1-AMPK-mTOR axis**. *Exp Mol Med* (2020) **52**. DOI: 10.1038/s12276-020-00525-3
35. Ciccarese F, Zulato E, Indraccolo S. **LKB1/AMPK pathway and drug response in cancer: A therapeutic perspective**. *Oxid Med Cell Longev* (2019) **2019**. DOI: 10.1155/2019/8730816
36. Delgado TC, Lopitz-Otsoa F, Martínez-Chantar ML. **Post-translational modifiers of liver kinase B1/serine/threonine kinase 11 in hepatocellular carcinoma**. *J Hepatocell Carcinoma* (2019) **6** 85-91. DOI: 10.2147/JHC.S169585
37. Matter MS, Decaens T, Andersen JB, Thorgeirsson SS. **Targeting the mTOR pathway in hepatocellular carcinoma: current state and future trends**. *J Hepatol* (2014) **60**. DOI: 10.1016/j.jhep.2013.11.031
38. Shaw RJ, Kosmatka M, Bardeesy N, Hurley RL, Witters LA, DePinho RA. **The tumor suppressor LKB1 kinase directly activates AMP-activated kinase and regulates apoptosis in response to energy stress**. *Proc Natl Acad Sci U.S.A.* (2004) **101**. DOI: 10.1073/pnas.0308061100
39. Xu C, Li YM, Sun B, Zhong FJ, Yang LY. **ATE1 inhibits liver cancer progression through RGS5-mediated suppression of wnt/β-catenin signaling**. *Mol Cancer Res* (2021) **19**. DOI: 10.1158/1541-7786.MCR-21-0027
40. Chen XZ, Zhang WK, Tang HB, Li XJ, Tian GH, Shang HC. **The ethanol supernatant extracts of liushenwan could alleviate nanodiethylnitrosamine-induced liver cancer in mice**. *Can J Gastroenterol Hepatol* (2018) **2018**. DOI: 10.1155/2018/6934809
41. Li XJ, Huang FZ, Wan Y, Li YS, Zhang WK, Xi Y. **Lipopolysaccharide stimulated the migration of NIH3T3 cells through a positive feedback between β-catenin and COX-2**. *Front Pharmacol* (2018) **9**. DOI: 10.3389/fphar.2018.01487
42. Park SY, Lee YK, Kim HJ, Park OJ, Kim YM. **AMPK interacts with β-catenin in the regulation of hepatocellular carcinoma cell proliferation and survival with selenium treatment**. *Oncol Rep* (2016) **35**. DOI: 10.3892/or.2015.4519
43. Park SY, Kim D, Kee SH. **Metformin-activated AMPK regulates β-catenin to reduce cell proliferation in colon carcinoma RKO cells**. *Oncol Lett* (2019) **17**. DOI: 10.3892/ol.2019.9892
44. Serra S, Chetty R. **p16**. *J Clin Pathol* (2018) **71**. DOI: 10.1136/jclinpath-2018-205216
45. LaPak KM, Burd CE. **The molecular balancing act of p16(INK4a) in cancer and aging**. *Mol Cancer Res* (2014) **12**. DOI: 10.1158/1541-7786.MCR-13-0350
46. Wang MJ, Chen JJ, Song SH, Su J, Zhao LH, Liu QG. **Inhibition of SIRT1 limits self-renewal and oncogenesis by inducing senescence of liver cancer stem cells**. *J Hepatocell Carcinoma* (2021) **8**. DOI: 10.2147/JHC.S296234
47. Liang X, Wang P, Gao Q, Xiang T, Tao X. **Endogenous LKB1 knockdown accelerates G(1)/S transition through p53 and p16 pathways**. *Cancer Biol Ther* (2010) **9**. DOI: 10.4161/cbt.9.2.10452
48. Liang X, Wang P, Gao Q, Tao X. **Exogenous activation of LKB1/AMPK signaling induces G**. *Mol Med Rep* (2014) **9**. DOI: 10.3892/mmr.2014.1916
49. Chen Y, Liu L, Xia L, Wu N, Wang Y, Li H. **TRPM7 silencing modulates glucose metabolic reprogramming to inhibit the growth of ovarian cancer by enhancing AMPK activation to promote HIF-1α degradation**. *J Exp Clin Cancer Res* (2022) **41** 44. DOI: 10.1186/s13046-022-02252-1
50. Seo J, Jeong DW, Park JW, Lee KW, Fukuda J, Chun YS. **Fatty-acid-induced FABP5/HIF-1 reprograms lipid metabolism and enhances the proliferation of liver cancer cells**. *Commun Biol* (2020) **3** 638. DOI: 10.1038/s42003-020-01367-5
51. Faubert B, Boily G, Izreig S, Griss T, Samborska B, Dong Z. **AMPK is a negative regulator of the warburg effect and suppresses tumor growth**. *Cell Metab* (2013) **17**. DOI: 10.1016/j.cmet.2012.12.001
|
---
title: Prenylated indole-terpenoids with antidiabetic activities from Penicillium
sp. HFF16 from the rhizosphere soil of Cynanchum bungei Decne
authors:
- Xijin Liu
- Fandong Kong
- Na Xiao
- Xiaoyu Li
- Mingyu Zhang
- Fujin Lv
- Xiaolin Liu
- Xiangchuan Kong
- Jing Bi
- Xinyi Lu
- Daqing Kong
- Gangping Hao
- Liman Zhou
- Guojun Pan
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC10018213
doi: 10.3389/fmicb.2023.1099103
license: CC BY 4.0
---
# Prenylated indole-terpenoids with antidiabetic activities from Penicillium sp. HFF16 from the rhizosphere soil of Cynanchum bungei Decne
## Abstract
Finding novel and effective suppression of hepatic glucagon response antidiabetic compounds is urgently required for the development of new drugs against diabetes. Fungi are well known for their ability to produce new bioactive secondary metabolites. In this study, four new prenylated indole-terpenoids [1-4], named encindolenes I-L, as well as a known analogue [5], were isolated from the fungus Penicillium sp. HFF16from the rhizosphere soil of Cynanchum bungei Decne. The structures of the compounds were elucidated by spectroscopic data and ECD analysis. In the antidiabetic activity assay, compounds 1-5 could inhibit glucagon-induced hepatic glucose output with EC50 values of 67.23, 102.1, 49.46, 25.20, and 35.96 μM, respectively, and decrease the intracellular cAMP contents in primary hepatocytes.
## Introduction
The liver plays a major role in whole body glucose metabolism by maintaining a balance between glucose production and glucose storage (Lewis et al., 2021; Zhang et al., 2022). Excessive hepatic glucose production contributes substantially to diabetes, and it is proposed that suppression of hepatic glucose production may provide therapeutic advantages for the control of diabetes (Xiao et al., 2017; Liao et al., 2021). During fasting, hepatic gluconeogenesis is the primary source of glucagon-mediated endogenous glucose production (Unger and Cherrington, 2012). Glucagon, a pancreas-derived hormone induced by fasting, promotes gluconeogenesis through induction of intracellular cAMP production. Glucagon promotes hepatic gluconeogenesis through upregulation of cAMP/PKA signaling pathway and prevents hypoglycemia (Zhang et al., 2019). Therefore, finding novel and effective inhibition of glucagon-mediated gluconeogenesis bioactive compounds are urgently required. Fungal secondary metabolites have been proven to be an important source of natural compounds with novel structures and unique activities, many of which contribute to drug discovery and are approved by the US Food and Drug Administration (Pan G. J. et al., 2021; Shankar and Sharma, 2022). The paxilline-type indole-diterpenoids are one of the largest classes of fungal indole-terpenoids (Kong et al., 2019), many of which have significant bioactivities. In our preliminary search for bioactive metabolites from Penicillium sp. HFF16, from the rhizosphere soil of Cynanchum bungei Decne from Mount Tai, China, nine new indole-terpenoids with weak anti-inflammatory activities and antidiabetic effects were investigated (Pan G. et al., 2021; Xiao et al., 2022). Considering such a significant work, Penicilliumsp. HFF16 was re-fermented and chemical investigation on its extracts revealed another four new indole-terpenoids [1-4] (Figure 1). All of the compounds exhibited moderate antidiabetic effects on glucagon-stimulated cAMP accumulation and hepatic glucose output in primary hepatocytes. Herein, the isolation, structural elucidation, and bioactivities of these compounds were described.
**Figure 1:** *The chemical structures of compounds 1-5.*
## General experimental procedures
Optical rotations were measured on an Anton PaarMCP-100 digital polarimeter, and UV spectra were measured on a Beckman DU 640 spectrophotometer. ECD data were collected using a JASCO J-715 spectropolarimeter. NMR spectra were recorded on a Bruckmercury Plus-400 spectrometers with TMS as an internal standard. HRESIMS spectra were recorded with a Micromass Autospec-Uitima-TOF. Infrared (IR) spectra were obtained on a FTIR-650 spectrometer. Semi-preparative high-performance liquid chromatography (HPLC) was carried out using an ODS column (YMC-pack ODS-A, 10 × 250 mm, 5 μm, 4 mL/min). Thin-layer chromatography (TLC) and column chromatography (CC) were performed on plates precoated with silica gel GF254 (10–40 μm, Yantai Jiangyou Silicone Development Co. Ltd).
## Fungal material and fermentation
The fungus Penicillium sp. HFF16 was isolated from the rhizosphere soil of Cynanchum bungei Decne, in Mount Tai, China in May 2020 and identified according to its morphological characteristics and ITS gene sequences (Pan G. et al., 2021). A reference culture of Penicillium sp. HFF16 maintained at -80°C is deposited in our laboratory. The isolate was cultured on the plates of PDA medium at 28°C for 4 days. Plugs of agar supporting mycelium growth were cut and transferred aseptically to the 10 × 250-mL Erlenmeyer flasks each containing 100 mL of liquid medium (potato 200 g, glucose 20 g per liter of tap water) and cultured at 28°C at 150 RPM for 3 days. The seed liquid was inoculated aseptically into the 200 × 1,000-mL Erlenmeyer flasks each containing rice medium (80 g rice, 100 mL of tap water) at 0.5–$1\%$ inoculation amount and incubated at room temperature under static conditions for 35 days.
## Extraction and isolation
The cultures (16 kg) were then extracted into 50 L of ethyl acetate (EtOAc) by soaking overnight. The extraction repeated for three times. The combined EtOAc extracts were dried under vacuum to produce 52.1 g of extract. The EtOAc extract was subjected to a silica gel column-vacuum liquid chromatography column, eluting with a stepwise gradient of 0, 9, 11, 15, 20, 30, 50, and $100\%$ EtOAc in petroleum ether (v/v), to give seven fractions (Fr. 1-7). Fraction 2 (17.2 g) was applied to ODS silica gel with gradient elution of CH3OH (MeOH)-H2O (1:5, 2:3, 3:2, 4:1, 1:0) to yield four subfractions (Fr. 2-1-Fr. 2-4). Fr. 2-1 (5.0 g) was applied to ODS silica gel with gradient elution of MeCN-H2O (1,4, 2:3, 3:2, 7:3, 4:1, 9:1, 9.5:1, and 0) to yield six tertiary fractions (Fr. 2-1-1-Fr. 2-1-6). Fr. 2-1-3 (0.81 g) was purified using semi-prep HPLC (isocratic system $93.6\%$ MeOH/H2O, v/v) to yield nine fourthiary fractions (Fr. 2-1-3-1-Fr. 2-1-3-9). Fr. 2-1-3-8 (49 mg) was purified using semi-prep HPLC (isocratic system $90\%$ MeCN/H2O, v/v) to give compounds 1 (tR10.1 min; 10 mg), 4 (tR6.1 min; 11 mg), and 5 (tR 8.1 min; 13 mg). Fr. 2-1-3-7 (62 mg) was purified using semi-prep HPLC (isocratic system $90\%$ MeCN/H2O, v/v) to give compounds 3 (tR5.8 min; 9 mg) and 2 (tR9.9 min; 11 mg).
Encindolene I [1]: white powder; [α]25 D-10 (c 0.1, MeOH); UV (MeOH) λmax (log ε): 288 (2.96), 233 (3.47) nm;IR (KBr) νmax: 3391, 2,961, 2,923, 1,667, 1,453, 1,260, 1,024 cm−1; ECD (MeOH) λmax 205 (−13.71), 229 (+4.43), 260 (+4.57), 325 (+6.60) nm. 1H and 13C NMR data, Tables 1, 2; HRESIMS m/z542.2862 [M + Na]+ (calcd for C32H41NO5Na, 542.2877).
Encindolene J [2]: white powder; [α]25 D-11 (c 0.1, MeOH); UV (MeOH) λmax (log ε): 283 (3.06), 231 (3.56) nm;IR (KBr) νmax: 3453, 2,953, 2,921, 1,670, 1,453, 1,375, 1,172 cm−1; ECD (MeOH) λmax 207 (−26.39), 226 (+5.56), 248 (+10.57), 302 (+2.26) nm. 1H and 13C NMR data, Tables 1, 2; HRESIMS m/z542.2864 [M + Na]+ (calcd for C32H41NO5Na, 542.2877).
Encindolene K [3]: white powder; [α]25 D + 354 (c 0.1, MeOH); UV (MeOH) λmax (log ε): 283 (3.06), 231 (3.50) nm;IR (KBr) νmax: 3427, 2,930, 1,660, 1,373, 1,298, 1,180 cm−1; ECD (MeOH) λmax 204 (−12.42), 223 (+4.43), 239 (−12.73), 325 (+18.33) nm. 1H and 13C NMR data, Tables 1, 2; HRESIMS m/z502.2947 [M + H]+ (calcd for C32H40NO4, 502.2952).
Encindolene L [4]: white powder; [α]25 D + 52 (c 0.1, MeOH); UV (MeOH) λmax (log ε): 288 (2.98), 233 (3.47) nm;IR (KBr) νmax: 3428, 2,929, 2,923, 1,660, 1,451, 1,297, 1,179 cm−1; ECD (MeOH) λmax 207 (−8.98), 225 (+6.60), 242 (−10.71), 325 (+16.67) nm. 1H and 13C NMR data, Tables 1, 2; HRESIMS m/z502.2950 [M + H]+ (calcd for C32H40NO4, 502.2952).
## Preparation of primary hepatocytes and cell viability assay
Primary hepatocytes were isolated from male C57BL/6 J mice (Jinan Pengyue Experimental Animal Breeding Co. Ltd) by an improved two-step collagenase infusion (Xiao et al., 2017). All experiments and animal care conducted in accordance with the Provision and General Recommendation of Chinese Experimental Animals Administration Legislation and were approved by the Animal Ethics Committee of Shandong Agriculture University. Primary mouse hepatocytes were maintained in DMEM medium with $10\%$ fetal bovine serum (FBS). After attachment, the cells incubated with 100 nM glucagon, as well as the tested compounds. After 24 h, MTT solution was added and incubated for 4 h. The purple crystals were dissolved with dimethylsulfoxide (DMSO) and the absorbance value was determined at 570 nm.
## Hepatic glucose production and intracellular cAMP measurement
Primary hepatocytes on 48-well plates were maintained in DMEM ($10\%$ FBS) medium. After attachment, the media was replaced with Krebs-Ringer HEPES buffer to fast the cells for 2 h. Then, the cells were cultured with glucose out media supplemented with 10 mM pyruvate, 100 nM glucagon, or with metformin (1 mM) and the tested compounds (1, 20, 40, 80, and 160 μm). After 6 h, the cell supernatant was collected for glucose analysis. For intracellular cAMP measurement, primary hepatocytes were treated with the tested compounds in the presence or absence of 100 nM glucagon for 4 h. cAMP was calculated in primary hepatocytes with an ELISA kit. All data were expressed as the mean ± SD from at least three independent experiments.
## Structure elucidation of compounds
Compound 1 was assigned the molecular formula C32H41NO5 by HRESIMS, with 13 degrees of double-bond equivalents. The 13C and HSQC NMR spectra (Table 1) of 1 revealed a total of 32 carbons including eight aromatic carbons (three protonated) corresponding to one indole moiety, four olefinic carbons attributed to two double bonds, four oxygenated sp3 carbons with one protonated, two sp3 quaternary carbons, six sp3 methylenes, one sp3 non-protonated methine, and six methyls. The presence of a prenyl group was demonstrated by HMBC correlations from the two methyls H3-4′ and H3-5′ (δH 1.74 and 1.75) to one olefinic quaternary carbon (δC 125.4) and one olefinic methine (δC 131.1) and COSY correlation between the olefinic proton H-2′ (δH 5.38) and the methylene protons H2-1′ (δH 3.37). The above data were quite similar to those of the known compound 3-methyl-2-butenylpaspaline [5] (Cole et al., 1977), with the main differences being the chemical shifts for the two oxygenated carbons C-18 and C-22, which were δC 98.4 and 78.5 for 1 while 104.4 and 88.0 for 5 (Cole et al., 1977). These data, as well as the less of a H2O in the molecule formula compared to that of 5 deduced from the HRESIMS data, suggested that the connection between C-18 and C-22 in 5 was cleaved through hydrolysis to afford 1. The HMBC and COSY data (Figure 2) further confirmed this deduction. The relative configuration of 1 was assigned by the analysis of its ROESY spectrum (Figure 3). ROESY correlations of H-11/H3-26/Hβ-13 suggested the same orientation of these protons and the trans-diaxial relationship of H3-26 and OH-14, while correlation of H3-27/Hα-16 indicated that these protons located at the face opposite to H3-26. ROESY correlations of H3-26/Hβ-17/H3-25 suggested the same face of these protons, indicating the α orientation of OH-18 and H-22. The experimental ECD spectrum (Figure 4) of 1 showed negative Cotton effects (CEs) around 206, 239, and 373 nm, and positive ones around 228, 220, and 326 nm, respectively (Figure 4), which were very similar to those for encindolenes D and E (Pan G. et al., 2021), two analogs isolated from the same fungus. This led to the assignment of the absolute configurations of 1 as shown in Figure 1.
**Figure 2:** *Selected HMBC and COSY correlations of compounds 1-4.* **Figure 3:** *Selected ROESY correlations of compounds 1-4.* **Figure 4:** *The experimental ECD spectra of compounds 1-4.*
Compound 2 was obtained as a white powder, and its molecular formula was determined to be the same as that of 1 according to the HRESIMS data, with a molecule of H2O less than 1. The NMR data of 2 were also quite similar to those of 1. The main differences between the 1H NMR spectra of them were that signals attributed to a 1,2,3-trisubstituted benzene ring in 2 replaced those corresponding to a 1,2,6-trisubstituted benzene ring in 1, indicating the location of the prenyl at C-2 or C-5 in 2. HMBC correlations from the methylene protons of the prenyl group H2-1′ to C-1, C-2, and C-3 in the indole group and COSY correlations of H-3/H-4/H-5 further confirmed this deduction. Their remaining substructures were determined to be identical according to the 2D NMR data. The relative configuration of 2 was deduced to be the same as that of 1 based on their similar NMR chemical shifts. ROESY correlations (Figure 3) of H-11/H3-26/Hβ-17/H-24 [25] suggested the face of these protons. ROESY correlation (Figure 3) of H3-27/Hα-16 indicated that these protons located at the face opposite to H3-26. These data further confirming the above deduction. The absolute configurations of 2 were also assigned as shown in Figure 1 by a comparison of its ECD spectrum with that of 1 (Figure 4), which showed great similarity.
The molecular formula of compound 3 was established as C32H39NO4 by HRESIMS, with one H2O less compared to 1 and 2. The NMR spectra of 3 were closely related to those of 2, indicating that 3 was also a prenylated indole-diterpenoid. A comparison of the NMR data between 2 and 3 revealed the absence of the dioxygenated non-protonated carbon C-18 and a methylene and the presence of an additional trisubstituted double bond in 2 compared to 3. COSY correlations of H2-16 with the olefinic proton H-17 and HMBC correlation from H2-16 to C-17 and C-18 suggested that dehydration occurred at C-17/C-18 in 2 to afford compound 3. The relative configuration of 3 was proposed to be the same as that of 2 based on a biosynthetic consideration, which was further confirmed by NOESY correlations of H-11/H3-26/Hβ-13 and H3-27/Hα-16. The ECD spectra of 3 were quite similar to those of 1 and 2 (Figure 4), thus assigning their same absolute configurations for the chiral carbons C-10, C-11, C-14, C-15, C-22, and C-23.
The molecular formula of compound 4 was established to be the same as that of 3 by HRESIMS. Their NMR data were also quite similar. A comparison of the NMR data of 4 with those of 1 revealed that they bear the same 3-prenylated indole moiety. The remaining NMR data of 4 were nearly identical to those of 3. The above data led to the determination of the structure of 4, and the only difference between it and 3 was the location of the prenyl group, which was C-2 in 4. HMBC correlations from H2-1′ to C-2, C-3, and C-4, as well as COSY correlations of H-4/H-5, further confirmed this deduction. The relative configuration of 4 was deduced to be the same as that of 3 by their similar NMR data (Tables 1, 2). ROESY correlations of H-11/H3-26/Hβ-13 and H3-27/Hα-16 further confirmed this deduction. The absolute configuration of 4 was also assigned to be the same as that of 3 by their similar ECD curves (Figure 4).
Compounds 1-5 are structurally closely related. Compounds 1 and 4 could be the dehydration products of 5, while 2 and 3 could be the dehydration products of another known compound paspalitrem C. Therefore, it is necessary to define whether these compounds were artificial products due to acidic dehydration during the purification process and find the reaction conditions of mutual transformation between them to lay a foundation for the accumulation of these compounds. The experiment of mutual transformation between these compounds was performed. The results indicated that compound 5 could be converted to compounds 1 and 4 in $0.1\%$ trifluoroacetic acid in methanol, and paspalitrem C, a previously isolated analog from the same fungus, can be converted to compounds 2 and 3 under the same conditions (Scheme 1). However, after the treatment of silica gel and C18, there is no structural transformation, but in other strong acids such as hydrochloric acid and sulfuric acid, the compound is basically degraded without effective results. These results indicated that the production of compounds 1-4 is probably the result of enzyme catalysis, but it cannot exclude the possibility that compounds 1-4 may be artificial products from 5 and paspalitrem C due to the slightly acidic growth environment in the late stage of cultivation. These results also suggested that the absolute configurations of all the chiral carbon except for C-18 in 1, 2, and 5 were the same.
**SCHEME 1:** *The transformation of 5 to 1 and 4 and paspalitrem C to 2 and 3 in acidic reaction condition.*
Until now, a total of 17 indole-terpenoids including compounds 1-5 and other twelve previously reported analogs such as encindolenes A-C, 18-O-methyl-encindolene A (Pan G. et al., 2021), encindolenes D-H (Xiao et al., 2022), paspalitrem C (Dorner et al., 1984), 7-methoxypaxilline (Ariantari et al., 2019), and 7-hydroxy-13-dehydroxypaxilline (Peter and Christopher, 1994) have been identified from Penicillium sp. HFF16, and the plausible biosynthetic pathway of the eight different skeletons was shown in Scheme 2. It was proposed that 7-hydroxy-13-dehydroxypaxilline was the main precursor of all the paxilline-type indole-terpenoids, which could undergo prenylation, dehydration, methoxylation, and cyclization occurred to afford the other analogs.
**SCHEME 2:** *Hypothetical biosynthetic pathway for the skeletons of the indole-terpenoids isolated from Penicillium sp. HFF16.*
## Antidiabetic activity assay
All the compounds were evaluated for cell viability at a concentration of 160 μM, and with this result, all the tested compounds (cell viability >$90\%$) were selected for subsequent glucose output inhibition experiment (Figure 5). Glucose output in response to all the nontoxic compounds was measured to assess the antidiabetic effects in hepatocytes. Glucagon promotes hepatic glycogenolysis and increases hepatic gluconeogenesis, and we showed that glucagon challenge increased hepatic glucose output. Compounds 1-5 inhibited hepatic glucose output and their EC50 values (67.23, 102.1, 49.46, 25.20, and 35.96 μM) were higher than that of the positive control metformin (EC50 = 5.09 μM). Cyclic AMP (cAMP) as an intracellular second messenger is crucial for glucagon-induced hepatic glucose production. Glucagon challenge increased intracellular cAMP content, while compounds 1-5 treatment suppressed cAMP accumulation in hepatocytes. The results suggested that tested compounds inhibited hepatic glucose output may be by suppression hepatic cAMP accumulation.
**Figure 5:** *Viability and antidiabetic effects of compounds 1-5 against primary hepatocytes. (A) Cell viability (B) hepatic glucose output level; (C) cAMP contents in primary hepatocytes treated with glucagon (100 nM). Metformin (1 mM) as the positive control. nsp > 0.05 vs. Blank, *p < 0.05 vs. Glucagon, #p < 0.05 vs. Blank.*
## Conclusion
In our previous study, nine new indole-diterpenoids were isolated from the secondary metabolites of Penicillium sp. HFF16 and evaluated their anti-inflammatory and hypoglycemic activities. The results showed that encindolene C had the best anti-inflammatory activity compared with other compounds in RAW.2647 cells stimulated by LPS. Through simple structural analysis, it was speculated that the existence of prenyl group was beneficial to the improvement of anti-inflammatory activity. In HepG2 cells stimulated by glucagon, encindolene L showed the best inhibitory activity on hepatic glycogen export compared with other compounds. Structural analysis showed that encindolene L also had a prenyl group and indole and diterpene did not form a fused ring structure, again suggesting the importance of prenyl group in improving the biological activity of compounds. In consideration of such valuable work and the fact that the compound belongs to the tryptophan pathway of biosynthesis, the strain is subjected to secondary fermentation after a small amount of tryptophan is added to a rice culture medium. Four new indole-diterpenoid encindolenes I-L containing prenyl moieties were isolated and identified. Hypoglycemic activity was evaluated by mouse primary hepatocytes, and the results showed that encindolenes I-L could inhibit the increase of cAMP production induced by glucagon and reduce hepatic glucose output, thus exerting hypoglycemic effect. From the structural analysis, it was found that the compounds containing semi-acetal group had the worst hypoglycemic activity and the dehydration compound had the best activity, suggesting that semi-acetal group was harmful to biological activity. The comparison of the structure and activity results of encindolenes I and J suggests that the different substitution positions of the prenyl group have a significant effect on the activity.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The animal study was reviewed and approved by All experiments and animal care conducted in accordance with the Provision and General Recommendation of Chinese Experimental Animals Administration Legislation and were approved by the Animal Ethics Committee of Shandong Agriculture University.
## Author contributions
NX and LZ contributed to bioactivity assay and revised the manuscript. GP conceived and designed the experiments and was involved in isolation of compounds. XjL, XyL, MZ, FL, and XlL contributed to isolation and collection of the NMR data of compounds. XK, JB, XyL, DK, and GH performed strain fermentation and extraction. FK supervised the study and prepared the manuscript. All authors contributed to the article and approved the submitted version.
## Funding
This study was financially supported by the Natural Science Foundation of Shandong Province (ZR2021MB087), the National Natural Science Foundation of China [82004014], the Shandong traditional Chinese Medicine Science and Technology Project (2021Q083), the Innovation and entrepreneurship training program for college students in Shandong Province [202210439005, 202214039008], the Specific research project of Guangxi for research bases and talents (AD18126005), and the Natural Science Foundation of Guangxi (2021GXNSFBA075036).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2023.1099103/full#supplementary-material
## References
1. Ariantari N. P., Ancheeva E., Wang C., Mándi A., Knedel T. O., Kurtán T.. **Indole Diterpenoids from an**. *J. Nat. Prod.* (2019) **82** 1412-1423. DOI: 10.1021/acs.jnatprod.8b00723
2. Cole R. J., Dorner J. W., Lansden J. A., Cox R. H., Pape C., Cunfer B.. **Paspalum staggers: isolation and identification of tremorgenic metabolites from sclerotia of**. *J. Agric. Food Chem.* (1977) **25** 1197-1201. DOI: 10.1021/jf60213a061
3. Dorner J. W., Cole R. J., Cox R. H., Cunfer B. M.. **Paspalitrem C, a new metabolite from sclerotia of**. *J. Agric. Food Chem.* (1984) **32** 1069-1071. DOI: 10.1021/jf00125a033
4. Kong F. D., Fan P., Zhou L. M., Ma Q. Y., Xie Q. Y., Zheng H. Z.. **Penerpenes A-D, four indole terpenoids with potent protein tyrosine phosphatase inhibitory activity from the marine-derived fungus**. *Org. Lett.* (2019) **21** 4864-4867. DOI: 10.1021/acs.orglett.9b01751
5. Lewis G. F., Carpentier A. C., Pereira S., Hahn M., Giacca A.. **Direct and indirect control of hepatic glucose production by insulin**. *Cell Metab.* (2021) **33** 709-720. DOI: 10.1016/j.cmet.2021.03.007
6. Liao W., Yang W., Shen Z., Ai W., Pan Q., Sun Y.. **Heme Oxygenase-1 regulates ferrous iron and Foxo1 in control of hepatic gluconeogenesis**. *Diabetes* (2021) **70** 696-709. DOI: 10.2337/db20-0954
7. Pan G. J., Li Y. L., Che X. Y., Tian D., Han W. J., Wang Z. M.. **New Thio-compounds and Monoterpenes with anti-inflammatory activities from the fungus**. *Front. Microbiol.* (2021) **12** 668938. DOI: 10.3389/fmicb.2021.668938
8. Pan G., Zhao Y., Ren S., Liu F., Xu Q., Pan W.. **Indole-Terpenoids with anti-inflammatory activities from**. *Front. Microbiol.* (2021) **12** 710364. DOI: 10.3389/fmicb.2021.710364
9. Peter G. M., Christopher M. W.. **Biosynthesis and transformation of tremorgenic indolediterpenoids by**. *Phytochemistry* (1994) **36** 1209-1217. DOI: 10.1016/S0031-9422(00)89639-9
10. Shankar A., Sharma K. K.. **Fungal secondary metabolites in food and pharmaceuticals in the era of multi-omics**. *Appl. Microbiol. Biotechnol.* (2022) **106** 3465-3488. DOI: 10.1007/s00253-022-11945-8
11. Unger R. H., Cherrington A. D.. **Glucagonocentric restructuring of diabetes: a pathophysiologic and therapeutic makeover**. *J. Clin. Invest.* (2012) **122** 4-12. DOI: 10.1172/JCI60016
12. Xiao N., Lou M. D., Lu Y. T., Yang L. L., Liu Q., Liu B.. **Ginsenoside Rg5 attenuates hepatic glucagon response via suppression of succinate-associated HIF-1α induction in HFD-fed mice**. *Diabetologia* (2017) **60** 1084-1093. DOI: 10.1007/s00125-017-4238-y
13. Xiao N., Xu Y., Zhang X., Li H., Zhang S., Xiao A.. **Anti-diabetic Indole-Terpenoids from**. *Front. Chem.* (2022) **9** 792810. DOI: 10.3389/fchem.2021.792810
14. Zhang R., Liu W., Zeng J., Meng J., Jiang H., Wang J.. **Niemann-pick C1-like 1 inhibitors for reducing cholesterol absorption**. *Eur. J. Med. Chem.* (2022) **230** 114111. DOI: 10.1016/j.ejmech.2022.114111
15. Zhang W. S., Pan A., Zhang X., Ying A., Ma G., Liu B. L.. **Inactivation of NF-κB2 (p52) restrains hepatic glucagon response via preserving PDE4B induction**. *Nat. Commun.* (2019) **10** 4303. DOI: 10.1038/s41467-019-12351-x
|
---
title: Association of maternal body composition and diet on breast milk hormones and
neonatal growth during the first month of lactation
authors:
- David Ramiro-Cortijo
- Pratibha Singh
- Gloria Herranz Carrillo
- Andrea Gila-Díaz
- María A. Martín-Cabrejas
- Camilia R. Martin
- Silvia M. Arribas
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC10018215
doi: 10.3389/fendo.2023.1090499
license: CC BY 4.0
---
# Association of maternal body composition and diet on breast milk hormones and neonatal growth during the first month of lactation
## Abstract
### Introduction
Preterm birth is associated with altered growth patterns and an increased risk of cardiometabolic diseases, with breast milk (BM) being a counteracting factor. Preterm infants also show alterations in adipokines and gut hormones influencing appetite and metabolism. Since these hormones are present in BM, it is possible that their levels may equilibrate deficiencies improving infant growth. We aimed to assess 1) the BM levels of ghrelin, resistin, leptin, insulin, peptide YY, and the gastrointestinal peptide in women with preterm and term labor; 2) the relationship between BM hormones and neonatal growth; and 3) the influence of maternal body composition and diet on these BM hormones.
### Methods
BM from 48 women (30 term and 18 preterm labor) was collected at days 7, 14, and 28 of lactation. Maternal body composition was evaluated by bioimpedance, and neonate anthropometric parameters were collected from medical records. The maternal dietary pattern was assessed by a 72-h dietary recall at days 7 and 28 of lactation. BM hormones were analyzed by the U-Plex Ultra-sensitive method. Data were analyzed using linear regression models. BM from women with preterm labor had lower ghrelin levels, with the other hormones being significantly higher compared to women with term delivery.
### Results
In premature infants, growth was positively associated with BM ghrelin, while, in term infants, it was positively associated with insulin and negatively with peptide YY. In the first week of lactation, women with preterm labor had higher body fat compared to women with term labor. In this group, ghrelin levels were positively associated with maternal body fat and with fiber and protein intake. In women with term labor, no associations between anthropometric parameters and BM hormones were found, and fiber intake was negatively associated with peptide YY.
### Discussion
Preterm labor is a factor influencing the levels of BM adipokines and gut hormones, with BM ghrelin being a relevant hormone for premature infant growth. Since ghrelin is lower in BM from women with preterm labor and the levels are associated with maternal fat storage and some dietary components, our data support the importance to monitor diet and body composition in women who gave birth prematurely to improve the BM hormonal status.
## Introduction
The estimated global prevalence of preterm labor (pregnancy <37 completed weeks of gestation) has been rising from $9.6\%$ in 2005 to $11\%$ in 2020 [1, 2], mainly in low- and middle-income countries [3]. Preterm infants have an increased risk of developing short-term comorbidities [4], as well as cardiometabolic diseases later in life [5]. Premature birth also results in an altered growth pattern [6], with a slower rate in the short term [7] but with a faster growth later compared to term-born infants, a process called catch-up growth [8]. At birth, premature infants also show alterations in the levels of adipokines and gut hormones that influence appetite and metabolism, such as leptin, adiponectin, and ghrelin, due to the inadequate adipose tissue storage and gastrointestinal maturation, which might be a risk factor for abnormal development and metabolic disorders later in life [9].
Breast milk (BM) is the gold standard for neonatal nutrition, providing not only macronutrients but also bioactive molecules. Metabolism- and appetite-regulating hormones have been identified in BM and may play a role in infant-feeding behavior, growth, and body composition (10–12). Furthermore, their presence in BM has been proposed to reduce the risk of cardiometabolic diseases [13, 14]. The hormones regulating the food intake and energy homeostasis present in BM can be synthesized and excreted from the mammary gland or can be transported from maternal plasma [15]. Insulin is synthetized by the pancreatic beta cells with a key role in glucose homeostasis control [16], with the maternal blood being the main source in BM [12]. Leptin is an adipokine synthesized mainly by white adipose tissue [17], being a long-term anorexigenic hormone regulating body fat storage [18]. Leptin can equally enter in the BM from maternal plasma and by the mammary gland synthesis [19, 20]. Resistin is another adipokine present in BM [21], synthetized mainly by adipocytes and macrophages [22]. The main role of resistin was described as an inflammatory mediator [22] antagonizing insulin actions [23]. In addition to the above-mentioned hormones, some peptides from the gastrointestinal system that contribute to growth and appetite control are also present in BM. Peptide YY is an anorexigenic hormone produced by the gut and small intestine, identified in BM [24, 25], its main source remaining unclear. Ghrelin is a short-term orexigenic hormone, with a role in growth by inducing growth hormone secretion [26, 27]. It is mainly synthesized by the stomach, although the pancreas, kidney, and placenta can produce it in modest amounts [28], being secreted in BM both by the mammary gland and maternal plasma [12, 29]. The gastrointestinal peptide (GIP) is an incretin synthesized by gut after nutrient-induced signaling. In experimental animals, The GIP increases glucagon secretion, reduces appetite and, in the long-term, reduces body weight [30]. BM GIP levels correlate with those in maternal plasma, suggesting that this is the main source [25].
Appetite hormones in BM are important factors not only for the neonatal appetite but also for neonatal growth regulation [24], especially in the first month of life. However, it is not clear if the levels of the appetite hormones in BM can affect growth, particularly in preterm infants. Another relevant aspect is the influence of maternal body composition and diet on BM hormone levels, mostly investigated in the context of maternal obesity and gestational diabetes [31, 32]. Since BM hormone levels can influence infant growth [33], the relationship with the maternal nutritional status is important for counseling breastfeeding women. In this context, we hypothesize that BM appetite–regulating hormones influence neonatal growth differently in term and premature infants. In this study, we aimed to assess in the first month postpartum 1) differences in the BM levels of appetite-related hormones between women with term and preterm labor, 2) the role of these BM hormones on neonatal growth, and 3) the influence of maternal body composition and diet on their BM levels.
## Study design and cohort
In this observational, longitudinal, and non-interventional study, the women were enrolled within the first 72 h postpartum at the Obstetrics and Gynecology and Neonatology Departments of Hospital Clínico San Carlos (HCSC, Madrid, Spain) from September 2019 to March 2020. The sample size was estimated considering that $11\%$ of women would develop premature delivery, indicated by the prevalence of prematurity in 2020 [1, 2] and with an error margin of $5\%$ and a statistical power of $80\%$. The estimated sample size would be 36 women. Considering this as an observational study, increasing the sample size versus the estimated could have benefits in suppressing the potential risk of bias. Maternal inclusion criteria were ≥18 years old, a good understanding of Spanish language, single pregnancy, and the absence of disease at the time of the study. Mothers with dietary restrictions (i.e., diet for competition sports, vegetarians, and vegans) were excluded from the study. The mothers who agreed to participate in the study signed informed consent. The final cohort included 48 women (Figure 1) who provided a BM sample at three time points, if possible (see below).
**Figure 1:** *Flowchart of the women enrollment in the study and sample size (n) with the inclusion and exclusion criteria.*
The present study was performed following the Declaration of Helsinki for studies on human subjects, and it was approved by the Ethical Committee of HCSC (Ref. $\frac{19}{393}$-E).
## Sociodemographic and gestational variables
Close to the enrollment day, the women filled a sociodemographic questionnaire including the maternal age (years), educational level (categorized as illiterate, middle school, high school, and university degree), employment situation (categorized as student, working, and unemployment), Spanish nationality (yes/no), and family size.
The obstetrical and labor clinical data were obtained from medical records and included the number of gestations (gravida) and previous abortions, use of assisted reproduction techniques (yes/no), vitamin D deficiency during this pregnancy (defined as 25-hydroxycholecalciferol blood levels <50 nmol/L in the second trimester), gestational hypothyroidism (defined as Thyroid Stimulating Hormone (TSH) blood levels >2.5 μU/ml in the first trimester and/or >3 μU/ml in the second and third trimesters), gestational anemia (hemoglobin levels <11 g/dl), gestational diabetes (defined as a positive result in the 100 g oral glucose tolerance test), preeclampsia (blood pressure >$\frac{160}{110}$ mmHg with proteinuria or thrombocytopenia after 20 weeks of gestation), use of antibiotic therapy during labor (yes/no), use of completed cycle of antenatal corticosteroids (yes/no), use of magnesium sulfate therapy (yes/no), type of labor (vaginal/C-section), and gestational age (weeks of gestation).
The neonate parameters were recorded from medical records and included the diagnoses of intrauterine growth restriction (defined as fetal growth <3rd percentile or <10th percentile with hemodynamic alterations), sex (male/female), Apgar score at 5 min, and Z-scores weight, length, and head circumference at birth considering to Fenton´s curves [34]. The neonates were categorized as preterm (born <37 weeks of gestation) or term (born ≥37 weeks of gestation).
## Maternal and neonatal anthropometric parameters
Maternal anthropometric parameters were measured at days 7 ± 1 and 28 ± 1 of lactation. Height (cm) was measured with a stadiometer (Seca 217, TAQ Sistemas Médicos, Madrid, Spain), and waist and hip circumferences (cm) were measured with an anthropometric tape with millimeter precision (Seca 201, TAQ Sistemas Médicos, Madrid, Spain). Body weight (kg), total body fat, and muscle mass (%) were assessed using a bioimpedance meter (Omron Healthcare HBF-514C Full Body Sensor W Scale, Madrid, Spain), according to the manufacturer’s instructions. From these parameters, the waist-to-hip index (waist/hip; WHI) and body mass index (BMI in kg/m2) were calculated.
The neonatal weight (g), length (cm), and head circumference (cm) were collected at 7, 14, and 28 days of life from medical records. From these data, the BMI was calculated following Olsen´s curves [35] and expressed in (g/cm2) × 10. Weight growth velocity was calculated as the exponential relationship between weight at birth and weight at the “j” point (Wj) as a function of time, according to Patel´s model [36, 37]. In this study, weight growth velocity was analyzed considering Wj as the discharge weight and expressed as (g/kg/day) using the described formula [38, 39]. The length and head circumference growth velocities were calculated as a linear model and expressed in cm/day, as previously published [38].
## Maternal 72-h dietary intake
Maternal dietary patterns were studied using the 72-h dietary recall (72hDR) questionnaire, a validated method to quantify an individual’s usual intake over a short period through open-ended questions (40–42). The 72hDR was obtained at days 7 ± 1 and 28 ± 1 of lactation. The women indicated the ingredients, food preparation methods, and quantities of everything they ingested during the three previous days, including any supplements and water intake. Women were instructed using the Spanish version of “Visual Guide to Food and Rations” [43] to obtain detailed information. The different foods and their ingredients were recorded manually and classified according to the meals. The composition of nutritional supplements was also included in the software. Thereafter, the analysis was carried out using DIAL software (version 3.11.9, Alce Ingeniería, Madrid, Spain) to calculate specific nutrient intake. The data provided information on energy (Kcal), carbohydrates (carbs.; g), protein (g), fat (g), saturated fatty acids (SFAs; g), monounsaturated fatty acids (MUFAs; g), polyunsaturated fatty acids (PUFAs; g), total cholesterol (mg), total dietary fiber (g), and water intake (ml). In addition, DIAL provided information regarding the intake of the following minerals: calcium (mg); iron (mg); iodine (mg); sodium (mg) and potassium (mg); and vitamins A (retinols, μg), B1 (thiamin; mg), B2 (riboflavin; mg), B3 (niacin; mg), B5 (pantothenic acid; mg), B6 (pyridoxines; mg), B9 (folic acids; μg), B12 (cobalamin; μg), biotin (μg), C (ascorbic acid; mg), D (μg), E (α-tocopherols; mg), and K (μg).
## Breast milk collection and processing to obtain defatted phase
A 1 ml volume of BM was collected at 7 ± 2, 14 ± 2, and 28 ± 2 days of lactation, if available. It was not possible to get colostrum samples for ethical reasons due to the small volume that can be obtained, which are reserved for the neonate. BM was collected by each woman by hand self-expression with an electric breast pump (Symphony® Medela, Barcelona, Spain). To collect the samples, the women washed their hands and cleaned their breast with a gauze with soap and water. BM was collected between 10:00 and 11:59 a.m., from both breasts (always after neonate feeding), pooled and immediately transferred to a glass bottle, and stored in a freezer. The time between extraction and processing took a maximum of 3 h. The sample was centrifuged three times (2,000 rpm for 5 min at 4°C) to obtain the defatted phase, minimizing turbidity. Glass serological pipettes were used to extract the aqueous layer, placed in a clean tube, and stored at −80°C until use. BM samples were analyzed within a month.
## Breast milk hormone detection
A Human U-Plex Ultra-Sensitive Meso Scale Discovery (MSD) Kit (Meso Scale Diagnostics, LCC, Rockville, MD, USA) was used to determine in the BM-defatted phase insulin, leptin, ghrelin, the GIP, and peptide YY following the manufacture protocol. Briefly, each linker vial was conjugated with 200 μl of biotinylated antibody and incubated for 30 min at room temperature. Then, 200 μl of the stop solution was added in each linker vial and incubated for 30 min at room temperature. To prepare a multiplex coating solution, the linker-conjugated antibodies were mixed in a clean tube obtaining a total volume of 6 ml, and 50 μl volume of the multiplex coating solution was added to each well. The plate was covered and incubated shaking for 1 h at room temperature. Thereafter, the plate was washed three times with 150 μl of $0.05\%$ Tween-20 Phosphate-Buffered Saline (PBS) 1X solution (v/v).
The standard curve was prepared according to manufacturer’s guidelines, and the BM-defatted samples were diluted 1.2-fold. A 50 μl volume of the standard curve or the diluted samples was added to the plate in duplicate. Thereafter, the plate was incubated under shaking for 2 h at room temperature. Then, the plate was washed three times with 150 μl of $0.05\%$ Tween-20 PBS 1X solution (v/v). A 50 μl volume of the detection antibody was added to each well, and the plate was incubated under shaking for 1 h at room temperature. Thereafter, the plate was washed three times with 150 μl of $0.05\%$ Tween-20 PBS 1X solution (v/v). Finally, 150 μl of the MSD-Gold read buffer reagent was added in each well. This is a highly sensitive method, in which the range and detection limits of the analyzed hormones were insulin = 0–736 pg/ml (0.32 pg/ml); leptin = 0–47,500 pg/ml (14.0 pg/ml); ghrelin = 0–2,710 pg/ml (1.70 pg/ml); GIP = 0–12,500 pg/ml (3.70 pg/ml); and peptide YY = 0–2,260 pg/ml (2.70 pg/ml).
The plate was read in the MESO QuickPlex SQ 120MM model 1300 system (Meso Scale Diagnostics, LCC, Rockville, MD, USA), and the data were extracted by the MSD discovery workbench analysis software. The hormones were reported in pg/ml, and the natural logarithmic transformation was used for the statistical analysis.
## Breast milk resistin detection
A Single Spot R-Plex Human Assay (Meso Scale Diagnostics, LCC, Rockville, MD, USA) for resistin was also run. Resistin could not be included in the multiplex assay since some of the reagents for this analyte had a cross-competitive reaction with the reagents of the multiplex plate. Considering the manufacturer’s instructions, a 25 μl volume of biotinylated antibody was added to each well, and the plate was covered and incubated under shaking for 1 h at room temperature. Thereafter, the plate was washed three times with 150 μl of $0.05\%$ Tween-20 PBS 1X solution (v/v). The standard curve was prepared according to the manufacturer’s guidelines, and the defatted BM samples were not diluted in this assay. A 25 μl volume of standard curve or BM-defatted samples were added to the plate in duplicate. The plate was incubated under shaking for 1 h at room temperature. Then, the plate was washed three times with 150 μl of $0.05\%$ Tween-20 PBS 1X solution (v/v). A 50 μl volume of detection antibody was added to each well, and the plate was incubated under shaking for 1 h at room temperature. Thereafter, the plate was washed three times with 150 μl of $0.05\%$ Tween-20 PBS 1X solution (v/v). Finally, a 150 μl volume of MSD-Gold read buffer reagent was added to each well. The resistin range was 0–2,500 pg/ml, and the low limit of detection was 0.13 pg/ml.
The plate was read in the MESO QuickPlex SQ 120MM model 1300 system (Meso Scale Diagnostics, LCC, Rockville, MD, USA), and the data were extracted by the MSD discovery workbench analysis software. Resistin was reported in pg/ml, and the natural logarithmic transformation was used for statistical analysis.
## Statistical analysis
Statistical analysis was performed with R software within RStudio interface (version 2022.07.1 + 554, 2022, R Core Team, Vienna; Austria) using rio, dplyr, compareGroups, ggpubr, devtools, stats, nlme, lme4, and ggplot2 packages.
Quantitative variables were expressed as the median and interquartile range [Q1; Q3], and qualitative variables were expressed as the relative frequency and sample size (n). In this study, we did not use methods to impute missing data. The probability (P) to show significance was established at value <$5\%$ in all analysis.
In the univariate analysis between term and preterm groups, the Mann–Whitney U test adjusted by Holm–Bonferroni multiple comparison was used for quantitative variables, and χ2 with Fisher correction was used for qualitative variables. In addition, Spearman-rho was used to assess correlations between the levels of BM hormones and maternal body composition. To analyze the neonatal growth curves and hormones over the lactation period, a two-way ANOVA was used to test the differences between groups (preterm/term), the day of lactation (d), and the interaction effect between the group and the day (g*d). In the growth curves, the interaction between the group and the sex was also reported.
In the multivariate analysis, two different models were used according to the strategy analysis to explore. Firstly, to explore the contribution of BM hormones to the neonatal growth pattern, linear regression models were used, with natural logarithmic transformation to normalize the weight, length, and head circumference. In this case, the prematurity and day subsets, adjusted by the neonatal sex and body fat of the women, was the strategy applied. Secondly, to detect if maternal body composition or dietary intakes were associated with BM hormones, the mixed models were used, considering woman by days as a random effect, and only those hormones showing association on neonatal growth were considered in this second analysis. In addition, the models were clustered by term or preterm infants and the anthropometry and dietary variables were normalized by natural logarithmic transformation. In all models, the coefficient (β) and standard error (SE) were reported.
## Maternal and neonatal characteristics at birth
There were no differences in sociodemographic characteristics between women with term and preterm labor. Regarding gestational variables, women with premature labor had higher rates of completed corticosteroid cycles and magnesium sulfate therapies than women with term labor. A higher proportion of male infants was observed in the preterm compared to the term group. No other differences in neonatal characteristics were detected (Table 1).
**Table 1**
| Unnamed: 0 | Term (n=30) | Preterm (n=18) | P |
| --- | --- | --- | --- |
| Maternal age (years) | 34.0 [32.2; 35.8] | 35.0 [25.8; 39.5] | 0.89 |
| Educational level | | | 0.26 |
| High school | 40.0% (12) | 66.7% (12) | |
| University | 43.3% (13) | 27.8% (5) | |
| Employment situation | | | 0.37 |
| Student | 0.0% (0) | 5.6% (1) | |
| Working | 63.3% (19) | 55.6% (10) | |
| Unemployment | 20.0% (6) | 33.3% (6) | |
| Nationality | | | 0.99 |
| Spanish | 56.7% (17) | 61.1% (11) | |
| Non-Spanish | 30.0% (9) | 33.3% (6) | |
| Family size | 4.0 [3.0; 4.0] | 4.0 [3.0; 5.0] | 0.98 |
| Gravida | 2.0 [2.0; 2.0] | 2.0 [2.0; 2.0] | 0.82 |
| Abortion | 0.0 [0.0; 0.8] | 0.0 [0.0; 1.0] | 0.25 |
| Assisted reproduction techniques | 3.3% (1) | 0.0% (0) | 0.99 |
| Vitamin D deficiency | 16.7% (5) | 27.8% (5) | 0.47 |
| Gestational hypothyroidism | 36.7% (11) | 38.9% (7) | 0.99 |
| Gestational anemia | 20.0% (6) | 22.2% (4) | 0.99 |
| Gestational diabetes | 23.3% (7) | 16.7% (3) | 0.72 |
| Preeclampsia | 6.7% (2) | 16.7% (3) | 0.36 |
| Antibiotic therapy | 33.3% (10) | 50.0% (9) | 0.42 |
| Corticosteroids (completed cycle) | 0.0% (0) | 61.1% (11) | <0.001 |
| Magnesium sulfate therapy | 0.0% (0) | 50.0% (9) | <0.001 |
| C-section | 26.7% (8) | 38.9% (7) | 0.57 |
| Gestational age (weeks) | 38.9 [38.0; 39.9] | 28.7 [27.2; 34.0] | <0.001 |
| Neonate sex (male) | 26.7% (8) | 61.1% (11) | 0.040 |
| Apgar 5 min | 10.0 [9.0; 10.0] | 9.0 [8.0; 10.0] | 0.11 |
| Birth weight (Z-score) | -0.03 [-0.62; 0.32] | 0.12 [-0.84; 0.80] | 0.54 |
| Birth length (Z-score) | -0.32 [-0.89; 0.17] | 0.14 [-0.60; 0.92] | 0.16 |
| Birth head circumference (Z-score) | -0.04 [-0.35; 0.50] | -0.22 [-0.67; 0.26] | 0.41 |
| Intrauterine growth restriction | 10.0% (3) | 0.0% (0) | 0.54 |
## Neonatal growth during the first month of life
We analyzed the growth pattern and velocity during the first month of life. The weight, length, BMI, and head circumference were significantly larger in term compared to preterm neonates at all time points (Figures 2A-C, S1A). However, growth velocities were not significantly different between groups (Figures 2D-F, S1B). The prematurity and days did not show interaction in any of the neonatal anthropometry parameters. No interactive effect was observed between prematurity and sex on growth (data not shown).
**Figure 2:** *Neonatal anthropometry pattern during the first month of life in weight (A), length (B), and head circumference (C) and weight gain (D), length gain (E), and head circumference gain (F). Data show the median and interquartile range [Q1; Q3]. Black lines show term infants (Birth n=30, 7 days n=22, 14 days n=17, 28 days n=15), and blue lines show preterm infants (Birth n=18, 7 days n=15, 14 days n=13, 28 days n=12). In the two way-ANOVA, the following were considered as factors: preterm/term (group = g), birth/7days/14 days/28 days (day = d), and interaction effect group and day (g * d). The P-value (P) was extracted by the Mann–Whitney U test in the boxplots.*
## Breast milk hormone levels in the first month of lactation
Figure 3 shows the levels of the hormones analyzed at the three time points during the first month of lactation. The “Birth” point was excluded from the analysis since BM was not obtained in the hours immediately after birth. When compared between two groups, the levels of insulin, leptin, resistin, the GIP, and peptide YY were significantly higher in BM from women with preterm compared with term labor, while ghrelin was significantly lower.
**Figure 3:** *Breast milk levels of insulin (A), leptin (B), ghrelin (C), resistin (D), the gastrointestinal peptide (GIP); (E), and peptide YY (F) during the first month of lactation. Data show the median and interquartile range [Q1; Q3]. Black lines show term infants (7 days n = 20, 14 days n = 18, 28 days n = 17), and blue lines show preterm infants (7 days n = 13, 14 days n = 10, 28 days n = 10). In the two way-ANOVA, the following were considered as factors: preterm/term (group = g), birth/7days/14 days/28 days (day = d), and the interaction effect group and day (g * d).*
With respect to variation over time along the first month of lactation, the BM levels of resistin significantly decreased, while GIP and peptide YY increased, with insulin, leptin, and ghrelin being stable levels. Only insulin levels showed an interaction effect between prematurity and days, indicating that, in BM from women with premature labor, insulin decreases during the first weeks postpartum, while it is stable in BM from mothers with term labor (Figure 3).
## Association between breast milk hormones and neonatal growth
To explore the contribution of BM hormones to the neonatal growth pattern, linear regression models were used adjusted by the sex and women fat mass percentage, preterm, and day in separated sets.
In term neonates, weight gain along lactation was associated with a decrease in BM peptide YY at 7 and 28 days of lactation and with an increase in insulin at day 7. In preterm neonates, weight gain was associated with the increase in BM ghrelin at day 14 (Figure 4A). In term neonates, length gain was associated with the increase in BM insulin levels at day 7. In preterm neonates, length gain was associated with the increase in BM ghrelin at day 14 (Figure 4B). In term neonates, the increase in head circumference was associated with a decrease in BM peptide YY at day 7 and 28 of lactation. In preterm neonates, there was no association between head circumference and BM hormones (Figure 4C).
**Figure 4:** *Linear regression models showing the association between breast milk (BM) hormones and neonatal weight (A), length (B), and head circumference (C). The neonatal growth variables were normalized by logarithmic transformation, and the coefficients (beta) for each day of lactation were separately analyzed. Data show beta and standard error (SE). The red lines mean significant association (P < 0.05), and non-significant associations are represented in black. All models were adjusted by sex and women fat mass percentage. GIP, gastrointestinal peptide.*
## Maternal body composition and dietary pattern during the first month of lactation
At 7 days, women with premature labor had significantly higher BMI and body fat and lower muscle percentage than women with term labor. However, at day 28, no statistical differences were detected between the groups in any of the anthropometric parameters (Table 2).
**Table 2**
| Unnamed: 0 | 7 days | 7 days.1 | 7 days.2 | 28 days | 28 days.1 | 28 days.2 |
| --- | --- | --- | --- | --- | --- | --- |
| Body composition | Term (n=21) | Preterm (n=16) | P | Term (n=18) | Preterm (n=11) | P |
| BMI (kg/m2) | 24.7[23.2; 27.5] | 28.8[26.1; 30.6] | 0.036 | 24.9[23.1; 28.3] | 28.7[25.0; 30.9] | 0.09 |
| WHI | 0.90[0.85; 0.91] | 0.90[0.84; 0.94] | 0.90 | 0.87[0.81; 0.89] | 0.88[0.84; 0.93] | 0.65 |
| Body fat mass (%) | 35.9[32.1; 41.6] | 42.4[37.4; 46.2] | 0.021 | 38.5[35.6; 42.0] | 40.8[37.8; 46.9] | 0.11 |
| Muscle mass (%) | 26.1[25.6; 28.9] | 24.6[22.8; 26.4] | 0.033 | 25.8[24.6; 27.3] | 25.9[22.9; 26.4] | 0.31 |
| Dietary intakes | Term (n=20) | Preterm (n=12) | P | Term (n=14) | Preterm (n=6) | P |
| Energy (Kcal) | 2,002[1,399; 2,400] | 2,109[1,882; 2,432] | 0.41 | 2,150[1,785; 2,248] | 1,820[1,528; 1,878] | 0.28 |
| Proteins (g) | 95.6[72.4; 106] | 87.6[78.0; 100] | 0.76 | 90.9[85.0; 106] | 90.0[79.9; 99.8] | 0.54 |
| Carbohydrates (g) | 204[168; 216] | 204[171; 268] | 0.71 | 194[162; 214] | 156[144; 186] | 0.23 |
| Fat (g) | 95.5[54.6; 114] | 89.4[77.2; 108] | 0.88 | 100[73.6; 105] | 76.2[58.6; 90.9] | 0.14 |
| SFAs (g) | 31.2[21.6; 40.2] | 31.0[22.6; 34.0] | 0.92 | 30.6[22.7; 38.4] | 26.2[21.4; 32.9] | 0.41 |
| MUFAs (g) | 42.2[21.7; 47.3] | 39.6[33.7; 45.0] | 0.82 | 40.5[32.5; 46.3] | 30.1[21.3; 41.6] | 0.14 |
| PUFAs (g) | 11.7[5.40; 14.9] | 12.6[10.9; 14.0] | 0.37 | 14.2[12.5; 16.1] | 11.6[8.50; 12.6] | 0.026 |
| Cholesterol (mg) | 418[271; 464] | 356[312; 361] | 0.28 | 378[308; 462] | 311[227; 394] | 0.30 |
| Fiber (g) | 19.3[13.2; 24.5] | 22.1[19.7; 28.0] | 0.44 | 21.8[15.4; 27.8] | 17.3[15.9; 21.1] | 0.74 |
| Water (ml) | 1,564[1,216; 2,362] | 1,451[1,349; 2,272] | 0.97 | 1,901[1,359; 2,868] | 1,654[1,261; 2,776] | 0.81 |
At day 7, the dietary pattern was not statistically different between women with or without term labor, neither in macronutrients nor in mineral or vitamin intakes. However, at day 28, the women with premature labor significantly decreased their PUFAs and vitamin E intake compared to women with term labor (Tables 2, S1). We had some missing data of the 72hDR questionnaires by day 28, due to the lack of answers by the participant, or the change of residence, among other reasons. Therefore, to assess the possible interference of these follow-up loses with the results, we performed a secondary intention to treat analysis comparing those women who only answered the 72hDR at day 7 with women who finished the study. This analysis showed that, in women with term delivery, those who did not complete the study had a lower PUFA intake (5.40 [4.70; 8.95] g) compared to those who completed the study (14.3 [12.1; 15.6] g; $$P \leq 0.004$$). However, in women with preterm delivery, those who did not complete the study had similar PUFA intake (12.9 [10.8; 14.9] g) compared to those who completed the study (12.5 [12.2; 12.7] g; $$P \leq 0.57$$).
## Association between maternal body composition and dietary intake with breast milk hormones
The mixed models, considering woman by days as a random effect, were built to study if maternal body composition (Figure 5A) or dietary intake (Figure 5B) were associated with BM hormones. Only the hormones that influenced the neonatal growth were considered in the analysis. In women with term labor, none of the anthropometric variables were associated with BM hormones. However, in women with premature labor, the BMI and body fat mass were positively associated with ghrelin levels in their BM (Figure 5A). This means that for each increased unit in the BMI or body fat mass, ghrelin levels also increased by 7.7 ± 0.8 and 5.6 ± 1.9 units, respectively. We also performed a correlation analysis across gestational ages to evaluate the influence of maternal body composition on BM hormones. We found a positive correlation between the BMI and body fat with insulin and leptin levels; the waist–hip index was also positively correlated with leptin, and body fat was positively correlated with the GIP. However, we did not detect correlations between any of the maternal body composition parameters and BM ghrelin across gestational ages (Figure S2). In addition, we analyzed the possible influences of maternal body composition on those BM hormones that did not influence neonatal growth. We only found a negative association between the percentage of muscle with BM insulin levels in women with premature labor (Figure S3A).
**Figure 5:** *Linear mixed models on the association between maternal anthropometry (A) and the maternal dietary pattern (B) on BM hormones that influenced neonatal growth in each cohort. Women by day were included as the random effect, and maternal variables were normalized by logarithmic transformation. Data show beta and SE. Red lines mean significant association (P < 0.05), and non-significant associations are represented in black. BMI, body mass index; WHI, waist-to-hip index; SFAs, saturated fatty acids; MUFAs, monounsaturated fatty acids; PUFAs, polyunsaturated fatty acids.*
Regarding the dietary pattern, in women with term labor, we did not detect any association with insulin levels. However, fiber intake was negatively associated with peptide BM YY (β = -0.2 ± 0.1; Figure 5B). In women with preterm labor, protein and fiber intake were positively associated with the levels of BM ghrelin (β = 2.6 ± 1.1 and β = 1.8 ± 0.5, respectively; Figure 5B). In addition, we analyzed the possible influences of the maternal diet on those BM hormones that did not influence neonatal growth. We only found a positive association between cholesterol intake with BM ghrelin levels in women with term labor (Figure S3B).
A secondary analysis in the preterm neonatal cohort demonstrated that the use of antenatal corticosteroids had a negative impact on insulin, leptin, ghrelin, resistin, and peptide YY levels for infant weight. Regarding length, a negative association was found with leptin, ghrelin, and resistin, and no association was observed with head circumference. On the other hand, the use of antenatal magnesium sulfate had a negative association on the GIP and peptide YY for head circumference (Table S2).
## Discussion
This article analyzes the levels of some BM hormones that influence appetite and metabolism comparing mothers with term and preterm delivery, focused on their association with the growth of their neonates during the first month of lactation. We also evaluate the influence of the maternal nutritional status on these BM hormones. We found important variations in the levels of the analyzed hormones between term and preterm delivery and a differential influence of these hormones on infant growth depending on the type of labor. In term infants, growth was positively affected by BM insulin levels, while peptide YY exerted a negative impact. On the other hand, in preterm infants, growth was mainly influenced by BM ghrelin levels, which had a positive association on body weight and length gain. Regarding the influence of maternal factors on BM hormone levels, our data indicate that, in women with term labor, body composition does not seem to affect BM hormones. However, in women who gave birth prematurely, fat mass was positively associated with BM ghrelin levels. With respect to the influence of diet, we detected that fiber intake affected some of the BM hormones which influenced neonatal growth, being negatively associated with the levels of peptide YY in women with term delivery and positively associated with ghrelin in women with preterm labor (summary shown in Figure 6). The present study demonstrates that preterm labor affects the content of appetite hormones in BM, being influenced by maternal body fat and some dietary components. Since these hormones exert an influence on infant growth, it would be important to monitor maternal nutrition and body composition during lactation.
**Figure 6:** *Summary of the main results of the study. Figure shows the maternal dietary components and body composition parameters that influence BM hormone levels and their association on neonatal growth. The arrows indicate associations between parameters. .*
BM is the gold standard for infant nutrition, and gaining knowledge in its components is critical to understand how it influences infant growth and development. This is particularly relevant in the population of preterm neonates, which exhibit a growth deficiency in the short term [7], followed by growth acceleration, which may contribute to the observed high metabolic risk of this population [9]. In this context, BM hormones that modulate metabolism and appetite are relevant since they may contribute to counteract these growth alterations and explain the reduction of risk to develop obesity and diabetes in individuals who were breastfeed [13, 14]. Even though BM is the best food for the neonate, in some situations, it is not possible to breastfeed, and, therefore, the knowledge gained about BM hormones can guide the pharmaceutical industry to develop a better infant formula.
## Differences in BM hormones between term and preterm delivery
Our first objective was to evaluate possible differences in hormones that influence appetite and metabolism in BM from women with term and preterm labor. We found significant differences in the levels of insulin, leptin, resistin, ghrelin, the GIP, and peptide YY between these groups, all being higher in women with preterm labor, except for ghrelin, which showed the opposite trend. The studied hormones can access BM from the maternal plasma or can be produced by lactocytes [15]. Maternal circulation seems to be the main source of insulin [12], due to the limited capacity of mammary epithelial cells to synthesize it [20], and of resistin, since its levels in BM correlate with those found in maternal plasma [21]. On the other hand, the mammary gland and maternal plasma contribute equally to BM ghrelin levels [12] and leptin can enter BM from maternal plasma [44] or can be synthesized by lactocytes [20]. Considering that the mammary gland has an ongoing development during the whole pregnancy [45], women with preterm delivery have an underdeveloped gland, with leaky tight junctions [46]. This may account, at least in part, for the observed differences, by modifying the passage of molecules from serum or the capacity of lactocyte synthesis. We did not analyze the plasma levels of the studied hormones, and, therefore, we cannot determine if they account for the differences observed between groups. This is a limitation of our study, and future work measuring hormone levels simultaneously in BM and in maternal plasma will allow to determine the relative contribution of lactocyte synthesis and passage from the blood. In addition, it is possible that, in the present study, the levels of some hormones may be underestimated since we used defatted BM and it has been shown that ghrelin [29] and leptin levels are positively correlated with BM fat content [20]. However, the differences between term and preterm BM are still valid since previous meta-analysis demonstrated similar fat content in the BM of women with term and preterm labor [47]. Overall, our study demonstrates that prematurity is a factor contributing to changes in the BM levels of relevant hormones that modulate appetite and infant growth.
## Influence of maternal nutrition and body composition on breast milk hormones
Maternal body composition and the dietary pattern have been previously shown to influence BM hormones. For example, a positive association between maternal BMI and BM leptin concentration is consistently found in most studies [24, 48, 49] and was also confirmed in our cohort, together with a positive correlation between the BMI, fat mass, and insulin levels. However, the evidence for an association between maternal body composition and other hormones in BM is lacking [48]. Our study also evidenced that ghrelin concentration was positively associated with maternal fat mass but only in BM from women with preterm labor. Higher fat mass was observed in this group (likely related to the interruption of pregnancy due to preterm labor, when fat accumulation is high), normalizing at the end of the first month. Moreover, there is evidence that ghrelin synthesis peaks around mid-pregnancy in association with an increase in maternal weight gain and fat storage during gestation [50]. In addition, it is possible that women with preterm labor accumulated more body fat due to the dysregulation of their appetite. Prematurity is a key factor for maternal psychological stress [51, 52], and exposure to a stressor has been shown to increase ghrelin levels in humans, which is considered a relevant hormone for stress-induced hyperphagia [53]. Therefore, it is possible that women with preterm labor, particularly those with higher body fat, may also have elevated plasma ghrelin, which would be interesting to confirm in future studies. Nevertheless, we found that BM ghrelin levels were lower in women with preterm delivery, compared to those with term labor. Since BM ghrelin is both released from maternal plasma and from the mammary gland [29], we suggest that the lower BM content in women with preterm labor is due to lactocyte immaturity [54].
We also analyzed the impact of the maternal diet on BM hormone levels. Overall, the diet between women with term or preterm delivery did not show marked differences. We only detected differences in PUFA intake at day 28, being lower in women with preterm delivery. This may be related to a worse dietary behavior in this group, as we have previously demonstrated [55]. However, we must consider a possible overestimation of PUFA intake in the women with term delivery due to the dropout of those with lower intake, as shown by the intention-to-treat analysis. With this consideration, it is necessary to control the potential high risk of significance bias. An interesting observation was the interaction between fiber intake and the levels of some BM hormones. In women with term delivery, fiber intake was negatively associated with BM peptide YY levels, while in women with premature labor, it was positively associated with BM ghrelin levels. This is a surprising result since ghrelin is an orexigenic hormone, which would be most likely decreased by fiber ingestion since it promotes satiation [56]. However, it has been demonstrated that different types of dietary fiber have different effects on satiety and the plasma levels of gut hormones, such as ghrelin and peptide YY [57]. Furthermore, the influence of fiber intake on gut satiety–related hormone release markedly differs depending on the population studied, their glucose homeostasis, and the chemical nature of the fiber (soluble versus insoluble) [58, 59]. Given the fact that ghrelin in BM comes from plasma and the mammary gland [29], further analysis assessing its level in maternal blood would shed light on the observed relationship between maternal fiber intake and ghrelin levels in BM.
## Influence of breast milk hormones on infant growth
Our third, and most important, aim was to assess the association of appetite-related BM hormone levels on infant growth. As expected, during the first month of lactation, the body weight, length, and head circumference were larger in term neonates. However, we did not find differences in neonatal Z-scores or growth velocities between term and preterm neonates. It has been proposed that preterm infants exhibit growth acceleration, being slower in those fed with BM [8]. However, we did not observe catch-up growth in the present study, which may be related to the fact that this growth acceleration is influenced by several factors, such as intrauterine growth restriction, birth weight, or comorbidities, among others [60], and may not take place until the pathological factor disappears, usually after the first year of life [61].
We observed that in term and preterm neonates, growth was influenced by different hormones. In term neonates, weight and length were positively associated with BM insulin levels. It has been proposed that BM insulin plays a role in both neonatal growth in the first month of life [62] and at 1 year of age [63], and high systemic insulin levels have been associated with larger growth [64]. However, it has been put forward that insulin BM levels and infant growth have a U-shape pattern and that intermediate concentrations of insulin in BM may be optimal to support infant metabolism, while insufficient or excessive insulin may impair this process [63]. In term infants, we also found that BM peptide YY levels were negatively associated with weight and head circumference during the first month of life. Since peptide YY is an anorexigenic signal, it is possible that BM content reduces appetite and thus energy intake. In preterm infants, growth was associated with a different pattern of hormones, the most relevant being ghrelin, which showed a positive correlation with growth. Our findings agree with the correlation between serum ghrelin and skinfold size observed in breastfed infants [65]. The growth-promoting effects of ghrelin could be related to its orexigenic actions, which may increase appetite in the neonate. Another possible effect is through the release of growth hormones [66]. Our observation of higher BM ghrelin levels in women with high body fat suggest that, in women with preterm labor, higher fat accumulation would be beneficial to increase this orexigenic hormone in their BM with a beneficial effect for premature infant growth. Our data showing the modulation of infant growth by BM hormones indicate that they must reach the plasma and, therefore, they must be absorbed. There is evidence that infants can enterally absorb leptin, including preterm neonates, shown by the correlation between leptin in BM and in infant plasma [67]. Leptin receptors have been located in human intestinal mucosae, even since the fetal stage [68], and, in rats, it has been demonstrated that oral leptin is absorbed by the neonatal epithelium, exerting biological effects [69]. Regarding how peptide hormones can reach the intestinal mucosae without degradation, it is possible that the lower pepsin secretion and the higher gastric pH in neonates hydrolyze less protein in the infant’s stomach [70, 71]. Given the substantial amount of evidence showing the role of BM hormones in neonatal growth, this aspect needs further attention. In addition to the effect of BM hormones, we cannot rule out the contribution of infant endogenous hormones for growth. However, it must be considered that, at birth, the levels of many of these hormones are altered in preterm infants [9], and the dysregulation in adipokines and gut-derived hormones may negatively contribute to immediate postnatal growth or program the development of metabolic disease later in life. Therefore, their source from maternal BM may have a regulatory effect, equilibrating levels. In our study, we must also take into account the fact that, in most cases, preterm infants were fed with supplements apart of BM, and we did not assess the relative amount of BM ingested, which is a limitation of our study. The use of supplements for preterm infants could be beneficial for growth if BM intake is low. However, if the mother can provide sufficient volume, attention should be focused on supporting maternal nutrition to improve BM macronutrients and bioactive molecules, such as hormones, which would, in turn, benefit infant growth.
We also analyzed the impact of antenatal corticosteroids and magnesium sulfate, which are commonly used in the context of prematurity, and there is controversy regarding some of their effects [72]. We found a negative influence of corticosteroid administration on infant weight and length through alterations in BM hormone levels. To the best of our knowledge, there is no information about the effect of antenatal corticosteroid treatment on BM hormone secretion, but it is known that they reduce the volume of BM [73]. This has also been confirmed in experimental animals, also showing that, together with reduced BM yield, neonatal growth is compromised [74]. Our data point in this direction, with a negative association of the use of antenatal corticoids and magnesium sulfate on neonatal growth through the BM hormones. This aspect deserves further studies.
## Conclusions
In conclusion, we found important variations in the content of BM hormones between term and preterm delivery, which may be related to different mammary gland development and maternal body composition. However, it must be considered that, in addition to preterm labor itself, other maternal influences, such as pregnancy complications or a woman’s genetic background, may also influence BM hormones. Our study also evidences that BM insulin is the main growth-driving hormone in term neonates, while ghrelin is relevant in premature infants. Given the fact that ghrelin is lower in preterm BM, being associated with maternal fat storage and maternal dietary components, our data support the importance to monitor diet and body composition in women who gave birth prematurely to improve the BM hormonal status. Our data also support a negative influence of corticosteroid administration on infant growth through alterations in BM hormone levels.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by Ethical Committee of Hospital Clínico San Carlos (Ref. $\frac{19}{393}$-E). Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.
## Author contributions
DR-C has contributed to the design and acquisition; performed the analysis and interpretation of data; and drafted the manuscript, tables, and figures. PS has contributed to the acquisition and interpretation of data and drafted the manuscript. GC and AG-D have contributed to the recruitment of infants and collection of samples. MM-C and CM have contributed to the interpretation of data and editing of the manuscript. SA has contributed to the conception, design, interpretation, and review of the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1090499/full#supplementary-material
## References
1. Beck S, Wojdyla D, Say L, Betran AP, Merialdi M, Requejo JH. **The worldwide incidence of preterm birth: A systematic review of maternal mortality and morbidity**. *Bull World Health Organ* (2010) **88**. DOI: 10.2471/BLT.08.062554
2. Walani SR. **Global burden of preterm birth**. *Int J Gynaecol Obstet* (2020) **150**. DOI: 10.1002/ijgo.13195
3. Blencowe H, Cousens S, Oestergaard MZ, Chou D, Moller A-B, Narwal R. **National, regional, and worldwide estimates of preterm birth rates in the year 2010 with time trends since 1990 for selected countries: A systematic analysis and implications**. *Lancet* (2012) **379**. DOI: 10.1016/S0140-6736(12)60820-4
4. Vogel JP, Chawanpaiboon S, Moller A-B, Watananirun K, Bonet M, Lumbiganon P. **The global epidemiology of preterm birth**. *Best Pract Res Clin Obstet Gynaecol* (2018) **52** 3-12. DOI: 10.1016/j.bpobgyn.2018.04.003
5. Patel RM. **Short- and long-term outcomes for extremely preterm infants**. *Am J Perinatol* (2016) **33**. DOI: 10.1055/s-0035-1571202
6. Jasper EA, Cho H, Breheny PJ, Bao W, Dagle JM, Ryckman KK. **Perinatal determinants of growth trajectories in children born preterm**. *PloS One* (2021) **16**. DOI: 10.1371/journal.pone.0245387
7. Niklasson A, Engstrom E, Hard A-L, Wikland KA, Hellstrom A. **Growth in very preterm children: A longitudinal study**. *Pediatr Res* (2003) **54** 899-905. DOI: 10.1203/01.PDR.0000091287.38691.EF
8. Toftlund LH, Halken S, Agertoft L, Zachariassen G. **Catch-up growth, rapid weight growth, and continuous growth from birth to 6 years of age in very-Preterm-Born children**. *Neonatology* (2018) **114**. DOI: 10.1159/000489675
9. Han L, Li B, Xu X, Liu S, Li Z, Li M. **Umbilical cord blood adiponectin, leptin, insulin, and ghrelin in premature infants and their association with birth outcomes**. *Front Endocrinol (Lausanne)* (2021) **12**. DOI: 10.3389/fendo.2021.738964
10. Ross MG, Desai M. **Developmental programming of appetite/satiety**. *Ann Nutr Metab* (2014) **64** 36-44. DOI: 10.1159/000360508
11. Badillo-Suárez PA, Rodríguez-Cruz M, Nieves-Morales X. **Impact of metabolic hormones secreted in human breast milk on nutritional programming in childhood obesity**. *J Mammary Gland Biol Neoplasia* (2017) **22**. DOI: 10.1007/s10911-017-9382-y
12. Suwaydi MA, Gridneva Z, Perrella SL, Wlodek ME, Lai CT, Geddes DT. **Human milk metabolic hormones: Analytical methods and current understanding**. *Int J Mol Sci* (2021) **22**. DOI: 10.3390/ijms22168708
13. Gunderson EP, Greenspan LC, Faith MS, Hurston SR, Quesenberry CP. **Breastfeeding and growth during infancy among offspring of mothers with gestational diabetes mellitus: A prospective cohort study**. *Pediatr Obes* (2018) **13** 492-504. DOI: 10.1111/ijpo.12277
14. Ortega-García JA, Kloosterman N, Alvarez L, Tobarra-Sánchez E, Cárceles-Álvarez A, Pastor-Valero R. **Full breastfeeding and obesity in children: A prospective study from birth to 6 years**. *Child Obes* (2018) **14**. DOI: 10.1089/chi.2017.0335
15. Savino F, Liguori SA, Fissore MF, Oggero R. **Breast milk hormones and their protective effect on obesity**. *Int J Pediatr Endocrinol* (2009) **2009**. DOI: 10.1155/2009/327505
16. Fu Z, Gilbert ER, Liu D. **Regulation of insulin synthesis and secretion and pancreatic beta-cell dysfunction in diabetes**. *Curr Diabetes Rev* (2013) **9** 25-53. DOI: 10.2174/157339913804143225
17. Zhang Y, Proenca R, Maffei M, Barone M, Leopold L, Friedman JM. **Positional cloning of the mouse obese gene and its human homologue**. *Nature* (1994) **372**. DOI: 10.1038/372425a0
18. Savino F, Liguori SA. **Update on breast milk hormones: leptin, ghrelin and adiponectin**. *Clin Nutr* (2008) **27**. DOI: 10.1016/j.clnu.2007.06.006
19. Kugananthan S, Gridneva Z, Lai CT, Hepworth AR, Mark PJ, Kakulas F. **Associations between maternal body composition and appetite hormones and macronutrients in human milk**. *Nutrients* (2017) **9**. DOI: 10.3390/nu9030252
20. Smith-Kirwin SM, O’Connor DM, de Johnston J, Lancey ED, Hassink SG, Funanage VL. **Leptin expression in human mammary epithelial cells and breast milk**. *J Clin Endocrinol Metab* (1998) **83**. DOI: 10.1210/jcem.83.5.4952
21. Ilcol YO, Hizli ZB, Eroz E. **Resistin is present in human breast milk and it correlates with maternal hormonal status and serum level of c-reactive protein**. *Clin Chem Lab Med* (2008) **46**. DOI: 10.1515/CCLM.2008.019
22. Park HK, Ahima RS. **Resistin in rodents and humans**. *Diabetes Metab J* (2013) **37**. DOI: 10.4093/dmj.2013.37.6.404
23. Steppan CM, Bailey ST, Bhat S, Brown EJ, Banerjee RR, Wright CM. **The hormone resistin links obesity to diabetes**. *Nature* (2001) **409**. DOI: 10.1038/35053000
24. Schueler J, Alexander B, Hart AM, Austin K, Larson-Meyer DE. **Presence and dynamics of leptin, GLP-1, and PYY in human breast milk at early postpartum**. *Obes (Silver Spring)* (2013) **21**. DOI: 10.1002/oby.20345
25. Berseth CL, Michener SR, Nordyke CK, Go VL. **Postpartum changes in pattern of gastrointestinal regulatory peptides in human milk**. *Am J Clin Nutr* (1990) **51**. DOI: 10.1093/ajcn/51.6.985
26. Devesa J. **The complex world of regulation of pituitary growth hormone secretion: The role of ghrelin, klotho, and nesfatins in it**. *Front Endocrinol (Lausanne)* (2021) **12**. DOI: 10.3389/fendo.2021.636403
27. Kojima M, Kangawa K. **Ghrelin: structure and function**. *Physiol Rev* (2005) **85** 495-522. DOI: 10.1152/physrev.00012.2004
28. van der Lely AJ, Tschöp M, Heiman ML, Ghigo E. **Biological, physiological, pathophysiological, and pharmacological aspects of ghrelin**. *Endocr Rev* (2004) **25**. DOI: 10.1210/er.2002-0029
29. Kierson JA, Dimatteo DM, Locke RG, Mackley AB, Spear ML. **Ghrelin and cholecystokinin in term and preterm human breast milk**. *Acta Paediatr* (2006) **95**. DOI: 10.1080/08035250600669769
30. Nauck MA, Quast DR, Wefers J, Pfeiffer AFH. **The evolving story of incretins (GIP and GLP-1) in metabolic and cardiovascular disease: A pathophysiological update**. *Diabetes Obes Metab* (2021) **23** 5-29. DOI: 10.1111/dom.14496
31. Sadr Dadres G, Whitaker KM, Haapala JL, Foster L, Smith KD, Teague AM. **Relationship of maternal weight status before, during, and after pregnancy with breast milk hormone concentrations**. *Obes (Silver Spring)* (2019) **27**. DOI: 10.1002/oby.22409
32. Rassie K, Mousa A, Joham A, Teede HJ. **Metabolic conditions including obesity, diabetes, and polycystic ovary syndrome: Implications for breastfeeding and breastmilk composition**. *Semin Reprod Med* (2021) **39**. DOI: 10.1055/s-0041-1732365
33. Leghi GE, Netting MJ, Lai CT, Narayanan A, Dymock M, Rea A. **Reduction in maternal energy intake during lactation decreased maternal body weight and concentrations of leptin, insulin and adiponectin in human milk without affecting milk production, milk macronutrient composition or infant growth**. *Nutrients* (2021) **13**. DOI: 10.3390/nu13061892
34. Fenton TR, Nasser R, Eliasziw M, Kim JH, Bilan D, Sauve R. **Validating the weight gain of preterm infants between the reference growth curve of the fetus and the term infant**. *BMC Pediatr* (2013) **13**. DOI: 10.1186/1471-2431-13-92
35. Olsen IE, Lawson ML, Ferguson AN, Cantrell R, Grabich SC, Zemel BS. **BMI curves for preterm infants**. *Pediatrics* (2015) **135**. DOI: 10.1542/peds.2014-2777
36. Patel AL, Engstrom JL, Meier PP, Kimura RE. **Accuracy of methods for calculating postnatal growth velocity for extremely low birth weight infants**. *Pediatrics* (2005) **116**. DOI: 10.1542/peds.2004-1699
37. Patel AL, Engstrom JL, Meier PP, Jegier BJ, Kimura RE. **Calculating postnatal growth velocity in very low birth weight (VLBW) premature infants**. *J Perinatol* (2009) **29**. DOI: 10.1038/jp.2009.55
38. Álvarez P, Ramiro-Cortijo D, Montes MT, Moreno B, Calvo MV, Liu G. **Randomized controlled trial of early arachidonic acid and docosahexaenoic acid enteral supplementation in very preterm infants**. *Front Pediatr* (2022) **10**. DOI: 10.3389/fped.2022.947221
39. Simon L, Hanf M, Frondas-Chauty A, Darmaun D, Rouger V, Gascoin G. **Neonatal growth velocity of preterm infants: The weight z-score change versus Patel exponential model**. *PloS One* (2019) **14**. DOI: 10.1371/journal.pone.0218746
40. Whitton C, Ho JCY, Tay Z, Rebello SA, Lu Y, Ong CN. **Relative validity and reproducibility of a food frequency questionnaire for assessing dietary intakes in a multi-ethnic Asian population using 24-h dietary recalls and biomarkers**. *Nutrients* (2017) **9**. DOI: 10.3390/nu9101059
41. Jacques S, Lemieux S, Lamarche B, Laramée C, Corneau L, Lapointe A. **Development of a web-based 24-h dietary recall for a French-Canadian population**. *Nutrients* (2016) **8**. DOI: 10.3390/nu8110724
42. Savard C, Lemieux S, Lafrenière J, Laramée C, Robitaille J, Morisset A-S. **Validation of a self-administered web-based 24-hour dietary recall among pregnant women**. *BMC Pregnancy Childbirth* (2018) **18** 112. DOI: 10.1186/s12884-018-1741-1
43. Candela Gómez C, Kohen VL, Nogueira TL, Editores Médicos SA. *Guia visual de alimentos y raciones* (2007) 180
44. Weyermann M, Beermann C, Brenner H, Rothenbacher D. **Adiponectin and leptin in maternal serum, cord blood, and breast milk**. *Clin Chem* (2006) **52**. DOI: 10.1373/clinchem.2006.071019
45. Biswas SK, Banerjee S, Baker GW, Kuo C-Y, Chowdhury I. **The mammary gland: Basic structure and molecular signaling during development**. *Int J Mol Sci* (2022) **23**. DOI: 10.3390/ijms23073883
46. Dallas DC, Smink CJ, Robinson RC, Tian T, Guerrero A, Parker EA. **Endogenous human milk peptide release is greater after preterm birth than term birth**. *J Nutr* (2015) **145**. DOI: 10.3945/jn.114.203646
47. Gidrewicz DA, Fenton TR. **A systematic review and meta-analysis of the nutrient content of preterm and term breast milk**. *BMC Pediatr* (2014) **14**. DOI: 10.1186/1471-2431-14-216
48. Andreas NJ, Hyde MJ, Gale C, Parkinson JRC, Jeffries S, Holmes E. **Effect of maternal body mass index on hormones in breast milk: A systematic review**. *PloS One* (2014) **9**. DOI: 10.1371/journal.pone.0115043
49. Andreas NJ, Hyde MJ, Herbert BR, Jeffries S, Santhakumaran S, Mandalia S. **Impact of maternal BMI and sampling strategy on the concentration of leptin, insulin, ghrelin and resistin in breast milk across a single feed: A longitudinal cohort study**. *BMJ Open* (2016) **6**. DOI: 10.1136/bmjopen-2015-010778
50. Garcés MF, Buell-Acosta JD, Ángel-Müller E, Parada-Baños AJ, Acosta-Alvarez J, Saavedra-López HF. **Study of the Ghrelin/LEAP-2 ratio in humans and rats during different phases of pregnancy**. *Int J Mol Sci* (2022) **23**. DOI: 10.3390/ijms23179514
51. Henderson J, Carson C, Redshaw M. **Impact of preterm birth on maternal well-being and women’s perceptions of their baby: A population-based survey**. *BMJ Open* (2016) **6**. DOI: 10.1136/bmjopen-2016-012676
52. Carson C, Redshaw M, Gray R, Quigley MA. **Risk of psychological distress in parents of preterm children in the first year: Evidence from the UK millennium cohort study**. *BMJ Open* (2015) **5**. DOI: 10.1136/bmjopen-2015-007942
53. McKay NJ, Giorgianni NR, Czajka KE, Brzyski MG, Lewandowski CL, Hales ML. **Plasma levels of ghrelin and GLP-1, but not leptin or amylin, respond to a psychosocial stressor in women and men**. *Horm Behav* (2021) **134**. DOI: 10.1016/j.yhbeh.2021.105017
54. Macias H, Hinck L. **Mammary gland development**. *Wiley Interdiscip Rev Dev Biol* (2012) **1**. DOI: 10.1002/wdev.35
55. Gila-Díaz A, Herranz Carrillo G, Arribas SM, Ramiro-Cortijo D. **Healthy habits and emotional balance in women during the postpartum period: Differences between term and preterm delivery**. *Children* (2021) **8**. DOI: 10.3390/children8100937
56. Slavin JL. **Dietary fiber and body weight**. *Nutrition* (2005) **21**. DOI: 10.1016/j.nut.2004.08.018
57. Weickert MO, Spranger J, Holst JJ, Otto B, Koebnick C, Möhlig M. **Wheat-fibre-induced changes of postprandial peptide YY and ghrelin responses are not associated with acute alterations of satiety**. *Br J Nutr* (2006) **96**. DOI: 10.1017/bjn20061902
58. Costabile G, Griffo E, Cipriano P, Vetrani C, Vitale M, Mamone G. **Subjective satiety and plasma PYY concentration after wholemeal pasta**. *Appetite* (2018) **125**. DOI: 10.1016/j.appet.2018.02.004
59. St-Pierre DH, Rabasa-Lhoret R, Lavoie M-È, Karelis AD, Strychar I, Doucet E. **Fiber intake predicts ghrelin levels in overweight and obese postmenopausal women**. *Eur J Endocrinol* (2009) **161** 65-72. DOI: 10.1530/EJE-09-0018
60. Radmacher PG, Looney SW, Rafail ST, Adamkin DH. **Prediction of extrauterine growth retardation (EUGR) in VVLBW infants**. *J Perinatol* (2003) **23**. DOI: 10.1038/sj.jp.7210947
61. Liu X, Luo B, Peng W, Xiong F, Yang F, Wu J. **Factors affecting the catch-up growth of preterm infants after discharge in China: A multicenter study based on the health belief model**. *Ital J Pediatr* (2019) **45** 87. DOI: 10.1186/s13052-019-0674-2
62. Fields DA, Demerath EW. **Relationship of insulin, glucose, leptin, IL-6 and TNF-α in human breast milk with infant growth and body composition**. *Pediatr Obes* (2012) **7**. DOI: 10.1111/j.2047-6310.2012.00059.x
63. Chan D, Goruk S, Becker AB, Subbarao P, Mandhane PJ, Turvey SE. **Adiponectin, leptin and insulin in breast milk: Associations with maternal characteristics and infant body composition in the first year of life**. *Int J Obes (Lond)* (2018) **42** 36-43. DOI: 10.1038/ijo.2017.189
64. Lind MV, Larnkjær A, Mølgaard C, Michaelsen KF. **Breastfeeding, breast milk composition, and growth outcomes**. *Nestle Nutr Inst Workshop Ser* (2018) **89** 63-77. DOI: 10.1159/000486493
65. Yiş U, Oztürk Y, Sişman AR, Uysal S, Soylu OB, Büyükgebiz B. **The relation of serum ghrelin, leptin and insulin levels to the growth patterns and feeding characteristics in breast-fed versus formula-fed infants**. *Turk J Pediatr* (2010) **52** 35-41. PMID: 20402065
66. Müller TD, Nogueiras R, Andermann ML, Andrews ZB, Anker SD, Argente J. **Ghrelin**. *Mol Metab* (2015) **4**. DOI: 10.1016/j.molmet.2015.03.005
67. Chatmethakul T, Schmelzel ML, Johnson KJ, Walker JR, Santillan DA, Colaizy TT. **Postnatal leptin levels correlate with breast milk leptin content in infants born before 32 weeks gestation**. *Nutrients* (2022) **14**. DOI: 10.3390/nu14245224
68. Aparicio T, Kermorgant S, Darmoul D, Guilmeau S, Hormi K, Mahieu-Caputo D. **Leptin and ob-Rb receptor isoform in the human digestive tract during fetal development**. *J Clin Endocrinol Metab* (2005) **90**. DOI: 10.1210/jc.2005-1498
69. Sánchez J, Oliver P, Miralles O, Ceresi E, Picó C, Palou A. **Leptin orally supplied to neonate rats is directly uptaken by the immature stomach and may regulate short-term feeding**. *Endocrinology* (2005) **146**. DOI: 10.1210/en.2005-0112
70. Palla MR, Harohalli S, Crawford TN, Desai N. **Progression of gastric acid production in preterm neonates: Utilization of in-vitro method**. *Front Pediatr* (2018) **6**. DOI: 10.3389/fped.2018.00211
71. Gan J, Bornhorst GM, Henrick BM, German JB. **Protein digestion of baby foods: Study approaches and implications for infant health**. *Mol Nutr Food Res* (2018) **62**. DOI: 10.1002/mnfr.201700231
72. Thevathasan I, Said JM. **Controversies in antenatal corticosteroid treatment**. *Prenat Diagn* (2020) **40**. DOI: 10.1002/pd.5664
73. Henderson JJ, Hartmann PE, Newnham JP, Simmer K. **Effect of preterm birth and antenatal corticosteroid treatment on lactogenesis II in women**. *Pediatrics* (2008) **121** e92-100. DOI: 10.1542/peds.2007-1107
74. Henderson JJ, Hartmann PE, Moss TJM, Doherty DA, Newnham JP. **Disrupted secretory activation of the mammary gland after antenatal glucocorticoid treatment in sheep**. *Reproduction* (2008) **136**. DOI: 10.1530/REP-08-0134
|
---
title: 'Association between napping and 24-hour blood pressure variability among university
students: A pilot study'
authors:
- Jie Dai
- Hua-ying Wu
- Xiao-dong Jiang
- Yong-jie Tang
- Hao-Kai Tang
- Li Meng
- Na Huang
- Jing-yu Gao
- Jian Li
- Julien S. Baker
- Chan-Juan Zheng
- Yi-De Yang
journal: Frontiers in Pediatrics
year: 2023
pmcid: PMC10018217
doi: 10.3389/fped.2023.1062300
license: CC BY 4.0
---
# Association between napping and 24-hour blood pressure variability among university students: A pilot study
## Abstract
### Background
Blood pressure variability (BPV) has been reported to be a predictor of cardiovascular and some cognitive diseases. However, the association between napping and BPV remains unknown. This study aimed to explore the association between napping and BPV.
### Materials and methods
A cross-sectional study including 105 university students was conducted. Participants’ 24 h ambulatory blood pressure monitoring (24 h ABPM) were measured, and napping behaviors were investigated. BPV were measured by the 24 h ABPM, included standard deviation (SD), coefficient of variation (CV), and average real variability (ARV).
### Results
Among the participants, $61.9\%$ reported daytime napping. We found that nap duration was significantly associated with daytime CV of diastolic blood pressure (DBP) ($r = 0.250$, $$P \leq 0.010$$), nighttime CV of systolic blood pressure (SBP) ($r = 0.217$, $$P \leq 0.026$$), 24 h WCV of DBP ($r = 0.238$, $$P \leq 0.014$$), 24 h ARV of SBP ($r = 0.246$, $$P \leq 0.011$$) and 24 h ARV of DBP ($r = 0.291$, $$P \leq 0.003$$). Compared with the no napping group, 24 h WCV of DBP, daytime CV of DBP, and daytime SD of DBP were significantly higher in participants with napping duration >60 min. With multiple regression analysis we found that nap duration was an independent predictor for 24 h ARV of SBP (β = 0.859, $95\%$ CI, 0.101–1.616, $$P \leq 0.027$$) and 24 h ARV of DBP (β = 0.674, $95\%$ CI, 0.173–1.175, $$P \leq 0.009$$).
### Conclusions
Napping durations are associated with BPV among university students. Especially those with napping durations >60 min had a significantly higher BPV than those non-nappers.
## Introduction
Blood pressure variability (BPV) is a concept used to characterize continuous dynamic fluctuations in blood pressure. BPV can be described in short-term variability (within a day, 24 h BPV) and long-term variability (between clinic visits over months and years, visit-to-visit BPV) [1]. There are some common indicators of BPV, including standard deviation (SD), range, coefficient of variation (CV), and average real variability (ARV) [2]. BPV has been reported to be risk a factor or associated with the progression of many diseases (3–7). Among patients with Parkinson's disease(PD), 24 h CV of DBP in the advanced PD group was significantly higher than that in the control group and the early PD group [3]. Also, BPV was found to be related to the severity of obstructive sleep apnea (OSA) [4]. Laure et al. showed that higher systolic BPV was associated with higher risk of dementia (HR = 1.23, $95\%$ CI, 1.01–1.50) and it could be a major clinical predictor of cognitive impairment and dementia [5]. Most importantly, sustained increases in BPV may reflect alterations in cardiovascular regulatory mechanisms or underlying pathological conditions [6]. Accumulating evidence suggests that sustained increase in BPV is associated with an increased risk of subclinical organ damage, cardiovascular events, and all-cause mortality [7].
Among university students, late bedtime and daytime napping are extremely prevalent, due to a variety of factors including academic stress, socialization, coffee intake, and uncontrolled use of social media [8]. A meta-analysis of sleep duration and sleep patterns among Chinese university students showed that the proportion of students who slept less than 6 h/day and 7 h/day (short sleep duration) was $8.4\%$ and $43.9\%$, respectively, and the mean bedtime was 00:51 AM [9]. Short sleep duration at night may contribute to a substantial increase in the duration and frequency of daytime napping which might also impact the incidence of cardiovascular diseases [10, 11]. Previous studies reported a 2.20-fold increase in the risk of heart failure in participants who napped <1.7 times per day compared to those who napped >1.7 times per day [12]. In another cohort study of older adults, the risk of hypertension was found to be 1.18 times higher in those with longer nap durations (≥90 min) than in non-nappers [13]. As we all know, the 24 h sleep-wake cycle is closely related to 24 h blood pressure fluctuations [14, 15]. Longer daytime napping was found to be linked with lower sleep quality in a global survey of athletes [16]. Participants with low sleep quality (inefficient sleep) have been reported to a high prevalence of elevated short-term BPV [17]. However, there is little evidence available that relates to the relationship between napping duration and BPV. In the present study, we hypothesized that napping duration potentially leads to abnormal fluctuations in blood pressure.
## Study population, sample size calculation and study procedures
Our study was conducted in a University in Hunan Province, China from 2020 to 2021. The study recruited university students through a convenience sampling method by flyers or posters. The sample size of the present correlation study was calculated using PASS 2021. Since no previous studies have reported the association between nap and BPV. We used the correlation coefficients between sleep and blood pressure in previous studies, previous studies reported correlations within the range of between 0.25 to 0.45 [18, 19]. Determination of the minimum sample size was calculated by the following estimates: (i) expected correlation between the two variables ($r = 0.35$), (ii) statistical power = $90\%$, (iii) Alpha = $5\%$, (iv) correlation coefficient of the null hypothesis ($r = 0.0$), (v) dropout rate = $20\%$. The sample size of 102 could provide $90\%$ statistical power to test the significant correlations.
In our study, we included university students who (i) provided written informed consent in the study, (ii) are university students during the 2020–2021 period; (iii) students with normal sleep cycle (no medical intern with shift work involved). The exclusion criteria included the following: (i) pregnancy, (ii) taking antihypertensive and hypnotics, (iii) hyperthyroidism, (iv) sleep disturbance or drinking caffeine on the day of the trial, (v) incomplete sleep records, (vi) 24 h ambulatory blood pressure monitoring was ineffective. General characteristics (including age, nation, sex, household income, birth, smoking, drinking, etc) were measured by questionnaire. Moreover, all participants underwent physical examinations (including weight, height, and blood pressure).They completed the 24 h ambulatory blood pressure monitoring (24 h ABPM). Weight was measured using calibrated body composition apparatus (TANITA-MC780MA) and height was measured using a stadiometer. The present study was approved by the Research Ethical Review Committee Board of Hunan Normal University (2019–88). Finally, 105 students were included in the final analysis.
## 24 h ABPM and blood pressure variability assessment
All participants were scheduled to undergo 24 h ABPM measurement using a noninvasive, validated device (Mobil-O-Graph NG, Germany). During the 24 h measurement period, the device was programmed to record readings of each parameter every 30 min during the nighttime (between 23:00 to 7:00) and every 15 min during the daytime (between 7:00–22:59). Recordings will be set to read missing values if they were outside systolic blood pressure (SBP) readings of 60–280 mmHg and diastolic blood pressure (DBP) readings of 30–190 mmHg. The recordings including at least 10 during the daytime and 5 during the nighttime were considered to be valid. [ 17]. Blood pressure values <$\frac{130}{80}$ mmHg over 24 h, <$\frac{135}{85}$ mmHg during the daytime and <$\frac{120}{70}$ mmHg at night were categorized as normotensive [20].
Total number of measurements was summarized and the following parameters of short BPV were calculated. Basic indicators (standard deviations (SD) and coefficients of variation (CV = SD* 100/BP) for SBP and DBP) were calculated by ABPM directly. The average real variability (ARV) calculates the average of the differences (in absolute value) between consecutive BP readings: ARV=1/(N−1)∑$K = 1$N−1|BPk+1−BPk| [2]. Weighted standard deviation (WSD) was calculated using the following formula: WSD = (daytime SD *number of hours awake) + (nighttime SD *number of hours asleep)/24 [21]. The weighted coefficients of variation (WCV) were calculated in the similar way.
## Sleep characteristics
Throughout the 24 h ABPM period, participants were asked to fill out a 24 h behavior diary that included daytime naps and nighttime sleep. Participants were asked to fill diary sleep records at the end of the day. Based on the diary, we obtained the following indicators: nap duration, nap frequency, bedtime, time of awakening, and duration of nighttime sleep.
## Statistical analysis
All statistical analyses were performed using IBM SPSS 20.0 for Windows (SPSS Inc., Chicago, IL, United States) and R Studio software. Participants’ 24 h ABPM indicators, basic characteristics and sleep were described using mean (±SD) and frequency (%). Pearson's correlation coefficient and multiple linear regression was used for analyzing the relationship between daytime nap duration and BPV. To test the potential nonlinear associations between napping and blood pressure variability (24 h ARV, WCV and day/night CV), we used restricted cubic spline regression models with age and sex adjusted. In addition, we divided the nap duration into three groups: no napping (0 min), 0–60 min, and >60 min. LSD-t test was conducted for pairwise comparison between groups to detect whether there were significant differences in BPV within the groups.
## General characteristics
A total of 105 university students (20 males and 85 females) were included for the present study. Their mean age was 18.84 ± 1.21 years, mean BMI was 21.01 ± 3.05 kg/m2, mean bedtime was 00:19, mean nighttime sleep duration 7.52 ± 1.33 h/day, and total sleep time 8.16 ± 1.44 h/day. The proportion of students who slept no more than 6 h/day and ≤7 h/day was $8.57\%$ and $48.57\%$. There were $61.9\%$ of participants reported daytime napping and the average duration of napping was 1.05 ± 0.56 h/day.
Among the 105 university students, $68.60\%$ had normotensive 24 h, daytime and also nighttime blood pressure. The mean 24 h SBP was 107.03 ± 6.81 mmHg, mean 24 h DBP 66.50 ± 5.32 mmHg, mean daytime SBP 108.66 ± 7.06 mmHg, mean daytime DBP 68.10 ± 5.70 mmHg, mean nighttime SBP 101.76 ± 7.21 mmHg, and mean nighttime DBP 61.83 ± 5.90 mmHg (Table 1).
**Table 1**
| Unnamed: 0 | Mean ± SD/N (%), (x ± s) |
| --- | --- |
| Age, year | 18.84 ± 1.21 |
| Male | 85 (80.95%) |
| BMI (Kg/m2) | 21.01 ± 3.05 |
| 24 h SBP (mmHg) | 107.03 ± 6.81 |
| 24 h DBP (mmHg) | 66.50 ± 5.32 |
| Daytime SBP (mmHg) | 108.66 ± 7.06 |
| Daytime DBP (mmHg) | 68.10 ± 5.70 |
| Nighttime SBP (mmHg) | 101.76 ± 7.21 |
| Nighttime DBP (mmHg) | 61.83 ± 5.90 |
| 24 h CV of SBP (%) | 10.45 ± 2.18 |
| 24 h CV of DBP (%) | 14.20 ± 2.68 |
| 24 h WCV of SBP | 9.77 ± 2.00 |
| 24 h WCV of DBP | 13.13 ± 2.56 |
| Daytime CV of SBP (%) | 9.99 ± 2.31 |
| Daytime CV of DBP (%) | 13.18 ± 3.08 |
| Nighttime CV of SBP (%) | 9.03 ± 2.48 |
| Nighttime CV of DBP (%) | 12.92 ± 3.59 |
| 24 h SD of SBP (mmHg) | 11.25 ± 2.51 |
| 24 h SD of DBP (mmHg) | 9.50 ± 1.72 |
| 24 h WSD of SBP | 10.52 ± 2.32 |
| 24 h WSD of DBP | 8.87 ± 1.79 |
| Daytime SD of SBP (mmHg) | 10.90 ± 2.71 |
| Daytime SD of DBP (mmHg) | 9.08 ± 2.17 |
| Nighttime SD of SBP (mmHg) | 9.27 ± 2.66 |
| Nighttime SD of DBP (mmHg) | 8.20 ± 2.70 |
| ARV of SBP | 9.91 ± 2.66 |
| ARV of DBP | 7.99 ± 1.87 |
| Bedtime | 00:19 |
| Nighttime sleep duration (hours/day) | 7.52 ± 1.33 |
| Nighttime sleep duration ≦6 h (%) | 9(8.57%) |
| Nighttime sleep duration ≦7 h (%) | 57 (48.57%) |
| Napping (%) | 65(61.90%) |
| Nap duration | Nap duration |
| no napping (0 min) | 40(38.10%) |
| 0–60 min | 39(37.14%) |
| >60 min | 26 (24.76%) |
| Nap duration for participants with napping (hours/day) | 1.05 ± 0.56 |
| Total sleep duration (hours/day) | 8.16 ± 1.44 |
## Correlation between napping and 24 h ABPM parameters
We found that nap duration was significantly correlated with daytime CV of DBP ($r = 0.250$, $$P \leq 0.010$$), nighttime CV of SBP ($r = 0.217$, $$P \leq 0.026$$), 24 h SD of DBP ($r = 0.193$, $$P \leq 0.036$$), daytime SD of DBP ($r = 0.201$, $$P \leq 0.040$$), nighttime SD of SBP ($r = 0.222$, $$P \leq 0.023$$), 24 h WCV of DBP ($r = 0.238$, $$P \leq 0.014$$), 24 h WSD of DBP ($r = 0.202$, $$P \leq 0.038$$), 24 h ARV of SBP ($r = 0.246$, $$P \leq 0.011$$) and 24 h ARV of DBP ($r = 0.291$, $$P \leq 0.003$$) (Table 2).
**Table 2**
| Unnamed: 0 | SBP | SBP.1 | DBP | DBP.1 |
| --- | --- | --- | --- | --- |
| | r | P | r | P |
| 24 h BP | 0.086 | 0.386 | −0.052 | 0.600 |
| Daytime BP | 0.052 | 0.596 | −0.005 | 0.958 |
| Nighttime BP | 0.134 | 0.173 | 0.089 | 0.364 |
| 24 h CV | 0.094 | 0.340 | 0.177 | 0.070 |
| 24 h WCV | 0.146 | 0.137 | 0.238 | 0.014* |
| Daytime CV | 0.087 | 0.380 | 0.250 | 0.010* |
| Nighttime CV | 0.217 | 0.026* | 0.056 | 0.572 |
| 24 h SD | 0.096 | 0.332 | 0.193 | 0.036* |
| 24 h WSD | 0.145 | 0.139 | 0.202 | 0.038* |
| Daytime SD | 0.090 | 0.362 | 0.201 | 0.040* |
| Nighttime SD | 0.222 | 0.023* | 0.074 | 0.450 |
| 24 h ARV | 0.246 | 0.011* | 0.291 | 0.003* |
## Comparison of 24 h ABPM indicators in different nap duration groups
The nap duration was categorized into three groups: no napping (0 min), 0–60 min, and >60 min. A pairwise comparison revealed that there was significant difference between the nap duration >60 min group and the no napping (0 min) group, and 24 h WCV of DBP, daytime CV of DBP, and daytime SD of DBP were higher in the nap duration >60 min group than the no napping group ($P \leq 0.05$) (Figure 1).
**Figure 1:** *Comparison of BPV between different nap duration groups. BPV, blood pressure variability; DBP, diastolic blood pressure; SBP, systolic blood pressure; SD, standard deviation; CV, coefficient of variation; W, Means weighted. *Means that the BPV indicator is significantly different from the no napping group (P < 0.05).*
## Test for the nonlinear associations between napping and blood pressure variability indicators
Using restricted cubic spline regression model, Figure 2 demonstrates the nonlinear correlation between napping and blood pressure variability indicators. With adjustment of age and sex, restricted cubic spline models suggested that the associations between napping and BPV (24 h ARV, WCV and day/night CV) of SBP/DBP were linear (P for nonlinearity = 0.990, 0.441, 0.919, 0.886, 0.958, 0.878, 0.341, and 0.859, respectively) (Figure 2).
**Figure 2:** *The restricted cubic spline analysis of association between nap duration and blood pressure variability. ARV, average real variability; DBP, diastolic blood pressure; SBP, systolic blood pressure; SD, standard deviation; CV, coefficient of variation; W, Means weighted. *P < 0.05. The cubic spline model was adjusted for age.*
## Multiple linear regression of BPV
Multiple linear regression analysis was performed to identify potential predictors of BPV. Nap duration was independently associated with nighttime CV of SBP and daytime CV of DBP, 24 h WCV of DBP, 24 ARV of SBP and 24 h ARV of DBP ($P \leq 0.05$). Longer nap duration was associated with higher daytime CV of DBP (β = 1.052, $95\%$ CI, 0.167–1.937, $$P \leq 0.020$$), higher CV of SBP at night (β = 0.785, $95\%$ CI, 0.062–1.509, $$P \leq 0.034$$), higher 24 h WCV of DBP (β = 0.827, $95\%$ CI, 0.088–1.565, $$P \leq 0.029$$), higher 24 h ARV of SBP (β = 0.859, $95\%$ CI, 0.101–1.616, $$P \leq 0.027$$) and higher 24 h ARV of DBP (β = 0.674, $95\%$ CI, 0.173–1.175, $$P \leq 0.009$$) (Tables 3, 4).
## Discussion
In the current study we found that in university students insufficient nighttime sleep and late bedtime were quite common. Longer nap duration was correlated with increased BPV, including daytime CV of DBP, nighttime CV of SBP, 24 h SD of DBP, daytime SD of DBP, nighttime SD of SBP, 24 h-WCV of DBP, 24 h-WSD of DBP, 24 h ARV of SBP and 24 h ARV of DBP. Especially for young adults with nap duration >60 min, 24 h WCV of DBP, daytime CV of DBP, and daytime SD of DBP were significantly higher than those without daytime napping. In addition, with Multiple regression analysis we found that nap duration was an independent predictor for nighttime CV of SBP, daytime CV of DBP, 24 h WCV of DBP, 24 ARV of SBP and 24 h ARV of DBP.
## Prevalence of insufficient nighttime sleep and late bedtime
In this study, we found that the average nighttime sleep duration of college students was 7.52 h higher than the 7.08 h reported in a meta-analysis on the sleep of Chinese college students [9]. The average nighttime sleep duration was within the recommended standard of 7–9 h. Yet, notably the percentage of participants with nighttime sleep duration ≤6 h was $8.57\%$ and $48.57\%$ for ≤7 h. Among university students, insufficient nighttime sleep has been shown to be detrimental to health [22, 23]. Adequate sleep is essential to rejuvenate students each day and help them with learning and memory processing [24]. Previous studies have reported that poor sleep quality was associated with poor academic performance [25]. Sleep deprivation was identified as a risk factor for poor mental health, even suicidal ideation [26]. Meanwhile, sleep deprivation was associated with higher levels of IL-1β, TNF-α, and IL-10 [27]. In addition, our study showed that university students generally go to bed at 12:00 PM and later, with $72.12\%$ of students going to bed at 12:00 PM and later. Previous studies have suggested that later bedtime (midnight or later) was related to adverse health outcomes [28]. The SBP and the prevalence of diabetes was significantly increased with the delay of bedtime, after 12pm in large community studies [29, 30]. Another large-scale association study from several countries found that late bedtime was associated with higher risk of general obesity (aOR = 1.20, $95\%$ CI, 1.12–1.29) and abdominal obesity (aOR = 1.20, $95\%$ CI, 1.12–1.28) compared to bedtime between 8:00 PM and 10:00 PM [31].
## Napping duration and BPV
Daytime napping is a common lifestyle in China [32]. However, there have been inconsistent findings about association between napping and health [33]. In some cross-sectional studies, it was reported that participants with daytime napping were more likely to develop obesity and diabetes than those without napping [34, 35]. Particularly, when combined with short sleep duration, daytime napping increased the risk of type 2 diabetes even more [36]. Napping also negatively affected renal health. The risk of microalbuminuria for participants with napping 0–1 h/day, 1–1.5 h/day and >1.5 h/day is 1.552, 1.301 and 1.567-fold compared with those without napping, respectively [37]. However, other studies have also reported that short daytime napping were also considered to be beneficial for physical health, in reducing the risk of cognitive decline in older adults [38]. A systematic review suggested that napping improved cognitive and physical performance and reduces fatigue in athletes [39]. There are also conflicting results regarding the effects of daytime naps on cardiovascular diseases [40]. Cao et al. showed that a much higher incidence of hypertension in those who took longer naps compared with those without napping [41]. In contrast, a cohort study in China showed that only prolonged daytime naps (≥30 min) increased the risk of cardiovascular events by $22\%$, and short daytime naps (<30 min) did not increase the risk [42]. Differently, another study showed that daytime naps >1 h reduced the risk of hypertension with an adjusted OR of 0.70 ($95\%$ CI, 0.51–0.97), and the protective effect of longer daytime naps was still found in the presence of adequate nighttime sleep [43].
In our study, $61.9\%$ of the university students reported participating in daytime napping. We did not find an association between abnormal BP and nap duration (Supplementary Table S1). It is probably the participants in our study are young and in a physical healthy condition. But we found a positive correlation between nap duration and BPV among university students. There are several explanations or potential mechanisms for the findings. Firstly, daytime napping may be associated with inefficient sleep (high level of sleep fragmentation) [44]. It has also been shown that inefficient sleep makes BPV increase [17]. Secondly, evidence has also linked prolonged daytime napping to inadequate nighttime sleep duration, induced strong inhibition of γ-aminobutyric acid receptors in the paraventricular nucleus of the hypothalamus, which could over-activate the sympathetic nervous system, leading to abnormal fluctuations in blood pressure and increased BPV (45–47). Thirdly, previous studies have also found that prolonged napping was associated with higher melatonin [48]. The blood pressure pattern during daytime napping is similar to that of nighttime sleep, with a rapid increase in blood pressure after daytime napping [49, 50]. Therefore, prolonged daytime napping may alter circadian rhythms, resulting in abnormal fluctuations in blood pressure. Our results revealed that there was a significant difference between the nap duration >60 min group and the no napping (0 min) group, and 24 h WCV of DBP, daytime CV of DBP and daytime SD of DBP were higher in the nap duration >60 min group than the no napping group ($P \leq 0.05$). Thus far, the majority of studies on sleep and BPV have focused on patients with obstructive sleep apnea and little evidence is available with respect to the relationship between napping and BPV among the general population. Further investigations demonstrating relationships between nap duration and BPV are warranted, especially larger prospective studies.
BPV can be an important predictor of the progression and severity of cardiac and vascular damage, and cardiovascular events and mortality, especially in high-risk cardiovascular populations [51]. Increased BPV can induce chronic myocardial inflammation, which exacerbates cardiac hypertrophy and myocardial fibrosis, leading to systolic dysfunction in the hypertensive heart. This may be related to mineralocorticoid receptor systems and activation of the local angiotensin II [52]. In women with pre-eclampsia, BPV was observed to be associated with right ventricular strain [53]. In patients with type 2 diabetes, E-selectin, an endothelial-specific molecule involved in vascular inflammation and cardiac metabolism, was positively associated with 24 h diastolic BPV($r = 0.238$) and daytime diastolic BPV ($r = 0.258$) [54]. Furthermore, recent studies have proposed that people with high short-term BPV are at high risk for hypertension and should be closely monitored [55]. Therefore, it is also necessary to explore the factors influencing BPV. However, there is no generally accepted gold standard index for blood pressure variability until now. In our study, we chose SD, CV and ARV as our BPV indicators. SD is the most commonly used indicator to assess blood pressure variability, which is able to reflect the dispersion of original BP readings [56]. The disadvantage is that it depends on the average level of BP and is susceptible to short-term blood pressure fluctuations [51]. However, CV is not affected by the mean level of blood pressure than SD and is suitable for comparison with different mean values [56]. When assessing 24 h blood pressure variability, the traditionally calculated 24 h SD and CV would be affected by circadian blood pressure variability, so the weighted 24 h SD and CV are considered better indicators [57]. Furthermore, ARV was proposed as a more reliable indicator for assessing BP variability, which was calculated as the mean of absolute differences over 24 h between consecutive BP measurements [51]. Not only is it not affected by the mean blood pressure level, but it also represents time series variability [58]. To eliminate the effect of the mean on the results, we selected the CV and ARV as the dependent variable for multiple linear regression analysis. With gender, age, BMI, bedtime, napping duration, and nighttime sleep time as independent variables, our study found that daytime napping was independently associated with nighttime CV of SBP, daytime CV of DBP, 24h WCV of DBP,24 h ARV of SBP, 24 ARV of DBP. Notably, in the regression model of 24 ARV of DBP, we found that males with a high BMI and long nap had higher ARV. For diastolic or systolic variability, it is inconclusive which is more valuable for predicting poor health outcomes. Systolic variability has received more attention than diastolic variability in studies. Higher systolic variability has been associated with increased risk of all-cause mortality, dementia, coronary heart diseases, stroke, and renal diseases [59, 60]. In studies of patients receiving intravenous thrombolysis (IVT) for acute ischemic stroke, higher 72 h BPV between prior IVT to 72 h after IVT had an increased risk of stroke outcome within 3 months, SBP (OR = 5.298, $95\%$ CI, 1.339–10.968) and DBP (OR = 6.397, $95\%$ CI, 1.576–25.958) [61]. However, it has been recommended that DBP may predict cardiovascular diseases risk in young adults to a greater extent than SBP [62]. Previous studies on daytime napping and blood pressure have mostly focused on elderly populations, and few studies have been reported on younger people. Overall, we obtained more positive results related to diastolic BPV (including daytime CV of DBP, daytime SD of DBP, etc.), especially when comparing different napping time groups. Diastolic variability relevance among youth remains to be further demonstrated.
## Limitations
Our findings help to elucidate the relationship between nap duration and BPV indicators, which will facilitate further in-depth studies. The present study also has some limitations. Firstly, ambulatory blood pressure measurement data were difficult to obtain, so the present study was conducted using a relatively small sample. Secondly, in the present study the sleep data is self-reported, there could be some recall bias. Future studies are recommended to assess napping characteristics more accurately by wearing a wrist actigraphy for multiple consecutive days. Also, our study did not take certain important factors into account, such as dietary information and physical activity that are related to sleep and BP. Future related studies should take these factors into consideration. Finally, this study is an observational study and can only provide correlations between sleep characteristics and BPV and cannot prove the causality.
## Conclusions
Short sleep duration and late bedtime are quite common among university students. Nap duration is independently associated with BPV among university students. Especially, those with daytime napping >60 min had a significantly higher BPV than those without daytime napping.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by the Research Ethical Review Committee Board of Hunan Normal University. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.
Written informed consent was obtained from the individual(s), and minor(s)’ legal guardian/next of kin, for the publication of any potentially identifiable images or data included in this article.
## Author contributions
All authors contributed substantially to the conception and design, reviewed the manuscript, and approved the submitted version of the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fped.2023.1062300/full#supplementary-material.
## References
1. Nardin C, Rattazzi M, Pauletto P. **Blood pressure variability and therapeutic implications in hypertension and cardiovascular diseases**. *High Blood Press Cardiovasc Prev* (2019) **26** 353-9. DOI: 10.1007/s40292-019-00339-z
2. Mena LJ, Felix VG, Melgarejo JD, Maestre GE. **24-hour blood pressure variability assessed by average real variability: a systematic review and meta-analysis**. *J Am Heart Assoc* (2017) **6** e006895. DOI: 10.1161/JAHA.117.006895
3. Shen L, Yang X, Lu W, Chen W, Ye X, Wu D. **24-hour ambulatory blood pressure alterations in patients with Parkinson’s disease**. *Brain Behav* (2022) **12** e2428. DOI: 10.1002/brb3.2428
4. Marrone O, Bonsignore MR. **Blood-pressure variability in patients with obstructive sleep apnea: current perspectives**. *Nat Sci Sleep* (2018) **10** 229-42. DOI: 10.2147/NSS.S148543
5. Rouch L, Cestac P, Sallerin B, Piccoli M, Benattar-Zibi L, Bertin P. **Visit-to-visit blood pressure variability is associated with cognitive decline and incident dementia: the S. Ages cohort**. *Hypertension* (2020) **76** 1280-8. DOI: 10.1161/HYPERTENSIONAHA.119.14553
6. Parati G, Torlasco C, Pengo M, Bilo G, Ochoa JE. **Blood pressure variability: its relevance for cardiovascular homeostasis and cardiovascular diseases**. *Hypertens Res* (2020) **43** 609-20. DOI: 10.1038/s41440-020-0421-5
7. Parati G, Ochoa JE, Lombardi C, Salvi P, Bilo G. **Assessment and interpretation of blood pressure variability in a clinical setting**. *Blood Press* (2013) **22** 345-54. DOI: 10.3109/08037051.2013.782944
8. Hershner SD, Chervin RD. **Causes and consequences of sleepiness among college students**. *Nat Sci Sleep* (2014) **6** 73-84. DOI: 10.2147/NSS.S62907
9. Li L, Wang YY, Wang SB, Li L, Lu L, Ng CH. **Sleep duration and sleep patterns in Chinese university students: a comprehensive meta-analysis**. *J Clin Sleep Med* (2017) **13** 1153-62. DOI: 10.5664/jcsm.6760
10. Cheungpasitporn W, Thongprayoon C, Srivali N, Vijayvargiya P, Andersen CA, Kittanamongkolchai W. **The effects of napping on the risk of hypertension: a systematic review and meta-analysis**. *J Evid Based Med* (2016) **9** 205-12. DOI: 10.1111/jebm.12211
11. Dashti HS, Daghlas I, Lane JM, Huang Y, Udler MS, Wang H. **Genetic determinants of daytime napping and effects on cardiometabolic health**. *Nat Commun* (2021) **12** 900. DOI: 10.1038/s41467-020-20585-3
12. Li P, Gaba A, Wong PM, Cui L, Yu L, Bennett DA. **Objective assessment of daytime napping and incident heart failure in 1140 community-dwelling older adults: a prospective, observational cohort study**. *J Am Heart Assoc* (2021) **10** e019037. DOI: 10.1161/JAHA.120.019037
13. Fu J, Zhang X, Moore JB, Wang B, Li R. **Midday nap duration and hypertension among middle-aged and older Chinese adults: a nationwide retrospective cohort study**. *Int J Environ Res Public Health* (2021) **18** 3680. DOI: 10.3390/ijerph18073680
14. Smolensky MH, Hermida RC, Castriotta RJ, Portaluppi F. **Role of sleep-wake cycle on blood pressure circadian rhythms and hypertension**. *Sleep Med* (2007) **8** 668-80. DOI: 10.1016/j.sleep.2006.11.011
15. Makarem N, Shechter A, Carnethon MR, Mullington JM, Hall MH, Abdalla M. **Sleep duration and blood pressure: recent advances and future directions**. *Curr Hypertens Rep* (2019) **21** 33. DOI: 10.1007/s11906-019-0938-7
16. Romdhani M, Rae DE, Nedelec M, Ammar A, Chtourou H, Al Horani R. **COVID-19 lockdowns: a worldwide survey of circadian rhythms and sleep quality in 3911 athletes from 49 countries, with data-driven recommendations**. *Sports Med* (2021) **52** 1433-48. DOI: 10.1007/s40279-021-01601-y
17. Liu X, Yan G, Bullock L, Barksdale DJ, Logan JG. **Sleep moderates the association between arterial stiffness and 24-hour blood pressure variability**. *Sleep Med* (2021) **83** 222-9. DOI: 10.1016/j.sleep.2021.04.027
18. Loredo JS, Nelesen R, Ancoli-Israel S, Dimsdale JE. **Sleep quality and blood pressure dipping in normal adults**. *Sleep* (2004) **27** 1097-103. DOI: 10.1093/sleep/27.6.1097
19. Culver MN, McMillan NK, Cross BL, Robinson AT, Montoye AH, Riemann BL. **Sleep duration irregularity is associated with elevated blood pressure in young adults**. *Chronobiol Int* (2022) **39** 1320-8. DOI: 10.1080/07420528.2022.2101373
20. Parati G, Stergiou G, O’Brien E, Asmar R, Beilin L, Bilo G. **European society of hypertension practice guidelines for ambulatory blood pressure monitoring**. *J Hypertens* (2014) **32** 1359-66. DOI: 10.1097/HJH.0000000000000221
21. Chen X, Zhu Y, Geng S, Li Q, Jiang H. **Association of blood pressure variability and intima-media thickness with white matter hyperintensities in hypertensive patients**. *Front Aging Neurosci* (2019) **11** 192. DOI: 10.3389/fnagi.2019.00192
22. Korostovtseva L, Bochkarev M, Sviryaev Y. **Sleep and cardiovascular risk**. *Sleep Med Clin* (2021) **16** 485-97. DOI: 10.1016/j.jsmc.2021.05.001
23. Cheung V, Yuen VM, Wong GTC, Choi SW. **The effect of sleep deprivation and disruption on DNA damage and health of doctors**. *Anaesthesia* (2019) **74** 434-40. DOI: 10.1111/anae.14533
24. Maheshwari G, Shaukat F. **Impact of poor sleep quality on the academic performance of medical students**. *Cureus* (2019) **11** e4357. DOI: 10.7759/cureus.4357
25. El Hangouche AJ, Jniene A, Aboudrar S, Errguig L, Rkain H, Cherti M. **Relationship between poor quality sleep, excessive daytime sleepiness and low academic performance in medical students**. *Adv Med Educ Pract* (2018) **9** 631-8. DOI: 10.2147/AMEP.S162350
26. Khader WS, Tubbs AS, Haghighi A, Athey AB, Killgore WDS, Hale L. **Onset insomnia and insufficient sleep duration are associated with suicide ideation in university students and athletes**. *J Affect Disord* (2020) **274** 1161-4. DOI: 10.1016/j.jad.2020.05.102
27. Zhai S, Tao S, Wu X, Zou L, Yang Y, Xie Y. **Associations of sleep insufficiency and chronotype with inflammatory cytokines in college students**. *Nat Sci Sleep* (2021) **13** 1675-85. DOI: 10.2147/NSS.S329894
28. Hu C, Zhang Y, Wang S, Lin L, Peng K, Du R. **Association of bedtime with the risk of non-alcoholic fatty liver disease among middle-aged and elderly Chinese adults with Pre-diabetes and diabetes**. *Diabetes Metab Res Rev* (2020) **36** e3322. DOI: 10.1002/dmrr.3322
29. Su Y, Li C, Long Y, He L, Ding N. **Association between bedtime at night and systolic blood pressure in adults in nhanes**. *Front Med (Lausanne)* (2021) **8** 734791. DOI: 10.3389/fmed.2021.734791
30. Yan B, Fan Y, Zhao B, He X, Yang J, Chen C. **Association between late bedtime and diabetes Mellitus: a large community-based study**. *J Clin Sleep Med* (2019) **15** 1621-7. DOI: 10.5664/jcsm.8026
31. Tse LA, Wang C, Rangarajan S, Liu Z, Teo K, Yusufali A. **Timing and length of nocturnal sleep and daytime napping and associations with obesity types in high-, middle-, and low-income countries**. *JAMA Netw Open* (2021) **4** e2113775. DOI: 10.1001/jamanetworkopen.2021.13775
32. Lan TY, Lan TH, Wen CP, Lin YH, Chuang YL. **Nighttime sleep, Chinese afternoon nap, and mortality in the elderly**. *Sleep* (2007) **30** 1105-10. DOI: 10.1093/sleep/30.9.1105
33. Mantua J, Spencer RMC. **Exploring the nap paradox: are mid-day sleep bouts a friend or foe?**. *Sleep Med* (2017) **37** 88-97. DOI: 10.1016/j.sleep.2017.01.019
34. Leger D, Torres MJ, Bayon V, Hercberg S, Galan P, Chennaoui M. **The association between physical and mental chronic conditions and napping**. *Sci Rep* (2019) **9** 1795. DOI: 10.1038/s41598-018-37355-3
35. Ciren W, Nima Q, Li Y, He R, Suolang D, Ciren Z. **Association of daytime napping with chronic diseases among tibetan people in China: a cross-sectional study**. *BMC Public Health* (2021) **21** 1810. DOI: 10.1186/s12889-021-11871-w
36. Leng Y, Cappuccio FP, Surtees PG, Luben R, Brayne C, Khaw KT. **Daytime napping, sleep duration and increased 8-year risk of type 2 diabetes in a British population**. *Nutr Metab Cardiovasc Dis* (2016) **26** 996-1003. DOI: 10.1016/j.numecd.2016.06.006
37. Ye Y, Zhang L, Yan W, Wang A, Wang W, Gao Z. **Self-reported sleep duration and daytime napping are associated with renal hyperfiltration and microalbuminuria in an apparently healthy Chinese population**. *PLoS One* (2019) **14** e0214776. DOI: 10.1371/journal.pone.0214776
38. Kitamura K, Watanabe Y, Nakamura K, Takano C, Hayashi N, Sato H. **Short daytime napping reduces the risk of cognitive decline in community-dwelling older adults: a 5-year longitudinal study**. *BMC Geriatr* (2021) **21** 474. DOI: 10.1186/s12877-021-02418-0
39. Souabni M, Hammouda O, Romdhani M, Trabelsi K, Ammar A, Driss T. **Benefits of daytime napping opportunity on physical and cognitive performances in physically active participants: a systematic review**. *Sports Med* (2021) **51** 2115-46. DOI: 10.1007/s40279-021-01482-1
40. Pan Z, Huang M, Huang J, Yao Z, Lin Z. **Association of napping and all-cause mortality and incident cardiovascular diseases: a dose-response meta analysis of cohort studies**. *Sleep Med* (2020) **74** 165-72. DOI: 10.1016/j.sleep.2020.08.009
41. Cao Z, Shen L, Wu J, Yang H, Fang W, Chen W. **The effects of midday nap duration on the risk of hypertension in a middle-aged and older Chinese population: a preliminary evidence from the tongji-dongfeng cohort study, China**. *J Hypertens* (2014) **32** 1993-8. DOI: 10.1097/HJH.0000000000000291
42. Wang L, Wang K, Liu LJ, Zhang YY, Shu HN, Wang K. **Associations of daytime napping with incident cardiovascular diseases and hypertension in Chinese adults: a nationwide cohort study**. *Biomed Environ Sci* (2022) **35** 22-34. DOI: 10.3967/bes2022.004
43. Huang M, Yang Y, Huang Z, Yuan H, Lu Y. **The association of nighttime sleep duration and daytime napping duration with hypertension in Chinese rural areas: a population-based study**. *J Hum Hypertens* (2021) **35** 896-902. DOI: 10.1038/s41371-020-00419-x
44. Goldman SE, Hall M, Boudreau R, Matthews KA, Cauley JA, Ancoli-Israel S. **Association between nighttime sleep and napping in older adults**. *Sleep* (2008) **31** 733-40. DOI: 10.1093/sleep/31.5.733
45. Dettoni JL, Consolim-Colombo FM, Drager LF, Rubira MC, Souza SB, Irigoyen MC. **Cardiovascular effects of partial sleep deprivation in healthy volunteers**. *J Appl Physiol (1985)* (2012) **113** 232-6. DOI: 10.1152/japplphysiol.01604.2011
46. Perry JC, Bergamaschi CT, Campos RR, Silva AM, Tufik S. **Interconnectivity of sympathetic and sleep networks is mediated through reduction of gamma aminobutyric acidergic inhibition in the paraventricular nucleus**. *J Sleep Res* (2014) **23** 168-75. DOI: 10.1111/jsr.12110
47. Kario K, Hettrick DA, Prejbisz A, Januszewicz A. **Obstructive sleep apnea-induced neurogenic nocturnal hypertension: a potential role of renal denervation?**. *Hypertension* (2021) **77** 1047-60. DOI: 10.1161/HYPERTENSIONAHA.120.16378
48. Lockley SW, Skene DJ, Tabandeh H, Bird AC, Defrance R, Arendt J. **Relationship between napping and melatonin in the blind**. *J Biol Rhythms* (1997) **12** 16-25. DOI: 10.1177/074873049701200104
49. Tanabe N, Iso H, Seki N, Suzuki H, Yatsuya H, Toyoshima H. **Daytime napping and mortality, with a special reference to cardiovascular disease: the jacc study**. *Int J Epidemiol* (2010) **39** 233-43. DOI: 10.1093/ije/dyp327
50. Stergiou GS, Mastorantonakis SE, Roussias LG. **Intraindividual reproducibility of blood pressure surge upon rising after nighttime sleep and siesta**. *Hypertens Res* (2008) **31** 1859-64. DOI: 10.1291/hypres.31.1859
51. Parati G, Ochoa JE, Lombardi C, Bilo G. **Blood pressure variability: assessment, predictive value, and potential as a therapeutic target**. *Curr Hypertens Rep* (2015) **17** 537. DOI: 10.1007/s11906-015-0537-1
52. Kai H, Kudo H, Takayama N, Yasuoka S, Aoki Y, Imaizumi T. **Molecular mechanism of aggravation of hypertensive organ damages by short-term blood pressure variability**. *Curr Hypertens Rev* (2014) **10** 125-33. DOI: 10.2174/1573402111666141217112655
53. Tadic M, Cuspidi C, Suzic Lazic J, Vukomanovic V, Mihajlovic S, Savic P. **Blood pressure variability correlates with right ventricular strain in women with gestational hypertension and preeclampsia**. *J Hum Hypertens* (2022) **36** 826-32. DOI: 10.1038/s41371-021-00580-x
54. Ciobanu DM, Bala C, Rusu A, Cismaru G, Roman G. **E-selectin is associated with daytime and 24-hour diastolic blood pressure variability in type 2 diabetes**. *Biomedicines* (2022) **10** 279. DOI: 10.3390/biomedicines10020279
55. Ozkan G, Ulusoy S, Arici M, Derici U, Akpolat T, Sengul S. **Does blood pressure variability affect hypertension development in prehypertensive patients?**. *Am J Hypertens* (2022) **35** 73-8. DOI: 10.1093/ajh/hpab125
56. Schutte AE, Kollias A, Stergiou GS. **Blood pressure and its variability: classic and novel measurement techniques**. *Nat Rev Cardiol* (2022) **19** 643-54. DOI: 10.1038/s41569-022-00690-0
57. Bilo G, Giglio A, Styczkiewicz K, Caldara G, Maronati A, Kawecka-Jaszcz K. **A new method for assessing 24-H blood pressure variability after excluding the contribution of nocturnal blood pressure fall**. *J Hypertens* (2007) **25** 2058-66. DOI: 10.1097/HJH.0b013e32829c6a60
58. Mena L, Pintos S, Queipo NV, Aizpurua JA, Maestre G, Sulbaran T. **A reliable index for the prognostic significance of blood pressure variability**. *J Hypertens* (2005) **23** 505-11. DOI: 10.1097/01.hjh.0000160205.81652.5a
59. Gosmanova EO, Mikkelsen MK, Molnar MZ, Lu JL, Yessayan LT, Kalantar-Zadeh K. **Association of systolic blood pressure variability with mortality, coronary heart disease, stroke, and renal disease**. *J Am Coll Cardiol* (2016) **68** 1375-86. DOI: 10.1016/j.jacc.2016.06.054
60. Chiu TJ, Yeh JT, Huang CJ, Chiang CE, Sung SH, Chen CH. **Blood pressure variability and cognitive dysfunction: a systematic review and meta-analysis of longitudinal cohort studies**. *J Clin Hypertens (Greenwich)* (2021) **23** 1463-82. DOI: 10.1111/jch.14310
61. He M, Wang H, Tang Y, Wang J, Cui B, Xu B. **Blood pressure undulation of peripheral thrombolysis period in acute ischemic stroke is associated with prognosis**. *J Hypertens* (2022) **40** 749-57. DOI: 10.1097/HJH.0000000000003070
62. Attard SM, Herring AH, Zhang B, Du S, Popkin BM, Gordon-Larsen P. **Associations between age, cohort, and urbanization with sbp and dbp in China: a population-based study across 18 years**. *J Hypertens* (2015) **33** 948-56. DOI: 10.1097/HJH.0000000000000522
|
---
title: Association of Area Deprivation With Primary Hypertension Diagnosis Among Youth
Medicaid Recipients in Delaware
authors:
- Carissa M. Baker-Smith
- Wei Yang
- Mary J. McDuffie
- Erin P. Nescott
- Bethany J. Wolf
- Cathy H. Wu
- Zugui Zhang
- Robert E. Akins
journal: JAMA Network Open
year: 2023
pmcid: PMC10018318
doi: 10.1001/jamanetworkopen.2023.3012
license: CC BY 4.0
---
# Association of Area Deprivation With Primary Hypertension Diagnosis Among Youth Medicaid Recipients in Delaware
## Key Points
### Question
Is there an association between neighborhood measures of deprivation and hypertension diagnosis in youth?
### Findings
In this cross-sectional study of 65 452 Delaware Medicaid-insured youths aged 8 to 18 years between 2014 and 2019, residence in neighborhoods with a higher area deprivation index was associated with primary hypertension diagnosis.
### Meaning
These findings suggest that there is an association between greater neighborhood deprivation and a diagnosis of primary hypertension in youths, which may be an important factor to consider in assessing the presence and prevalence of hypertension in youths.
## Abstract
This cross-sectional study assesses the association between neighborhood measures of deprivation and diagnosis of primary hypertension in Medicaid-insured children and adolescents in Delaware.
### Importance
The association between degree of neighborhood deprivation and primary hypertension diagnosis in youth remains understudied.
### Objective
To assess the association between neighborhood measures of deprivation and primary hypertension diagnosis in youth.
### Design, Setting, and Participants
This cross-sectional study included 65 452 Delaware Medicaid-insured youths aged 8 to 18 years between January 1, 2014, and December 31, 2019. Residence was geocoded by national area deprivation index (ADI).
### Exposures
Higher area deprivation.
### Main Outcomes and Measures
The main outcome was primary hypertension diagnosis based on International Classification of Diseases, Ninth Revision and Tenth Revision codes. Data were analyzed between September 1, 2021, and December 31, 2022.
### Results
A total of 65 452 youths were included in the analysis, including 64 307 ($98.3\%$) without a hypertension diagnosis (30 491 [$47\%$] female and 33 813 [$53\%$] male; mean [SD] age, 12.5 (3.1) years; 12 500 [$19\%$] Hispanic, 25 473 [$40\%$] non-Hispanic Black, 24 565 [$38\%$] non-Hispanic White, and 1769 [$3\%$] other race or ethnicity; 13 029 [$20\%$] with obesity; and 31 548 [$49\%$] with an ADI ≥50) and 1145 ($1.7\%$) with a diagnosis of primary hypertension (mean [SD] age, 13.3 [2.8] years; 464 [$41\%$] female and 681 [$59\%$] male; 271 [$24\%$] Hispanic, 460 [$40\%$] non-Hispanic Black, 396 [$35\%$] non-Hispanic White, and 18 [$2\%$] of other race or ethnicity; 705 [$62\%$] with obesity; and 614 [$54\%$] with an ADI ≥50). The mean (SD) duration of full Medicaid benefit coverage was 61 [16] months for those with a diagnosis of primary hypertension and 46.0 (24.3) months for those without. By multivariable logistic regression, residence within communities with ADI greater than or equal to 50 was associated with $60\%$ greater odds of a hypertension diagnosis (odds ratio [OR], 1.61; $95\%$ CI 1.04-2.51). Older age (OR per year, 1.16; $95\%$, CI, 1.14-1.18), an obesity diagnosis (OR, 5.16; $95\%$ CI, 4.54-5.85), and longer duration of full Medicaid benefit coverage (OR, 1.03; $95\%$ CI, 1.03-1.04) were associated with greater odds of primary hypertension diagnosis, whereas female sex was associated with lower odds (OR, 0.68; $95\%$, 0.61-0.77). Model fit including a Medicaid-by-ADI interaction term was significant for the interaction and revealed slightly greater odds of hypertension diagnosis for youths with ADI less than 50 (OR, 1.03; $95\%$ CI, 1.03-1.04) vs ADI ≥50 (OR, 1.02; $95\%$ CI, 1.02-1.03). Race and ethnicity were not associated with primary hypertension diagnosis.
### Conclusions and Relevance
In this cross-sectional study, higher childhood neighborhood ADI, obesity, age, sex, and duration of Medicaid benefit coverage were associated with a primary hypertension diagnosis in youth. Screening algorithms and national guidelines may consider the importance of ADI when assessing for the presence and prevalence of primary hypertension in youth.
## Introduction
It is estimated that hypertension is prevalent in 4 out of every 100 youth1,2,3 but underdiagnosed in $74\%$ of cases.3,4 Hypertension can begin in childhood and is a predominant cause of target organ damage in childhood5,6 and cardiovascular disease (CVD) in adulthood.7 A diagnosis of hypertension is made in individuals younger than 13 years when the average blood pressure (BP) across 2 to 3 visits equals or exceeds the 95th percentile; in individuals 13 years of age or older, hypertension is diagnosed when the average BP equals or exceeds $\frac{130}{80}$ mm Hg.8 Knowledge of factors associated with hypertension diagnosis in children and adolescents is essential to improving long-term cardiovascular outcomes.7,9 Previously published studies have highlighted a multitude of risk factors associated with primary hypertension development, including prenatal exposures,10,11 race,12,13 ethnicity,12,13 physical inactivity,14 increased sodium intake,15 lack of sufficient green space,16 environmental chemical exposures,17 ambient temperature,17 exposure to violence18 and crime,19,20 stress,21 low family income-to-poverty ratio,12 obesity,22 and lack of insurance.23 Studies outside the US have identified associations between parental education, occupation, employment, crowding, and home ownership and hypertension development in youth.24 However, to date, no US studies have evaluated the association between area deprivation index (ADI), an index of neighborhood-level socioeconomic factors, and primary hypertension diagnosis among insured US children and adolescents.
Physician recognition of hypertension is poor, even when BP values are consistent with the diagnosis.3 Current factors known to be associated with greater physician recognition of hypertension in youth include older age, male sex, higher BP, and obesity.3 However, the association between a child’s neighborhood-level deprivation and hypertension diagnosis has not been previously explored.
In this study, we assess the association between ADI and primary hypertension diagnosis in youth. We hypothesize that a higher ADI, even among Medicaid-insured youths, is associated with a greater likelihood of primary hypertension diagnosis.
## Methods
We performed a cross-sectional analysis of data from Delaware Medicaid recipients collected between 2014 and 2019. The Nemours Children’s Health and University of Delaware institutional review boards approved the study activities. A consent waiver was granted given that patients could not be identified directly or through identifiers. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
## Cohort
We analyzed Delaware Medicaid administrative data, including medical claims, eligibility, and client data. The Delaware Division of Medicaid and Medical Assistance provides weekly data updates to the University of Delaware Center for Community Research and Service (CCRS). The CCRS accesses data for research and uses internal procedures to safeguard the data. For this study, CCRS policy scientists (M.J.M. and E.P.N.,) coded and provided aggregate data to the primary investigator (C.M.B.-S.).
## Study Population
Age of the patients included age at the end of the Medicaid enrollment year. We selected youths aged 8 to 18 years, similarly to a previously published study,12 due to reports of higher prevalence of hypertension among youths aged 8 to 9 years2 and to capture key developmental points for the diagnosis of hypertension from midchildhood through adolescence. We included youths with at least 1 health care visit (inpatient, outpatient, emergency department, long-term care, dental, or home health) and at least 1 month of full Medicaid insurance coverage between January 1, 2014, and December 31, 2019. We excluded pregnant youths, using approximately 388 pregnancy codes specified by the Office of Population Affairs.25 Of 80 414 Delaware Medicaid-insured youths aged 8 to 18 years, 69 139 ($85.6\%$) had geocoded addresses. Geocode failures for 11 275 were secondary to post office box addresses, missing addresses, or absence of an available geocode. A total of 970 youths were removed because of ADI suppression due to high group quarters, low population, or low housing communities. Of the remaining youths, 2717 were removed because of a pregnancy diagnosis. A total of 65 452 youths were included in the final analysis (Figure 1). Mean (SD) duration of full Medicaid benefit coverage was 46.0 (24.3) months. Mean (SD) ADI was 50 [19]. A total of 32 162 ($49\%$) had an ADI greater than or equal to 50, and 33 290 ($51\%$) had an ADI less than 50.
**Figure 1.:** *Delaware Medicaid-Insured Youths Aged 8 to 18 Years, 2014-2019ADI indicates area of deprivation index; ADI suppression, geocoded addresses with no available ADI due to residence within high–group quarters population, low population, or low housing regions.*
## Outcomes
The primary outcome was a diagnosis of primary hypertension by International Classification of Diseases, Ninth Revision (ICD-9) and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes, excluding a secondary hypertension diagnosis. We excluded secondary diagnoses because these conditions tend be less common26 and tend to present before age 6 years.8 Excluded from the primary hypertension outcome group were youths with ICD-9 codes 402, 403, 404, and 405 and ICD-10 codes I15 and I11.0. Included in the primary hypertension outcome group were youths with ICD-9 codes 401.0, 401.1, and 401.9 and ICD-10 codes I10 and I11.9. Choosing to prioritize sample size, no more than 1 encounter with a given ICD-9 or ICD-10 code was required for inclusion. We also did not specify the type of visit (eg, outpatient, inpatient, emergency department) or clinician type.
## Independent Variables
Each patient’s primary enrollment address, at the time of their first health care visit between 2014 and 2019, was geocoded using ArcGIS, version 10.8.1 (Esri). This census block group geocode was then used to link each participant with their 2014-2019 American Community Survey national ADI value by census block group.27,28 Updated every 5 years, the ADI is a composite index of 17 variables used to describe the socioeconomic status (SES) of the community in which one lives (Box). The ADI is based on weighted US census data and is not determined by race, ethnicity, or individual SES. It provides a geospatial description of neighborhood deprivation and ranking of area deprivation at the census block group level (eg, neighborhood) from 1 to 100, where an ADI of 100 represents communities with the most deprivation.
Data for the ADI are obtained from the American Community Survey.28,29 Delaware’s median national ADI is 38 (lower quartile, 25; upper quartile, 52).30 *In this* analysis, we included a dichotomous outcome variable of ADI greater than or equal to 50 [1] vs less than 50 [0], where ADI of 50 represents the upper quartile of the state’s national ADI and mean for the sample population.
Obesity diagnosis was ascertained using ICD-9 codes 278.00 and 278.01 and ICD-10 codes E66* (obesity). All youths with obesity had a Z code for body mass index (BMI, calculated by weight in kilograms divided by height in meters squared) greater than the 95th percentile or for BMI 30 and above (Z68.30-45). Any child with a Z code for BMI greater than the 95th percentile also had an ICD-9 or ICD-10 code for obesity.
Patient demographic characteristics (sex, race and ethnicity, and age) as described in the pooled *Medicaid data* set were included in the analysis. Delaware *Medicaid data* include self-reported race. For this study, race and ethnicity were categorized according to the following groups: Hispanic, non-Hispanic Black, non-Hispanic White, and other. Due to small individual subgroup sample size, youths with race and ethnicity described as Hawaiian; Indian or Alaskan; Native American; other, not Hispanic; or Pacific Islander were categorized as other.
We included a descriptor of the duration of Medicaid benefit coverage in months. An interaction term for ADI50 by months of Medicaid coverage was included to assess for an association between ADI50 and duration of Medicaid benefit coverage given that individual SES might influence the community in which one lives (ADI) and duration of Medicaid coverage. An additional interaction between ADI50 and obesity was assessed given publications highlighting an association between neighborhood deprivation and obesity.16,31
## Statistical Analysis
Data analysis was performed between September 1, 2021, and December 31, 2022. We provided descriptive statistics for demographic variables and used the χ2 test for comparison of categorical variables and t test for comparison of means of continuous variables (2-sided null hypothesis at an a priori significance level of $P \leq .05$). Based on univariable analyses, we selected age, biological sex, race and ethnicity (Hispanic, non-Hispanic Black, and non-Hispanic White), total months receiving Medicaid, national ADI50 (1 = at or above the state’s upper quartile of 50 vs 0 = below the state’s upper quartile of 50), and obesity diagnosis to model hypertension diagnosis. We ensured model fit using goodness-of-fit tests (deviance and Pearson correlation). All statistical analyses were performed using SAS, version 9.4 software (SAS Institute Inc). Receiver operating characteristic curve analysis was performed and area under the curve calculated for the final model.
## Primary Hypertension Diagnosis
We identified 1145 ($1.7\%$) youths with a diagnosis of primary hypertension (mean [SD] age, 13.3 [2.8] years; 464 [$41\%$] female and 681 [$59\%$] male; 271 [$24\%$] Hispanic, 460 [$40\%$] non-Hispanic Black, 396 [$35\%$] non-Hispanic White, and 18 [$2\%$] of other race or ethnicity) and 64 307 ($98.3\%$) without a hypertension diagnosis (30 491 [$47\%$] female and 33 813 [$53\%$] male; mean [SD] age, 12.5 [3.1] years; 12 500 [$19\%$] Hispanic, 25 473 [$40\%$] non-Hispanic Black, 24 565 [$38\%$] non-Hispanic White, and 1769 [$3\%$] other race or ethnicity) (Table 1). Prevalence of primary hypertension was highest among youths aged 13 to 18 years. Among the youths with a primary hypertension diagnosis, 614 ($54\%$) had an ADI greater than or equal to 50. The mean (SD) duration of Medicaid coverage was 60.6 (16.0) months for youths with a diagnosis of primary hypertension vs 46.0 (24.3) months for participants without a diagnosis of primary hypertension. Youth with a primary hypertension diagnosis were more likely to have an obesity diagnosis (705 [$62\%$] vs 13 029 [$20\%$]) (Table 1).
**Table 1.**
| Unnamed: 0 | No. (%) | No. (%).1 | P value |
| --- | --- | --- | --- |
| | Hypertension diagnosis | No hypertension diagnosis | P value |
| No. of youths | 1145 (1.7) | 64 307 (98.3) | |
| Age, mean (SD), y | 13.3 (2.8) | 12.5 (3.1) | <.001 |
| ADI | | | .002 |
| ≥50 | 614 (54) | 31 548 (49) | .002 |
| <50 | 531 (46) | 32 759 (51) | .002 |
| Duration of full Medicaid beneficiary coverage, mean (SD), mo | 60.6 (16.0) | 46.0 (24.3) | <.001 |
| National ADI | 52.7 (19.2) | 50.2 (18.8) | |
| Obesity diagnosis | | | <.001 |
| Yes | 705 (62) | 13 029 (20) | <.001 |
| No | 440 (38) | 51 278 (80) | <.001 |
| Race and ethnicity | | | <.001 |
| Hispanic | 271 (24) | 12 500 (19) | <.001 |
| Non-Hispanic Black | 460 (40) | 25 473 (40) | <.001 |
| Non-Hispanic White | 396 (35) | 24 565 (38) | <.001 |
| Otherb | 18 (2) | 1769 (3) | <.001 |
| Sex | | | <.001 |
| Female | 464 (41) | 30 491 (47) | <.001 |
| Male | 681 (59) | 33 813 (53) | <.001 |
## Multivariable Logistic Regression
According to multivariable analysis, patient-level factors associated with a greater likelihood of a primary hypertension diagnosis included ADI greater than or equal to 50 (odds ratio [OR], 1.61; $95\%$ CI, 1.04-2.51; $$P \leq .03$$), older age (OR, 1.16; $95\%$ CI, 1.14-1.18 for each additional year of age; $P \leq .001$), and obesity diagnosis (OR, 5.16; $95\%$ CI, 4.54-5.85; $P \leq .001$). An interaction between ADI50 and obesity was not significant. Female sex was associated with a lower likelihood of a primary hypertension diagnosis (OR, 0.68; $95\%$ CI, 0.61-0.77; $P \leq .001$) (Table 2 and Figure 2). Further analysis of the association between duration of Medicaid coverage and hypertension diagnosis when ADI was greater than or equal to 50 vs when ADI was less than 50 revealed ORs of 1.02 ($95\%$ CI, 1.02-1.03) and 1.03 ($95\%$ CI, 1.03-1.04), respectively (eTable in Supplement 1). The area under the curve for the final model was 0.79 (eFigure in Supplement 1).
## Discussion
Cardiovascular disease remains the leading cause of death in the US.32 *Hypertension is* the leading modifiable risk factor for CVD development.7 *In this* study, we highlight a significant association between neighborhood deprivation and hypertension in youth.
Disparities in CVD prevalence plague the most vulnerable populations.33 We know and consider individual factors that influence risk for hypertension,8,22,34 yet we devise strategies for screening and diagnosis of hypertension that do not routinely consider neighborhood-level risk.3,8,33,35,36 *In this* cross-sectional study, we assessed the association between neighborhood measures of deprivation and diagnosis of primary hypertension in 65 452 Delaware youths insured by Medicaid. Consistent with our study hypothesis, we found that residence within a high-deprivation neighborhood was associated with $60\%$ greater odds of a hypertension diagnosis. The association between ADI and primary hypertension diagnosis was rivaled only by a diagnosis of obesity, which was associated with 5 times greater odds of a primary hypertension diagnosis.
Our finding of an association between neighborhood and primary hypertension in youth is supported by previously published studies assessing the association between neighborhood deprivation and hypertension diagnosis in adults, environmental factors and hypertension diagnosis,37,38,39 childhood opportunity and cardiometabolic disease in midchildhood,16 and perinatal environmental exposures and BP in early childhood.17 According to the Residential Environment and Coronary Heart Disease cohort study of 5941 participants aged 30 to 79 years recruited in 2007-2008, systolic BP increased independently and regularly with both decreasing individual education and decreasing residential neighborhood education.39 In the Prospective Epidemiological Study of Myocardial Infarction, a European study surveying 7850 men aged 50 to 60 years, decreasing neighborhood education level was also associated with increased systolic BP.38 Early-life exposures to components of the built environment, including access to natural spaces, traffic, air pollution, noise, and meteorology, may be associated with hypertension prevalence at a young age. The Human Early Life Exposome project, a *European consortium* study, identified a positive association between levels of air pollution during the prenatal and postnatal periods and BP.17 High ambient temperature exposure and noise exposure during pregnancy were also associated with higher BP in offspring. There were limitations to these findings, including that the authors could not exclude a seasonal temperature effect and were unable to identify a noise dose-response trend.
The Child Opportunity Index combines 29 neighborhood-level indicators into a single composite measure to describe 3 domains of child health: education, health and environment, and social and economic. Aris et al16 geocoded the residential addresses of 743 youths and found that independent of individual and family SES, youths who resided in communities with higher overall opportunity (higher Child Opportunity Index, lower deprivation) had lower metabolic risk scores in midchildhood (7.9 years) and early adolescence (13.1 years).
The sensitive period model proposes that negative health exposures have a greater effect when they take place during sensitive developmental periods and influence later outcomes.40 Longitudinal studies of early-life exposure to low neighborhood SES and association with adult BP from the New England Family Study found that neighborhood deprivation may contribute to an increased risk for hypertension later in life.41 The study, however, did not assess the association between neighborhood SES and hypertension during childhood. Policies such as community redlining may also be adversely associated with BP and cardiovascular health status. Among adults, this process has been associated with higher rates of CVD and hypertension and poorer cardiometabolic health outcomes.42 Not all published studies, however, have identified an association between neighborhood deprivation and hypertension diagnosis in youth. Ribeiro et al24 assessed the relative index of inequality (RII), a summary measure of relative inequality that expresses the risk ratio of an outcome, comparing the outcome of the lowest social hierarchy group with the outcome of the highest social hierarchy group. In this study of 7459 youths aged 4 years living in Portugal, the RII for hypertension was greater than 1 with a $95\%$ CI that did not include 1 when the SES indicator was parental (mother or father) education, occupation, income, or home ownership. The European Deprivation Index (EDI) was used in this study as another indicator of SES. The EDI is a cross-culture deprivation index used in 5 European countries. In Portugal, the EDI includes 8 variables. The EDI did not provide as detailed a description of the poverty level of the neighborhood households.43 *Using this* EDI, Ribeiro et al did not identify an RII for hypertension when the socioeconomic position evaluated was the EDI.
The ADI is specific to a smaller region and describes deprivation at the census block group level. The ADI provides a much more detailed description of the SES of the neighborhood than the EDI and at a more specific level (eg, census block group) than the Child Opportunity Index. In addition, the ADI is updated every 5 years.27 *Insurance status* is an important determinant of access to care.23 Without health insurance, many persons cannot receive regular medical care, and given that the diagnosis of hypertension relies on detection of a BP above an established threshold on 2 or more separate occasions,8 hypertension diagnoses can either be delayed or missed in the absence of insurance coverage. In this study of 65 452 Delaware Medicaid recipients aged 8 to 18 years, we assessed whether, beyond access to health insurance, which all the youths had, there was an association between duration of Medicaid insurance coverage and odds of a diagnosis of hypertension. We found greater odds of a primary hypertension diagnosis for each additional month of Medicaid insurance coverage that was slightly different in ADI less than 50 (OR, 1.033; $95\%$ CI 1.027-1.038; $P \leq .001$) vs ADI greater than or equal to 50 (OR, 1.024; $95\%$ CI, 1.019-1.029; $P \leq .001$) (eTable in Supplement 1). Without further data, we are unable to determine the exact reason for this difference.
Racial and ethnic disparities in hypertension prevalence have been well described, particularly in adults.7,32 Hypertension prevalence has been reported as highest among non-Hispanic Black men and women ($56.5\%$ and $55.3\%$, respectively) according to 2015-2018 National Health and Nutrition Examination Survey data.7 However, studies in children and adolescents have highlighted that racial and ethnic differences in hypertension prevalence are no longer significant when adjusted for BMI.2 Older studies, however, have suggested that even after adjusting for BMI, Hispanic males have higher odds of hypertension vs White males (OR, 1.21; $95\%$ CI, 1.07-1.37; $P \leq .01$).44 We found no association between race and ethnicity and hypertension diagnosis.
## Limitations
This study has several limitations. First, we used an administrative database. Although the sample was large and diagnoses were available, BP and BMI values were not available. We know that $74\%$ of cases of pediatric hypertension are underdiagnosed.3,4 Therefore, although our estimated prevalence of primary hypertension was similar to that of previously published reports,2,3 there may have been an underestimation of the true prevalence of hypertension in this population. Second, selection bias may have occurred, as we were forced to exclude 11 275 youths due to absence of geocoding. Nevertheless, our findings regarding the association between ADI and primary hypertension may remain valid given that we do not suspect a higher proportion of screen failures in more affluent communities and given that many of the youths with geocode failures may have resided within high-deprivation neighborhoods. Third, we excluded 2717 youths because of a pregnancy diagnosis. This exclusion also may not have contributed to selection bias as the exclusion was uniform and represented a low percentage of the total sample. Fourth, we assumed that youths had the same ADI throughout the 5-year study period. We had no reason to suspect that a significant number of youths changed addresses from a low- to a high- or a high- to a low-deprivation neighborhood during the 5-year study. Fifth, while we used very simplified categories for race and ethnicity, the distinctions used are those commonly reported in the literature. Future studies will more fully explore the association between distinct racial and ethnic groups and diagnosis of primary hypertension.
## Conclusions
The findings from this cross-sectional study highlight the significant association between neighborhood deprivation and a diagnosis of primary hypertension among Medicaid-insured youths and the importance of considering neighborhood-related factors, such as ADI, when diagnosing hypertension. Future studies are needed to further elucidate the association between the 17 components of ADI and hypertension development and diagnosis in youth. Screening algorithms and national guidelines may consider the importance of ADI when assessing for the presence and prevalence of primary hypertension in youth.
## References
1. Song P, Zhang Y, Yu J. **Global prevalence of hypertension in children: a systematic review and meta-analysis**. *JAMA Pediatr* (2019.0) **173** 1154-1163. DOI: 10.1001/jamapediatrics.2019.3310
2. Goulding M, Goldberg R, Lemon SC. **Differences in blood pressure levels among children by sociodemographic status**. *Prev Chronic Dis* (2021.0) **18**. DOI: 10.5888/pcd18.210058
3. Hansen ML, Gunn PW, Kaelber DC. **Underdiagnosis of hypertension in children and adolescents**. *JAMA* (2007.0) **298** 874-879. DOI: 10.1001/jama.298.8.874
4. Kaelber DC, Liu W, Ross M. **Diagnosis and medication treatment of pediatric hypertension: a retrospective cohort study**. *Pediatrics* (2016.0) **138**. DOI: 10.1542/peds.2016-2195
5. Urbina EM, Lande MB, Hooper SR, Daniels SR. **Target organ abnormalities in pediatric hypertension**. *J Pediatr* (2018.0) **202** 14-22. DOI: 10.1016/j.jpeds.2018.07.026
6. Haley JE, Woodly SA, Daniels SR. **Association of blood pressure-related increase in vascular stiffness on other measures of target organ damage in youth**. *Hypertension* (2022.0) **79** 2042-2050. DOI: 10.1161/HYPERTENSIONAHA.121.18765
7. Virani SS, Alonso A, Aparicio HJ. **Heart disease and stroke statistics-2021 update: a report from the American Heart Association**. *Circulation* (2021.0) **143** e254-e743. DOI: 10.1161/CIR.0000000000000950
8. Flynn JT, Kaelber DC, Baker-Smith CM. **Clinical practice guideline for screening and management of high blood pressure in children and adolescents**. *Pediatrics* (2017.0) **140**. DOI: 10.1542/peds.2017-1904
9. Theodore RF, Broadbent J, Nagin D. **Childhood to early-midlife systolic blood pressure trajectories: early-life predictors, effect modifiers, and adult cardiovascular outcomes**. *Hypertension* (2015.0) **66** 1108-1115. DOI: 10.1161/HYPERTENSIONAHA.115.05831
10. Warembourg C, Maitre L, Tamayo-Uria I. **Early-life environmental exposures and blood pressure in children**. *J Am Coll Cardiol* (2019.0) **74** 1317-1328. DOI: 10.1016/j.jacc.2019.06.069
11. Fraser A, Nelson SM, Macdonald-Wallis C, Sattar N, Lawlor DA. **Hypertensive disorders of pregnancy and cardiometabolic health in adolescent offspring**. *Hypertension* (2013.0) **62** 614-620. DOI: 10.1161/HYPERTENSIONAHA.113.01513
12. Hardy ST, Sakhuja S, Jaeger BC. **Trends in blood pressure and hypertension among US children and adolescents, 1999-2018**. *JAMA Netw Open* (2021.0) **4**. DOI: 10.1001/jamanetworkopen.2021.3917
13. Lo JC, Sinaiko A, Chandra M. **Prehypertension and hypertension in community-based pediatric practice**. *Pediatrics* (2013.0) **131** e415-e424. DOI: 10.1542/peds.2012-1292
14. Aguilar-Cordero MJ, Rodríguez-Blanque R, Leon-Ríos X, Expósito Ruiz M, García García I, Sánchez-López AM. **Influence of physical activity on blood pressure in children with overweight/obesity: a randomized clinical trial**. *Am J Hypertens* (2020.0) **33** 131-136. DOI: 10.1093/ajh/hpz174
15. He FJ, MacGregor GA. **Importance of salt in determining blood pressure in children: meta-analysis of controlled trials**. *Hypertension* (2006.0) **48** 861-869. DOI: 10.1161/01.HYP.0000245672.27270.4a
16. Aris IM, Rifas-Shiman SL, Jimenez MP. **Neighborhood Child Opportunity Index and adolescent cardiometabolic risk**. *Pediatrics* (2021.0) **147**. DOI: 10.1542/peds.2020-018903
17. Warembourg C, Nieuwenhuijsen M, Ballester F. **Urban environment during early-life and blood pressure in young children**. *Environ Int* (2021.0) **146**. DOI: 10.1016/j.envint.2020.106174
18. Kapur G, Stenson AF, Chiodo LM. **Childhood violence exposure predicts high blood pressure in Black American young adults**. *J Pediatr* (2022.0) **248** 21-29.e1. DOI: 10.1016/j.jpeds.2022.05.039
19. Gooding HC, Milliren CE, Austin SB, Sheridan MA, McLaughlin KA. **Child abuse, resting blood pressure, and blood pressure reactivity to psychosocial stress**. *J Pediatr Psychol* (2016.0) **41** 5-14. DOI: 10.1093/jpepsy/jsv040
20. Tung EL, Wroblewski KE, Boyd K, Makelarski JA, Peek ME, Lindau ST. **Police-recorded crime and disparities in obesity and blood pressure status in Chicago**. *J Am Heart Assoc* (2018.0) **7**. DOI: 10.1161/JAHA.117.008030
21. Suglia SF, Koenen KC, Boynton-Jarrett R. **Childhood and adolescent adversity and cardiometabolic outcomes: a scientific statement from the American Heart Association**. *Circulation* (2018.0) **137** e15-e28. DOI: 10.1161/CIR.0000000000000536
22. Zhao W, Mo L, Pang Y. **Hypertension in adolescents: the role of obesity and family history**. *J Clin Hypertens (Greenwich)* (2021.0) **23** 2065-2070. DOI: 10.1111/jch.14381
23. Huguet N, Larson A, Angier H. **Rates of undiagnosed hypertension and diagnosed hypertension without anti-hypertensive medication following the Affordable Care Act**. *Am J Hypertens* (2021.0) **34** 989-998. DOI: 10.1093/ajh/hpab069
24. Ribeiro AI, Fraga S, Correia-Costa L, McCrory C, Barros H. **Socioeconomic disadvantage and health in early childhood: a population-based birth cohort study from Portugal**. *Pediatr Res* (2020.0) **88** 503-511. DOI: 10.1038/s41390-020-0786-9
25. 25Contraceptive provision measures. Office of Population Affairs. Accessed February 13, 2023. https://opa.hhs.gov/claims-data-sas-program-instructions. *Contraceptive provision measures*
26. Nugent JT, Young C, Funaro MC. **Prevalence of secondary hypertension in otherwise healthy youths with a new diagnosis of hypertension: a meta-analysis**. *J Pediatr* (2022.0) **244** 30-37.e10. DOI: 10.1016/j.jpeds.2022.01.047
27. Kind AJH, Buckingham WR. **Making neighborhood-disadvantage metrics accessible—the Neighborhood Atlas**. *N Engl J Med* (2018.0) **378** 2456-2458. DOI: 10.1056/NEJMp1802313
28. Singh GK. **Area deprivation and widening inequalities in US mortality, 1969-1998**. *Am J Public Health* (2003.0) **93** 1137-1143. DOI: 10.2105/AJPH.93.7.1137
29. 29About the Neighborhood Atlas. Center for Health Disparities Research, University of Wisconsin School of Medicine and Public Health. Accessed January 1, 2021. https://www.neighborhoodatlas.medicine.wisc.edu/. *About the Neighborhood Atlas*
30. 30Area Deprivation Index (ADI). Delaware.gov. Accessed February 13, 2023. https://myhealthycommunity.dhss.delaware.gov/data-dictionary/dataset-metadata/ADI. *Area Deprivation Index (ADI)*
31. Nguyen TH, Götz S, Kreffter K, Lisak-Wahl S, Dragano N, Weyers S. **Neighbourhood deprivation and obesity among 5656 pre-school children-findings from mandatory school enrollment examinations**. *Eur J Pediatr* (2021.0) **180** 1947-1954. DOI: 10.1007/s00431-021-03988-2
32. Tsao CW, Aday AW, Almarzooq ZI. **Heart disease and stroke statistics—2022 update: a report from the American Heart Association**. *Circulation* (2022.0) **145** e153-e639. DOI: 10.1161/CIR.0000000000001052
33. Fuller-Rowell TE, Curtis DS, Klebanov PK, Brooks-Gunn J, Evans GW. **Racial disparities in blood pressure trajectories of preterm children: the role of family and neighborhood socioeconomic status**. *Am J Epidemiol* (2017.0) **185** 888-897. DOI: 10.1093/aje/kww198
34. Falkner B. **Recent clinical and translational advances in pediatric hypertension**. *Hypertension* (2015.0) **65** 926-931. DOI: 10.1161/HYPERTENSIONAHA.114.03586
35. Whelton PK, Carey RM, Aronow WS. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *J Am Coll Cardiol* (2018.0) **71** e127-e248. DOI: 10.1016/j.jacc.2017.11.006
36. Chakraborty S, Brijnath B, Dermentzis J, Mazza D. **Defining key questions for clinical practice guidelines: a novel approach for developing clinically relevant questions**. *Health Res Policy Syst* (2020.0) **18** 113. DOI: 10.1186/s12961-020-00628-3
37. Brook RD, Weder AB, Rajagopalan S. **“Environmental hypertensionology” the effects of environmental factors on blood pressure in clinical practice and research**. *J Clin Hypertens (Greenwich)* (2011.0) **13** 836-842. DOI: 10.1111/j.1751-7176.2011.00543.x
38. Chaix B, Ducimetière P, Lang T. **Residential environment and blood pressure in the PRIME study: is the association mediated by body mass index and waist circumference?**. *J Hypertens* (2008.0) **26** 1078-1084. DOI: 10.1097/HJH.0b013e3282fd991f
39. Chaix B, Bean K, Leal C. **Individual/neighborhood social factors and blood pressure in the RECORD cohort study: which risk factors explain the associations?**. *Hypertension* (2010.0) **55** 769-775. DOI: 10.1161/HYPERTENSIONAHA.109.143206
40. Ben-Shlomo Y, Kuh D. **A life course approach to chronic disease epidemiology: conceptual models, empirical challenges and interdisciplinary perspectives**. *Int J Epidemiol* (2002.0) **31** 285-293. DOI: 10.1093/ije/31.2.285
41. Jimenez MP, Wellenius GA, Subramanian SV. **Longitudinal associations of neighborhood socioeconomic status with cardiovascular risk factors: a 46-year follow-up study**. *Soc Sci Med* (2019.0) **241**. DOI: 10.1016/j.socscimed.2019.112574
42. Motairek I, Lee EK, Janus S. **Historical neighborhood redlining and contemporary cardiometabolic risk**. *J Am Coll Cardiol* (2022.0) **80** 171-175. DOI: 10.1016/j.jacc.2022.05.010
43. Guillaume E, Pornet C, Dejardin O. **Development of a cross-cultural deprivation index in five European countries**. *J Epidemiol Community Health* (2016.0) **70** 493-499. DOI: 10.1136/jech-2015-205729
44. Rosner B, Cook N, Portman R, Daniels S, Falkner B. **Blood pressure differences by ethnic group among United States children and adolescents**. *Hypertension* (2009.0) **54** 502-508. DOI: 10.1161/HYPERTENSIONAHA.109.134049
|
---
title: Effect of Sleep Changes on Health-Related Quality of Life in Healthy Children
authors:
- Rachael W. Taylor
- Jillian J. Haszard
- Rosie Jackson
- Silke Morrison
- Dean W. Beebe
- Kim A. Meredith-Jones
- Dawn E. Elder
- Barbara C. Galland
journal: JAMA Network Open
year: 2023
pmcid: PMC10018327
doi: 10.1001/jamanetworkopen.2023.3005
license: CC BY 4.0
---
# Effect of Sleep Changes on Health-Related Quality of Life in Healthy Children
## Abstract
This secondary analysis examines the self-rated aspects of health-related quality of life, including physical and psychological well-being, social and peer support, and coping at school, among healthy children who were enrolled in a sleep manipulation trial.
## Key Points
### Question
Does losing sleep affect health-related quality of life in children?
### Findings
In this secondary analysis of a randomized crossover trial involving 100 healthy children aged 8 to 12 years, receiving 39 minutes less sleep per night resulted in significantly lower physical and overall well-being, ability to cope well at school, and total health-related quality of life, especially in children with at least a 30-minute difference in sleep.
### Meaning
Findings of this secondary analysis of a randomized clinical trial indicate that ensuring children receive sufficient good-quality sleep is an important child health issue.
### Importance
Little is known regarding the effect of poor sleep on health-related quality of life (HRQOL) in healthy children.
### Objective
To determine the effect of induced mild sleep deprivation on HRQOL in children without major sleep issues.
### Design, Setting, and Participants
This prespecified secondary analysis focused on HRQOL, a secondary outcome of the Daily Rest, Eating, and Activity Monitoring (DREAM) randomized crossover trial of children who underwent alternating weeks of sleep restriction and sleep extension and a 1-week washout in between. The DREAM trial intervention was administered at participants’ homes between October 2018 and March 2020. Participants were 100 children aged 8 to 12 years who lived in Dunedin, New Zealand; had no underlying medical conditions; and had parent- or guardian-reported normal sleep (8-11 hours/night). Data were analyzed between July 4 and September 1, 2022.
### Interventions
Bedtimes were manipulated to be 1 hour later (sleep restriction) and 1 hour earlier (sleep extension) than usual for 1 week each. Wake times were unchanged.
### Main Outcomes and Measures
All outcome measures were assessed during both intervention weeks. Sleep timing and duration were assessed using 7-night actigraphy. Children and parents rated the child’s sleep disturbances (night) and impairment (day) using the 8-item Pediatric Sleep Disturbance and 8-item Sleep-Related Impairment scales of the Patient-Reported Outcomes Measurement Information System questionnaire. Child-reported HRQOL was assessed using the 27-item KIDSCREEN questionnaire with 5 subscale scores and a total score. Both questionnaires assessed the past 7 days at the end of each intervention week. Data were presented as mean differences and $95\%$ CIs between the sleep restriction and extension weeks and were analyzed using intention to treat and an a priori difference in sleep of at least 30 minutes per night.
### Results
The final sample comprised 100 children (52 girls [$52\%$]; mean [SD] age, 10.3 [1.4] years). During the sleep restriction week, children went to sleep 64 ($95\%$ CI, 58-70) minutes later, and sleep offset (wake time) was 18 ($95\%$ CI, 13-24) minutes later, meaning that children received 39 ($95\%$ CI, 32-46) minutes less of total sleep per night compared with the sleep extension week in which the total sleep time was 71 ($95\%$ CI, 64-78) minutes less in the per-protocol sample analysis. Both parents and children reported significantly less sleep disturbance at night but greater sleep impairment during the day with sleep restriction. Significant standardized reductions in physical well-being (standardized mean difference [SMD], −0.28; $95\%$ CI, −0.49 to −0.08), coping in a school environment (SMD, −0.26; $95\%$ CI, −0.42 to −0.09), and total HRQOL score (SMD, −0.21; $95\%$ CI, −0.34 to −0.08) were reported by children during sleep restriction, with an additional reduction in social and peer support (SMD, −0.24; $95\%$ CI, −0.47 to −0.01) in the per-protocol sample analysis.
### Conclusions and Relevance
Results of this secondary analysis of the DREAM trial indicated that even 39 minutes less of sleep per night for 1 week significantly reduced several facets of HRQOL in children. This finding shows that ensuring children receive sufficient good-quality sleep is an important child health issue.
### Trial Registration
Australian New Zealand Clinical Trials Registry: ACTRN12618001671257
## Introduction
While inadequate or poor-quality sleep has been associated with a wide range of adverse physical and psychological health outcomes in infants, children, and adolescents,1,2,3,4,5 interest is growing regarding the association of sleep with more global indices of health, such as health-related quality of life (HRQOL).6 The HRQOL is a widely used concept with many definitions and measures7 and generally encompasses 3 main domains: physical, mental, and social health.8 To date, most research has examined the association of more medically related sleep issues, such as obstructive sleep apnea,9 insomnia,10 and other sleep disorders,11 with HRQOL in clinical samples of children or has included outcomes that are only indirectly associated with HRQOL, such as cognitive testing12 and mood regulation.13 *Evidence is* emerging that HRQOL is also adversely affected in community samples of children with milder sleep health issues, such as parent-reported sleep problems14,15 or sleep initiation or maintenance difficulties.16 However, little research has determined whether HRQOL is a factor in sleep in otherwise healthy children. A 2016 systematic review6 reported an inverse association between sleep duration and HRQOL in 3 studies but rated such evidence as very low quality. Although several recent large studies have reported associations between sleep and HRQOL in children,17,18,19 these studies have all been cross-sectional and unable to determine causality. Few studies have used objective measures of sleep, which can differ markedly from measures obtained from questionnaires.20,21 *As a* recent study showed, device-measured sleep had no association with HRQOL, whereas self-reported ratings of poor sleep quantity and quality were associated with lower HRQOL.19 To our knowledge, no experimental studies have yet determined the effect of manipulating sleep on HRQOL in healthy children to the extent that might represent the levels of mild sleep deprivation that many children may experience today.22 The aim of this secondary analysis was to determine the effect of mild sleep deprivation (induced via the Daily Rest, Eating, and Activity Monitoring [DREAM], a home-based sleep manipulation trial) on HRQOL in children without major sleep issues.
## Study Design
The DREAM randomized crossover trial investigated how mild sleep deprivation influenced eating behaviors and activity patterns in children aged 8 to 12 years in the naturalistic home environment. Detailed information on this trial is provided in the protocol23 (Supplement 1) and a previous article on the main outcomes.24 This prespecified secondary analysis focused on HRQOL, a secondary outcome of the trial. No sample size calculations were undertaken because HRQOL was a secondary outcome, but this study was sufficiently powered to detect relevant differences in the primary outcomes.23,24 The DREAM trial was approved by the University of Otago Human Ethics Committee, and written informed consent (by parents or guardians) or assent (by children) was obtained at the first visit after a verbal explanation of the protocol and an opportunity to ask any questions. We followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.25
## Participants
Healthy children were recruited by advertisement between October 2018 and March 2020 and were eligible to participate if they were aged 8 to 12 years; lived in the wider *Dunedin area* in New Zealand; had no underlying medical conditions or medications that could affect sleep; and scored 39 or lower, which indicated no major sleep problems, on the Sleep Disturbance Scale for Children.26 Only children with parent (or guardian)-reported time in bed of 8 to 11 hours per night were included to ensure that the intervention (restriction and extension of time in bed) did not place them in the not recommended category of sleep duration for this age group (ie, <7 or >12 hours per night).27 Parents and children were emailed separate written information sheets prior to enrolling.
## Randomization and Masking
Children were randomized to the order in which they underwent sleep restriction and sleep extension weeks and were stratified by age group (8-10 years or 11-12 years) and sex (male or female). Randomization to order was generated by one of us (J.J.H., the study biostatistician) using random block lengths in Stata, version 17.0 (StataCorp LLC) in a 1:1 allocation and then was uploaded to a research management program randomization module (REDCap; Vanderbilt University).28 Randomization was undertaken by the research staff following baseline measures. Participants or intervention deliverers could not be blinded to the intervention group (sleep extension or sleep restriction), but those who were analyzing the accelerometry data and the biostatistician were all blinded to the intervention allocation and worked using code A or B to denote group randomization order rather than sleep extension or restriction.
## Procedures
To achieve mild sleep deprivation, children were asked to go to bed 1 hour earlier than usual for 1 week (sleep extension) and 1 hour later than usual for 1 week (sleep restriction), separated by a 1-week washout to allow sufficient time for children to return to their usual sleeping habits before the next intervention week.29 Means of usual bed and wake times, determined from a 7-day sleep diary at baseline, were calculated separately for weekdays and weekend days. Researchers discussed with parents whether these means reflected usual bed and wake times and adjusted the means if required (eg, if a child was late to bed on 1 night, this was adjusted to usual as indicated by the parent). Wake times were kept constant to mirror daily life restrictions (eg, school start times), and interventions were administered only during the school term.
Researchers worked with families during a single problem-solving session (typically 5-10 minutes) to identify any barriers to changing bedtimes, such as being at a sporting activity until close to the extension bedtime, that limited the ability to complete prebed activities in a timely fashion. A suggestion to counteract this barrier could have been to prepare the dinner meal earlier that day or to complete school assignments in the morning. Families received daily personalized bedtime text reminders during each intervention week.
## Measures Collected at Baseline
Demographic data were obtained from the parent, including age, sex, and race and ethnicity of the child; presence of any siblings; and maternal educational level (the index used in New Zealand; and all participating parents or guardians identified as mothers). Area-level deprivation for a family was measured with the 2018 New Zealand Index of Deprivation, an index based on the New Zealand Census of Population and *Dwellings data* that reflects the extent of material and social deprivation used to construct deciles from 1 (least deprived) to 10 (most deprived).30 Parents completed the 22-item Children’s Sleep Hygiene Scale,31 which assesses the regularity of behaviors that might support or interfere with sleep and has a score range of 1 to 6 for each item, with higher scores indicating better sleep hygiene (subscale and total scores are calculated as means). This information was mainly collected to provide feedback to families but was occasionally used to inform discussions about bedtimes if an individual sleep hygiene practice was uncommon. Duplicate measures of height (in centimeters) and weight (in kilograms) were collected at baseline using standard procedures,23 and body mass index z scores were calculated using the World Health Organization Child Growth Reference data.32 Children wore an accelerometer (ActiGraph wGT3X-BT; ActiGraph LLC), which was set to 30 hz at initialization and downloaded with 15-second epochs, on their right hip 24 hours a day for 1 week to measure sleep timing (onset and offset), sleep duration (total sleep time), and sleep quality (number of awakenings, wake after sleep onset, and sleep efficiency). Actigraphy data were analyzed using a count-scaled algorithm developed in MATLAB (MathWorks Inc) that estimated sleep onset and offset as well as awakenings that were specific to each individual for each day.33,34
## Measures Collected During Each Intervention Week
Actigraphy data were collected for each child for the duration of each intervention week, as occurred at baseline. Questionnaire data were collected at the end of each intervention week (day 8) during an assessment session.23 Children and their parents each completed the 8-item Pediatric Sleep Disturbance and 8-item Sleep-Related Impairment scales of the Patient-Reported Outcomes Measurement Information System (PROMIS) questionnaire.35 The responses provided a subjective assessment of the difficulties the child had in falling and staying asleep (disturbance) and daytime sleepiness and the effects on functioning (impairment). Questionnaire items referred to occurrences over the past week and used 5 frequency response options: never, almost never, sometimes, almost always, or always. The Cronbach α for the sleep disturbance scale was α = 0.87 in children and α = 0.85 in parents. The Cronbach α for the sleep impairment scale was α = 0.91 in children and α = 0.94 in parents.
Children also completed the 27-item KIDSCREEN questionnaire, which assessed HRQOL over the past week, with a score range of 1 (never) to 5 (always) or 1 (not at all) to 5 (extremely), as appropriate.36 Children provided answers based on 2 response scales (eg, Thinking about the last week, have you been in a good mood? with answer options of never, seldom, quite often, very often, or always; or Thinking about last week, have you been happy at school? with answer options of not at all, slightly, moderately, very, or extremely). The KIDSCREEN-27 questionnaire produced a total score (score range: 2.4-5.0) and 5 subscale scores: physical well-being (5 items, with score range: 1.6-5.0; Cronbach α = 0.78), psychological well-being (7 items, with score range: 2.6-5.0; Cronbach α = 0.84), autonomy and parental relations (7 items, with score range: 1.7-5.0; Cronbach α = 0.80), social and peer support (4 items, with score range: 2.0-5.0; Cronbach α = 0.87), and school environment (4 items, with score range: 1.8-5.0; Cronbach α = 0.82). Children were assisted with completing the questionnaires if required; in practice, assistance was rarely needed.
## Statistical Analysis
All analyses were undertaken in Stata, version 17.0 (StataCorp LLC). Effects of mild sleep deprivation were estimated by mixed-effects regression models, with the child as a random effect. Mean differences and $95\%$ CIs were determined for sleep restriction compared with sleep extension. Standardized mean differences (SMDs) and $95\%$ CIs were also calculated using a pooled SD. Residuals of models were plotted and visually assessed for homoskedasticity and normality. Complete case analyses were undertaken using the full sample (intention to treat) and were restricted to those who met the per-protocol a priori definition of a difference in sleep of at least 30 minutes per night.
Two-sided $P \leq .05$ indicated statistical significance. Data were analyzed between July 4 and September 1, 2022.
## Results
The Figure illustrates that just 5 children withdrew from participation due to health ($$n = 3$$) and COVID-19 restrictions ($$n = 2$$), leaving a final sample of 100 children. The children included 52 girls ($52\%$) and 48 boys ($48\%$) and had a mean (SD) age of 10.3 (1.4) years; of these children, 24 ($24\%$) had overweight and 16 ($16\%$) had obese status. Of the parents, 47 ($47\%$) reported having a university degree or higher educational level (Table 1).30,31,32,37 **Figure.:** *CONSORT Flow DiagramHRQOL indicates health-related quality of life; SDSC, Sleep Disturbance Scale for Children.* TABLE_PLACEHOLDER:Table 1.
Baseline sleep characteristics (from actigraphy) showed that children received a mean (SD) total sleep time of 8 hours and 59 minutes (45 minutes) per night and a high level of sleep efficiency (mean [SD], $95.8\%$ [$4.3\%$]), waking less than once a night on average (mean [SD], 0.71 [0.61] awakenings) (Table 1). Baseline levels of sleep hygiene were good, with mean scores for all subscales of the Children’s Sleep Hygiene Scale being greater than 4 (with 6 as the possible maximum), indicating that the children regularly followed sleep hygiene practices.
Table 2 shows the differences in child sleep between the 2 intervention weeks (with data from the PROMIS questionnaire and actigraphy). With restricted opportunities to sleep, both children (mean difference, −0.6; $95\%$ CI, −0.8 to −0.5) and parents (mean difference, −0.7; $95\%$ CI, −0.8 to −0.5) reported relative reductions in sleep disturbances, suggesting that sleep was more consolidated when sleep opportunities were restricted rather than extended. However, both children (mean difference, 0.4; $95\%$ CI, 0.3-0.6) and parents (mean difference, 0.8; $95\%$ CI, 0.6-0.9) also reported that the child felt more impaired during the day under the sleep restriction condition. Findings were broadly similar when limited to those children who met the a priori difference in sleep across the 2 intervention conditions. The SMDs indicated that the effects on sleep impairment and disturbance were moderate (eg, child-reported sleep impairment: SMD, 0.6; $95\%$ CI, 0.4-0.8) to large (eg, parent-reported sleep disturbance: SMD, –1.1; $95\%$ CI, –1.3 to –0.9).
**Table 2.**
| Unnamed: 0 | No. of participants | Mean (SD) | Mean (SD).1 | Mean difference (95% CI) | Standardized mean difference (95% CI)a |
| --- | --- | --- | --- | --- | --- |
| | No. of participants | Sleep extension | Sleep restriction | Mean difference (95% CI) | Standardized mean difference (95% CI)a |
| Full sample | Full sample | Full sample | Full sample | Full sample | Full sample |
| PROMIS questionnaireb | | | | | |
| Child reported | | | | | |
| Sleep disturbance scale | 100 | 2.3 (0.8) | 1.7 (0.7) | −0.6 (−0.8 to −0.5) | −0.8 (−1.0 to −0.6) |
| Sleep impairment scale | 100 | 1.4 (0.5) | 1.8 (0.8) | 0.4 (0.3 to 0.6) | 0.6 (0.4 to 0.8) |
| Parent reported | | | | | |
| Sleep disturbance scale | 96 | 2.0 (0.6) | 1.3 (0.4) | −0.7 (−0.8 to −0.5) | −1.1 (−1.3 to −0.9) |
| Sleep impairment scale | 99 | 1.3 (0.4) | 2.1 (0.7) | 0.8 (0.6 to 0.9) | 1.0 (0.8 to 1.1) |
| Actigraphyc | | | | | |
| Sleep onset at night, h:min | 96 | 20:54 (0:45) | 21:58 (0:36) | 64 (58 to 70) | 1.2 (1.1 to 1.4) |
| Sleep offset in morning, h:min | 96 | 6:37 (0:33) | 6:55 (0:32) | 18 (13 to 24) | 0.6 (0.4 to 0.7) |
| Total sleep time, mind | 96 | 556 (40) | 517 (41) | −39 (−46 to −32) | −0.9 (−1.0 to −0.7) |
| Sleep variability, %e | 96 | 10.0 (4.4) | 8.9 (5.5) | −1.1 (−2.4 to 0.2) | −0.2 (−0.5 to 0.1) |
| Weekday-to-weekend differences, min | 88 | −8.3 (60.5) | −13.9 (53.7) | −5.5 (−22.3 to 11.3) | −0.1 (−0.4 to 0.2) |
| WASO, min | 96 | 26.2 (23.7) | 17.5 (18.0) | −8.7 (−12.9 to −4.4) | −0.4 (−0.6 to −0.2) |
| Sleep efficiency, %f | 96 | 95.6 (4.0) | 96.7 (3.3) | 1.2 (0.4 to 1.9) | 0.3 (0.1 to 0.5) |
| No. of awakenings | 96 | 0.7 (0.6) | 0.5 (0.5) | −0.2 (−0.3 to −0.1) | −0.3 (−0.6 to −0.1) |
| Per-protocol sample | | | | | |
| PROMIS questionnaireb | | | | | |
| Child reported | | | | | |
| Sleep disturbance scale | 59 | 2.4 (0.7) | 1.8 (0.7) | −0.6 (−0.8 to −0.4) | −0.8 (−1.0 to −0.5) |
| Sleep impairment scale | 59 | 1.4 (0.5) | 1.9 (0.8) | 0.4 (0.2 to 0.7) | 0.6 (0.3 to 0.9) |
| Parent reported | | | | | |
| Sleep disturbance scale | 56 | 2.0 (0.7) | 1.3 (0.3) | −0.7 (−0.9 to −0.6) | −1.2 (−1.4 to −0.9) |
| Sleep impairment scale | 59 | 1.3 (0.4) | 2.1 (0.7) | 0.8 (0.6 to 1.0) | 1.0 (0.8 to 1.13) |
| Actigraphyc | | | | | |
| Sleep onset at night, h:min | 59 | 20:53 (0:45) | 22:05 (0:36) | 71 (64 to 78) | 1.3 (1.2 to 1.4) |
| Sleep offset in morning, h:min | 59 | 6:40 (0:32) | 6:49 (0:32) | 10 (4 to 16) | 0.3 (0.1 to 0.5) |
| Total sleep time, mind | 59 | 563 (40) | 502 (36) | −61 (−68 to −54) | −1.3 (−1.4 to −1.1) |
| Sleep variability, %e | 59 | 9.0 (3.6) | 9.3 (6.1) | 0.3 (−1.4 to 2.0) | 0.0 (−0.4 to 0.3) |
| Weekday-to-weekend differences, min | 54 | −9.2 (52.8) | −14.2 (52.2) | −5.0 (−24.4 to 14.4) | −0.1 (−0.4 to 0.3) |
| WASO, min | 59 | 23.7 (22.4) | 19.3 (17.1) | −4.5 (−9.7 to 0.8) | −0.2 (−0.5 to 0.0) |
| Sleep efficiency, %f | 59 | 96.0 (3.8) | 96.3 (3.2) | 0.3 (−0.6 to 1.2) | 0.1 (−0.2 to 0.3) |
| No. of awakenings | 59 | 0.7 (0.5) | 0.6 (0.5) | −0.1 (−0.2 to 0.1) | −0.2 (−0.4 to 0.1) |
Accelerometry data indicated that children went to sleep an average of 64 ($95\%$ CI, 58-70) minutes later each night over the sleep restriction week compared with the sleep extension week, with a smaller difference (as anticipated) in sleep offset (mean difference, 18; $95\%$ CI, 13-24 minutes). These differences in sleep timing meant that children received 39 ($95\%$ CI, 32-46) minutes less of total sleep time each night during the sleep restriction week. These differences were magnified in the per-protocol sample, which had a nightly difference in total sleep time of 71 ($95\%$ CI, 64-78) minutes. As observed from the child and parent reports, more consolidated sleep was apparent, with small reductions in the number of awakenings (mean difference, −0.2; $95\%$ CI, −0.3 to −0.1) and wake after sleep onset (mean difference, −8.7 minutes; $95\%$ CI, −12.9 to −4.4 minutes) as well as a small improvement in sleep efficiency (mean difference, $1.2\%$; $95\%$ CI, $0.4\%$-$1.9\%$) in the full sample (Table 2). By contrast, sleep variability (variation in total sleep time or difference in total sleep time between weekdays and weekends) did not differ significantly. Significant differences in sleep consolidation were not apparent in the per-protocol sample for any measure.
Table 3 presents the mean differences for HRQOL in the sleep restriction week compared with the sleep extension week. Children reported significantly lower scores for physical well-being (SMD, −0.28; $95\%$ CI, −0.49 to −0.08) and ability to cope well in the school environment (SMD, −0.26; $95\%$ CI, −0.42 to −0.09), leading to total HRQOL scores that were significantly lower when tired (SMD, −0.21; $95\%$ CI, −0.34 to −0.08). Reductions in psychological well-being (mean difference, −0.09; $95\%$ CI, −0.19 to 0.02), social and peer support (mean difference, −0.13; $95\%$ CI, −0.26 to 0.01), and autonomy and parental relations (mean difference, −0.08; $95\%$ CI, −0.19 to 0.03) were observed but were not statistically significant. Differences in HRQOL were generally magnified in the per-protocol sample, with the reduction in social and peer support (SMD, −0.24; $95\%$ CI, −0.47 to −0.01) also being statistically significant. The SMDs indicated that these effects on HRQOL were small (eg, total well-being: SMD, –0.21; $95\%$ CI, –0.34 to –0.08).
**Table 3.**
| Unnamed: 0 | No. of participants | Mean (SD) | Mean (SD).1 | Mean difference (95% CI)a | Standardized mean difference (95% CI)b |
| --- | --- | --- | --- | --- | --- |
| | No. of participants | Sleep extension | Sleep restriction | Mean difference (95% CI)a | Standardized mean difference (95% CI)b |
| Full sample | Full sample | Full sample | Full sample | Full sample | Full sample |
| KIDSCREEN-27 questionnairec | | | | | |
| Physical well-being subscale score | 99 | 3.8 (0.7) | 3.7 (0.7) | −0.20 (−0.34 to −0.05) | −0.28 (−0.49 to −0.08) |
| Psychological well-being subscale score | 98 | 4.3 (0.5) | 4.2 (0.6) | −0.09 (−0.19 to 0.02) | −0.15 (−0.32 to 0.03) |
| Autonomy and parental relations subscale score | 97 | 3.9 (0.8) | 3.8 (0.8) | −0.08 (−0.19 to 0.03) | −0.10 (−0.25 to 0.04) |
| Social and peer support subscale score | 100 | 4.3 (0.8) | 4.2 (0.8) | −0.13 (−0.26 to 0.01) | −0.15 (−0.32 to 0.01) |
| School environment subscale score | 97 | 4.2 (0.7) | 4.0 (0.8) | −0.19 (−0.31 to −0.07) | −0.26 (−0.42 to −0.09) |
| Total score | 100 | 4.1 (0.6) | 4.0 (0.6) | −0.12 (−0.20 to −0.04) | −0.21 (−0.34 to −0.08) |
| Per-protocol sample | Per-protocol sample | Per-protocol sample | Per-protocol sample | Per-protocol sample | Per-protocol sample |
| KIDSCREEN-27 questionnairec | | | | | |
| Physical well-being subscale score | 58 | 3.8 (0.7) | 3.6 (0.7) | −0.20 (−0.38 to −0.01) | −0.29 (−0.57 to −0.01) |
| Psychological well-being subscale score | 57 | 4.3 (0.5) | 4.2 (0.7) | −0.11 (−0.25 to 0.03) | −0.18 (−0.42 to 0.06) |
| Autonomy and parental relations subscale score | 57 | 3.9 (0.7) | 3.8 (0.9) | −0.12 (−0.26 to 0.02) | −0.14 (−0.32 to 0.03) |
| Social and peer support subscale score | 59 | 4.4 (0.7) | 4.2 (0.8) | −0.19 (−0.37 to −0.01) | −0.24 (−0.47 to −0.01) |
| School environment subscale score | 58 | 4.2 (0.7) | 3.9 (0.9) | −0.21 (−0.38 to −0.05) | −0.27 (−0.48 to −0.06) |
| Total score | 59 | 4.1 (0.6) | 3.9 (0.6) | −0.15 (−0.26 to −0.04) | −0.25 (−0.43 to −0.06) |
## Discussion
Results of this secondary analysis of the DREAM trial demonstrated that even relatively small reductions in nightly sleep duration can have a considerable effect on HRQOL in children. These children received 39 minutes less sleep per night between sleep conditions over only 1 week. This loss of sleep resulted in significant reductions in the children’s physical well-being, overall well-being, and ability to cope well in a school environment. For those who achieved the a priori difference in sleep of at least 30 minutes per night, additional reductions in well-being associated with less social and peer support were also observed. While these differences may generally be considered as small but not trivial,38 the reductions in multiple aspects of HRQOL were observed after only 1 week of less sleep. As such, we believe these findings are clinically and statistically significant and require confirmation over the longer term.
It is difficult to compare the findings with those reported in the literature because most previous experimental studies12,13 appeared to involve children with clinical sleep issues, where greater benefits might be expected. In young children with obstructive sleep apnea, significant improvements in HRQOL were observed, principally in school and physical function domains, after adenotonsillectomy.9 In children with milder sleep issues (several parent-reported moderate sleep problems at school entry), a brief clinician-delivered intervention showed short-term (3-month) improvements in psychosocial HRQOL even though the proportion of children with sleep problems had not decreased. However, benefits to HRQOL were not maintained at follow-up (12 months), and any corresponding changes in sleep duration were not reported.39 Other research that manipulated sleep in children focused more on emotional or cognitive outcomes rather than on HRQOL, reporting consistent adverse effects on mood and smaller effects on emotion and some cognitive test scores.4,12,13 We believe the findings of this trial add considerable value to the existing cross-sectional literature that shows the association of sleep duration with HRQOL in children17,18,19 because causality can now be inferred.
Previous analyses of the DREAM trial provided some insight into why these relatively modest changes in sleep might affect HRQOL, although it is difficult to disentangle which behavioral changes might influence HRQOL the most as they tend to be interrelated in children.40 We found that when children slept less, they ate substantially more calories, particularly in the evenings, all of which came from noncore foods (generally those with poor nutritional quality) rather than from core foods, such as fruit and vegetables,41 which are associated with higher HRQOL.42 Children replaced this loss of sleep mostly with sedentary time and, to a lesser degree, light activity.43 Given that these activity patterns were measured using accelerometry, we do not know which specific behaviors might have changed, although anecdotally, parents reported more screen time during the sleep restriction week. *In* general, greater HRQOL has been associated with higher amounts of activity and lower amounts of sedentary time in children.44,45
## Strengths and Limitations
This study has several strengths. These strengths primarily centered on the randomized crossover trial design, wherein each child participant was able to act as their own control, with the sleep extension week providing the conditions for more sleep opportunity compared with the sleep restriction week that provided less sleep opportunity. We determined the difference in sleep on the basis of actigraphy rather than subjective measures of sleep duration, which should provide greater accuracy.20 However, it was interesting to observe greater sleep efficiency when sleep was restricted, suggesting more consolidation of sleep, which was supported by children and parent reports of less sleep disturbance in the PROMIS questionnaire. This finding likely reflects an increased homeostatic drive to sleep as a consequence of sleep restriction.46 *It is* important to note that we specifically tested sleep restriction vs sleep extension conditions to ensure the best opportunity to create a difference in true sleep, as the DREAM trial was a mechanistic study that aimed to determine the effect of mild sleep deprivation on eating and activity behaviors associated with obesity in children.23,24 We purposely chose to restrict sleep by a relatively small amount in an effort to mimic clinical levels of mild sleep deprivation,22 which we believed had greater applicability for public health than more severe sleep deficits often used in other sleep manipulation trials.47,48 This study also has limitations. It focused on a secondary outcome of interest, for which we did not undertake specific power calculations. Although data collection had to be concluded slightly earlier than anticipated due to COVID-19 restrictions, the total dropout rate was low at $5\%$ compared with the projected rate of at least $20\%$,23 ensuring a robust sample size and reduced risk of attrition bias. The sample was not diverse, which may limit extrapolation to other groups. As the trial was a mechanistic study, we cannot comment on the effect of sleep loss over the long term and its implications for HRQOL in children because a consistently linear association between less sleep and worse HRQOL cannot be assumed in the absence of such evidence. Furthermore, we had a single measure of HRQOL from the children; confirmatory data from different measures or parents would have been an advantage.
## Conclusions
In this secondary analysis of the DREAM randomized crossover trial of sleep manipulation, we showed that after only 1 week of receiving 39 minutes less sleep per night between sleep conditions, children reported significantly lower HRQOL in terms of their physical and overall well-being and ability to cope well at school. These findings highlight that ensuring children receive sufficient good-quality sleep is an important child health issue.
## References
1. Miller MA, Kruisbrink M, Wallace J, Ji C, Cappuccio FP. **Sleep duration and incidence of obesity in infants, children, and adolescents: a systematic review and meta-analysis of prospective studies**. *Sleep* (2018.0) **41**. DOI: 10.1093/sleep/zsy018
2. Rudnicka AR, Nightingale CM, Donin AS. **Sleep duration and risk of type 2 diabetes**. *Pediatrics* (2017.0) **140**. DOI: 10.1542/peds.2017-0338
3. Peach H, Gaultney JF, Reeve CL. **Sleep characteristics, body mass index, and risk for hypertension in young adolescents**. *J Youth Adolesc* (2015.0) **44** 271-284. DOI: 10.1007/s10964-014-0149-0
4. Tomaso CC, Johnson AB, Nelson TD. **The effect of sleep deprivation and restriction on mood, emotion, and emotion regulation: three meta-analyses in one**. *Sleep* (2021.0) **44**. DOI: 10.1093/sleep/zsaa289
5. Vriend J, Davidson F, Rusak B, Corkum P. **Emotional and cognitive impact of sleep restriction in children**. *Sleep Med Clin* (2015.0) **10** 107-115. DOI: 10.1016/j.jsmc.2015.02.009
6. Chaput JP, Gray CE, Poitras VJ. **Systematic review of the relationships between sleep duration and health indicators in school-aged children and youth**. *Appl Physiol Nutr Metab* (2016.0) **41** S266-S282. DOI: 10.1139/apnm-2015-0627
7. Wallander JL, Koot HM. **Quality of life in children: a critical examination of concepts, approaches, issues, and future directions**. *Clin Psychol Rev* (2016.0) **45** 131-143. DOI: 10.1016/j.cpr.2015.11.007
8. Ravens-Sieberer U, Erhart M, Wille N, Wetzel R, Nickel J, Bullinger M. **Generic health-related quality-of-life assessment in children and adolescents: methodological considerations**. *Pharmacoeconomics* (2006.0) **24** 1199-1220. DOI: 10.2165/00019053-200624120-00005
9. Garetz SL, Mitchell RB, Parker PD. **Quality of life and obstructive sleep apnea symptoms after pediatric adenotonsillectomy**. *Pediatrics* (2015.0) **135** e477-e486. DOI: 10.1542/peds.2014-0620
10. Combs D, Goodwin JL, Quan SF, Morgan WJ, Shetty S, Parthasarathy S. **Insomnia, health-related quality of life and health outcomes in children: a seven year longitudinal cohort**. *Sci Rep* (2016.0) **6** 27921. DOI: 10.1038/srep27921
11. Hart CN, Palermo TM, Rosen CL. **Health-related quality of life among children presenting to a pediatric sleep disorders clinic**. *Behav Sleep Med* (2005.0) **3** 4-17. DOI: 10.1207/s15402010bsm0301_3
12. Sadeh A, Gruber R, Raviv A. **The effects of sleep restriction and extension on school-age children: what a difference an hour makes**. *Child Dev* (2003.0) **74** 444-455. DOI: 10.1111/1467-8624.7402008
13. Gruber R, Cassoff J, Frenette S, Wiebe S, Carrier J. **Impact of sleep extension and restriction on children’s emotional lability and impulsivity**. *Pediatrics* (2012.0) **130** e1155-e1161. DOI: 10.1542/peds.2012-0564
14. Williamson AA, Mindell JA, Hiscock H, Quach J. **Longitudinal sleep problem trajectories are associated with multiple impairments in child well-being**. *J Child Psychol Psychiatry* (2020.0) **61** 1092-1103. DOI: 10.1111/jcpp.13303
15. Williamson AA, Zendarski N, Lange K. **Sleep problems, internalizing and externalizing symptoms, and domains of health-related quality of life: bidirectional associations from early childhood to early adolescence**. *Sleep* (2021.0) **44**. DOI: 10.1093/sleep/zsaa139
16. Sundell AL, Angelhoff C. **Sleep and its relation to health-related quality of life in 3-10-year-old children**. *BMC Public Health* (2021.0) **21** 1043. DOI: 10.1186/s12889-021-11038-7
17. Qin Z, Wang N, Ware RS, Sha Y, Xu F. **Lifestyle-related behaviors and health-related quality of life among children and adolescents in China**. *Health Qual Life Outcomes* (2021.0) **19** 8. DOI: 10.1186/s12955-020-01657-w
18. Wong CKH, Wong RS, Cheung JPY. **Impact of sleep duration, physical activity, and screen time on health-related quality of life in children and adolescents**. *Health Qual Life Outcomes* (2021.0) **19** 145. DOI: 10.1186/s12955-021-01776-y
19. Xiao Q, Chaput JP, Olds T. **Sleep characteristics and health-related quality of life in 9- to 11-year-old children from 12 countries**. *Sleep Health* (2020.0) **6** 4-14. DOI: 10.1016/j.sleh.2019.09.006
20. Galland B, Meredith-Jones K, Terrill P, Taylor R. **Challenges and emerging technologies within the field of pediatric actigraphy**. *Front Psychiatry* (2014.0) **5** 99. DOI: 10.3389/fpsyt.2014.00099
21. Alfano CA, Patriquin MA, De Los Reyes A. **Subjective-objective sleep comparisons and discrepancies among clinically-anxious and healthy children**. *J Abnorm Child Psychol* (2015.0) **43** 1343-1353. DOI: 10.1007/s10802-015-0018-7
22. Ford K, Kelly PT, Williamson R, Hlavac M. **Sleep habits of intermediate-aged students: roles for the students, parents and educators**. *N Z Med J* (2020.0) **133** 59-66. PMID: 32242179
23. Ward AL, Galland BC, Haszard JJ. **The effect of mild sleep deprivation on diet and eating behaviour in children: protocol for the Daily Rest, Eating, and Activity Monitoring (DREAM) randomized cross-over trial**. *BMC Public Health* (2019.0) **19** 1347. DOI: 10.1186/s12889-019-7628-x
24. Morrison S, Galland BC, Haszard JJ. **Eating in the absence of hunger in children with mild sleep loss: a randomized crossover trial with learning effects**. *Am J Clin Nutr* (2021.0) **114** 1428-1437. DOI: 10.1093/ajcn/nqab203
25. Dwan K, Li T, Altman DG, Elbourne D. **CONSORT 2010 statement: extension to randomised crossover trials**. *BMJ* (2019.0) **366** l4378. DOI: 10.1136/bmj.l4378
26. Bruni O, Ottaviano S, Guidetti V. **The Sleep Disturbance Scale for Children (SDSC): construction and validation of an instrument to evaluate sleep disturbances in childhood and adolescence**. *J Sleep Res* (1996.0) **5** 251-261. DOI: 10.1111/j.1365-2869.1996.00251.x
27. Hirshkowitz M, Whiton K, Albert SM. **National Sleep Foundation’s updated sleep duration recommendations: final report**. *Sleep Health* (2015.0) **1** 233-243. DOI: 10.1016/j.sleh.2015.10.004
28. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. **Research Electronic Data Capture (REDCap)–a metadata-driven methodology and workflow process for providing translational research informatics support**. *J Biomed Inform* (2009.0) **42** 377-381. DOI: 10.1016/j.jbi.2008.08.010
29. Short MA, Weber N, Reynolds C, Coussens S, Carskadon MA. **Estimating adolescent sleep need using dose-response modeling**. *Sleep* (2018.0) **41**. DOI: 10.1093/sleep/zsy011
30. Atkinson J, Salmond C, Crampton P. *NZDep2018 Index of Deprivation, Interim Research Report* (2019.0)
31. LeBourgeois MK, Giannotti F, Cortesi F, Wolfson AR, Harsh J. **The relationship between reported sleep quality and sleep hygiene in Italian and American adolescents**. *Pediatrics* (2005.0) **115** 257-265. DOI: 10.1542/peds.2004-0815H
32. de Onis M, Onyango AW, Borghi E, Siyam A, Nishida C, Siekmann J. **Development of a WHO growth reference for school-aged children and adolescents**. *Bull World Health Organ* (2007.0) **85** 660-667. DOI: 10.2471/BLT.07.043497
33. Meredith-Jones K, Williams SM, Galland BC, Kennedy G, Taylor RW. **24hr Accelerometry: impact of sleep-screening methods on estimates of physical activity and sedentary time**. *J Sports Sci* (2016.0) **34** 679-685. DOI: 10.1080/02640414.2015.1068438
34. Smith C, Galland B, Taylor R, Meredith-Jones K. **ActiGraph GT3X+ and Actical wrist and hip worn accelerometers for sleep and wake indices in young children using an automated algorithm: validation with polysomnography**. *Front Psychiatry* (2020.0) **10** 958. DOI: 10.3389/fpsyt.2019.00958
35. Forrest CB, Meltzer LJ, Marcus CL. **Development and validation of the PROMIS Pediatric Sleep Disturbance and Sleep-Related Impairment item banks**. *Sleep* (2018.0) **41**. DOI: 10.1093/sleep/zsy054
36. Ravens-Sieberer U, Herdman M, Devine J. **The European KIDSCREEN approach to measure quality of life and well-being in children: development, current application, and future advances**. *Qual Life Res* (2014.0) **23** 791-803. DOI: 10.1007/s11136-013-0428-3
37. 37Health New Zealand, Ministry of Health. HISO 10001: 2017 ethnicity data protocols. Accessed January 15, 2021. https://www.health.govt.nz/publication/hiso-100012017-ethnicity-data-protocols. (2017.0)
38. Schäfer T, Schwarz MA. **The meaningfulness of effect sizes in psychological research: differences between sub-disciplines and the impact of potential biases**. *Front Psychol* (2019.0) **10** 813. DOI: 10.3389/fpsyg.2019.00813
39. Quach J, Hiscock H, Ukoumunne OC, Wake M. **A brief sleep intervention improves outcomes in the school entry year: a randomized controlled trial**. *Pediatrics* (2011.0) **128** 692-701. DOI: 10.1542/peds.2011-0409
40. Davison J, Stewart-Knox B, Connolly P, Lloyd K, Dunne L, Bunting B. **Exploring the association between mental wellbeing, health-related quality of life, family affluence and food choice in adolescents**. *Appetite* (2021.0) **158**. DOI: 10.1016/j.appet.2020.105020
41. Morrison S
42. Wu XY, Zhuang LH, Li W. **The influence of diet quality and dietary behavior on health-related quality of life in the general population of children and adolescents: a systematic review and meta-analysis**. *Qual Life Res* (2019.0) **28** 1989-2015. DOI: 10.1007/s11136-019-02162-4
43. Morrison S, Haszard JJ, Galland BC. **Where does the time go when children don’t sleep? A randomized crossover study**. *Obesity (Silver Spring)*. DOI: 10.1002/oby.23615
44. Marker AM, Steele RG, Noser AE. **Physical activity and health-related quality of life in children and adolescents: a systematic review and meta-analysis**. *Health Psychol* (2018.0) **37** 893-903. DOI: 10.1037/hea0000653
45. Shi J, Wang X, Wu Q. **The joint association of physical activity and sedentary behavior with health-related quality of life among children and adolescents in Mainland China**. *Front Public Health* (2022.0) **10**. DOI: 10.3389/fpubh.2022.1003358
46. Borbély AA, Achermann P. **Sleep homeostasis and models of sleep regulation**. *J Biol Rhythms* (1999.0) **14** 557-568. DOI: 10.1177/074873099129000894
47. Hart CN, Carskadon MA, Considine RV. **Changes in children’s sleep duration on food intake, weight, and leptin**. *Pediatrics* (2013.0) **132** e1473-e1480. DOI: 10.1542/peds.2013-1274
48. Beebe DW, Simon S, Summer S, Hemmer S, Strotman D, Dolan LM. **Dietary intake following experimentally restricted sleep in adolescents**. *Sleep* (2013.0) **36** 827-834. DOI: 10.5665/sleep.2704
|
---
title: Controlling therapeutic protein expression via inhalation of a butter flavor
molecule
authors:
- Adrian Bertschi
- Bozhidar-Adrian Stefanov
- Shuai Xue
- Ghislaine Charpin-El Hamri
- Ana Palma Teixeira
- Martin Fussenegger
journal: Nucleic Acids Research
year: 2023
pmcid: PMC10018347
doi: 10.1093/nar/gkac1256
license: CC BY 4.0
---
# Controlling therapeutic protein expression via inhalation of a butter flavor molecule
## Abstract
Precise control of the delivery of therapeutic proteins is critical for gene- and cell-based therapies, and expression should only be switched on in the presence of a specific trigger signal of appropriate magnitude. Focusing on the advantages of delivering the trigger by inhalation, we have developed a mammalian synthetic gene switch that enables regulation of transgene expression by exposure to the semi-volatile small molecule acetoin, a widely used, FDA-approved food flavor additive. *The* gene switch capitalizes on the bacterial regulatory protein AcoR fused to a mammalian transactivation domain, which binds to promoter regions with specific DNA sequences in the presence of acetoin and dose-dependently activates expression of downstream transgenes. Wild-type mice implanted with alginate-encapsulated cells transgenic for the acetoin gene switch showed a dose-dependent increase in blood levels of reporter protein in response to either administration of acetoin solution via oral gavage or longer exposure to acetoin aerosol generated by a commercial portable inhaler. Intake of typical acetoin-containing foods, such as butter, lychees and cheese, did not activate transgene expression. As a proof of concept, we show that blood glucose levels were normalized by acetoin aerosol inhalation in type-I diabetic mice implanted with acetoin-responsive insulin-producing cells. Delivery of trigger molecules using portable inhalers may facilitate regular administration of therapeutic proteins via next-generation cell-based therapies to treat chronic diseases for which frequent dosing is required.
## INTRODUCTION
Trigger-inducible gene switches are important tools in biology research and for synthetic biology applications, including gene and cell-based therapies. In recent years, the range of synthetic gene regulatory systems responsive to small molecules (e.g. caffeine [1], spearmint [2], menthol [3]) or physical stimuli (e.g. heating [4]), light [5] or electrical signals [6]) has increased considerably, offering multiple options for specific in vivo gene regulation. To achieve clinical relevance, such systems should be strictly controllable by nontoxic and inexpensive inducers that can be easily administered in order to encourage patient compliance. Furthermore, in the case of food components, the range of activation concentrations should be higher than the endogenous levels following food intake or inadvertent external stimulation. Small molecule triggers have been delivered by oral ingestion [7,8], by transdermal application [9], and by inhalation [2]. Among them, inhalation is a particularly attractive strategy for systemic administration of (semi-)volatile molecules, enabling rapid absorption owing to the large surface area of the pulmonary alveoli connected to a vast network of blood capillaries. A further advantage is the avoidance of potential degradation by trigger-metabolizing gut bacteria, in contrast to orally administered molecules. However, no synthetic gene regulatory system that can be activated by inhalation has yet met all the criteria for practical application. For instance, the spearmint inducible system [2] can be unintentionally activated by exposures occurring during everyday life.
In this work, we focused on the volatile small molecule acetoin to engineer a mammalian gene switch that can be activated in vivo by inhalation. Acetoin is a non-toxic food additive. It is actively metabolized in mammals to the secondary alcohol 2,3-butanediol, which in turn is reduced to 2-butanone or 2-butanol and eventually cleared from the body with a half-life of about 0.9 h [10]. When humans are exposed to large quantities of ethanol, a slight increase of blood acetoin can be detected [11], but as the production of acetoin is slower than its reduction, there is no accumulation of acetoin. Using a small trigger molecule that is actively degraded would enable better control of dosing and therefore better regulation of the activity of the target gene circuit.
Several bacteria, mostly from the Enterobacteriaceae family, produce acetoin as by-product when growing in glucose media, and use acetoin as an energy source when glucose is depleted [12]. Acetoin binds to the regulatory protein AcoR, leading to induction of acetoin-metabolizing enzymes [13]. Structurally, AcoR contains a helix-turn-helix DNA binding domain, an ATP-binding domain, an acetoin binding pocket and a sigma54 recruiting domain [14]. The natural operator sequence to which the helix-turn-helix domain binds is located approximately 80 to 100 bp upstream of the start codon at a palindromic region of around 30 bp [15,16], but the nature of the interaction of AcoR with the DNA has not been fully characterized. In the presence of acetoin, AcoR shows increased binding affinity to the DNA and initiates transcription of the regulated genes.
In order to develop a synthetic gene circuit to control therapeutic gene expression using a commercial portable inhaler, we generated a stable HEK293T cell line that constitutively expresses AcoR from *Bacillus subtilis* fused to a mammalian transactivator (PmPGK-VP16f-type-AcoR-pAbGH) and a transgene under the control of the operon of the acetoin-responsive AcoR (POAcoR-PhCMVmin-SEAP-pAbGH). AcoR is only activated by high concentrations of acetoin, and is not activated even by natural products containing substantial amounts of acetoin. To test the potential utility of this system, we first induced the engineered cells in vitro either by direct exposure to acetoin in the medium or by exposure to acetoin vapor. Then, as a proof of concept, we translated the findings to an in vivo mouse model, in which gene expression was induced by exposure to acetoin aerosol generated with a commercial portable inhaler.
## Plasmid design and molecular cloning
All plasmids used and constructed in this study are listed in Supplementary Table 1. Plasmids were designed using Benchling (www.benchling.com) and sequences were verified at Microsynth AG, Switzerland.
The bacterial strain XL10 gold K12 E. coli (Stratagene) was used to propagate the plasmids. Bacteria were grown at 37°C in Lurie-Bertani lysogeny broth under shaker-aeration. Plasmid DNA was extracted after an 8–16 h incubation period using a Zippy Plasmid Mini-Prep Kit (Cat. No. D4037, Zymo Research) according to the manufacturer's instructions. For PCR reactions Phusion High-Fidelity DNA polymerase (Cat. No. F530, ThermoFisher Scientific, Rheinach, Switzerland) was used according to the manufacturer's instructions. For phosphorylation and annealing of two oligonucleotides, T4 Polynucleotide Kinase (Cat. No. M0201, New England BioLabs) was used in T4 DNA ligase buffer (Cat. No. B69, ThermoFisher Scientific) according to the manufacturer's instructions.
Plasmid DNA or purified PCR-amplified sequences (300–1000 ng) were digested using 2 units of standard restriction enzyme (New England BioLabs) and calf intestinal alkaline phosphatase (Quick CIP, Cat. No. M0525, New England BioLabs) for an incubation period of 2 h at the manufacturer's recommended temperature. The digestion products were purified by standard agarose gel electrophoresis. DNA fragments were extracted using the Zymoclean Gel DNA Recovery Kit (Cat. No. D4002, Zymo Research) according to the manufacturer's instructions. Purified DNA fragments were then ligated to the counterpart with the matching overhang using T4 DNA ligase (Cat. No. EL0011, ThermoFisher Scientific) in T4 DNA ligase buffer (Cat. No. B69, ThermoFisher Scientific) for at least 20 min before transformation into competent bacteria. To do so, 20–50 μl of competent bacteria was added to 10 μl of the ligation mix and heat-shocked at 42°C for 45 s. Heat-shocked bacteria were plated onto ampicillin-containing LB-agar and incubated for 16 h, then a colony was picked and grown as previously described.
Chemicals. Ethanol (EtOH; Cat. no. 51976), dimethyl sulfoxide (DMSO; Cat. no. D4540), acetoin (Cat. no. W200808), lactic acid (Cat. no. L7022), pyruvate (Cat. no. P3662), 2,3-butanedione (Cat. no. B0753), and 4-hydroxy-3-hexanone (Cat. no. CDS000463) were purchased from Sigma-Aldrich (Buchs, Switzerland).
Cell culture and transfection. Human embryonic kidney cells (HEK-293T, ATCC: CRL-11268), HeLa cells (HeLa, ATCC: CCL-2), bone marrow-derived immortalized mesenchymal stem/stromal cells (hMSC-TERT) cells [17], Hep G2 cells (HEPG2, ATCC: HB-8065), baby hamster kidney cells (BHK, ATCC: CCL-10) and HT-1080 cells (HT1080, ATCC: CCL-121) were cultivated in Dulbecco's modified Eagle's medium (Gibco™ DMEM, Cat. no. 31966–021, ThermoFisher Scientific), with $1\%$ (v/v) streptomycin/penicillin (Gibco™ Penicillin-Streptomycin, Cat. no. 15070–063, ThermoFisher Scientific) and supplemented with $10\%$ (v/v) fetal bovine serum (FBS, Cat. no. F7524, Sigma-Aldrich). Chinese hamster ovary cells (CHO-K1, ATCC: CCL-61) and A549 cells (A549, ATCC: CCL-185) were cultivated in Ham's F-12K (Kaighn's) medium (Gibco™ F-12K, Cat. no. 21127–022, ThermoFisher Scientific) with $1\%$ (v/v) streptomycin/penicillin (Gibco™ Penicillin-Streptomycin, Cat. no. 15070-063, ThermoFisher Scientific) and supplemented with $10\%$ (v/v) fetal bovine serum (FBS, Cat. no. F7524, Sigma-Aldrich). For optimal consistency between different transient transfections, the protocol was standardized. 15 000 cells were transfected overnight with a total of 150 ng of DNA per well in 96-well plates, using a ratio of DNA to polyethyleneimine (PEI, Cat. No. 24765-2, Polysciences) of 1:6. The PEI (MW 40 000) stock solution was 1 mg/mL in ddH2O. The next morning the medium was exchanged for inducer-containing medium and after 24 h induction (unless otherwise stated) the culture supernatant was collected for quantification of secreted reporter protein.
SEAP reporter assay. SEAP reporter assay was performed as previously described [18]. Briefly, culture supernatant containing SEAP was heat-inactivated at 65°C for 30 min. Then, 20 μl of inactivated medium was diluted with 80 μl of water. Next, 80 μl of 2× SEAP assay buffer (20 mM homoarginine, 1 mM MgCl2, $21\%$(v/v) diethanolamine, pH 9.8) and 20 μl of substrate solution (120 mM p-nitrophenyl phosphate in 2× SEAP assay buffer (Acros Organics BVBA)) was added and the time-dependent increase in absorbance at 405 nm was measured using a Tecan Infinite M1000 microplate reader over a period of 30 min. The presented values are representative measurements of three independent experiments.
Western-blot analysis. HEK293T cells were seeded and transfected with FLAG-tagged AcoR-VP16f-type from different Bacillales under the control of the constitutive PPKG promoter. At 48 h after transfection, cells were harvested, centrifuged at 1000 × g for 2 min, washed twice with ice-cold PBS and lysed using RIPA buffer supplemented with a protease inhibitor cocktail (Pierce™ Protease Inhibitor Tablets, Cat. no. A32963, ThermoFisher Scientific) on ice for 20 min with repeated vortexing. The lysate was then centrifuged at 12 000 × g for 30 min at 4°C and the clear lysate was transferred into a new tube. Protein concentration was determined using a quantification kit (Pierce™ BCA Protein Assay Kit, Cat. no. 23227, ThermoFisher Scientific). 5X reduced Laemmli sample buffer was added and the mixture was heated for 5 min at 95°C. An aliquot containing 10 μg of protein was loaded on SDS-PAGE. After SDS-PAGE, the gel was blocked using $10\%$ skimmed milk in TBST buffer at room temperature for 1 h. FLAG-specific primary antibody (monoclonal ANTI-FLAG® M2 antibody produced in mouse, Cat. no. F1804, Sigma-Aldrich) and β-actin specific primary antibody (β-Actin (13E5) Rabbit mAb, Cat. no. 4970, Cell Signaling) were used at 1:1000 dilution. Secondary HRP-conjugated goat-anti-rabbit (HRP-AffiniPure Goat Anti-Rabbit IgG (H + L), Cat. no. 111–035-144, Jackson Immunoresearch) and anti-mouse (HRP-AffiniPure Goat Anti-Mouse IgG (H + L), Cat. no. 115-035-003, Jackson Immunoresearch) antibodies were used at a dilution of 1:10 000. A protein ladder (PageRulerTM Plus Prestained Protein Ladder, 10–250 kDa, Cat. no. 26619, ThermoFisher Scientific) was used as protein molecular weight marker. An imaging platform (FUSION Pulse TS, Cat. no. 37480003, Vilber, France) was used to develop blots before they were analyzed using Adobe Illustrator.
RNA sequencing. HEK293T cells were seeded and transfected as previously described. Cells were collected 24 h after induction with acetoin or control medium. RNA was then isolated using a Quick-RNA MiniPrep Kit (Cat. no. R1055, Zymo Research). The isolated RNA was prepared for sequencing with a TruSeq stranded mRNA Illumina HT kit v2 and sequenced with a NextSeq 500 using Illumina RTA v 2.11.3 with 76 cycles. Illumina sequencing data were demultiplexed and primary analysis was performed using a Snakemake workflow, which includes Trimmomatic (v 0.35), alignment to the GRCh38 genome with Hisat2 (v 2.1.0), Samtools (v 1.9) to sort and index the alignment BAM files, and Counts from the Subread package (v 2.0.1) to count reads in the gene ranges, using human Ensembl annotation v105. The count vectors for all samples were combined into a table, which was then subjected to the secondary analysis in R.
Quality control and sample consistency were confirmed with PCA, using R package PCATools. The count table was processed for secondary (statistical) analysis with R scripts using EdgeR (v3.32), affording lists of genes ranked for differential expression by P-value. Benjamini–Hochberg adjusted P-value was used to estimate the false discovery rate. Pathway enrichment analysis was performed with GeneGo Metacore. ( *Primary analysis* workflow repository: https://github.com/michalogit/snake_hisat/) Stable cell line generation. For stable cell line production, we employed the Tier-3 vector described by Haellman et al. [ 19] using the Sleeping Beauty transposon protocol [20]. 25 000 HEK293T cells were seeded into a well of a 6-well plate and transfected with pAB800 (2000 ng), pAB801 (400 ng) and pTS395 (400 ng) overnight. Next morning, the transfection mix was replaced by standard cell culture medium. After 24 h, antibiotic selection was started by addition of 2 μg ml−1 puromycin (Cat. no. A1113803; ThermoFisher Scientific) and 10 μgml−1 blasticidin (Cat. no. A1113903; ThermoFisher Scientific). Single cells were sorted using standard FACS according to the expression level of the integrated fluorophores. Single cells were seeded into wells of a 96-well plate containing conditioned media, where they were grown for two weeks before testing their performance. Similarly, the AIGESIns monoclonal cell line was generated by transfecting HEK-293T cells with the plasmids pAB804, pBS894 and pTS395 and selected with 100 μg ml−1 zeocin (Cat. no. R25005; ThermoFisher Scientific).
Animal experiments. For intraperitoneal implantation into mice, the transgenic HEK cells were alginate-encapsulated using an Inotech Encapsulator Research Unit IE-50R (EncapBioSystem Inc., Greifensee, Switzerland) with a 200 μm nozzle, a vibration frequency of 1,024 Hz, 1200 V and a flow rate of 400 units using a 25 ml syringe [21]. Functionality of encapsulated cells was tested in vitro and capsule homogeneity as well as integrity was confirmed by means of microscopy. 25 000–50 000 capsules per mouse containing 200 cells per capsule were implanted intraperitoneally. All experiments involving animals were performed according to the directive of the European Community Council ($\frac{2010}{63}$/EU), approved by the French Republic (project no. DR2018-40v5, APAFIS #16753), and carried out by Ghislaine Charpin-El Hamri (license no. 69266309) at the Institut Universitaire de Technologie of the Université Claude Bernard Lyon 1, F-696226, Villeurbanne Cedex, France.
In vivo induction protocols. Encapsulated cells were implanted intraperitoneally into C57BL/6 mice (female, 4–6 weeks, Janvier) in 1 ml of MOPS buffer. For the experimental type-1 diabetes study, we used C57BL/6 mice (male, 4–6 weeks, Janvier) treated with 70 mg/kg per mouse per day streptozocin (STZ, Cat. No. S0130, Sigma-Aldrich) for 4 days to deplete the insulin production of beta cells [22]. Cells were implanted 24 h prior to the induction for all experiments. The mice were then treated with 0–7500 mg/kg acetoin dissolved in 100 mM Tris–HCl (Tris base, Cat. No. 200923, Biosolve; HCl, Cat. No. 320331, Sigma-Aldrich) at pH 7.2 by oral gavage or given lychees, liquid butter or cheddar cheese, purchased at a local supermarket. Lychees in the form of a smoothie and butter were given by oral gavage (200 μl per mouse), while cheese was presented as a block after removal of other food sources; 18 g of cheese was consumed overnight by 6 mice. For induction of the encapsulated cells via an inhaler (EMSER, Sidro AG, Rheinfelden, Switzerland), a solution of 600 mg/ml acetoin in 100 mM Tris–HCl was vaporized into a 11.5 × 6 × 8 cm box with an induction cycle of 150 s on, 150 s off for 0–30 min. All data was collected at 12 h after the induction, unless otherwise stated.
GTT Assay. Mice were fasted for 8 h before i.p. injection with a glucose solution (1.5 g/kg). After the glucose injection, blood glucose level was monitored every 15 min for 60 min and every 30 min for another 60 min using a glucometer (Contour XT, BAYER HealthCare, Leverkusen, Germany) to generate a glycemia profile over the time course of 120 min.
Insulin ELISA. Ultrasensitive Mouse Insulin ELISA (Cat. No. 10-1132-01, Mercodia) was used according to the manufacturer's instructions to measure blood insulin levels. Absorbance was measured at 450 nm using a Tecan Infinite M1000 microplate reader.
Inflammation marker assays. Encapsulated cells were implanted into TD1 mice, cells were induced by oral gavage application of acetoin [5 g/kg] body weight at 2 and 4 days after cell implantation and blood serum was collected at 12 h after the second acetoin application. Blood serum was then analyzed using a mouse IL-6 high-sensitivity ELISA kit (BMS603HS, Invitrogen), a mouse INF alpha high-sensitivity ELISA kit (BMS607HS, Invitrogen) and a mouse IFN gamma ELISA kit (BMS606-2, Invitrogen), according to the manufacturer's instructions.
Statistical analysis. Variation was determined using standard deviation and presented as error bars within the graphs. Standard error of mean was used to present the variation of the in vivo data. To examine significance, statistical evaluation was conducted using the unpaired two-tailed student t-test and one-way ANOVA analysis for comparison of two datasets or multiple datasets respectively. Results are indicated in the graph as followed: *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$ All statistical evaluations were conducted using the implemented algorithms of Graphpad Prism 8 (GraphPad Software Inc., San Diego, California, USA) Data availability. All plasmids and data generated in this study are available on request. Requests for materials should be made to the corresponding author.
## Design and in vitro evaluation of the acetoin-inducible gene switch
To engineer an acetoin-responsive mammalian transactivator (AceA), we fused the *Bacillus subtilis* AcoR protein C’-terminally to the minimal acidic activation domain of the herpes simplex virus VP16 transactivation domain (VP16f-type). To test the ability of AceA to regulate gene expression in mammalian cells, we designed a synthetic promoter consisting of the full 30 bp operator sequence previously reported for AcoR [16] upstream of a minimal version of the human cytomegalovirus immediate-early promoter (PhCMVmin) and used it to drive expression of human placental secreted alkaline phosphatase (SEAP) as a reporter protein (Figure 1A). Human embryonic kidney (HEK293-T) cells co-transfected with this reporter plasmid (PhCMVmin-SEAP-pA; pAB008) and a construct encoding constitutive expression of AceA driven by the mouse phosphoglycerate kinase 1 promoter (PmPGK-AceA; pAB003) showed induction of SEAP expression in the presence of acetoin (Figure 1B).
**Figure 1.:** *(A) Schematic overview of the design and function of the acetoin-inducible gene switch (AIGES). The Bacillus subtilis-derived acetoin transcriptional regulator AcoR was C’-terminally fused to a minimal version of the VP16 viral transactivation domain f-type (VP16f-type). This fusion protein (VP16f-type-AcoR), which we called acetoin-responsive mammalian transactivator AceA, is controlled by the constitutive human phosphoglycerate kinase promoter (PmPGK) (pAB003, PmPGK-AceA-pA). A reported AcoR binding region was cloned 5’ to the minimal human cytomegalovirus immediate early promoter (PhCMVmin) to control the expression of the reporter protein, secreted alkaline phosphatase (SEAP), in an acetoin-responsive manner (pAB008, OAceAPhCMVmin-SEAP-pA). AceA interacts with acetoin and binds the response element to start transcription of SEAP. (B) SEAP levels in the supernatants of HEK cells transfected with the AIGES plasmids (pAB003 and pAB008), cultured in medium either containing acetoin (10 mM), or in acetoin-free medium, and then left for 24 h before SEAP measurement. Due to the volatility of acetoin, induced and non-induced cells were cultured in separate plates and kept in different shelves inside the incubator. (C) Characterization of the binding motif of AcoR within the reported binding region. SEAP secretion from HEK cells co-transfected with pAB003 and the natural operator sequence of Bacillus subtilis (pAB008), the consensus sequence of five Bacillales (pAB101) or variants of truncations of the palindromic consensus sequence (pAB102-pAB111), which comprises two homologous pairs of binding sites, 1 and 2, and in combination with 1’ and 2’ form a palindromic repeat. (D) The specificity of AceA for acetoin was demonstrated by adding structurally similar molecules and measuring the reporter response. Toxicity of all tested compounds was measured using a resazurin cell viability assay. All experiments were performed in triplicates and repeated at least three times.*
To characterize the binding region of AceA, we analyzed the AcoR operator site consensus sequence (pAB101) from five different Bacillales (B. subtilis, B. licheniformis, B. pseudomonas, B. cereus and B. pumilus) using Jalview [23] and replaced different regions with a confirmed non-binding nucleotide sequence of the same length. The consensus sequence reveals two regions in each strand, designated 1 and 2 or 1’ and 2’, that form a palindrome (Figure 1C). Removal of either 1 or 2 results in the loss of AceA-regulated gene expression and inducibility by acetoin. Replacing both 1’ and 2’ results in the loss of induction, but the same level of basal expression is retained. The replacement of either 1’ or 2’ does not impair the induction of gene expression by acetoin. However, the replacement of 2’ increased the basal expression by a factor of 5. Based on these results, we selected the full consensus sequence as the AceA response element to place upstream of a minimal promoter controlling reporter gene expression (OAceAPhCMVmin-SEAP-pA; pAB101).
Next, to evaluate the specificity of the AceA towards acetoin, we tested if a set of structurally similar molecules can activate reporter gene expression in HEK293T cells transfected with the acetoin-inducible gene switch (AIGES). None of the tested compounds increased SEAP production (Figure 1D). Resazurin assay showed that the tested compounds had no effect on cell viability at the concentration of 10 mM. These results indicate high selectivity of AceA for acetoin, which is desirable for precise regulation of gene expression. Furthermore, RNA sequencing only revealed a total of 156 differentially expressed genes out of over 60 000 analyzed genes with a false discovery rate <$20\%$. Both the fold changes and expression intensities of those genes were low, suggesting that acetoin had little or no effect other cellular pathways (Supplementary Figure S1).
## Optimization of the acetoin-inducible gene switch
With the aim of further enhancing the performance of the AIGES, we next tested various modifications of AceA and the reporter construct. First, we compared the performance of the AcoR proteins from B. subtilis, B. licheniformis, B. cereus and B. pumilus fused to the VP16f-type transactivation domain in HEK293T cells co-transfected with the reporter plasmid. Except for the B. cereus protein, all other AcoR proteins showed acetoin-responsiveness and B. subtilis AcoR showed the greatest induction (Figure 2A). Although HEK293T cells expressed the proteins derived from various Bacillales in different quantities (Supplementary Figure S2), the most highly expressed protein did not afford the highest reporter protein expression in the functional assay (Figure 2A). The observed differences in induction of the reporter are most likely due to a combination of different protein expression levels and different affinities of the AcoR variants for the DNA binding sites.
**Figure 2.:** *(A) Comparison of the ability of AcoR proteins from different Bacillales fused to VP16f-type to activate SEAP expression in the presence of acetoin. (B) Effect of C- and N-terminal fusion of the viral transactivators VPR, VP16 and VP16f-type on SEAP expression. (C) VP16f-type-AcoR fusion under the control of different constitutive promoters and its effect on SEAP expression. (D) Influence of different spacers between the operator site and the minimal promoter. (E) Effect of multiple operator sites on SEAP expression. (F) Impact of different minimal promoters on the performance of the acetoin-inducible gene switch. (G) Acetoin-inducible gene expression across various mammalian cell lines transfected with pAB003 and the reporter pAB101. SEAP activity was assayed after 24 h incubation of HEK and CHO-K1 cells, and after 72 h incubation for the remaining cell lines. (H) Orthogonality of the acetoin and vanillic acid gene switches. HEK cells were co-transfected with both the switch ON AIGES system and the switch OFF system based on vanillic acid (VA) responsive transcription factor (PhCMV-VanR-VP16-pA), which in the absence of VA binds to a synthetic promoter controlling NLuc expression (OVanRPmin-NLuc-pA). The dose-response of SEAP and NLuc reporter proteins was evaluated in medium containing acetoin and vanillic acid. All experiments were performed in triplicate and repeated at least three times.*
Next, we fused different transactivation domains, namely VPR, VP16 and VP16f-type C- and N-terminally to AcoR to screen for the best fold induction (Figure 2B). We then placed VP16f-type-AcoR under the control of the different constitutive promoters mPGK, hSV40 and hCMV and evaluated SEAP expression (Figure 2C). To optimize the reporter plasmid, we constructed plasmids with different spacers between the minimal promoter and the AceA binding sequence (Figure 2D), or with one to four tandem repeats of the binding sequence (Figure 2E), or with different minimal promoters (Figure 2F). Increasing the distance from the binding sequence to the minimal promoter, as well as increasing the number of binding sequences increased the expression level, but led to lower fold changes due to higher basal expression of the reporter protein. Among the different minimal promoters tested, we selected hCMVmin for the follow-up experiments, as it showed the highest expression level of the reporter protein together with good fold induction.
Finally, we transfected eight different mammalian cell lines with the best-performing acetoin gene switch to evaluate its broader applicability. All the cell lines tested showed increased SEAP production upon exposure to acetoin (Figure 2G).
In order to examine the functionality of the acetoin-based gene switch in combination with another flavor-based gene switch, the vanillic acid VanR-VP16 switch OFF system, we co-expressed AceA, VanR-VP16 and two reporter proteins, SEAP under the control of AceA and nano luciferase (NLuc) under the control of VanR, in the same cells. With increasing concentrations of both flavor molecules (vanillic acid and acetoin) the expression of NLuc decreased while the expression of SEAP increased, demonstrating the compatibility of the two flavor-based transcriptional regulation systems (Figure 2H). Overall SEAP expression is reduced due to higher protein loads and potential squelching when both systems are used simultaneously, as both systems rely on the acidic viral transactivator VP16 [24,25].
## Generation of transgenic stable cell line for the acetoin gene switch
We established a HEK293T cell population with both components of the acetoin-inducible gene switch (PmPGK-AceA (pAB800) and OAceAPhCMVmin-SEAP (pAB801)) stably integrated in their genome using the Sleeping Beauty transposase system [20]. Monoclonal cell lines were isolated and screened to select the one (HEK-AIGES) with the highest induced expression and lowest basal expression of the reporter protein SEAP (Supplementary Figure S3). We measured the dose-response curve of the HEK-AIGES cell line in response to acetoin (Figure 3A). The response was approximately linear between 3 mM and 33 mM with an EC50 of 10.22 mM. Cells cultured with different acetoin concentrations in this range showed no significant differences in cell viability, confirming that acetoin is not cytotoxic in this range (Supplementary Figure S4). To examine the time-dependence of reporter expression after exposure to acetoin, we measured SEAP expression over a period of 32 h after exposure to acetoin. SEAP production was observed after 8 h, and was significantly upregulated between 12 and 16 h post exposure (Figure 3B). Off-kinetics of the HEK-AIGES cell line was tested by exposing the cells to acetoin (10 mM) for 0, 8, 16 or 24 h. After the exposure, the medium was exchanged to acetoin-free medium and reporter gene production was measured over a period of 48 h (Figure 3C). Cells were co-transfected with a second reporter gene (NLuc) controlled by the constitutive promotor SV40 to ensure the off-kinetic observations were not affected by loss of cell viability over time due to overgrowth of the cells (Supplementary Figure S5). The overall induction level was positively correlated with the induction time.
**Figure 3.:** *(A) Characterization of the SEAP levels produced by the stable monoclonal cell line HEK-AIGES when cultured in the presence of different acetoin concentrations. (B) Activation dynamics after exposure of HEK-AIGES cells to acetoin at different concentrations (0, 5, 10 and 50 mM). SEAP expression was monitored over a period of 32 h following the exposure to acetoin, with a sampling interval of 4 h. (C) HEK-AIGES cells were exposed to 10 mM acetoin for 0, 8, 16 and 24 h before the medium was exchanged to standard cell culture medium without acetoin, and SEAP expression was monitored over a period of 48 h at intervals of 8 h. (D) To test for reversibility, HEK-AIGES cells were exposed to acetoin (10 mM) or no acetoin for 8 h before the medium was exchanged. Cells were grown for an additional 16 h, then trypsinized and reseeded into fresh medium to maintain a stable cell number of 15 000 cells per well of a 96-well plate. Reseeding as well as media changes removed all SEAP in the medium, resetting the SEAP levels. The medium in which the cells were reseeded either contained acetoin (10 mM) or no acetoin. After 8 h the medium was exchanged again. This was repeated over a period of 120 h using two different protocols, starting either with acetoin (ON-OFF-ON-OFF-ON) or no acetoin (OFF-ON-OFF-ON-OFF). Samples were taken every 8 h. (E) Picture and heatmap of SEAP expression in a 96-well plate after volatile induction. Solid acetoin (10 mg) was placed in one dry well of a 96-well plate while other wells contained HEK-AIGES cells. SEAP level in each well was measured 24 h after acetoin addition. All experiments were performed in triplicate and repeated at least three times.*
To assess the reversibility of the gene switch, we repeatedly cultured the stable monoclonal cell line in the presence (ON) or absence (OFF) of acetoin for 8 h, then changed the medium to acetoin-free medium and followed SEAP expression for the next 16 h. After each 24-h cycle, we reseeded the cells to maintain a stable cell number before starting the next cycle. The results of five successive 24-h cycles using acetoin ON-OFF-ON-OFF-ON and OFF-ON-OFF-ON-OFF patterns are shown in Figure 3D. The results confirm that expression of SEAP can be reversibly controlled over multiple cycles.
Finally, we assessed whether AIGES cells would also be responsive to acetoin vapor generated by diffusion of solid acetoin. We placed acetoin powder in a middle well of 96-well plates. The acetoin spontaneously diffused and induced SEAP expression in cells seeded in the surrounding wells, in a distance-dependent manner (Figure 3E).
## In vivo performance of the AIGES to treat T1DM.
To test the potential therapeutic applicability of the acetoin-inducible gene network, we injected alginate-encapsulated HEK-AIGES cells intraperitoneally (i.p.) in mice and measured the SEAP levels in the bloodstream in response to administration of acetoin by different means. Mice treated with an acetoin solution delivered via oral gavage showed increased blood SEAP levels in a dose-dependent manner (Figure 4A). To examine the kinetics of our gene switch, we examined acetoin depletion (Supplementary Figure S6), as well as SEAP and insulin kinetics after a single dose of acetoin [5 mg/kg] (Supplementary Figure S7). The intake of different acetoin-containing foods, namely lychees [26], cheddar cheese [27,28] and butter [29], did not activate SEAP expression (Figure 4B). Cell implants were also tested at three and five days after injection to check the longevity of the encapsulated cells (Supplementary Figure S8). Cells retrieved after being implanted were compared to cells cultured in vitro (Supplementary Figure S9) and inflammation markers were measured in mouse serum at five days after implantation (Supplementary Figure S10). No loss in functionality or signs of inflammation were observed. Mice exposed to acetoin aerosol generated by a portable inhaler containing an acetoin-rich solution (Figure 4C) for 450, 675 or 900 s showed an exposure time-dependent increase of blood SEAP levels (Figure 4D). In order to test the applicability of inhaler-based induction of the AIGES system in a disease model, we engineered a stable cell line with acetoin-responsive insulin production and chose the best monoclonal cell line (AIGESIns) based on basal and total expression (Supplementary Figure S11) and tested for dose-dependent insulin production capacity (Supplementary Figure S12). We then injected type-I diabetic mice with encapsulated AIGESIns cells and exposed them to an acetoin-containing aerosol generated by an inhaler. Mice exposed to acetoin aerosol showed lower initial and peak glucose levels, as well as a faster decrease in blood glucose level after the spike in GTT assay, resembling the behavior seen in the non-diabetes control group (Figure 4E). To evaluate the full capacity of the AIGESIns cells we induced mice by exposure to acetoin aerosol for 900 s, which resulted in high insulin levels (Figure 4F) and corresponding low fasting-glucose levels (Figure 4G) at 24 h post treatment.
**Figure 4.:** *(A) SEAP levels in the bloodstream of wild-type mice with intraperitoneally implanted encapsulated HEK-AIGES cells following acetoin administration. Acetoin was delivered via oral gavage at the indicated doses. Mice without HEK-AIGES cell implants were used as negative controls. (B) SEAP production when mice were fed lychees, cheese or butter, which naturally contain acetoin. The lowest functional dosage of oral gavage-applied acetoin was used as a positive control. (C) Image showing the setup for inhaler-based induction. (D) Inhaler-based in vivo induction was conducted by vaporizing an acetoin containing solution (600 mg/ml) with repeated intervals of switching the inhaler on for 150 s and off for 150 s, for a total time of 900, 1350 and 1800 s, resulting in induction times of 450, 675 and 900 s. Waiting times were included to minimize oversaturation of the air with moisture. (E) Glucose tolerance test was performed by administering 1.5 g/kg glucose solution i.p. and the glycemia profile was generated by monitoring blood glucose levels over a period of 120 min. (F) Insulin levels were measured 24 h after acetoin treatment by assessing the serum insulin levels with an ELISA kit. (G) Blood glucose levels were measured before and 24 h after acetoin treatment. For panels (E), (F) and (G) acetoin administration was performed with the 900-s inhaler protocol. All experiments were performed twice with 8 mice per treatment group and 6 mice per control group. To test for significance, statistical analysis using one-way ANOVA was performed comparing all datasets of one representative biological replicate to the uninduced dataset (0 mM acetoin) in panels A–D.*
## DISCUSSION
There have been major advances in cell-based therapies in recent years, with several engineered T cell therapies approved to treat blood cancers, and many more in the pipeline (30–34). Some of these therapies need to be controlled by an external trigger to increase their safety. Ideally, a gene regulation system for cell-based therapies would be orthogonal and finely tunable by a non-toxic inducer that has a short half-life time in vivo and is physiologically inert and/or clinically licensed. While clinically licensed drugs are designed to have therapeutic effects (35–37), non-orthogonal inducers such as vitamins [38] and metabolites such as amino acids [39,40] have to be used at concentrations well above physiological levels to ensure control over gene expression. However, prolonged exposures to toxic or physiologically unnatural inducer levels pose the risk of severe side effects. To increase patient compliance and to avoid the need for medically trained persons to administer the inducer, a simple mode of administration of the inducer is essential for cell-based therapy.
Inhalation is an ancient route of drug administration. The first evidence of inhalation therapy was traced back 4000 years in India [41], but it was not until the middle of the 19th century when its significance was discovered for the treatment of lung diseases with the invention of the glass bulb nebulizer [41]. Today, inhalers are integral to today's treatment options for asthma or chronic obstructive pulmonary disease [42]. Indeed, many products are now commercially available to ease the administration of a drug or small molecule, including nicotine inhalers for nicotine reduction treatment [43].
The key to a successful implementation of gene- or cell-based therapies is to increase not only the patient's life expectancy, but also quality of life. As these therapies replace lost functionality, such as the expression of proteins or peptide hormones, a simple means of controlling their expression is critical, as it will be a regular part of the patient's life. In addition, it is of great importance that engineered cells are not triggered to produce the therapeutic protein unintentionally by any external stimulus. This is the main drawback of physically induced systems based on triggers such as heat, light or sound. Even though induction of these systems is very simple, avoiding unintended stimulation would have a drastic impact on the patient's life. *Controlling* gene expression via inhalation is therefore attractive. Previous inducible gene regulatory systems based on gaseous inducers are either too toxic for in vivo use (e.g. acetaldehyde [44]) or were designed to react to external stimuli present in cosmetic products or consumer goods used in everyday life (e.g. spearmint essential oil or chewing gum containing spearmint [2]), and therefore require constant attentiveness.
In contrast, our AIGES system is characterized by high inducibility and tight control without unspecific induction, combined with an easy route of trigger administration, and thus could be a very attractive option for future gene- and cell-therapies. The use of an inducer molecule not available from natural sources is advantageous, though relatively large amounts of the inducer are needed to activate gene transcription. One approach to enable good activation with a smaller quantity of inducer would be to select the location of the implant for optimal induction properties. Further studies are also needed to examine cell implant survival, in particular to avoid fibrotic overgrowth, a typical foreign-body response [45,46].
Given that the acetoin inducer is inexpensive and is already an FDA-approved food additive, AIGES could be a useful tool to regulate gene expression for therapeutic protein production in biphasic production processes where the cell expansion phase is decoupled from the protein production phase [47]. The potential of AIGES for use in cell-based therapies was confirmed here in a proof-of-concept experiment using a type-1 diabetes mouse model, in which it effectively normalized blood glucose levels.
## DATA AVAILABILITY
All plasmids and data generated in this study are available on request. Requests for materials should be made to the corresponding author.
## SUPPLEMENTARY DATA
Supplementary Data are available at NAR Online.
## FUNDING
European Research Council advanced grant [ElectroGene, no. 785800]; Swiss National Science Foundation NCCR Molecular Systems Engineering (in part). Funding for open access charge: ETH Zürich.
Conflict of interest statement. None declared.
## References
1. Bojar D., Scheller L., Hamri G.C.-E., Xie M., Fussenegger M.. **Caffeine-inducible gene switches controlling experimental diabetes**. *Nat. Commun.* (2018) **9** 2318. PMID: 29921872
2. Wang H., Xie M., Charpin-El Hamri G., Ye H., Fussenegger M. **Treatment of chronic pain by designer cells controlled by spearmint aromatherapy**. *Nat. Biomed. Eng.* (2018) **2** 114-123. PMID: 31015627
3. Bai P., Liu Y., Xue S., Hamri G.C.-E., Saxena P., Ye H., Xie M., Fussenegger M.. **A fully human transgene switch to regulate therapeutic protein production by cooling sensation**. *Nat. Med.* (2019) **25** 1266-1273. PMID: 31285633
4. Stefanov B.A., Teixeira A.P., Mansouri M., Bertschi A., Krawczyk K., Hamri G.C.E., Xue S., Fussenegger M.. **Genetically encoded protein thermometer enables precise electrothermal control of transgene expression**. *Adv. Sci.* (2021) **8** 2101813
5. Mansouri M., Hussherr M.-D., Strittmatter T., Buchmann P., Xue S., Camenisch G., Fussenegger M.. **Smart-watch-programmed green-light-operated percutaneous control of therapeutic transgenes**. *Nat. Commun.* (2021) **12** 3388. PMID: 34099676
6. Krawczyk K., Xue S., Buchmann P., Charpin-El-Hamri G., Saxena P., Hussherr M.-D., Shao J., Ye H., Xie M., Fussenegger M.. **Electrogenetic cellular insulin release for real-time glycemic control in type 1 diabetic mice**. *Science* (2020) **368** 993-1001. PMID: 32467389
7. Gitzinger M., Kemmer C., Fluri D.A., El-Baba Daoud, Weber M., Fussenegger M. **The food additive vanillic acid controls transgene expression in mammalian cells and mice**. *Nucleic Acids Res.* (2012) **40** e37-e37. PMID: 22187155
8. Ye H., Hamri Charpin-El, Zwicky G., Christen K., Folcher M., Fussenegger M. **Pharmaceutically controlled designer circuit for the treatment of the metabolic syndrome**. *Proc. Natl. Acad. Sci. U.S.A.* (2013) **110** 141. PMID: 23248313
9. Gitzinger M., Kemmer C., El-Baba M.D., Weber W., Fussenegger M.. **Controlling transgene expression in subcutaneous implants using a skin lotion containing the apple metabolite phloretin**. *Proc. Natl. Acad. Sci. U.S.A.* (2009) **106** 10638. PMID: 19549857
10. Dietz F.K., Rodriguez-Giaxola M., Traiger G.J., Stella V.J., Himmelstein K.J.. **Pharmacokinetics of 2-butanol and its metabolites in the rat**. *J. Pharmacokinet. Biopharm.* (2005) **9** 553-576
11. Otsuka M., Harada N., Itabashi T., Ohmori S.. **Blood and urinary levels of ethanol, acetaldehyde, and C4 compounds such as diacetyl, acetoin, and 2,3-butanediol in normal male students after ethanol ingestion**. *Alcohol* (1999) **17** 119-124. PMID: 10064379
12. Huang M., Fred Oppermann-Sanio, Steinbüchel A. **Biochemical and molecular characterization of the**. *J. Bacteriol.* (1999) **181** 3837-3841. PMID: 10368162
13. Krüger N., Steinbüchel A.. **Identification of acoR, a regulatory gene for the expression of genes essential for acetoin catabolism in Alcaligenes eutrophus H16**. *J. Bacteriol.* (1992) **174** 4391-4400. PMID: 1378052
14. Zhang L.L., Fan G., Li X., Ren J.N., Huang W., Pan S.Y., He J.. **Identification of functional genes associated with the biotransformation of limonene to trans-dihydrocarvone in**. *J. Sci. Food Agric.* (2021) **102** 3297-3307. PMID: 34800295
15. Ali Ould, Bignon N., Rapoport J., Debarbouille M. **Regulation of the acetoin catabolic pathway is controlled by sigma L in**. *J. Bacteriol.* (2001) **183** 2497-2504. PMID: 11274109
16. Peng Q., Zhao X., Wen J., Huang M., Zhang J., Song F.. **Transcription in the acetoin catabolic pathway is regulated by AcoR and CcpA in**. *Microbiol. Res.* (2020) **235** 126438. PMID: 32088504
17. Simonsen J.L., Rosada C., Serakinci N., Justesen J., Stenderup K., Rattan S.I.S., Jensen T.G., Kassem M.. **Telomerase expression extends the proliferative life-span and maintains the osteogenic potential of human bone marrow stromal cells**. *Nat. Biotechnol.* (2002) **20** 592-596. PMID: 12042863
18. Berger J., Hauber J., Hauber R., Geiger R., Cullen B.R.. **Secreted placental alkaline phosphatase: a powerful new quantitative indicator of gene expression in eukaryotic cells**. *Gene* (1988) **66** 1-10. PMID: 3417148
19. Haellman V., Strittmatter T., Bertschi A., Stücheli P., Fussenegger M.. **A versatile plasmid architecture for mammalian synthetic biology (VAMSyB)**. *Metab. Eng.* (2021) **66** 41-50. PMID: 33857582
20. Mátés L., Chuah M.K., Belay E., Jerchow B., Manoj N., Acosta-Sanchez A., Grzela D.P., Schmitt A., Becker K., Matrai J.. **Molecular evolution of a novel hyperactive Sleeping Beauty transposase enables robust stable gene transfer in vertebrates**. *Nat. Genet.* (2009) **41** 753-761. PMID: 19412179
21. Xie M., Ye H., Wang H., Hamri Charpin-El, Lormeau G., Saxena C., Stelling P., Fussenegger M. **β-cell–mimetic designer cells provide closed-loop glycemic control**. *Science* (2016) **354** 1296-1301. PMID: 27940875
22. Wu K.K., Huan Y.. **Diabetic atherosclerosis mouse models**. *Atherosclerosis* (2007) **191** 241-249. PMID: 16979174
23. Waterhouse A.M., Procter J.B., Martin D.M.A., Clamp M., Barton G.J.. **Jalview Version 2—a multiple sequence alignment editor and analysis workbench**. *Bioinformatics* (2009) **25** 1189-1191. PMID: 19151095
24. Cahill M.A., Ernst W.H., Janknecht R., Nordheim A.. **Regulatory squelching**. *FEBS Lett.* (1994) **344** 105-108. PMID: 8187867
25. Matis C., Chomez P., Picard J., Rezsohazy R.. **Differential and opposed transcriptional effects of protein fusions containing the VP16 activation domain**. *FEBS Lett.* (2001) **499** 92-96. PMID: 11418119
26. Xiao Z., Lu J.R.. **Generation of acetoin and its derivatives in foods**. *J. Agric. Food Chem.* (2014) **62** 6487-6497. PMID: 25000216
27. Iwasawa A., Suzuki-Iwashima A., Iida F., Shiota M.. **Effects of flavor and texture on the desirability of cheddar cheese during ripening**. *Food Sci. Technol. Res.* (2014) **20** 23-29
28. Andiç S., Tunçtürk Y., Boran G., Preedy V.. *Processing and Impact on Active Components in Food* (2015) 231-239
29. Winter M., Stoll M., Warnhoff E., Greuter F., Büchi G.. **Volatile carbonyl constituents of dairy butter**. *J. Food Sci.* (1963) **28** 554-561
30. Leibovitch J.N., Tambe A.V., Cimpeanu E., Poplawska M., Jafri F., Dutta D., Lim S.H.. **l-glutamine, crizanlizumab, voxelotor, and cell-based therapy for adult sickle cell disease: hype or hope?**. *Blood Rev.* (2022) 100925. PMID: 34991920
31. Maude S.L., Frey N., Shaw P.A., Aplenc R., Barrett D.M., Bunin N.J., Chew A., Gonzalez V.E., Zheng Z., Lacey S.F.. **Chimeric Antigen Receptor T Cells for Sustained Remissions in Leukemia**. *N. Engl. J. Med.* (2014) **371** 1507-1517. PMID: 25317870
32. Turtle C.J., Hanafi L.-A., Berger C., Gooley T.A., Cherian S., Hudecek M., Sommermeyer D., Melville K., Pender B., Budiarto T.M.. **CD19 CAR–T cells of defined CD4+: CD8+ composition in adult B cell ALL patients**. *J. Clin. Invest.* (2016) **126** 2123-2138. PMID: 27111235
33. Grupp S.A., Kalos M., Barrett D., Aplenc R., Porter D.L., Rheingold S.R., Teachey D.T., Chew A., Hauck B., Wright J.F.. **Chimeric antigen receptor–modified T cells for acute lymphoid leukemia**. *N. Engl. J. Med.* (2013) **368** 1509-1518. PMID: 23527958
34. Turtle C.J., Hanafi L.-A., Berger C., Hudecek M., Pender B., Robinson E., Hawkins R., Chaney C., Cherian S., Chen X.. **Immunotherapy of non-Hodgkin's lymphoma with a defined ratio of CD8+ and CD4+ CD19-specific chimeric antigen receptor–modified T cells**. *Sci. Transl. Med.* (2016) **8** 355ra116
35. Gossen M., Freundlieb S., Bender G., Müller G., Hillen W., Bujard H.. **Transcriptional activation by tetracyclines in mammalian cells**. *Science* (1995) **268** 1766-1769. PMID: 7792603
36. Palli S.R., Kapitskaya M.Z., Potter D.W.. **The influence of heterodimer partner ultraspiracle/retinoid X receptor on the function of ecdysone receptor**. *FEBS J.* (2005) **272** 5979-5990. PMID: 16302963
37. Tascou S., Sorensen T.-K., Glénat V., Wang M., Lakich M.M., Darteil R., Vigne E., Thuillier V.. **Stringent rosiglitazone-dependent gene switch in muscle cells without effect on myogenic differentiation**. *Mol. Ther.* (2004) **9** 637-649. PMID: 15120324
38. Weber W., Lienhart C., Daoud-El Baba M., Fussenegger M. **A biotin-triggered genetic switch in mammalian cells and mice**. *Metab. Eng.* (2009) **11** 117-124. PMID: 19271268
39. Hartenbach S., Daoud-El Baba M., Weber W., Fussenegger M. **An engineered L-arginine sensor of Chlamydia pneumoniae enables arginine-adjustable transcription control in mammalian cells and mice**. *Nucleic Acids Res.* (2007) **35** e136-e136. PMID: 17947334
40. Bacchus W., Lang M., El-Baba M.D., Weber W., Stelling J., Fussenegger M.. **Synthetic two-way communication between mammalian cells**. *Nat. Biotechnol.* (2012) **30** 991-996. PMID: 22983089
41. Grossman J.. **The evolution of inhaler technology**. *J. Asthma* (1994) **31** 55-64. PMID: 8175626
42. Geller D.E.. **Comparing clinical features of the nebulizer, metered-dose inhaler, and dry powder inhaler**. *Respir. Care* (2005) **50** 1313-1322. PMID: 16185367
43. Schneider N.G., Olmstead R.E., Franzon M.A., Lunell E.. **The nicotine inhaler**. *Clin. Pharmacokinet.* (2001) **40** 661-684. PMID: 11605715
44. Weber W., Rimann M., Spielmann M., Keller B., Baba M.D.-E., Aubel D., Weber C.C., Fussenegger M.. **Gas-inducible transgene expression in mammalian cells and mice**. *Nat. Biotechnol.* (2004) **22** 1440-1444. PMID: 15502819
45. Vaithilingam V., Fung C., Ratnapala S., Foster J., Vaghjiani V., Manuelpillai U., Tuch B.E.. **Characterisation of the xenogeneic immune response to microencapsulated fetal pig islet-like cell clusters transplanted into immunocompetent C57BL/6 Mice**. *PLoS One* (2013) **8** e59120. PMID: 23554983
46. Vegas A.J., Veiseh O., Gürtler M., Millman J.R., Pagliuca F.W., Bader A.R., Doloff J.C., Li J., Chen M., Olejnik K.. **Long-term glycemic control using polymer-encapsulated human stem cell-derived beta cells in immune-competent mice**. *Nat. Med.* (2016) **22** 306-311. PMID: 26808346
47. Deer J.R., Allison D.S.. **High-level expression of proteins in mammalian cells using transcription regulatory sequences from the Chinese hamster EF-1α gene**. *Biotechnol. Progr.* (2004) **20** 880-889
|
---
title: 'The Effectiveness of an App (Insulia) in Recommending Basal Insulin Doses
for French Patients With Type 2 Diabetes Mellitus: Longitudinal Observational Study'
journal: JMIR Diabetes
year: 2023
pmcid: PMC10018375
doi: 10.2196/44277
license: CC BY 4.0
---
# The Effectiveness of an App (Insulia) in Recommending Basal Insulin Doses for French Patients With Type 2 Diabetes Mellitus: Longitudinal Observational Study
## Abstract
### Background
For patients with type 2 diabetes (T2D), calculating the daily dose of basal insulin may be challenging. Insulia is a digital remote monitoring solution that uses clinical algorithms to recommend basal insulin doses. A predecessor device was evaluated in the TeleDiab-2 randomized controlled trial, showing that a higher percentage of patients using the app achieved their target fasting blood glucose (FBG) level compared to the control group, and insulin doses were adjusted to higher levels without hypoglycemia.
### Objective
This study aims to analyze how the glycemic control of Insulia users has evolved when using the app in a real-life setting in France.
### Methods
A retrospective observational analysis of data collected through the device in adult French patients with T2D treated with basal insulin and oral antihyperglycemic agents using the system for ≥6 months was conducted. Analyses were descriptive and distinguished the results in a subpopulation of regular and compliant users of the app. Glycemic outcomes were estimated considering the percentage of patients who achieved their individualized FBG target between 5.5 and 6 months following the initiation of device use, the frequency of hypoglycemia resulting in a treatment change over the 6-month period of exposure, and the evolution of the average hemoglobin A1c (HbA1c) level over the same period.
### Results
Of the 484 users, 373 ($77.1\%$) performed at least one dose calculation. A total of 221 ($59.2\%$) users were men. When app use started, the mean age, BMI, HbA1c, and basal insulin dose were 55.8 (SD 11.9) years, 30.6 (SD 5.9) kg/m2, $10.1\%$ (SD $2.0\%$), and 25.5 (SD 15.8) IU/day, respectively. Over a median use duration of 5.0 ($95\%$ CI 3.8-5.7) months, patients used the system 5.8 (SD 1.6) times per week on average, and $73.4\%$ of their injected doses were consistent with the app’s suggested doses. Among regular and compliant user patients ($$n = 91$$, ≥5 measurements/week and ≥$80\%$ adherence to calculated doses), $60\%$ ($\frac{55}{91}$) achieved the FBG target (±$5\%$) at 6 months (5.5-6 months) versus $51.5\%$ ($\frac{145}{282}$) of the other patients ($$P \leq .15$$). There was an increase in the proportion of patients achieving their target FBG for regular and compliant users (+$1.86\%$ every 2 weeks) without clear improvement in other patients. A logistic model did not identify the variables that were significantly associated with this outcome among regular and compliant users. In the overall population, the incidence of reported hypoglycemia decreased simultaneously (–$0.16\%$/month). Among 82 patients, the mean HbA1c decreased from $9.9\%$ to $7.2\%$ at 6 months.
### Conclusions
An improvement in glycemic control as measured by the percentage of patients reaching their FBG individualized target range without increasing hypoglycemic risk was observed in patients using the Insulia app, especially among regular users following the dose recommendations of the algorithm.
## Introduction
For patients with type 2 diabetes (T2D), achieving recommended glycemic targets remains difficult, especially in people treated with basal insulin. One of the reasons for this difficulty is related to the challenge of titrating insulin doses. Insulia is a digital solution combining a smartphone app for basal insulin dose suggestions and a web portal accessible to professionals to personalize and manage patients’ treatments remotely. Beyond remote monitoring of basal insulin therapy, the app uses the data entered by the patients to calculate the recommended basal insulin dose according to the objectives set by the patient’s physician. This dose calculation is triggered by the patient’s request.
A predecessor device to Insulia called Diabeo-Basal was evaluated in the TeleDiab-2 study [1]. This randomized controlled trial evaluated the efficacy and safety of two remote monitoring systems to optimize basal insulin initiation in patients with poorly controlled T2D (hemoglobin A1c [HbA1c] $7.5\%$-$10\%$). A total of 191 participants (mean age 58.7 years, mean HbA1c $8.9\%$) were randomized into three groups: group 1 (standard care, $$n = 63$$), group 2 (interactive voice response system, $$n = 64$$), and group 3 (Insulia app software, $$n = 64$$). After 4 months of follow-up, HbA1c reduction was significantly higher in the remote monitoring groups (group 2: –$1.44\%$ and group 3: –$1.48\%$ vs group 1: –$0.92\%$; $$P \leq .002$$). In addition, twice as many patients in the telemonitoring groups achieved their target fasting blood glucose (FBG) level as in the control group, and insulin doses were adjusted to higher levels. No severe hypoglycemia was observed in the remote monitoring groups, and the frequency of mild hypoglycemia was similar in all groups.
Consequently, Insulia was available by prescription and used in France as part of a nationwide program financing new health remote monitoring systems (Expérimentations de télémédecine pour l'amélioration des parcours en santé [ETAPES] program: National Experiments on Remote Diabetes Monitoring [2]) since the end of 2020. Despite the potential benefits for patients suggested by the TeleDiab-2 study, Insulia, as with other apps offering insulin dose calculation, carries a risk of incorrect dose recommendations, which could lead to suboptimal disease control [3]. In this field, perhaps even more than for other health products, it is necessary to complement experimental results with the analysis of monitoring data on both the clinical efficacy and safety of the apps during their use in real life.
The data entered by the patients and their physicians in the Insulia app are collected on a dedicated computer platform. This study aims to analyze this database to determine how the glycemic control of Insulia users has evolved when using the app in a real-life setting in France.
## Overview
The Insulia app is presented in Multimedia Appendix 1. From the Insulia app’s home screen, patients can enter their blood glucose monitoring, hypoglycemia symptoms, and insulin doses. Insulia takes this data into account to recommend personalized doses in real time. Each recommended dose is accompanied by an explanation of how it was calculated. Data is automatically sent to the health care team so that they can monitor the patient’s progress and even adjust the treatment. The ETAPES program funds the device for 6 months for patients with T2D diagnosed more than 12 months ago, who are 18 years or older, with an HbA1c ≥$9\%$ on two measurements taken within a 6-month interval, and treated with insulin. It also funds the device for a maximum of 3 months in patients with T2D diagnosed for more than 12 months who were 18 years or older at the time of insulin initiation when their HbA1c level was <$9\%$ at two measurements taken within a 6-month interval.
A retrospective observational study was conducted using data collected through the Insulia device in adult patients with T2D who were treated with basal insulin and oral antihyperglycemic regimen and who were enrolled as users of the solution for 6 months or more by September 30, 2021.
A subpopulation of regular and compliant users was identified. These are patients who have used the device for at least 6 months without interruption with at least 5 dose calculations per week on average during the study period and for whom more than $80\%$ of their injected insulin doses corresponded to the recommended doses.
Glycemic outcomes were estimated considering the percentage of patients who achieved their individualized glycemic target (average FBG level ±$5\%$) between 5.5 and 6 months following the initiation of the device use.
Other criteria included the frequency of hypoglycemia resulting in a change in treatment over the 6-month period of exposure and the evolution of the average HbA1c level over the same period of time (+/– 1.5 months). HbA1c level was not considered as the primary glycemic outcome as the collection of this data is not mandatory in the app, which determines the insulin doses to be administered based on FBG levels.
A multivariate regression analysis was finally conducted on the achievement of the FBG objective, including a subgroup analysis considering the patients from the center with the most patients versus other patients to identify a possible center effect.
## Ethical Considerations
This study was conducted in accordance with Regulation (EU) $\frac{2016}{679}$ of the European Parliament and of the Council of April 27, 2016, on the protection of natural persons with regard to the processing of personal data and the free movement of such data. Informed consent of the patients was not requested as the data analyzed were fully anonymized. A full privacy impact assessment was conducted on July 29, 2021.
## Results
The Insulia database included 484 patients enrolled as users of the app for 6 months or more. Among them, 111 patients did not conduct any dose calculation with the device or did not indicate any basal insulin dose injected since their registration on the app. Consequently, 373 patients were considered in the main analysis. Among them, 91 ($24.4\%$) patients were identified as regular and compliant users over a 6-month period.
The characteristics of the patients are described in Table 1. On average, they were aged 55.8 (SD 11.9) years, and $59.2\%$ ($$n = 221$$) were men. At the time of their first use of the Insulia device, $48.6\%$ ($$n = 181$$) of them had a BMI ≥30 kg/m2 (average BMI 30.6 kg/m2). The mean HbA1c level was $10.1\%$ (SD $2\%$). The individual FBG target ranged from 70-100 to 100-150 mg/dL. The target ranges were 80-130 mg/dL for $33.5\%$ ($$n = 125$$) of the patients and 80-120 mg/dL for $30\%$ ($$n = 112$$) of the patients. The first calculated basal insulin dose averaged 25.5 IU with significant variability (between 4 IU and 92 IU according to the patients). Among compliant and regular users ($$n = 91$$, $24.4\%$ of the patients over the 6 months of observation), the HbA1c level at baseline was slightly lower compared to other patients ($9.6\%$ vs $10.3\%$; $$P \leq .002$$), and the FBG target was slightly more stringent, with a higher proportion of patients having an FBG target in the range of 80-120 mg/dL and a lower proportion in the range of 100-150 mg/dL.
The percentage of patients defined as regular and compliant users evolved during the 6-month period of the study with a progressive disaffection of the patients from the second month of use. A similar evolution was observed considering only regular use of the device (at least 5 dose calculations per week; Figure 1).
Figure 2 shows the percentage of Insulia users with an average FBG in their individualized target range over 15-day periods according to whether they are regular and compliant users. The percentages are calculated on the number of patients still using the device over the 15-day period. We observed an increase in the proportion of patients achieving their target FBG for regular and compliant users (+$1.86\%$ every 2 weeks). No clear improvement was observed in other patients (irregular or not compliant users of the app).
After 6 months of use (5.5 to 6 months), the FBG target was achieved in $60\%$ ($\frac{55}{91}$) of the regular and compliant users versus $51.5\%$ ($\frac{145}{282}$) of the other patients ($$P \leq .15$$), although the FBG target was slightly more stringent for the regular and compliant users.
Variables available at baseline (age, gender, BMI, HbA1c level, and insulin dose at Insulia initiation) were tested in a logistic model to explain potential factors associated with achieving the individualized FBG target at 6 months among regular and compliant users ($$n = 91$$; Figure 3). None of these variables were significantly associated with this outcome.
Figure 4 presents the evolution of the HbA1c level over time for patients having at least 2 measurements regardless of the time elapsed between these two measurements (all patients, $$n = 182$$, and all regular and compliant users, $$n = 69$$). In both cases, a slight but significant decrease in HbA1c values of –$0.155\%$ and –$0.161\%$ per month, respectively, was observed.
Data were available at baseline and 6 months for only 82 patients. The mean HbA1c level decreased from $9.9\%$ to $7.2\%$ after 6 months (±1.5 months) of app use, with no significant difference according to the degree of Insulia use.
Table 2 shows the numbers and proportions of patients who reported in the app that they had at least one change in treatment because of hypoglycemia, defined as a blood glucose measurement <70 mg/dL and whether this hypoglycemia was symptomatic or not. A favorable trend (–$0.16\%$ per month) was observed but not statistically significant due to the low number of patients.
Finally, to identify a possible center effect, a subgroup analysis was conducted on the sample of patients enrolled in the principal investigating center (Centre Hospitalier Sud Francilien [CHSF], Corbeil-Essonnes) where 68 ($18.3\%$) patients of the overall population considered in the analysis ($$n = 373$$) were enrolled. The baseline characteristics of those patients were similar to those of other centers, excluding a higher average HbA1c level at the time of the first basal insulin dose calculation ($10.9\%$, SD $2.3\%$ vs $10.0\%$, SD $1.9\%$). Interestingly, individualized FBG targets were also often less strict in this center than in other centers (patients with a target between 100-150 mg/dL: $\frac{20}{68}$, $29.4\%$ vs $\frac{12}{305}$, $3.9\%$), and over the first 6 months, patients at the CHSF were less often regular and compliant users than at the other centers ($\frac{9}{68}$, $13.2\%$ vs $\frac{82}{305}$, $26.9\%$; $$P \leq .02$$). Despite these discrepancies, the percentage of Insulia app users achieving their FBG target after 6 months was not different considering the overall population ($$P \leq .77$$) or only regular and compliant users ($$P \leq .75$$), excluding a center effect on the results.
## Principal Findings
More technologies are being developed to assist in outpatient insulin dosing [4], but few of them are intended to adjust long-acting insulins for patients with T2D. After reviewing patients’ data, medical history, comorbidities, and current treatment, providers formulate initial insulin dose and titration plans. A target blood glucose range is defined individually as other criteria including adjustment period, low blood glucose threshold, and maximum total daily doses. Patients are supposed to log their FBG readings and episodes of hypoglycemia events or administered insulin doses. Based on these inputs, the system recalculates the next appropriate dose of basal insulin. Franc et al [1] reported on the TeleDiab-2 trial that, at month 4, twice as many patients using such a device compared to the control group achieved an HbA1c level <$7\%$ ($29.8\%$ vs $12.5\%$). Other similar devices have also shown positive results [5,6]. However, the translation of clinical trial results into real life often raises a series of questions that lead to an interest in conducting postmarketing observational studies of products. This is especially the case when the assessed technology is strongly dependent on the involvement of the patients who use it as well as on the nature of the support implemented by the professionals around the technology.
Following the marketing of the Insulia device in France, we aimed to examine the results obtained in real life by the users of such a solution. This study was conducted based on data collected through the system itself, which constitutes a methodological limitation due to the relatively large number of missing data on some outcomes (ie, HbA1c level evolution). Nevertheless, some results were of interest. First, the device cannot be expected to have a positive effect if it is not used by the patient. About one-fifth of the patients did not use it after the first inscription on the device, and among users, only $37\%$ ($\frac{138}{373}$) were still regular users, and $24.4\%$ ($\frac{91}{373}$) were regular and compliant users after 6 months. We noted a progressive disaffection of the patients with time; even in the first 3 months, only half of the patients were regular users and slightly more than one-third of the patients were both regular and compliant users.
As anticipated, the impact on glycemic control was significantly better among regular and compliant users, with an individualized FBG target achieved in $60\%$ ($\frac{55}{91}$) of patients after 6 months versus $51.5\%$ ($\frac{145}{282}$) in other patients using Insulia less frequently. The trend in the proportion of patients achieving their target FBG (+$1.86\%$ every 2 weeks for regular and compliant users) and the HbA1c level decrease (from $9.9\%$ to $7.2\%$) after 6 months (±1.5 months) of app use are in the same direction as the results obtained in clinical trials but were clearly less favorable. The patient selection and close monitoring generally implemented in clinical trials is one possible explanation for this situation. Another explanation probably lies in the fact that, in the TeleDiab-2 study as well as the Bergenstal et al [7] study, the included patients benefited from sustained human support, which was not necessarily the case in our real-life observational study. This probably reflects the importance of the support that must be provided to patients when a tool such as *Insulia is* offered to them. It also highlighted the necessity to conduct as often as possible pragmatic trials to estimate the added value of such devices.
## Conclusion
In real life, an improvement in glycemic control as measured by the percentage of patients reaching their FBG individualized target range without increasing the incidence of hypoglycemia was observed in patients regularly using the Insulia app and following the dose recommendations of the algorithm. However, these results should be confirmed on a larger population as no significant difference according to the degree of Insulia use was observed considering HbA1c level results.
## References
1. Franc S, Joubert M, Daoudi A, Fagour C, Benhamou P, Rodier M, Boucherie B, Benamo E, Schaepelynck P, Guerci B, Dardari D, Borot S, Penfornis A, D'Orsay G, Mari K, Reznik Y, Randazzo C, Charpentier G. **Efficacy of two telemonitoring systems to improve glycaemic control during basal insulin initiation in patients with type 2 diabetes: The TeleDiab-2 randomized controlled trial**. *Diabetes Obes Metab* (2019) **21** 2327-2332. DOI: 10.1111/dom.13806
2. **Arrêté du 23 décembre 2020 portant cahiers des charges des expérimentations relatives à la prise en charge par télésurveillance mis en œuvre sur le fondement de l’article 54 de la loi no 2017- 1836 de financement de la sécurité sociale pour 2018**. *Ministère de la Santé et de la Prévention* (2020)
3. Huckvale K, Adomaviciute S, Prieto JT, Leow MKS, Car J. **Smartphone apps for calculating insulin dose: a systematic assessment**. *BMC Med* (2015) **13** 106. DOI: 10.1186/s12916-015-0314-7
4. Cui L, Schroeder PR, Sack PA. **Inpatient and outpatient technologies to assist in the management of insulin dosing**. *Clin Diabetes* (2020) **38** 462-473. DOI: 10.2337/cd20-0054
5. Grdinovac K, Robbins D, Lavenbarg T, Levin P, Sysko R. **iSage: successful basal insulin titration managed by a prescription-only digital therapy for T2DM**. *Diabetes* (2019) **68** 122-LB. DOI: 10.2337/db19-122-lb
6. Vogt L, Thomas A, Fritzsche G, Heinke P, Kohnert K, Salzsieder E. **Model-based tool for personalized adjustment of basal insulin supply in patients with intensified conventional insulin therapy**. *J Diabetes Sci Technol* (2019) **13** 928-934. DOI: 10.1177/1932296818823020
7. Bergenstal RM, Johnson M, Passi R, Bhargava A, Young N, Kruger DF, Bashan E, Bisgaier SG, Isaman DJM, Hodish I. **Automated insulin dosing guidance to optimise insulin management in patients with type 2 diabetes: a multicentre, randomised controlled trial**. *Lancet* (2019) **393** 1138-1148. DOI: 10.1016/S0140-6736(19)30368-X
|
---
title: 'Menstrual Tracking Mobile App Review by Consumers and Health Care Providers:
Quality Evaluations Study'
journal: JMIR mHealth and uHealth
year: 2023
pmcid: PMC10018377
doi: 10.2196/40921
license: CC BY 4.0
---
# Menstrual Tracking Mobile App Review by Consumers and Health Care Providers: Quality Evaluations Study
## Abstract
### Background
Women’s menstrual cycle is an important component of their overall health. Physiological cycles and associated symptoms can be monitored continuously and used as indicators in various fields. Menstrual apps are accessible and can be used to promote overall female health. However, no study has evaluated these apps’ functionality from both consumers’ and health care providers’ perspectives. As such, the evidence indicating whether the menstrual apps available on the market provide user satisfaction is insufficient.
### Objective
This study was performed to investigate the key content and quality of menstrual apps from the perspectives of health care providers and consumers. We also analyzed the correlations between health care provider and consumer evaluation scores. On the basis of this analysis, we offer technical and policy recommendations that could increase the usability and convenience of future app.
### Methods
We searched the Google Play Store and iOS App Store using the keywords “period” and “menstrual cycle” in English and Korean and identified relevant apps. An app that met the following inclusion criteria was selected as a research app: nonduplicate; with >10,000 reviews; last updated ≤180 days ago; relevant to this topic; written in Korean or English; available free of charge; and currently operational. App quality was evaluated by 6 consumers and 4 health care providers using Mobile Application Rating Scale (MARS) and user version of the Mobile Application Rating Scale (uMARS). We then analyzed the correlations among MARS scores, uMARS scores, star ratings, and the number of reviews.
### Results
Of the 34 apps, 31 ($91\%$) apps could be used to predict the menstrual cycle, and 2 ($6\%$) apps provided information pertinent to health screening. All apps that scored highly in the MARS evaluation offer a symptom logging function and provide the user with personalized notifications. The “Bom Calendar” app had the highest MARS (4.51) and uMARS (4.23) scores. The MARS (2.22) and uMARS (4.15) scores for the “Menstrual calendar—ovulation & pregnancy calendar” app were different. In addition, there was no relationship between MARS and uMARS scores ($r = 0.32$; $$P \leq .06$$).
### Conclusions
We compared consumer and health care provider ratings for menstrual apps. Continuous monitoring of app quality from consumer and health care provider perspectives is necessary to guide their development and update content.
## Background
Women’s menstrual cycles are important to their overall health [1,2] and are characterized by predictable and recurring symptoms. Continuous tracking of the menstrual cycle can aid in health management. Monitoring systems are needed to optimize menstrual health and provide easily accessible health information for women [3].
Mobile health (mHealth) apps facilitate personalized health monitoring and management [4], and menstrual apps are the most important mHealth apps for women. Such apps are typically highly accessible for most women and provide indicators relevant to various health domains [5].
Typically, menstrual apps also allow the user to log their symptoms, mood changes, and body temperature, and they visually represent statistical data via graphs and tables. Some apps offer professional-level information through communities and links, that is, they promote women’s health care by facilitating smooth communication with medical staff through information-sharing services [6]. The apps’ menstrual cycle tracking functions facilitate health care planning and management in various domains, including contraception, fertility, preparation with respect to pregnancy and ensuring adequate “menstrual supplies,” leisure activities, and travel [7].
Women use these functions for various purposes, such as tracking pregnancy, preventing pregnancy, and managing menstruation periods [8]. The functions desired by consumers depend on the intended purpose of the app. Currently, indicators to help consumers identify and select menstrual apps according to the desired functions are lacking [9]. Recently developed systems recommend apps based on consumer requirements [10,11]. Menstrual apps with various functions have been developed, and related research is being actively conducted. However, most of the existing studies are content reviews or expert evaluations [12-15]. Consumer-centered studies have also started to appear [5,16-18], but quality evaluations of menstrual apps remain scarce.
Menstrual apps are directly relevant to women’s health, so it is necessary for experts and health care providers to evaluate these apps [8]. Health care provider quality evaluations can contribute to the app development, which is important to ensure that consumers have access to high-quality apps [19]. Consumer quality evaluations provide feedback on apps, such as consumer preferences (eg, for easy-to-use content) [20,21]. Consistent use of an app is critical given the recurring nature of the menstrual cycle, but research on this topic is lacking from the developer’s standpoint. To promote sustained app use, evaluations from consumers and health care providers’ perspective are required. However, it is unclear whether current commercially available apps satisfy the quality standards of consumers and health care providers. If we find the differences between consumers and health care providers, they need to be discussed importantly to develop apps that can be used to promote women’s health that satisfy both consumers’ wants and health care providers’ needs.
## Purpose
This study examines currently available menstrual apps in term of their main contents and quality based on evaluations by health care providers and consumers. We also correlated the two sets of evaluation scores. The study’s findings could improve the utility and convenience of future mHealth apps.
## App Selection
We searched for keywords related to the development and evaluation of menstrual apps commonly included in previous studies, such as “period” and “menstrual cycle” in both English and Korean. The Google Play Store and the iOS App Store were searched from April 8 to April 15, 2021. Up to 150 apps were screened for each keyword. Since then, to secure the appropriated number of apps that can be statistically analyzed, the following app inclusion criteria have been set, based on previous studies [14,22,23]: A total of 1127 menstrual apps were initially identified via the keyword search, and 34 apps (Android: $$n = 28$$; iPhone: $$n = 6$$) met all of the study criteria and were included in the final analysis (Figure 1).
**Figure 1:** *Flow diagram of the app review and selection process. C1-7 denotes Criterion 1-7.*
The operating system (OS) of apps is linked with the app store, so that the app is updated simultaneously with updates of the OS [32]. Therefore, the app version and function may vary depending on the timing of the OS update. The number of reviews and star ratings differed among the apps in this study. For example, Bom Calendar had star ratings of 4.8 and 4.4 for the Android and iOS versions, respectively (23,437 and 76,258 reviews, respectively). In instances where the same apps were available for different OSs, each version was considered to be a unique app in the analysis.
## Analysis of App Contents
To examine the apps’ main contents, we performed a pilot study of 14 representative apps. The apps’ main contents were classified as follows: menstrual cycle management, education and knowledge, sharing information, and notifications. Frequency analysis was performed to determine the number of apps providing these functions.
Most apps ($$n = 31$$, $91\%$) offered a menstrual cycle prediction function. Some apps ($$n = 14$$, $41\%$) offered menstruation and fertility period notifications, while others had no specific functions ($$n = 3$$, $9\%$). Most apps were confidential ($$n = 29$$, $85\%$), allowed data export ($$n = 28$$, $82\%$), and had a log-in function ($$n = 25$$, $74\%$). However, few apps provided education or knowledge ($$n = 10$$, $29\%$), screening-related information ($$n = 2$$, $6\%$), or advice ($$n = 4$$, $12\%$; Table 2).
**Table 2**
| Contents | Contents.1 | Contents.2 | App, n (%) | App, n (%).1 | App, n (%).2 | App, n (%).3 |
| --- | --- | --- | --- | --- | --- | --- |
| | | | Android (n=28) | Android (n=28) | iPhone (n=6) | Total (N=34) |
| Menstrual cycle management | Menstrual cycle management | Menstrual cycle management | Menstrual cycle management | Menstrual cycle management | Menstrual cycle management | Menstrual cycle management |
| | Symptoms (pain) | 4 (14) | 4 (14) | 0 (0) | 0 (0) | 4 (12) |
| | Additional symptom | 18 (64) | 18 (64) | 6 (100) | 6 (100) | 24 (71) |
| Ovulation management | Ovulation management | Ovulation management | Ovulation management | Ovulation management | Ovulation management | Ovulation management |
| | Calculate pregnancy probability | 4 (14) | 4 (14) | 0 (0) | 0 (0) | 4 (12) |
| | Contraception methods | 3 (11) | 3 (11) | 0 (0) | 0 (0) | 3 (9) |
| | Both | 12 (43) | 12 (43) | 6 (100) | 6 (100) | 18 (53) |
| Last update (months) | Last update (months) | Last update (months) | Last update (months) | Last update (months) | Last update (months) | Last update (months) |
| | <3 | 21 (75) | 21 (75) | 5 (83) | 5 (83) | 26 (76) |
| | 3~6 | 7 (25) | 7 (25) | 1 (17) | 1 (17) | 8 (24) |
| Function s | Function s | Function s | Function s | Function s | Function s | Function s |
| | Graphical chart | 17 (61) | 17 (61) | 4 (67) | 4 (67) | 21 (62) |
| | Lock | 23 (82) | 23 (82) | 6 (100) | 6 (100) | 29 (85) |
| | Advice provision | 2 (7) | 2 (7) | 2 (33) | 2 (33) | 4 (12) |
| | Data export | 23 (82) | 23 (82) | 5 (83) | 5 (83) | 28 (82) |
| | Predictions | 25 (89) | 25 (89) | 6 (100) | 6 (100) | 31 (91) |
| | Log-in | 20 (71) | 20 (71) | 5 (83) | 5 (83) | 25 (74) |
| Education or knowledge | Education or knowledge | Education or knowledge | Education or knowledge | Education or knowledge | Education or knowledge | Education or knowledge |
| | General health information | 5 (18) | 5 (18) | 1 (17) | 1 (17) | 6 (18) |
| | Personalized information | 1 (4) | 1 (4) | 2 (33) | 2 (33) | 3 (9) |
| | Both | 1 (4) | 1 (4) | 0 (0) | 0 (0) | 1 (3) |
| | Health screening | 1 (4) | 1 (4) | 1 (17) | 1 (17) | 2 (6) |
| Sharing information (with health care professionals) | Sharing information (with health care professionals) | Sharing information (with health care professionals) | Sharing information (with health care professionals) | Sharing information (with health care professionals) | Sharing information (with health care professionals) | Sharing information (with health care professionals) |
| | All information | 4 (14) | 4 (14) | 3 (50) | 3 (50) | 7 (21) |
| | Only information specified by the consumer | 1 (4) | 1 (4) | 0 (0) | 0 (0) | 1 (3) |
| Visualization | Visualization | Visualization | Visualization | Visualization | Visualization | Visualization |
| | Menstruation or ovulation | 2 (7) | 2 (7) | 0 (0) | 0 (0) | 2 (6) |
| | Menstrual cycle | 9 (32) | 9 (32) | 0 (0) | 0 (0) | 9 (26) |
| | All data | 16 (57) | 16 (57) | 6 (100) | 6 (100) | 22 (65) |
| Notifications | Notifications | Notifications | Notifications | Notifications | Notifications | Notifications |
| | Menstruation or fertility | 2 (7) | 2 (7) | 0 (0) | 0 (0) | 2 (56) |
| | Both | 14 (50) | 14 (50) | 0 (0) | 0 (0) | 14 (41) |
| | Personalized alarms | 9 (32) | 9 (32) | 6 (100) | 6 (100) | 15 (44) |
| Other features | Other features | Other features | Other features | Other features | Other features | Other features |
| | Community | 4 (14) | 4 (14) | 2 (33) | 2 (33) | 6 (18) |
| | Shopping | 2 (7) | 2 (7) | 1 (17) | 1 (17) | 3 (9) |
## Evaluation of App Quality
The quality of 34 menstrual apps were evaluated using Mobile Application Rating Scale (MARS) and user version of the Mobile Application Rating Scale (uMARS). The MARS was developed to evaluate mobile apps; it is a reliable tool with a high internal consistency and interrater reliability [24]. The uMARS was subsequently developed, based on the MARS, to allow consumers to evaluate the quality of mHealth apps. uMARS has excellent internal consistency [25]. Both scales comprise the following five categories: engagement, aesthetics, functionality, information, and subjective quality (Table 1). The MARS has been used to evaluate various health care–related apps, such as apps related to chronic disease, COVID-19, and physical activity, as well as apps related to allergy, hepatitis treatment support, and breast cancer [14,15,19,22,23,26,27]. The uMARS has recently been used to evaluate various types of mHealth apps, including apps pertaining to weight loss and nutrition, rheumatic diseases, and the management of ankylosing spondylitis [28-31].
**Table 1**
| Measures | Measures.1 | Characteristics |
| --- | --- | --- |
| Objective measures | Objective measures | Objective measures |
| | Engagement | Entertainmenta,bInteresta,bCustomizationa,bInteractivitya,bTarget groupa,b |
| | Functionality | Performancea,bEase of usea,bNavigationa,bGestural designa,b |
| | Information | Accuracy of app designaGoalaCredibilityaEvidence basesaQuality of informationa,bQuantity of informationa,bVisual informationa,bCredibility of sourceb |
| | Aesthetics | Layouta,bGraphicsa,bVisual appeala,b |
| Subjective measures | Subjective measures | Subjective measures |
| | Subjective quality | Recommendationa,bFrequency of usea,bPayment for expensesa,bStar ratinga,b |
A total of 6 consumers completed the uMARS between July 9 and July 22, 2021, and 4 nurses (ie, health care providers) majoring in health care and working in medical centers completed the MARS between July 21 and July 30, 2021. Each evaluator was asked to use the app for more than 10 minutes every day and the evaluation was conducted in a blind test. The apps were randomly assigned to evaluators to prevent bias related to subjectivity. Each app was crossevaluated by at least two evaluators.
The MARS and uMARS scores of all apps were obtained through the average of the evaluator’s evaluation scores. The average MARS and uMARS scores (ie, health care provider and consumer scores, respectively) were 3.06 (SD 0.62) and 3.33 (SD 0.57), respectively. The iPhone “Bom Calendar” app had the highest score for both the MARS (4.51, SD 0.22) and uMARS (4.23, SD 0.27). The Android “Period calendar—Women’s menstrual calendar❤” app had the second lowest score for both the MARS (2.05, SD 0.45) and uMARS (2.09, SD 0.05), despite its high star rating of 4.8. The Android “Menstrual calendar—ovulation & pregnancy calendar” app showed contrasting scores between the MARS (2.22, SD 0.07) and uMARS (4.15, SD 0.46); it had the third highest overall uMARS score, with an engagement score of 3.86 (SD 0.49), functionality score of 4.04 (SD 0.49), aesthetics score of 4.25 (SD 0.47), information score of 4.25 (SD 0.47), and subjective quality score of 4.37 (SD 0.41). However, for the MARS, its overall score was the third lowest, with an engagement score of 2.40 (SD 0.20), functionality score of 2.50 (SD 0.50), aesthetics score of 2.00 (SD 0.00), information score of 2.58 (SD 0.42), and subjective quality score of 1.63 (SD 0.38; Table 3 and Table 4). MARS and uMARS scores of all 34 apps are presented in Multimedia Appendix 1.
## Comparative Analysis of Consumer and Health Care Provider Data
The MARS results represent health care providers’ perspectives, as stated above. The uMARS results, star ratings, and number of reviews were analyzed from the consumers’ perspective. After normalization, Pearson correlation was used to correlate app content, MARS and uMARS scores, star ratings, and reviews and to correlate perspectives of health care providers and consumers. We identified the top and bottom five apps based on the MARS and uMARS scores and compared app preferences between consumers and health care providers. P values <.05 were considered significant. R software (version 4.1.2; R Core Team,) was used for the analysis.
## Further Comparative Analysis of Consumer and Health Care Provider Data
Table 5 shows the results of correlation analysis of the MARS and uMARS scores, star ratings, number of reviews, and app content. The number of reviews was not correlated with app content and menstrual cycle management ($r = 0.53$; $$P \leq .001$$), and visualization ($r = 0.51$; $$P \leq .002$$) had the highest correlation with star ratings. Among the evaluation scores, the highest correlation was found between uMARS and notification ($r = 0.39$; $$P \leq .02$$) as well as between MARS and ovulation date management ($r = 0.49$; $$P \leq .003$$).
Multimedia Appendix 2 shows content comparison between the top and bottom five apps in consumer and health care provider evaluation scores. Personalized alarms could be set in the top five apps. In addition, they provided a function to visualize all information through a calendar or to specify and manage ovulation days. On the contrary, in the bottom five apps did not have functions for managing or predicting the menstrual cycle.
Figure 2 shows the correlations among the MARS and uMARS scores, star ratings, and number of reviews to compare the perspective of health care providers and consumers. Figure 2 shows no correlation between MARS and uMARS scores of the health care providers and consumers ($r = 0.32$; $$P \leq .06$$). uMARS scores and star rating ($r = 0.11$; $$P \leq .54$$) as well as uMRAS scores and number of reviews ($r = 0.07$; $$P \leq .67$$) also showed no significant correlations. The number of reviews and star rating ($r = 0.39$; $$P \leq .02$$) showed a very low correlation.
## Principal Findings and Comparison With Prior Work
Interest in mHealth apps has recently increased, and new apps are continuously being developed, including menstrual apps [33]. However, the needs of consumers and health care providers are different, and studies evaluating whether the available menstrual apps satisfy both of these groups are difficult to find. This study obtained quality evaluations of relevant apps from consumers and health care providers using the uMARS and MARS, respectively; the consistency of the evaluations of these two groups in terms of key app contents was analyzed. The health care providers valued engagement, functionality, and aesthetics when evaluating apps, while consumers valued aesthetics and information provision the most.
The MARS and uMARS scores were not correlated in this study. For example, the app with the third lowest MARS score had the third highest uMARS score. A significant difference was observed in app aesthetics scores between the consumers and health care providers; the scores for this attribute showed the largest group difference. Previous studies on health services have reported disparities between consumers and health care providers, and these results affect the implementation of consumer-centered services. Data from health care providers can provide a basis for high-quality apps [20], while consumer data serves as feedback on app quality [19]. To ensure high app quality and consumer satisfaction, app quality should be continuously monitored from the perspective of both health care providers and consumers. Monitoring can identify the needs of consumers and health care providers, which can in turn help in app development and update [10]. This study focused on evaluating app quality using MARS and uMARS for consistency, but could be extended to qualitative studies, including interviews, to collect in-depth answers in the future [34].
According to our findings, the uMARS (ie, consumer) scores, star ratings, and number of reviews were unrelated variables. The uMARS allows for direct assessment of mHealth apps and is a reliable measure of app quality [25]. However, reviews and star ratings are subjective indicators [24]. The currently available mHealth apps have been evaluated in a simplistic manner, such as through star ratings and reviews, even though they differ significantly with respect to content; thus, appropriate guidelines to aid app selection are lacking. By using the uMARS to guide app selection, the limitations of reviews and star ratings can be overcome such that consumers will likely select more useful apps to meet their particular needs. However, it is difficult for consumers to evaluate the quality of the apps using uMARS every time they download one. New indicators to guide app selection are needed so that consumers can make decisions based on objective evaluation results.
Most of the five apps that scored highly in the quality evaluation in this study included personalized monitoring functions. Menstrual apps are becoming increasingly popular, and apps that include self-monitoring functions and provide related information are continuously being developed [28]. Personalized monitoring can improve user well-being by encouraging them to check for signs and symptoms of health issues. Health-related information can also promote consumer health. mHealth apps that include personalized content of this nature are particularly useful for consumers [23,35]. However, only a few of the apps evaluated in this study facilitated consultations with specialists or provided information relating to women’s health. To increase the utility of apps, notifications, symptom recording functions, and the provision of knowledge should be prioritized.
mHealth apps should provide customized content for individual consumers [36]. However, the five bottom-scoring apps in this study did not meet the needs of the women who were using them. Consumers use apps to predict menstrual cycles and ovulation dates, and to monitor their general health [37]. However, most of the five bottom-scoring apps did not provide content enhancing consumer convenience, such as functions for menstrual cycle and ovulation day management, and some apps also lacked predictive functions. Such apps must be updated to include content allowing for the prediction and management of menstrual cycles based on accumulated menstrual cycle- and health-related information.
Menstrual apps collects personal information from consumers, such as name, date of birth, menstrual cycle, and medical history [38]. Personal data must be protected because it is sensitive information [12,39], but some apps do not provide locking functions, and few apps provide icon change functions for protecting personal information. Most fitness apps that record the number of steps do not consider privacy issues, and the data protection of mHealth apps related to women’s health is typically poor [39]. Therefore, regulations pertaining to app management of private data are necessary [40,41]. In fact, there are existing regulations protecting personal information, such as the European Unions’s General Data Protection Regulation, but no standard regulations are enforced worldwide. mHealth apps that protect personal information tend to be favored by consumers [21]. App developers should improve data protection–related functions to protect the personal information of consumers.
The MARS and uMARS were developed specifically for evaluating mHealth apps that aim to improve consumers’ health. Meanwhile, menstrual apps were designed to help consumers keep track of their current health rather than improve it. Health apps must provide solutions customized to individual consumers [36]; reliability may be key in this respect [42]. This study’s results indicate that current health apps do not fully meet consumers’ requirements or desires with respect to content. To better identify consumer objectives and take account of them during the development of menstrual apps, a new evaluation scale is needed to evaluate menstrual apps.
## Limitations
This study had several limitations. First, the results cannot be generalized to all menstrual apps because a small number of evaluators evaluated only the most popular apps. Second, the apps were selected based on App Store searches with a limited timeframe. Updates to apps may result in differences between the analyzed content and that in the future. Third, the database is not an electronic database but the App Store. The App Store’s app recommendation function may have compromised an inconsistent search accuracy.
## Conclusions
In this study, consumer and health care provider ratings of menstrual apps were obtained using validated scales. Consumer preferred app had high scores of aesthetics and information, and evaluation scores differed between consumers and health care providers. The findings highlight the importance of consumer participation in menstrual app development and evaluation. This study is significant in that it is the first to compare health care providers’ and consumers’ menstrual app quality ratings. We expect our results to guide future mHealth app development and provide consumers with information on menstrual app content and quality. To provide high-quality apps for consumers, continuous quality evaluation research needs to be conducted, and the perspectives of both consumers and health care providers should be taken into account.
## References
1. Bae J, Park S, Kwon J. **Factors associated with menstrual cycle irregularity and menopause**. *BMC Womens Health* (2018) **18** 36. DOI: 10.1186/s12905-018-0528-x
2. Critchley HO, Babayev E, Bulun SE, Clark S, Garcia-Grau I, Gregersen PK, Kilcoyne A, Kim JJ, Lavender M, Marsh EE, Matteson KA, Maybin JA, Metz CN, Moreno I, Silk K, Sommer M, Simon C, Tariyal R, Taylor HS, Wagner GP, Griffith LG. **Menstruation: science and society**. *AJOG* (2020) **223** 624-664. DOI: 10.1016/j.ajog.2020.06.004
3. Park H. **Health behavior and health promotion in women**. *Health and Welfare Policy Forum* (2021) 2-4
4. Derbyshire E, Dancey D. **Smartphone medical applications for women's health: what is the evidence-base and feedback?**. *Int J Telemed Appl* (2013) **2013** 782074. DOI: 10.1155/2013/782074
5. Moglia ML, Nguyen HV, Chyjek K, Chen KT, Castaño PM. **Evaluation of smartphone menstrual cycle tracking applications using an adapted APPLICATIONS scoring system**. *Obstet Gynecol* (2016) **127** 1153-1160. DOI: 10.1097/AOG.0000000000001444
6. Levy J, Romo-Avilés N. **"A good little tool to get to know yourself a bit better": a qualitative study on users' experiences of app-supported menstrual tracking in Europe**. *BMC Public Health* (2019) **19** 1213. DOI: 10.1186/s12889-019-7549-8
7. Epstein DA, Lee NB, Kang JH, Agapie E, Schroeder J, Pina LR, Fogarty J, Kientz JA, Munson SA. **Examining menstrual tracking to inform the design of personal informatics tools**. *Proc SIGCHI Conf Hum Factor Comput Syst* (2017) **2017** 6876-6888. DOI: 10.1145/3025453.3025635
8. Earle S, Marston HR, Hadley R, Banks D. **Use of menstruation and fertility app trackers: a scoping review of the evidence**. *BMJ Sex Reprod Health* (2021) **47** 90-101. DOI: 10.1136/bmjsrh-2019-200488
9. DeNicola N, Marko K. **Connected health and mobile apps in obstetrics and gynecology**. *Obstet Gynecol Clin North Am* (2020) **47** 317-331. DOI: 10.1016/j.ogc.2020.02.008
10. Lee J, Kim J. **Can menstrual health apps selected based on users' needs change health-related factors? A double-blind randomized controlled trial**. *J Am Med Inform Assoc* (2019) **26** 655-666. DOI: 10.1093/jamia/ocz019
11. Lee J, Kim J. **Method of app selection for healthcare providers based on consumer needs**. *Comput Inform Nurs* (2018) **36** 45-54. DOI: 10.1097/CIN.0000000000000399
12. Dantas LO, Carvalho C, Pena C, Breda CC, Driusso P, Ferreira CHJ, Bø Kari. **Mobile health technologies for the monitoring of menstrual cycle: a systematic review of online stores in Brazil**. *J Obstet Gynaecol Res* (2022) **48** 5-14. DOI: 10.1111/jog.15082
13. Starič KD, Trajkovik V, Belani H, Vitagliano A, Bukovec P. **Smart phone applications for self-monitoring of the menstrual cycle: a review and content analysis**. *Clin Exp Obstet Gynecol* (2019) **46** 731-735. DOI: 10.12891/ceog4830.2019
14. Davalbhakta S, Advani S, Kumar S, Agarwal V, Bhoyar S, Fedirko E, Misra DP, Goel A, Gupta L, Agarwal V. **A systematic review of smartphone applications available for corona virus disease 2019 (COVID19) and the assessment of their quality using the Mobile Application Rating Scale (MARS)**. *J Med Syst* (2020) **44** 164. DOI: 10.1007/s10916-020-01633-3
15. Rodrigues AT, Sousa CT, Pereira J, Figueiredo IV, Lima TDM. **Mobile applications (Apps) to support the hepatitis c treatment: a systematic search in app stores**. *Ther Innov Regul Sci* (2021) **55** 152-162. DOI: 10.1007/s43441-020-00201-8
16. Gambier-Ross K, McLernon DJ, Morgan HM. **A mixed methods exploratory study of women's relationships with and uses of fertility tracking apps**. *Digit Health* (2018) **4** 2055207618785077. DOI: 10.1177/2055207618785077
17. Starling MS, Kandel Z, Haile L, Simmons RG. **User profile and preferences in fertility apps for preventing pregnancy: an exploratory pilot study**. *Mhealth* (2018) **4** 21. DOI: 10.21037/mhealth.2018.06.02
18. Saparamadu AADNS, Fernando P, Zeng P, Teo H, Goh A, Lee JMY, Lam CWL. **User-centered design process of an mhealth app for health professionals: case study**. *JMIR Mhealth Uhealth* (2021) **9** e18079. DOI: 10.2196/18079
19. Gómez-Martínez-Sagrera P, Escudero-Vilaplana V, Collado-Borrell R, Villanueva-Bueno C, Gómez-Centurión I, Herranz-Alonso A, Sanjurjo-Sáez M. **Mobile apps for hematological conditions: review and content analysis using the Mobile App Rating Scale**. *JMIR Mhealth Uhealth* (2022) **10** e32826. DOI: 10.2196/32826
20. Nicholas J, Fogarty AS, Boydell K, Christensen H. **The reviews are in: a qualitative content analysis of consumer perspectives on apps for bipolar disorder**. *J Med Internet Res* (2017) **19** e105. DOI: 10.2196/jmir.7273
21. van Haasteren A, Gille F, Fadda M, Vayena E. **Development of the mHealth App Trustworthiness checklist**. *Digit Health* (2019) **5** 2055207619886463. DOI: 10.1177/2055207619886463
22. Mandracchia F, Llauradó E, Tarro L, Valls RM, Solà R. **Mobile phone apps for food allergies or intolerances in app stores: systematic search and quality assessment using the Mobile App Rating Scale (MARS)**. *JMIR Mhealth Uhealth* (2020) **8** e18339. DOI: 10.2196/18339
23. Karasneh RA, Al-Azzam SI, Alzoubi KH, Muflih SM, Hawamdeh SS. **Smartphone applications for period tracking: rating and behavioral change among women users**. *Obstet Gynecol Int* (2020) **2020** 2192387-9. DOI: 10.1155/2020/2192387
24. Stoyanov SR, Hides L, Kavanagh DJ, Zelenko O, Tjondronegoro D, Mani M. **Mobile app rating scale: a new tool for assessing the quality of health mobile apps**. *JMIR Mhealth Uhealth* (2015) **3** e27. DOI: 10.2196/mhealth.3422
25. Stoyanov SR, Hides L, Kavanagh DJ, Wilson H. **Development and validation of the user version of the Mobile Application Rating Scale (uMARS)**. *JMIR Mhealth Uhealth* (2016) **4** e72. DOI: 10.2196/mhealth.5849
26. Lee JB, Woo H. **Quality evaluation of smartphone applications for physical activity promotion**. *KOSHEP_EN* (2019) **36** 67-76. DOI: 10.14367/kjhep.2019.36.4.67
27. Wright A. **Evaluation of two mobile health apps for patients with breast cancer using the Mobile Application Rating Scale**. *Mhealth* (2021) **7** 60. DOI: 10.21037/mhealth-20-161
28. Bardus M, Ali A, Demachkieh F, Hamadeh G. **Assessing the quality of mobile phone apps for weight management: user-centered study with employees from a Lebanese university**. *JMIR Mhealth Uhealth* (2019) **7** e9836. DOI: 10.2196/mhealth.9836
29. Li Y, Ding J, Wang Y, Tang C, Zhang P. **Nutrition-related mobile apps in the china app store: assessment of functionality and quality**. *JMIR Mhealth Uhealth* (2019) **7** e13261. DOI: 10.2196/13261
30. Lambrecht A, Vuillerme N, Raab C, Simon D, Messner E, Hagen M, Bayat S, Kleyer A, Aubourg T, Schett G, Hueber A, Knitza J. **Quality of a supporting mobile app for rheumatic patients: patient-based assessment using the user version of the Mobile Application Scale (uMARS)**. *Front Med (Lausanne)* (2021) **8** 715345. DOI: 10.3389/fmed.2021.715345
31. Song Y, Chen H. **Evaluating chinese mobile health apps for ankylosing spondylitis management: systematic app search**. *JMIR Mhealth Uhealth* (2021) **9** e27234. DOI: 10.2196/27234
32. Novac O, Novac M, Gordan C, Berczes T, Bujdosó G. **Comparative study of Google Android, Apple iOS and Microsoft Windows Phone mobile operating systems**. (2017) 154-159. DOI: 10.1109/emes.2017.7980403
33. Symul L, Wac K, Hillard P, Salathé M. **Assessment of menstrual health status and evolution through mobile apps for fertility awareness**. *NPJ Digit Med* (2019) **2** 64. DOI: 10.1038/s41746-019-0139-4
34. Namey E, Guest G, McKenna K, Chen M. **Evaluating Bang for the Buck**. *AJE* (2016) **37** 425-440. DOI: 10.1177/1098214016630406
35. Zhao J, Freeman B, Li M. **Can mobile phone apps influence people's health behavior change? An evidence review**. *J Med Internet Res* (2016) **18** e287. DOI: 10.2196/jmir.5692
36. Anderson K, Emmerton LM. **Contribution of mobile health applications to self-management by consumers: review of published evidence**. *Aust Health Rev* (2016) **40** 591-597. DOI: 10.1071/AH15162
37. Adnan T, Coull BA, Jukic AM, Mahalingaiah S. **The real-world applications of the symptom tracking functionality available to menstrual health tracking apps**. *Curr Opin Endocrinol Diabetes Obes* (2021) **28** 574-586. DOI: 10.1097/MED.0000000000000682
38. Tangari G, Ikram M, Ijaz K, Kaafar MA, Berkovsky S. **Mobile health and privacy: cross sectional study**. *BMJ* (2021) **373** n1248. DOI: 10.1136/bmj.n1248
39. Alfawzan N, Christen M, Spitale G, Biller-Andorno N. **Privacy, data sharing, and data security policies of women’s mhealth apps: scoping review and content analysis**. *JMIR Mhealth Uhealth* (2022) **10** e33735. DOI: 10.2196/33735
40. Eng DS, Lee JM. **The promise and peril of mobile health applications for diabetes and endocrinology**. *Pediatr Diabetes* (2013) **14** 231-8. DOI: 10.1111/pedi.12034
41. Costa Figueiredo M, Huynh T, Takei A, Epstein DA, Chen Y. **Goals, life events, and transitions: examining fertility apps for holistic health tracking**. *JAMIA Open* (2021) **4** ooab013. DOI: 10.1093/jamiaopen/ooab013
42. Pichon A, Jackman K, Winkler I, Bobel C, Elhadad N. **The messiness of the menstruator: assessing personas and functionalities of menstrual tracking apps**. *J Am Med Inform Assoc* (2022) **29** 385-399. DOI: 10.1093/jamia/ocab212
|
---
title: 'Identification of Postpartum Depression in Electronic Health Records: Validation
in a Large Integrated Health Care System'
journal: JMIR Medical Informatics
year: 2023
pmcid: PMC10018380
doi: 10.2196/43005
license: CC BY 4.0
---
# Identification of Postpartum Depression in Electronic Health Records: Validation in a Large Integrated Health Care System
## Abstract
### Background
The accuracy of electronic health records (EHRs) for identifying postpartum depression (PPD) is not well studied.
### Objective
This study aims to evaluate the accuracy of PPD reporting in EHRs and compare the quality of PPD data collected before and after the implementation of the International Classification of Diseases, Tenth Revision (ICD-10) coding in the health care system.
### Methods
Information on PPD was extracted from a random sample of 400 eligible Kaiser Permanente Southern California patients’ EHRs. Clinical diagnosis codes and pharmacy records were abstracted for two time periods: January 1, 2012, through December 31, 2014 (International Classification of Diseases, Ninth Revision [ICD-9] period), and January 1, 2017, through December 31, 2019 (ICD-10 period). Manual chart reviews of clinical records for PPD were considered the gold standard and were compared with corresponding electronically coded diagnosis and pharmacy records using sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Kappa statistic was calculated to measure agreement.
### Results
Overall agreement between the identification of depression using combined diagnosis codes and pharmacy records with that of medical record review was strong (κ=0.85, sensitivity $98.3\%$, specificity $83.3\%$, PPV $93.7\%$, NPV $95.0\%$). Using only diagnosis codes resulted in much lower sensitivity ($65.4\%$) and NPV ($50.5\%$) but good specificity ($88.6\%$) and PPV ($93.5\%$). Separately, examining agreement between chart review and electronic coding among diagnosis codes and pharmacy records showed sensitivity, specificity, and NPV higher with prescription use records than with clinical diagnosis coding for PPD, $96.5\%$ versus $72.0\%$, $96.5\%$ versus $65.0\%$, and $96.5\%$ versus $65.0\%$, respectively. There was no notable difference in agreement between ICD-9 (overall κ=0.86) and ICD-10 (overall κ=0.83) coding periods.
### Conclusions
PPD is not reliably captured in the clinical diagnosis coding of EHRs. The accuracy of PPD identification can be improved by supplementing clinical diagnosis with pharmacy use records. The completeness of PPD data remained unchanged after the implementation of the ICD-10 diagnosis coding.
## Introduction
Postpartum depression (PPD), major or minor depressive episodes occurring within 12 months after childbirth, is a common obstetric complication in the United States, with a prevalence of $13.2\%$ in 2018 [1]. The American College of Obstetricians and Gynecologists recommends all obstetrics care providers conduct comprehensive screening for PPD and anxiety disorders using a validated instrument for each patient separately during their postpartum visit [2]. Meanwhile, the American Academy of Pediatrics recommended routine PPD screening to be integrated at well-child visits (1-, 2-, 4-, and 6-month infant visits) [3]. The US Preventive Services Task Force also supports the provision of depression screening during postpartum visits, citing moderate net benefits for identifying those affected and recommending referrals to counseling interventions [4]. It is important to identify those with PPD because undetected or untreated depressive episodes can negatively impact the patient and their infant’s health and well-being. For instance, about $9\%$ of pregnancy-related deaths were due to mental health conditions [5]. Early PPD was also associated with increased behavior disturbances in the infant [6]. Moreover, other potential risk factors, including a prior history of depression, depression and anxiety episodes during pregnancy, preterm birth and lower infant birth weight, traumatic birth experience, stressful life events during early postpartum, and low social support, have been linked with PPD [7-9].
Health systems previously used the International Classification of Diseases, Ninth Revision (ICD-9), an official coding system to identify hospital-related diagnoses and procedures in the United States [10]. However, the Kaiser Permanente health systems shifted to using the International Classification of Diseases, Tenth Revision (ICD-10) codes after October 1, 2015, which has significant improvements over ICD-9 for many clinical codes [11]. However, Stewart et al [12] concluded that there is a need to perform a validation of diagnosis codes for each mental health condition following the ICD-10 transition. Colvin et al [13] used a data linkage of national pharmacy records and hospital admission information to identify patients with major depressive episodes in pregnancy but found the use of either source alone to be inadequate.
While there are multiple validated scales to screen for PPD, like the Patient Health Questionnaire (9-item) and the Edinburgh Postnatal Depression Scale, validation of these measures has been performed using ICD-9 or ICD-10 diagnostic codes as the gold standard [14,15]. Several studies have also developed machine learning algorithms using electronic health record (EHR) data to create risk-based models and examined whether they can predict PPD in large health care systems, relying on PPD ascertained using ICD-9 or ICD-10 codes [16,17]. However, the accuracy of ICD-9 and ICD-10 codes as the gold standard in ascertaining PPD has not been established previously. Prior validation of ICD-9 and ICD-10 found high positive predictive values (PPVs) for ascertaining general depression ($89.7\%$ and $89.5\%$, respectively), but these were not specific to the postpartum period [18]. This study aimed to assess the validity of ascertaining PPD diagnosis using the EHR from a large integrated health care delivery system, Kaiser Permanente Southern California (KPSC).
## Cohort Selection
We identified a random sample of 400 women with live birth records in the Air Pollution and Pregnancy Complications in Complex Urban Environments (APPCUE) study [19] between January 1, 2008, and December 31, 2018, within KPSC, a large integrated health system. The APPCUE study was a retrospective cohort study conducted in collaboration between KPSC and the University of California, Irvine with access to KPSC’s comprehensive EHRs. The APPCUE study included all singleton births at KPSC facilities. The EHRs contain patient-level data from out- and inpatient clinical care, including ICD-9, Clinical Modification or ICD-10, Clinical *Modification diagnosis* and procedure codes, as well as pharmacy and laboratory test records. From 236,759 pregnancies during the study period, we excluded pregnancies resulting in nonlive births ($$n = 8422$$) and patients who were not members from the start of their pregnancy through a 1-year postpartum period ($$n = 70$$,836) to have a complete medical history for this validation study. Of the remaining 157,501 pregnancies, we selected a random sample of 400. Simple random sampling was used to select 100 patients from groups based on EHR data: those without any diagnostic or pharmacy use record for PPD, those with only a diagnostic code for PPD, those with only a pharmacy record indicating treatment for PPD, and those with both diagnostic and pharmacy indications. Additionally, each sample was evenly split (50 each) between the ICD-9 diagnosis code era (date of delivery 2012-2014) and the ICD-10 era [2017-2019].
Table 1 shows the distribution of the APPCUE study cohort as well as the overall KPSC birth cohort during the study period. Nearly half ($\frac{194}{400}$, $48.5\%$) were Latina, most ($\frac{379}{400}$, $94.8\%$) received prenatal care starting in the first trimester, and most ($\frac{354}{400}$, $88.5\%$) delivered at 37 weeks of gestation or later. The study sample generally has very similar characteristics to the APPCUE study cohort overall and all KPSC births during the period, though there are some differences relative to all deliveries in the state of California, notably a higher percentage of non-Hispanic White mothers ($\frac{113}{400}$, $28.3\%$ vs 372,$\frac{037}{2}$,874,396, $12.9\%$), older mothers ($\frac{259}{400}$, $64.8\%$ age ≥30 years vs 1,465,$\frac{998}{2}$,874,396, $50.0\%$), and generally higher educational attainment ($\frac{199}{400}$, $49.8\%$ with at least a college degree vs 1,047,$\frac{594}{2}$,874,396, $36.5\%$).
**Table 1**
| Characteristics | Characteristics.1 | Characteristics.2 | Chart review samplea (N=400), n (%) | Chart review samplea (N=400), n (%).1 | APPCUEb study population (N=157,501), n (%) | APPCUEb study population (N=157,501), n (%).1 | P value | P value.1 | All KPSC births (N=236,759), n (%) | All KPSC births (N=236,759), n (%).1 | P value.2 | P value.3 | All California State birthsc (N=2,874,396), n (%) | All California State birthsc (N=2,874,396), n (%).1 | P value.4 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Maternal age (years) | Maternal age (years) | Maternal age (years) | Maternal age (years) | Maternal age (years) | Maternal age (years) | Maternal age (years) | 0.42 | .42 | | | .01 | .01 | | | <.001 |
| | <20 | 10 (2.5) | 10 (2.5) | 4665 (3.0) | 4665 (3.0) | | | 6804 (3.0) | 6804 (3.0) | | | 144,945 (5.0) | 144,945 (5.0) | | |
| | 20-29 | 131 (32.8) | 131 (32.8) | 56,679 (36.0) | 56,679 (36.0) | | | 92,203 (40.4) | 92,203 (40.4) | | | 1,263,453 (44.0) | 1,263,453 (44.0) | | |
| | 30-34 | 153 (38.3) | 153 (38.3) | 54,810 (34.8) | 54,810 (34.8) | | | 75,633 (33.1) | 75,633 (33.1) | | | 843,010 (29.3) | 843,010 (29.3) | | |
| | ≥35 | 106 (26.5) | 106 (26.5) | 41,347 (26.3) | 41,347 (26.3) | | | 53,697 (23.5) | 53,697 (23.5) | | | 622,988 (21.7) | 622,988 (21.7) | | |
| Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | 0.31 | .31 | | | .29 | .29 | | | <.001 |
| | Non-Hispanic White | 113 (28.3) | 113 (28.3) | 39,219 (24.9) | 39,219 (24.9) | | | 55,218 (24.2) | 55,218 (24.2) | | | 372,037 (12.9) | 372,037 (12.9) | | |
| | Non-Hispanic Black | 32 (8.0) | 32 (8.0) | 10,862 (6.9) | 10,862 (6.9) | | | 16,207 (7.1) | 16,207 (7.1) | | | 68,195 (2.4) | 68,195 (2.4) | | |
| | Hispanic | 194 (48.5) | 194 (48.5) | 78,853 (50.1) | 78,853 (50.1) | | | 117,162 (51.3) | 117,162 (51.3) | | | 1,356,354 (47.2) | 1,356,354 (47.2) | | |
| | Asian/Pacific Islander | 47 (11.8) | 47 (11.8) | 22,783 (14.5) | 22,783 (14.5) | | | 31,318 (13.7) | 31,318 (13.7) | | | 213,499 (7.4) | 213,499 (7.4) | | |
| | Others/unknown | 14 (3.5) | 14 (3.5) | 5784 (3.7) | 5784 (3.7) | | | 8432 (3.7) | 8432 (3.7) | | | 864,311 (30.1) | 864,311 (30.1) | | |
| Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | 0.36 | .36 | | | .20 | .20 | | | <.001 |
| | Less than high school | 9 (2.2) | 9 (2.2) | 4355 (2.8) | 4355 (2.8) | | | 6925 (3.0) | 6925 (3.0) | | | 435,360 (15.1) | 435,360 (15.1) | | |
| | High school graduate | 83 (20.8) | 83 (20.8) | 35,411 (22.5) | 35,411 (22.5) | | | 55,598 (24.4) | 55,598 (24.4) | | | 694,118 (24.1) | 694,118 (24.1) | | |
| | Some college | 99 (24.8) | 99 (24.8) | 32,616 (20.7) | 32,616 (20.7) | | | 50,153 (22.0) | 50,153 (22.0) | | | 558,288 (19.4) | 558,288 (19.4) | | |
| | Bachelor’s/associate’s degree | 126 (31.5) | 126 (31.5) | 54,293 (34.5) | 54,293 (34.5) | | | 75,849 (33.2) | 75,849 (33.2) | | | 729,896 (25.4) | 729,896 (25.4) | | |
| | Master’s degree/above | 73 (18.3) | 73 (18.3) | 27,388 (17.4) | 27,388 (17.4) | | | 34,556 (15.1) | 34,556 (15.1) | | | 317,698 (11.1) | 317,698 (11.1) | | |
| | Missing | 10 (2.5) | 10 (2.5) | 3438 (2.2) | 3438 (2.2) | | | 5256 (2.3) | 5256 (2.3) | | | 139,036 (4.8) | 139,036 (4.8) | | |
| Household income (US $) | Household income (US $) | Household income (US $) | Household income (US $) | Household income (US $) | Household income (US $) | Household income (US $) | 0.64 | .64 | | | .25 | .25 | | | —d |
| | <30,000 | 16 (4.0) | 16 (4.0) | 5194 (3.3) | 5194 (3.3) | | | 8318 (3.6) | 8318 (3.6) | | | — | — | | |
| | 30,000-49,999 | 90 (22.5) | 90 (22.5) | 39,969 (25.4) | 39,969 (25.4) | | | 61,562 (27.0) | 61,562 (27.0) | | | — | — | | |
| | 50,000-69,999 | 124 (31.0) | 124 (31.0) | 47,864 (30.4) | 47,864 (30.4) | | | 69,844 (30.6) | 69,844 (30.6) | | | — | — | | |
| | 70,000-89,999 | 82 (20.5) | 82 (20.5) | 32,486 (20.6) | 32,486 (20.6) | | | 45,469 (19.9) | 45,469 (19.9) | | | — | — | | |
| | ≥90,000 | 88 (22.0) | 88 (22.0) | 31,925 (20.3) | 31,925 (20.3) | | | 42,782 (18.7) | 42,782 (18.7) | | | — | — | | |
| Prenatal care initiation | Prenatal care initiation | Prenatal care initiation | Prenatal care initiation | Prenatal care initiation | Prenatal care initiation | Prenatal care initiation | 0.52 | .52 | | | <.001 | <.001 | | | <.001 |
| | First trimester | 379 (94.8) | 379 (94.8) | 147,017 (93.3) | 147,017 (93.3) | | | 199,866 (87.5) | 199,866 (87.5) | | | 2,386,232 (83.0) | 2,386,232 (83.0) | | |
| | No or late care | 20 (5.0) | 20 (5.0) | 9860 (6.3) | 9860 (6.3) | | | 26,966 (11.8) | 26,966 (11.8) | | | 442,493 (15.4) | 442,493 (15.4) | | |
| | Missing | 1 (0.2) | 1 (0.2) | 624 (0.4) | 624 (0.4) | | | 1505 (0.7) | 1505 (0.7) | | | 45,671 (1.6) | 45,671 (1.6) | | |
| Smoking during pregnancy | Smoking during pregnancy | Smoking during pregnancy | 23 (5.8) | 23 (5.8) | 6420 (4.1) | 6420 (4.1) | 0.09 | .09 | 10,256 (4.5) | 10,256 (4.5) | .16 | .16 | 46,977 (1.6) | 46,977 (1.6) | <.001 |
| Gestational age (weeks) | Gestational age (weeks) | Gestational age (weeks) | Gestational age (weeks) | Gestational age (weeks) | Gestational age (weeks) | Gestational age (weeks) | 0.13 | .13 | | | .08 | .08 | | | .16 |
| | <34 | 14 (3.5) | 14 (3.5) | 3412 (2.2) | 3412 (2.2) | | | 4779 (2.1) | 4779 (2.1) | | | 66,099 (2.3) | 66,099 (2.3) | | |
| | 34-36 | 32 (8.0) | 32 (8.0) | 9865 (6.3) | 9865 (6.3) | | | 13,933 (6.1) | 13,933 (6.1) | | | 180,352 (6.3) | 180,352 (6.3) | | |
| | ≥37 | 354 (88.5) | 354 (88.5) | 144,192 (91.5) | 144,192 (91.5) | | | 209,553 (91.8) | 209,553 (91.8) | | | 2,624,620 (91.3) | 2,624,620 (91.3) | | |
| | Missing | 0 (0.0) | 0 (0.0) | 32 (0.0) | 32 (0.0) | | | 72 (0.0) | 72 (0.0) | | | 3325 (0.1) | 3325 (0.1) | | |
## Outcomes
EHR outcomes were determined by the presence of PPD diagnosis codes in inpatient or outpatient encounters in the 12 months after delivery, new prescription order, or pharmacy dispense for the treatment of PPD. Diagnosis codes during the ICD-9 coding period were 300.4, 309.0, and 311 and during the ICD-10 period were F32.9, F33.0, F33.2, F33.3, F33.41, F33.9, F34.1, F43.21, and F53.0. Medications included were bupropion, Celexa, citalopram, Cymbalta, desvenlafaxine, duloxetine, Effexor, escitalopram, fluoxetine, Lexapro, paroxetine, Paxil, Pristiq, Prozac, sertraline, venlafaxine, Wellbutrin, and Zoloft.
Gold standard PPD outcomes were determined by review of health records by trained research personnel, who documented any diagnosis or finding of PPD in the record, including in free-text encounter notes, as well as any prescription given for the treatment of PPD. These included new prescriptions for the treatment of PPD. PPD diagnosis and medication were documented independently, both for the EHR data and the chart review. A mother was considered to have PPD if she had either a diagnosis or a prescription noted in the EHR within 1 year postpartum.
The overall agreement of EHR-identified PPD (based on either a diagnosis or a prescription) with medical record review was high, with a kappa of $84.7\%$ ($95\%$ CI $78.8\%$-$90.6\%$). The EHR identified 281 of 286 cases (sensitivity $98.3\%$, $95\%$ CI $96.0\%$-$99.4\%$) while maintaining high specificity ($95.0\%$, $95\%$ CI $88.7\%$-$98.4\%$), PPV ($93.7\%$, $95\%$ CI $90.3\%$-$96.1\%$), and NPV ($95.0\%$, $95\%$ CI $88.7\%$-$98.4\%$). There was little difference in the overall agreement between the ICD-9 coding era (κ=$86.0\%$, $95\%$ CI $78.0\%$-$94.0\%$) and the ICD-10 era (κ=$83.4\%$, $95\%$ CI $74.8\%$-$92.1\%$; Table 2).
**Table 2**
| Unnamed: 0 | Unnamed: 1 | TPa, n | TNb, n | FPc, n | FNd, n | Sensitivity, % (95% CI) | Specificity, % (95% CI) | PPVe, % (95% CI) | NPVf, % (95% CI) | Kappa (95% CI) | AUCg |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records | Combined electronic diagnosis codes and pharmacy records |
| | Overall | 281 | 95 | 19 | 5 | 98.3 (96.0-99.4) | 83.3 (75.2-89.7) | 93.7 (90.3-96.1) | 95.0 (88.7-98.4) | 0.85 (0.79-0.91) | 0.91 |
| | 2012-2014 | 141 | 48 | 9 | 2 | 98.6 (95.0-99.8) | 84.2 (72.1-92.5) | 94.0 (88.9-97.2) | 96.0 (86.3-99.5) | 0.86 (0.78-0.94) | 0.91 |
| | 2017-2019 | 140 | 47 | 10 | 3 | 97.9 (94.0-99.6) | 82.5 (70.1-91.3) | 93.3 (88.1-96.8) | 94.0 (83.5-98.7) | 0.83 (0.75-0.92) | 0.90 |
| ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only |
| | Overall | 187 | 101 | 13 | 99 | 65.4 (59.6-70.9) | 88.6 (81.3-93.8) | 93.5 (89.1-96.5) | 50.5 (43.4-57.6) | 0.44 (0.36-0.52) | 0.77 |
| | 2012-2014 | 94 | 51 | 6 | 49 | 65.7 (57.3-73.5) | 89.5 (78.5-96.0) | 94.0 (87.4-97.8) | 51.0 (40.8-61.1) | 0.45 (0.34-0.56) | 0.78 |
| | 2017-2019 | 93 | 50 | 7 | 50 | 65.0 (56.6-72.8) | 87.7 (76.3-94.9) | 93.0 (86.1-97.1) | 50.0 (39.8-60.2) | 0.43 (0.32-0.54) | 0.76 |
| Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only |
| | Overall | 194 | 108 | 6 | 92 | 67.8 (62.1-73.2) | 94.7 (88.9-98.0) | 97.0 (93.6-98.9) | 54.0 (46.8-61.1) | 0.51 (0.43-0.59) | 0.81 |
| | 2012-2014 | 97 | 54 | 3 | 46 | 67.8 (59.5-75.4) | 94.7 (85.4-98.9) | 97.0 (91.5-99.4) | 54.0 (43.7-64.0) | 0.51 (0.40-0.62) | 0.81 |
| | 2017-2019 | 97 | 54 | 3 | 46 | 67.8 (59.5-75.4) | 94.7 (85.4-98.9) | 97.0 (91.5-99.4) | 54.0 (43.7-64.0) | 0.51 (0.40-0.62) | 0.81 |
Electronic diagnosis records alone were not able to accurately identify PPD, only identifying 187 of 286 cases (sensitivity $65.4\%$, $95\%$ CI $59.6\%$-$70.9\%$), with low NPV ($50.5\%$, $95\%$ CI $43.4\%$-$57.6\%$). PPV ($93.5\%$, $95\%$ CI $89.1\%$-$96.5\%$) and specificity ($88.6\%$, $95\%$ CI $81.3\%$-$93.8\%$) were high, however (Table 2). Results were similar when using EHR prescription records alone (sensitivity $67.8\%$, $95\%$ CI $62.1\%$-$73.2\%$; specificity $94.7\%$, $95\%$ CI $88.9\%$-$98.0\%$; PPV $97.0\%$, $95\%$ CI $93.6\%$-$98.9\%$; NPV $54.0\%$, $95\%$ CI $46.8\%$-$61.1\%$).
Considering only medication data, the reliability of EHR data for identifying prescriptions for PPD was high, with an overall kappa of $92.5\%$ ($95\%$ CI $88.8\%$-$96.2\%$). Agreement was very high in both the ICD-9 (κ=$92.0\%$, $95\%$ CI $86.6\%$-$97.4\%$) and ICD-10 eras (κ=$93.0\%$, $95\%$ CI $87.9\%$-$98.1\%$; Table 3). Sensitivity, specificity, PPV, and NPV were all at or above $96\%$ (Table 3).
Agreement for ICD diagnostic codes between EHR and manual chart review was much lower overall (κ=$55.0\%$, $95\%$ CI $47.1\%$-$62.9\%$; Table 3). The PPV was high ($90.0\%$, $95\%$ CI $85.0\%$-$93.8\%$), with sensitivity lower ($72.0\%$, $95\%$ CI $66.0\%$-$77.5\%$) and specificity and NPV much lower (both $65.0\%$, $95\%$ CI $58.0\%$-$71.6\%$; Table 3). Agreement was similar between the ICD-9 (κ=$58.0\%$, $95\%$ CI $47.1\%$-$68.9\%$) and ICD-10 (κ=$52.0\%$, $95\%$ CI $40.5\%$-$63.5\%$) eras (Table 3).
**Table 3**
| Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | TPa, n | bTN, n | bTN, n.1 | FPc, n | FPc, n.1 | FNd, n | FNd, n.1 | Sensitivity, % (95% CI) | Specificity, % (95% CI) | Specificity, % (95% CI).1 | PPVe, % (95% CI) | NPVf, % (95% CI) | NPVf, % (95% CI).1 | Kappa (95% CI) | Kappa (95% CI).1 | AUCg |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only | ICD-9h/ICD-10i diagnosis codes only |
| | Overall | 180 | 180 | 130 | 20 | 20 | 70 | 70 | 72.0 (66.0-77.5) | 72.0 (66.0-77.5) | 86.7 (80.2-91.7) | 90.0 (85.0-93.8) | 90.0 (85.0-93.8) | 65.0 (58.0-71.6) | 0.55 (0.47-0.63) | 0.55 (0.47-0.63) | 0.79 | 0.79 |
| | 2012-2014 | 92 | 92 | 66 | 8 | 8 | 34 | 34 | 73.0 (64.4-80.5) | 73.0 (64.4-80.5) | 89.2 (79.8-95.2) | 92.0 (84.8-96.5) | 92.0 (84.8-96.5) | 66.0 (55.8-75.2) | 0.58 (0.47-0.69) | 0.58 (0.47-0.69) | 0.78 | 0.78 |
| | 2017-2019 | 88 | 88 | 64 | 12 | 12 | 36 | 36 | 71.0 (62.1-78.8) | 71.0 (62.1-78.8) | 84.2 (74.0-91.6) | 88.0 (80.0-93.6) | 88.0 (80.0-93.6) | 64.0 (53.8-73.4) | 0.52 (0.41-0.64) | 0.52 (0.41-0.64) | 0.81 | 0.81 |
| Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only | Pharmacy records only |
| | Overall | 192 | 192 | 193 | 8 | 8 | 7 | 7 | 96.5 (92.9-98.6) | 96.5 (92.9-98.6) | 96.0 (92.3-98.3) | 96.0 (92.3-98.3) | 96.0 (92.3-98.3) | 96.5 (92.9-98.6) | 0.93 (0.89-0.96) | 0.93 (0.89-0.96) | 0.96 | 0.96 |
| | 2012-2014 | 96 | 96 | 96 | 4 | 4 | 4 | 4 | 96.0 (90.1-98.9) | 96.0 (90.1-98.9) | 96.0 (90.1-98.9) | 96.0 (90.1-98.9) | 96.0 (90.1-98.9) | 96.0 (90.1-98.9) | 0.92 (0.87-0.97) | 0.92 (0.87-0.97) | 0.97 | 0.97 |
| | 2017-2019 | 96 | 96 | 97 | 4 | 4 | 3 | 3 | 97.0 (91.4-99.4) | 97.0 (91.4-99.4) | 96.0 (90.2-98.9) | 96.0 (90.1-98.9) | 96.0 (90.1-98.9) | 97.0 (91.5-99.4) | 0.93 (0.88-0.98) | 0.93 (0.88-0.98) | 0.96 | 0.96 |
## Quality Assurance
Multiple individuals were trained on reviewing charts, and a double chart review was performed at the beginning of data collection as a training exercise and near the middle and at the end of data collection to verify data quality and consistency. At each point, eight charts were randomly selected for review by two abstractors. In case of disagreement on the findings, abstractors met with the trainer to determine the correct result.
During the training process, 8 charts were independently reviewed by two chart abstractors. Their assessments of medication use for PPD agreed for all 8 records ($100\%$), while the assessment of a diagnostic finding agreed for 7 ($88\%$). After training was complete, another 8 records were independently reviewed. All 8 ($100\%$) agreed in their findings for both medications and diagnoses.
## Statistical Analysis
The patient population was described in terms of demographics, smoking status, prenatal care, and birth weight using percentages. These characteristics were also described for the study population of the APPCUE study [19] and all live births among KPSC members and the state of California during the study timeframe. The chi-square test was used to compare the distribution of characteristics in the study sample to the APPCUE population, all KPSC births, and the California birth cohort.
Manual chart review findings were treated as the true PPD status. The sensitivity, specificity, PPV, and negative predictive value (NPV) of the electronic records to identify true PPD status were calculated and presented as a percentage and $95\%$ exact binomial CI. Agreement between electronic records and manual review was calculated using the kappa statistic, which adjusts for agreement expected due to random chance, and its $95\%$ CI. The area under the receiver operating characteristic curve was calculated. Each measure was calculated overall and within the ICD-9 and ICD-10 coding eras separately. There was no missing data for PPD status; those without documented PPD diagnosis or medication were taken to not have PPD. For patient characteristics, a missing category was included when presenting the data.
The primary analysis focused on the ability of EHRs to capture PPD, while secondary analyses examined the agreement of diagnosis and prescription records separately. The sample size was selected so that the expected width of the CIs for sensitivity and PPV would be at most $10\%$ for the full sample and $13\%$ for the ICD-9 and ICD-10 periods if the true sensitivity and PPV were $80\%$. Higher sensitivity and PPV would yield narrower CIs. The STARD (Standards for Reporting Diagnostic Accuracy Studies) guidelines were followed. All analyses were performed in SAS version 9.4 (SAS Institute).
## Ethics Approval
The study was approved by the institutional review board of KPSC and received a waiver for informed consent (IRB 12110).
## Principal Findings
This validation study demonstrated the potential to improve the accuracy of PPD case identification from an EHR when using diagnosis codes in conjunction with pharmacy records. The combination of clinical codes and prescription pharmacy records yielded much greater sensitivity and NPV, with no notable loss in specificity or PPV, compared with using either the diagnosis codes or pharmacy records alone. Using either record alone would result in significant undercounting, each missing about one-third of those with PPD, compared to the $95\%$ identified using both together. Furthermore, we observed no significant difference in the ICD-9 and ICD-10 codes in terms of ascertaining PPD cases.
We found that electronic records of PPD diagnosis were not a reliable indicator of PPD diagnostic findings identified through chart review, relative to pharmacy records. Pharmacy records have both a sensitivity and specificity much higher than that seen for diagnosis codes.
The quality of data extracted from EHRs for pharmacoepidemiologic research has been proven to be valuable. Although using clinical diagnosis codes for perinatal epidemiology studies has limitations, the use of KPSC’s comprehensive pharmacy use records enhances the identification of PPD cases (sensitivity $98.3\%$, specificity $95.0\%$, PPV $93.7\%$, and NPV $95.0\%$).
While switching from ICD-9 to ICD-10 coding created some complexity, we did not see a significant difference in the accuracy of the electronic diagnosis records between the ICD-9 and ICD-10 coding eras. This is reassuring, as studies would not need to be limited to one era or the other for the sake of accuracy. Additionally, the prevalence of PPD identified in both periods is essentially the same, suggesting that both ICD-9 and ICD-10 coding systems identify patients with PPD at the same rate, negating any need to adjust prevalence estimates to account for the difference.
Accurate characterization of those with PPD is crucial to performing valid research on this condition. Many researchers rely on electronic records due to a lack of access to detailed patient histories or a lack of time to review these records. Our study suggests that researchers can accurately identify PPD from EHRs using both diagnosis and pharmacy records.
## Comparison to Prior Work
Prior research validating diagnosis codes for identifying general depression found the PPV to be similar to that seen in our study ($89.7\%$ for ICD-9 and $89.5\%$ for ICD-10), but these were not specific to the postpartum period [18]. These findings highlight the continuing debate regarding the use of diagnosis codes alone for epidemiological studies. Our study concurs with prior findings that the sensitivity and specificity of case ascertainment can be improved by concurrently using both diagnosis and pharmacy records [13]. Therefore, researchers should not rely exclusively on either diagnostic codes or pharmacy records for PPD case ascertainment.
## Strengths and Limitations
There are some potential limitations to this study. First, while the KPSC EHR is comprehensive, it may not capture care received outside the system if it is not submitted for reimbursement. Specifically, members may receive mental health counseling from non-KPSC providers, and a PPD diagnosis made in that setting may not be entered into the KPSC medical record, resulting in a potentially missed PPD diagnosis and an underestimate of the sensitivity of diagnosis coding. However, these diagnoses may still be identified during regular clinical care within KPSC, hence limiting the number of potentially missed diagnoses.
Second, misclassification is also possible as variables were ascertained from clinical diagnosis codes and pharmacy record notes. In addition, there is the potential for misclassification of PPD within the data sources if women are unaware of the condition, do not seek medical care, or the diagnosis or treatment is not recorded in the clinical notes. Any completely undocumented cases would result in an underestimate of PPD in the population, though its potential effect on our validation is unknown. Finally, due to the small number of records reviewed in some groups, we were not able to look for differences in medical record accuracy within subsets of the population, including by age and race/ethnicity. If differences are present, this will limit the generalizability of these findings to other populations with different demographics.
Strengths of this study include the comprehensive medical record and chart review conducted to identify PPD in this patient population. The training and validation of the chart review process helped to ensure that the gold standard PPD identification was accurate.
## Conclusions
This validation study of PPD that was carried out in a large integrated health care system in Southern California has demonstrated that PPD data ascertainment based on a combination of diagnosis codes and prescription medication records from the EHR is highly accurate for pharmacoepidemiologic studies. Neither diagnosis codes alone nor prescription records alone are sufficient to capture PPD cases.
## Data Availability
Most of the data that support the findings of this study are available on request from the corresponding author. The complete data set is not publicly available due to privacy, institutional approval, and/or ethical restrictions. Study data come from patient electronic health records and birth certificates from the state of California. Data from patient health records cannot be shared without signed confidentiality agreements. Some of the data that support the findings of this study are available from the state of California. Restrictions apply to the availability of these data, which were used under license and approval for this study. Data can be made available by the authors provided that all required approvals are obtained from the departments in the state that oversees the use of state vital records data. To obtain California birth certificate data, researchers can email [email protected] or visit their website [21]. Requests for data may be sent to JS ([email protected]) and DG ([email protected]).
## References
1. Bauman BL, Ko JY, Cox S, D'Angelo Mph DV, Warner L, Folger S, Tevendale HD, Coy KC, Harrison L, Barfield WD. **Vital signs: postpartum depressive symptoms and provider discussions about perinatal depression - United States, 2018**. *MMWR Morb Mortal Wkly Rep* (2020.0) **69** 575-581. DOI: 10.15585/mmwr.mm6919a2
2. **ACOG Committee Opinion No. 757: screening for perinatal depression**. *Obstet Gynecol* (2018.0) **132** e208-e212. DOI: 10.1097/AOG.0000000000002927
3. Earls MF, Yogman MW, Mattson G, Rafferty J. **Incorporating recognition and management of perinatal depression into pediatric practice**. *Pediatrics* (2019.0) **143** e20183259. DOI: 10.1542/peds.2018-3259
4. Siu AL, Bibbins-Domingo K, Grossman DC, Baumann LC, Davidson KW, Ebell M, García FAR, Gillman M, Herzstein J, Kemper AR, Krist AH, Kurth AE, Owens DK, Phillips WR, Phipps MG, Pignone MP. **Screening for depression in adults: US Preventive Services Task Force recommendation statement**. *JAMA* (2016.0) **315** 380-7. DOI: 10.1001/jama.2015.18392
5. **Pregnancy-related deaths: data from 14 U.S. Maternal Mortality Review Committees, 2008-2017**. *Centers for Disease Control and Prevention* (2019.0)
6. Wrate RM, Rooney AC, Thomas PF, Cox JL. **Postnatal depression and child development. A three-year follow-up study**. *Br J Psychiatry* (1985.0) **146** 622-7. DOI: 10.1192/bjp.146.6.622
7. Lancaster CA, Gold KJ, Flynn HA, Yoo H, Marcus SM, Davis MM. **Risk factors for depressive symptoms during pregnancy: a systematic review**. *Am J Obstet Gynecol* (2010.0) **202** 5-14. DOI: 10.1016/j.ajog.2009.09.007
8. Robertson E, Grace S, Wallington T, Stewart DE. **Antenatal risk factors for postpartum depression: a synthesis of recent literature**. *Gen Hosp Psychiatry* (2004.0) **26** 289-95. DOI: 10.1016/j.genhosppsych.2004.02.006
9. Cook N, Ayers S, Horsch A. **Maternal posttraumatic stress disorder during the perinatal period and child outcomes: a systematic review**. *J Affect Disord* (2018.0) **225** 18-31. DOI: 10.1016/j.jad.2017.07.045
10. **International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)**. *Centers for Disease Control and Prevention* (2021.0)
11. **ICD-10-CM Browser Tool**. *Centers for Disease Control and Prevention* (2021.0)
12. Stewart CC, Lu CY, Yoon TK, Coleman KJ, Crawford PM, Lakoma MD, Simon GE. **Impact of ICD-10-CM transition on mental health diagnoses recording**. *EGEMS (Wash DC)* (2019.0) **7** 14. DOI: 10.5334/egems.281
13. Colvin L, Slack-Smith L, Stanley FJ, Bower C. **Are women with major depression in pregnancy identifiable in population health data?**. *BMC Pregnancy Childbirth* (2013.0) **13** 63. DOI: 10.1186/1471-2393-13-63
14. Pereira AT, Bos SC, Marques M, Maia BR, Soares MJ, Valente J, Gomes AA, Macedo A, de Azevedo MHP. **The postpartum depression screening scale: is it valid to screen for antenatal depression?**. *Arch Womens Ment Health* (2011.0) **14** 227-38. DOI: 10.1007/s00737-010-0178-y
15. Smith-Nielsen J, Matthey S, Lange T, Væver MS. **Validation of the Edinburgh Postnatal Depression Scale against both DSM-5 and ICD-10 diagnostic criteria for depression**. *BMC Psychiatry* (2018.0) **18** 393. DOI: 10.1186/s12888-018-1965-7
16. Hochman E, Feldman B, Weizman A, Krivoy A, Gur S, Barzilay E, Gabay H, Levy J, Levinkron O, Lawrence G. **Development and validation of a machine learning-based postpartum depression prediction model: a nationwide cohort study**. *Depress Anxiety* (2021.0) **38** 400-411. DOI: 10.1002/da.23123
17. Betts KS, Kisely S, Alati R. **Predicting postpartum psychiatric admission using a machine learning approach**. *J Psychiatr Res* (2020.0) **130** 35-40. DOI: 10.1016/j.jpsychires.2020.07.002
18. Fiest KM, Jette N, Quan H, St Germaine-Smith C, Metcalfe A, Patten SB, Beck CA. **Systematic review and assessment of validated case definitions for depression in administrative data**. *BMC Psychiatry* (2014.0) **14** 289. DOI: 10.1186/s12888-014-0289-5
19. Sun Y, Li X, Benmarhnia T, Chen J, Avila C, Sacks DA, Chiu V, Slezak J, Molitor J, Getahun D, Wu J. **Exposure to air pollutant mixture and gestational diabetes mellitus in Southern California: results from electronic health record data of a large pregnancy cohort**. *Environ Int* (2022.0) **158** 106888. DOI: 10.1016/j.envint.2021.106888
20. **Natality information: live births**. *CDC WONDER*
21. **Committee for the Protection of Human Subjects**. *California Health and Human Services*
|
---
title: 'Outcomes of Hallux Amputation Versus Partial First Ray Resection in People
with Non-Healing Diabetic Foot Ulcers: A Pragmatic Observational Cohort Study'
authors:
- Virginie Blanchette
- Louis Houde
- David G. Armstrong
- Brian M. Schmidt
journal: The International Journal of Lower Extremity Wounds
year: 2022
pmcid: PMC10018408
doi: 10.1177/15347346221122859
license: CC BY 4.0
---
# Outcomes of Hallux Amputation Versus Partial First Ray Resection in People with Non-Healing Diabetic Foot Ulcers: A Pragmatic Observational Cohort Study
## Body
Diabetes mellitus (DM) is one of the most common chronic diseases worldwide. DM-related foot complications such as peripheral arterial disease, diabetic foot infection (DFI), diabetic foot ulcer (DFU) and minor or major lower extremity amputation (LEA) reduce the quality of life and lead to premature death.1,2 Personal, societal and economic burdens of DFUs highlight the importance to support prevention strategies for the at-risk population as well as effective treatments that will prevent DFU recurrence, re-amputation or other complications such DFI and death.3,4 Indeed, DFI is involved in $58\%$ of DFU and approximately $50\%$ of these infected patients are affected with PAD. PAD is highly predictive of LEA.5–8 Approximately $17\%$ to $30\%$ of people with a DFU will ultimately require a LEA and patients with DFI have 155 times greater risk of LEA than patients without associated infection.3,7,9,10 *It is* estimated that $85\%$ of all DM-related LEA are preceded by a DFU but sometime, LEAs are an inevitable treatment. 11 The key components of successful limb salvage are to achieve a DFU-free, plantigrade foot that is functional with treatments that have minimum impact on a patient’s global health. A successful LEA is i) the complete eradication of nonviable tissue to optimize the patient healing potential, ii) reduce the risk of DFU recurrence (or new DFU onset) and iii) avoid the need for extended local wound care or repeat surgical interventions.12,13 The goal of isolated partial-foot amputation, such as a hallux amputation and a partial first ray resection, is to maintain bipedal ambulatory status and function.14,15 Minor LEA are preferred to major LEA because of their association with less morbidity and mortality.16,17 The forefoot has been reported as the most frequent location of DFI in DM. 18 Furthermore, the metatarsophalangeal joint of the hallux, including sesamoid bones, is more complex from an anatomical perspective than the lesser metatarsophalangeal joints. Such differences in anatomy might impact surgical outcomes. 18 However, first ray amputations (eg, hallux disarticulation and/or partial first ray amputation) impact a patient’s gait pattern because of the absence of the propulsive phase provided by now altered medial column of the foot.19,20 Although those procedures seem to affect gait less than a more proximal LEA, published studies have reported that patients who undergo partial first ray resection often progress to requiring a more proximal repeat LEA.13,21 Moreover, following hallux amputation, subsequent higher level of amputation is frequently observed due to new infected DFU associated diabetes limited joint mobility and new ambulatory pattern because of the amputated hallux. 22 Furthermore, the literature comparing outcomes following hallux amputation or partial first ray resection are limited. 15 In similar context, the choice to perform one of these two surgeries is attributable to the clinician’s decision according to their experience, to the patient’s DFU characteristics and patient’s preference through informed consent. Hence, guidelines are suggesting clinical decision based on several factors (eg, functional, infection and vascular status, bone quality, presence of infection, etc) with the intent to preserve as much of the limb as possible.23–28 The aim was to determine the most definitive surgery between hallux amputation and partial first ray resection for patients with infected ulcer (+ /-) osteomyelitis involving the first ray who were followed for 1-year postoperatively. Our primary objective was to compare DFU events (at the surgical site and/or the ipsilateral foot only) at 3-, 6- and 12-months following the surgical intervention in patients who had hallux amputation or partial first ray resection. Our secondary aim was to compare other outcomes between both cohorts (eg, infection, re-amputation, death). We hypothesized hallux amputation would be most definitive and result in less complications during the 1-year follow up, in line with similar trends from previous studies.19,21 It have been reported that patients who undergo partial first ray resection often progress to requiring a more proximal re-amputation. 21
## Abstract
There are few data comparing outcomes after hallux amputation or partial first ray resection after diabetic foot ulcer (DFU). In a similar context, the choice to perform one of these two surgeries is attributable to clinician preference based on experience and characteristics of the patient and the DFU. Therefore, the purpose of this study was to determine the more definitive surgery between hallux amputation and partial first ray resection. We abstracted data from a cohort of 70 patients followed for a 1-year postoperative period to support clinical practice. We also attempted to identify patient characteristics leading to these outcomes. Our results suggested no statistical difference between the type of surgery and outcomes such as recurrence of DFU and amputation at 3, 6, and 12 months or death. However, there was a statistically significantly increased likelihood of re-ulceration for patients with CAD who underwent hallux amputation ($$p \leq 0.02$$). There was also a significantly increased likelihood of re-ulceration for people with depression or a history when the partial ray resection was performed ($$p \leq 0.02$$). Patients with prior amputation showed a higher probability of undergoing another re-amputation with partial ray resection ($$p \leq 0.01$$). Although the trends that emerge from this project are limited to what is observed in this statistical context, where the number of patients included and the number of total observations per outcome were limited, it highlights interesting data for future research to inform clinical decisions to support best practices for the benefit of patients.
## Materials and Methods
We performed a observational cohort investigation (retrospective; level of evidence III) which mined and analyzed big data, with coding, from single unified Electronic Medical Records (EMR) at University of Michigan Health System, a large tertiary academic health system overseeing the care of more than 80,000 patients with DM. 29 Between 2016 and 2020, 70 patients from which 26 had hallux amputation and 44, a partial first ray resection, were retrieved from database and followed for longitudinal outcomes on a one-year period. According to sample size calculation, 38 to 216 patients are sufficient power for confidence interval between 90–$95\%$ in the conservative proportion of LEA ($17\%$). 30 All patients underwent comprehensive medical treatment and surgical intervention by a multidisciplinary team, which included five board-certified podiatric surgeons (for the amputations), nurses, vascular surgeons, and structured and targeted diabetic foot care according to the International Working Group on Diabetic Foot recommendations (IWGDF). 24 *Inclusion criteria* were adult DM patients age ≥18 with a concomitant diabetic foot surgery whether hallux amputation or partial first ray resection that EMR reported data over a 1-year period. Our EMR mining system was programmed to include limb salvage procedural codes, based on Common Procedure Terminology (CPT) for higher-level amputations (CPT 84.13-84.19), minor lower extremity amputations (CPT 84.10-84.12). The hallux amputation is defined as the level of amputation distal to the first metatarsophalangeal, including the hallux and the joint.15,17,31 Partial first ray resection is defined as the primary amputation of the hallux phalanxes and at least a part of the first metatarsus, distal to the first metatarsal–cuneiform joint and excluded additional digital amputations.14,17,21
## Outcomes Measures
Data collected included demographic information (eg, age, sex, race, body mass index, coronary heart disease, hypertension, etc (Table 1). The outcome measures were related DFU healing after the LEA on a 1-year period. DFU healing was defined as a continuous, viable epithelial covering over the entire previously open wound, subsequently within 2 months with no new ulcerations. Complications associated with each surgical approach (DFU at 3-, 6- and 12 months, re-amputation at 3-,6-and 12- months and death) were also collected.
**Table 1.**
| Characteristics | Total(n = 70) | Partial First Ray Resection(n = 44) | Hallux amputation(n = 26) | P-value |
| --- | --- | --- | --- | --- |
| Age, years, mean ± SD | 57.4 ± 11.0 | 56.3 ± 10.6 | 59.3 ± 11.2 | 0.27 |
| Sex, % Men (n) | 85.7 (60) | 86.4 (38) | 84.6 (22) | 0.84 |
| Race, % (n) Caucasian Others† | 78.6 (55)21.4 (15) | 70.5 (31)29.5 (13) | 92.3 (24)7.7 (2) | 0.03* |
| BMI (kg/m2) mean ± SD | 32.8 ± 7.0 | 32.4 ± 7.0 | 33.4 ± 7.2 | 0.58 |
| Previous Amputation % (n) | 27.1 (19) | 27.3 (12) | 26.9 (7) | 0.22 |
| Presence of SIRS % (n) | 11.4 (8) | 11.4 (5) | 11.5 (3) | 0.80 |
| IDSA Classification 25 % (n) 1 : None 2 : Mild 3 : Moderate 4 : Severe | 1.4 (1)41.4 (29)45.7 (32)11.4 (8) | 0 (0)36.4 (16)50 (22)13.6 (6) | 3.8 (1)50.0 (13)38.5 (10)7.7 (2) | > 0.5 |
| Presence of OM‡ | 91.3 (63) | 93.2 (41) | 88.0 (22) | 0.25 |
| CAD % (n) | 44.3 (31) | 45.5 (20) | 42.3 (11) | 0.19 |
| HTN % (n) | 35.7 (25) | 34.1 (15) | 38.4 (10) | 0.72 |
| CKD stage, % (n) Stage 0 (no CKD) (GFR > 90 mL/min) Stage 1 (GFR = 60–89 mL/min) Stage 2 (GFR = 45–59 mL/min) Stage 3 (GFR = 30–44 mL/min) Stage 4 (GFR = 15–29 mL/min) Stage 5 CKD (GFR <15 mL/min) | 57.1 (40) 35.7 (25) 1.4 (1) 2.9 (2) 1.4(1) 1.4 (1) | 63.6 (28) 27.3 (12) 0 (0) 2.5 (2) 2.3 (1) 2.3 (1) | 46.2 (12) 50.0 (13) 3.8 (1) 0 (0) 0 (0) 0 (0) | > 0.5 |
| Smoking, % (n) | 32.9 (23) | 31.8 (14) | 34.6 (9) | 0.81 |
| DPN | 91.4 (64) | 93.2 (41) | 88.5 (23) | 0.50 |
| PLT, K/uL, mean ± SD | 267.8 ± 118.4 | 277.5 ± 133.9 | 251.4 ± 86.4 | 0.52 |
| HBG g/dL, mean ± SD | 11.7 ± 1.8 | 11.4 ± 2.0 | 12.1 ± 1.3 | 0.02* |
| ESR mm/hr, mean ± SD# | 73.6 ± 36.3 | 82.35 ± 37.3 | 58.9 ± 27.7 | 0.04* |
| MCV fI, mean ± SD | 86.4 ± 6.6 | 85.6 ± 6.4 | 87.7 ± 6.9 | 0.82 |
| Glucose mg/dL, mean ± SD¶ | 206.5 ± 125.1 | 228.1 ± 136.4 | 170.9 ± 95.6 | 0.08 |
| C-RP mg/dL, mean ± SD¥ | 11.5 ± 9.7 | 13.09 ± 9.5 | 8.65 ± 9.6 | 0.98 |
| TBI, mean ratio (amputation side when possible)§ | 0.55 ± 0.24 | 0.51 ± 0.25 | 0.63 ± 0.17 | 0.12 |
| Non compressible vessel due to calcification§ | 22.2 (12) | 22.9 (8) | 22.2 (4) | 0.99 |
| Previous revascularization, % (n) | 11.4 (8) | 15.0 (6) | 7.7 (2) | 0.45 |
| Depression& % (n) | 15.7 (11) | 15.9 (7) | 15.4 (4) | 0.95 |
| Ulcer classification (UT) 35 previous to amputation, % (n) 1B 2B 2C 2D 3A 3B 3C 3D 4B 4D Missing data | 2.9 (2)17.1 (12)1.9 (1)7.1 (5)4.3 (3)27.1 (19)1.9 (1)8.6 (6)12.9 (9)4.3 (3)12.9 (9) | 2.3 (1)9.1 (4)0 (0)11.4 (5)2.3 (1)27.3 (12)0 (0)11.4 (5)18.1 (8)4.5 (2)13.6 (6) | 3.8 (1)30.8 (8)3.8 (1)0 (0)7.7 (2)26.9 (7)3.8 (1)3.8 (1)3.8 (1)3.8 (1)11.5 (3) | > 0.05 |
## Data Analysis
Demographic data were analyzed using descriptive statistics. To compare the grades both groups, the characteristics were analyzed using chi square (χ2). Re-ulceration and re-amputation (or better ulcer-free and amputation-free survival) are time-dependent measures that can be reported as Kaplan-Meier curves. However, our retrospective data have allowed only time estimates (in months; not precise, as they were agglomerated). Since we cannot be very precise related to the time, which is important in Kaplan-Meier curves, we performed Mann-Whitney U test (non-parametric) and Friedman test on the independent samples to compare the means of the quantitative variables related to the outcomes. When the sample sizes were not sufficient to the accurate p-value we did adjustment using a bootstrap method. We performed a multivariate logistic regression per variable for patients’ characteristics known to be predictor factors for DFU and LEA according to the literature and our previous work. 32 Odd ratio was the association measure for continuous data. The χ2 was used to measure the independence of the dichotomous and multinomials variables between surgical type (hallux amputation or partial ray resection), the outcomes (cumulative re-ulceration or re-amputation) related to the variable interest. Odd ratios cannot be calculated in this statistical context. This was expressed using proportion. The death as outcomes could not be assessed with the regression because there were too few events for the sample size. P-value inferior to 0.05 was considered a significant association between outcomes and those factors in this analysis. This study is reported according to the STROCSS 2019 guidelines. 33 It was approved by the Institutional Review Board (HUM00108607) and it was completed in accordance with the ethical standards of the Ethics Committee. We used SPSS Statistics software 27 (IBM Corp, New York, United States) to perform the analysis.
## Demographics and Clinical Characteristics
A total of 70 patients who underwent first ray amputation surgery or hallux amputation were included in the study. The total cohort is mainly Caucasian ($78.8\%$) male ($85.7\%$) with an average age of 57.4 years (Table 1). DFU clinical presentation during hospital admission was primarily used to determine necessity of operative intervention. Ten patients ($38.4\%$) in the hallux cohort and fifteen patients ($34.1\%$) in the partial first ray cohort had index DFU on the left foot requiring surgical intervention. Neuropathic wound etiologies accounted for $92.2\%$ and $88.5\%$ in the hallux and partial first ray amputation cohorts, respectively. Although we had missing data for the vascular component, calcified vessels accounted for $22.2\%$ and limited accurate reporting of vascular status. It is known at least $11.4\%$ of the cohort had prior revascularization and ischemia was mild to moderate. 34 However, the majority of DFU were classified according to the University of Texas classification which accounts for an ischemic component of the index DFU (ie, class C or D). 35 All patients except one (in the partial ray resection cohort) were ambulatory prior to the amputation.
Pre-operative imaging was obtained in all patients to assist in operative planning. Radiographs were obtained in all patients and advanced imaging via magnetic resonance imaging (MRI) was obtained in 25 ($56.8\%$) and 18 ($69.2\%$) in the partial first ray and hallux amputations cohorts, respectively ($p \leq 0.05$). The rates of OM diagnosed was ($93.2.0\%$ v. $88.0\%$, $$p \leq 0.25$$). Prior to amputation, Charlson Comorbidity Index (CCI) values (5.4 ± 3.5 v. 4.7 ± 2.6; $p \leq 0.05$), IDSA classification at time of admission (2.5 ± 0.7 v. 2.8 ± 0.7; $p \leq 0.32$), leukocyte count (9.4 ± 4.6 v. 12 ± 6.9; $p \leq 0.05$) were similar. Patient characteristics were relatively similar and did not reach statistical significance ($p \leq 0.05$) for all variables (Table 1). Inflammatory markers including erythrocyte sedimentation rate (ESR) and C-reactive protein (C-RP) demonstrated divergence in our population. ESR demonstrated increased elevation in the partial ray group versus the hallux cohort (58.9 ± 27.7 v. 82.4 ± 37.3 $$p \leq 0.04$$), but the acute phase reactant C-RP did not demonstrate a difference (8.6 ± 9.6 v. 13.1 ± 9.5; $$p \leq 0.98$$). The partial first ray resection group was more ethnically diverse ($29.5\%$ v. $7.2\%$; $$p \leq 0.03$$) and also had a lower hemoglobin level (11.4 ± 2.0 v. 12.1 ± 1,3; $$p \leq 0.02$$).
## Outcomes
Of the 70 patients, all had defined primary outcomes at 1 year (Table 2). In the hallux amputation group, six ($23\%$), three ($12\%$), and two ($8\%$) developed ulcer recurrence within 3-, 6-, and 12 months post-operatively, respectively. Similarly, in the partial ray group, 16 ($36\%$), 8 ($18\%$), and 6 ($14\%$) developed re-ulceration within 3-, 6-, and 12-months postoperative follow-up, respectively. The difference among cohorts did not reach statistical significance. Re-amputation occurred in 0, 3 ($12\%$), and 0 patients and 6($14\%$), 5($11\%$), and 1 ($2\%$) in the hallux and partial first ray amputation groups, respectively, at 3-, 6-, and 12-months follow-up periods. The difference in rate of re-amputation was not significant at any time point in longitudinal follow-up. Additionally, two patients in the hallux amputation group and four in the partial first ray amputation group died; no deaths were related to surgical intervention or foot infection.
**Table 2.**
| Unnamed: 0 | 3-month ulcer n (%) | 3-month amputation n (%) | 6-month ulcer n (%) | 6-month amputation n (%) | 12-month ulcer n (%) | 12-month amputation n (%) | Death† n (%) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Partial First Ray Resection (n = 44) | 16 (36) | 6 (14) | 8 (18) | 5 (11) | 6 (14) | 1 (2) | 4 (9) |
| Hallux Amputation (n = 26) | 6 (23) | 0 (0) | 3 (12) | 3 (12) | 2 (8) | 0 (0) | 2 (8) |
| p-value | 0.295 | 0.078 | 0.521 | 1 | 0.701 | 1 | 1 |
## Factors Associated with Outcomes and Surgical Procedures
Although the association was not statically significant for chronic kidney disease (CKD) (Table 3), a trend was observed in the association between having a re-ulceration at one-year and having undergone partial ray resection amputation versus hallux amputation (OR 4.15 vs 0.53; $p \leq 0.05$). In terms of baseline demographic, clinical and laboratory characteristics, only three factors were found to influence outcomes with statistically significant differences (Table 4). Patients had a higher probability of re-ulceration in the hallux amputation cohort ($54.5\%$; $$p \leq 0.02$$) if they had coronary artery disease (CAD). The same was not true in the partial ray resection cohort ($45.0\%$; $$p \leq 0.96$$). For patients with current or a history of depression (not specified in the EMR), the partial first ray resection cohort had more re-ulcerations ($85.7\%$; $$p \leq 0.02$$) compared with the likelihood of the hallux cohort ($50\%$; $$p \leq 0.37$$). A higher probability to have a re-amputation was found for patient in the partial ray resection cohort ($58.3\%$; $$p \leq 0.01$$) compared to the other cohort probability ($14.3\%$; $$p \leq 0.79$$) when they presented with a prior history of amputation.
## Discussion
This study reported outcome difference between hallux amputation and partial first ray resection in a retrospective patient cohort of 70 patients followed on 1-year postoperative period and intended to support decision-making. Although the groups were slightly different at the baseline, especially related to two laboratory tests (HBG and ESR), the characteristics of the DFU, age and sex were similar. HBG and ESR, respectively associated with anemia and infection, are recognized as markers of morbidity and mortality in patients with DFU and to increase amputation risk.36,37 Moreover, there was a greater population’s diversity in the partial ray resection cohort, which could have also influenced the results. Indeed, it is well known that some ethnicity undergo more major amputations. 38 Recent studies have demonstrated that American Africans have more minor LEA when they have DFU infection, but there is less LEA in the Asian population.39,40 *As a* result, we would have expected to observe more outcomes in the partial ray resection group. However, our results did not show significant differences related to re-ulceration, re-amputation, or death. Thus, our results are partially in agreement with those of a previous study specifically on partial first ray amputation, which reported this type of surgery often progresses to a more proximal LEA and increases the risk of DFU. 19 In the context of this study, we identified factors such as depression and CAD are associated with more re-ulceration depending on the type of surgery. Patient with previous amputation was also associated with more re-amputation in the partial ray resection group which is consistent with previous study. 21 Moreover, depression was also highlighted as a predictor to LEA. 41 While not statistically significant, a partial ray first resection with CKD can lead to more re-ulceration compared to the hallux amputation (OR 4.15 vs 0.53). Our results are again consistent with a previous systematic review. 42 Therefore, these findings suggest a partial first ray resection should be avoided in patients with the following characteristics: CKD, depression or a history, and a previous amputation. It may support, to some extent when the presentation of the infection permits, the clinical decision to avoid this surgical procedure to reduce the likelihood of a poor (future) prognosis.
Overall, approximately $59\%$ of patients had a re-ulceration and $21\%$ had a re-amputation within one year in our cohort. In parallel with earlier literature reports which demonstrate approximately $60\%$ of patients will need further LEA and $46\%$ will have an DFU recurrence.13,43,44 However, the mortality rate of approximately $9\%$ was lower than the one reported in a recent systematic review (approximately $20\%$). 42 This positive finding can be justified by the diabetic foot management at our institution including a specialized service with a team approach to diabetic foot disease including podiatry. 45 This approach has been recognized to improve diabetic foot outcomes and enhance quality of care.46,47 *Although this* is a hypothesis, the lower mortality rate should be further explored, particularly as the data from this project did not allow for differentiation of major and minor LEA as outcomes. It is recognized that mortality and poor quality of life are higher in DM patients who undergo major LEAs. 3 This type of data would have been informative and represents a limitation.
There are also other limitations to this study. First, this is an observational study; therefore, there is no control group and some missing data (Table 1). Second, providers chose surgical intervention based on clinical appearance and radiographic findings. There was no structured algorithm to guide surgeons in their decision-making, and thus the dataset was dependent on standard of care as described by IWGDF. However, there were only five board-certified surgeons involved and reduced bias in decision-making and limited excessive heterogeneity. In fact, the design of the study is pragmatic in that it aims to answer a practical clinical question to support decision-making and potentially is helpful to guide therapy.
More specific continuous measurement variables, such as albumin and (absolute) toe pressures, were not available for comparison and a better understanding of the vascular and healing potential are essential. However, these are not routinely performed in inpatient assessment at our institution. In addition, analysis was complicated by missing data but also because of the low number of events at each time of follow-up. Additional information on these variables collected at uniform timelines could provide improved granularity into optimal procedure selection for a given patient. The statistical context limits the generalizability of the results. However, further prospective study in this area could also inform, in addition to health outcomes, about benefit, harms, adverse events and satisfaction or other patient-related outcomes to better support shared-clinical decisions (between patients and providers) in DFI context. This study highlighted future hypotheses exploration such as whether the complication rates of hallux amputations are worse first in a particular population (ie, with CAD or other comorbidities /risk factors), and thus whether these individuals should have a partial first ray amputation at the first place to achieve the best outcome.
To date, the decision to perform a partial first ray amputation or hallux amputation (disarticulation) was based on provider decision-making and not evidence-based medicine with respect to outcomes. Our cohort, although small ($$n = 70$$), demonstrates no significant difference in patient outcomes at one-year following surgical intervention. This included outcomes such as re-ulceration, re-amputation, and death. When faced with an infected ulcer (+ /-) osteomyelitis involving the first ray, if the infection can be eradicated through the removal of additional bone (partial first ray instead of hallux amputation), this decision is supported by evidence to be as safe as a hallux disarticulation without additional long-term sequelae of the operation from this study. However, consideration should be given when the patient outlined characteristics identified by this study. From an overall perspective, lower mortality at 1-year of our cohort supports the importance of team management of this health issue.
## Conclusion
This study highlights interesting data to inform clinical decisions to support best practices for the benefit of patients with respect to osteomyelitis in the first ray. Future research should guide surgeons in their decision-making to incorporate evidence-based medicine approaches to diabetic foot infections before intervention rather than to continue to operate blindly with respect to eventual clinical outcomes.
## References
1. Armstrong DG, Swerdlow MA, Armstrong AA, Conte MS, Padula WV, Bus SA. **Five year mortality and direct costs of care for people with diabetic foot complications are comparable to cancer**. *J Foot Ankle Res.* (2020.0) **13** 1-4. PMID: 31956341
2. Sothornwit J, Srisawasdi G, Suwannakin A, Sriwijitkamol A. **Decreased health-related quality of life in patients with diabetic foot problems**. *Diabetes, Metab Syndr Obes: Targets Ther.* (2018.0) **11** 35-43
3. Armstrong DG, Boulton AJ, Bus SA. **Diabetic foot ulcers and their recurrence**. *N Engl J Med.* (2017.0) **376** 2367-2375. PMID: 28614678
4. Soo BP, Rajbhandari S, Egun A, Ranasinghe U, Lahart IM, Pappachan JM. **Survival at 10 years following lower extremity amputations in patients with diabetic foot disease**. *Endocrine*
5. Hicks CW, Selvin E. **Epidemiology of peripheral neuropathy and lower extremity disease in diabetes**. *Curr Diab Rep.* (2019.0) **19** 1-8. PMID: 30637535
6. Prompers L, Schaper N, Apelqvist J. **Prediction of outcome in individuals with diabetic foot ulcers: Focus on the differences between individuals with and without peripheral arterial disease. The EURODIALE study**. *Diabetologia* (2008.0) **51** 747-755. PMID: 18297261
7. Prompers L, Huijberts M, Apelqvist J. **High prevalence of ischaemia, infection and serious comorbidity in patients with diabetic foot disease in Europe. Baseline results from the eurodiale study**. *Diabetologia* (2007.0) **50** 18-25. PMID: 17093942
8. Gazzaruso C, Gallotti P, Pujia A, Montalcini T, Giustina A, Coppola A. **Predictors of healing, ulcer recurrence and persistence, amputation and mortality in type 2 diabetic patients with diabetic foot: A 10-year retrospective cohort study**. *Endocrine* (2021.0) **71** 59-68. PMID: 32712853
9. Rastogi A, Goyal G, Kesavan R. **Long term outcomes after incident diabetic foot ulcer: Multicenter large cohort prospective study (EDI-FOCUS investigators) epidemiology of diabetic foot complications study: Epidemiology of diabetic foot complications study**. *Diabetes Res Clin Pract.* (2020.0) **162** 108113. PMID: 32165163
10. Lavery LA, Armstrong DG, Wunderlich RP, Mohler MJ, Wendel CS, Lipsky BA. **Risk factors for foot infections in individuals with diabetes**. *Diabetes Care* (2006.0) **29** 1288-1293. PMID: 16732010
11. Hamilton EJ, Davis WA, Siru R, Baba M, Norman PE, Davis TM. **Temporal trends in incident hospitalization for diabetes-related foot ulcer in type 2 diabetes: The fremantle diabetes study**. *Diabetes Care* (2021.0) **44** 722-730. PMID: 33441420
12. Wukich DK, Raspovic KM. **What role does function play in deciding on limb salvage versus amputation in patients with diabetes?**. *Plast Reconstr Surg.* (2016.0) **138** 188S-195S. PMID: 27556759
13. Borkosky SL, Roukis TS. **Incidence of re-amputation following partial first ray amputation associated with diabetes mellitus and peripheral sensory neuropathy: A systematic review**. *Diabet Foot Ankle.* (2012.0) **3** 12169
14. Kadukammakal J, Yau S, Urbas W. **Assessment of partial first-ray resections and their tendency to progress to transmetatarsal AmputationsA retrospective study**. *J Am Podiatr Med Assoc.* (2012.0) **102** 412-416. PMID: 23001735
15. Murdoch DP, Armstrong DG, Dacus JB, Laughlin TJ, Morgan CB, Lavery LA. **The natural history of great toe amputations**. *J Foot Ankle Surg.* (1997.0) **36** 204-208. PMID: 9232500
16. Rathnayake A, Saboo A, Malabu UH, Falhammar H. **Lower extremity amputations and long-term outcomes in diabetic foot ulcers: A systematic review**. *World J Diabetes.* (2020.0) **11** 391-399. PMID: 32994867
17. van Netten JJ, Bus SA, Apelqvist J. **Definitions and criteria for diabetic foot disease**. *Diabetes Metab Res Rev.* (2020.0) **36**
18. Cecilia-Matilla A, Lázaro-Martínez JL, Aragón-Sánchez J, García-Álvarez Y, Chana-Valero P, Beneit-Montesinos JV. **Influence of the location of nonischemic diabetic forefoot osteomyelitis on time to healing after undergoing surgery**. *Int J Low Extrem Wounds.* (2013.0) **12** 184-188. PMID: 24043680
19. Aprile I, Galli M, Pitocco D. **Does first ray amputation in diabetic patients influence gait and quality of life?**. *J Foot Ankle Surg.* (2018.0) **57** 44-51. PMID: 29268902
20. Mann RA, PoPPEN NK, O'Konski M. **Amputation of the great toe. A clinical and biomechanical study**. *Clin Orthop Relat Res.* (1988.0) **226** 192-205
21. Borkosky SL, Roukis TS. **Incidence of repeat amputation after partial first ray amputation associated with diabetes mellitus and peripheral neuropathy: An 11-year review**. *J Foot Ankle Surg.* (2013.0) **52** 335-338. PMID: 23540756
22. El-Hilaly R, Elshazly O, Amer A. **The role of a total contact insole in diminishing foot pressures following partial first ray amputation in diabetic patients**. *The Foot* (2013.0) **23** 6-10. PMID: 23266131
23. Hingorani A, LaMuraglia GM, Henke P. **The management of diabetic foot: A clinical practice guideline by the society for vascular surgery in collaboration with the American podiatric medical association and the society for vascular medicine**. *J Vasc Surg.* (2016.0) **63** 3S-21S. PMID: 26804367
24. **Practical guidelines on the prevention and management of diabetic foot disease (IWGDF 2019 update)**. *Diab Metab Res Rev* (2020.0)
25. Lipsky BA, Berendt AR, Cornia PB. **2012 Infectious diseases society of America clinical practice guideline for the diagnosis and treatment of diabetic foot infections**. *Clin Infect Dis.* (2012.0) **54**
26. Weledji EP, Fokam P. **Treatment of the diabetic foot–to amputate or not?**. *BMC Surg.* (2014.0) **14** 1-6. PMID: 24401085
27. Thorud JC, Jupiter DC, Lorenzana J, Nguyen TT, Shibuya N. **Reoperation and reamputation after transmetatarsal amputation: A systematic review and meta-analysis**. *J Foot Ankle Surg.* (2016.0) **55** 1007-1012. PMID: 27475711
28. Skoutas D, Papanas N, Georgiadis G. **Risk factors for ipsilateral reamputation in patients with diabetic foot lesions**. *Int J Low Extrem Wounds.* (2009.0) **8** 69-74. PMID: 19443895
29. Schmidt BM, Wrobel JS, Munson M, Rothenberg G, Holmes CM. **Podiatry impact on high-low amputation ratio characteristics: A 16-year retrospective study**. *Diabetes Research & Clinical Practice* (2017.0) **126** 272-277. DOI: 10.1016/j.diabres.2017.02.008.
30. Charan J, Biswas T. **How to calculate sample size for different study designs in medical research?**. *Indian J Psychol Med.* (2013.0) **35** 121-126. DOI: 10.4103/0253-7176.116232.
31. Chu YJ, Li XW, Wang PH. **Clinical outcomes of toe amputation in patients with type 2 diabetes in tianjin, China**. *Int Wound J.* (2016.0) **13** 175-181. PMID: 24629051
32. Schmidt BM, McHugh JB, Patel RM, Wrobel JS. **Prospective analysis of surgical bone margins after partial foot amputation in diabetic patients admitted with moderate to severe foot infections**. *Foot Ankle Spec.* (2019.0) **12** 131-137. PMID: 29644884
33. Agha RA, Borrelli MR, Vella-Baldacchino M. **The STROCSS statement: Strengthening the reporting of cohort studies in surgery**. *Int J Surg.* (2017.0) **46** 198-202. PMID: 28890409
34. Sr JL M, Conte MS, Armstrong DG. **The society for vascular surgery lower extremity threatened limb classification system: Risk stratification based on wound, ischemia, and foot infection (WIfI)**. *J Vasc Surg.* (2014.0) **59** 220-234.e2. PMID: 24126108
35. Lavery LA, Armstrong DG, Harkless LB. **Classification of diabetic foot wounds**. *J Foot Ankle Surg.* (1996.0) **35** 528-531. PMID: 8986890
36. Chuan F, Zhang M, Yao Y, Tian W, He X, Zhou B. **Anemia in patients with diabetic foot ulcer: Prevalence, clinical characteristics, and outcome**. *Int J Low Extrem Wounds.* (2016.0) **15** 220-226. PMID: 27440798
37. Yesil S, Akinci B, Yener S. **Predictors of amputation in diabetics with foot ulcer: Single center experience in a large turkish cohort**. *Hormones* (2009.0) **8** 286-295. PMID: 20045802
38. Tan T-W, Calhoun EA, Knapp SM. **Rates of diabetes-related Major amputations among racial and ethnic minority adults following medicaid expansion under the patient protection and affordable care act**. *JAMA Network Open* (2022.0) **5**
39. Tan T-W, Shih C-D, Concha-Moore KC. **Disparities in outcomes of patients admitted with diabetic foot infections**. *PLoS One* (2019.0) **14**
40. Narres M, Kvitkina T, Claessen H. **Incidence of lower extremity amputations in the diabetic compared with the non-diabetic population: A systematic review**. *PloS one* (2017.0) **12**
41. O’Neill SM, Kabir Z, McNamara G, Buckley CM. **Comorbid depression and risk of lower extremity amputation in people with diabetes: Systematic review and meta-analysis**. *BMJ Open Diabetes Research and Care* (2017.0) **5**
42. Yammine K, Assi C. **A meta-analysis of the types and outcomes of conservative excisional surgery for recalcitrant or infected diabetic toe ulcers**. *Foot Ankle Spec.* (2020.0) **13** 152-160. PMID: 31216881
43. Vassallo IM, Gatt A, Cassar K, Papanas N, Formosa C. **Healing and mortality rates following toe amputation in type 2 diabetes mellitus**. *Exp Clin Endocrinol Diabetes.* (2021.0) **129** 438-442. PMID: 31207664
44. Collins PM, Joyce DP, O’Beirn ES. **Re-amputation and survival following toe amputation: Outcome data from a tertiary referral centre**. *Irish Journal of Medical Science (1971-)*
45. 45Schmidt BM, Holmes CM, Ye W, Pop-Busui R. A tale of two eras: Mining big data from electronic health records to determine limb salvage rates with podiatry. Curr Diabetes Rev. 2019;15(6):497–502.
46. Blanchette V, Brousseau-Foley M, Cloutier L. **Effect of contact with podiatry in a team approach context on diabetic foot ulcer and lower extremity amputation: Systematic review and meta-analysis**. *J Foot Ankle Res.* (2020.0) **13** 1-12. PMID: 31956341
47. Monteiro-Soares M, Vale-Lima J, Martiniano J, Pinheiro-Torres S, Dias V, Boyko EJ. **A systematic review with meta-analysis of the impact of access and quality of diabetic foot care delivery in preventing lower extremity amputation**. *J Diabetes Complicat.*
|
---
title: SIRT3 regulates mitochondrial biogenesis in aging-related diseases
authors:
- Hongyan Li
- Zhiyou Cai
journal: Journal of Biomedical Research
year: 2023
pmcid: PMC10018414
doi: 10.7555/JBR.36.20220078
license: CC BY 4.0
---
# SIRT3 regulates mitochondrial biogenesis in aging-related diseases
## Abstract
Sirtuin 3 (SIRT3), the main family member of mitochondrial deacetylase, targets the majority of substrates controlling mitochondrial biogenesis via lysine deacetylation and modulates important cellular functions such as energy metabolism, reactive oxygen species production and clearance, oxidative stress, and aging. Deletion of SIRT3 has a deleterious effect on mitochondrial biogenesis, thus leading to the defect in mitochondrial function and insufficient ATP production. Imbalance of mitochondrial dynamics leads to excessive mitochondrial biogenesis, dampening mitochondrial function. Mitochondrial dysfunction plays an important role in several diseases related to aging, such as cardiovascular disease, cancer and neurodegenerative diseases. Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) launches mitochondrial biogenesis through activating nuclear respiratory factors. These factors act on genes, transcribing and translating mitochondrial DNA to generate new mitochondria. PGC1α builds a bridge between SIRT3 and mitochondrial biogenesis. This review described the involvement of SIRT3 and mitochondrial dynamics, particularly mitochondrial biogenesis in aging-related diseases, and further illustrated the role of the signaling events between SIRT3 and mitochondrial biogenesis in the pathological process of aging-related diseases.
Mitochondrial biogenesis is a complex biological process, requiring cooperation with nuclear and mtDNA to generate enormous mitochondrial enzymes and proteins. Besides, mitochondrial biogenesis is regulated by plenty of external factors and internal molecules. CR: caloric restriction; PGC1α: peroxisome proliferator-activated receptor gamma coactivator 1-alpha; TFAM: mitochondrial transcription factor A; mtDNA: mitochondrial DNA; AMPK: adenosine 5′-monophosphate (AMP)-activated protein kinase; NO: nitric oxide; SIRT1: sirtuin 1; SIRT3: sirtuin 3; ERRα: estrogen-related receptor alpha; CAMP: cyclic adenosine monophosphate; NRF$\frac{1}{2}$: nuclear respiratory factor $\frac{1}{2.}$
PGC1α/SIRT3 axis exerts a neuroprotective role in neurodegenerative diseases through maintaining normal mitochondrial function, ROS balance, and energy production. Impaired mitochondrial fusion/fission cycle is also a key component in the pathogenesis of neurodegenerative diseases. PGC1α: peroxisome proliferator-activated receptor gamma coactivator 1-alpha; SIRT3: sirtuin 3; mtDNAcn: mtDNA copy number; ROS: reactive oxygen species; SOD2: superoxide dismutase 2; UCP-2: uncoupling protein 2.
A: Under external stress, cardiovascular system is sheltered by mitochondrial biogenesis regulated by SIRT3 and mitochondrial fusion controlled by PGC1α. B: Hyperglycemia disrupted PGC1α/SIRT3 axis, along with OXPHOS and a ROS defense system decline. Besides, abnormal mitochondrial dynamics are involved in the progression of the two diseases. PGC1α: peroxisome proliferator-activated receptor gamma coactivator 1-alpha; SIRT1: sirtuin 1; SIRT3: sirtuin 3; mtDNAcn: mtDNA copy number; ROS: reactive oxygen species; IRS: insulin receptor substrate; Mfn2: mitofusin 2; OXPHOS: oxidative phosphorylation; EC: endothelial cells.
## Introduction
A research based on population showed that the aging population is gradually increasing throughout the world[1–2]. However, there is a general physical and functioning decline with aging. Aging is an independent risk factor for diseases. Most common aging-related diseases include Parkinson's disease (PD), Alzheimer's disease (AD), type 2 diabetes mellitus (T2DM), cardiovascular disease, and cancer[2–3].
Mitochondria, a major source of energy, are important organelles to fuel the human body's functions. Thereby, mitochondrial dysfunction is viewed as a potential regulator of the aging process[4]. Although plenty of factors, including abnormal mitochondrial quality control and aging[4], contribute to mitochondrial dysfunction, our attention is paid to mitochondrial biogenesis, yielding new mitochondria in response to various stress. Besides, mitochondrial dynamics will be discussed briefly.
Mitochondrial function is modulated by a series of enzymes, the majority of which are also deacetylated by sirtuin 3 (SIRT3). SIRT3 targets many substrates controlling mitochondrial biogenesis through lysine deacetylation and modulates important cellular functions such as mitochondrial sugar, fat, and amino acid metabolism as well as reactive oxygen species (ROS) production and clearance[5].
This review summarizes recent research on mitochondrial biogenesis and SIRT3, further illustrating the effects of SIRT3 on mitochondrial biogenesis in aging-related diseases.
## SIRT3
Sirtuins are a family of NAD+‐dependent histone deacetylases/mono-ADP-ribosyl transferase enzymes (SIRT1–SIRT7) and highly conserved in both bacteria and humans during the evolution[6–7]. Sirtuins have been reported to regulate a wide variety of biological processes such as metabolism, mitochondria homeostasis, genomic stability, DNA repair, ROS homeostasis, and aging[8–9]. SIRT3 is primarily localized in the mitochondria and functions as a major mitochondrial deacetylase that targets a growing number of substrates involved in metabolic homeostasis, mitochondrial dynamics, mitochondrial unfolded protein response, and oxidative stress[8]. In addition, SIRT3 protein is highly expressed in the brain, heart, liver, and brown adipose tissue[10]. SIRT3 controls global mitochondrial lysine acetylation level, and the sum of acetyl modifications are critical for these biological processes[10]. For instance, SIRT3 deacetylates numerous mitochondrial metabolic enzymes such as isocitrate dehydrogenase and malate dehydrogenase in the tricarboxylic acid cycle[11], complex Ⅰ, complex Ⅱ, and complex Ⅲ in the electron transport chain[12], and long-chain acyl-CoA dehydrogenase involved in the fatty acid oxidation pathway[13]. SIRT3 regulates these biosynthetic pathways involved in glucose and lipid metabolism and provides adequate energy to cells or tissues. SIRT3 deacetylates enzymes related to cellular anti-oxidative defense capacity, like superoxide dismutase 2 (SOD2) associated with ROS homeostasis[14], to prevent excessive ROS accumulation. SIRT3 can also stimulate the peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) controlling mitochondrial biogenesis via adenosine 5′-monophosphate (AMP)-activated protein kinase (AMPK) pathway to maintain mitochondrial quality and quantity[15]. Accordingly, SIRT3 is essential for sustaining anti-oxidative system, mitochondrial integrity, and normal mitochondrial function as the basis of physiological processes.
It is well known that SIRT3 is strongly involved in the aging process. *Subsequent* genetic studies found the relationship between single nucleotide polymorphisms (SNPs) of the SIRT3 gene and longevity[16–17]. Three types of SNPs occurring in SIRT3 (G477T, variable number tandem repeat and V208I) have been discovered in human and SIRT3-knockout mice, and they can modify human health and lifespan[16,18–19].
Upregulated SIRT3 increases energy generation to meet high energy demand during calorie restriction (CR), fasting, and exercise[17]. Accumulating evidence confirms that SIRT3 attenuation or ablation accelerates the development of aging-related diseases, including cancers, metabolic syndromes, cardiovascular diseases, and neurodegenerative diseases[12,20–21]. For example, SIRT3 protects dopaminergic neurons from degeneration and necrosis by regulating mitochondrial quality control, reducing mitochondrial oxidative stress, and downregulating α-synuclein (α-syn) in PD[21–24]. FOXO3 deacetylated by SIRT3 can maintain a balance between mitochondrial fission/fusion by inducing a series of FOXO3-dependent genes expression[25]. This process is beneficial in delaying the progression of PD. In response to stress situations, Ku70 is deacetylated by SIRT3, then binds to Bax, which protects cardiomyocyte against apoptosis induced by Bax[26]. SIRT3 was indicated to inhibit p53 activity, rescuing cells from apoptosis[27]. In PTEN-deficient cells, p53 at lysines 320 and 382 is modified by SIRT3, deteriorating the condition of non-small cell lung cancer[28]. Conversely, overexpression of SIRT3 in human glioma cells can promote tumor progression through Ku70-BAX interaction[29].
## Mitochondrial biogenesis
Mitochondrial biogenesis is a complex process of mitochondrial growth and division. It requires the synthesis of proteins encoded by nuclear and mitochondrial (mt) DNA, besides the biogenesis of new organellar structures[30–31]. The nuclear genome encodes enormous mitochondrial enzymes and proteins, thus regulating transcriptional networks[32]. The mitochondrial DNA (mtDNA) is responsible for essential components of the electron transport chain as well as all rRNAs and tRNAs[33–34]. Biological processes, including transcription and replication from both nuclear and mtDNA, must be coordinated with each other to generate new mitochondrion in response to high-demand for energy[35] (Fig. 1).
**Figure 1:** *Regulators of mitochondrial biogenesis.*
It is well known that PGC1α is a primary regulator of mitochondrial biogenesis through activating nuclear respiratory factors (NRFs) and nuclear receptor subfamily (e.g., estrogen-related receptor alpha [ERRα]), expressing in tissues and organs with high-energy-requirements[35–36]. Because they regulate the expression of mitochondrial respiratory subunits and mitochondrial transcription factors including mitochondrial transcription factor A (TFAM) and mitochondrial transcription factor B (mtTFB) isoform genes[37]. *These* genes drive the transcription and replication of mtDNA. PGC1α drives multiple transcription factors (NRFs, ERRα, etc.), thus creating new mitochondria in response to various stress conditions[38].
The regulation of mitochondrial biogenesis is affected by multiple internal and external factors through PGC1α[39]. Because PGC1α, one member of the transcriptional coactivators family, controls various aspects of mitochondrial biogenesis, including initiation of respiratory chain and fatty acid oxidation genes, an increase of mitochondrial number, and augmentation of mitochondrial respiratory capacity[33]. At the level of molecule, factors involved in activation of PGC1α include nitric oxide, AMPK, CREB, SIRT1, NRF-1and NRF-2[39–40]. However, RIP140 and p160 myb binding protein take part in the inhibition of PGC1α and thus suppress mitochondrial biogenesis[39]. At the level of the organism, hormones also influence mitochondrial biogenesis. Under oxidative stress, the thyroid hormone triiodothyronine can drive mitochondrial biogenesis and respiration process in various tissues of mouse[41–43], and steroid hormones stimulate production of mitochondrial proteome[39]. Hormones are capable of controlling mitochondrial biogenesis by affecting PPARGC1A level[38]. Besides, physiological environment changes are also reported to modulate mitochondrial biogenesis such as exercise, CR, hypoxia, stress, and temperature[40], and such observations were shown in rats and humans[44–45].
In addition to mitochondrial biogenesis, mitochondrial fission (mito-fission) is a process in which mitochondria are broken up into smaller fragments, thereby producing two daughter mitochondria to increase mitochondrial mass during cell self-renewal[46]. During mito-fission, the damaged mitochondrion is removed by mitophagy, whereas the healthy one steps into the next phase, which is mitochondrial biogenesis[47]. Mito-fission is launched by endoplasmic reticulum (ER). Because once ER contacts with mitochondria, the sites at which dynamin-related protein 1 (DRP1) is recruited and assembles will be established[48]. DRP1, cutting the inner and outer membranes, also recruits some proteins aiding in splitting into two segments. Those proteins consist of mitochondrial receptor protein 1 (Fis1), mitochondrial fission factor (Mff), and mitochondrial dynamic proteins (MIDs)[47,49–50]. When cells are exposed to stresses, mito-fission bears the responsibility to maintain mitochondrial network functioning properly and promotes mitochondrial trafficking through segregating defective mitochondria and preserving the normal one[50].
Facilitating mitochondrial biogenesis would be advantageous in many disease models, but once the balance of mitochondrial dynamics (mito-fission, mitophagy, and mitochondrial biogenesis) is destroyed, excessive mitochondrial biogenesis is detrimental to cells for extreme oxygen consumption[38].
## Dysregulation of mitochondrial biogenesis in aging-related diseases
Aging brings a general decline in physiological functions. This process is featured by mitochondrial decay and decrease of oxidative phosphorylation (OXPHOS) capacity, along with changes in mitochondrial morphology and mitochondrial content (number and protein levels)[51]. Defective mitochondria, a hallmark of cellular aging, include malfunction of mitochondrial biogenesis, abnormal mitochondrial dynamics and trafficking, aberrant autophagy function, and transcriptional dysregulation[52–53]. The ability of mitochondrial biogenesis is in slow and progressive decline with age. Hence, mitochondrial biogenesis has been viewed as a target for delaying aging and extending lifespan[35,53–54]. Mitochondrial biogenesis, a self-renewal route, aims to generate new mitochondria from the existing one, in order to meet energy requirements and maintain the dynamic circulation of mitochondria through collaborating with mitochondrial autophagy[55]. Therefore, mitochondrial biogenesis plays an important role in maintaining homeostasis of the mitochondrial mass and function, malfunction of which is associated with aging, neurodegenerative diseases, metabolic diseases, and cancers[56].
Tremendous efforts have been made to discover the role of mitochondrial biogenesis-related proteins and genes under pathological condition. Accumulative literature has reported that mtDNA deletion, mutation and damage result in mitochondrial dysfunction among different tissues or within the same tissue, eventually responsible for aging and age-related neurodegenerative diseases[57–58]. The accumulation of mtDNA mutations triggers neuronal loss in the substantia nigra of patients with PD or rat PD models[59–60], and negatively affects neuronal mitochondrial energy and synapse in the frontal cortex and hippocampus of AD patients[61]. Substantial evidence similarly shows that mtDNA damage is closely related to Huntington's disease (HD)[53,62–63].
In another way, enhancement of mtDNA repair alleviates lung endothelial barrier dysfunction induced by donation after circulatory death related ischemia-reperfusion injury[64]. mtDNA copy number (mtDNAcn), regulated by transcriptional and translational factors, represents the mitochondrial abundance within a cell and varies with cellular energy requirement[65]. Alternations in mtDNAcn are associated with both increased and decreased disease burdens[51]. Some researchers hold the view that the upregulation of mtDNAcn due to overexpression of TFAM prolongs lifespan in the mouse suffering from mitochondrial diseases or myocardial infarction[66–67]. On the contrary, increased mtDNAcn negatively impacts the replication and transcription of mitochondrial proteins, accompanied by nucleoid enlargement[68]. These changes harm mitochondrial functions.
In brief, damaged mitochondrial biogenesis contributes to an accumulation of old or dysfunctional mitochondria and the progression of various diseases.
## Interaction between SIRT3 and mitochondrial biogenesis in aging-related diseases
At the subcellular level, SIRT3 mainly localizes in mitochondria and deacetylates many mitochondrial metabolic proteins, and such observation was identified by mass spectrometry[69]. Likewise, mitochondrial metabolism such as fatty acid metabolism, glycolysis, and the tricarboxylic acid cycle is rich in acetylated proteins. SIRT3 plays a major role in sustaining mitochondrial bioenergetics. Thus, the pathway of SIRT3 and mitochondrial biogenesis is activated to maintain normal mitochondrial function and protect cell from death under pathological condition. In this part, we reviewed mitochondrial biogenesis with SIRT3 in three major aging-related diseases: cardiovascular diseases, neurodegenerative diseases, and T2DM. Additionally, the role of mitochondrial dynamics in the course of these diseases was summarized (Fig. 2 and Fig. 3).
**Figure 2:** *Interaction between SIRT3 and PGC1α in neurodegenerative diseases.* **Figure 3:** *Interaction between SIRT3 and PGC1α in cardiovascular diseases and type 2 diabetes mellitus.*
## Neurodegenerative diseases
Neurodegenerative diseases include AD, PD, HD, and amyotrophic lateral sclerosis (ALS). Featured by high-energy demand, neurons are vulnerable to oxidative stressors. Thus, the decrease of ATP level will result in neuronal death in patients with neurodegenerative diseases[70]. Mitochondrial malfunction and oxidative damage are pivotal contributors to the development of neurodegenerative diseases[71–73].
The mtDNA or nuclear DNA mutations are reported to cause diseases related to neuronal degeneration[71]. The mtDNA replication machinery is more error-prone than that of genomic DNA, because the mtDNA polymerase lacks the function of proofreading[34,74]. Besides, DNA repair system is less reliable. The mtDNA damage can impair respiratory chain protein synthesis, reduce the efficiency of energy production and overproduce ROS. An increase in ROS impairs mtDNA in the same manner and misfolds proteins, vice versus[75]. These processes are related to mitochondrial dysfunction, eventually accelerating the development of diseases such as HD, AD, ALS, and hypertension[72,74]. However, SIRT3 possesses the ability to repair mtDNA damage, preserve mitochondrial function and protect apoptotic cell against death through deacetylating a variety of protein targets. These proteins consist of endonuclease Ⅷ-like 1, endonuclease Ⅷ-like 1 (NEIL2), 8-oxoguanine-DNA glycosylase 1 (OGG1)[76], mutY DNA glycosylase (MUTYH), apurinic/apyrimidinic endonuclease (APE1), and DNA ligase 3 (LIG3)[77], which modulate the activity of mtDNA base excision repair (BER), consequently taking charge of removing damaged bases. A large body of findings support that mtDNA repair machinery is impaired in neurodegenerative disorders[78–80]. In Caenorhabditis elegans PD models, incomplete BER results in genomic stress, promoting neuronal loss and further driving pathological process of age-related neurodegeneration[81]. Evidence supporting a critical role for DNA repair deficiencies in AD demonstrated that compromised DNA repair is a driving force of neuronal dysfunction and loss, because of damaged mitophagy, metabolic disturbance, and energy deprivation[82]. Research in ALS patients confirmed the perspective that altered mtDNA (mtDNA mutations and deletions) was observed in spinal neurons of ALS, driving synaptic dysfunction and facilitating motor neuron degeneration[83]. Altogether, studies concerning mtDNA alternations or ineffective BER demonstrate they have tight connection with neurodegenerative diseases.
SIRT3 and PGC-1α play a neuroprotective role in ALS model through preventing mitochondrial fragmentation and neuronal cell apoptosis[84]. SIRT3 is beneficial to maintaining ROS balance between production and clearance, overexpression of which suppresses ROS accumulation. PGC1α, strongly detected in tissues with high requirement for energy, not only induces antioxidant defense gene expression such as SOD2 and uncoupling protein 2 to stop oxidative damage and mitochondrial destruction[85], but also modulates SIRT3 expression to achieve ROS homeostasis. In mutant SOD1 mice, motor neurons were found to develop degeneration due to aberrated mitochondria aggregation in neuronal axons and dendrites[72]. Knockout of SIRT3 suppresses antioxidant gene expression, leading to mitochondrial disequilibrium of antioxidant defense controlled by PGC1α[84]. Additionally, research in neuro-2a cell and SIRT3-knockout mice demonstrated that loss of SIRT3 had a deleterious effect on SOD2 and ATP synthase β acetylation levels targeting functional sites (SOD2-K130 and ATP synthase b-K485), critical for the regulation of ROS and ATP levels[86]. ROS aggregation and reduction in ATP are responsible for neuronal death in PD mice. Another insight is that PGC1a is a regulator of SIRT3 through interacting with ERRα to protect against dopaminergic neuronal death, a hallmark of PD[86].
The imbalance of mitochondrial dynamics is a key component in the pathogenesis of neurodegenerative diseases such as PD and AD[87–88]. In PD, α-syn aggregation in the neuron is capable to perturb the cycle of mitochondrial fusion and fission because of its deleterious effect on mitochondrial associated proteins including mitofusin 1 and 2 (Mfn$\frac{1}{2}$), DRP1, and MFF, correlative with an increase in aberrant mitochondria and inefficiency of neuronal signaling[89–90]. Mitophagy, functioning as degrading damaged mitochondria, is also under the adverse influence of α-syn in PD[88]. Interestingly, SIRT3 depends on optic atrophy 1 (OPA1), a mitochondrial fusion regulator, to maintain normal mitochondrial dynamics, ultimately preventing cell apoptosis via the effect on cytochrome c translocation and mitochondrial respiratory efficacy[91]. Recently, research in the HD model reveals that upregulated SIRT3 contributes to mitochondrial fusion, rather than fission, leading to remodeled mitochondrial function, dynamics, and distribution in neural cells, further exerting neuroprotective effect. Decline in DRP1 and Fis1 levels caused by SIRT3 suppresses mitochondrial fission and biogenesis, and it means that mitochondrial mass isn't affected in HD[92]. Overall, SIRT3 has certain impacts on neurodegenerative diseases through the modulation of mitochondrial dynamics.
## Cardiovascular diseases
A small number of mitochondria exist in cardiomyocytes, one of the highest energy consuming cell types[93–94]. Not surprisingly, mitochondria in the function of vascular endothelial cells (VECs) as a metabolic signaling regulator triggering cell proliferation or apoptosis instead of energy supply[93]. Aberrant mitochondrial function is a contributing factor to endothelial dysfunction, responsible for cardiovascular diseases such as cardiac hypertrophy and atherosclerosis.
SIRT3 plays a protective role in myocardial cells via different signaling pathways to promote mitochondrial biogenesis. SIRT3 protects cardiomyocytes from ischemia reperfusion injury by preventing mitochondria mediated apoptosis, and this effect has been related to the activation of the AMPK pathway[15]. One explanation is that AMPK, an energy sensor of cells, drives mitochondrial biogenesis machinery in response to energy demand[40]. This viewpoint was supported by research in H9c2 cell, overexpression of SIRT3 induced corresponding changes in expression of mtDNA encoded genes, SOD2 expression and activity through the AMPKα-PGC1α axis, contributing to mitochondrial biogenesis and protecting myocardial cells[95]. Another function of PGC1α in VECs is to modulate several antioxidant enzymes, strengthening ROS defenses[96]. Furthermore, PGC1α controls some genes expression related to fatty acid oxidation, the tricarboxylic acid cycle, electron transport chain, and oxidative phosphorylation[97], promoting vascular endothelial growth factor (VEGF) expression[92] and protects against apoptosis via the effect of 15-hydroxyeicosatetraenoic acid[98]. This process mentioned above accelerates VEC proliferation and activates vessel sprouting in response to environmental stimulus (e.g., CR and hypoxia)[93].
Respiratory chain generates amounts of cellular energy, along with free radicals and ROS as by-products. Minimal ROS are conducive to signal transduction in physiological surroundings, whereas elevated ROS damages the adjacent cell structures, and alters DNA, proteins, and other molecules, further causing cell death[34]. It's well documented that excessive mitochondrial ROS give rise to VEC dysfunction, leading to formation of atherogenesis and cardiac hypertrophy[34,74]. Mitochondria-targeted esculetin-induced SIRT3, not SIRT1 overexpression, a new lipoxygenase inhibitor, protects endothelial cells from death through AMPK-mediated nitric oxide pathway, thus attenuating plaque formation[99]. Another perspective is that SIRT1 and SIRT3 cooperate to drive the antiaging effects under CR condition[100]. Increased expression of SIRT1 deacetylates and activates PGC1α in response to nutrient limitation, which further reduces ROS production, and promotes antioxidant environment through coactivating transcription of SIRT3[100–101]. This relationship is beneficial for longevity because of elevated mitochondrial biogenesis and ROS detoxification.
Therefore, mitochondrial dynamics play a critical role in VECs. Fusion proteins such as mfn1, mfn2, and OPA1 are associated with VEGF-mediated angiogenesis and angiogenic function[102–103]. The Mfn2 unregulated by PGC1α and PGC1β can promote mitochondrial fusion[104]. After knocking out the mfn2 gene in the heart of the mouse, calcium (Ca2+) fails to transfer from sarcoplasmic reticulum to mitochondria, followed by disrupted Ca2+ signaling, and diminished cardiac contractility function[105]. Evidence indicates that lacking cardiac-specific mfn2 undermines cellular autophagy and impairs mitochondrial network function, ultimately leading to aberrant left ventricular function[106].
## Type 2 diabetes mellitus
T2DM is featured by deficient mitochondrial function and ROS. SIRT3 deacetylates and modifies several mitochondrial protease activities to manage mitochondrial functions and maintain redox homeostasis. The reduction of SIRT3 function contributes to the development of insulin resistance (IR), and hallmark of the pathogenesis of T2DM[107–108]. Evidence from experiment in cultured human endothelial cells demonstrated that SIRT3 deficiency is implicated in endothelial IR because of a drop in phosphorylation of Akt and endothelial nitric oxide synthase[109]. Similarly, overexpression of SIRT3 ameliorates negative effect of pancreatic β-cell involving malfunction and apoptosis induced by palmitate[110]. β-cell dysfunction is related to dysregulation of insulin synthesis and insulin deficiency, further leading to hyperglycemia[111–112]. Hyperglycemia in turn promotes mitochondria fragmentation, and increases ROS production[113].
In pre-diabetic model, the researcher found that PGC1α/SIRT3 axis of testis was disrupted, accompanied by mtDNAcn decline[114]. PGC1α and Sirt3 are implicated in various biological processes such as mitochondrial biogenesis, functional OXPHOS, and an active ROS defense system[101]. The impaired PGC1α/SIRT3 axis compromises respiratory capacity, and promotes oxidative stress. Additionally, mitochondrial dynamics is not only the link between impaired mitochondrial function and IR, but also implicated in the development of T2D[115–117]. Diabetes susceptible cybrid cell model has demonstrated that IR is a consequence of abnormal mitochondrial dynamics, because upregulating Mfn1/*Mfn2* genes and depressing DRP1/Fis1 can remodel mitochondrial network, repairing the IR signaling[118]. According to some viewpoints, both insulin signaling and insulin sensitivity are manipulated by Mfn2. In Mfn2 KO mouse, glucose tolerance and IRS-Akt pathway were impaired, whereas hepatic glucose production was enhanced[115]. Reduced Mfn2 expression was also found in T2D patients' muscles.
## Conclusions and perspectives
This review summarizes the involvement of SIRT3 and mitochondrial biogenesis in aging-related diseases. Based on SIRT3 regulation of mitochondrial function, we reviewed that there are intersecting signaling pathways between SIRT3 and mitochondrial biogenesis such as AMPK-PGC1α axis, SIRT1-PGC1α axis, and PGC1α/SIRT3 axis in the development of aging-related diseases including neurodegenerative diseases[19], cardiovascular diseases[12] and T2DM[114]. Impaired mitochondrial dynamics such as mitochondrial fusion and fission participate in the process of aging-related disease progression as well. Notably, manipulation of SIRT3, mitochondrial biogenesis, and mitochondrial dynamics offers novel therapeutic options for these aging-related diseases. However, other aspects of mitochondrial dysfunction such as abnormal mitochondrial quality control, mitochondrial homeostasis imbalance, and mitophagy dysfunction also account for aging-related disease progression and development. Thus, future investigations on SIRT3-mediated mitochondrial function may aid in providing new pathways during the aging process.
## References
1. **Epidemiology of aging**. *Radiol Clin North Am* (2008) **46** 643-652. DOI: 10.1016/j.rcl.2008.07.005
2. **Age-related diseases and clinical and public health implications for the 85 years old and over population**. *Front Public Health* (2017) **5** 335. DOI: 10.3389/fpubh.2017.00335
3. **Aging and age-related diseases: from mechanisms to therapeutic strategies**. *Biogerontology* (2021) **22** 165-187. DOI: 10.1007/s10522-021-09910-5
4. **The role of mitochondria in aging**. *J Clin Invest* (2018) **128** 3662-3670. DOI: 10.1172/JCI120842
5. **SIRT3 expression decreases with reactive oxygen species generation in rat cortical neurons during early brain injury induced by experimental subarachnoid hemorrhage**. *Biomed Res Int* (2016) **2016** 8263926. DOI: 10.1155/2016/8263926
6. **Characterization of five human cDNAs with homology to the yeast SIR2 gene: sir2-like proteins (sirtuins) metabolize NAD and may have protein ADP-ribosyltransferase activity**. *Biochem Biophys Res Commun* (1999) **260** 273-279. DOI: 10.1006/bbrc.1999.0897
7. **Sirtuins and FoxOs in osteoporosis and osteoarthritis**. *Bone* (2019) **121** 284-292. DOI: 10.1016/j.bone.2019.01.018
8. **Sirtuins in cancer: a balancing act between genome stability and metabolism**. *Mol Cells* (2015) **38** 750-758. DOI: 10.14348/molcells.2015.0167
9. **SIRT3 regulation of mitochondrial oxidative stress**. *Exp Gerontol* (2013) **48** 634-639. DOI: 10.1016/j.exger.2012.08.007
10. **Mammalian Sir2 homolog SIRT3 regulates global mitochondrial lysine acetylation**. *Mol Cell Biol* (2007) **27** 8807-8814. DOI: 10.1128/MCB.01636-07
11. **The mTOR/PGC-1α/SIRT3 pathway drives reductive glutamine metabolism to reduce oxidative stress caused by ISKNV in CPB cells**. *Microbiol Spectr* (2022) **10** e0231021. DOI: 10.1128/spectrum.02310-21
12. **Emerging role of SIRT3 in mitochondrial dysfunction and cardiovascular diseases**. *Free Radic Res* (2019) **53** 139-149. DOI: 10.1080/10715762.2018.1549732
13. **Mouse SIRT3 attenuates hypertrophy-related lipid accumulation in the heart through the deacetylation of LCAD**. *PLoS One* (2015) **10** e0118909. DOI: 10.1371/journal.pone.0118909
14. **SIRT3 protects hepatocytes from oxidative injury by enhancing ROS scavenging and mitochondrial integrity**. *Cell Death Dis* (2017) **8** e3158. DOI: 10.1038/cddis.2017.564
15. **SirT3 activates AMPK-related mitochondrial biogenesis and ameliorates sepsis-induced myocardial injury**. *Aging* (2020) **12** 16224-16237. DOI: 10.18632/aging.103644
16. **A novel VNTR enhancer within the**. *Genomics* (2005) **85** 258-263. DOI: 10.1016/j.ygeno.2004.11.003
17. **Forever young: SIRT3 a shield against mitochondrial meltdown, aging, and neurodegeneration**. *Front Aging Neurosci* (2013) **5** 48. DOI: 10.3389/fnagi.2013.00048
18. **Variability of the**. *Exp Gerontol* (2003) **38** 1065-1070. DOI: 10.1016/S0531-5565(03)00209-2
19. **SIRT3 deficiency and mitochondrial protein hyperacetylation accelerate the development of the metabolic syndrome**. *Mol Cell* (2011) **44** 177-190. DOI: 10.1016/j.molcel.2011.07.019
20. **SIRT3 regulates progression and development of diseases of aging**. *Trends Endocrinol Metab* (2015) **26** 486-492. DOI: 10.1016/j.tem.2015.06.001
21. **SIRT3 regulation of mitochondrial quality control in neurodegenerative diseases**. *Front Aging Neurosci* (2019) **11** 313. DOI: 10.3389/fnagi.2019.00313
22. **Sirtuin 3 rescues neurons through the stabilisation of mitochondrial biogenetics in the virally-expressing mutant α-synuclein rat model of parkinsonism**. *Neurobiol Dis* (2017) **106** 133-146. DOI: 10.1016/j.nbd.2017.06.009
23. **Alpha-synuclein-induced mitochondrial dysfunction is mediated**. *Mol Neurodegener* (2020) **15** 5. DOI: 10.1186/s13024-019-0349-x
24. **Mitochonic acid-5 attenuates TNF-α-mediated neuronal inflammation**. *J Cell Physiol* (2019) **234** 22172-22182. DOI: 10.1002/jcp.28783
25. **SIRT3 deacetylates FOXO3 to protect mitochondria against oxidative damage**. *Free Radic Biol Med* (2013) **63** 222-234. DOI: 10.1016/j.freeradbiomed.2013.05.002
26. **SIRT3 is a stress-responsive deacetylase in cardiomyocytes that protects cells from stress-mediated cell death by deacetylation of Ku70**. *Mol Cell Biol* (2008) **28** 6384-6401. DOI: 10.1128/MCB.00426-08
27. **SirT3 and p53 deacetylation in aging and cancer**. *J Cell Physiol* (2017) **232** 2308-2311. DOI: 10.1002/jcp.25669
28. **SIRT3 deacetylates and promotes degradation of P53 in PTEN-defective non-small cell lung cancer**. *J Cancer Res Clin Oncol* (2018) **144** 189-198. DOI: 10.1007/s00432-017-2537-9
29. **Sirt3 enhances glioma cell viability by stabilizing Ku70-BAX interaction**. *Onco Targets Ther* (2018) **11** 7559-7567. DOI: 10.2147/OTT.S172672
30. **ATG12 deficiency leads to tumor cell oncosis owing to diminished mitochondrial biogenesis and reduced cellular bioenergetics**. *Cell Death Differ* (2020) **27** 1965-1980. DOI: 10.1038/s41418-019-0476-5
31. **Regulation of mitochondrial biogenesis and PGC-1α under cellular stress**. *Mitochondrion* (2013) **13** 134-142. DOI: 10.1016/j.mito.2013.01.006
32. **Mitochondrial biogenesis and dynamics in the developing and diseased heart**. *Genes Dev* (2015) **29** 1981-1991. DOI: 10.1101/gad.269894.115
33. **Transcriptional integration of mitochondrial biogenesis**. *Trends Endocrinol Metab* (2012) **23** 459-466. DOI: 10.1016/j.tem.2012.06.006
34. **Mitochondrial dysfunction in atherosclerosis**. *DNA Cell Biol* (2019) **38** 597-606. DOI: 10.1089/dna.2018.4552
35. **Mitochondrial quality control in cardiac microvascular ischemia-reperfusion injury: new insights into the mechanisms and therapeutic potentials**. *Pharmacol Res* (2020) **156** 104771. DOI: 10.1016/j.phrs.2020.104771
36. **New insights into the role of mitochondria in cardiac microvascular ischemia/reperfusion injury**. *Angiogenesis* (2020) **23** 299-314. DOI: 10.1007/s10456-020-09720-2
37. **Nuclear control of respiratory chain expression by nuclear respiratory factors and PGC-1-related coactivator**. *Ann N Y Acad Sci* (2008) **1147** 321-334. DOI: 10.1196/annals.1427.006
38. **After the banquet: mitochondrial biogenesis, mitophagy, and cell survival**. *Autophagy* (2013) **9** 1663-1676. DOI: 10.4161/auto.24135
39. **Mitochondrial biogenesis and healthy aging**. *Exp Gerontol* (2008) **43** 813-819. DOI: 10.1016/j.exger.2008.06.014
40. **Regulation of mitochondrial biogenesis**. *Essays Biochem* (2010) **47** 69-84. DOI: 10.1042/bse0470069
41. **Thyroid hormone (T**. *Autophagy* (2019) **15** 131-150. DOI: 10.1080/15548627.2018.1511263
42. **Thyroid hormone stimulation of autophagy is essential for mitochondrial biogenesis and activity in skeletal muscle**. *Endocrinology* (2016) **157** 23-38. DOI: 10.1210/en.2015-1632
43. **Thyroid hormone induction of mitochondrial activity is coupled to mitophagy**. *Autophagy* (2015) **11** 1341-1357. DOI: 10.1080/15548627.2015.1061849
44. **Data on mitochondrial function in skeletal muscle of old mice in response to different exercise intensity**. *Data Brief* (2016) **7** 1519-1523. DOI: 10.1016/j.dib.2016.04.043
45. **Biochemical adaptations in muscle. Effects of exercise on mitochondrial oxygen uptake and respiratory enzyme activity in skeletal muscle**. *J Biol Chem* (1967) **242** 2278-2282. DOI: 10.1016/S0021-9258(18)96046-1
46. **The diverse functions of GAPDH: views from different subcellular compartments**. *Cell Signal* (2011) **23** 317-323. DOI: 10.1016/j.cellsig.2010.08.003
47. **Mitochondria in cancer**. *Cell Stress* (2020) **4** 114-146. DOI: 10.15698/cst2020.06.221
48. **Dietary copper and high saturated and trans fat intakes associated with cognitive decline**. *Arch Neurol* (2006) **63** 1085-1088. DOI: 10.1001/archneur.63.8.1085
49. **Fusion and fission: interlinked processes critical for mitochondrial health**. *Annu Rev Genet* (2012) **46** 265-287. DOI: 10.1146/annurev-genet-110410-132529
50. **Mitochondrial fission, fusion, and stress**. *Science* (2012) **337** 1062-1065. DOI: 10.1126/science.1219855
51. **Mitochondrial DNA copy number in human disease: the more the better?**. *FEBS Lett* (2021) **595** 976-1002. DOI: 10.1002/1873-3468.14021
52. **The many faces of autophagy dysfunction in Huntington's disease: from mechanism to therapy**. *Drug Discov Today* (2014) **19** 963-971. DOI: 10.1016/j.drudis.2014.02.014
53. **Age-related mitochondrial alterations in brain and skeletal muscle of the YAC128 model of Huntington disease**. *npj Aging Mech Dis* (2021) **7** 26. DOI: 10.1038/s41514-021-00079-2
54. **Ripk3 promotes ER stress-induced necroptosis in cardiac IR injury: a mechanism involving calcium overload/XO/ROS/mPTP pathway**. *Redox Biol* (2018) **16** 157-168. DOI: 10.1016/j.redox.2018.02.019
55. **Melatonin protects cardiac microvasculature against ischemia/reperfusion injury**. *J Pineal Res* (2017) **63** e12413. DOI: 10.1111/jpi.12413
56. **Mitochondrial biogenesis: an update**. *J Cell Mol Med* (2020) **24** 4892-4899. DOI: 10.1111/jcmm.15194
57. **Mitochondrial DNA deletions are abundant and cause functional impairment in aged human substantia nigra neurons**. *Nat Genet* (2006) **38** 518-520. DOI: 10.1038/ng1778
58. **Somatic mtDNA mutations cause aging phenotypes without affecting reactive oxygen species production**. *Proc Natl Acad Sci U S A* (2005) **102** 17993-17998. DOI: 10.1073/pnas.0508886102
59. **Increased mitochondrial DNA deletions in substantia nigra dopamine neurons of the aged rat**. *Curr Aging Sci* (2014) **7** 155-160. DOI: 10.2174/1874609808666150122150850
60. **High levels of mitochondrial DNA deletions in substantia nigra neurons in aging and Parkinson disease**. *Nat Genet* (2006) **38** 515-517. DOI: 10.1038/ng1769
61. **Alzheimer's brains harbor somatic mtDNA control-region mutations that suppress mitochondrial transcription and replication**. *Proc Natl Acad Sci U S A* (2004) **101** 10726-10731. DOI: 10.1073/pnas.0403649101
62. **Huntington's disease and mitochondrial DNA deletions: event or regular mechanism for mutant huntingtin protein and CAG repeats expansion?!**. *Cell Mol Neurobiol* (2007) **27** 867-875. DOI: 10.1007/s10571-007-9206-5
63. **Mitochondrial DNA damage is a hallmark of chemically induced and the R6/2 transgenic model of Huntington's disease**. *DNA Repair (Amst)* (2009) **8** 126-136. DOI: 10.1016/j.dnarep.2008.09.004
64. **Enhanced mitochondrial DNA repair resuscitates transplantable lungs donated after circulatory death**. *J Surg Res* (2020) **245** 273-280. DOI: 10.1016/j.jss.2019.07.057
65. **Mitochondrial DNA copy number and biogenesis in different tissues of early- and late-lactating dairy cows**. *J Dairy Sci* (2016) **99** 1571-1583. DOI: 10.3168/jds.2015-9847
66. **Over-expression of**. *Biochem Biophys Res Commun* (2010) **401** 26-31. DOI: 10.1016/j.bbrc.2010.08.143
67. **Overexpression of mitochondrial transcription factor a ameliorates mitochondrial deficiencies and cardiac failure after myocardial infarction**. *Circulation* (2005) **112** 683-690. DOI: 10.1161/CIRCULATIONAHA.104.524835
68. **High mitochondrial DNA copy number has detrimental effects in mice**. *Hum Mol Genet* (2010) **19** 2695-2705. DOI: 10.1093/hmg/ddq163
69. **The mitochondrial acylome emerges: proteomics, regulation by sirtuins, and metabolic and disease implications**. *Cell Metab* (2018) **27** 497-512. DOI: 10.1016/j.cmet.2018.01.016
70. **Proteasomal dysfunction in sporadic Parkinson's disease**. *Neurology* (2006) **66** S37-S49. DOI: 10.1212/wnl.66.10_suppl_4.s37
71. **Mitochondria take center stage in aging and neurodegeneration**. *Ann Neurol* (2005) **58** 495-505. DOI: 10.1002/ana.20624
72. **Mitochondrial dysfunction and its role in motor neuron degeneration in ALS**. *Mitochondrion* (2005) **5** 77-87. DOI: 10.1016/j.mito.2005.01.002
73. **Mitochondrial defect in muscle precedes neuromuscular junction degeneration and motor neuron death in**. *Acta Neuropathol* (2019) **138** 123-145. DOI: 10.1007/s00401-019-01988-z
74. **Mitochondrial dysfunction in vascular wall cells and its role in atherosclerosis**. *Int J Mol Sci* (2021) **22** 8990. DOI: 10.3390/ijms22168990
75. **SirT3 regulates the mitochondrial unfolded protein response**. *Mol Cell Biol* (2014) **34** 699-710. DOI: 10.1128/MCB.01337-13
76. **Interaction of Sirt3 with OGG1 contributes to repair of mitochondrial DNA and protects from apoptotic cell death under oxidative stress**. *Cell Death Dis* (2013) **4** e731. DOI: 10.1038/cddis.2013.254
77. **Sirt3 regulates the level of mitochondrial DNA repair activity through deacetylation of NEIL1, NEIL2, OGG1, MUTYH, APE1 and LIG3 in colorectal cancer**. *Pol Przegl Chir* (2019) **92** 1-4. DOI: 10.5604/01.3001.0013.5539
78. **Mitochondrial DNA damage and repair in neurodegenerative disorders**. *DNA Repair (Amst)* (2008) **7** 1110-1120. DOI: 10.1016/j.dnarep.2008.03.012
79. **Mitochondrial DNA repair: a critical player in the response of cells of the CNS to genotoxic insults**. *Neuroscience* (2007) **145** 1249-1259. DOI: 10.1016/j.neuroscience.2006.10.002
80. **DNA repair, mitochondria, and neurodegeneration**. *Neuroscience* (2007) **145** 1318-1329. DOI: 10.1016/j.neuroscience.2006.08.061
81. **Base excision repair causes age-dependent accumulation of single-stranded DNA breaks that contribute to Parkinson disease pathology**. *Cell Rep* (2021) **36** 109668. DOI: 10.1016/j.celrep.2021.109668
82. **DNA damage-induced neurodegeneration in accelerated ageing and Alzheimer's disease**. *Int J Mol Sci* (2021) **22** 6748. DOI: 10.3390/ijms22136748
83. **ALS spinal neurons show varied and reduced mtDNA gene copy numbers and increased mtDNA gene deletions**. *Mol Neurodegener* (2010) **5** 21. DOI: 10.1186/1750-1326-5-21
84. **Mutant SOD1**. *Neurobiol Dis* (2013) **51** 72-81. DOI: 10.1016/j.nbd.2012.07.004
85. **PGC-1**. *Oxid Med Cell Longev* (2020) **2020** 1452696. DOI: 10.1155/2020/1452696
86. **PGC-1α/ERRα-sirt3 pathway regulates DAergic neuronal death by directly deacetylating SOD2 and ATP synthase β**. *Antioxid Redox Signal* (2016) **24** 312-328. DOI: 10.1089/ars.2015.6403
87. **Mitochondrial dynamics and Parkinson's disease: focus on parkin**. *Antioxid Redox Signal* (2012) **16** 935-949. DOI: 10.1089/ars.2011.4105
88. **Intracellular and intercellular mitochondrial dynamics in Parkinson's disease**. *Front Neurosci* (2019) **13** 930. DOI: 10.3389/fnins.2019.00930
89. **Pathogenic alpha-synuclein aggregates preferentially bind to mitochondria and affect cellular respiration**. *Acta Neuropathol Commun* (2019) **7** 41. DOI: 10.1186/s40478-019-0696-4
90. **SUMOylation, aging and autophagy in neurodegeneration**. *NeuroToxicology* (2018) **66** 53-57. DOI: 10.1016/j.neuro.2018.02.015
91. **SIRT3-dependent mitochondrial dynamics remodeling contributes to oxidative stress-induced melanocyte degeneration in vitiligo**. *Theranostics* (2019) **9** 1614-1633. DOI: 10.7150/thno.30398
92. **Mitochondrial SIRT3 confers neuroprotection in Huntington's disease by regulation of oxidative challenges and mitochondrial dynamics**. *Free Radic Biol Med* (2021) **163** 163-179. DOI: 10.1016/j.freeradbiomed.2020.11.031
93. **Mitochondria in endothelial cells: sensors and integrators of environmental cues**. *Redox Biol* (2017) **12** 821-827. DOI: 10.1016/j.redox.2017.04.021
94. **The large apparent work capability of the blood-brain barrier: a study of the mitochondrial content of capillary endothelial cells in brain and other tissues of the rat**. *Ann Neurol* (1977) **1** 409-417. DOI: 10.1002/ana.410010502
95. **Sirt3 ameliorates mitochondrial dysfunction and oxidative stress through regulating mitochondrial biogenesis and dynamics in cardiomyoblast**. *Cell Signal* (2022) **94** 110309. DOI: 10.1016/j.cellsig.2022.110309
96. **PGC-1α regulates the mitochondrial antioxidant defense system in vascular endothelial cells**. *Cardiovasc Res* (2005) **66** 562-573. DOI: 10.1016/j.cardiores.2005.01.026
97. **Transcriptional control of cardiac fuel metabolism and mitochondrial function**. *Cold Spring Harb Symp Quant Biol* (2011) **76** 175-182. DOI: 10.1101/sqb.2011.76.011965
98. **PGC-1**. *Respir Physiol Neurobiol* (2015) **205** 84-91. DOI: 10.1016/j.resp.2014.10.015
99. **Mitochondria-targeted esculetin alleviates mitochondrial dysfunction by AMPK-mediated nitric oxide and SIRT3 regulation in endothelial cells: potential implications in atherosclerosis**. *Sci Rep* (2016) **6** 24108. DOI: 10.1038/srep24108
100. **The SirT3 divining rod points to oxidative stress**. *Mol Cell* (2011) **42** 561-568. DOI: 10.1016/j.molcel.2011.05.008
101. **Sirtuin 3, a new target of PGC-1α, plays an important role in the suppression of ROS and mitochondrial biogenesis**. *PLoS One* (2010) **5** e11707. DOI: 10.1371/journal.pone.0011707
102. **Mitofusins are required for angiogenic function and modulate different signaling pathways in cultured endothelial cells**. *J Mol Cell Cardiol* (2011) **51** 885-893. DOI: 10.1016/j.yjmcc.2011.07.023
103. **OPA1 and angiogenesis: beyond the fusion function**. *Cell Metab* (2020) **31** 886-887. DOI: 10.1016/j.cmet.2020.04.014
104. **Mitochondrial dynamics as a bridge between mitochondrial dysfunction and insulin resistance**. *Arch Physiol Biochem* (2009) **115** 1-12. DOI: 10.1080/13813450802676335
105. **Mitochondrial fusion and fission proteins as novel therapeutic targets for treating cardiovascular disease**. *Eur J Pharmacol* (2015) **763** 104-114. DOI: 10.1016/j.ejphar.2015.04.056
106. **G-protein β2 subunit interacts with mitofusin 1 to regulate mitochondrial fusion**. *Nat Commun* (2010) **1** 101. DOI: 10.1038/ncomms1099
107. **Sirtuins and type 2 diabetes: role in inflammation, oxidative stress, and mitochondrial function**. *Front Endocrinol (Lausanne)* (2019) **10** 187. DOI: 10.3389/fendo.2019.00187
108. **Fatty liver is associated with reduced SIRT3 activity and mitochondrial protein hyperacetylation**. *Biochem J* (2011) **433** 505-514. DOI: 10.1042/BJ20100791
109. **SIRT3 deficiency induces endothelial insulin resistance and blunts endothelial-dependent vasorelaxation in mice and human with obesity**. *Sci Rep* (2016) **6** 23366. DOI: 10.1038/srep23366
110. **SIRT3 overexpression attenuates palmitate-induced pancreatic β-cell dysfunction**. *PLoS One* (2015) **10** e0124744. DOI: 10.1371/journal.pone.0124744
111. **Pancreatic β-cells and type 2 diabetes development**. *Curr Diabetes Rev* (2017) **13** 108-121. DOI: 10.2174/1573399812666151020101222
112. **Pancreatic β-cells in type 1 and type 2 diabetes mellitus: different pathways to failure**. *Nat Rev Endocrinol* (2020) **16** 349-362. DOI: 10.1038/s41574-020-0355-7
113. **Mitochondrial fission triggered by hyperglycemia is mediated by ROCK1 activation in podocytes and endothelial cells**. *Cell Metab* (2012) **15** 186-200. DOI: 10.1016/j.cmet.2012.01.009
114. **Pre-diabetes alters testicular PGC1-α/SIRT3 axis modulating mitochondrial bioenergetics and oxidative stress**. *Biochim Biophys Acta* (2014) **1837** 335-344. DOI: 10.1016/j.bbabio.2013.12.008
115. **Mitofusin 2 (Mfn2) links mitochondrial and endoplasmic reticulum function with insulin signaling and is essential for normal glucose homeostasis**. *Proc Natl Acad Sci U S A* (2012) **109** 5523-5528. DOI: 10.1073/pnas.1108220109
116. **Mfn2 deficiency links age-related sarcopenia and impaired autophagy to activation of an adaptive mitophagy pathway**. *EMBO J* (2016) **35** 1677-1693. DOI: 10.15252/embj.201593084
117. **Mitochondrial fission contributes to mitochondrial dysfunction and insulin resistance in skeletal muscle**. *Mol Cell Biol* (2012) **32** 309-319. DOI: 10.1128/MCB.05603-11
118. **The causal role of mitochondrial dynamics in regulating insulin resistance in diabetes: link through mitochondrial reactive oxygen species**. *Oxid Med Cell Longev* (2018) **2018** 7514383. DOI: 10.1155/2018/7514383
|
---
title: 'Gut microbiota links with cognitive impairment in amyotrophic lateral sclerosis:
A multi-omics study'
authors:
- Zhenxiang Gong
- Li Ba
- Jiahui Tang
- Yuan Yang
- Zehui Li
- Mao Liu
- Chun Yang
- Fengfei Ding
- Min Zhang
journal: Journal of Biomedical Research
year: 2023
pmcid: PMC10018415
doi: 10.7555/JBR.36.20220198
license: CC BY 4.0
---
# Gut microbiota links with cognitive impairment in amyotrophic lateral sclerosis: A multi-omics study
## Abstract
Recently, cognitive impairments (CI) and behavioral abnormalities in patients with amyotrophic lateral sclerosis (ALS) have been reported. However, the underlying mechanisms have been poorly understood. In the current study, we explored the role of gut microbiota in CI of ALS patients. We collected fecal samples from 35 ALS patients and 35 healthy controls. The cognitive function of the ALS patients was evaluated using the Edinburgh Cognitive and Behavioral ALS Screen. We analyzed these samples by using 16S rRNA gene sequencing as well as both untargeted and targeted (bile acids) metabolite mapping between patients with CI and patients with normal cognition (CN). We found altered gut microbial communities and a lower ratio of Firmicutes/Bacteroidetes in the CI group, compared with the CN group. In addition, the untargeted metabolite mapping revealed that 26 and 17 metabolites significantly increased and decreased, respectively, in the CI group, compared with the CN group. These metabolites were mapped to the metabolic pathways associated with bile acids. We further found that cholic acid and chenodeoxycholic acid were significantly lower in the CI group than in the CN group. In conclusion, we found that the gut microbiota and its metabolome profile differed between ALS patients with and without CI and that the altered bile acid profile in fecal samples was significantly associated with CI in ALS patients. These results need to be replicated in larger studies in the future.
ALS: amyotrophic lateral sclerosis; BMI: body mass index; ALSFRS-R: revised ALS Functional Rating Scale; ECAS: Edinburgh Cognitive and Behavioral ALS Screen.
A: Comparison of microbial α diversity (assessed by Sobs, Ace and Chao index). B: Comparison of microbial β diversity (assessed by PCoA). C: Relative abundance of the microbial community at the phylum level. D and E: Distinguishing bacterial taxa identified by LEfSe analysis. Cladogram showing the phylogenetic distribution of the bacterial lineages. Circles indicate phylogenetic levels from domain to genus. The diameter of each circle is proportional to the abundance of the bacterial group (D). The letters show the distinguishing taxa with a LDA score >3 (E). F: Comparison of fecal metabolites. Among 14 108 metabolites detected, 350 increased in ALS ($P \leq 0.05$), while 339 were reduced in ALS ($P \leq 0.05$), compared with the HC group. G and H: The classification of 41 distinguishing fecal metabolites ($P \leq 0.05$, VIP score >1) were identified in database. Compounds with biological roles in KEGG (G). Superclass metabolites in HMDB database (H). I: In KEGG, 41 distinguishing fecal metabolites between ALS and HC were mapped to five metabolic pathways. J: Relative concentrations of 41 distinguishing fecal metabolites ($P \leq 0.05$, VIP score >1) between ALS and HC. ns: not significant; ALS: amyotrophic lateral sclerosis; HC: healthy controls; PCoA: Principal coordinates analysis; LEfSe: the Linear discriminant analysis of effect size; VIP: Variable Importance in the Projection; KEGG: Kyoto Encyclopedia of Genes and Genomes; HMDB: Human Metabolome Database.
A–C: Cognitive function of patients with ALS in the current study. Frequency of patients with CI in five subdomains of ECAS (A). Frequency of patients with CI and behavioral impairments (B). Coexistence of CI and behavioral impairments in patients with ALS (C). D: Comparison of microbial α diversity between the CN and CI groups (assessed by Sobs, Ace, and Chao index). E: Comparison of microbial β diversity between the CN and CI groups (assessed by PCoA). F and G: Relative abundance of the microbial community at phylum level between CN and CI. H and I: Distinguishing bacterial taxa between the CN and CI groups was identified by LEfSe analysis. Cladogram showing the phylogenetic distribution of the bacterial lineages. Circles indicate phylogenetic levels from domain to genus. The diameter of each circle is proportional to the abundance of the bacterial group (H). The letters show the distinguishing taxa with a LDA score >3 (I). ns: not significant; ECAS: Edinburgh Cognitive and Behavioral ALS Screen; CN: normal cognition; CI: cognitive impairments; BVN: behavior normal; BVI: behavior impaired; PCoA: Principal coordinates analysis; LEfSe: the Linear discriminant analysis of effect size.
A: Comparison of fecal metabolites. Three hundred and thirty metabolites increased ($P \leq 0.05$), and 369 metabolites decreased ($P \leq 0.05$), respectively, in the CI group, compared with the CN group. B: In OPLS-DA, obvious separation indicates a different structure of fecal metabolites between the CN and CI groups. C: Relative concentrations of 43 distinguishing fecal metabolites ($P \leq 0.05$, VIP score >1) between the CN and CI groups. D and E: The classification of 43 distinguishing fecal metabolites. HMDB Superclass metabolites (D). HMDB Class metabolites (E). F: In KEGG, 43 distinguishing fecal metabolites between the CN and CI groups were mapped to 6 metabolic pathways. CN: normal cognition; CI: cognitive impairments; OPLS-DA: Orthogonal partial least squares discriminate analysis; VIP: Variable Importance in the Projection; HMDB: Human Metabolome Database; KEGG: Kyoto Encyclopedia of Genes and Genomes.
A: The correlation between 10 metabolites with the highest concentrations and 10 bacterial genera with the highest relative abundance (*$P \leq 0.05$). B: Comparison of genus belonging to family ruminococcaceae between the CN and CI groups. SD: standard deviation; PBA: primary bile acids; SBA: secondary bile acids; CN: normal cognition; CI: cognitive impairments.
## Introduction
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder characterized by a progressive degeneration of the upper and lower motor neurons[1]. Motor functions of ALS patients would usually deteriorate within three to five years, and they might suffer from respiratory failure in the late stages[1]. ALS is devastating to both patients and caregivers due to an absence of curative treatment. Many neuroscientists and neurologists have observed that cognitive and behavioral abnormalities are common among ALS patients[2]. Approximately $35\%$ of ALS patients have comorbid cognitive impairments (CI)[3]. Cognitive decline severely impairs treatment compliance among patients and causes a burden on caregivers[4], leading to shortened survival spans and dampening the quality of life[5]. Cognitive decline in ALS patients primarily manifests with executive, language, and visual space dysfunction[6]. A subpopulation of ALS patients could develop behavioral impairments, including loss of interest, apathy, absence of insight, irritability, and aggression[7–8]. However, the underlying mechanisms remain unknown.
*Among* genes related to the pathogenesis of ALS, hexanucleotide GGGGCC repeat expansions in the C9orf72 gene explain the association between ALS and frontotemporal dementia. About $5\%$–$15\%$ of ALS patients satisfy the diagnostic criteria for frontotemporal dementia[9]. However, this C9orf72 mutation barely occurs in Chinese ALS patients[10–11]. Our clinical data suggested that about $38.7\%$ of ALS patients exhibited CI and that $31.5\%$ had behavioral abnormalities[12]. Therefore, we proposed an alternative mechanism other than genetic susceptibility that may contribute to the impaired cognition in ALS patients.
Recent studies have elucidated a link between gut microbiota and CI in neurodegenerative disorders, such as Alzheimer's (AD) and Parkinson's disease[13–14]. The altered gut microbiota play a role in the CI of patients suffering from neurodegenerative diseases[15]. A few microbiomics studies combined with metabolomics have shedded light on cognitive impairment mechanisms in patients with neurodegenerative diseases, including regulating microglial function using secreted metabolites and neurotransmitters[16–17]. Although several global study groups have reported altered gut microbiota in ALS patients compared with healthy controls, the conclusions are inconsistent[18–22]. To date, no studies have uncovered the mechanistic links between CI and gut microbiota in ALS patients.
In the current study, we first conducted 16S rRNA sequencing of fecal samples from both ALS patients and healthy controls. Then we further compared gut microbiota and untargeted fecal metabolites of the ALS patients with and without CI. Biostatistical analyses revealed that the extracted metabolites significantly differed between the two groups and that these metabolites were mapped to metabolic pathways associated with bile acids. We attempted to identify a link between the gut microbiota and CI in ALS patients, providing clues for future mechanistic studies.
## Study design, registrations, and patient consents
We conducted a case-control study of 35 patients with ALS and 35 age- and sex-matched healthy controls and also explored the changes in gut microbiota and fecal metabolites between ALS patients with and without cognitive decline. The current study was approved by the ethics committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology (TJ-IRB20201219). All participants provided written informed consent, and the study was conducted following the Declaration of Helsinki.
## Subject recruitment and fecal sample collection
The study design is described in Fig. 1. Briefly, 35 ALS patients diagnosed with the revised El Escorial criteria[23] and 35 age- and sex-matched healthy controls were recruited between September 2020 and December 2021. Exclusion criteria were incorporated to exclude diseases with a clear impact on the gut microbiota (Supplementary Table 1, available online). Specifically, patients with severe dysphagia were excluded to reduce the biases in nutrient intake. Two fecal samples of each participant were collected from the first bowel moment of the day, and the samples were collected in 15 mL falcon tubes and stored at −80 ℃.
**Figure 1:** *The research flow chart.*
## Assessment of motor function and cognitive function in ALS patients
The Revised ALS Functional Rating Scale (ALSFRS-R) was performed to assess disease severity of ALS patients[24]. The Edinburgh Cognitive and Behavioral ALS Screen (ECAS) is specifically designed to identify cognitive and behavioral changes in ALS patients[6]. The ECAS evaluates three ALS-specific cognitive domains (language, verbal fluency, and executive function) and two ALS-non-specific cognitive domains (memory and visuospatial function). In the current study, we screened ALS patients with CI and ALS patients with normal cognition (CN), based on the cut-off value of total ECAS scores in Chinese population[6]. Additionally, ECAS includes a brief questionnaire to detect behavioral abnormalities, completed by primary caregivers. Patients with an abnormality in at least one behavioral domain were considered to have behavioral impairments (BVI); otherwise, behavior normal (BVN). Two trained investigators performed the ALSFRS-R and the Chinese version of ECAS.
## 16S rRNA gene sequencing and microbiome analysis
The amplicon sequencing of the 16S rRNA gene was conducted by the Shanghai Majorbio Bio-Pharm Technology Co., Ltd. (China), as previously described[25]. Briefly, the DNA was extracted (Cat. No. DP712, Tiangen, Beijing, China) and the V3-V4 variable region was amplified with barcode-indexed primers (338F and 806R). Then, amplicons were sequenced using the Illumina MiSeq platform (PE300). The raw sequences were processed as previously described[25]. The sequence data has been deposited in the National Genomics Data Center (https://ngdc.cncb.ac.cn/, PRJCA006335). Sobs, Chao, and Ace index were calculated (Mothur, v1.30.2) to evaluate the richness and evenness (α diversity) of the microbial communities[26–27]. Principal coordinate analysis (PCoA) and permutational multivariate analysis of variance were used for the comparison of β diversity based on the Bray-Curtis distance[28]. The linear discriminant analysis (LDA) of effect size (LEfSe) revealed representative bacterial taxa in each group[29], and the threshold was an LDA score >3.
## Fecal metabolite detection and metabolomic analysis
The Shanghai Majorbio Bio-Pharm Technology Co., Ltd. performed the detection of fecal metabolites. The metabolites were separated using an ExionLC AD System (AB Sciex, Framingham, USA). The Ultra Performance Liquid Chromatography (UPLC) system was coupled with a quadrupole time-of-flight mass spectrometer (AB SCIEX-Triple TOF 5600+, Framingham, USA) and electrospray ionization source. The data acquisition was performed in the Data Dependent Acquisition mode. The metabolites were annotated based on the Human Metabolome Database (HMDB) and Metlin mass spectral database. The orthogonal partial least squares discriminate analysis (OPLS-DA) showed the similarity of fecal metabolites among different groups. Representative metabolites from each group were selected based on the Variable Importance in the Projection (VIP) value in OPLS-DA and the P-value within the comparison of concentrations (VIP score >1, $P \leq 0.05$). The metabolites were mapped to biochemical pathways in the Kyoto Encyclopedia of Genes and Genomes (KEGG) for exploring metabolic functions. In the bile acids-targeted detection, the ion fragments were identified in AB SCIEX quantification software OS and manually checked. The concentration of 46 bile acids was evaluated based on the standard curve plotted using the mass spectral peak area as the ordinate and the analytic concentration.
## Statistical analysis
Continuous variables are presented as mean (SD) or median (including $25\%$ quartile and $75\%$ quartile). In comparing the two groups, Student's t-test was used for normally distributed data, Mann-Whitney U-test was used for non-normally distributed data, and Fisher's exact probability test was employed for categorical data. Spearman's correlation coefficient was calculated in the correlation analysis. $P \leq 0.05$ was considered statistically significant. Statistical analysis was performed using SPSS (version 23.0). Graphs were drawn using R (version 3.4.1) and GraphPad Prism 7.0.
## Demographic and clinical characteristics
The same number of ALS patients and healthy controls (35 each) were enrolled. No significant difference was observed among age, body mass index (BMI), and sex ratio between the cases and controls (Table 1). Among the 35 ALS patients, 32 had a spinal onset, and three had a bulbar onset. The median disease duration in these patients with ALS was 10 months, and the median score of ALSFRS-R was 43.
**Table 1**
| Parameters | Participants | Participants.1 | Participants.2 | Participants.3 | P-value1 | P-value2 |
| --- | --- | --- | --- | --- | --- | --- |
| Parameters | ALS (N=35) | HC (N=35) | CN (N=19) | CI (N=10) | P-value1 | P-value2 |
| Demographic information | Demographic information | Demographic information | Demographic information | Demographic information | Demographic information | Demographic information |
| Age (years) | 54 (8) | 54 (9) | 50 (6) | 59 (10) | 0.967a | 0.030a |
| Female/Male | 14/21 | 14/21 | 8/11 | 4/6 | 1c | 0.615c |
| BMI (kg/m2) | 22.15 (2.69) | 22.77 (2.11) | 22.35 (2.97) | 22.02 (3.01) | 0.291a | 0.776a |
| Clinical characteristics | Clinical characteristics | Clinical characteristics | Clinical characteristics | Clinical characteristics | Clinical characteristics | Clinical characteristics |
| Duration (months) | 10 (6, 24) | – | 7 (4, 15) | 24 (12, 24) | – | 0.006b |
| ALSFRS-R | 43 (39, 46) | – | 43 (39, 46) | 43 (39, 45) | – | 0.626b |
| ALSFRS-R, Bulb scores | 12 (10, 12) | – | 12 (11, 12) | 11 (9, 12) | – | 0.129b |
| ALSFRS-R, Respiration scores | 12 (12, 12) | – | 12 (12, 12) | 12 (12, 12) | – | 0.490b |
| ECAS | – | – | 97 (11) | 59 (17) | – | - |
| Data are presented as mean (SD) or median (25% quantile, 75% quantile). P-values were calculated by Student's t-testa, Mann-Whitney U-testb or Fisher's exact-testc. P-value1: ALS vs. HC; P-value2: CN vs. CI. ALS: amyotrophic lateral sclerosis; HC: healthy control; CN: normal cognition; CI: cognitive impairments; BMI: body mass index; ALSFRS-R: revised ALS Functional Rating Scale; ECAS: the Edinburgh Cognitive and Behavioral ALS Screen; SD: standard deviation. | Data are presented as mean (SD) or median (25% quantile, 75% quantile). P-values were calculated by Student's t-testa, Mann-Whitney U-testb or Fisher's exact-testc. P-value1: ALS vs. HC; P-value2: CN vs. CI. ALS: amyotrophic lateral sclerosis; HC: healthy control; CN: normal cognition; CI: cognitive impairments; BMI: body mass index; ALSFRS-R: revised ALS Functional Rating Scale; ECAS: the Edinburgh Cognitive and Behavioral ALS Screen; SD: standard deviation. | Data are presented as mean (SD) or median (25% quantile, 75% quantile). P-values were calculated by Student's t-testa, Mann-Whitney U-testb or Fisher's exact-testc. P-value1: ALS vs. HC; P-value2: CN vs. CI. ALS: amyotrophic lateral sclerosis; HC: healthy control; CN: normal cognition; CI: cognitive impairments; BMI: body mass index; ALSFRS-R: revised ALS Functional Rating Scale; ECAS: the Edinburgh Cognitive and Behavioral ALS Screen; SD: standard deviation. | Data are presented as mean (SD) or median (25% quantile, 75% quantile). P-values were calculated by Student's t-testa, Mann-Whitney U-testb or Fisher's exact-testc. P-value1: ALS vs. HC; P-value2: CN vs. CI. ALS: amyotrophic lateral sclerosis; HC: healthy control; CN: normal cognition; CI: cognitive impairments; BMI: body mass index; ALSFRS-R: revised ALS Functional Rating Scale; ECAS: the Edinburgh Cognitive and Behavioral ALS Screen; SD: standard deviation. | Data are presented as mean (SD) or median (25% quantile, 75% quantile). P-values were calculated by Student's t-testa, Mann-Whitney U-testb or Fisher's exact-testc. P-value1: ALS vs. HC; P-value2: CN vs. CI. ALS: amyotrophic lateral sclerosis; HC: healthy control; CN: normal cognition; CI: cognitive impairments; BMI: body mass index; ALSFRS-R: revised ALS Functional Rating Scale; ECAS: the Edinburgh Cognitive and Behavioral ALS Screen; SD: standard deviation. | Data are presented as mean (SD) or median (25% quantile, 75% quantile). P-values were calculated by Student's t-testa, Mann-Whitney U-testb or Fisher's exact-testc. P-value1: ALS vs. HC; P-value2: CN vs. CI. ALS: amyotrophic lateral sclerosis; HC: healthy control; CN: normal cognition; CI: cognitive impairments; BMI: body mass index; ALSFRS-R: revised ALS Functional Rating Scale; ECAS: the Edinburgh Cognitive and Behavioral ALS Screen; SD: standard deviation. | Data are presented as mean (SD) or median (25% quantile, 75% quantile). P-values were calculated by Student's t-testa, Mann-Whitney U-testb or Fisher's exact-testc. P-value1: ALS vs. HC; P-value2: CN vs. CI. ALS: amyotrophic lateral sclerosis; HC: healthy control; CN: normal cognition; CI: cognitive impairments; BMI: body mass index; ALSFRS-R: revised ALS Functional Rating Scale; ECAS: the Edinburgh Cognitive and Behavioral ALS Screen; SD: standard deviation. |
## Microbial and metabolomic differences between ALS patients and healthy controls
According to the 16S rRNA gene data, there was no difference in the α diversity ($P \leq 0.05$) between ALS patients and healthy controls (Fig. 2A). Regarding β diversity, PCoA analysis depicted a significant difference between ALS patients and healthy controls ($P \leq 0.05$) (Fig. 2B). Compared with the healthy control group, the phylum Proteobacteria significantly decreased in the ALS patient group (Fig. 2C). The Firmicutes to Bacteroidetes (F/B) ratio was comparable between ALS patients and healthy controls (Supplementary Table 2, available online). Additionally, the *Lefse analysis* revealed distinctive bacterial species of each group and the subordination of bacterial species (Fig. 2D and E).
**Figure 2:** *Difference in microbial community and fecal metabolites between ALS patients and healthy controls.*
For the untargeted metabolite mapping, 29 ALS patients and 23 healthy controls were included. A total of 14108 metabolites were detected, among which 41 statistically different metabolites were identified from the KEGG and HMDB database ($P \leq 0.05$) (Fig. 2F–H). Twenty-two metabolites increased, and 19 metabolites decreased, respectively, in the ALS patient group, compared with the healthy control group (Fig. 2J). In the KEGG pathway enrichment analysis, those 41 metabolites were mapped to five metabolic pathways ($P \leq 0.05$), including the retrograde endocannabinoid signaling, inflammatory mediator regulation of the transient receptor potential channels, sphingolipid, nicotinamide, and thiamine metabolism (Fig. 2I).
## Microbial difference between ALS patients with and without cognitive impairment
Among 29 ALS patients who underwent the ECAS, 10 ($34.48\%$) had CI, and 19 patients ($65.52\%$) had normal cognition (Fig. 3A and B), according to the cut-off score of ECAS in Chinese ALS patients[6]. Fourteen patients ($48.28\%$) exhibited at least a single type of BVI (Fig. 3B), and six patients showed CI combined with abnormal behaviors (Fig. 3C). The microbial α diversity was similar between the CN and CI groups ($P \leq 0.05$) (Fig. 3D). The microbial β diversity was significantly different between the CN and CI groups based on the PCoA ($P \leq 0.05$) (Fig. 3E). The phylum Bacteroidetes ($P \leq 0.05$) and Proteobacteria ($P \leq 0.05$) elevated, while Firmicutes and Actinobacteria reduced in the CI group ($P \leq 0.05$) compared with the CN group (Fig. 3F and G). The F/B ratio was significantly lower in the CI group than that in the CN group ($P \leq 0.05$) (Supplementary Table 3, available online). In Lefse analysis, the CI group revealed a higher abundance of bacterial species belonging to Bacteroidia and Verrucomicrobiae classes, while the CN group showed a higher abundance of bacterial species belonging to the class Erysipelotrichia (Fig. 3H and I).
**Figure 3:** *Difference in the microbial community between patients with CN and CI.*
## Fecal metabolite difference between patients with and without CI
The OPLS-DA analysis separates the CN and CI groups into discrete distributions, indicating distinctive metabolomic characteristics (Fig. 4B). Among 43 metabolites identified in the KEGG and HMDB, 26 metabolites decreased and 17 metabolites increased in the CI group (Fig. 4C–E), respectively, compared with the CN group. Those 43 metabolites were mapped to six metabolic pathways, including the sulfur relay system, antifolate resistance, vitamin B6 metabolism, bile secretion, primary bile acid biosynthesis, and secondary bile acid biosynthesis (Fig. 4F). Patients with CI showed unique metabolic characteristics in fecal samples. Notably, three of six distinctive metabolic pathways were linked to bile acid metabolism.
**Figure 4:** *Difference in fecal metabolites between patients with CN and CI.*
## Bile acids profile changed in patients with CI
Based on our findings in the untargeted metabolic mapping (Fig. 4F), we further undertook targeted bile acids quantification between the CN and CI groups. Primary bile acids (PBAs) include cholic acid (CA), chenodeoxycholic acid (CDCA), and their taurine or glycine bound derivatives. Secondary bile acids (SBAs) were the PBA derivates, including deoxycholic acid (DCA, a CA derivative), lithocholic acid (LCA, a CDCA derivative), ursodeoxycholic acid (UDCA), and α-muricholic acid and their conjugated forms[30]. In total, 46 kinds of bile acids were identified. Compared with the CN group, PBAs were significantly lower in the CI group, including CA, CDCA, GlyCA, GlyCDCA, and TaurCDCA ($P \leq 0.05$) (Table 2 and Supplementary Table 5 [available online]).
**Table 2**
| Bile acids | Classification | Group | Group.1 | P-value |
| --- | --- | --- | --- | --- |
| Bile acids | Classification | CN (ng/g) | CI (ng/g) | P-value |
| Cholic acid | PBA | 17476.93 (3392.46, 31117.48) | 2298.26 (402.65, 15086.72) | 0.035 |
| Chenodeoxycholic acid | PBA | 13405.17 (4045.43, 23799.13) | 2222.42 (541.38, 13613.2) | 0.017 |
| Glycocholic acid | PBA | 440.68 (163.06, 1255.82) | 99.83 (39.4, 577.28) | 0.022 |
| Glycochenodeoxycholic acid | PBA | 746.89 (398.69, 1124.03) | 218.68 (100.81, 428.34) | 0.004 |
| Taurochenodeoxycholic acid | PBA | 220.9 (120.53, 407.36) | 77.85 (32.10, 223.88) | 0.022 |
| Allocholic Acid | SBA | 1591.32 (212.42, 2767.59) | 152.17 (74.74, 1402.8) | 0.039 |
| Alpha-Muricholic acid | SBA | 36.88 (19.05) | 17.45 (14.45) | 0.009 |
| Beta-Muricholic acid | SBA | 988.52 (398.5, 2676.31) | 265.4 (226.12, 1015.36) | 0.048 |
| Chenodeoxycholic Acid 24-Acyl-β-D-glucuronide | SBA | 3.14 (1.22, 9.08) | 0.39 (0.20, 4.19) | 0.039 |
| Glycohyocholic acid | SBA | 2.4 (1.36, 5.34) | 0.39 (0.34, 1.03) | 0.001 |
| Glycoursodeoxycholic acid | SBA | 268.64 (106.39, 602.2) | 28.84 (23.06, 267.43) | 0.008 |
| Lithocholic acid 3-sulfate | SBA | 2337.19 (941.51, 8830.24) | 461.87 (122.87, 3939.11) | 0.035 |
| Norcholic acid | SBA | 131.32 (84.62, 299.36) | 55.98 (35.71, 157.43) | 0.044 |
| Taurohyodeoxycholic acid | SBA | 25.91 (10.84, 120.13) | 2.44 (0.94, 33.75) | 0.013 |
| Tauroursodeoxycholic acid | SBA | 24.58 (7.75, 95.88) | 2.04 (1.15, 32.88) | 0.019 |
| Data are expressed as median (25% quantile, 75% quantile). P-values were calculated by Student's t test or Mann-Whitney U-test. CN: normal cognition; CI: cognitive impairments; PBA: primary bile acid; SBA: secondary bile acid. | Data are expressed as median (25% quantile, 75% quantile). P-values were calculated by Student's t test or Mann-Whitney U-test. CN: normal cognition; CI: cognitive impairments; PBA: primary bile acid; SBA: secondary bile acid. | Data are expressed as median (25% quantile, 75% quantile). P-values were calculated by Student's t test or Mann-Whitney U-test. CN: normal cognition; CI: cognitive impairments; PBA: primary bile acid; SBA: secondary bile acid. | Data are expressed as median (25% quantile, 75% quantile). P-values were calculated by Student's t test or Mann-Whitney U-test. CN: normal cognition; CI: cognitive impairments; PBA: primary bile acid; SBA: secondary bile acid. | Data are expressed as median (25% quantile, 75% quantile). P-values were calculated by Student's t test or Mann-Whitney U-test. CN: normal cognition; CI: cognitive impairments; PBA: primary bile acid; SBA: secondary bile acid. |
## The correlation between the altered bile acids and gut microbial species
We analyzed the correlations between the abundance of the top 10 microbial genus species and the concentration of the top 10 abundant metabolites. CDCA and CA were associated with eight out of 10 top-expressing genera (Fig. 5A). Bacterial species belonging to the family Ruminococcaceae and genus *Clostridium are* the known bacteria that play a critical role in transforming PBAs to SBAs, converting CA to DCA, and CDCA to LCA[31–32]. Further, we compared the relative abundance of all genera belonging to the family Ruminococcaceae between the CN group and the CI group, 10 of 11 genera belonging to the Ruminococcaceae family increased in the CI group (Fig. 5B).
**Figure 5:** *Difference in bacterial species related to bile acids metabolism between patients with CN and CI.*
## Microbial and metabolomic differences between ALS patients with and without behavioral abnormalities
In the current study, 14 ALS patients were divided into patients with behavior impairments (BVI) and 15 ALS patients were divided into patients without behavior impairments (BVN) as described in the methods. There was also no difference in α diversity and β diversity between the BVN and BVI groups ($P \leq 0.05$) (Supplementary Fig. 1A and B, available online). Both the abundance of four dominant phyla and the ratio of F/B were comparable between the BVN and BVI groups ($P \leq 0.05$) (Supplementary Fig. 1C and Supplementary Table 4, available online). Additionally, there was no bacterial species with a LDA score >3 in LEfSe analysis. In comparison of metabolites, there were 22 metabolites with significant differences in concentration between the BVI and BVI groups (Supplementary Fig. 1D, available online). However, these metabolites were not significantly mapped to any metabolic pathway. In the quantitative determination of bile acids, only dehydrocholic acid was higher in BVI group compared with the BVN group (Supplementary Table 6, available online).
## Discussion
In the current study, we found altered gut microbiota and lowered PBAs levels in fecal samples of ALS patients with CI, compared with patients with normal cognition. Decreased PBAs were associated with increased bacteria that could consume PBAs. The current data suggests that the disruption of gut microbiota-related bile acids metabolism is associated with cognitive decline in ALS.
## Gut microbiota and fecal metabolites differed between ALS patients with and without CI
In the current study, we found evident differences in gut microbiota and fecal metabolites between CN and CI patients, but not between BVN and BVI patients. Firmicutes and Bacteroidetes belong to bacterial phyla, lower F/B ratio indicates intestinal homeostasis dysregulation[33]. We found that the F/B ratio was comparable between ALS patients and healthy controls; however, it was significantly lower in the CI group than that in the CN group. Previously, various gut microbiota studies have been conducted in ALS patients (reviewed in Table 3). In the current study, the changes of gut microbiota between ALS patients and healthy controls have some similarities with those found in previous studies, with a similar α diversity but a different β diversity between ALS patients and healthy controls[34–38]. In untargeted metabolic mapping, we found disturbed profiles of lipids and lipid-like molecules in CI patients, including CA CDCA, 1-tetradecanoyl-2-(9Z-tetradecenoyl)-glycero-3-phosphoserine, and 1,3-dihydroxypropan-2-yl (9Z)-tetradec-9-enoate. Particularly, three of six mapped metabolic pathways in KEGG were related to bile acid metabolism, suggesting a potential link between bile acids and cognition in ALS patients. A recent study was conducted with both metagenomics and metabolomics analyses in 10 ALS patients and 10 healthy controls, providing limited conclusive findings[36]. In spite of previous studies about the ALS gut microbiota, to the best of our knowledge, we are the first to reveal the differences in gut microbiota and fecal metabolites between patients with and without cognitive decline using a multi-omics approach.
**Table 3**
| References | Subjects (N) | Subjects (N).1 | Subjects (N).2 | Study methods | Main results |
| --- | --- | --- | --- | --- | --- |
| References | ALS | HC | NDC | Study methods | Main results |
| Fang et al, 2016[34] | 6 | 5 | 0 | High-throughput sequencing (V3–V4 region) | 1. The β-diversity was different between ALS and HC. 2. Dorea genus increased in ALS, and Oscillibacter, Anaerostripes, Lachnospiraceae genus decreased in ALS. |
| Rowin et al, 2017[35] | 5 | 96 | 0 | 1. PCR for specific bacterial species 2. Quantitative determination of short chain fatty acids (SCFAs) in fecal samples using mass spectrometry | 1. The diversity of the microbiome decreased in ALS (not clearly stated whether α-diversity or β-diversity).2. A low Firmicutes/Bacteroidetes (F/B) ratio was found in 3 ALS patients. 3. The level of SCFAs decreased in 1 ALS patient. |
| Brenner et al, 2017[21] | 25 | 32 | 0 | 454-pyrosequencing (V3–V6 region) | 1. Both α-diversity and β-diversity were comparable between ALS and HC. 2. The total number of operational taxonomic units increased in ALS. 3. The relative abundance of the uncultured Ruminococcaceae genus was different between ALS and HC. |
| Zhai et al, 2019[25] | 8 | 8 | 0 | 1. High-throughput sequencing (V4–V5 region)2. Determination of endotoxin, SCFAs, NO2-N/NO3-N, and g-aminobutyric acid in fecal samples using enzyme-linked immunosorbent assay (ELISA) kits | 1. The F/B ratio and Methanobrevibacter genus showed an increased tendency in ALS.2. Faecalibacterium and Bacteroides genus (beneficial micro-organisms) decreased in ALS.3. No significant difference in levels of endotoxin, SCFAs, NO2-N/NO3-N, and g-aminobutyric acid was found between ALS and HC. |
| Blacher et al, 2019[20] | 37 | 29 | 0 | Shotgun sequencing | 1. The microbiome composition was different between ALS and HC.2. Only a marginally significant difference in the abundances of specific bacterial species was found between ALS and HC.3. ALS microbiomes decreased significantly in the global bacterial gene content related to nicotinamide metabolism. |
| Zeng et al, 2020[36] | 20 | 20 | 0 | 1. High-throughput sequencing (V4 region) 2. Shotgun sequencing (10 ALS and 10 HC) 3. Untargeted metabolome using liquid chromatography mass spectrometry (LC-MS) (10 ALS and 10 HC) | 1. The α-diversity (Shannon index) was different between ALS and HC.2. Bacteroidetes phylum increased in ALS.3. Firmicutes phylum, Kineothrix, Parabacteroides, Odoribacter, Sporobacter, Eisenbergiella, Mannheimia, Anaerotruncus, and unclassified Porphyromonadaceae genus decreased in ALS. 4. Enterococcus columbae positively correlated with 2-(1-ethoxyethoxy) propanoic acid and 3,7-dihydroxy-12-oxocholanoic acid. |
| Ngo et al, 2020[37] | 49 | 51 | 0 | High-throughput sequencing (V6–V8 region) | 1. The fecal microbiome was not significantly different between ALS and HC. 2. A greater risk for earlier death was reported in ALS patients with increased richness and diversity of the microbiome, and in those with a higher F/B ratio. |
| Nicholson et al, 2020[38] | 66 | 61 | 12 | Shotgun sequencing | 1. Comparable α-diversity and β-diversity between ALS and HC2. The relative abundance of the dominant butyrate-producing microbial members decreased in ALS. |
| Di Gioia et al, 2021[19] | 43 | 44 | 0 | High-throughput sequencing (V3–V4 region) | 1. The α-diversity (Chao1 index) and β-diversity were different between ALS and HC. 2. Microbial members of the Cyanobacteria phylum increased in ALS.3. Microbial members of Clostridiaceae family decreased in ALS. |
| Niccolai, et al, 2021[22] | 19 | 9 | 0 | 1. High-throughput sequencing (V3–V4 region)2. Determination of 30 kinds of cytokines (test kits) and SCFAs (gas chromatography and mass spectrometry) | 1. The F/B ratio decreased in ALS. 2. The relative abundance of butyrate-producing microbial members decreased in ALS.3. Interleukin-2 (IL-2) and IL-1β increased in ALS. 4. IL-21 decreased in patients with a fast progression. |
| Note: only the results regarding the fecal microbiome and metabolome were summarized in this table. ALS: amyotrophic lateral sclerosis; HC: healthy control; NDC: neurodegenerative control; PCR: polymerase chain reaction; SCFAs: short chain fatty acids; F/B: Firmicutes/Bacteroidetes; IL: interleukin. | Note: only the results regarding the fecal microbiome and metabolome were summarized in this table. ALS: amyotrophic lateral sclerosis; HC: healthy control; NDC: neurodegenerative control; PCR: polymerase chain reaction; SCFAs: short chain fatty acids; F/B: Firmicutes/Bacteroidetes; IL: interleukin. | Note: only the results regarding the fecal microbiome and metabolome were summarized in this table. ALS: amyotrophic lateral sclerosis; HC: healthy control; NDC: neurodegenerative control; PCR: polymerase chain reaction; SCFAs: short chain fatty acids; F/B: Firmicutes/Bacteroidetes; IL: interleukin. | Note: only the results regarding the fecal microbiome and metabolome were summarized in this table. ALS: amyotrophic lateral sclerosis; HC: healthy control; NDC: neurodegenerative control; PCR: polymerase chain reaction; SCFAs: short chain fatty acids; F/B: Firmicutes/Bacteroidetes; IL: interleukin. | Note: only the results regarding the fecal microbiome and metabolome were summarized in this table. ALS: amyotrophic lateral sclerosis; HC: healthy control; NDC: neurodegenerative control; PCR: polymerase chain reaction; SCFAs: short chain fatty acids; F/B: Firmicutes/Bacteroidetes; IL: interleukin. | Note: only the results regarding the fecal microbiome and metabolome were summarized in this table. ALS: amyotrophic lateral sclerosis; HC: healthy control; NDC: neurodegenerative control; PCR: polymerase chain reaction; SCFAs: short chain fatty acids; F/B: Firmicutes/Bacteroidetes; IL: interleukin. |
## Gut microbiota correlated with cognition by altering fecal bile acids profile in ALS
Based on our findings of altered gut microbiota and in bile acids-related metabolic pathways between CN and CI, we further performed a quantification of fecal bile acids. Decreases in CA, CDCA, and conjugated forms with glycine or taurine and increases in several secondary bile acids were found in the CI group, compared with the CN group. The higher ratios of CA/DCA (CI: 0.63, CN: 0.10) and CDCA/LCA (CI: 0.48, CN: 0.07) demonstrated a higher efficiency of conversion from CA to DCA, CDCA to LCA in the CI group than in the CN group. In humans, PBAs (mainly CA and CDCA) are synthesized in the liver and excreted to the small intestine[30]. As PBAs move from the small intestine to the colon, they are converted to SBAs by the biotransformation of the resident microbial community, and the key step is the 7α-dehydroxylation reaction[32]. In the colon, nearly $100\%$ of CA and CDCA are converted to DCA and LCA, respectively. However, only a few known bacteria, all from the Ruminococcaceae family and Clostridium genus, perform the 7α-dehydroxylation[39]. Interestingly, 11 genera belonging to the family Ruminococcaceae were identified in the current study, and 10 genera increased in the CI group. According to these findings, we propose that the higher efficiency of conversion from PBAs to SBAs in the CI group probably results from alterations in gut microbial communities. However, since 16S sequencing is limited in the annotation of bacteria at the species level[40], we failed to further link the bile acids metabolism to specific bacterial strains. More in-depth basic and clinical studies, such as metagenomic studies and animal intervention studies, need to be further conducted to verify this hypothesis.
In the current study, ALS patients with CI presented with decreased fecal CA and CDCA, leading us to think about whether the impaired cognitive function is potentially linked to altered bile acid metabolic profiles. Bile acids are essential products of cholesterol metabolism, and in addition to playing a key role in lipid metabolism and absorption, recent studies suggest that the bile acid metabolism is associated with cognitive function[41]. In bile duct ligation mice, oral administration of obeticholic acid normalized memory function, prevented hippocampal network deficits, and reversed neuronal senescence by activating the farnesoid X receptor[42]. Interestingly, CDCA is an endogenous activator of the farnesoid X receptor[43], but whether CDCA can protect cognition by activating the farnesoid X receptor needs to be further explored. It was also reported that oral CDCA supplementation ameliorates neurotoxicity and cognitive deterioration via enhancing insulin signaling in AD model rats[44]. In clinical studies, AD patients exhibited significantly lower serum CA and higher DCA levels, compared with healthy controls. The ratio of DCA/CA was significantly correlated with the severity of cognitive decline[45]. Elevated DCA was also found in diabetes patients with cognitive impairment[46]. Both CA and CDCA could diffuse across the blood-brain barrier, generating aligned concentration in the brain and peripheral tissue[47]. Therefore, oral administration of either CA or CDCA could potentially be used as therapeutics for improving cognition. Clinical efficacy of such an approach for slowing cognitive decline in ALS patients could be tested in clinical trials in the future.
## Conclusions
In the current study, we found altered gut microbial communities and bile acid metabolism in ALS patients, highlighting a possible role in the pathogenesis of cognitive decline in ALS patients. To our knowledge, we are the first to reveal connections between gut microbiota and CI in ALS patients. Based on the findings of our multi-omics approach, we propose that novel therapeutics could target bile acid metabolites with the aim of reducing CI in ALS patients. The study could be further strengthened by investigating bile acid metabolites in serum and cerebrospinal fluid, and the current findings need to be verified in studies with a larger sample size, particularly in longitudinal studies that may observe the dynamic changes of gut microbiota and metabolomics overtime.
## References
1. **Amyotrophic lateral sclerosis**. *Lancet* (2011.0) **377** 942-955. DOI: 10.1016/S0140-6736(10)61156-7
2. **Microbiota impacts on chronic inflammation and metabolic syndrome - related cognitive dysfunction**. *Rev Endocr Metab Disord* (2019.0) **20** 473-480. DOI: 10.1007/s11154-019-09537-5
3. **ALS-specific cognitive and behavior changes associated with advancing disease stage in ALS**. *Neurology* (2018.0) **91** e1370-e1380. DOI: 10.1212/WNL.0000000000006317
4. **Neurobehavioral dysfunction in ALS has a negative effect on outcome and use of PEG and NIV**. *Neurology* (2012.0) **78** 1085-1089. DOI: 10.1212/WNL.0b013e31824e8f53
5. **Executive dysfunction is a negative prognostic indicator in patients with ALS without dementia**. *Neurology* (2011.0) **76** 1263-1269. DOI: 10.1212/WNL.0b013e318214359f
6. **The edinburgh cognitive and behavioural ALS screen in a Chinese amyotrophic lateral sclerosis population**. *PLoS One* (2016.0) **11** e0155496. DOI: 10.1371/journal.pone.0155496
7. **Screening for cognitive and behavioural impairment in amyotrophic lateral sclerosis: frequency of abnormality and effect on survival**. *J Neurol Sci* (2017.0) **376** 16-23. DOI: 10.1016/j.jns.2017.02.061
8. **Changes in cognition and behaviour in amyotrophic lateral sclerosis: nature of impairment and implications for assessment**. *Lancet Neurol* (2013.0) **12** 368-380. DOI: 10.1016/S1474-4422(13)70026-7
9. 9Balendra R, Isaacs AMC9orf72-mediated ALS and FTD: multiple pathways to diseaseNat Rev Neurol201814954455810.1038/s41582-018-0047-230120348. *Nat Rev Neurol* (2018.0) **14** 544-558. DOI: 10.1038/s41582-018-0047-2
10. **Genetic epidemiology of amyotrophic lateral sclerosis: a systematic review and meta-analysis**. *J Neurol, Neurosurg, Psychiatry* (2017.0) **88** 540-549. DOI: 10.1136/jnnp-2016-315018
11. 11He J, Tang L, Benyamin B, et alC9orf72 hexanucleotide repeat expansions in Chinese sporadic amyotrophic lateral sclerosisNeurobiol Aging20153692660.e12660.e810.1016/j.neurobiolaging.2015.06.002. *Neurobiol Aging* (2015.0) **36** 2660.e1-2660.e8. DOI: 10.1016/j.neurobiolaging.2015.06.002
12. **Plasma uric acid helps predict cognitive impairment in patients with amyotrophic lateral sclerosis**. *Front Neurol* (2021.0) **12** 789840. DOI: 10.3389/fneur.2021.789840
13. **Sodium oligomannate therapeutically remodels gut microbiota and suppresses gut bacterial amino acids-shaped neuroinflammation to inhibit Alzheimer's disease progression**. *Cell Res* (2019.0) **29** 787-803. DOI: 10.1038/s41422-019-0216-x
14. **Regulation of neurotransmitters by the gut microbiota and effects on cognition in neurological disorders**. *Nutrients* (2021.0) **13** 2099. DOI: 10.3390/nu13062099
15. **Neuroinflammatory remodeling of the anterior cingulate cortex as a key driver of mood disorders in gastrointestinal disease and disorders**. *Neurosci Biobehav Rev* (2022.0) **133** 104497. DOI: 10.1016/j.neubiorev.2021.12.020
16. **Gut microbiota and their neuroinflammatory implications in Alzheimer's disease**. *Nutrients* (2018.0) **10** 1765. DOI: 10.3390/nu10111765
17. **Gut microbiota, probiotic interventions, and cognitive function in the elderly: a review of current knowledge**. *Nutrients* (2021.0) **13** 2514. DOI: 10.3390/nu13082514
18. **Gut microbiota in ALS: possible role in pathogenesis?**. *Expert Rev Neurother* (2019.0) **19** 785-805. DOI: 10.1080/14737175.2019.1623026
19. **A prospective longitudinal study on the microbiota composition in amyotrophic lateral sclerosis**. *BMC Med* (2020.0) **18** 153. DOI: 10.1186/s12916-020-01607-9
20. **Potential roles of gut microbiome and metabolites in modulating ALS in mice**. *Nature* (2019.0) **572** 474-480. DOI: 10.1038/s41586-019-1443-5
21. **The fecal microbiome of ALS patients**. *Neurobiol Aging* (2018.0) **61** 132-137. DOI: 10.1016/j.neurobiolaging.2017.09.023
22. 22Niccolai E, Di Pilato V, Nannini G, et al. The Gut Microbiota-Immunity Axis in ALS: A Role in Deciphering Disease Heterogeneity?[J]. Biomedicines, 2021, 9(7).
23. **El Escorial revisited: revised criteria for the diagnosis of amyotrophic lateral sclerosis**. *Amyotroph Lateral Scler Other Motor Neuron Disord* (2000.0) **1** 293-299. DOI: 10.1080/146608200300079536
24. **The ALSFRS-R: a revised ALS functional rating scale that incorporates assessments of respiratory function. BDNF ALS Study Group (Phase Ⅲ)**. *J Neurol Sci* (1999.0) **169** 13-21. DOI: 10.1016/s0022-510x(99)00210-5
25. **Intestinal microbiota composition in patients with amyotrophic lateral sclerosis: establishment of bacterial and archaeal communities analyses**. *Chin Med J* (2019.0) **132** 1815-1822. DOI: 10.1097/CM9.0000000000000351
26. **Comparison of mothur and QIIME for the analysis of rumen microbiota composition based on 16S rRNA amplicon sequences**. *Front Microbiol* (2018.0) **9** 3010. DOI: 10.3389/fmicb.2018.03010
27. **Analysing microbial community composition through amplicon sequencing: from sampling to hypothesis testing**. *Front Microbiol* (2017.0) **8** 1561. DOI: 10.3389/fmicb.2017.01561
28. **Persistence of antibiotic resistance genes from river water to tap water in the Yangtze River Delta**. *Sci Total Environ* (2020.0) **742** 140592. DOI: 10.1016/j.scitotenv.2020.140592
29. **Metagenomic biomarker discovery and explanation**. *Genome Biol* (2011.0) **12** R60. DOI: 10.1186/gb-2011-12-6-r60
30. **Bile salt biotransformations by human intestinal bacteria**. *J Lipid Res* (2006.0) **47** 241-259. DOI: 10.1194/jlr.R500013-JLR200
31. **Increase in fecal primary bile acids and dysbiosis in patients with diarrhea-predominant irritable bowel syndrome**. *Neurogastroenterol Motil* (2012.0) **24** 513-520. DOI: 10.1111/j.1365-2982.2012.01893.x
32. **Interaction of gut microbiota with bile acid metabolism and its influence on disease states**. *Appl Microbiol Biotechnol* (2017.0) **101** 47-64. DOI: 10.1007/s00253-016-8006-6
33. **What is the healthy gut microbiota composition? A changing ecosystem across age, environment, diet, and diseases**. *Microorganisms* (2019.0) **7** 14. DOI: 10.3390/microorganisms7010014
34. **Evaluation of the microbial diversity in amyotrophic lateral sclerosis using high-throughput sequencing**. *Front Microbiol* (2016.0) **7** 1479. DOI: 10.3389/fmicb.2016.01479
35. **Gut inflammation and dysbiosis in human motor neuron disease**. *Physiol Rep* (2017.0) **5** e13443. DOI: 10.14814/phy2.13443
36. **The alteration of gut microbiome and metabolism in amyotrophic lateral sclerosis patients**. *Sci Rep* (2020.0) **10** 12998. DOI: 10.1038/s41598-020-69845-8
37. **Progression and survival of patients with motor neuron disease relative to their fecal microbiota**. *Amyotroph Lateral Scler Frontotemporal Degener* (2020.0) **21** 549-562. DOI: 10.1080/21678421.2020.1772825
38. **The human gut microbiota in people with amyotrophic lateral sclerosis**. *Amyotroph Lateral Scler Frontotemporal Degener* (2021.0) **22** 186-194. DOI: 10.1080/21678421.2020.1828475
39. **Dysbiosis-induced secondary bile acid deficiency promotes intestinal inflammation**. *Cell Host Microbe* (2020.0) **27** 659-670.e5. DOI: 10.1016/j.chom.2020.01.021
40. **Evaluation of 16S rRNA gene sequencing for species and strain-level microbiome analysis**. *Nat Commun* (2019.0) **10** 5029. DOI: 10.1038/s41467-019-13036-1
41. **Intestinal crosstalk between bile acids and microbiota and its impact on host metabolism**. *Cell Metab* (2016.0) **24** 41-50. DOI: 10.1016/j.cmet.2016.05.005
42. **Anti-cholestatic therapy with obeticholic acid improves short-term memory in bile duct-ligated mice**. *Am J Pathol* (2022.0) **193** 11-26. DOI: 10.1016/j.ajpath.2022.09.005
43. **Bile acid nuclear receptor FXR and digestive system diseases**. *Acta Pharm Sin B* (2015.0) **5** 135-144. DOI: 10.1016/j.apsb.2015.01.004
44. **Chenodeoxycholic acid ameliorates AlCl**. *Molecules* (2019.0) **24** 1992. DOI: 10.3390/molecules24101992
45. **Altered bile acid profile associates with cognitive impairment in Alzheimer's disease-An emerging role for gut microbiome**. *Alzheimers Dement* (2019.0) **15** 76-92. DOI: 10.1016/j.jalz.2018.07.217
46. **Metabolic footprint of diabetes: a multiplatform metabolomics study in an epidemiological setting**. *PLoS One* (2010.0) **5** e13953. DOI: 10.1371/journal.pone.0013953
47. **Bile acid signaling pathways from the enterohepatic circulation to the central nervous system**. *Front Neurosci* (2017.0) **11** 617. DOI: 10.3389/fnins.2017.00617
|
---
title: Capsaicin protects against septic acute liver injury by attenuation of apoptosis
and mitochondrial dysfunction
authors:
- Atefeh Ghorbanpour
- Sepide Salari
- Tourandokht Baluchnejadmojarad
- Mehrdad Roghani
journal: Heliyon
year: 2023
pmcid: PMC10018474
doi: 10.1016/j.heliyon.2023.e14205
license: CC BY 4.0
---
# Capsaicin protects against septic acute liver injury by attenuation of apoptosis and mitochondrial dysfunction
## Abstract
Capsaicin is the main pungent bioactive constituent in red chili with promising therapeutic properties due to its anti-oxidative and anti-inflammatory effects. No evidence exists on the beneficial effect of capsaicin on apoptosis and mitochondrial function in acute liver injury (ALI) under septic conditions. For inducing septic ALI, lipopolysaccharide (LPS, 50 μg/kg) and d-galactose (D-Gal, 400 mg/kg) was intraperitoneally injected and capsaicin was given orally at 5 or 20 mg/kg. Functional markers of liver function and mitochondrial dysfunction were determined as well as hepatic assessment of apoptotic, oxidative, and inflammatory factors. Capsaicin at the higher dose appropriately decreased serum level of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in addition to reducing hepatic level of malondialdehyde (MDA), reactive oxygen species (ROS), nitrite, NF-kB, TLR4, IL-1β, TNF-α, caspase 3, DNA fragmentation and boosting sirtuin 1, Nrf2, superoxide dismutase (SOD) activity, and heme oxygenase (HO-1). These beneficial effects of capsaicin were associated with reversal and/or improvement of gene expression for pro-apoptotic Bax, anti-apoptotic Bcl2, mitochondrial and metabolic regulators PGC-1α, sirtuin 1, and AMPK, and inflammation-associated factors. Additionally, capsaicin exerted a hepatoprotective effect, as revealed by its reduction of liver histopathological changes. These findings evidently indicate hepatoprotective property of capsaicin under septic conditions that can be attributed to its down-regulation of oxidative and inflammatory processes besides its potential to attenuate mitochondrial dysfunction and apoptosis.
## Graphical abstract
Image 1
## Introduction
Acute liver injury (ALI) and ensuing hepatic failure is recognized a severe health complication worldwide with high incidence of morbidity and mortality [1]. ALI occurs following exposing to numerous damaging agents such as alcohol, lipopolysaccharide (LPS), carbon tetrachloride (CCL4), and acetaminophen [2,3]. The underlying mechanisms proposed for the development of ALI include multiple factors comprising excessive generation of reactive oxygen species [4], unmanaged inflammation [5,6] and dysregulated apoptotic process [7]. Exposure of d-galactosamine (D-Gal)-sensitized rodents to LPS causes critical liver injury that is usually used for modeling ALI [8,9]. LPS challenge is associated with derangement of mitochondrial biogenesis and oxidative metabolism, leading to mitochondrial dysfunction and induction of liver injury in septic conditions [10].
Natural products with potential to ameliorate oxidative stress and inflammation may be of potential therapeutic benefit to control and prevent ALI [11]. Capsaicin (8-methyl-N-vanillyl-trans-6-nonenamide) is the main pungent bioactive constituent in red chili in the genus Capsicum that has been suggested as a promising therapeutic agent [12]. It possesses some curative effects for the treatment of arthritis, diabetic neuropathy, gastric lesions, and cardiac excitability [13,14]. Capsaicin has also shown potent anti-oxidative effects which is independent of TRPV1 receptor activation [15]. Additionally, capsaicin could exert a hepatoprotective effect against concanavalin A-induced hepatic damage through ameliorating oxidative and inflammatory events [16]. Moreover, protective effect of capsaicin against 2,3,7,8-tetrachlorodibenzo-p-dioxini-induced oxidative damage of heart, liver, and kidney tissues has been shown [17]. Besides, dietary capsaicin is capable to alleviate hepatic oxidative stress and apoptosis in rats on high fat diet through balancing oxidant-antioxidant status [18]. Meanwhile, capsaicin can alleviate LPS-induced inflammatory cytokine generation including IL-1β, IL-6 and TNF-α, partly mediated though inhibition of NF-κB [19] and is capable to exert protective effect in the liver and lung tissues against LPS injury that is mediated via appropriate modulation of oxidative status and alleviation of inflammation [20]. Capsaicin can also prevent acute kidney injury through attenuation of mitochondrial dysfunction linked to Nrf2 activation [21]. There is currently no research evidence on the beneficial effect of capsaicin on apoptosis and mitochondrial biogenesis and function in septic models of ALI. Hence, this study was designed and conducted to assess hepatoprotective effect of capsaicin in LPS/D-Gal model of ALI with emphasis on its beneficial effect on attenuation of mitochondrial dysfunction and apoptosis.
## Animals
Male mice (C57BL/6 strain; $$n = 40$$; obtained from Razi Institute of Karaj, Iran) were kept for 1 week for being adapted to animal house conditions. All animals were kept at stipulated conditions (22–23 °C, 42–$48\%$ humidity, 12-h lighting photoperiod, and with free access to food and water). All procedures conducted on animals were according to NIH protocols that were approved by NIMAD Institute (IR.NIMAD.REC.1397.163).
## Experimental procedures
Animals were randomly divided into 4 testing groups using random number table as follows: control, LPS/D-Gal, and LPS/D-Gal groups receiving capsaicin (Cat #M2028, >$95\%$, SigmaAldrich, USA) at doses of 5 or 20 mg/kg. Mice in LPS/D-Gal group had intraperitoneal injection of a combination of LPS (50 μg/kg) isolated from E. coli (Cat #L2630, >$95\%$, SigmaAldrich, USA) and D-Gal hydrochloride (Cat #G0500, >$99\%$, SigmaAldrich, Germany) at a dose of 400 mg/kg (dissolved in normal saline) [22]. Treatment groups received capsaicin (p.o. through the gavage needle) daily for 3 days till 1 h before LPS/D-Gal injection. Dose of capsaicin was chosen from its efficacy in amelioration of alcohol-induced ALI [23]. After 6 h, mice were deeply anesthetized with ketamine (150 mg/kg) and after drawing blood samples through the heart were killed and their liver samples were collected for biochemical or histological assessment.
## Measurement of serum activity of ALT and AST
Blood samples were drawn from the heart under deep anesthesia with ketamine-HCl. The blood samples were kept at room temperature for 30 min and were then centrifuged at 3000×g for 10 min to separate serum samples. Serum activity of ALT (Cat # 1022003, Pars Azmun Co., Tehran, Iran) and AST (Cat # 97203232, Pars Azmun Co., Tehran, Iran) was measured per provided instructions of kits.
## Hepatic evaluation of oxidative stress-associated parameters
After preparing liver homogenate samples using 150 mM Tris-HCl buffer (pH 7.4) and centrifuging them, the obtained supernatant was used for measurement of oxidative stress factors. Levels of MDA, which is known as an index of lipid peroxidation, were measured using MDA assay reagent containing trichloroacetic acid (Cat #T4885, >$99\%$, SigmaAldrich, USA) and 2-thiobarbituric acid (Cat #T5500, >$98\%$, SigmaAldrich, USA). Nitrite level as an end-product of nitric oxide (NO) catabolism was evaluated using Griess protocol with its reagent containing sulfanilamide (Cat #S9251, >$98\%$, SigmaAldrich, USA) and N-(1-Naphthyl)ethylenediamine dihydrochloride (Cat # 33461, >$98\%$, SigmaAldrich, USA) in an acidic medium [24]. Activity of SOD was obtained with the help of its specific kit (Cat # 706002, Cayman Chemical, USA). Catalase activity was determined in accordance to Claiborne's method in which disappearance of peroxide was followed spectrophotometrically at 240 nm using potassium phosphate buffer (pH 7) and 0.059 M hydrogen peroxide [25,26]. Bradford method was used for measurement of total protein [27] using its specific kit (Cat # KBRF96, Kiazist, Hamadan, Iran).
## Determination of hepatic IL-1β, TNF-α, sirtuin 1, HO-1, IL-10, NF-kB, Nrf2, and TLR4
Liver tissue levels of these factors were determined using sandwich Elisa protocol with antibodies or kits as follows: IL-1β (Cat # RAB0274, SigmaAldrich, USA), TNF-α (Cat # RAB0477, SigmaAldrich, USA), sirtuin 1 (Cat # sc-74465, Santa Cruz Biotechnology, Inc., USA), HO-1 (Cat # sc-390991, Santa Cruz Biotechnology, Inc., USA), IL-10 (Cat # sc-365858, Santa Cruz Biotechnology, Inc., USA), NF-kB (Cat # sc-8008, Santa Cruz Biotechnology, Inc., USA), Nrf2 (Cat # sc-722, Santa Cruz Biotechnology, Inc., USA), and TLR4 (Cat # SAB5700684, SigmaAldrich, USA).
## Real-time qPCR
Total RNA was extracted from the liver tissue with Kiazol reagent (Cat # KZOL50, Kiazist, Iran) and reverse transcribed into cDNA with the SYBR Green qPCR Master Mix (Cat # MM2042, Sinaclon, Iran). Forward and reverse primers for the related genes were designed using Primer Express software (Applied Biosystems, USA) and sequences were as follows: TLR4, CAAGAACATAGATCTGAGCTTCAACCC (forward), GCTGTCCAATAGGGAAGCTTTCTAGAG (Reverse); Bcl2, GACTGAGTACCTGAACCGGCATC (forward), CTGAGCAGCGTCTTCAGAGACA (reverse); Bax, CGAATTGGCGATGAACTGGA (forward), CAAACATG TCAGCTGCCACAC (reverse); NF-kB p65, GAGGCACGAGGCTCCTTTTCT (forward), GTAGCTGCATGGAGACTCGAACA (reverse); PGC-1α, TATGGAGTGACATAGAGTGTGCT (forward), GTCGCTACACCACTTCAATCC (reverse); Nrf2, TTGGCAGAGACATTCCCAT (forward), GCTGCCACCGTCACTGGG (reverse), HO-1, CACGCATATACCCGCTACCT (forward), CCAGAGTGTTCATTCGAGCA (reverse); AMPK, GTCAAAGCCGACCCAATGATA (forward), CGTACACGCAAATAATAGGGGTT (reverse); Sirtuin 1, TGATTGGCACCGATCCTCG (forward), CCACAGCGTCATATCATCCAG (reverse), beta actin, ACTGCCGCATCCTCTTCCT (forward), TCAACGTCACACTTCATGATGGA (reverse). Thermal cycling conditions were denaturation at 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Relative differences were expressed using cycle time (Ct) values and data were expressed as a fold change relative to the control and according to 2^-ΔΔCt.
## Assessment of apoptosis
For estimation of apoptotic process, liver level of DNA fragmentation (using Cell Death Detection ELISA Plus kit (Cat # 11774425001, Roche, USA) and caspase 3 activity [28] were determined.
## Estimation of MMP
MMP as an indicator of mitochondrial integrity was determined according to a previous study with Rhodamine 123 (Cat #R8004, SigmaAldrich, USA) as the detecting probe [29]. In this test, supernatant was re-centrifuged at 10,000 rpm for 15 min and 20 μl of rhodamine 123 solution (10 μmol/L) and 180 μL of PBS was added to the formed precipitate. After stirring, it was transferred to 96-cell microplate and incubated at 37 °C for 30 min. Finally, MMP was determined after excitation at 488 nm and emission at 525 nm.
## Histopathology of the liver
Liver sections (a thickness of 5 μm) were stained using H&E routine protocol. Histological changes were assessed in randomly selected microscopic fields at a magnification of 200. Severity of hepatic damage was graded according to a four-point scale from 0 to 3, according to no evidence of damage, moderate to severe damage with widespread nuclear pyknosis, loss of intercellular borders and severe necrosis with hemorrhage and neutrophil infiltration [30]. For TLR4 immunohistochemistry, liver sections were incubated with primary TLR4 antibody (TLR4 (Cat # SAB5700684, SigmaAldrich, USA) and then with secondary HRP-conjugated antibody (Cat # SAB3700852, SigmaAldrich, USA). Every analysis was repeated twice and its data was averaged.
## Statistical analysis
All data were expressed as mean ± SEM and statistically analyzed by one-way ANOVA and Tukey tests after verification of parametric distribution of data by Shapiro-Wilk test. Statistical significance was accepted at P less than 0.05.
## The effect of capsaicin on serum activity of ALT and AST
The enzymes ALT and AST are two reliable indicators for estimation of liver function and their notable elevation is the biochemical basis for diagnosing liver damage [31]. The beneficial effect of capsaicin on serum activity of ALT and AST are shown in Fig. 1. Statistical analysis with one-way ANOVA indicated significant inter-group differences regarding ALT (F [3,24] = 22.57, $p \leq 0.001$) (Fig. 1A) and AST (F [3,24] = 19.60, $p \leq 0.001$) (Fig. 1B). LPS/D-Gal group has a significantly elevated level of the serum activity of these enzymes relative to the control ($p \leq 0.001$). In contrast, capsaicin pretreatment at a dose of 20 mg/kg significantly decreased serum activities of ALT ($p \leq 0.01$) and AST ($p \leq 0.01$) as compared to the LPS/D-Gal-challenged group. Besides, capsaicin at the lower dose of 5 mg/kg did not exert such beneficial effects at a significant level. Fig. 1Serum indicators of liver function comprising ALT (A) and AST (B). Data are shown in means ± SEM. Experiments were done in duplicate. c, cc, and ccc indicate p values lower than 0.05, 0.01, and 0.001, respectively (relative to the control); d and dd indicate p values less than 05 and 0.01, respectively (relative to the LPS/D-Gal group). $$n = 7$$ per group. Fig. 1
## The effect of capsaicin on oxidative stress-related factors
LPS/D-Gal challenge increases hepatic oxidative stress that is confirmed by elevated levels of MDA (as the final product of lipid peroxidation process) and ROS and lower levels of some antioxidants [30]. Hence, we explored the effect of capsaicin on the liver levels of some oxidative stress-related indices in this model of ALI. Statistical analysis of oxidative stress data showed significant differences between the group for MDA (F [3,24] = 13.49, $p \leq 0.001$), nitrite (F [3,24] = 16.31, $p \leq 0.001$), catalase (F [3,24] = 3.68, $p \leq 0.05$), SOD (F [3,24] = 9.18, $p \leq 0.001$). Mice challenged with LPS/D-Gal had significantly elevated levels of MDA (Fig. 2A) ($p \leq 0.01$) and nitrite (Fig. 2B) ($p \leq 0.001$) as compared to relevant data of the control group. Our findings also showed significantly lower levels of catalase activity (Fig. 2C) ($p \leq 0.05$) and SOD activity (Fig. 2D) ($p \leq 0.01$) when compared with comparable findings of the control group. On the contrary, capsaicin pretreatment of LPS/D-Gal group at 20 mg/kg, but not at 5 mg/kg, significantly and suitably lowered level of MDA ($p \leq 0.01$) and nitrite ($p \leq 0.01$) and improved level of SOD ($p \leq 0.01$) and with no significant improvement of the enzyme catalase. Fig. 2Hepatic levels of oxidative stress-associated indices consisting of MDA (A), nitrite (B), catalase activity (C), SOD activity (D), HO-1 (E), and sirtuin 1 (F). c, cc, and ccc indicate p values lower than 0.05, 0.01, and 0.001, respectively (relative to the control); d and dd indicate p values less than 0.05 and 0.01, respectively (relative to the LPS/D-Gal group). $$n = 7$$ per group. Experiments were done in duplicate. All data are presented as mean ± SEM.Fig. 2 Part of enhancement of antioxidant elements following administration of natural products in models of disorders is mediated through an enhancement of nuclear translocation of Nrf2 [32]. Thus, we also measured nuclear level of Nrf2 (Fig. 3A) and its mRNA gene expression (Table 1). Our results showed no significant change of Nrf2 level or its gene expression in LPS/D-Gal group. On the contrary, there was a significant elevation of Nrf2 (F [3,24] = 5.06, $p \leq 0.01$) ($p \leq 0.05$) and its gene expression ($p \leq 0.05$) due to administration of capsaicin at a dose of 20 mg/kg to LPS/D-Gal group. Nrf2 cascade is associated with HO-1 with critical function in prevention of oxidative stress and inflammation. In this study, HO-1 level (Fig. 2E) (F [3,24] = 11.82, $p \leq 0.01$) ($p \leq 0.01$) and its gene expression (Table 2) ($p \leq 0.01$) was significantly lower in LPS/D-Gal-injured group versus the control group. Conversely, capsaicin at 20 mg/kg significantly increased HO-1 level ($p \leq 0.05$) and its gene expression ($p \leq 0.05$) versus the vehicle-treated injured group. Fig. 3Hepatic levels of Nrf2 (A), NF-kB (B), TLR4 (C), IL-1β (D), TNFα (E), and IL-10 (F). c, cc, and ccc indicate p values lower than 0.05, 0.01, and 0.001, respectively (relative to the control); d and dd indicate p values less than 05 and 0.01, respectively (relative to the LPS/D-Gal group). All data are presented as mean ± SEM. Experiments were done in duplicate. $$n = 7$$ per group. Fig. 3Table 1The effect of capsaicin on mitochondrial and apoptotic indices in the liver tissue following LPS/D-Gal-induced hepatotoxicity. Table 1ControlLPS/D-GalLPS/D-Gal + Capsaicin5LPS/D-Gal + Capsaicin20ApoptosisDNA fragmentation0.37 ± 0.050.89 ± 0.08ccc0.82 ± 0.09cc0.57 ± 0.08dCaspase 3 activity0.38 ± 0.040.92 ± 0.07ccc0.94 ± 0.08ccc0.63 ± 0.05cdMitochondrial dysfunctionMMP (AFU)100 ± 0.0053.8 ± 4.757.1 ± 5.474.2 ± 5.1dData are shown as means ± SE. c, cc, and ccc suggest $p \leq 0.05$, 0.01, and 0.001, respectively (as compared to the control); d suggest $p \leq 05$ (as compared to the LPS/D-Gal group). AFU indicates arbitrary fluorescence unit. $$n = 7$$ per experimental group. Experiments for DNA fragmentation and caspase 3 were done in duplicate and for MMP were conducted once. Table 2The effect of capsaicin on gene expression of ΑΜPK, Bax, Bcl2, PGC-1α, sirtuin 1, HO-1, NF-κB, TLR4, and Nrf2 in the liver tissue following LPS/D-Gal-induced hepatotoxicity. Table 2ControlLPS/D-GalLPS/D-Gal + Capsaicin5LPS/D-Gal + Capsaicin20Relative mRNA expressionBax0.27 ± 0.030.98 ± 0.12ccc0.75 ± 0.10c0.55 ± 0.09dBcl21.14 ± 0.040.53 ± 0.09cc0.61 ± 0.10cc0.92 ± 0.09dTLR40.57 ± 0.041.37 ± 0.10ccc1.15 ± 0.11cc0.97 ± 0.08cdNF-κB0.68 ± 0.051.49 ± 0.12ccc1.25 ± 0.13c0.91 ± 0.10ddNrf21.09 ± 0.041.27 ± 0.081.42 ± 0.10c1.63 ± 0.08ccdHO-11.17 ± 0.050.64 ± 0.11cc0.73 ± 0.09c1.03 ± 0.08dPGC-1α0.95 ± 0.040.41 ± 0.09cc0.53 ± 0.08cc0.75 ± 0.08dSirtuin 11.73 ± 0.050.85 ± 0.12ccc1.14 ± 0.13c1.59 ± 0.12ddAMPK1.19 ± 0.060.94 ± 0.111.27 ± 0.141.48 ± 0.12dData are shown in means ± SEM. c, cc, and ccc indicate p values lower than 0.05, 0.01, and 0.001, respectively (relative to the control); d and dd indicate p values less than 05 and 0.01, respectively (relative to the LPS/D-Gal group). $$n = 7$$ per experimental group.
There is a crosstalk between oxidative stress, sirtuin 1, and inflammation. In this regard, during oxidative and inflammatory events, level of sirtuin 1 decreases and vice versa [33]. Our results showed significant and marked fall of sirtuin 1 level (Fig. 2F) (F [3,24] = 7.84, $p \leq 0.001$) ($p \leq 0.01$) and its gene expression (Table 2) ($p \leq 0.001$). In contrast, giving capsaicin orally at a dose of 20 mg/kg, but not at the lower dose of 5 mg/kg, was capable to properly and significantly elevate hepatic level of sirtuin 1 ($p \leq 0.05$) and also its gene expression ($p \leq 0.01$).
## The effect of capsaicin treatment on inflammation-associated factors
To properly assess involvement of inflammation following LPS/D-Gal challenge and to evaluate capsaicin efficacy, we measured liver level of nuclear level of NF-kB (Fig. 3B) and its mRNA gene expression (Table 2), level of TLR4 (Fig. 3C) and its gene expression (Table 2), (IL-1β) (Fig. 3D), TNF-α (Fig. 3E) and IL-10 (Fig. 3F). Our obtained data showed elevated levels of NF-kB (F [3,24] = 8.92) ($p \leq 0.001$) and its gene expression ($p \leq 0.001$), TLR4 (F [3,24] = 11.57) ($p \leq 0.001$) and its gene expression ($p \leq 0.001$), IL-1β (F [3,24] = 8.53) ($P \leq 0.01$), and TNF-α (F [3,24] = 9.25) ($p \leq 0.001$) besides non-significant and slight increase of anti-inflammatory factor IL-10 β (F [3,24] = 2.77) in the LPS/D-Gal group. In contrast, capsaicin pretreatment of LPS/D-Gal-challenged group at a dose of 20 mg/kg, but not at a dosage of 5 mg/kg, reduced hepatic levels of NF-kB ($p \leq 0.05$) and its gene expression ($p \leq 0.01$), TLR4 ($p \leq 0.05$) and its gene expression ($p \leq 0.05$), IL-1β ($p \leq 0.05$), and TNF-α ($p \leq 0.01$) and properly elevated hepatic level of IL-10 ($p \leq 0.05$) when these findings are compared with the LPS/D-Gal group.
## The effect of capsaicin treatment on hepatic apoptosis-associated factors
We take into account caspase 3 activity and DNA fragmentation (Table 1) as known apoptotic indices [34]. Statistical analysis of apoptosis-associated data indicated significant differences between the groups for caspase 3 (F [3,24] = 11.75, $p \leq 0.001$) and DNA fragmentation (F [3,24] = 10.38, $p \leq 0.001$). The LPS/D-Gal group had a significantly elevated activity of caspase 3 ($p \leq 0.001$) and also higher level of DNA fragmentation ($p \leq 0.001$) versus the control group. On the contrary, capsaicin administration to LPS/D-Gal-challenged group at a dose of 20 mg/kg significantly and suitably reduced DNA fragmentation ($p \leq 0.05$) and also activity of caspase 3 ($p \leq 0.05$) when comparing these findings with relevant data of the LPS/D-Gal group. To further evaluate the beneficial effect of capsaicin on apoptosis at gene expression level, we determined liver mRNA for pro-apoptotic factor Bax and anti-apoptotic factor Bcl2, as shown in Table 2. Analysis of data showed significantly higher gene expression for Bax ($p \leq 0.001$) and lower gene expression for Bcl2 ($p \leq 0.01$) in LPS/D-Gal group. In addition, capsaicin given at a dose of 20 mg/kg significantly reduced *Bax* gene expression ($p \leq 0.05$) and raised *Bcl2* gene expression ($p \leq 0.05$), indirectly indicating lower rate of apoptosis in the liver tissue.
## The effect of capsaicin on mitochondrial homeostasis and biogenesis
Development of mitochondrial dysfunction in disease conditions leads to liver injury [35]. During a septic insult, mitochondrial biogenesis is upset which is indicated by lower expression for PGC-1α as its key regulator [36]. In this study, to have an evaluation of mitochondrial biogenesis and health status, we measured MMP (Table 1) in addition to determination of gene expression for PGC-1α and AMPK (Table 2). Analysis of data showed lower level of MMP (F [3,24] = 9.83) ($p \leq 0.01$) besides reduction of PGC-1α ($p \leq 0.01$) and AMPK ($p \leq 0.05$) in the LPS/D-Gal group as compared to the control group. Contrarily, capsaicin at a dose of 20 mg/kg appropriately prevented MMP fall ($p \leq 0.05$) and PGC-1α ($p \leq 0.05$) and AMPK ($p \leq 0.05$) reduction.
## The effect of capsaicin on liver histopathology
Fig. 4A shows histopathological results based on a 0–3 scoring according to severity of injury in different groups and typical photomicrographs of the liver tissue. Microscopic assessment of liver tissue from the control group showed normal structure such as clear central vein, distinct hepatocytes, and normal sinusoidal spaces. On the contrary, liver tissue of LPS/D-Gal group had robust pathologic changes, as shown by widespread necrotic areas, neutrophil infiltration and presence of inflammatory cells and even derangement of liver cell organization as compared with the control group. Accordingly, severity of pathological score for LPS/D-Gal group was significantly higher when compared to the control group ($p \leq 0.001$). These inappropriate changes were less evident in capsaicin-pretreated LPS/D-Gal groups, so pathological score in LPS/D-Gal + capsaicin20 group (but not LPS/D-Gal + capsaicin5 group) was significantly lower than LPS/D-Gal group ($p \leq 0.01$) (F [3,16] = 17.32, $p \leq 0.001$).Fig. 4Severity of histopathological changes (on the basis of a 0–3 scoring) and typical histopathologic alterations of the liver tissue sections stained with H&E protocol (A) and TLR4 immunoreactivity (IRA) using immunohistochemistry method (B). ( A) Control, (B) LPS/D-Gal, (C) LPS/D-Gal + Capsaicin at a dose of 5 mg/kg, and (D) LPS/D-Gal + Capsaicin at a dose of 20 mg/kg $$n = 5$$ per group. Solid red arrows indicate infiltration of defensive cells and solid black arrows show derangement of liver cell organization with abnormal and expanded sinusoidal spaces and/or cytoarchitectural disarrangement. Experiments were done once and every analysis was repeated twice and its data was averaged for each specimen. ( For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)Fig. 4 Evaluation of TLR4 immunoreactivity in our experimental groups indicated a significant inter-group differences (F [3,16] = 11.69, $p \leq 0.001$) (Fig. 4B). Further Tukey post-test showed a markedly higher TLR4 IRA in LPS/D-Gal group versus the control group ($p \leq 0.001$). Conversely, this immunoreactivity for TLR4 was significantly lower in capsaicin20-treated LPS/D-Gal group when compared to vehicle-treated LPS/D-Gal group ($p \leq 0.05$).
## Discussion
This research study was conducted to show possible hepatoprotective effect of capsaicin in LPS/D-Gal model of ALI. Administration of capsaicin at the higher dose (20 mg/kg) to LPS/D-Gal-injured animals decreased liver functional markers besides its reduction of oxidative stress, inflammation, apoptosis, and mitochondrial dysfunction.
Liver is a vital tissue for the detoxification of toxic substances and excessive exposure to toxicants is associated with hepatic injury [37]. Liver damage due to a challenge of LPS/D-Gal in rodents is regarded a reliable model to test the possible efficacy of new drugs, especially natural products, to prevent ultimate liver failure [38]. To ameliorate sepsis-induced tissue injury, inhibition of inflammatory and oxidative stress factors is of paramount significance [39]. To manage and treat disorders of liver and its associated complications, research studies have emphasized on finding of novel agents with antioxidant and anti-inflammatory potential. Hence, we selected capsaicin as the main effective ingredient of red chili to explore its possible efficacy in LPS/D-Gal model of ALI.
The activity of aminotransferase enzymes ALT and AST is taken as reliable and consistent indicators of liver function. ALT and AST changes indicate somehow hepatocellular dysfunction. Hence, blood levels of these enzymes increase following liver diseases [40]. Similarly, in this study, serum activities of AST and ALT significantly were higher after LPS/D-GAL. On the contrary, capsaicin administration given at 20 mg/kg ameliorated these changes which is indicative of its suppression of hepatic disturbance. In agreement with our finding, it has been shown that capsaicin has hepatoprotective effect in mice on a high-fat diet, as shown by lower levels of ALT and AST, that is exerted partly through its alleviation of mitochondrial oxidative stress and improvement of mitochondrial function and bioenergetics [41].
Furthermore, we also showed protective potential of capsaicin on the liver tissue, as demonstrated by its amelioration of hemorrhagic areas, necrosis, neutrophil infiltration and prevention of liver cell disarrangement. In line with this finding, it has been shown that capsaicin is capable to exert a protective effect in the liver tissue in carbon tetrachloride (CCl4) model of hepatotoxicity in the rat [42]. Additionally, other studies have also recently shown protective effect of capsaicin in other tissues under toxic conditions that is partly due to its antioxidant and anti-inflammatory potential [21,43].
Earlier studies have shown the involvement of oxidative stress in LPS/D-Gal-induced ALI [11]. Obtained findings of this study showed significant elevation of MDA and nitrite and reduction of catalase and SOD activity following LPS-D-Gal that is in agreement with past studies [44,45]. On the contrary, capsaicin alleviated oxidative stress, as was shown by reversing some of these changes. To support our obtained data, it has been shown that capsaicin due to its anti-oxidative potential is capable to exert a hepatoprotective effect against concanavalin A through suppressing oxidative stress and inflammation [16]. In another study, it was shown that capsaicin could ameliorate alcohol-induced ALI, partly via improvement of hepatic antioxidant status [23]. In addition, Nrf2 is known as an important redox-sensing transcription factor that governs the gene expression of endogenous antioxidants [46]. Nrf2 itself control and induces the expression of HO-1 with important roles in cellular antioxidant axis [47]. Studies have shown that activation of Nrf2/HO-1 pathway produces a protective effect in LPS/D-Gal model of liver injury [48]. In this study, capsaicin administration significantly improved level and gene expression of Nrf2 in LPS/D-Gal group which is responsible for part of its anti-oxidative potential. In agreement with this finding, it has been shown that capsaicin can prevent contrast-associated kidney damage, partly via activation of Nrf2 cascade [21].
Sirtuin 1 is a nicotinamide adenine dinucleotide (NAD)-associated deacetylase that plays important tasks in prevention of apoptosis, DNA injury, and imbalance of mitochondrial metabolism [49]. Sirtuin 1 protects living cells against oxidative injury via Nrf2 pathway [50] and its upregulation attenuates inflammatory events through suppression of NF-κB cascade [51]. Hence, sirtuin 1 upregulation can mitigate oxidative and inflammatory events [52]. In addition, down-regulation of sirtuin 1 in the liver tissue has been reported following LPS/D-Gal [53]. In our study, capsaicin was capable to raise gene expression of sirtuin 1 besides preventing reduction of its tissue level. In line with this finding, capsaicin can alleviate intermittent high glucose-induced endothelial senescence via elevating sirtuin 1 and proper regulation of TRPV1/AMPK pathways [54].
LPS/D-Gal challenge is associated with activation of Kupffer cells with ensuing generation and release of inflammatory cytokines encompassing IL-1β, IL-6, and TNFα [55]. In this study, LPS/D-Gal injection caused significant elevation of inflammatory factors besides elevation of NF-kB and TLR4, clearly denoting the occurrence of inflammation in the liver. Contrarily, capsaicin administration alleviated inflammation severity, as was apparent by lower quantities of proinflammatory cytokines. Anti-inflammatory potential of capsaicin in acetaminophen-induced ALI has been reported before [56].
LPS/D-Gal also increases hepatic apoptosis, as shown by higher rate of DNA fragmentation and higher activity of caspase 3 [57] that was also shown in this study. In contrast, capsaicin was able to significantly alleviate hepatic levels of these apoptotic markers that is also confirmed in earlier studies [18]. In addition, capsaicin was capable to reduce gene expression of pro-apoptotic factor Bax and to elevate gene expression for anti-apoptotic factor Bcl2 and theses alterations produce lower level of apoptosis in the liver tissue. Of relevance to our findings on apoptosis, capsaicin can protect against LPS-induced acute lung injury via down-regulation of caspase 3 and Bax expression and up-regulation of Bcl-2 besides its proper regulation of NF-κB/PI3K/AKT/mTOR cascades [43].
Changes in mitochondrial metabolism and biogenesis play a pivotal role in different diseases [58]. Past studies have indicated that mitochondrial dysfunction is a key factor in the development of ALI [59,60]. A challenge of LPS is associated with disturbance of mitochondrial dynamics and biogenesis in the liver [61]. Expression or level of PGC-1α (PPARγ coactivator-1α) as a key regulator of mitochondrial function decreases during LPS-induced ALI [62]. In this study, LPS/D-Gal-provoked ALI reduced MMP in addition to down-regulation of PGC-1α and its gene expression which has also been reported separately in earlier studies [63,64]. Capsaicin pretreatment significantly improved dysfunction of mitochondrial function, as indicated by higher level of MMP and greater level and gene expression of PGC-1α. In agreement with this finding, previous studies have shown that capsaicin can protect cardiomyocytes against LPS injury via improvement of mitochondrial function [65]. In addition, capsaicin affects lipogenesis in HepG2 cells via activating and/or up-regulating AMPK/PGC-1α [66]. Besides PGC-1α and sirtuin 1 as the central regulators of mitochondrial biogenesis, AMPK signaling also plays a pivotal role in regulation of mitochondrial biogenesis, inflammation, and apoptosis and it is essential for maintenance of cell homeostasis [67]. In the current study, capsaicin properly elevated gene expression of AMPK following LPS/D-Gal challenge. In support of our finding, it has been shown that part of protective effect of capsaicin in LPS model of cardiomyocyte damage is through regulation of AMPK/mTOR pathways and in this way can inhibit oxidative stress and inflammation as well as its maintenance of mitochondrial function and autophagy augmentation [65].
Lack of further histochemical studies including NF-kB and Nrf2 immunohistochemistry and absence of Western blotting experiments were some limitations of the present study which may be taken into account in future relevant studies.
To conclude, this study indicated hepatoprotective property of capsaicin under septic conditions that can be attributed to its down-regulation of oxidative and inflammatory processes besides its potential to attenuate mitochondrial dysfunction and apoptosis. This maybe of potential benefit in clinical settings after further studies.
## Author contribution statement
Tourandokht Baluchnejadmojarad and Mehrdad Roghani conceived and designed the study, analyzed and interpreted the data, and wrote the paper; Atefeh Ghorbanpour and Sepide Salari performed the experiments, analyzed and interpreted the data, and wrote the paper.
## Funding statement
This research project was financially supported in 2018 by National Institute for Medical Research Development (NIMAD) of Iran (grant number 965431).
## Data availability statement
Data sets of this study are available from the corresponding author on reasonable request.
## Conflict of interest statement
The authors declare no conflict of interest.
## Additional information
No additional information is available for this paper.
## References
1. García-Cortés M., Ortega-Alonso A., Andrade R.J.. **Safety of treating acute liver injury and failure**. *Expet Opin. Drug Saf.* (2022) **21** 191-203
2. Stravitz R.T., Lee W.M.. **Acute liver failure**. *Lancet* (2019) **394** 869-881. PMID: 31498101
3. Xu R., Qiu S., Zhang J., Liu X., Zhang L., Xing H., You M., Wang M., Lu Y., Zhang P., Zhu J.. **Silibinin schiff base derivatives counteract CCl(4)-induced acute liver injury by enhancing anti-inflammatory and antiapoptotic bioactivities**. *Drug Des. Dev. Ther.* (2022) **16** 1441-1456
4. Peng X., Yang Y., Tang L., Wan J., Dai J., Li L., Huang J., Shen Y., Lin L., Gong X., Zhang L.. **Therapeutic benefits of apocynin in mice with lipopolysaccharide/D-galactosamine-induced acute liver injury via suppression of the late stage pro-apoptotic AMPK/JNK pathway**. *Biomed. Pharm.* (2020) **125**
5. Liu Z., Wang X., Li L., Wei G., Zhao M.. *Mitochondrial Function, and Inflammation, Oxidative medicine and cellular longevity 2020* (2020)
6. Al-Harbi N.O., Nadeem A., Al-Harbi M.M., Zoheir K.M.A., Ansari M.A., El-Sherbeeny A.M., Alanazi K.M., Alotaibi M.R., Ahmad S.F.. **Psoriatic inflammation causes hepatic inflammation with concomitant dysregulation in hepatic metabolism via IL-17A/IL-17 receptor signaling in a murine model**. *Immunobiology* (2017) **222** 128-136. PMID: 27773660
7. Helal M.G., Samra Y.A.. **Irbesartan mitigates acute liver injury, oxidative stress, and apoptosis induced by acetaminophen in mice**. *J. Biochem. Mol. Toxicol.* (2020)
8. Rousta A.M., Mirahmadi S.M., Shahmohammadi A., Ramzi S., Baluchnejadmojarad T., Roghani M.. **S-allyl cysteine, an active ingredient of garlic, attenuates acute liver dysfunction induced by lipopolysaccharide/d-galactosamine in mouse: underlying mechanisms**. *J. Biochem. Mol. Toxicol.* (2020)
9. Ansari M.A., Raish M., Bin Jardan Y.A., Ahmad A., Shahid M., Ahmad S.F., Haq N., Khan M.R., Bakheet S.A.. **Sinapic acid ameliorates D-galactosamine/lipopolysaccharide-induced fulminant hepatitis in rats: role of nuclear factor erythroid-related factor 2/heme oxygenase-1 pathways**. *World J. Gastroenterol.* (2021) **27** 592-608. PMID: 33642831
10. Wang P.F., Xie K., Cao Y.X., Zhang A.. **Hepatoprotective effect of mitochondria-targeted antioxidant mito-TEMPO against lipopolysaccharide-induced liver injury in mouse**. *Mediat. Inflamm.* (2022)
11. Li Z., Feng H., Han L., Ding L., Shen B., Tian Y., Zhao L., Jin M., Wang Q., Qin H., Cheng J., Liu G.. **Chicoric acid ameliorate inflammation and oxidative stress in Lipopolysaccharide and d-galactosamine induced acute liver injury**. *J. Cell Mol. Med.* (2020) **24** 3022-3033. PMID: 31989756
12. Kang C., Wang B., Kaliannan K., Wang X., Lang H., Hui S., Huang L., Zhang Y., Zhou M., Chen M., Mi M.. **Gut microbiota mediates the protective effects of dietary capsaicin against chronic low-grade inflammation and associated obesity induced by high-fat diet**. *mBio* (2017) **8** e00470. PMID: 28536285
13. Ilie M.A., Caruntu C., Tampa M., Georgescu S.R., Matei C., Negrei C., Ion R.M., Constantin C., Neagu M., Boda D.. **Capsaicin: physicochemical properties, cutaneous reactions and potential applications in painful and inflammatory conditions**. *Exp. Ther. Med.* (2019) **18** 916-925. PMID: 31384324
14. Sanati S., Razavi B.M., Hosseinzadeh H.. **A review of the effects of Capsicum annuum L. and its constituent, capsaicin, in metabolic syndrome**. *Iran. J. Basic Med. Sci.* (2018) **21** 439-448. PMID: 29922422
15. Chaudhary A., Gour J.K., Rizvi S.I.. (2019) 1-7
16. Zhang H., Bai Y., Gao M., Zhang J., Dong G., Yan F., Ma Q., Fu X., Zhang Q., Li C., Shi H., Ning Z., Dai J., Li Z., Ming J., Xue Q., Si C., Xiong H.. **Hepatoprotective effect of capsaicin against concanavalin A-induced hepatic injury via inhibiting oxidative stress and inflammation**. *Am. J. Tourism Res.* (2019) **11** 3029-3038
17. Doğan M.F., Başak Türkmen N., Taşlıdere A., Şahin Y., Çiftçi O.. **The protective effects of capsaicin on oxidative damage-induced by 2,3,7,8-tetrachlorodibenzo-p-dioxin in rats**. *Drug Chem. Toxicol.* (2021) 1-8
18. Tanrikulu-Kucuk S., Basaran-Kucukgergin C., Sogut I., Tuncdemir M., Dogru-Abbasoglu S., Seyithanoglu M., Kocak H., Oner-Iyidogan Y.. **Dietary curcumin and capsaicin: relationship with hepatic oxidative stress and apoptosis in rats fed a high fat diet**. *Adv. Clin. Exp. Med. : Off. Organ Wroclaw Med. Univer.* (2019) **28** 1013-1020
19. Tang J., Luo K., Li Y., Chen Q., Tang D., Wang D., Xiao J.. **Capsaicin attenuates LPS-induced inflammatory cytokine production by upregulation of LXRalpha**. *Int. Immunopharm.* (2015) **28** 264-269
20. Abdel-Salam O.M., Abdel-Rahman R.F., Sleem A.A., Farrag A.R.. **Modulation of lipopolysaccharide-induced oxidative stress by capsaicin**. *Inflammopharmacology* (2012) **20** 207-217. PMID: 22127606
21. Ran F., Yang Y., Yang L., Chen S., He P., Liu Q., Zou Q., Wang D., Hou J., Wang P.. **Capsaicin prevents contrast-associated acute kidney injury through activation of Nrf2 in mice**. *Oxid. Med. Cell. Longev.* (2022) **2022**
22. Li L., Yin H., Zhao Y., Zhang X., Duan C., Liu J., Huang C., Liu S., Yang S., Li X.. **Protective role of puerarin on LPS/D-Gal induced acute liver injury via restoring autophagy**. *Am. J. Tourism Res.* (2018) **10** 957-965
23. Koneru M., Sahu B.D., Mir S.M., Ravuri H.G., Kuncha M., Mahesh Kumar J., Kilari E.K., Sistla R.. **Capsaicin, the pungent principle of peppers, ameliorates alcohol-induced acute liver injury in mice via modulation of matrix metalloproteinases**. *Can. J. Physiol. Pharmacol.* (2018) **96** 419-427. PMID: 29053935
24. Afshin-Majd S., Khalili M., Roghani M., Mehranmehr N., Baluchnejadmojarad T.. **Carnosine exerts neuroprotective effect against 6-hydroxydopamine toxicity in hemiparkinsonian rat**. *Mol. Neurobiol.* (2015) **51** 1064-1070. PMID: 24939694
25. Raoufi S., Baluchnejadmojarad T., Roghani M., Ghazanfari T., Khojasteh F., Mansouri M.. **Antidiabetic potential of salvianolic acid B in multiple low-dose streptozotocin-induced diabetes**. *Pharm. Biol.* (2015) **53** 1803-1809. PMID: 25885938
26. Claiborne A.. (1985)
27. Bradford M.M.. **A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding**. *Anal. Biochem.* (1976) **72** 248-254. PMID: 942051
28. Movsesyan V.A., Yakovlev A.G., Dabaghyan E.A., Stoica B.A., Faden A.I.. **Ceramide induces neuronal apoptosis through the caspase-9/caspase-3 pathway**. *Biochem. Biophys. Res. Commun.* (2002) **299** 201-207. DOI: 10.1016/S0006-291X(02)02593-7
29. Baluchnejadmojarad T., Mohamadi-Zarch S.M., Roghani M.. **Safranal, an active ingredient of saffron, attenuates cognitive deficits in amyloid β-induced rat model of Alzheimer's disease: underlying mechanisms**. *Metab. Brain Dis.* (2019) **34** 1747-1759. PMID: 31422512
30. Lyu Z., Ji X., Chen G., An B.. **Atractylodin ameliorates lipopolysaccharide and d-galactosamine-induced acute liver failure via the suppression of inflammation and oxidative stress**. *Int. Immunopharm.* (2019) **72** 348-357
31. Ali S.A., Sharief N.H., Mohamed Y.S.. **Hepatoprotective activity of some medicinal plants in Sudan**. *Alternative Med.* (2019)
32. Liu Z., Dou W., Zheng Y., Wen Q., Qin M., Wang X., Tang H., Zhang R., Lv D., Wang J., Zhao S.. **Curcumin upregulates Nrf2 nuclear translocation and protects rat hepatic stellate cells against oxidative stress**. *Mol. Med. Rep.* (2016) **13** 1717-1724. PMID: 26676408
33. Salminen A., Kaarniranta K., Kauppinen A.. **Crosstalk between Oxidative stress and SIRT1: impact on the aging process**. *Int. J. Mol. Sci.* (2013) **14** 3834-3859. PMID: 23434668
34. Stanojlovic M., Gusevac Stojanovic I., Zaric M., Martinovic J., Mitrovic N., Grkovic I., Drakulic D.. **Progesterone protects prefrontal cortex in rat model of permanent bilateral common carotid occlusion via progesterone receptors and akt/erk/eNOS**. *Cell. Mol. Neurobiol.* (2020) **40** 829-843. PMID: 31865501
35. Chen J., Wang D., Zong Y., Yang X.. **DHA protects hepatocytes from Oxidative injury through gpr120/ERK-mediated mitophagy**. *Int. J. Mol. Sci.* (2021) **22**
36. Tran M., Tam D., Bardia A., Bhasin M., Rowe G.C., Kher A., Zsengeller Z.K., Akhavan-Sharif M.R., Khankin E.V., Saintgeniez M., David S., Burstein D., Karumanchi S.A., Stillman I.E., Arany Z., Parikh S.M.. **PGC-1α promotes recovery after acute kidney injury during systemic inflammation in mice**. *J. Clin. Invest.* (2011) **121** 4003-4014. PMID: 21881206
37. Al-Harbi N.O., Imam F., Nadeem A., Al-Harbi M.M., Iqbal M., Ahmad S.F.. **Carbon tetrachloride-induced hepatotoxicity in rat is reversed by treatment with riboflavin**. *Int. Immunopharm.* (2014) **21** 383-388
38. Yang S., Kuang G., Zhang L., Wu S., Zhao Z., Wang B., Yin X., Gong X., Wan J.. **Mangiferin attenuates LPS/D-GalN-Induced acute liver injury by promoting HO-1 in kupffer cells**. *Front. Immunol.* (2020) **11** 285. PMID: 32158448
39. Al-Harbi N.O., Nadeem A., Ahmad S.F., Alanazi M.M., Aldossari A.A., Alasmari F.. **Amelioration of sepsis-induced acute kidney injury through inhibition of inflammatory cytokines and oxidative stress in dendritic cells and neutrophils respectively in mice: role of spleen tyrosine kinase signaling**. *Biochimie* (2019) **158** 102-110. PMID: 30599182
40. Neag M.A., Catinean A., Muntean D.M., Pop M.R., Bocsan C.I., Botan E.C., Buzoianu A.D.. **Probiotic Bacillus spores protect against acetaminophen induced acute liver injury in rats**. *Nutrients* (2020) **12**
41. Sekeroglu V., Aydin B., Atli Sekeroglu Z., Ozdener Kompe Y.. **Hepatoprotective effects of capsaicin and alpha-tocopherol on mitochondrial function in mice fed a high-fat diet**. *Biomed. Pharm.* (2018) **98** 821-825
42. Hassan M.H., Edfawy M., Mansour A., Hamed A.A.. **Antioxidant and antiapoptotic effects of capsaicin against carbon tetrachloride-induced hepatotoxicity in rats**. *Toxicol. Ind. Health* (2012) **28** 428-438. PMID: 21859771
43. Chen H., Li N., Zhan X., Zheng T., Huang X., Chen Q., Song Z., Yang F., Nie H., Zhang Y., Zheng B., Gong Q.. **Capsaicin protects against lipopolysaccharide-induced acute lung injury through the HMGB1/NF-κB and PI3K/AKT/mTOR pathways**. *J. Inflamm. Res.* (2021) **14** 5291-5304. PMID: 34703269
44. Bian X., Liu X., Liu J., Zhao Y., Li H., Zhang L., Li P., Gao Y.. **Hepatoprotective effect of chiisanoside from Acanthopanax sessiliflorus against LPS/D-GalN-induced acute liver injury by inhibiting NF-kappaB and activating Nrf2/HO-1 signaling pathways**. *J. Sci. Food Agric.* (2019) **99** 3283-3290. PMID: 30552777
45. Li M., Wang S., Li X., Jiang L., Wang X., Kou R., Wang Q., Xu L., Zhao N., Xie K.. **Diallyl sulfide protects against lipopolysaccharide/d-galactosamine-induced acute liver injury by inhibiting oxidative stress, inflammation and apoptosis in mice**. *Food Chem. Toxicol. : Int. J. Pub. British Indus. Biolog. Res. Ass.* (2018) **120** 500-509
46. Kasai S., Shimizu S., Tatara Y., Mimura J., Itoh K.. **Regulation of Nrf2 by mitochondrial reactive Oxygen species in physiology and pathology**. *Biomolecules* (2020) **10**
47. Habtemariam S.. **The Nrf2/HO-1 Axis as targets for flavanones: neuroprotection by pinocembrin, naringenin, and eriodictyol**. *Med. Cell. Longev* (2019)
48. Li Z., Feng H., Wang Y., Shen B., Tian Y., Wu L., Zhang Q., Jin M., Liu G.. **Rosmarinic acid protects mice from lipopolysaccharide/d-galactosamine-induced acute liver injury by inhibiting MAPKs/NF-kappaB and activating Nrf2/HO-1 signaling pathways**. *Int. Immunopharm.* (2019) **67** 465-472
49. Jiang W., Zhang X., Hao J., Shen J., Fang J., Dong W., Wang D., Zhang X., Shui W., Luo Y., Lin L., Qiu Q., Liu B., Hu Z.. **SIRT1 protects against apoptosis by promoting autophagy in degenerative human disc nucleus pulposus cells**. *Sci. Rep.* (2014) **4** 7456. PMID: 25503852
50. Li Y., Xu W., McBurney M.W., Longo V.D.. **SirT1 inhibition reduces IGF-I/IRS-2/Ras/ERK1/2 signaling and protects neurons**. *Cell Metabol.* (2008) **8** 38-48
51. Yang H., Zhang W., Pan H., Feldser H.G., Lainez E., Miller C., Leung S., Zhong Z., Zhao H., Sweitzer S., Considine T., Riera T., Suri V., White B., Ellis J.L., Vlasuk G.P., Loh C.. **SIRT1 activators suppress inflammatory responses through promotion of p65 deacetylation and inhibition of NF-κB activity**. *PLoS One* (2012) **7**
52. Yang R., Song C., Chen J., Zhou L., Jiang X., Cao X., Sun Y., Zhang Q.. **Limonin ameliorates acetaminophen-induced hepatotoxicity by activating Nrf2 antioxidative pathway and inhibiting NF-κB inflammatory response via upregulating Sirt1**. *Phytomedicine* (2020) **69**
53. Li L., Wang H., Zhao S., Zhao Y., Chen Y., Zhang J., Wang C., Sun N., Fan H.. **Paeoniflorin ameliorates lipopolysaccharide-induced acute liver injury by inhibiting oxidative stress and inflammation via SIRT1/FOXO1a/SOD2 signaling in rats**. *Phytother Res.* (2022) **36** 2558-2571. PMID: 35570830
54. Zhu S.L., Wang M.L., He Y.T., Guo S.W., Li T.T., Peng W.J., Luo D.. **Capsaicin ameliorates intermittent high glucose-mediated endothelial senescence via the TRPV1/SIRT1 pathway**. *Phytomedicine* (2022) **100**
55. Yang P., Zhou W., Li C., Zhang M., Jiang Y., Jiang R., Ba H., Li C., Wang J., Yin B., Gong F., Li Z.. **Kupffer-cell-expressed transmembrane TNF-α is a major contributor to lipopolysaccharide and D-galactosamine-induced liver injury**. *Cell Tissue Res.* (2016) **363** 371-383. PMID: 26267221
56. Zhan X., Zhang J., Chen H., Liu L., Zhou Y., Zheng T., Li S., Zhang Y., Zheng B., Gong Q.. **Capsaicin alleviates acetaminophen-induced acute liver injury in mice**. *Clin. Immunol.* (2020) **220**
57. Lian L.H., Jin X., Wu Y.L., Cai X.F., Lee J.J., Nan J.X.. **Hepatoprotective effects of Sedum sarmentosum on D-galactosamine/lipopolysaccharide-induced murine fulminant hepatic failure**. *J. Pharmacol. Sci.* (2010) **114** 147-157. PMID: 20838028
58. Sanchis-Gomar F., García-Giménez J.L., Gómez-Cabrera M.C., Pallardó F.V.. **Mitochondrial biogenesis in health and disease. Molecular and therapeutic approaches**. *Curr. Pharmaceut. Des.* (2014) **20** 5619-5633
59. Niu B., Lei X., Xu Q., Ju Y., Xu D., Mao L., Li J., Zheng Y., Sun N., Zhang X., Mao Y., Li X.. **Protecting mitochondria via inhibiting VDAC1 oligomerization alleviates ferroptosis in acetaminophen-induced acute liver injury**. *Cell Biol. Toxicol.* (2022) **38** 505-530. PMID: 34401974
60. Rostami A., Baluchnejadmojarad T., Roghani M.. **Sinapic acid ameliorates paracetamol-induced acute liver injury through targeting oxidative stress and inflammation**. *Mol. Biol. Rep.* (2022) **49** 4179-4191. PMID: 35279777
61. Lee S.B., Kang J.W., Kim S.J., Ahn J., Kim J., Lee S.M.. **Afzelin ameliorates D-galactosamine and lipopolysaccharide-induced fulminant hepatic failure by modulating mitochondrial quality control and dynamics**. *Br. J. Pharmacol.* (2017) **174** 195-209. PMID: 27861739
62. Xu Y., Chen J., Yu X., Tao W., Jiang F., Yin Z., Liu C.. **Protective effects of chlorogenic acid on acute hepatotoxicity induced by lipopolysaccharide in mice**. *Inflamm. Res.* (2010) **59** 871-877. PMID: 20405164
63. Lou G., Li A., Cen Y., Yang Q., Zhang T., Qi J., Chen Z., Liu Y.. **Selonsertib, a potential drug for liver failure therapy by rescuing the mitochondrial dysfunction of macrophage via ASK1-JNK-DRP1 pathway**. *Cell Biosci.* (2021) **11** 9. PMID: 33413667
64. Wang L., Wang X., Kong L., Wang S., Huang K., Wu J., Wang C., Sun H., Liu K., Meng Q.. **Isoliquiritigenin alleviates LPS/D-GalN-induced acute liver failure by activating the PGC-1α/Nrf2 pathway to reduce oxidative stress and inflammatory response**. *Int. Immunopharm.* (2021) **100**
65. Qiao Y., Wang L., Hu T., Yin D., He H., He M.. **Capsaicin protects cardiomyocytes against lipopolysaccharide-induced damage via 14-3-3γ-mediated autophagy augmentation**. *Front. Pharmacol.* (2021) **12**
66. Bort A., Sánchez B.G., Mateos-Gómez P.A., Díaz-Laviada I., Rodríguez-Henche N.. **Capsaicin targets lipogenesis in HepG2 cells through AMPK activation, AKT inhibition and PPARs regulation**. *Int. J. Mol. Sci.* (2019) **20**
67. He L., Zhou X., Huang N., Li H., Tian J., Li T., Yao K., Nyachoti C.M., Kim S.W., Yin Y.. **AMPK regulation of glucose, lipid and protein metabolism: mechanisms and nutritional significance**. *Curr. Protein Pept. Sci.* (2017) **18** 562-570. PMID: 27356941
|
---
title: 'The I148M PNPLA3 variant mitigates niacin beneficial effects: How the genetic
screening in non-alcoholic fatty liver disease patients gains value'
authors:
- Erika Paolini
- Miriam Longo
- Marica Meroni
- Giada Tria
- Annalisa Cespiati
- Rosa Lombardi
- Sara Badiali
- Marco Maggioni
- Anna Ludovica Fracanzani
- Paola Dongiovanni
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC10018489
doi: 10.3389/fnut.2023.1101341
license: CC BY 4.0
---
# The I148M PNPLA3 variant mitigates niacin beneficial effects: How the genetic screening in non-alcoholic fatty liver disease patients gains value
## Abstract
### Background
The PNPLA3 p.I148M impact on fat accumulation can be modulated by nutrients. Niacin (Vitamin B3) reduced triglycerides synthesis in in vitro and in vivo NAFLD models.
### Objectives
In this study, we aimed to investigate the niacin-I148M polymorphism crosstalk in NAFLD patients and examine niacin’s beneficial effect in reducing fat by exploiting hepatoma cells with different PNPLA3 genotype.
### Design
We enrolled 172 (Discovery cohort) and 358 (Validation cohort) patients with non-invasive and histological diagnosis of NAFLD, respectively. Dietary niacin was collected from food diary, while its serum levels were quantified by ELISA. Hepatic expression of genes related to NAD metabolism was evaluated by RNAseq in bariatric NAFLD patients ($$n = 183$$; Transcriptomic cohort). Hep3B (148I/I) and HepG2 (148M/M) cells were silenced (siHep3B) or overexpressed (HepG2I148+) for PNPLA3, respectively.
### Results
In the Discovery cohort, dietary niacin was significantly reduced in patients with steatosis ≥ 2 and in I148M carriers. Serum niacin was lower in subjects carrying the G at risk allele and negatively correlated with obesity. The latter result was confirmed in the Validation cohort. At multivariate analysis, the I148M polymorphism was independently associated with serum niacin, supporting that it may be directly involved in the modulation of its availability. siHep3B cells showed an impaired NAD biosynthesis comparable to HepG2 cells which led to lower niacin efficacy in clearing fat, supporting a required functional protein to guarantee its effectiveness. Conversely, the restoration of PNPLA3 Wt protein in HepG2I148+ cells recovered the NAD pathway and improved niacin efficacy. Finally, niacin inhibited de novo lipogenesis through the ERK$\frac{1}{2}$/AMPK/SIRT1 pathway, with the consequent SREBP1-driven PNPLA3 reduction only in Hep3B and HepG2I148M+ cells.
### Conclusions
We demonstrated a niacin-PNPLA3 I148M interaction in NAFLD patients which possibly pave the way to vitamin B3 supplementation in those with a predisposing genetic background.
## Introduction
Non-alcoholic fatty liver disease (NAFLD) is the chronic liver disorder with the highest prevalence worldwide whose pathogenesis is mainly related to the presence of obesity and type 2 diabetes (T2D). NAFLD comprises a plethora of clinical conditions ranging from simple steatosis to necroinflammation and ballooning which together define non-alcoholic steatohepatitis (NASH) that in turn may evolve to fibrosis and then to cirrhosis and hepatocellular carcinoma (HCC).
The environment is not the only predisposing factor to NAFLD which has a strong hereditable component and we previously demonstrated that hepatic fat is the main driver of the progression to end-stage liver damages in genetically predisposed individuals [1]. In the last decade, the rs738409 (p.I148M; c.C444G) missense variation in Patatin-like phospholipase domain-containing 3 (PNPLA3) has been consistently associated with the entire spectrum of NAFLD in different populations-based studies (2–4). PNPLA3 is the strongest genetic predictor of NAFLD, showing a high prevalence among NASH patients of whom $34\%$ carry the mutant allele in homozygous. Heterozygous CG carriers have showed a prevalence of NASH and fibrosis higher than CC patients but lower than those carrying the GG genotype. Moreover, the homozygosity for PNPLA3 risk allele has been associated with more than 2-fold greater risk to develop NASH and cirrhosis, with up to a 12-fold increased risk for HCC and with an 18-fold increase in liver-related mortality [5, 6], thus suggesting a dose-dependent allele risk.
PNPLA3 protein localizes on lipid droplets (LD) surface and has triacylglycerol hydrolase and acyltransferase activities, which promote LDs remodeling in hepatocytes and hepatic stellate cells [7]. The 148M substitution leads to a reduced fatty acid hydrolysis yielding an impaired mobilization of triglycerides (TG) which accumulate in the liver [8, 9]. It has been showed that the methionine-isoleucine substitution at residue 148 in the PNPLA3 protein does not impact on the orientation of the catalytic dyad. However, the longer side chain of methionine blocks the access of fatty acids to the catalytic site, thus hampering PNPLA3 hydrolytic activity [10]. Kumari et al. demonstrated that the I148M polymorphism increases the hepatic lipogenic activity and TG synthesis [7]. Consistently, in a murine model, the overexpression of the mutated PNPLA3 (I148M) in the liver increased fatty acids and TG formation, impaired TAG hydrolysis and weakened the remodeling of TAG long-chain polyunsaturated fatty acids (PUFAs) [11]. Moreover, it has been demonstrated that I148M variant disrupts ubiquitylation and proteasomal degradation of PNPLA3 thus resulting in the accumulation of mutated protein on lipid droplets and impaired lipids mobilization [12]. Finally, it has been shown that carbohydrates may induce the accumulation of mutant PNPLA3 on LD surface thus worsening hepatic fat content [13].
To date there are no approved pharmacological treatments for the management of NAFLD and the clinical recommendations rely on lifestyle changes including daily exercise and healthy diet. Several drugs (anti-inflammatory, anti-fibrotic agents, and metabolic modulators), partially improving NASH activity and fibrosis, are under investigations even though they have produced marginally positive results [14]. Although advanced stages of NAFLD are irreversible, isolated hepatic steatosis and early NASH offer a “therapeutic window” for targeted interventions based on nutritional and lifestyle modifications. Indeed, effective and sustained weight loss has been associated with marked improvement in glycemic control, hepatic insulin sensitivity, transaminases and liver histology [15].
It has been demonstrated that the response to diet may differ accordingly to the individual genetic background [15]. A crosstalk between PNPLA3 rs738409 variant and nutrition has been already assessed. A nutrigenetic analysis revealed that the hepatic fat fraction in GG carriers is strongly influenced by carbohydrates and dietary sugar whose consumption may induce sterol-regulatory element binding protein-1C (SREBP1c) and, in turn, the expression of PNPLA3 mutated protein thus exacerbating fat deposition [16]. In addition, hepatic fat accumulation can be modulated by the interaction between PNPLA3 I148M variant and dietary omega 6/omega 3 PUFAs and diet supplemented with n-3 respect to n-6 PUFA could provide a targeted therapy in NAFLD subjects who are homozygous for the PNPLA3 G allele [17].
It has also been demonstrated that a higher intake of several micronutrients including niacin (nicotinic acid or vitamin B3) could modulate hepatic steatosis. Pharmacological doses of niacin have favorable effects on lipid parameters, increasing high-density lipoprotein cholesterol (HDL-C) and decreasing low-density lipoprotein cholesterol (LDL-C), TG and lipoprotein (a) [18]. Niacin administration in Sprague–Dawley rats fed high-fat diet significantly reduced hepatic and serum TG, lipid peroxidation thus ameliorating steatosis [19]. In HepG2 cells treated with palmitic acid (PA) to mimic steatosis, niacin supplementation reduced the expression of acyl-CoA diacylglycerol acyltransferase 2 (DGAT2) responsible for the committing step of TG synthesis, the production of ROS and inflammation by inhibiting interleukin-8 (IL-8) [20]. Li et al. revealed that DGAT2 inhibition reduced nuclear localization of SREBP1c, which is involved in de novo lipogenesis (DNL) and in the transcriptional regulation of PNPLA3, thus providing a possible mechanism through which niacin could ameliorate hepatic fat accumulation [21]. Additionally, niacin may improve steatosis by inhibiting DNL through the activation of ERK$\frac{1}{2}$/AMPK/SIRT1 pathway. SIRT1 is involved in the transcriptional regulation of SREBP1c paralleling the epigenetic regulation of PNPLA3 gene and together with the NAD driven-AMPK activation downregulate DNL, suggesting a possible gene-nutrient interplay and providing another strategy across which niacin could exert its beneficial role [22, 23].
Consistently, patients with hypertriglyceridemia and treated with niacin (Niacin ER, trade name: Niaspan) showed a significant reduction of liver and visceral fat whereas the DGAT2 rs3060 and rs101899116 variants were associated with a smaller decrease in liver fat content in response to niacin [24]. Finally, Linder et al. evaluated whether dietary niacin intake predicts change of liver fat content during a lifestyle intervention. Among fat compartments, the hepatic one showed the largest decrease and about half of NAFLD patients reached a steatosis resolution thus suggesting that niacin-fortified foods may represent a valuable strategy to treat the disease, taking into consideration the individual genetic background [25].
Therefore, this study aimed to assess the dietary and circulating niacin levels in NAFLD patients stratified according to the presence of the I148M variant which represents the stronger genetic predictor related to steatosis whose impact on fat deposition may be modulated by nutrients. Moreover, we examined the efficacy of niacin in reducing fat accumulation by exploiting Hep3B and HepG2 cells, which are wild-type and homozygous for the I148M mutation, respectively.
## Discovery cohort
We enrolled patients affected by NAFLD ($$n = 172$$; Discovery cohort), of whom the presence of steatosis was non-invasively evaluated through ultrasound echography using a convex 3.5 MHz probe and by FibroScan®, at the Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico, Milano. FibroScan® is a rapid and painless ultrasound (US)-based technique which emits low-frequency (50 mHz) vibrations into the liver, creating a propagating shear wave. This is detected by a pulse-echo acquisition that calculates its velocity which, in turn, is proportional to the stiffness of the tissue passed through. The Controlled attenuation parameter (CAP) value ≥ 248 estimated the presence of liver steatosis, whereas the liver stiffness measurement (LSM) value ≥ 7.0 and ≥ 6.2 kPa defined a significant liver fibrosis [26, 27]. Patients were genotyped to assess the presence of the rs738409 C > G (p.I148M) PNPLA3 variant as previously described (28–30), and the population was consistent with Hardy-*Weinberg equilibrium* ($$p \leq 0.13$$). Moreover, in order to assess dietary habits, we requested patients to carefully compile a food diary for 3 weeks. Kilocalories (kcal) of micro- and macro-nutrients were calculated with MètaDieta software1 and are listed in Supplementary Table 1. Dietary products containing niacin are shown in Supplementary Table 2. Demographic, anthropometric and clinicopathological features of the Discovery cohort are shown in Supplementary Table 3.
## Validation cohort
The Validation cohort includes 358 NAFLD unrelated patients of European descent who were consecutively enrolled at the Metabolic Liver Diseases outpatient service at Fondazione IRCCS Cà Granda, Ospedale Maggiore Policlinico, Milan, Italy. Inclusion criteria were the availability of liver biopsies performed for suspected NASH or severe obesity, DNA samples, and clinical data. Individuals with excessive alcohol intake (men, > 30 g/day; women, > 20 g/day), viral and autoimmune hepatitis, or other causes of liver disease were excluded. Patients were stratified according to both the presence of the rs738409 C > G (p.I148M) PNPLA3 variant and the population was not consistent with Hardy-*Weinberg equilibrium* ($$p \leq 0.03$$). The study conformed to the Declaration of Helsinki and was approved by the Institutional Review Board of the Fondazione Ca’ Granda IRCCS of Milan and relevant Institutions. All participants gave written informed consent. Patients’ genotyping and histological evaluation are presented in the Supplementary materials and methods. Demographic, anthropometric and clinicopathological features of the Discovery cohort are shown in Supplementary Table 4.
## Transcriptomic cohort
RNA-seq was performed in a subset of 183 severely obese patients (31 without and 152 with NAFLD) in whom a percutaneous liver biopsy was performed during bariatric surgery at Fondazione IRCCS Cà Granda, Ospedale Policlinico, Milan, Italy [31]. The study was conformed to the Declaration of Helsinki and approved by the Institutional Review Boards and their Ethics Committees. All participants gave written informed consent. Clinical characteristics of patients of whom RNA-seq data was available are presented in Supplementary Table 5. RNA-seq mapping descriptive statistics, detailed protocol, data analysis approach, patients’ genotyping and histological assessment are described in the Supplementary materials and methods.
## Measurement of circulating niacin
Niacin levels were evaluated in sera of both Discovery ($$n = 172$$) and Validation ($$n = 358$$) cohorts, collected at the time of NAFLD diagnosis. Niacin was quantified through the “ID-Vit Niacin” assay (Immunodiagnostik AG., Germany), based on a microbiological method which measures the total free niacin contained in the serum. The assay exploits microtiter plates covered with Lactobacillus plantarum, which uses niacin to grow. Serum samples were incubated at 37°C for 48 h and then turbidity was measured at λ = 610–630 nm with a spectrophotometer. The amount of niacin in the serum was directly proportional to the bacterial growth.
## In vitro models and treatments
To investigate the possible interaction between PNPLA3 and niacin metabolism, we compared two human hepatoma cell lines with a different genetic background (Hep3B, HepG2) and commonly used to study liver metabolism in vitro. The Hep3B cells are wild-type for the PNPLA3 gene, while the HepG2 cells carry the rs738409 C > G (I148M) PNPLA3 polymorphism in homozygosity. Both cell lines were cultured in Dulbecco’s modified eagle’s medium (DMEM) containing $10\%$ fetal bovine serum (FBS), 100 U/L penicillin, 100 U/L streptomycin and $1\%$ L-glutamine (Life Technologies-ThermoFisher Scientific, Waltham, MA, USA) and maintained at 37°C and $5\%$ CO2. To mimic human steatosis, both cell types were exposed to PA at 0.25 mM, whereas to assess the niacin efficacy on fat accumulation in hepatocytes, they were supplemented with a mixture of PA and niacin (PA + NIA) at 0.5 mM for 24 hours. Treatments were freshly prepared and administered when appropriate.
## RNA interference
Hep3B cells were transiently transfected for 48 h by pooling three different target-specific siRNA oligo duplexes directed against the exons 2, 4, and 5 of the human PNPLA3 (siHep3B) in order to improve the gene-silencing efficiency and at a final concentration of 10 μM (MyBioSource, Inc., San Diego, CA, USA). Cyclophilin B (10 μM) was used as a scramble negative control (Horizon Discovery, Waterbeach, UK).
## PNPLA3 llentiviral overexpression
PNPLA3 was stably overexpressed in HepG2 cells through pLenti-C-mGFP lentiviral vector which were engineered to express a complete Open Reading Frame (ORF) fused with green fluorescent protein (GFP) tag (henceforth HepG2I148+). We seeded 3 × 104 cells in 24-well plates and incubated at 37°C and $5\%$ CO2 overnight. Multiplicity of infection (MOI) was set at 2.5 and the amount of lentiviral particles for the transduction were calculated according to the manufactures‘instruction (OriGene, Rockville, USA). Lentiviral particles were added to pre-warmed cultured media for 24 h. To introduce PNPLA3-GFP tagged protein, HepG2 cells were transduced with PNPLA3 Human Tagged ORF Clone Lentiviral Particle (OriGene, Rockville, USA) containing a molecular sequence which aligns with the PNPLA3 mRNA (gene accession number: NM_025225).
## Statistical analysis
For descriptive statistics, continuous variables were reported as mean and standard deviation or median and interquartile range for highly skewed biological variables. Variables with skewed distribution were logarithmically transformed before analysis. Differences between two groups were calculated with non-parametric Wilcoxon test, followed by post hoc pairwise comparison. One-way non-parametric ANOVA (Kruskal–Wallis) followed by post hoc Dunn’s multiple comparison test was used to compare multiple groups adjusted for the number of comparisons. P-values < 0.05 were considered statistically significant. Statistical analyses were performed using JMP 16.0 (SAS, Cary, NC) and Prism software (version 6, GraphPad Software).
## The PNPLA3 I148M variant modulates both dietary and serum niacin levels in NAFLD patients and more so in the presence of obesity
Genetic variations may modulate the response to therapeutic approaches and vitamins in NAFLD patients. Among the latter, niacin has been reported to protect against severe steatosis although its interplay with genetics remains to be elucidated. The assessment of food diary in 172 patients with a non-invasive assessment of NAFLD revealed that niacin was the micronutrient with the lowest dietary levels in carriers of the PNPLA3 I148M variation ($$p \leq 0.04$$ CG/GG vs CC; Supplementary Table 1). Therefore, in the attempt to investigate the impact of the rs738409 C > G PNPLA3 genotype on niacin metabolism, we assessed the vitamin absorption in the serum of NAFLD patients belonging to the Discovery cohort. Alimentary and circulating niacin levels were then correlated with clinical-pathological features of NAFLD subjects in order to evaluate a possible gene-environment interaction.
In the Discovery cohort, $\frac{114}{172}$ patients ($66.27\%$) had severe steatosis (grade 2-3) and $\frac{104}{172}$ ($60.46\%$) carried the PNPLA3 p.I148M variant (Supplementary Table 3). At bivariate analysis, niacin intake was lower in NAFLD subjects with steatosis ≥ 2 compared to those with low-grade or no steatosis ($p \leq 0.05$ at Wilcoxon, adj $$p \leq 0.02$$ vs steatosis < 2, Figure 1A) and in PNPLA3 CG/GG carriers ($p \leq 0.05$ at Wilcoxon, adj $$p \leq 0.02$$ vs PNPLA3 CC, Figure 1B).
**FIGURE 1:** *The PNPLA3 I148M variant affects alimentary and serum niacin levels in the discovery cohort. (A,B) Niacin intake was reduced in NAFLD patients with steatosis ≥ 2 and PNPLA3 CG/GG mutation at bivariate analysis (*p < 0.05 at Wilcoxon test, vs steatosis < 2 and vs PNPLA3 CC). (C) Circulating niacin levels were lower in PNPLA3 I148M carriers at bivariate analysis compared to non-carriers (*p < 0.05 at Wilcoxon test). (D) Negative correlation between serum niacin concentration and body mass index (BMI). (E,F) Bivariate analysis shows the trend of dietary niacin intake in NAFLD patients stratified according to both the PNPLA3 G at-risk allele and obesity. In panel (F), NAFLD patients were even subcategorized by the presence of steatosis < 2 and steatosis ≥ 2. (G,H) The PNPLA3 at-risk genotype affected niacin absorption in a subgroup of obese NAFLD subjects and with steatosis ≥ 2 (p = 0.005 at ANOVA, *p < 0.05 vs BMI < 30 CC; p = 0.04 at ANOVA, *p < 0.05 vs steatosis ≥ 2 with BMI < 30 CC).*
Similarly, serum niacin was reduced in individuals carrying the PNPLA3 CG/GG mutation ($p \leq 0.05$ at Wilcoxon, adj $$p \leq 0.03$$ vs PNPLA3 CC, Figure 1C) although no significant differences in its levels emerged among patients with steatosis < 2 and ≥ 2 (Supplementary Figure 1A), possibly suggesting that the presence of the rs738409 C > G PNPLA3 genotype rather than hepatic fat accumulation may influence niacin absorption or metabolism. Moreover, at correlation analysis, serum niacin negatively associated with body mass index (BMI) ($p \leq 0.01$, Figure 1D) supporting the hypothesis that niacin availability may be even swayed by environmental risk factors as obesity.
Therefore, to deepen the impact of obesity on niacin metabolism, we stratified the Discovery cohort according to BMI ≥ 30 kg/m2 and to the presence of the I148M variant. We found that dietary niacin showed a trend of reduction in patients carrying the PNPLA3 at-risk G allele and BMI ≥ 30 kg/m2 (Figure 1E) whereas by stratifying patients according to the presence of severe steatosis the lower levels of niacin in subjects with the PNPLA3 CG/GG mutation didn’t vary across BMI (Figure 1F). We next assessed circulating niacin and we observed that its serum levels tended to be lower in G allele carriers and the reduction became significant in those with BMI ≥ 30 kg/m2 ($$p \leq 0.005$$ at ANOVA, adj $$p \leq 0.03$$ vs BMI < 30 CC, Figure 1G). This effect was highly emphasized in the subcategory of patients with steatosis ≥ 2 ($$p \leq 0.04$$ at ANOVA, adj $$p \leq 0.03$$ vs steatosis ≥ 2 with BMI < 30 CC, Figure 1H), probably due to the additive weight of PNPLA3 I148M variant and obesity in niacin absorption.
*At* generalized linear model adjusted for sex, age, steatosis ≥ 2 and alimentary niacin, the presence of the PNPLA3 CG/GG mutation was independently associated with serum niacin in NAFLD individuals with BMI ≥ 30 kg/m2 (β = –0.21, $95\%$ CI: –0.41–0.004, $$p \leq 0.04$$, Table 1), thereby supporting that PNPLA3 p.I148M aminoacidic substitution may be directly involved in the alteration of systemic niacin availability and that the effect attributable to this mutation may be amplified by adiposity.
**TABLE 1**
| Unnamed: 0 | Circulating niacin (μ g/μ L) | Circulating niacin (μ g/μ L).1 | Circulating niacin (μ g/μ L).2 | Circulating niacin (μ g/μ L).3 | Circulating niacin (μ g/μ L).4 | Circulating niacin (μ g/μ L).5 |
| --- | --- | --- | --- | --- | --- | --- |
| | BMI < 30 | BMI < 30 | BMI < 30 | BMI ≥ 30 | BMI ≥ 30 | BMI ≥ 30 |
| | β | 95% CI | P-value* | β | 95% CI | P-value* |
| Sex, M | –0.02 | –0.14 to 0.10 | 0.73 | 0.03 | –0.19 to 0.26 | 0.74 |
| Age, years | –0.006 | –0.014 to 0.002 | 0.15 | 0.0002 | –0.01 to 0.01 | 0.97 |
| Steatosis ≥ 2, yes | –0.03 | –0.12 to 0.07 | 0.57 | –0.12 | –0.31 to 0.07 | 0.21 |
| Dietary niacin (kcal) | –0.12 | –0.29 to 0.04 | 0.15 | –0.19 | –0.44 to 0.05 | 0.12 |
| PNPLA3 G allele, yes | –0.03 | –0.13 to 0.06 | 0.46 | –0.21 | –0.41 to 0.004 | 0.04 |
## The additive effect of the PNPLA3 I148M mutation and obesity impacts on circulating niacin in patients with biopsy-proven NAFLD
To validate the results obtained in the Discovery cohort, we measured serum niacin in a larger cohort including $$n = 358$$ subjects with histological assessment of NAFLD, of whom dietary niacin was not available (Validation cohort). In this cohort, steatosis ≥ 2 was diagnosed in $\frac{232}{358}$ patients ($64.80\%$), whereas obesity (BMI ≥ 30 kg/m2) and the PNPLA3 I148M variant were identified in $\frac{194}{358}$ ($54.18\%$) and $\frac{255}{358}$ ($71.2\%$), respectively (Supplementary Table 4). Circulating niacin concentration was inversely correlated with BMI ($p \leq 0.0001$, Figure 2A) and the lowest levels were observed in individuals affected by obesity and carrying the PNPLA3 CG/GG mutation thus confirming the results obtained in the Discovery cohort ($$p \leq 0.0001$$ at ANOVA, adj $p \leq 0.05$ vs PNPLA3 CC with either BMI < 30 or BMI ≥ 30; adj $p \leq 0.01$ vs PNPLA3 CG/GG with BMI < 30, Figure 2B).
**FIGURE 2:** *Obesity heightens PNPLA3 genetic risk on niacin absorption in the validation cohort. (A) Negative correlation between serum niacin concentration and body mass index (BMI). (B,C) The presence of the PNPLA3 I148M variant combined with obesity reduced niacin absorption in biopsied NAFLD subjects, independently of steatosis severity (p = 0.004 at ANOVA, *p < 0.05 vs PNPLA3 CG/GG with/without obesity and vs PNPLA3 CC with BMI ≥ 30). p = 0.0001 at ANOVA, **p < 0.01 vs. PNPLA3 CG/GG without obesity.*
Moreover, serum niacin levels were decreased in presence of both obesity and the PNPLA3 G risk allele, thereby resembling what observed in the Discovery cohort although we didn’t observe any variance between patients with steatosis < 2 and ≥ 2 ($$p \leq 0.004$$ at ANOVA, adj $p \leq 0.05$ vs PNPLA3 CG/GG with/without obesity and vs PNPLA3 CC with BMI ≥ 30, Figure 2C). In addition, no relevant differences were found in serum niacin levels by stratifying patients according to the severity of necroinflammation, fibrosis and NAS (Supplementary Figures 1B–D), suggesting that whole spectrum of NAFLD per se may not influence niacin absorption.
At multivariate analysis adjusted for sex, age and steatosis ≥ 2 the association between the PNPLA3 I148M variant and lower serum niacin remained strongly significant in NAFLD individuals who belonged to the Validation cohort with BMI ≥ 30 kg/m2 (β = –0.24, $95\%$ CI: –0.43–0.06, $$p \leq 0.009$$, Table 2), thus corroborating the hypothesis that the PNPLA3 p.I148M missense variation may be a genetic modifier of vitamin B3 metabolism.
**TABLE 2**
| Unnamed: 0 | Circulating niacin (μg/μL) | Circulating niacin (μg/μL).1 | Circulating niacin (μg/μL).2 | Circulating niacin (μg/μL).3 | Circulating niacin (μg/μL).4 | Circulating niacin (μg/μL).5 |
| --- | --- | --- | --- | --- | --- | --- |
| | BMI < 30 | BMI < 30 | BMI < 30 | BMI ≥ 30 | BMI ≥ 30 | BMI ≥ 30 |
| | β | 95% CI | P-value* | β | 95% CI | P-value* |
| Sex, M | –0.02 | –0.11 to 0.06 | 0.60 | –0.02 | –0.22 to 0.17 | 0.80 |
| Age, years | 0.0001 | –0.006 to 0.007 | 0.96 | 0.0002 | –0.01 to 0.015 | 0.97 |
| Steatosis ≥ 2, yes | 0.07 | –0.009 to 0.16 | 0.08 | –0.09 | –0.26 to 0.08 | 0.29 |
| PNPLA3 G allele, yes | –0.02 | –0.11 to 0.06 | 0.56 | –0.24 | –0.43 to 0.06 | 0.009 |
## The PNPLA3 I148M variation modulates hepatic enzymes of NAD metabolism in NAFLD patients
The results obtained in the Discovery and Validation cohorts have suggested that niacin availability, the primary source of NAD synthesis, may be affected in patients with the PNPLA3-driven genetic predisposition to develop NAFLD and more so in obese ones. In order to evaluate whether the presence of the PNPLA3 I148M variant may even interfere with hepatic NAD metabolism, we assessed the expression of genes involved in NAD biosynthetic pathways as well as NAD/NADH-dependent enzymes through RNA-seq analysis performed in 183 biopsied NAFLD patients (Transcriptomic cohort) who underwent bariatric surgery.
At bivariate analysis, the hepatic nicotinate phosphoribosyl-transferase 1 (NAPRT1) mRNA levels, the main enzyme involved in niacin conversion into NAD precursors, were lower in carriers of the PNPLA3 G allele compared to wild-type group ($p \leq 0.01$ at Wilcoxon, adj $$p \leq 0.0026$$ vs PNPLA3 CC, Figure 3A). Conversely, the expression of NAD synthetase 1 (NADSYN1) and nicotinamide phosphoribosyl-transferase (NMNAT1), alternatively producing NAD from tryptophan and the salvage pathway, was increased in I148M PNPLA3 carriers compared to non-carriers ($p \leq 0.05$ at Wilcoxon, adj $$p \leq 0.01$$ and adj $$p \leq 0.03$$,vs PNPLA3 CC, respectively, Figures 3B, C) possibly due to a compensatory mechanism to provide NAD in the liver.
**FIGURE 3:** *The PNPLA3 G allele impairs hepatic NAD metabolism in the transcriptomic cohort independently of steatosis severity. (A–C) Hepatic NAPRT1 mRNA levels were reduced in the presence of the PNPLA3 CG/GG mutation (**p < 0.01 at Wilcoxon test vs PNPLA3 CC), while NADSYN1 and NMNAT1 expression were upregulated in PNPLA3 CG/GG carriers (*p < 0.05 at Wilcoxon test vs PNPLA3 CC). (D–F) Hepatic NAPRT1, NADSYN1, and NMNAT1 expression were modulated in response to PNPLA3 CG/GG mutation, but regardless of steatosis grade (NAPRT1: p = 0.0004 at ANOVA, **p < 0.01 vs PNPLA3 CC with steatosis < 2; *p < 0.05 vs PNPLA3 CC with steatosis ≥ 2; NADSYN: p = 0.0001 at ANOVA, *p < 0.05 vs PNPLA3 CC with steatosis < 2 and ≥ 2; NMNAT1: p = 0.0001 at ANOVA, **p < 0.01 PNPLA3 CC with steatosis < 2, *p < 0.05 vs PNPLA3 CC with ≥ 2). (G–L) The mRNA levels of NAD-utilizing complexes (MDH1/2. IDH3B, PDHA1/B and OGDB) were decreased in presence of the PNPLA3 G at-risk allele (*p < 0.05 and **p < 0.01 at Wilcoxon vs PNPLA3 CC).*
Similarly to what observed in the Discovery and Validation cohorts stratified according to steatosis grade, the mRNA levels of NAPRT1, NADSYN1 and NMNAT1 were not affected by the severity of steatosis ($$p \leq 0.0004$$ at ANOVA, adj $$p \leq 0.003$$ and $$p \leq 0.02$$ PNPLA3 CG/GG with steatosis < 2 and ≥ 2 vs PNPLA3 CC with steatosis < 2 and ≥ 2, respectively; Figures 3D; $$P \leq 0.0001$$ at ANOVA, adj $$p \leq 0.004$$ and $$p \leq 0.03$$ PNPLA3 CG/GG with steatosis < 2 and ≥ 2 vs PNPLA3 CC with steatosis < 2 and ≥ 2, respectively Figures 3E, F), supporting that the PNPLA3 genetic variant more than hepatic fat accumulation influences the niacin-dependent NAD metabolism.
Moreover, the hepatic mRNA levels of the NAD/NADH-dependent enzymes as the malate dehydrogenase $\frac{1}{2}$ (MDH$\frac{1}{2}$), the isocitrate dehydrogenase [NAD] subunit beta (IDH3B), the pyruvate dehydrogenase E1 subunits alpha A1 and beta (PDHA1/B) and the oxoglutarate dehydrogenase (OGDH) were significantly reduced in carriers of the PNPLA3 G allele (adj $p \leq 0.05$ and adj $$p \leq 0.0001$$ at Wilcoxon vs PNPLA3 CC genotype, Figures 3G–L), suggesting that the rs738409 C > G PNPLA3 at-risk genotype may affect vitamin B3 metabolism by modulating the expression of NAD/NADH-consuming genes.
## The PNPLA3 loss-of-function impairs NAD metabolism in hepatoma cells
Evidence in NAFLD patients highlighted that niacin availability and the hepatic NAD biosynthesis are altered by the presence of the I148M polymorphism. Therefore, to explore the possible interaction between PNPLA3 and niacin metabolism, we compared NAD biosynthetic rate in Hep3B and HepG2 cells, which carried the PNPLA3 CC and GG genotype, respectively. Moreover, we induced the PNPLA3 silencing in Hep3B cells (siHep3B) in order to evaluate whether its loss-of-function may impair NAD production. Finally, we overexpressed the PNPLA3 *Wt* gene (HepG2I148+) in HepG2 cells attempting to elucidate whether the re-introduction of the PNPLA3 Wt form may restore vitamin B3 efficacy in reducing fat accumulation.
As expected, PNPLA3 mRNA and protein levels were reduced in siHep3B by around $50\%$ ($p \leq 0.05$ and $p \leq 0.01$ vs scramble, Figure 4A), while they were significantly increased after the lentiviral transduction in HepG2I148+ cells ($p \leq 0.01$ vs HepG2, Figure 4B).
**FIGURE 4:** *PNPLA3 loss-of-function dampens NAD synthesis in hepatocytes after niacin administration. (A,B) PNPLA3 mRNA and protein levels were assessed by qRT-PCR and Western blot, respectively, in hepatoma cells. (C) NAD content was assessed in Hep3B, siHep3B, HepG2 and HepG2I148+ cells through NAD/NADH Colorimetric/Fluorometric Assay Kit at both baseline and after niacin exposure. (D–F)
NAPRT1, NADSYN and NMNAT1 mRNA levels evaluated by qRT-PCR in hepatoma cells (Hep3B, siHep3B, HepG2 and HepG2I148+) with or without niacin treatment. For gene expression, data were normalized to ACTB housekeeping gene and expressed as fold increase (Arbitrary Unit-AU) compared to control group. For Western blot, data were normalized to vinculin housekeeping protein and expressed as fold increase (AU) compared to control group. For violin plot, data were expressed as median concentration (thick dashed lines) and interquartile range (dotted lines). Adjusted *p < 0.05 and **p < 0.01.*
At baseline, Hep3B cells showed higher NAD content compared to siHep3B ones ($p \leq 0.01$ vs siHep3B; Figure 4C), indicating that the PNPLA3 deficiency may affect NAD production. Likewise, HepG2 cells exhibited lower NAD concentration, matching the levels measured in siHep3B cells, whereas in HepG2I148+ model it was comparable to Hep3B. After niacin exposure, NAD concentration increased in Hep3B cells but not in siHep3B ones ($p \leq 0.05$ vs Hep3B untreated, Figure 4C) supporting that PNPLA3 silencing may modify the response to niacin supply. Furthermore, intracellular NAD content remain unchanged in HepG2 cells after niacin administration, showing a similar range of that observed in siHep3B ones, while it was increased in HepG2I148+ model ($p \leq 0.05$ vs HepG2I148+ untreated, Figure 4C).
Furthermore, in HepG2 cells the NAPRT1 expression was reduced compared to both HepG2I148+ and Hep3B Wt models ($p \leq 0.01$ and $p \leq 0.05$ vs HepG2I148+, $p \leq 0.01$ and $p \leq 0.05$ vs Hep3B; Figures 4C, D), thereby sustaining that the PNPLA3 I148M variation may be involved in the impairment of the canonical NAD synthesis. Consistently with the increased NAD content upon niacin treatment, NAPRT1 expression was induced in Hep3B cells ($p \leq 0.05$ vs Hep3B untreated, Figure 4D), whereas it was not modified in siHep3B ones. Similarly, HepG2 cells didn’t show an increment of NAPRT1 mRNA levels after niacin exposure, while its expression was slightly induced in the PNPLA3 overexpressed cells, resembling the results obtained in Hep3B.
Moreover, both siHep3B and HepG2 cells showed higher basal expression of NADSYN1 and NMNAT1 than Hep3B and HepG2I148+ cells ($p \leq 0.01$ vs Hep3B; $p \leq 0.05$ vs HepG2I148+; Figures 4E, F), possibly to compensate NAD shortage and corroborating the results obtained in the Transcriptomic cohort.
As concern the alternative pathways of NAD biosynthesis, both Hep3B and HepG2I148+ displayed high NADSYN1 and NMNAT1 mRNA levels after niacin exposure ($p \leq 0.05$ vs HepG2I148+ untreated, Figures 4E, F). Contrariwise, siHep3B and HepG2 cells reduced the expression of NADSYN1 and NMNAT1 enzymes (NADSYN1: $p \leq 0.05$ vs HepG2 untreated, Figure 4E; NMNAT1: $p \leq 0.01$ vs siHep3B untreated, Figure 4F) probably due to a negative feedback loop exerted by niacin or its metabolites on the alternative pathways.
Thus, these findings may support that the presence of the PNPLA3 loss-of-function induced by the silencing or the I148M variant may impair the canonical via of NAD biosynthesis at both baseline and after niacin supplementation. In support of this hypothesis, our in vitro results have suggested that the re-establishment of the PNPLA3 functional protein seems to rescue the vitamin B3 metabolism in hepatocytes, thereby sustaining the link between PNPLA3 and NAD availability.
## The PNPLA3 loss-of-function mitigates the beneficial role of niacin in reducing triglycerides synthesis
It has been previously demonstrated that niacin improves hepatic steatosis by downregulating DGAT2 and reducing TG synthesis which, in turn, lead to a decreased oxidative stress [19, 20]. In order to evaluate whether the presence of the I148M polymorphism may disturb niacin effectiveness on intracellular fat content, we treated hepatoma cells with PA alone or combined with NIA for 24 h.
At ORO staining, both Hep3B and HepG2I148+ cells accumulated less lipid droplets in response to PA administration rather than siHep3B and HepG2 cell lines, possibly due to the efficient PNPLA3 hydrolytic activity (Figure 5A). PA exposure enhanced the intracellular TG content and induced DGAT2 upregulation in all experimental models ($p \leq 0.01$ vs Hep3B, siHep3B, HepG2 and HepG2I148+ untreated, Figures 5B, C). However, siHep3B and HepG2 cells exhibited an exacerbated lipid accumulation and DGAT2 induction as a consequence of low lipid clearance induced by PNPLA3 silencing or I148M variant, respectively ($p \leq 0.01$ and $p \leq 0.05$ vs Hep3B + PA or vs HepG2I148+ + PA, Figures 5B, C).
**FIGURE 5:** *(A) Evaluation of LDs formation was assessed in hepatoma cells (Hep3B, siHep3B, HepG2, and HepG2I148+) after PA challenge and NIA treatment by ORO staining (100× magnification). Nuclei were counterstained by hematoxylin. (B) TG content was measured in cell lysates (Hep3B, siHep3B, HepG2, and HepG2I148+) through Triglycerides Colorimetric/Fluorometric Assay Kit. (C)
DGAT2 mRNA levels were quantified in hepatoma cells with or without niacin treatment after PA challenge (Hep3B, siHep3B, HepG2, and HepG2I148+) by qRT-PCR. For gene expression, data were normalized to ACTB housekeeping gene and expressed as fold increase (Arbitrary Unit-AU) compared to control group. For violin plot, data were expressed as median concentration (thick dashed lines) and interquartile range (dotted lines). Adjusted *p < 0.05 and **p < 0.01.*
After niacin administration, lipid overload was reduced in all in vitro models, albeit this effectiveness was mitigated in siHep3B and HepG2 cells, supporting that the presence of a non-functional PNPLA3 protein may interfere with niacin protective role (Figure 5A). In keeping with the ORO staining, niacin administration strongly reduced the intracellular TG content ($p \leq 0.01$ vs Hep3B + PA, siHep3B + PA, HepG2 + PA and HepG2I148+ + PA, Figure 5B) and the mRNA levels of DGAT2 in all cell lines ($p \leq 0.05$ vs Hep3B + PA and HepG2 + PA, $p \leq 0.01$ vs siHep3B and HepG2I148+ + PA, Figure 5C), showing the greatest effect in Hep3B and HepG2I148+ models (Figures 5B, C) and sustaining that the presence of PNPLA3 in the WT form may improve the niacin efficacy on fat overload clearance.
Consistently with the worsened fat accumulation, siHep3B and HepG2 cells showed higher ER-oxidative injury compared to those induced by PA in Wt cellular models, by increasing the mRNA levels of Activating Transcription Factor 4-6 (ATF4, ATF6) and Glucose-regulated protein 78 (GRP78) and enhancing the production of hydrogen peroxide (H2O2) and malondi- aldehyde (MDA) ($p \leq 0.05$ and $p \leq 0.01$ vs Hep3B + PA and HepG2I148+ + PA; Supplementary Figures 2A–E).
Moreover, we found that niacin treatment strongly counteracted the negative effects of PA on ATF4 ($p \leq 0.05$ vs PA, $p \leq 0.01$ vs PA, Supplementary Figure 2A), ATF6 ($p \leq 0.05$ vs PA, $p \leq 0.01$ vs PA, Supplementary Figure 2B), GRP78 expression ($p \leq 0.01$ vs PA, Supplementary Figure 2C) and oxidative injury (H2O2: $p \leq 0.05$ vs PA, $p \leq 0.01$ vs PA, Supplementary Figure 2D; MDA: $p \leq 0.05$ vs PA, $p \leq 0.01$ vs PA, Supplementary Figure 2E) in all the in vitro models. Although these findings have suggested that the effect of niacin in reducing TG synthesis may differ accordingly to the PNPLA3 genetic background in our experimental models, we could speculate that the impact on hepatocellular toxicity may be more a consequence of the reduced fat accumulation rather than dependent by PNPLA3.
## The PNPLA3 loss-of-function promotes de novo lipogenesis by altering niacin-induced ERK1/2/AMPK/SIRT1 pathway
Few studies underlined that niacin could inhibit DNL through the activation of extracellular regulated kinase $\frac{1}{2}$, AMP-activated protein kinase and sirtuin1 (ERK$\frac{1}{2}$/AMPK/SIRT1) pathway [20, 32]. SIRT1, whose activity is dependent of NAD+ availability, is involved in the transcriptional regulation of SREBP-1c and in the epigenetic regulation of PNPLA3 gene by nutritional factors [33]. Therefore, we investigated whether niacin beneficial effects on DNL may be even affected by the loss of PNPLA3 hydrolytic activity.
In Hep3B and HepG2I148+ cells, niacin exposure promoted a marked phosphorylation of ERK$\frac{1}{2}$ (pERK$\frac{1}{2}$; $p \leq 0.05$ vs Hep3B + PA, $p \leq 0.01$ vs HepG2I148+ + PA, Figures 6A, B) and AMPK (pAMPK; $p \leq 0.01$ vs Hep3B + PA, $p \leq 0.01$ vs HepG2I148+ + PA, Figures 6A–C), while they were mildly activated in siHep3B and HepG2 models (Figures 6A–C). Consistently, SIRT1 mRNA levels were upregulated in Hep3B and HepG2I148+ cells ($p \leq 0.05$ vs Hep3B + PA, $p \leq 0.05$ vs HepG2I148+ + PA, Figure 6D), whereas its expression was not significantly changed between siHep3B and HepG2 models at baseline and after treatment with PA or PA + NIA (Figure 6D).
**FIGURE 6:** *The PNPLA3-niacin crosstalk may occur via ERK1/2/AMPK/SIRT1 pathway. (A–C) Phospho-p44/42 MAPK (ERK1/2) (Thr202/Tyr204), p44/42 MAPK (ERK1/2), Phospho-AMPK and AMPK protein levels were evaluated by Western blot in hepatoma cells (Hep3B, siHep3B, HepG2 and HepG2I148+) after PA challenge and niacin treatment. (D–H)
SIRT1, ACC, FASn, SREBP1 and PNPLA3 mRNA levels were assessed in hepatoma cells by qRT-PCR. (I) Schematic figure of the putative interplay between the canonical pathway of NAD biosynthesis, which is dependent of niacin availability, and PNPLA3 genotype. For gene expression, data were normalized to ACTB housekeeping gene and expressed as fold increase (Arbitrary Unit-AU) compared to control group. For Western blot, data were normalized to vinculin housekeeping protein and expressed as fold increase (AU) compared to control group. Adjusted *p < 0.05 and **p < 0.01.*
Consequently to the induction of SIRT1, Hep3B and HepG2I148+ cells showed a huge downregulation of genes involved in DNL after niacin administration, as acetyl-CoA carboxylase (ACC, $p \leq 0.05$ vs Hep3B + PA, $p \leq 0.05$ vs HepG2I148+ + PA, Figure 6E), fatty acid synthase (FASn, $p \leq 0.05$ vs Hep3B + PA, $p \leq 0.05$ vs HepG2I148+ + PA, Figure 6F), and SREBP1 ($p \leq 0.05$ vs Hep3B + PA, $p \leq 0.05$ vs HepG2I148+ + PA, Figure 6G). In contrast, siHep3B and HepG2 cells showed a mild or no reduction DNL-related genes after niacin exposure (Figures 6E–G), thereby strengthening the theory that niacin efficacy on lipid clearance may be impaired by a non-functional PNPLA3 protein.
In keeping with SREBP1 downregulation, the mRNA levels of PNPLA3 were reduced after niacin supplementation in Hep3B and HepG2I148+ models ($p \leq 0.01$ vs Hep3B + PA, $p \leq 0.01$ vs HepG2I148+ + PA, Figure 6H) but not in siHep3B and HepG2 ones, corroborating that the possible gene-nutrient interaction between PNPLA3 and vitamin B3 occurs via the ERK$\frac{1}{2}$/AMPK/SIRT1 signaling and the subsequent PNPLA3 transcriptional regulation mediated by SREBP1 (Figure 6I).
## Discussion
Environmental risk factors, among which obesity, and genetic variations such as the I148M PNPLA3 polymorphism strongly contribute to NAFLD pathogenesis. It has been established that PNPLA3 may be modulated by nutrients thus affecting the response to dietary interventions [34, 35]. Niacin has been proposed for NAFLD management as it reduces TG synthesis thus improving steatosis in both mice and patients [19, 20]. Therefore, we investigated the interplay between the I148M polymorphism and niacin absorption/metabolism in NAFLD patients and in vitro models.
We firstly examined food diary in 172 subjects with a non-invasive NAFLD diagnosis (Discovery cohort). We identified a cluster of micronutrients, mostly foods containing fibers and proteins, which appeared less ingested in I148M carriers, thereby offering the possibility to introduce them for a personalized nutritional intervention. Among them, niacin, enriched in fruits, vegetables, meat, and fish, resulted the least consumed. Accordingly, we found that serum niacin was lower in presence of I148M PNPLA3 variation independently of dietary niacin, suggesting that it may be implied in niacin systemic availability, possibly affecting its absorption and metabolism.
Notably, serum niacin was inversely correlated with body weight and the lowest levels were observed in patients carrying the I148M variant with a BMI ≥ 30 kg/m2 both in the Discovery and Validation cohorts, thus supporting a cumulative effect of obesity and the mutation in impairing niacin availability. A relation between niacin intake and BMI changes was previously described by Linder et al. who observed a higher niacin-dependent reduction of hepatic steatosis in NAFLD patients who lost weight during a period of diet counselling and physical activity [25]. Consistently, an increased diet-related adiposity was associated with an amplified PNPLA3 genetic risk of fatty liver. In particular, carbohydrates up-regulate mutant PNPLA3 on lipid droplets surfaces, thus hindering their hydrolysis [34].
To deepen whether the I148M variant may influence niacin metabolism, we evaluated the hepatic NAD biosynthesis and we found that NAPRT1, involved in the canonical NAD pathway, was reduced in subjects with the PNPLA3 G risk allele independently of steatosis grade, mirroring the low serum niacin levels observed in both Discovery and Validation cohorts. Conversely, the NADSYN1 and NMNAT1 mRNA levels, implicated in the alternative and recovery NAD signaling, respectively, increased in I148M carriers as a possible compensatory mechanism to provide hepatic NAD. Alterations of NAD metabolism have been previously associated with NAFLD although no evidence is available regarding the role of genetics in this pathway. According to our findings, Penke et al. have shown that the increased NAD salvage pathway is involved in hepatic steatosis and supplementation with NAD precursors may aid to attenuate disease progression by promoting the NAD+-dependent Sirt1 activation thus highlighting the importance to maintain a sufficient hepatic NAD availability [23].
Our findings in NAFLD patients have suggested that the I148M variation may possibly influence niacin canonical turnover in the liver. To assess the mechanisms through which the PNPLA3 loss-of-function impacts on niacin metabolism, we exploited Hep3B (Wt) and HepG2 (I148M) cells, in which we silenced PNPLA3 or overexpressed the Wt protein, respectively. Hep3B cells exhibited higher levels of NAD which, as expected, increased after niacin exposure. Conversely, siHep3B and HepG2 cells showed lower NAD concentration compared to Hep3B cells and its content did not increase after niacin administration, suggesting an aberrant response to niacin supplementation in presence of the PNPLA3 loss-of-function. The PNPLA3 Wt overexpression in HepG2I148+ model restored both basal and niacin-induced NAD production, suggesting the requirement of a functional PNPLA3 protein to rescue the NAD synthesis. In keeping with the increased NAD levels, niacin exposure promoted the canonical via by upregulating NAPRT1 mRNA in both Hep3B and HepG2I148+cells, but not in siHep3B and HepG2 ones in which we observed an upregulation of the alternative pathways of NAD biosynthesis, thus resembling the transcriptomic data. In sum, our results have suggested that the PNPLA3 loss-of-function inhibits NAD production from niacin paralleled by the increase of NAD-related alternative pathways.
Niacin can reduce fat accumulation by downregulating TG synthesis via DGAT2 in cell cultures, rodents, and humans [24, 36, 37]. DGAT2 inhibition was even associated with lower DNL due to less SREPB1-1c nuclear translocation and, consequently, transcriptional regulation of its lipogenic targets, among which PNPLA3 [21]. We found that niacin supplementation improved the intracellular fat content by targeting DGAT2 in all experimental models and it may be due to a combination of reduced TG synthesis and DNL. Such findings were consistent with those reported by Ganji et al. and Blond et al., who demonstrated that niacin treatment directly inhibits DGAT2 activity and reduced TG content in HepG2, HuH7 and primary hepatocytes, and with those of Li et al. who provided the link among DGAT2 inhibition and DNL [21, 24, 36, 37]. Another study has pointed out that 39 patients with dyslipidemia improved lipid profile, exhibited lower visceral/subcutaneous fat and ameliorated hepatic fat content after niacin treatment [24, 38]. Here, we showed that PNPLA3 deficiency may affect niacin efficacy on TG synthesis as the siHep3B and HepG2 cells displayed less reduction of hepatocellular TG compared to Hep3B and HepG2I148+ models.
Furthermore, it has been demonstrated that niacin reverses oxidative stress and inflammation, in cells and animal models, possibly due to its effect in clearing lipids [38]. We observed greater ER-oxidative stress in siHep3B and HepG2 cells than Hep3B and HepG2I148+ models after PA challenge, likely exacerbated by the presence of a non-functional PNPLA3 protein. Another mechanism by which niacin could reduce fat accumulation foresees the activation of the ERK$\frac{1}{2}$/AMPK/SIRT1 signaling [32, 39]. Several studies revealed that phospho-AMPK and treatments with NAD+ precursors promote SIRT1 activation, which, in turn, decrease DNL [22, 23, 32, 39]. Ye et al. have demonstrated that niacin treatment in obese HFD-fed mice hampered the transcriptional activity of SREBP1 through ERK$\frac{1}{2}$/AMPK activation with the consequent reduction of hepatic and plasma TG content [40]. We found that niacin promoted the ERK$\frac{1}{2}$ and AMPK phosphorylation paralleled by SIRT1 mRNA upregulation more in Hep3B and HepG2I148+ rather than siHep3B and HepG2 cells, suggesting that the PNPLA3 loss-of-function may reduce the inhibition of DNL induced by niacin. Consistently, niacin promoted a significant reduction of SREBP1, ACC and FASn expression in Hep3B and HepG2I148+ cells, but not in siHep3B and HepG2 models. Recently, it has been proven that the AMPK/SIRT1 may hamper the SREBP1 binding to PNPLA3 promoter [33] and as above mentioned SREBP1 is involved in the transcriptional regulation of PNPLA3 after carbohydrates loading [34]. We found that PNPLA3 mRNA levels were decreased in Hep3B and HepG2I148+ after niacin supplementation, and this effect may be ascribable to the low SREBP1 expression. Conversely, niacin did not alter PNPLA3 expression in HepG2 cells possibly suggesting that the crosstalk between PNPLA3 and niacin may occur through SREBP1.
To date, niacin was proposed at pharmacologic doses in the range of 1,500–2,000 mg (Niaspan®) for the treatment of dyslipidemia and prevention of cardiovascular complications [37]. A clinical trial carried out in 39 hyper-triglyceridemic patients with steatosis has showed a reduction of liver fat by $47\%$ and liver enzymes when treated with Niaspan for 6 months [37].
Lifestyle interventions have shown a great efficacy to improve hepatic steatosis and they currently represent a valid approach for NAFLD management. Preclinical studies have demonstrated that dietary intake of NAD precursors may ameliorate fatty liver by boosting the hepatic NAD metabolism. Accordingly, our study revealed a novel nutrigenetic regulation of the PNPLA3 gene by niacin and underlined how the genetic screening which is useful in terms of costs and non-invasiveness gains value for a personalized approach in NAFLD patients. By looking at a translational perspective, either the nutritional supplementation with niacin or the increased consumption of niacin-fortified foods may represent an alternative option to overcome the low niacin levels observed in genetically predisposed NAFLD patients, even more with the co-occurrence of obesity.
## Conclusions
The I148M PNPLA3 variant, the strongest genetic predictor of NAFLD onset and progression, may undergo nutritional regulation and its deleterious effect could be worsened by environmental factors as obesity. Niacin, belonging to Vitamin B class, has been suggested for NAFLD management as it reduces hepatic fat content and inflammation. In this study, we highlighted a potential interaction between the presence of mutant PNPLA3 and niacin metabolism. Dietary evaluation through food diary pointed out that NAFLD subjects, among which around $60\%$ carried the I148M variant and > $40\%$ was affected by obesity, exhibited an unbalanced diet with low intake of vitamins, fibers, and proteins. In patients, our results have supported that the PNPLA3 CG/GG genotype was associated with lower niacin availability in the serum and, at hepatic levels with altered expression of enzymes involved in NAD biosynthesis promoting more the alternative pathways than the canonical via. Even in hepatocytes, the presence of PNPLA3 loss-of-function limited the NAD production through the canonical pathway after niacin supplementation as well as it dampened niacin efficacy on fat accumulation and oxidative stress, thus sustaining the possible gene-nutrient crosstalk. In sum, vitamin B3 supplements or niacin-fortified foods should be recommended for NAFLD patients with a predisposing genetic background, amplified by adiposity.
## Data availability statement
The ethical approval of the study does not allow to publicly share individual patients’ genetic data. All data, code and materials used in the analysis are available upon reasonable request for collaborative studies regulated by materials/data transfer agreement (MTA/DTA) to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by Fondazione IRCCS Cà Granda CE: 164_2019. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
EP and ML: study design, data analysis and interpretation, and manuscript drafting. MMe: data analysis and interpretation and manuscript drafting. GT: data analysis. AC, RL, SB, and MMa: patients recruitment and characterization. AF: discussion and manuscript revision. PD: study design, manuscript drafting, data analysis and interpretation, study funding, supervision, and has primary responsibility for final content. All authors read and approved the final manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1101341/full#supplementary-material
## References
1. Dongiovanni P, Stender S, Pietrelli A, Mancina R, Cespiati A, Petta S. **Causal relationship of hepatic fat with liver damage and insulin resistance in nonalcoholic fatty liver.**. (2018) **283** 356-70. DOI: 10.1111/joim.12719
2. Rotman Y, Koh C, Zmuda JM, Kleiner DE, Liang TJ. **The association of genetic variability in patatin-like phospholipase domain-containing protein 3 (PNPLA3) with histological severity of nonalcoholic fatty liver disease.**. (2010) **52** 894-903. DOI: 10.1002/hep.23759
3. Valenti L, Al-Serri A, Daly A, Galmozzi E, Rametta R, Dongiovanni P. **Homozygosity for the patatin-like phospholipase-3/adiponutrin I148M polymorphism influences liver fibrosis in patients with nonalcoholic fatty liver disease.**. (2010) **51** 1209-17. DOI: 10.1002/hep.23622
4. Santoro N, Kursawe R, D’Adamo E, Dykas D, Zhang C, Bale A. **A common variant in the patatin-like phospholipase 3 gene (PNPLA3) is associated with fatty liver disease in obese children and adolescents.**. (2010) **52** 1281-90. PMID: 20803499
5. Salameh H, Hanayneh M, Masadeh M, Naseemuddin M, Matin T, Erwin A. **PNPLA3 as a genetic determinant of risk for and severity of non-alcoholic fatty liver disease spectrum.**. (2016) **4** 175-91. PMID: 27777887
6. Stender S, Loomba R. **PNPLA3 genotype and risk of liver and all-cause mortality.**. (2020) **71** 777-9. PMID: 31954067
7. Kumari M, Schoiswohl G, Chitraju C, Paar M, Cornaciu I, Rangrez A. **Adiponutrin functions as a nutritionally regulated lysophosphatidic acid acyltransferase.**. (2012) **15** 691-702. DOI: 10.1016/j.cmet.2012.04.008
8. BasuRay S, Wang Y, Smagris E, Cohen JC, Hobbs HH. **Accumulation of PNPLA3 on lipid droplets is the basis of associated hepatic steatosis.**. (2019) **116** 9521-6. PMID: 31019090
9. Vilar-Gomez E, Pirola C, Sookoian S, Wilson L, Liang T, Chalasani N. **PNPLA3 rs738409 and risk of fibrosis in NAFLD: exploring mediation pathways through intermediate histological features.**. (2022) **76** 1482-94. DOI: 10.1002/hep.32491
10. Wilson PA, Gardner SD, Lambie NM, Commans SA, Crowther DJ. **Characterization of the human patatin-like phospholipase family.**. (2006) **47** 1940-9. PMID: 16799181
11. Li JZ, Huang Y, Karaman R, Ivanova P, Brown H, Roddy T. **Chronic overexpression of PNPLA3I148M in mouse liver causes hepatic steatosis.**. (2012) **122** 4130-44. DOI: 10.1172/JCI65179
12. BasuRay S, Smagris E, Cohen JC, Hobbs HH. **The PNPLA3 variant associated with fatty liver disease (I148M) accumulates on lipid droplets by evading ubiquitylation.**. (2017) **66** 1111-24. DOI: 10.1002/hep.29273
13. Valenti L, Dongiovanni P. **Mutant PNPLA3 I148M protein as pharmacological target for liver disease.**. (2017) **66** 1026-8. PMID: 28586091
14. Tsugawa Y, Jena A, Orav E, Blumenthal D, Tsai T, Mehtsun W. **Age and sex of surgeons and mortality of older surgical patients: observational study. BMJ. 2018;361:k1343.**. (2018) **108** 695-700. DOI: 10.1136/bmj.k1343
15. Meroni M, Longo M, Rustichelli A, Dongiovanni P. **Nutrition and genetics in NAFLD: the perfect binomium.**. (2020) **21**. DOI: 10.3390/ijms21082986
16. Davis JN, Lê K, Walker R, Vikman S, Spruijt-Metz D, Weigensberg M. **Increased hepatic fat in overweight hispanic youth influenced by interaction between genetic variation in PNPLA3 and high dietary carbohydrate and sugar consumption.**. (2010) **92** 1522-7. DOI: 10.3945/ajcn.2010.30185
17. Santoro N, Savoye M, Kim G, Marotto K, Shaw M, Pierpont B. **Hepatic fat accumulation is modulated by the interaction between the rs738409 variant in the PNPLA3 gene and the dietary omega6/omega3 PUFA intake.**. (2012) **7**. DOI: 10.1371/journal.pone.0037827
18. Carlson LA. **Nicotinic acid: the broad-spectrum lipid drug. a 50th anniversary review.**. (2005) **258** 94-114. DOI: 10.1111/j.1365-2796.2005.01528.x
19. Ganji SH, Kukes GD, Lambrecht N, Kashyap ML, Kamanna VS. **Therapeutic role of niacin in the prevention and regression of hepatic steatosis in rat model of nonalcoholic fatty liver disease.**. (2014) **306** G320-7. DOI: 10.1152/ajpgi.00181.2013
20. Ganji SH, Kashyap ML, Kamanna VS. **Niacin inhibits fat accumulation, oxidative stress, and inflammatory cytokine IL-8 in cultured hepatocytes: impact on non-alcoholic fatty liver disease.**. (2015) **64** 982-90. DOI: 10.1016/j.metabol.2015.05.002
21. Li C, Li L, Lian J, Watts R, Nelson R, Goodwin B. **Roles of Acyl-CoA:diacylglycerol acyltransferases 1 and 2 in triacylglycerol synthesis and secretion in primary hepatocytes.**. (2015) **35** 1080-91
22. Gariani K, Menzies K, Ryu D, Wegner C, Wang X, Ropelle E. **Eliciting the mitochondrial unfolded protein response by nicotinamide adenine dinucleotide repletion reverses fatty liver disease in mice.**. (2016) **63** 1190-204. DOI: 10.1002/hep.28245
23. Revollo JR, Körner A, Mills K, Satoh A, Wang T, Garten A. **Nampt/PBEF/Visfatin regulates insulin secretion in beta cells as a systemic NAD biosynthetic enzyme.**. (2007) **6** 363-75. DOI: 10.1016/j.cmet.2007.09.003
24. Hu M, Chu W, Yamashita S, Yeung D, Shi L, Wang D. **Liver fat reduction with niacin is influenced by DGAT-2 polymorphisms in hypertriglyceridemic patients.**. (2012) **53** 802-9. DOI: 10.1194/jlr.P023614
25. Linder K, Willmann C, Kantartzis K, Machann J, Schick F, Graf M. **Dietary niacin intake predicts the decrease of liver fat content during a lifestyle intervention.**. (2019) **9**. DOI: 10.1038/s41598-018-38002-7
26. Castera L, Forns X, Alberti A. **Non-invasive evaluation of liver fibrosis using transient elastography.**. (2008) **48** 835-47. PMID: 18334275
27. Boursier J, Zarski J, de Ledinghen V, Rousselet M, Sturm N, Lebail B. **Determination of reliability criteria for liver stiffness evaluation by transient elastography.**. (2013) **57** 1182-91. PMID: 22899556
28. Meroni M, Dongiovanni P, Longo M, Carli F, Baselli G, Rametta R. **Mboat7 down-regulation by hyper-insulinemia induces fat accumulation in hepatocytes.**. (2020) **52**. DOI: 10.1016/j.ebiom.2020.102658
29. Longo M, Meroni M, Paolini E, Erconi V, Carli F, Fortunato F. **TM6SF2/PNPLA3/MBOAT7 loss-of-function genetic variants impact on NAFLD development and progression both in patients and in In vitro models.**. (2022) **13** 759-88. DOI: 10.1016/j.jcmgh.2021.11.007
30. Meroni M, Longo M, Paolini E, Lombardi R, Piciotti R, Francione P. **MAFLD definition underestimates the risk to develop HCC in genetically predisposed patients.**. (2022) **291** 374-6. DOI: 10.1111/joim.13396
31. Baselli GA, Dongiovanni P, Rametta R, Meroni M, Pelusi S, Maggioni M. **Liver transcriptomics highlights interleukin-32 as novel NAFLD-related cytokine and candidate biomarker.**. (2020) **69** 1855-66. DOI: 10.1136/gutjnl-2019-319226
32. Ye L, Cao Z, Lai X, Shi Y, Zhou N. **Niacin ameliorates hepatic steatosis by inhibiting de novo lipogenesis via a GPR109A-Mediated PKC–ERK1/2–AMPK signaling pathway in C57BL/6 mice fed a high-fat diet.**. (2019) **150** 672-84. DOI: 10.1093/jn/nxz303
33. Xu X, Deng X, Chen Y, Xu W, Xu F, Liang H. **SIRT1 mediates nutritional regulation of SREBP-1c-driven hepatic PNPLA3 transcription via modulation of H3k9 acetylation.**. (2022) **44**. DOI: 10.1186/s41021-022-00246-1
34. Huang Y, He S, Li J, Seo Y, Osborne T, Cohen J. **A feed-forward loop amplifies nutritional regulation of PNPLA3.**. (2010) **107** 7892-7. DOI: 10.1073/pnas.1003585107
35. Mondul A, Mancina R, Merlo A, Dongiovanni P, Rametta R, Montalcini T. **PNPLA3 I148M variant influences circulating retinol in adults with nonalcoholic fatty liver disease or obesity.**. (2015) **145** 1687-91. PMID: 26136587
36. Ganji SH, Tavintharan S, Zhu D, Xing Y, Kamanna V, Kashyap M. **Niacin noncompetitively inhibits DGAT2 but not DGAT1 activity in HepG2 cells.**. (2004) **45** 1835-45. DOI: 10.1194/jlr.M300403-JLR200
37. Blond E, Rieusset J, Alligier M, Lambert-Porcheron S, Bendridi N, Gabert L. **Nicotinic acid effects on insulin sensitivity and hepatic lipid metabolism: an in vivo to in vitro study.**. (2014) **46** 390-6. DOI: 10.1055/s-0034-1372600
38. Kashyap ML, Ganji S, Nakra NK, Kamanna VS. **Niacin for treatment of nonalcoholic fatty liver disease (NAFLD): novel use for an old drug?**. (2019) **13** 873-9. DOI: 10.1016/j.jacl.2019.10.006
39. Guarino M, Dufour J-F. **Nicotinamide and NAFLD: is there nothing new under the sun?**. (2019) **9**. DOI: 10.3390/metabo9090180
40. Li Y, Xu S, Mihaylova M, Zheng B, Hou X, Jiang B. **AMPK phosphorylates and inhibits SREBP activity to attenuate hepatic steatosis and atherosclerosis in diet-induced insulin-resistant mice.**. (2011) **13** 376-88. DOI: 10.1016/j.cmet.2011.03.009
|
---
title: Quantifying factors that affect polygenic risk score performance across diverse
ancestries and age groups for body mass index
authors:
- Daniel Hui
- Brenda Xiao
- Ozan Dikilitas
- Robert R. Freimuth
- Marguerite R. Irvin
- Gail P. Jarvik
- Leah Kottyan
- Iftikhar Kullo
- Nita A. Limdi
- Cong Liu
- Yuan Luo
- Bahram Namjou
- Megan J. Puckelwartz
- Daniel Schaid
- Hemant Tiwari
- Wei-Qi Wei
- Shefali Verma
- Dokyoon Kim
- Marylyn D. Ritchie
journal: Pacific Symposium on Biocomputing. Pacific Symposium on Biocomputing
year: 2023
pmcid: PMC10018532
license: CC BY 4.0
---
# Quantifying factors that affect polygenic risk score performance across diverse ancestries and age groups for body mass index
## Abstract
Polygenic risk scores (PRS) have led to enthusiasm for precision medicine. However, it is well documented that PRS do not generalize across groups differing in ancestry or sample characteristics e.g., age. Quantifying performance of PRS across different groups of study participants, using genome-wide association study (GWAS) summary statistics from multiple ancestry groups and sample sizes, and using different linkage disequilibrium (LD) reference panels may clarify which factors are limiting PRS transferability. To evaluate these factors in the PRS generation process, we generated body mass index (BMI) PRS (PRSBMI) in the Electronic Medical Records and Genomics (eMERGE) network ($$n = 75$$,661). Analyses were conducted in two ancestry groups (European and African) and three age ranges (adult, teenagers, and children). For PRSBMI calculations, we evaluated five LD reference panels and three sets of GWAS summary statistics of varying sample size and ancestry. PRSBMI performance increased for both African and European ancestry individuals using cross-ancestry GWAS summary statistics compared to European-only summary statistics ($6.3\%$ and $3.7\%$ relative R2 increase, respectively, pAfrican=0.038, pEuropean=6.26×10−4). The effects of LD reference panels were more pronounced in African ancestry study datasets. PRSBMI performance degraded in children; R2 was less than half of teenagers or adults. The effect of GWAS summary statistics sample size was small when modeled with the other factors. Additionally, the potential of using a PRS generated for one trait to predict risk for comorbid diseases is not well understood especially in the context of cross-ancestry analyses – we explored clinical comorbidities from the electronic health record associated with PRSBMI and identified significant associations with type 2 diabetes and coronary atherosclerosis. In summary, this study quantifies the effects that ancestry, GWAS summary statistic sample size, and LD reference panel have on PRS performance, especially in cross-ancestry and age-specific analyses.
## Introduction
Polygenic risk scores (PRS) provide individualized genetic estimates of a phenotype by aggregating genetic effects across hundreds or thousands of loci, typically from genome-wide association studies (GWAS). PRS are potentially a powerful source of increased prediction performance, even when combined with family history [1,2]. However, in recent years it has become increasingly apparent that performance of PRS is substantially reduced when the ancestry of the individuals in whom prediction is being done differs from the ancestry of the individuals from the GWAS used to generate SNP weights used for PRS construction. For instance, when using GWAS from European ancestry individuals, the prediction accuracy of polygenic scores in individuals of African or Hispanic/Latino ancestry have a relative performance of $25\%$ and $65\%$ compared to performance in European ancestry individuals [3]. Additionally, evidence exists suggesting that for some traits, such as adiposity traits, this disparity may be further exacerbated by environmental, demographic, or social risk factors (including age, physical activity, smoking status, and alcohol use (4–7)). For example, differences in the genetic architecture of body mass index (BMI) have been shown to differ between age groups (8–11). Thus, the performance of PRS for BMI is also affected by the age of the individuals used in the GWAS and the study data where the PRS is evaluated [12]. Broad-sense heritability estimates for BMI in adults ranges from $40\%$−$90\%$ when estimated in adults of different cohorts even of homogeneous ancestry [13]; even if heritability estimates are similar across populations, genetic architecture and enrichment for variants in different functional categories may still differ [14,15].
Several outstanding questions surrounding PRS, especially within the context of adiposity traits and BMI, warrant further investigation. For instance, when cross-ancestry summary statistics (i.e., those including individuals of multiple ancestry groups in the GWAS) are available, can they be used to improve prediction performance in individuals from one or more different ancestry groups? We need a more thorough evaluation of the potential prediction performance gain (or loss) in African ancestry individuals when cross-ancestry GWAS summary statistics are used to estimate the SNP weights. In addition, we need to improve our understanding of the impact of the composition of the linkage disequilibrium (LD) reference panel in combination with cross-ancestry GWAS summary statistics on PRS prediction performance. For prediction of BMI specifically, how does prediction performance differ for individuals in different age groups, especially those who are not adults (i.e., less than age 18)? Additionally, how much these different variables impact the PRS performance when considered together is important to explore. Developing a deeper understanding of which features (ancestry of individuals in the GWAS, ancestry of the individuals generating the LD references panel, ancestry of the study data, age of the study data) have the greatest impact of PRS performance will help the field develop future studies and strategies around clinical risk prediction with PRS. The degree to which increased GWAS sample size increases prediction performance regardless of these other factors is also important to determine. Finally, there is potential for using a PRS generated for one trait to predict risk for comorbid traits. Understanding how much the different elements of PRS generation affects associations with clinical comorbidities of obesity is of great importance for precision medicine.
We comprehensively investigated the influence of these factors on the performance of PRS using the Electronic Medical Records and Genomics (eMERGE) Network dataset. eMERGE is an NIH funded consortium that combines participants from multiple electronic health record (EHR) linked biobanks [16]. In the present study, we included 75,661 individuals of diverse ancestry and age ($14\%$ African ancestry, $55\%$ female, and $12\%$ children age < 13). These individuals were from the eMERGE III imputed array dataset ($$n = 83$$,717) (dbGaP Study: phs001584.v2.p2), estimated European or African ancestry, and had BMI measurements available. For these analyses, we used published BMI GWAS summary statistics from the GIANT (Genetic Investigation of ANthropometric Traits) consortium, an international consortium that primarily studies anthropometric traits, which included participants (max $$n = 339$$,224, mean N per variant=226,960) from European, African, and Asian ancestry groups [17]. We also used summary statistics from a European ancestry BMI GWAS [18] in UK Biobank (UKBB) individuals ($$n = 339$$,721), which was conducted using both the full sample size of the European ancestry UKBB, as well as after down-sampling to the same number of individuals in the GIANT GWAS. This comparison allowed us to better evaluate whether it was the ancestry composition or the sample size of the dataset where the GWAS summary statistics were derived that affected the results of the PRS performance. We calculated PRS for BMI (PRSBMI) across 90 different combinations of analyses (described more in Methods) – six different groupings based on ancestry and age, five different LD reference panels (of varying ancestry and from three different cohorts), and the three mentioned sets of GWAS summary statistics. We then statistically compared the different sets of analyses to see what factors most influence PRSBMI performance across various groupings of individuals based on ancestry and age. Lastly, we also tested the association of the best performing PRSBMI with common comorbidities across ancestry groups to identify the clinical relevance of the PRSBMI in phenotypes derived from an Electronic Health Record (EHR). Investigation of these variables elucidates our understanding of the factors that affect PRS performance and transferability across ancestries and populations, especially within the context of BMI, as well as the potential of using PRSBMI to predict risk for comorbid disease.
## Overall study design
The electronic Medical Records and Genomics (eMERGE) network dataset is an NIH funded consortium that combines participants from multiple electronic health record (EHR) linked biobanks. In this study, we included 75,661 individuals with available genetic and phenotypic data. The individuals in the eMERGE dataset include multiple ancestry groups – genetically inferred ancestry was assigned by the eMERGE consortium [16] – and a large age distribution ($14\%$ African ancestry, $19\%$ less than age 18, Figure 1). Briefly, we calculated PRSBMI for all individuals within each combination where the following elements of the model varied: 1) LD panels that differed in ancestry, 2) GWAS summary statistics with variable ancestry composition, and 3) GWAS summary statistics for two different sample sizes. The details for each of these are provided more below. The data was also split by ancestry and age group, and we statistically compared PRSBMI performance between all the different groups – in total, 90 sets of PRSBMI were calculated separately and then compared. We first estimated the effect and significance of each variable (i.e., ancestry of GWAS summary statistics and test data, LD panel ancestry, size of GWAS summary statistics, and age of test individuals) on PRS performance. Next, we estimated how much each variable affects PRSBMI performance when all are modeled together, and finally we analyzed the potential clinical associations by testing the PRSBMI for association with common comorbid conditions from the EHR. For the primary results related to LD panel or ancestry of summary statistics and test data, we restricted analyses to adults as the other age groups were limited in sample size. In the following sections, we describe all these elements in more detail.
## Summary statistics to generate PRSBMI
We obtained published GWAS summary statistics from the GIANT consortium [17] to use as one set of BMI GWAS summary statistics. Up to 322,154 adults of European ancestry, as well as an additional 17,072 adults of non-European ancestry (adults of African, East Asian, and South Asian ancestry), were included in the GIANT GWAS analysis.
For the second set of summary statistics, we performed a GWAS in the individuals of European ancestry from the UK Biobank (UKBB). Individuals were first filtered by low quality samples (sex mismatch between genetically inferred and self-reported, variant missingness > $5\%$), relatedness (no 2nd degree relatives or higher), and within the White British ancestry subset (with these individuals being defined by UKBB and selected based on self-reports and genetically determined ancestry) [18]; a total of 377,921 individuals initially remained. Variants were filtered on imputation quality score (using the INFO metric [19]) > 0.30, and minor allele frequency > $1\%$ within this subset of individuals. In addition, we generated a second set of GWAS summary statistics from the UKBB, where we randomly down-sampled individuals to the sample size in the GIANT GWAS dataset ($$n = 226$$,960). In each UKBB GWAS, data processing and modeling were performed similarly as in the GIANT GWAS – summary statistics were calculated using linear regression, with age, age2, sex, and the first 5 genetic principal components (PCs) included as covariates. BMI, defined as weight in kilograms divided by squared height in meters, was first inverse-rank normal transformed.
After calculation of BMI GWAS summary statistics in each of the two datasets of UKBB individuals of European ancestry, we harmonized variants across all datasets used (UKBB, eMERGE, GIANT, and 1000 Genomes Phase 3). For the remainder of downstream analyses, we kept only those variants that were present in all datasets, and additionally excluded any strand-ambiguous SNPs (alleles A/T or C/G), and retained only biallelic variants; in total, 2,014,457 variants were retained for analyses.
## LD reference panels
Five different LD reference panels were used for each set of PRSBMI calculations: 1) all of 1000 Genomes (1KGAll) ($$n = 2$$,504), 2) 1000 Genomes European ancestry (1KGEUR) ($$n = 503$$), 3) 5,000 randomly selected European ancestry individuals from the UK Biobank (UKBBEUR), 4) 5,000 randomly selected individuals from all of UK Biobank (UKBBAll), and 5) up to 5,000 randomly selected individuals from the dataset for which PRSBMI were being calculated for in the eMERGE dataset (referred to as test data henceforward). These panels were chosen to test for differences in ancestry distribution and sample size on PRS performance.
## PRS software
For each comparison set, PRSBMI were calculated using pruning and thresholding method via PRSice v2.1.9 [20]. We chose to use PRSice due to the flexibility it provides in choosing external LD panels and allowed us to easily include multi-ancestry LD panels in our analyses. Default parameters were used in all analyses (clumping performed in 250 kb windows using an R2 of 0.1, p-value step size of 0.00005 between p-values of.0001 up to.10 and step size of.0001 between p-values of.10 up to.50).
## Statistical comparisons
Incremental R2 for PRSBMI was calculated by subtracting the R2 using a model with only the covariates from the R2 of the model using the covariates and the PRSBMI (the default option in PRSice). Statistical differences between model performances from different iterations were determined using the Wilcoxon rank-sum test to compare the distributions of the squared residuals generated from the model for all individuals in the iteration; for comparisons between the same set of individuals, the paired Wilcoxon rank-sum test was used. When testing which of the five LD panels performed the best, we used a Bonferroni-corrected threshold of $\frac{0.05}{10}$ = 0.005 (ten comparisons between five LD panels). When comparing the best performing PRSBMI across ancestries and summary statistics using their best LD panel, we used a Bonferroni threshold of $\frac{0.05}{25}$ = 0.002 (25 comparisons between the five LD panels used).
## Proportion of variance explained by each individual variable
We modeled all evaluated features together in the following linear regression model: R2∼LDpanel+NSumstat+AgeTest+AncestrySumstat+AncestryTest+AncestrySumstat∗AncestryTest Where the Sumstat subscript is defined as a set of GWAS summary statistics, and the Test subscript is defined as a set of test individuals that PRS prediction is being assessed in. We quantified the variance in R2 that could be explained by each of these different variables using type II sum of squares from ANOVA. The sum of squares of variables involving ancestry were summed together; an interaction term between summary statistics ancestry and test data ancestry was included to identify whether the ancestry of summary statistics and test data matched.
## Association of PRSBMI with comorbidities
We selected the ten most frequent Phecodes [21] from the EHR data in the eMERGE dataset (which includes obesity as a positive control) to test their association with the PRSBMI. For each Phecode, individuals were classified as a case for the condition if there was at least one occurrence of the respective Phecode in their EHR record; individuals were classified as a control for that condition if there was no occurrence of the Phecode. This classification is a rule-of-one instance of a Phecode to define case status. For each eMERGE ancestry subgroup, we selected the best performing PRSBMI i.e., the PRSBMI with the highest R2, and tested the association of the PRSBMI with these ten clinical conditions using a logistic regression model. PRSBMI was first mean-centered and standard deviation was set to 1. Sex, age, age2, and the first five genetic PCs were included as covariates.
## Data visualization
The ‘ggplot2’ R package was used for plotting, with the ‘geom_signif’ package used to include significance bars. The association results were plotted using PheWAS-View [22].
## Effect of LD panel
For adults of African ancestry, when using the down-sampled UKBB GWAS summary statistics, using either cross-ancestry or African ancestry test data LD panels significantly improved PRSBMI performance compared to European ancestry LD panels (Figure 2). When using the UKBB summary statistics, the top PRSBMI R2 was 0.0140 using the test data as LD panel, while the second-best performing LD panel (UKBB European) had an R2 of 0.0109 ($$p \leq 4.94$$×10−20). When using the GIANT summary statistics, the top PRSBMI R2 was 0.0149 using 1KGAll as the reference panel. The PRSBMI calculated using the best European ancestry panel (1KGEUR) resulted in a R2 of 0.0141, but this difference between these two reference panels was not Bonferroni significant ($$p \leq 0.037$$). However, the 1KGall LD panel performed significantly better than the two UKBB LD panels (UKBBAll: R2 = 0.0134, $$p \leq 3.65$$×10−5; UKBB European: R2 = 0.0128, $$p \leq 3.65$$×10−9). The test data LD panel performed the second-best with an R2 of 0.0142, and significantly outperformed the UKBB European LD panel ($$p \leq 4.78$$×10−5). For adults of European ancestry, we observed more significant differences in performance when using the GIANT summary statistics compared to the down-sampled UKBB summary statistics. The 1KGAll LD panel performed the best with a R2 of 0.0612. It also significantly outperformed all other LD panels (1KGEUR: R2 = 0.0560, $$p \leq 5.54$$×10−104; Test data: R2 = 0.0564, $$p \leq 6.50$$×10−67; UKBBAll: R2 = 0.0561, 8.09×10−107; UKBBEUR: R2 = 0.0561, $$p \leq 3.02$$×10−77). We note that this increase was larger when using the GIANT summary statistics but was still present when using the UKBB summary statistics. When using the UKBB summary statistics, the choice of LD panel had a much smaller impact on prediction performance. While the 1KGAll LD panel performed the best, the difference in performance was much less significant between the next best performing LD panel (R21KGAll = 0.0590, R2UKBBAll = 0.0583, $$p \leq 3.48$$×10−4). The difference between the best and worst performing scores – LD panel using 1KG all versus 1KG European – was also much less significant ($$p \leq 1.15$$×10−12). These results suggest that the choice of LD panel particularly matters when calculating PRSBMI using cross-ancestry GWAS, or for African ancestry individuals when the GWAS summary statistics are derived from European ancestry individuals.
However, we did observe a slight decrease in the impact of the choice of LD panel when using the full UKBB summary statistics for adults; again, the largest differences were observed in adults of African ancestry, but differences in performance across LD panels were not as significant. The test LD panel performed second best with the 1KGEUR LD panel performing best (R2Test = 0.0197, R21KGEUR = 0.0200, $$p \leq 0.18$$). The 1KGAll LD panel was the worst performing LD panel with an R2 of 0.0185, and this difference between the 1KGEUR LD panel was significant after multiple hypothesis correction ($$p \leq 5.08$$×10−7).
## Effect of summary statistics and ancestry of test data
As expected, the R2 values of the PRSBMI were significantly higher when calculated for European ancestry adults than adults of African ancestry, even when using the cross-ancestry GIANT summary statistics (Figure 2). When using the GIANT summary statistics, the best performing PRSBMI in adults of European ancestry had an R2 of 0.0612, which was significantly higher than the R2 from the best performing PRSBMI in African ancestry adults (R2 = 0.0149, $p \leq 4.9$×10−324).
In African ancestry adults, the R2 when using the GIANT summary statistics was higher than the R2 when using the down-sampled UKBB summary statistics with their respective best LD panel (GIANT (1KGAll LD panel): R2 = 0.0149, UKBB (test data LD panel): R2 = 0.0140; $$p \leq 0.038$$). This difference was not statistically significant after multiple hypothesis correction. However, the GIANT summary statistics with the 1KGAll LD panel did significantly outperform the UKBB summary statistics with all other LD panels. When keeping the LD panel constant, the PRSBMI calculated using the GIANT summary statistics resulted in higher R2 than using the UKBB summary statistics for all LD panels except for the test data LD panel, and this difference was statistically significant for the 1KGAll ($$p \leq 1.55$$×10−33), 1KGEUR ($$p \leq 6.78$$×10−18), and UKBBAll ($$p \leq 1.28$$×10−15) LD panels. Somewhat surprisingly, we observed higher R2 values for European ancestry adults when using the cross-ancestry GIANT summary statistics versus the down-sampled European UKBB summary statistics (R2GIANT = 0.0612 versus R2UKBB = 0.0590), with this difference being statistically significant ($$p \leq 6.26$$×10−4); the best performing LD panel for both set of summary statistics was 1KGAll.
We also compared prediction performance in all individuals using the full ($$n = 377$$,921) European UKBB GWAS versus the European UKBB GWAS down-sampled to GIANT’s sample size ($$n = 226$$,960) (Figure 2, Supplemental Table 1). For consistency, UKBB European individuals were used for the European test ancestry comparisons, and for the African ancestry comparisons the test sets (i.e., African ancestry LD panels) were used as LD panels. Uniformly across test ancestry and age groups, we observed higher and statistically significant increases in R2.
## Prediction performance across different age groups
Across different ancestries and summary statistics, we broadly observed similar R2 values for adults and teenagers, with substantially reduced performance in children (Supplemental Figure 1). R2 values in children were consistently less than half of that in adults and teenagers, with differences in R2 values for adults and teenagers being minimal (except in the case of African ancestry individuals using the GIANT summary statistics, with teenagers having more than double the R2 of adults). Somewhat surprisingly, teenagers consistently had higher R2 than adults across all analyses, although these differences were much less significant than those compared with children.
## Proportion of variance explained by each assessed factor
While we observed significant differences due to ancestry, age, and number of individuals used to calculate summary statistics, we aimed to quantify the effect of these different variables on PRSBMI performance when considered together (Table 1). We observed that $89.5\%$ of the variance in PRSBMI R2 could be explained using these variables, indicating that the majority of the effects of LD panel, ancestry, age, and sample size could be explained through linear relationships with PRSBMI R2. In the context of these comparisons, the ancestry of the summary statistics or test data accounts for $55.1\%$ of the variance explained in PRSBMI R2. Choice of LD panel and age of test individuals accounted for similar amounts of variance explained in PRSBMI R2 ($16.5\%$ and $15.9\%$, respectively), while the number of individuals used to calculate the GWAS summary statistics only accounted for $1.9\%$ of variance explained in PRSBMI R2. Per previous sections, while number of individuals used for summary statistics resulted in significant differences in PRSBMI performance, its overall impact when modeled jointly with all the other factors in the context of these analyses seemed to be small.
## PRSBMI association with comorbid traits
To determine whether the PRSBMI was associated with clinical comorbidities, we performed a Phenome-Wide Association Study for ten clinical conditions (Supplemental Table 2, described more in Methods). Here, the PRSBMI was tested for association with diagnosis codes (Phecodes) to evaluate whether the polygenic background for BMI associates with these clinical diagnoses. The PRSBMI was significantly associated with several of the most frequent Phecodes in eMERGE, particularly in European adults (Figure 3a). As expected, obesity had the strongest association with PRSBMI in all ancestry groups (pEUR < 4.9×10−324; pAFR = 5.17×10−8); this was a positive control. In European ancestry individuals, the best performing PRSBMI was also significantly positively associated with type 2 diabetes (pEUR = 1.04×10−102), essential hypertension (pEUR = 7.12×10−56), coronary atherosclerosis (pEUR = 3.61×10−26), hyperlipidemia (pEUR = 4.38×10−16), depression (pEUR = 1.95×10−13), hypercholesteremia (pEUR = 3.64×10−15), asthma (pEUR = 3.13×10−13), and diverticulosis (pEUR = 0.0017). These associations were less statistically significant in African ancestry individuals, which had much lower sample size, and many associations were no longer significant after Bonferroni correction. Only type 2 diabetes (pAFR = 1.2×10−5) and coronary atherosclerosis (pAFR = 0.001) were significantly associated with the PRSBMI in African ancestry adults. We also looked at the prevalence of each condition per PRS quintile for the most significantly associated conditions (Figure 3b). The case prevalence generally increased in higher PRSBMI quintile groups for conditions significantly associated with the PRSBMI, a trend matching the results we obtained from the association analysis. Phenotypes with downward trends were not significantly associated with PRSBMI, and low sample sizes in earlier quintile groups may have contributed to this seemingly decreasing prevalence. We performed similar analyses in teens and children but identified no statistically significant associations (results not shown). The much smaller sample sizes of the Phecodes in these age groups may have also contributed to the lack of statistically significant results – most of these diagnoses are adult-onset conditions.
## Discussion
Somewhat unintuitively, African ancestry LD panels performed best for African ancestry individuals, regardless of whether European ancestry or cross-ancestry GWAS summary statistics were used. We observed minimal impact of the choice of LD panel when both test data and summary statistics were of European ancestry. These results suggest that as long as either the test data or GWAS summary statistics are of similar ancestry, or the test data and LD panel are of similar ancestry, the difference in PRS performance may be minimal as compared to if all the GWAS summary statistics, test data, and LD panel are all of the same ancestry. We also observed significantly decreased PRS performance in children compared to adults and teens, with the GWAS used in this study being conducted on adult populations.
While the findings in this study highlight many important strategies for performing PRS in different ancestry and age groups, there are limitations that should be addressed in future studies.
First, inclusion of analyses that evaluate how different proportions of non-European ancestry individuals affect the prediction performance of PRS would be useful. The GIANT summary statistics we used in this study are only about $6\%$ non-European ancestry. It may be useful to see how the PRS prediction performance changes in both non-European and European ancestry datasets as a function of the proportion of non-European ancestry samples included in the GWAS. Such analyses may be possible by combining African ancestry individuals from these different datasets. These analyses will be possible once larger datasets that include non-European ancestry cohorts are publicly available or could be tested by analyzing other traits with larger African ancestry GWAS. Future analyses could also include sex-stratified GWAS and comparison sets to evaluate the influence of sex on PRSBMI performance. Finally, repeating these types of analyses with different PRS methods would be useful as novel PRS methods are being developed on a regular basis, many of which incorporate ancestry in different ways.
Overall, this study demonstrates the importance of expanding non-European ancestry data resources for PRS, specifically in the generation of GWAS summary statistics and LD reference panels. Failure to do so reduces the impact of PRS in diverse populations and increases the potential for continued health disparities, especially in precision medicine where genetics is being integrated into clinical care.
## Data and code availability
Code supporting the current study are available from the corresponding author on request.
## References
1. Truong B, Zhou X, Shin J, Li J, van der Werf JHJ, Le TD. **Efficient polygenic risk scores for biobank scale data by exploiting phenotypes from inferred relatives**. *Nat Commun* (2020.0) **11** 3074. PMID: 32555176
2. Margaux LA, Hujoel Po-Ru Loh, Benjamin M. Neale, Alkes L. *. Incorporating family history of disease improves polygenic risk scores in diverse populations*
3. Martin AR, Kanai M, Kamatani Y, Okada Y, Neale BM, Daly MJ. **Clinical use of current polygenic risk scores may exacerbate health disparities**. *Nat Genet* (2019.0) **51** 584-91. PMID: 30926966
4. Rask-Andersen M, Karlsson T, Ek WE, Johansson Å. **Gene-environment interaction study for BMI reveals interactions between genetic factors and physical activity, alcohol consumption and socioeconomic status**. *PLoS Genet* (2017.0) **13**
5. Robinson MR, English G, Moser G, Lloyd-Jones LR, Triplett MA, Zhu Z. **Genotype-covariate interaction effects and the heritability of adult body mass index**. *Nat Genet* (2017.0) **49** 1174-81. PMID: 28692066
6. Sulc J, Mounier N, Günther F, Winkler T, Wood AR, Frayling TM. **Quantification of the overall contribution of gene-environment interaction for obesity-related traits**. *Nat Commun* (2020.0) **11** 1385. PMID: 32170055
7. Justice AE, Winkler TW, Feitosa MF, Graff M, Fisher VA, Young K. **Genome-wide meta-analysis of 241,258 adults accounting for smoking behaviour identifies novel loci for obesity traits**. *Nat Commun* (2017.0) **8** 14977. PMID: 28443625
8. Helgeland Ø, Vaudel M, Juliusson PB, Lingaas Holmen O, Juodakis J, Bacelis J. **Genome-wide association study reveals dynamic role of genetic variation in infant and early childhood growth**. *Nat Commun* (2019.0) **10** 4448. PMID: 31575865
9. Vogelezang S, Bradfield JP, Ahluwalia TS, Curtin JA, Lakka TA, Grarup N. **Novel loci for childhood body mass index and shared heritability with adult cardiometabolic traits. PLoS Genet**. *Oct* (2020.0) **16**
10. Couto Alves A, De Silva NMG, Karhunen V, Sovio U, Das S, Taal HR. **GWAS on longitudinal growth traits reveals different genetic factors influencing infant, child, and adult BMI**. *Sci Adv* (2019.0) **5**
11. Choh AC, Lee M, Kent JW, Diego VP, Johnson W, Curran JE. **Gene-by-age effects on BMI from birth to adulthood: the Fels Longitudinal Study. Obes Silver Spring Md**. (2014.0) **22** 875-81
12. Mostafavi H, Harpak A, Agarwal I, Conley D, Pritchard JK, Przeworski M. **Variable prediction accuracy of polygenic scores within an ancestry group**. *eLife* (2020.0) **9**
13. Elks CE, den Hoed M, Zhao JH, Sharp SJ, Wareham NJ, Loos RJF. **Variability in the heritability of body mass index: a systematic review and meta-regression**. *Front Endocrinol* (2012.0) **3** 29
14. Galinsky KJ, Reshef YA, Finucane HK, Loh PR, Zaitlen N, Patterson NJ. **Estimating cross-population genetic correlations of causal effect sizes**. *Genet Epidemiol* (2019.0) **43** 180-8. PMID: 30474154
15. Shi H, Gazal S, Kanai M, Koch EM, Schoech AP, Siewert KM. **Population-specific causal disease effect sizes in functionally important regions impacted by selection**. *Nat Commun* (2021.0) **12** 1098. PMID: 33597505
16. Stanaway IB, Hall TO, Rosenthal EA, Palmer M, Naranbhai V, Knevel R. **The eMERGE genotype set of 83,717 subjects imputed to ~40 million variants genome wide and association with the herpes zoster medical record phenotype**. *Genet Epidemiol* (2019.0) **43** 63-81. PMID: 30298529
17. Locke AE, Kahali B, Berndt SI, Justice AE, Pers TH, Day FR. **Genetic studies of body mass index yield new insights for obesity biology**. *Nature* (2015.0) **518** 197-206. PMID: 25673413
18. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K. **The UK Biobank resource with deep phenotyping and genomic data**. *Nature* (2018.0) **562** 203-9. PMID: 30305743
19. Howie BN, Donnelly P, Marchini J. **A flexible and accurate genotype imputation method for the next generation of genome-wide association studies**. *PLoS Genet* (2009.0) **5**
20. Euesden J, Lewis CM, O’Reilly PF. **PRSice: Polygenic Risk Score software**. *Bioinforma Oxf Engl* (2015.0) **31** 1466-8
21. Wu P, Gifford A, Meng X, Li X, Campbell H, Varley T. **Mapping ICD-10 and ICD-10-CM Codes to Phecodes: Workflow Development and Initial Evaluation**. *JMIR Med Inform* (2019.0) **7** e14325. PMID: 31553307
22. Pendergrass SA, Dudek SM, Crawford DC, Ritchie MD. **Visually integrating and exploring high throughput Phenome-Wide Association Study (PheWAS) results using PheWAS-View**. *BioData Min* (2012.0) **5** 5. PMID: 22682510
|
---
title: 'Implementation and use of technology-enabled blood pressure monitoring and
teleconsultation in Singapore’s primary care: a qualitative evaluation using the
socio-technical systems approach'
authors:
- Sok Huang Teo
- Evelyn Ai Ling Chew
- David Wei Liang Ng
- Wern Ee Tang
- Gerald Choon Huat Koh
- Valerie Hui Ying Teo
journal: BMC Primary Care
year: 2023
pmcid: PMC10018584
doi: 10.1186/s12875-023-02014-8
license: CC BY 4.0
---
# Implementation and use of technology-enabled blood pressure monitoring and teleconsultation in Singapore’s primary care: a qualitative evaluation using the socio-technical systems approach
## Abstract
### Background
Telemedicine is becoming integral in primary care hypertension management, and is associated with improved blood pressure control, self-management and cost-effectiveness. This study explored the experiences of patients and healthcare professionals and their perceived barriers and facilitators in implementing and using a technology-enabled blood pressure monitoring intervention with teleconsultation in the Singapore primary care setting.
### Methods
This was a qualitative study embedded within the Primary Technology-Enhanced Care Hypertension pilot trial. Patients were selected purposively and invited to participate by telephone; healthcare professionals involved in the trial were invited to participate by email. Individual semi-structured interviews were conducted in English or Mandarin with thirteen patients and eight healthcare professionals. Each interview was audio-recorded and transcribed verbatim. Data were analyzed inductively to identify emergent themes which were then grouped into the dimensions of the socio-technical systems model to study the interactions between the technical, individual and organizational factors involved in the process.
### Results
Several emergent themes were identified. The factors involved in the implementation and use of the intervention are complex and interdependent. Patients and healthcare professionals liked the convenience resulting from the intervention and saw an improvement in the patient-provider relationship. Patients appreciated that the intervention helped form a habit of regular blood pressure monitoring, improved their self-management, and provided reassurance that they were being monitored by the care team. Healthcare professionals found that the intervention helped to manage workload by freeing up time for other urgent matters. Nevertheless, participants highlighted challenges with usability of the equipment and management portal, data access, and some expressed technology anxiety. Participants suggested patient segmentation for the intervention to be more targeted, wished for a more user-friendly equipment and proposed allocating more resources to the intervention.
### Conclusions
The implementation and use of telemedicine for hypertension management can engender various benefits and challenges to patients, healthcare professionals and the healthcare system. Stakeholder feedback gathered on the sociotechnical aspects of the technology should be taken into consideration to guide the design, implementation and evaluation of future telemedicine interventions in primary care.
### Trial registration
This study was registered on ClinicalTrials.gov on October 9, 2018. ID: NCT03698890.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12875-023-02014-8.
## Background
Hypertension is one of the most prevalent chronic diseases in Singapore, affecting $35.5\%$ of adults aged between 18 and 74 years old [1]. It is commonly managed in primary care in Singapore and globally [2]. In the Global Burden of Disease Study 2019, elevated blood pressure (BP) was one of the top five risks of attributable deaths and a key risk factor for ischemic heart disease and stroke, which were the leading causes of disability-adjusted life years of people aged 50 years and above [3, 4].
Besides lifestyle modifications and pharmacological treatment if indicated, telemedicine has increasingly been considered as an integral part of hypertension management in primary care to improve access to services, quality of care, productivity and prevention of cardiovascular diseases [5]. Telemedicine refers to the systematic provision of healthcare services remotely via information and communications technology, and includes tele-collaboration, tele-treatment, tele-monitoring and tele-support in Singapore [6]. It has been used in the management of chronic diseases in primary care, such as asthma, diabetes, hypertension and multimorbidity, to improve disease control and quality of life [7–11].
The most well-received telemedicine application for hypertension management is BP telemonitoring, which allows BP readings to be transmitted remotely from the patient’s home to the physician [12]. Studies have shown that BP telemonitoring, along with telecounseling and management by a team of healthcare professionals (HCPs), is associated with reductions in BP, healthcare utilization, mortality and cost, and improvements in patient self-management, empowerment, quality of life and patient-provider relationship [13–15]. The Telemonitoring and Self Management in the Control of Hypertension (TASMINH2) and the Telemonitoring and Self-management in Hypertension (TASMINH4) trials in the UK [16–18] and the Scale-Up BP study in Scotland [19] showed that BP telemonitoring is feasible in primary care, facilitates better medication management and well-received by patients and HCPs, who are important stakeholders in its implementation. Nevertheless, a multitude of cultural practices, technical, individual and organizational factors which differ between countries can influence the safe and effective implementation and use of telemedicine.
Singapore is a multi-ethnic and urbanized country in Asia, with a small population size of 5.7 million people and high mobile phone and internet penetration rates [20, 21]. Like many developed countries, *Singapore is* facing the challenges posed by an ageing population, increasing chronic disease prevalence, and rising healthcare cost. Primary healthcare in Singapore comprises unsubsidized private clinics and subsidized public institutions (polyclinics), with the latter managing most chronic disease patients. Telemedicine was already in use for chronic disease management in primary care before the COVID-19 pandemic in the form of nurse-based tele-support services [22]. The pandemic has propelled its use to unprecedented levels, as it enabled the continuation of chronic disease management while adhering to safe distancing measures.
The Primary Technology-Enhanced Care – Hypertension (PTEC-HT) pilot trial showed that there is an opportunity for telemedicine to improve clinical outcomes and cost-effectiveness of chronic disease management programs in Singapore’s primary care. PTEC-HT was a quasi-experimental trial conducted in a polyclinic in central Singapore with patients recruited from September 2018 to March 2019. The study methodology and main results have been reported elsewhere [23]. 217 patients with hypertension (office BP ≥ $\frac{140}{90}$ mmHg) or hypertension with hyperlipidemia were recruited to the intervention or usual care group and followed up for 6 months. All patients were cared for by healthcare teams called teamlets, each comprising two family physicians, a care manager who is a nurse trained in chronic disease management, and a care coordinator who is a lay person that follows up with patients on their appointments. Each participant in the intervention group was loaned a Bluetooth-enabled home BP monitor and mobile data network connecting gateway device, and was asked to monitor BP at least once weekly. The BP readings were automatically uploaded to a web-based portal for clinical management. Patients with well-controlled BP were reviewed regularly through telephone consultations with the care managers (scheduled teleconsultations). Care managers would contact patients if unexpectedly high readings were detected to check on their condition (unscheduled teleconsultations). Medications were adjusted over the phone (medication titration) if the physicians deemed it clinically necessary. The intervention of BP telemonitoring combined with teleconsultation was found to have improved BP control and patient satisfaction, and was cost-effective.
The current qualitative study, embedded within the PTEC-HT trial, explored the experiences and perspectives of patients and HCPs on the barriers and facilitators to the implementation and use of the intervention in the Singapore primary care setting.
## Recruitment of interview participants
To capture a range of experiences of implementing and using the intervention, the research team identified HCPs who were involved in the implementation and patients who had remained in the trial to participate in the current study. Twenty patients were selected purposively and invited by telephone, and a maximum variation sample was sought based on age and sex characteristics. Ten HCPs were identified and invited via email. Due to the scope of the ethical approval, patients who withdrew from the trial could not be recruited into the current qualitative study.
## Data collection
Individual semi-structured interviews were chosen to elicit participant thoughts and feelings about the implementation of the BP telemonitoring intervention [24]. Interview questions were developed based on the study aims, literature review and discussions within the research team. The topic guide (Additional file 1) was pilot-tested and covered experiences and opinions on using the telemonitoring equipment, BP self-monitoring, teleconsultation, medication adjustments and lifestyle modifications. Additional probes were used to clarify and discover in-depth information. Data collection and analysis were carried out concurrently. The topic guide was refined iteratively as needed, including adding prompts and questions, based on emerging findings from preceding interviews.
All participants provided written consent for this study. The interviews were conducted face-to-face with each participant between March and September 2019, each interview lasted 20–60 minutes. Two investigators (SHT and EALC) who are trained in qualitative methods conducted the interviews in English or Mandarin in an enclosed room in the polyclinic. Each interview was audio-recorded and transcribed verbatim. Repeat interviews were not carried out. Transcripts were anonymized to ensure data confidentiality.
Ethical approval was received from the National Healthcare Group Domain Specific Review Board.
## Analysis
Each transcript was checked for accuracy against the recording. Data analysis was conducted following the framework method [25].
Two investigators (SHT and EALC) read and re-read the transcripts independently to familiarize with the data and independently coded the data. SHT is a research fellow with a master’s degree in public health and EALC is a research fellow with PhD in Communications. Data were analyzed inductively to examine the perspectives of different participants, highlighting similarities and differences, and generating unanticipated insights [26]. Data gathered from patients and HCPs were analyzed separately. A reflexivity journal was maintained by the investigators to record their views before the analysis. Initial codes were discussed and a working analytical framework was agreed upon. Constant comparison was employed to ensure consistency in coding. Differences in opinion were resolved through consensus on the main themes and subthemes that emerged from the data. An Excel spreadsheet was used to generate a matrix to chart the data with the cases in rows, codes in columns and summarized data in the cells. After coding all transcripts, relationships among categories were explored to raise the analytical level from categorical to thematic. Thematic data saturation was reached after the 21st interview of both patients and HCPs, as no new themes emerged from the data by then. The different codes, categories, and themes were grouped into a coding tree chart that illustrated the patient and provider perspectives on the BP telemonitoring and teleconsultation intervention.
The themes were further mapped to the dimensions of the socio-technical systems (STS) model to understand the relationships between the sociotechnical aspects of the intervention [26]. The STS model (Fig. 1) was developed by Sittig and Singh to comprehensively study the interconnectedness between the key factors that influence the success of health information technology interventions within complex adaptive healthcare systems like primary care. It consists of eight dimensions [27, 28]:[1] Hardware and software – This dimension focuses on the technical aspects required to run the applications, including the physical equipment and software.[2] Clinical content – This dimension considers the data collected that are useful for patients and HCPs to manage patient care and inform clinical decisions.[3] Human-computer interface – This dimension includes aspects of the system that users can see, touch or hear as they interact with the application.[4] People – This dimension represents the patients and HCPs who are involved in the implementation and use of the intervention, and how the intervention helps them think and makes them feel.[5] Workflow and communication – This dimension focuses on the interaction between the patients and HCPs to accomplish patient care as well as the impact of the intervention on the HCP workflows.[6] Internal organizational policies, procedures, and culture – This dimension can affect every other dimension in the model, as organizational policies and procedures can affect the availability of hardware and software and clinical workflows related to the intervention.[7] External rules and regulations – This dimension focuses on the external regulations and priorities that can facilitate or impede the implementation and use of the intervention.[8] System measurement and monitoring – This dimension focuses on measuring and monitoring the availability, usage, effectiveness and outcomes of the intervention regularly. Fig. 1Socio-technical systems (STS) model by Sittig and Singh The latter two dimensions were omitted from this study as the analysis did not find any relevance of the data to these dimensions.
NVivo 12 (QSR International Pty Ltd) was used for data management and analysis. This study is reported according to the *Consolidated criteria* for Reporting Qualitative research (COREQ) checklist [29].
## Results
Thirteen patients and eight HCPs were enrolled into the current study. Participant characteristics are shown in Tables 1 and 2 for patients and HCPs, respectively. $61.5\%$ of patients interviewed were males and mean age was 55.7 years (range from 35 to 73 years). Most patients were Chinese, attained pre-university or tertiary education level and working full-time. We interviewed two family physicians, one nurse clinician, three care managers and two care coordinators. Most HCPs were females and the number of years in practice ranged from 1.5 to more than 40 years. Table 1Participant characteristics: Patients ($$n = 13$$)Characteristicn (%)Gender Male8 (61.5)Ethnicity Chinese12 (92.3) Malay1 (7.7)Highest education level Primary1 (7.7) Secondary3 (23.1) Pre-University2 (15.4) Tertiary7 (53.8)*Employment status* Working part-time2 (15.4) Working full-time9 (69.2) Retired2 (15.4)Table 2Participant characteristics: Staff ($$n = 8$$)Characteristicn (%)Gender Female7 (87.5)Ethnicity Chinese8 (100.0)Profession Family physician—Approved medication titration; provided clinical input2 (25.0) Nurse clinician—Supported backend coordination of implementation of intervention1 (12.5) Care manager—Performed teleconsultation and monitoring of BP readings3 (37.5) Care coordinator—Trained participants to use remote BP monitor; provided follow-up technical support2 (25.0)Years in practice 1–42 (25.0) 5–91 (12.5) 10–142 (25.0) 15–191 (12.5) 20 and above2 (25.0) We reported our key findings classified into themes presented by the applicable STS dimensions. The emergent codes, categories, and themes were represented in a coding tree chart (Fig. 2). Representative quotes are illustrated in Additional file 2 and Fig. 3 summarizes the study findings mapped to the STS model. Fig. 2Coding tree of patient and provider perspectives on the intervention, mapped to the STS dimensionsFig. 3Study findings mapped to the socio-technical systems model. Green boxes represent emergent themes and subthemes, green arrows represent additional relationships identified from the study
## Usability and functionality
Six patients commented that the devices were fast and easy to use, with clear indications and instructions by the care team. Four patients were below 50 years old, two were above 60 years old, and all were in full-time or part-time employment.
Nevertheless, patients encountered technical challenges with the devices. Four patients were frustrated that the cuff did not fit them well. F035 (male, 67 years old) lamented that it was “really big, stiff!”, and F009 (female, 47 years old) was frustrated as “it keeps dropping down…By the time I can take the measurement, I think my blood pressure shoot up already.” S003 (care manager) acknowledged the issue, “So the first batch they were all quite stiff, very hard to apply. So no matter how you do it, it’s always quite loose…” Cuffs were subsequently replaced for some patients during the initial stage of the trial to prevent inaccurate readings and loss of motivation from patients.
Many patients also experienced challenges with the gateway, which they referred to as “phone”. They described the connection issues and that it would switch to flight mode unknowingly as the option was close to the power button, thereby affecting the transmission of readings. Some found the gateway cumbersome as it added to the number of devices they had to manage. The care coordinators, whom patients usually approached for troubleshooting, recalled that some patients withdrew from the trial due to the gateway issues. S001 escalated some issues to the vendor, while S007 was concerned that these performance issues might affect patients’ confidence in the intervention and relationship with the care team “like they lose trust in our treatment because the devices don’t work well or not as accurate as how they expect it to be.” When asked about how the issues could be ameliorated, most patients, care managers and care coordinators expressed preference for a more user-friendly machine that could integrate BP measurement and data transmission more seamlessly. Patients who provided suggestions were all professionals with tertiary education, including IT professionals, managers, and teachers. F022 who was working in senior management, remarked that the intervention was outdated and advised the team to obtain user experience feedback as the intervention was being implemented and enhanced. User experience is something that’s very important…put on a hat of a profile of the client, whether it’s a particular age group, then you try to empathize how that person feels…Maybe some of them are not so technologically-savvy…go through with them…and make it simpler. Interaction can be just a one-touch button, to make it more user-friendly. Things should be intuitive. ( F022, male, 64 years old)
## Availability of records
The transmitted BP readings were only accessible through the web-based clinical management portal by the HCPs, and patients did not know how to retrieve the history from the BP monitor. Therefore, F096 and F099 kept their own record through methods such as taking a video or photo of the readings displayed on the monitor, or manually in a book.
Three patients expressed desire to have access to their readings, so that they could share with other healthcare providers, or monitor the BP trends and motivate themselves to adopt a healthier lifestyle. I think it’ll be good if you can aggregate the readings over a month and say…I think you’re doing very well or not doing too well, and maybe nudge you, you may like to exercise or you might want to control your diet, less salt…I think it is a behavioral thing that can nudge you to a more healthy lifestyle…I think every participant would benefit. ( F022, male, 64 years old) They also suggested having an integrated mobile application where patients could readily communicate with the care team or access their BP data and other useful information, such as contact numbers of the care team and frequently asked questions. HCPs interviewed concurred as patients could feel empowered and a sense of ownership when they have access to their own data.
A concern arising from having data access, and with consultation and medication titration being done over the phone, was data confidentiality as unauthorized personnel may impersonate as HCPs. Patients proposed verification methods, such as two-factor authentication or using the device serial number, to ensure that the call is genuinely from the polyclinic.
## Usability of clinical management portal
The care coordinators and care managers who regularly used the portal for enrolling patients and monitoring patients’ BP noted that its functions and interface were manual and not user-friendly. This slowed their work down and they took some time to adapt to it, affecting the feasibility of the intervention.
Not the easiest to navigate… it’s not very neat and tidy. The buttons are like everywhere… So not the most intuitive to use, took me awhile to get used to it. ( S003, female, care manager) It’s not that bad but it’s not very straight-forward…But of course can improve much more… the speed is not very fast... Due to PDPA [Personal Data Protection Act], it keeps logging us out halfway… It feels like you wasted all your documents. A little bit complicated... A lot of info. ( S007, female, care coordinator)
## Impact on users
The participants described the emotional to physical to behavioral impact that they experienced since they started using the intervention.
## Caused technology anxiety
Three patients, who were above 60 years old and were retired or working part-time, described how they became more anxious since participating in the trial. Their concerns were—afraid to see high BP readings, errors with transmission of readings, unable to reach polyclinic hotline, and found it a chore to remember to measure BP. Two of these three patients had uncontrolled BP (≥ $\frac{140}{90}$ mmHg) after 6 months of trial participation. Staff also recalled encountering anxious patients, some of whom even withdrew from the trial.
[It is] okay but also sometimes gives you anxiety. Like if high, you [get] scared… Sometimes it could be a chore, you have to remember to measure. ( F099, female, 64 years old) Sometimes the hotline can be very busy, and if you are very urgent, you can’t get through, you can get panicked. ( F122, female, 73 years old) Three patients also feared losing or damaging the BP monitor, as they perceived it to belong to the government and compensation would be a consequence. Due to this, F007 (male, 47 years old) kept the devices in the box after each measurement, which became a barrier and “makes it more cumbersome” to take regular BP readings since it was kept out of sight.
## Personal schedules
Most patients and HCPs appreciated the convenience brought about by the intervention. It saved patients’ clinic visits, travelling and waiting time, especially for the working individuals. They could also monitor at a time that suited their schedules. Patients recounted that they no longer had to record BP readings manually on a monitoring sheet, and HCPs could monitor the readings from the portal easily without having to wait for the patients to bring the physical copy at their next clinic visit. One thing is because I can easily monitor it at home, only one week once, out of my own convenience, how I want to check it. Sometimes I do it after I came back from work, after I have a wash, relax myself… You just have to press a button, it got transmitted... for you. Rather than you have to jot down…the timing… (F096, female, 46 years old) During the intervention period, patients received scheduled and unscheduled teleconsultations from the clinic to follow up on their BP readings. While F029 (male, 49 years old) liked that the timing was flexible as “at that time [when they called] I was busy so I rescheduled…”, three other patients described how scheduled calls were not punctual despite having made appointments, or that the calls did not match their work schedules, resulting in rejected or missed calls. Office hours and sometimes, you are in the midst of doing something, and then the call come in. So, it’s like a lot of interruption…Your care team, they have a job to do. They follow up and all. I mean that’s good. But for people working, I think it’s a little bit inconvenient. Especially the few times when they called me, I was in the meeting with my boss. So, I have to say, “No, I’m busy now. I cannot talk to you." ( F119, male, 58 years old) The care managers, who usually made the calls, mentioned that calls to most patients were completed smoothly. However, there were instances of unfriendly remarks from patients or unresponsiveness and they had to make multiple calls to patients which added to their workload. Sometimes they work same times as us…they will need to keep the hand phones [aside] so they can’t answer… It’s not that they don’t want to [answer]. ( S003, female, care manager)
## Behavior change
Six patients, mostly working individuals, reported that the intervention had helped instill a habit, routine and discipline of measuring BP regularly, which would otherwise not be possible even if they had owned a BP monitor. Even though you can buy [a BP monitor], you don't have the discipline to check, but with this thing, you know that you have to check one week once, so there is no excuse that you don't have to check…that is one thing good about it... The responsibility is there... (F096, female, 46 years old) Patients became more motivated and improved their self-management and health literacy of hypertension from the advice provided by the care team. They grew more aware of their own triggers of BP fluctuations and ways to maintain a healthy BP in terms of diet, exercise and emotions. F022 felt that participation in the trial allowed patients to have “more ownership of managing their own healthcare”. S003 noted that patients had shown “good outcomes”, and S007 found that whether patients were new to home BP monitoring or not, the intervention “gives them an exposure to start monitoring themselves…Empowered, maybe.”
## Patient profile
The participants identified the patient profiles that are more suitable for the intervention than others. These include personal schedules, adaptability, age and digital literacy, health literacy and social support, and physical condition. They felt that the intervention would be more appropriate for certain individuals – busy, retired, adaptable or accepting of technology, possess some level of health and digital literacy regardless of age, or have social support if they are unable to self-manage. Participants also highlighted that the intervention would benefit those with poor BP control and had a risk or history of stroke to keep them motivated and prevent their condition from worsening.
Among the study participants, those who were younger and received tertiary education were more accepting of teleconsultation and phone medication titration. Contrastingly, the older patients preferred traditional face-to-face mode of care for better communication or when discussions involved medications. F035 preferred physical consultations as he could obtain the test reports from the clinic and that “through the phone, maybe communication breakdown” as the elderly may have hearing difficulties. HCPs agreed that physical consultations might be more useful for elderly patients, or for those who required monitoring of medication compliance. S003 felt that “patients may feel more confident to have a face-to-face [consultation] especially those like elderly patients” and that people in Singapore are generally “not used to talk to a person over the phone asking many health-related questions”.
## Communication process
The participants discussed several aspects of the intervention relating to internal workflows and engagement between patients and HCPs.
## Workload management
Through the intervention, HCPs could receive patients’ BP readings remotely and conduct phone medication titration quickly if warranted. This raised conflicting views from HCPs about the impact on their workload.
In the initial stages of the trial, S005 (female, care manager) had to “find pockets of time” to check each patient’s BP readings monthly for fluctuations through the portal and phone calls, “whether it is urgent or not”. Subsequently, a Jobstack function was created and helped the healthcare team prioritize certain patients for closer monitoring and unscheduled teleconsultations. This freed up some time for them to focus on other tasks. Since then, S005, S006 and S007 saw better workload management as they could spend their clinic time attending to patients with more complex needs, while conducting phone medication titration and teleconsultation for patients with stable BP.At individual level, the time that I usually spend for my patient [usual consult] compared to now for this [intervention], it has reduced a little. But overall, if I see it as a whole, for the 120 patients, it does help to ease out the appointment for other patients. If I get a lot of them to be on teleconsult, so I free up some of the appointments for other more complex cases. ( S006, male, family physician) On the contrary, the intervention become overwhelming for some HCPs. S002 found an increase in workload as she had to manage patients with clinic appointments and phone follow-up with the trial participants concurrently. It’s quite packed every day… Sometimes [the time] I spend [on] the call is very long, so the patient actually in the clinic was waiting. These are the difficulties. ( S002, female, care manager) One family physician felt that the effort involved in teleconsultation and phone medication titration in this pilot trial was not acknowledged but was hopeful that the workload might improve in future. All [done] remotely, [patients] don’t have to come back. So in a way, [they] save time, right? Because they don’t come in and take our slot. But the bad thing is that it is not accounted for, it’s just extra work... Maybe in the initial phase, there will be increased workload, but in the long run once everyone is optimized… there may be a decrease in the workload. ( S004, female, family physician)
## Patient-provider engagement
The patients and HCPs reflected on the engagement between each other relating to the relationship built and support provided during the intervention.
Patients reported a sense of security, as they knew that they were being monitored closely by the healthcare team, who would alert them of any irregular or missing BP readings. F035 (male, 67 years old) even exclaimed “I feel happy!” when he received an unscheduled teleconsultation for a rise in BP. This also motivated him to measure BP regularly and look out for triggers. HCPs concurred that issues with patients’ BP could be addressed in a timelier manner through this intervention as the healthcare team could intervene quickly, if needed, through a teleconsultation or phone medication titration. HCPs felt that their relationship with patients improved, as the ready technical support rendered by the care coordinators, frequent phone calls and encouragement by the doctors and care managers, and BP improvements contributed to the building of rapport and between patient and HCPs.
Conversely, four patients were expecting to receive feedback on their BP readings from the healthcare team during the trial, but they did not seem to have received any phone calls. Two were aged 47–49 years old and two were aged 64–73 years old. This affected their confidence in the patient-provider engagement, as F007 wondered “if someone is looking at my records or just for the sake of collecting data”. They preferred more regular interaction with the healthcare team to understand HCP workflows in relation to monitoring BP readings, provide feedback on the process and receive advice on their BP readings and lifestyle management.
F029 (male, 49 years old) recounted that she had to return to the clinic to refill her medications as the doctor increased the dosage over phone medication titration. However, the pharmacy was not informed and she “had to wait for quite some time because I guess they need to clarify with the doctor… So if that link is there, it’s more convenient.” This also highlighted that the healthcare teams need to work with other clinic staff who might not be part of but still support the core team, such as the pharmacist.
## Internal policy
HCPs interviewed, especially the care managers and care coordinators, were concerned about the limited manpower and time available, which would also affect the workload of the intervention. They wondered whether adjustments to internal allocation of resources would be possible in order for the intervention to sustain.
If we have a mobile phone to call the patients, if patients keep on texting or messaging, then the nurse will be very stressed also. So I don’t know whether in the future [there] will be a coordinator [who] have some nursing background [and] able to man the phone… (S002, female, care manager) Maybe give us a bit of extra time every day to just handle things like this. I would say, minimum maybe 15 minutes a day? If not, half an hour would be better, to go through the cases. ( S004, female, family physician) During the initial recruitment stage, like on standby, we can have more care coordinators and care managers in the same area to prepare to recruit patients. Because sometimes when it’s just one care coordinator, it gets piled up more and more. Then every recruitment could get very long, half an hour to an hour... But at the same time, it’s quite hard because the doctors and care managers also have to see other patients… So that part in workflow, it’s more on team communication also. ( S007, female, care coordinator)
## Discussion
The purpose of this study was to assess the barriers and facilitators to the implementation and use of the PTEC-HT BP telemonitoring system with teleconsultation, to inform future expansion of the intervention. Our analysis showed that the technical, human, workflow and organizational factors involved were complex and interdependent, and the STS model provided a comprehensive framework to understand the relationship between these dimensions.
Our study findings further substantiated the patient satisfaction survey results of the PTEC-HT pilot trial, in which patients agreed that it was convenient to record and share BP measurements with the healthcare team, were motivated to monitor BP weekly and were satisfied with the advice provided through teleconsultations [23]. Overall, patients had a positive experience using the intervention although they encountered technical difficulties and expressed concerns relating to data security and technology anxiety. Through implementing the intervention, HCPs found that it helped with workload management and useful for patients, despite challenges with usability and resource allocation. Suggestions to overcome these obstacles were provided, for the intervention to be feasible and sustainable in Singapore’s primary care setting.
## Comparison with prior work
The themes that emerged from the current study were largely consistent with the literature which reported that technological usability, reduction of in-office visits, positive impact on self-management and improved patient-provider relationship were the most frequently reported facilitators for adoption of telemedicine [30]. However, several differences exist.
## Hardware and software, human–computer interface, people
Our study participants found the intervention easy-to-use, and patients could save time travelling to and waiting in the hectic polyclinic setting, or recording BP readings manually. HCPs also had quicker access to BP readings and could conduct teleconsultation, if necessary, thereby improving their efficiency. Fletcher et al. reported that patients liked the ability to be able to monitor BP whenever they wanted as they were more relaxed outside the clinic environment [31]. This was reflected by a patient in this study who attributed the convenience to having the ability to monitor any time she preferred in a week.
Patients in our study saw changes in their attitudes and behaviors in managing hypertension, like previous studies conducted in the US and UK [16–19, 32, 33]. They appreciated that the intervention helped to form a habit of regular BP self-monitoring and improved their self-management. Improvements in BP readings and lifestyle behaviors coupled with timely feedback from the care team, increased their awareness, motivation and sense of responsibility for their own condition.
Participants experienced technical challenges with using the equipment and clinical management portal. Firstly, issues with the cuff and gateway and having to handle multiple cumbersome devices caused frustration and affected patients’ motivation to continue the intervention. They preferred the functions of measuring, recording and transmission of BP readings to be integrated in one machine. Similar to interventions targeting home blood glucose and BP telemonitoring for patients with diabetes, the equipment was not “plug and play” for some patients and higher digital literacy and frequent assistance from the care team was needed to overcome these obstacles [32]. Secondly, while HCPs were able to receive patients’ BP readings quickly through the management portal, they did not find it intuitive. Additionally, they had to check through BP readings of all patients in the initial stage. These translated to additional workload and time to navigate the portal and affected their efficiency, until the Jobstack function was created to help with prioritizing patients for closer management.
Patients and clinicians were found to be uncomfortable with the interpretation of self-monitoring measurements due to the difference between home and office readings [31]. Some of our study participants were ambivalent due to the technical challenges, while some perceived home BP readings to be more accurate than office readings which could otherwise be confounded by white coat hypertension. Besides, the Ministry of Health Clinical Practice Guidelines for hypertension indicated home BP monitoring for monitoring BP in diagnosed patients. The guidelines proposed that home BP monitoring is cheaper, more widely available, easily repeatable and shows day-to-day variability, and it can be offered to committed patients to improve treatment adherence via positive data feedback [34].
## Clinical content, people
Patients in the current study wished to have access to their data for sharing with other HCPs or self-management, and some form of feedback or advice even if their readings were satisfactory. These could encourage them to become more engaged as active managers of their healthcare thereby facilitating self-management.
However, while most patients were comfortable with data being shared between patients and HCPs, some were concerned about data privacy and security especially when the intervention involved medication titration. This is a common concern reported in the literature, and it was expected due to the health data security breaches previously reported in Singapore [22, 35, 36]. Singapore has the most comprehensive telemedicine guidelines in Southeast Asia that provides advice on various aspects of telemedicine usage, and is comparable with other countries [37]. An example is the Personal Data Protection Act (PDPA) mentioned by S007, which governs the collection, use, disclosure and care of personal data in Singapore. Unfortunately, it created a barrier for HCPs as they were logged out from the portal frequently possibly to prevent unauthorized access. These external regulations can help to inform the development of organizational policies, which would be important for allocating budget and stipulating policies and workflows for authentication and data backups, and help HCPs provide safe and effective care to patients.
## People
There was variability in the use of the intervention by certain populations in this study. Older patients experienced technology anxiety as they were afraid to see abnormal readings or were worried that the readings did not get transmitted. They also preferred face-to-face consultations possibly for the human connection [38, 39]. Contrastingly, younger patients or those with higher education level and busy lifestyle welcomed telemonitoring, teleconsultation and phone medication titration. Perhaps a common issue in Asia, a Malaysian study also found that older patients found the telemonitoring device unfamiliar and had difficulties using it [40]. This could be due to the different levels of health literacy, digital literacy and technology acceptance, as shown in a study on perception towards digital health among elderly Singaporeans [41]. Sin et al. found that only a slim majority of patients with diabetes and/or hypertension were willing to adopt telemedicine, and care satisfaction with using telemedicine was a significant factor [35]. Although *Singapore is* technologically advanced, levels of comfort with health technologies still vary considerably.
Participants also proposed higher utility of the intervention by specific populations, and this suggests the need for patient segmentation to craft and target strategies to assist patients with various needs to achieve maximum benefit. The personae proposed by Low et al. or the Bloem-Stalpers model could be considered to segment patients according to their socioeconomic and sociodemographic characteristics, acceptance of technology, or acceptance and perceived control of their condition [39, 42].
## Workflow and communication, people, organizational policies and culture
Whereas patients and HCPs were reported to be concerned about telemedicine hindering the patient-provider relationship [30], those in our study felt that the frequent support and encouragement provided to the patients improved this relationship. The intervention complemented the relationship, rather than interposing as an intruder, by providing timely and accurate information via telephone calls in between clinic consultations [43]. Patients felt reassured that their health was being closely monitored, through the quick feedback from the healthcare team especially when abnormal readings were recorded. However, the practice of conducting teleconsultations or tele-support was not consistent as although patients were informed upon enrolment to expect calls from the healthcare team during the intervention period, some did not receive any and might perceive this as a lack of accountability by the institution.
While other studies mainly focused on the doctor as the provider of the intervention, team-based care was provided at the polyclinic as teamlets. However, one patient experienced confusion as there was a lack of communication with the pharmacists, who usually supported the teamlets. Coordinating workflows with other support staff ensures patient safety in the form of timely supply of patients’ medications and can also help to maintain patients’ motivation for the intervention.
Although there was divergence of HCPs’ opinions on whether the intervention helped with workload management, they agreed on the need to allocate more resources to prevent burnout, and for the intervention to be sustainable in the long run. Although geographical barrier is less of a concern in Singapore than in other countries where healthcare may not readily accessible, the COVID-19 pandemic has created an invisible barrier between HCPs and patients in the form of safe distancing and reallocation of clinic manpower to support COVID-19 operations and other priorities. Telemedicine, with adequate planning of resources and workflows, can bridge this barrier by sustaining BP management and providing care to patients even in times when physical visits are not possible [44].
The study populations of previous studies were mainly from the West and may not be translated to the Singapore or Asian context. The interventions in these studies included monitoring, medication titration and contact initiated by patients. However, the PTEC-HT intervention required HCPs to initiate contact with patients through scheduled or unscheduled teleconsultation, which was not available in the referenced studies. This provided sufficient support right from the start, especially for patients with technology anxiety or low health or digital literacy. Teleconsultation can provide reassurance, improve patient education and health communication and contribute to behavior change, addressing the areas for future research raised by Fletcher et al. [ 31].
As we focus on patient-centredness in the implementation and use of telemedicine interventions, we can incorporate principles of service design and human factor approaches to achieve greater impact and better experience for our patients and HCPs [45, 46]. Study participants suggested that it is imperative to obtain and incorporate user feedback earlier in the development of user-friendly telemedicine interventions which prioritize effectiveness, patient safety, data confidentiality and the needs of patients and HCPs.
## Strengths and limitations
This study was conducted within a quasi-experimental trial using qualitative methodologies. It provided insights into the views of patients and HCPs on the implementation and use of BP telemonitoring system with teleconsultation in a real-world setting. Integration with routine clinical practice and the ability to adapt features of the intervention encouraged uptake by both patients and HCPs. Using semi-structured interviews allowed us to build rapport with the participants and obtain rich data about their experiences. We interviewed patients and HCPs to obtain a diversity of views as they played different roles in the process.
Applying the STS model in the analysis allowed us to study the intervention on a more holistic level and identified some complexities and important considerations as we plan and improve the implementation and use of telemedicine interventions in primary care. The dimensions of “external rules, regulations and pressures” and “system measurement and monitoring” were not addressed in this study as participants did not express any related views. Future research could consider interviews with developers of the intervention and healthcare administrators to understand the impact of any external forces and the expected outcomes of the intervention.
There have been several other models developed to study health technologies [27, 43], but few provide a holistic framework to do so. The exception is the nonadoption, abandonment, scale-up, spread and sustainability (NASSS) framework by Greenhalgh et al., which is very flexible and useful to study technology implementation at the micro, meso and macro-level [47]. While the current study was more limited in scope which focused more on the users’ perspectives and the impact of the intervention on the users, communication and practice workflow in one polyclinic, a longitudinal qualitative approach using the NASSS framework could be used to study the adaptation, scale-up and adoption or non-adoption of the intervention over time in more polyclinics. The scaling of the PTEC-HT program with more integrated features is currently underway [48].
Notwithstanding the efforts to obtain a representative sample of patients to reflect demographics of the clinic, ethnic diversity was limited. Self-selection might also have occurred from voluntary participation. Compared to other participants in the PTEC-HT pilot trial, patients in this study were younger (mean age 50.5 vs 56.3 years) and a larger proportion attained tertiary education qualification ($53.8\%$ vs $25.2\%$), and they might have been more willing and able to share their views.
The patient interviews were conducted when the patients had completed the 6-month trial in hope to maximize their experience and elucidate their perspectives of what did or did not work. However, there was a possibility that some patients could not recall details from the early months of their participation accurately. Future research may also involve views from patients who rejected participation or withdrew from the trial to understand their reasons and concerns. As the HCPs were part of the teamlet tasked to implement the intervention, they might have been supporters of the intervention so their views might not be representative of other primary care HCPs.
As this was a qualitative study conducted in one public primary care clinic in Singapore before the COVID-19 pandemic, transferability of findings to other primary care clinics or non-primary care settings is limited as operational procedures may differ between institutions. However, the challenges identified from this study, such as digital literacy and data security, are still largely relevant even as we are adapting our healthcare services since the pandemic happened [49].
## Conclusions
Our study provides comprehensive perspectives on the barriers and facilitators to the implementation and use of a BP telemedicine intervention in a complex fast-paced primary care setting. It can engender various benefits and challenges to patients, healthcare professionals and the healthcare system. An intervention that is well-designed and successfully implemented can be an enabler to improve healthcare delivery, patient journey and safety, as well as avoid unplanned hospital admissions and visits. It is hoped that the insights gleaned can help to guide various stakeholders understand the importance of the various dimensions as they embark on telemedicine program design and implementation.
## Supplementary Information
Additional file 1. Additional file 2.
## References
1. 1.Epidemiology & Disease Control Division and Policy, Research & Surveillance Group, Ministry of Health and Health Promotion Board, Singapore. National Population Health Survey 2020 (Household interview and health examination); 2021 November [cited 2021 Dec 30]. 185 p. Available from: https://www.hpb.gov.sg/docs/default-source/default-document-library/nphs-2020-survey-report.pdf?sfvrsn=ece50aa9_0.
2. Finley CR, Chan DS, Garrison S, Korownyk C, Kolber MR, Campbell S. **What are the most common conditions in primary care?**. *Can Fam Physician* (2018.0) **64** 832-840. PMID: 30429181
3. 3.GBD 2019 Diseases and Injuries Collaborators. Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2020;396(10258):1204–22.
4. 4.GBD 2019 Risk Factors Collaborators. Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2020;396(10258):1223–49.
5. Wang JG, Li Y, Chia YC, Cheng HM, Minh HV, Siddique S. **Telemedicine in the management of hypertension: evolving technological platforms for blood pressure telemonitoring**. *J Clin Hypertens (Greenwich)* (2021.0) **23** 435-439. DOI: 10.1111/jch.14194
6. 6.Ministry of Health, Singapore. National Telemedicine Guidelines; 2015 January [cited 2022 Oct 24]. Available from: https://www.moh.gov.sg/docs/librariesprovider5/resources-statistics/guidelines/moh-cir-06_2015_30jan15_telemedicine-guidelines-rev.pdf.
7. Kew KM, Cates CJ. **Home telemonitoring and remote feedback between clinic visits for asthma**. *Cochrane Database Syst Rev* (2016.0) **2016** CD011714. PMID: 27486836
8. Chongmelaxme B, Lee S, Dhippayom T, Saokaew S, Chaiyakunapruk N, Dilokthornsakul P. **The effects of telemedicine on asthma control and patients’ quality of life in adults: a systematic review and meta-analysis**. *J Allergy Clin Immunol Pract* (2019.0) **7** 199-216.e11. DOI: 10.1016/j.jaip.2018.07.015
9. Wu C, Wu Z, Yang L, Zhu W, Zhang M, Zhu Q. **Evaluation of the clinical outcomes of telehealth for managing diabetes - a PRISMA-compliant meta-analysis**. *Medicine* (2018.0) **97** e12962. DOI: 10.1097/MD.0000000000012962
10. Ma Y, Zhao C, Zhao Y, Lu J, Jiang H, Cao Y. **Telemedicine application in patients with chronic disease: a systematic review and meta-analysis**. *BMC Med Inform Decis Mak* (2022.0) **22** 105. DOI: 10.1186/s12911-022-01845-2
11. Kraef C, van der Meirschen M, Free C. **Digital telemedicine interventions for patients with multimorbidity: a systematic review and meta-analysis**. *BMJ Open* (2020.0) **10** e036904. DOI: 10.1136/bmjopen-2020-036904
12. Omboni S, Ferrari R. **The role of telemedicine in hypertension management: focus on blood pressure telemonitoring**. *Curr Hypertens Rep* (2015.0) **17** 1-13
13. Omboni S. **Connected Health in Hypertension Management**. *Front Cardiovasc Med* (2019.0) **6** 76. DOI: 10.3389/fcvm.2019.00076
14. Purcell R, McInnes S, Halcomb EJ. **Telemonitoring can assist in managing cardiovascular disease in primary care: a systematic review of systematic reviews**. *BMC Fam Pract* (2014.0) **15** 43. DOI: 10.1186/1471-2296-15-43
15. Margolis KL, Asche SE, Dehmer SP, Bergdall AR, Green BB, Sperl-Hillen JM. **Long-term outcomes of the effects of home blood pressure telemonitoring and pharmacist management on blood pressure among adults with uncontrolled hypertension: follow-up of a cluster randomized clinical trial**. *JAMA Netw Open* (2018.0) **1** e181617. DOI: 10.1001/jamanetworkopen.2018.1617
16. 16.Jones MI, Greenfield SM, Bray EP, Baral-Grant S, Hobbs FD, Holder R, et al. Patients’ experiences of self-monitoring blood pressure and self-titration of medication: the TASMINH2 trial qualitative study. Br J Gen Pract. 2012;62(595):e135–42.
17. 17.Jones MI, Greenfield SM, Bray EP, Hobbs FR, Holder R, Little P, et al. Patient self-monitoring of blood pressure and self-titration of medication in primary care: the TASMINH2 trial qualitative study of health professionals’ experiences. Br J Gen Pract. 2013;63(611):e378–85.
18. Grant S, Hodgkinson J, Schwartz C, Bradburn P, Franssen M, Hobbs FDR. **Using mHealth for the management of hypertension in UK primary care: an embedded qualitative study of the TASMINH4 randomised controlled trial**. *Br J Gen Pract* (2019.0). DOI: 10.3399/bjgp19X704585
19. Hammersley V, Parker R, Paterson M, Hanley J, Pinnock H, Padfield P. **Telemonitoring at scale for hypertension in primary care: an implementation study**. *PLoS Med* (2020.0) **17** e1003124. DOI: 10.1371/journal.pmed.1003124
20. 20.Department of Statistics Singapore. Infocomm and Media. 2022 March [cited 2022 Oct 24]. Available from: https://www.singstat.gov.sg/find-data/search-by-theme/industry/infocomm-and-media/latest-data.
21. 21.The World Bank Group. Individuals using the Internet (% of population) - Singapore. 2022 [cited 2022 Oct 24]. Available from: https://data.worldbank.org/indicator/IT.NET.USER.ZS?locations=SG&most_recent_value_desc=true.
22. 22.Wong SH, Cheng WR, Aw J. Medical, legal and ethical issues arising from the use of telemedicine in the primary care setting in Singapore. Singapore Med J. 2022. 10.11622/smedj.2022026.
23. 23.Teo VHY, Teo SH, Burkill SM, Wang Y, Chew EAL, Ng DWL, et al. Effects of technology-enabled blood pressure monitoring in primary care: a quasi-experimental trial. J Telemed Telecare. 2021;1357633X211031780.
24. 24.Adams WC. Conducting semi-structured interviews. In: Newcomer KE, Hatry HP, Wholey JS, editors. Handbook of practical program evaluation [Internet]. United States of America: John Wiley & Sons, Inc; 2015 [cited 2022 Oct 24]. Chapter 19. Available from: https://onlinelibrary.wiley.com/doi/book/10.1002/9781119171386.
25. Gale NK, Heath G, Cameron E, Rashid S, Redwood S. **Using the framework method for the analysis of qualitative data in multi-disciplinary health research**. *BMC Med Res Methodol* (2013.0) **13** 117. DOI: 10.1186/1471-2288-13-117
26. Braun V, Clarke V. **Using thematic analysis in psychology**. *Qual Res Psychol* (2006.0) **3** 77-101. DOI: 10.1191/1478088706qp063oa
27. Sittig DF, Singh H. **A new sociotechnical model for studying health information technology in complex adaptive healthcare systems**. *Qual Saf Health Care* (2010.0) **19** i68-74. DOI: 10.1136/qshc.2010.042085
28. Wesley DB, Schubel L, Hsiao CJ, Burn S, Howe J, Kellogg K. **A socio-technical systems approach to the use of health IT for patient reported outcomes: patient and healthcare provider perspectives**. *J Biomed Inform: X* (2019.0) **4** 100048. DOI: 10.1016/j.yjbinx.2019.100048
29. Tong A, Sainsbury P, Craig J. **Consolidated criteria for reporting qualitative research (COREQ): a 32-item checklist for interviews and focus groups**. *Int J Qual Health Care* (2007.0) **19** 349-357. DOI: 10.1093/intqhc/mzm042
30. Palacholla RS, Fischer N, Coleman A, Agboola S, Kirley K, Felsted J. **Provider- and patient-related barriers to and facilitators of digital health technology adoption for hypertension management: scoping review**. *JMIR Cardio* (2019.0) **3** e11951. DOI: 10.2196/11951
31. Fletcher BR, Hinton L, Hartmann-Boyce J, Roberts NW, Bobrovitz N, McManus RJ. **Self-monitoring blood pressure in hypertension, patient and provider perspectives: a systematic review and thematic synthesis**. *Patient Educ Couns* (2016.0) **99** 210-219. DOI: 10.1016/j.pec.2015.08.026
32. Koopman RJ, Wakefield BJ, Johanning JL, Keplinger LE, Kruse RL, Bomar M. **Implementing home blood glucose and blood pressure telemonitoring in primary care practices for patients with diabetes: lessons learned**. *Telemed J E Health* (2014.0) **20** 253-260. DOI: 10.1089/tmj.2013.0188
33. Baratta J, Brown-Johnson C, Safaeinili N, Rosas LG, Palaniappan L, Winget M. **Patient and health professional perceptions of telemonitoring for hypertension management: qualitative study**. *JMIR Forma Res* (2022.0) **6** e32874. DOI: 10.2196/32874
34. 34.Ministry of Health, Singapore. Hypertension - MOH Clinical Practice Guidelines 1/2017. 2017 November [cited 2022 Oct 24]. Available from: https://www.moh.gov.sg/docs/librariesprovider4/guidelines/cpg_hypertension-booklet---nov-2017.pdf.
35. Sin DYE, Guo X, Yong DWW, Tiu TY, Moey PKS, Muller-Riemenschneider F. **Assessment of willingness to tele-monitoring interventions in patients with type 2 diabetes and/or hypertension in the public primary healthcare setting**. *BMC Med Inform Decis Mak* (2020.0) **20** 11-35. DOI: 10.1186/s12911-020-1024-4
36. 36.Hossai I, Ang YN, Chng HT, Wong PS. Patients’ attitudes towards mobile health in Singapore: a cross-sectional study. mHealth. 2019;5:34.
37. Sabrina MI, Defi IR. **Telemedicine guidelines in South East Asia - a scoping review**. *Front Neurol* (2021.0) **11** 581649. DOI: 10.3389/fneur.2020.581649
38. Schwamm LH. **Telehealth: seven strategies to successfully implement disruptive technology and transform health care**. *Health Aff (Millwood)* (2014.0) **33** 200-206. DOI: 10.1377/hlthaff.2013.1021
39. Low STH, Sakhardande PG, Lai YF, Long ADS, Kaur-Gill S. **Attitudes and perceptions toward healthcare technology adoption among older adults in Singapore: a qualitative study**. *Front Public Health* (2021.0) **9** 588590. DOI: 10.3389/fpubh.2021.588590
40. 40.Abdullah A, Liew SM, Hanafi NS, Ng CJ, Lai PSM, Chia YC, et al. What influences patients’ acceptance of a blood pressure telemonitoring service in primary care? A qualitative study Patient Prefer Adherence. 2016;10:99–106.
41. Teo CL, Chee ML, Koh KH, Tseng R, Majithia S, Thakur S. **COVID-19 awareness, knowledge and perception towards digital health in an urban multi-ethnic Asian population**. *Sci Rep* (2021.0) **11** 10795. DOI: 10.1038/s41598-021-90098-6
42. Bloem S, Stalpers J, Groenland EAG, van Montfort K, van Raaij WF. **Segmentation of health-care consumers: paychological determinants of subjective health and other person-related variables**. *BMC Health Serv Res* (2020.0) **20** 726. DOI: 10.1186/s12913-020-05560-4
43. 43.Chew EAL, Teo SH, Tang WE, Ng DWL, Koh GCH, Teo VHY. Trust and uncertainty in the implementation of a pilot remote blood pressure monitoring program in primary care: qualitative study of patient and healthcare professional view. JMIR Hum Factors. 2023;10:e36072.
44. 44.Kok TWK, Chong SJ, Yau WKJ, Kumar PR, Chua SBR. Nationwide implementation of a centralised telemedicine platform in Singapore to fight the COVID-19 pandemic. J Telemed Telecare. 2022;1357633X221122890.
45. 45.Shaw J, Agarwal P, Desveaux L, Palma DC, Stamenova V, Jamieson T, et al. Beyond “implementation”: digital health innovation and service design. NPJ Digit Med. 2018;1:48.
46. Turner P, Kushniruk A, Nohr C. **Are we there yet? Human factors knowledge and health information technology - the challenges of implementation and impact**. *Yearb Med Inform* (2017.0) **26** 84-91. DOI: 10.15265/IY-2017-014
47. 47.Greenhalgh T, Wherton J, Papoutsi C, Lynch J, Hughes G, A’Court C. Beyond adoption: a new framework for theorizing and evaluation nonadoption, abandonment, and challenges to the scale-up, spread, and sustainability of health and care technologies. J Med Internet Res. 2017;19(11): e367.
48. 48.Integrated Health Information Systems, Singapore. Primary Tech-Enhanced Care (PTEC): telehealth programme enables patients to better manage their health conditions from home. 2022 [cited 2022 Oct 24]. Available from: https://www.ihis.com.sg/Project_Showcase/covid-19/Pages/ptec-telehealth-patients-health-condition-vsm.aspx.
49. Yoon S, Goh H, Chan A, Malhotra R, Visaria A, Matchar D. **Spillover effects of COVID-19 on essential chronic care and ways to foster health system resilience to support vulnerable non-COVID patients: a multistakeholder study**. *JAMDA* (2022.0) **23** 7-14. PMID: 34848198
|
---
title: Synthetic multiantigen MVA vaccine COH04S1 and variant-specific derivatives
protect Syrian hamsters from SARS-CoV-2 Omicron subvariants
authors:
- Felix Wussow
- Mindy Kha
- Taehyun Kim
- Minh Ly
- Marcal Yll-Pico
- Swagata Kar
- Mark G. Lewis
- Flavia Chiuppesi
- Don J. Diamond
journal: NPJ Vaccines
year: 2023
pmcid: PMC10018591
doi: 10.1038/s41541-023-00640-y
license: CC BY 4.0
---
# Synthetic multiantigen MVA vaccine COH04S1 and variant-specific derivatives protect Syrian hamsters from SARS-CoV-2 Omicron subvariants
## Abstract
Emerging SARS-CoV-2 Omicron subvariants continue to disrupt COVID-19 vaccine efficacy through multiple immune mechanisms including neutralizing antibody evasion. We developed COH04S1, a synthetic modified vaccinia Ankara vector that co-expresses Wuhan-Hu-1-based spike and nucleocapsid antigens. COH04S1 demonstrated efficacy against ancestral virus and Beta and Delta variants in animal models and was safe and immunogenic in a Phase 1 clinical trial. Here, we report efficacy of COH04S1 and analogous Omicron BA.1- and Beta-specific vaccines to protect Syrian hamsters from Omicron subvariants. Despite eliciting strain-specific antibody responses, all three vaccines protect hamsters from weight loss, lower respiratory tract infection, and lung pathology following challenge with Omicron BA.1 or BA.2.12.1. While the BA.1-specifc vaccine affords consistently improved efficacy compared to COH04S1 to protect against homologous challenge with BA.1, all three vaccines confer similar protection against heterologous challenge with BA.2.12.1. These results demonstrate efficacy of COH04S1 and variant-specific derivatives to confer cross-protective immunity against SARS-CoV-2 Omicron subvariants.
## Introduction
Despite unprecedented mass vaccination and availability of effective COVID-19 vaccines, SARS-CoV-2 remains a threat to human health due to the continuous emergence of variants of concern (VOC) with ability to evade vaccine-induced immunity. Following emergence of Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), and Delta (B.1.617)1,2, Omicron subvariants have emerged as predominant SARS-CoV-2 VOC, which includes BA.1, BA.2, and BA.2 sub-lineages BA.2.12.1, BA.4/BA.5, and BA.2.753–8, as well as the currently dominating BA.2-derived Omicron subvariant XBB.1.59. Omicron subvariants have exceptional capacity to evade neutralizing antibodies (NAb) due to numerous mutations in the spike (S) protein that dramatically exceed S mutations of other earlier occurring VOC3–9, thereby posing a unique challenge for COVID-19 vaccination. Several studies report reduced clinical effectiveness against Omicron variants by approved COVID-19 vaccines10–13, which were designed to elicit protective immunity based on the S protein of the Wuhan-Hu-1 reference strain14–18. Waning COVID-19 vaccine efficacy against major VOC can be counteracted by repeated booster vaccination10–12, while updated COVID-19 booster vaccines with altered or variant-matched antigen design have been developed to specifically enhance the stimulation of cross-protective immunity against emerging SARS-CoV-2 VOC19,20.
We developed COVID-19 vaccine COH04S1, a multiantigen synthetic modified vaccinia Ankara (sMVA) vector that co-expresses Wuhan-Hu-1-based S and nucleocapsid (N) antigens21,22. COH04S1 afforded protection against SARS-CoV-2 ancestral virus and Beta and Delta variants in Syrian hamster and non-human primate models and was safe and immunogenic in a Phase 1 clinical trial in healthy adults22–24. COH04S1 is currently being tested in multiple Phase 2 clinical trials in healthy volunteers and in cancer patients (NCT4639466, NCT4977024). In this report, we assessed the efficacy of COH04S1 and analogous Omicron BA.1 or Beta-specific vaccines to protect against Omicron subvariants in Syrian hamsters. We found that the three vaccines elicited strain-specific antibody responses, while all three vaccines conferred efficacy to protect hamsters from weight loss, lower respiratory tract infection, and lung pathology following challenge with Omicron BA.1 or BA.2.12.1. In contrast to the Beta-specific vaccine, the BA.1-specific vaccine conferred consistently improved efficacy compared to COH04S1 to protect against homologous challenge with Omicron BA.1. However, neither the Beta- nor the BA.1-specific vaccine afforded significantly improved efficacy compared to COH04S1 to protect against heterologous challenge with BA.2.12.1. These results demonstrate that COH04S1 and analogous variant-specific derivatives confer cross-protective immunity against SARS-CoV-2 Omicron subvariants.
## Vaccine-elicited ancestral- and variant-specific humoral immune responses
Syrian hamsters were vaccinated with COH04S1 or analogous vaccine constructs with Omicron BA.1 or Beta (B.1.351) sequence-modified S and N antigens, termed COH04S529 or COH04S351, respectively (Fig. 1a, b and Supplementary Table 1). Hamsters were vaccinated twice in a four-week interval by intramuscular route. All three sMVA vaccines elicited potent post-prime binding antibody responses to S and N antigens of SARS-CoV-2 ancestral virus, Beta variant, and Omicron variants BA.1, BA.2, BA.2.12.1 and BA$\frac{.4}{5}$ (Fig. 1c, d). COH04S1 and the Beta-specific COH04S351 stimulated significantly higher ancestral-specific S antibody titers than the BA.1-specific COH04S529 vaccine. Consistent with the modified antigen sequences, the Beta-specific COH04S351 vaccine elicited significantly higher Beta-specific S antibody responses than COH04S1 and COH04S529, and the BA.1-specific COH04S529 vaccine elicited significantly higher BA.1-specific S antibody responses than COH04S1 and COH04S351. In addition, COH04S529 vaccination resulted in increased Omicron BA.2-specific S antibody titers compared to COH04S1 vaccination as well as increased BA.2.12.1-specific S antibody titers compared to COH04S1 or COH04S351 vaccination. In contrast, BA.4/BA.5-specific S antibody titers were comparable between the three vaccine groups (Fig. 1c). No significant differences were observed for ancestral- and variant-specific N antibody titers between the three vaccine groups (Fig. 1d).Fig. 1COH04S1 and Omicron BA.1 and Beta sequence-modified vaccines elicit strain-specific antibody responses against SARS-CoV-2 ancestral virus and VOC.a. Vaccine constructs. COH04S1, COH04S529, and COH04S351 are sMVA-vectored COVID-19 vaccines co-expressing S and N antigens based on the Wuhan-Hu-1 reference strain or Omicron BA.1 (Ο) or Beta (β) variants, respectively. The antigen sequences were inserted into the MVA deletion sites 2 (Del2) and 3 (Del3) as indicated. b. Study design. Hamsters were vaccinated twice with COH04S1 ($$n = 20$$), COH04S351 ($$n = 20$$), or COH04S529 ($$n = 20$$) by intramuscular route as indicated (black arrows). Hamsters vaccinated with empty sMVA vector ($$n = 10$$), or unvaccinated hamsters ($$n = 10$$) were used as controls. Blood samples were collected at day 14 and 42 (red arrows) after the first and second vaccination. COH04S1-, COH04S529-, and COH04S351-vaccinated hamsters were challenged intranasally at day 42 with Omicron subvariants BA.1 or BA.2.12.1 ($$n = 10$$/group). sMVA control animals were challenged with BA.1. Unvaccinated controls were challenged with BA.2.12.1. Post-challenge body weight changes were recorded daily for 8 days. Lung tissue and nasal turbinates for viral load measurements and lung histopathology were collected at days 4 and 8 post-challenge ($$n = 5$$/group/timepoint). IgG endpoint titers. S-specific (c) and N-specific (d) binding antibody titers to ancestral virus, Beta and Omicron subvariants BA.1, BA.2, BA.2.12.1, BA.4, and BA.5 were measured in serum samples of vaccine and control groups at day 14 (d14) post-prime vaccination via ELISA. Dotted lines indicate lower limit of detection (LLOD). # indicates significantly higher IgG titers in vaccine groups compared to controls. e. NAb titers. NAb titers were measured in serum samples of vaccine and control groups at day 42 (d42) post-second vaccination via PRNT assay against ancestral SARS-CoV-2 (WA/01), and Omicron BA.1 and BA.2.12.1. Dotted line indicates LLOD. Values below the LLOD are indicated as half the LLOD. Data are presented as box plots extending from 25th to 75th percentiles, with lines indicating medians, and whiskers going from minimum to maximum values. Two-way ANOVA with Tukey’s multiple comparison test was used in c-e following log transformation. * 0.05 < $p \leq 0.01$, **0.01 < $p \leq 0.001$, ***0.001 < $p \leq 0.0001$, ****$p \leq 0.0001.$ When not indicated, differences are not significant ($p \leq 0.05$).
At 2 weeks post-vaccination prior to virus challenge, NAb responses were measured by plaque reduction neutralization titer (PRNT) assay against ancestral SARS-CoV-2 (USA-WA$\frac{1}{2020}$) and Omicron BA.1 or BA.2.12.1. NAb responses against ancestral SARS-CoV-2 were measured in all three vaccine groups, although ancestral-specific NAb titers in COH04S1- and COH04S351-vaccinated hamsters were significantly higher than those in COH04S529-vaccinated animals (Fig. 1e). Consistent with the Omicron immune evasion capacity, BA.1-specific NAb responses in COH04S1-vaccinated animals were either very low or undetectable. In contrast, COH04S529-vaccinated hamsters showed potent Omicron BA.1-specific NAb titers that significantly exceeded BA.1-specific NAb titers measured in COH04S1- and COH04S351-vaccinated animals (Fig. 1e), consistent with the BA.1-specific vaccine modification. BA.1-specific NAb titers measured in COH04S351-vaccinated animals were of intermediate level compared to those measured in COH04S1- and COH04S529-vaccinated hamsters. In contrast to BA.1-specific NAb responses, BA.2.12.1-specific NAb titers measured in all three vaccine groups were either very low or undetectable, indicating limited cross-neutralizing activity against Omicron BA.2.12.1 by the vaccine-elicited responses. Interestingly, BA.2.12.1-specific NAb titers measured in COH04S351-vaccinated animals were slightly elevated compared to those measured in COH04S1- and COH04S529-vaccinated hamsters, although these differences in BA.2.12.1-specific responses between the three vaccine groups were not statistically significant. These results demonstrate that COH04S1 and Omicron BA.1- and Beta-specific vaccine derivatives elicit strain-specific antibody responses against ancestral SARS-CoV-2, Beta variant, and Omicron subvariants.
## Vaccine protection against Omicron BA.1 or BA.2.12.1 virus-induced weight loss
At two-week post vaccination, hamsters were intranasally challenged with Omicron variant BA.1 or BA.2.12.1 and body weight was measured for 8 days. Hamsters vaccinated with sMVA without inserted antigens and unvaccinated hamsters were used as controls. Control animals showed progressive weight loss following challenge with BA.1 or BA.2.12.1, with maximum weight loss between 4-$8\%$ at day 6 post vaccination. In contrast, body weight of hamsters vaccinated with COH04S1 or the variant-specific vaccines remained relatively stable or increased gradually over the 8-day observation period following challenge with either BA.1 or BA.2.12.1 (Fig. 2a, b). Notably, both COH04S529- and COH04S351-vaccinated animals showed consistently higher body weight than COH04S1-vaccinated hamsters following challenge with BA.1, suggesting improved vaccine efficacy through both the BA.1 and Beta-specific antigen modification to protect against BA.1-induced weight loss, although these differences in weight between vaccine groups were only significant at day 3 post-challenge (Fig. 2a). In contrast, similar body weight changes were measured for all three vaccine groups following challenge with BA.2.12.1 (Fig. 2b), indicating similar levels of protection against BA.2.12.1-indcued weight loss by all three vaccines. These results show that COH04S1 and Omicron BA.1 and Beta sequence-modified vaccines protect hamsters from weight loss following challenge with Omicron BA.1 and BA.2.12.1.Fig. 2COH04S1 and Omicron BA.1- and Beta-modified vaccines protect hamsters from weight loss following challenge with Omicron BA.1 or BA.2.12.1.Body weight of COH04S1, COH04S529-, COH04S351-vaccinated animals was measured daily for 8 days following challenge with Omicron variants BA.1 (a) and BA.2.12.1 (b). Hamsters vaccinated with empty sMVA vector or unvaccinated hamsters were used as controls. Weight loss compared to day 0 is reported as mean ± SEM. Two-way ANOVA followed by Tukey’s multiple comparison test was used to compare group mean values at each timepoint following log transformation. Color-coded asterisks indicate significant difference compared to controls unless specified. * 0.05 < $p \leq 0.01$, **0.01 < $p \leq 0.001$, ***0.001 < $p \leq 0.0001$, ****$p \leq 0.0001.$ When not indicated, differences are not significant ($p \leq 0.05$). Black asterisks indicate time of sacrifice ($$n = 5$$/group).
## Vaccine protection against Omicron BA.1 or BA.2.12.1 respiratory tract infection
At days 4 and 8 post challenge, viral loads were measured in lung tissue and nasal turbinates by quantification of SARS-CoV-2 genomic RNA (gRNA) and sub-genomic RNA (sgRNA) to assess the magnitude of total and replicating virus at lower and upper respiratory tracts. Compared to the high lung viral loads measured in control animals, significantly reduced gRNA and sgRNA levels were measured in the lungs of all three vaccine groups at day 4 and 8 following virus challenge with BA.1 or BA.2.12.1 (Fig. 3a–d), demonstrating potent efficacy of all three vaccines to control lower respiratory tract infection by BA.1 or BA.2.12.1. In addition, in contrast to all or most control animals, all animals of the vaccine groups had undetectable sgRNA in the lungs at day 8 following BA.1 or BA.2.12.1 virus challenge (Fig. 3c, d), indicating complete control of lower respiratory tract infection. Notably, unlike COH04S351-vaccinated animals, COH04S529-vaccinated animals had significantly reduced lung viral loads compared to COH04S1-vaccinated animals at day 4 following challenge with BA.1 (Fig. 3a, c), consistent with improved immediate viral control of BA.1 lower respiratory tract infection through the BA.1-specific vaccine adaptation. Moreover, in contrast to COH04S1- and COH04S351-vaccinated animals, all COH04S529-vaccinated animals had undetectable sgRNA at day 4 following BA.1 challenge (Fig. 3c). No significant differences in lung viral loads were observed between the three vaccine groups at day 8 following challenge with BA.1 (Fig. 3a, c). In addition, no significant differences in lung viral loads were observed between the three vaccine groups at day 4 or 8 following BA.2.12.1 virus challenge (Fig. 3b, d) indicating comparable vaccine efficacy to protect against lower respiratory infection caused by BA.2.12.1. Compared to the potent vaccine efficacy to protect against Omicron BA.1 or BA.2.12.1 infection in the lower respiratory tract, the vaccine efficacy to control infection of the Omicron variants in the upper respiratory tract was less evident (Fig. 4a–d). While viral loads measured in nasal turbinates of all three vaccine groups following virus challenge with BA.1 were consistently lower than those of controls (Fig. 4a, c), these differences in nasal viral loads were not statistically significant. Moreover, no evident differences in nasal viral loads were observed between the vaccine groups and controls following challenge with BA.2.12.1 (Fig. 4b, d). The precise reason for the low vaccine efficacy to prevent upper respiratory tract infection is unclear, although this may suggest limited protective mucosal immunity at the respiratory epithelium. Other explanations such as an excessive viral challenge dose, as suggested by the relatively high viral loads in both the vaccine and control groups, may also account for this observation. These results demonstrate efficacy of COH04S1 and Omicron BA.1- and Beta sequence-modified vaccines to protect hamsters against lower respiratory tract infection by Omicron BA.1 and BA.2.12.1.Fig. 3COH04S1 and Omicron BA.1- and Beta-modified vaccines protect hamsters from lower respiratory tract infection following virus challenge with Omicron BA.1 or BA.2.12.1.SARS-CoV-2 genomic RNA (gRNA, a, b) and sub-genomic RNA (sgRNA, c, d) copies were quantified by qPCR in lung tissue of COH04S1, COH04S351, and COH04S529 vaccine and control groups at days 4 and 8 following challenge with Omicron subvariants BA.1 (a, c) and BA.2.12.1 (b, d). Lines indicate median RNA copies. Two-way ANOVA followed by Tukey’s multiple comparison test was used following log transformation. * 0.05 < $p \leq 0.01$, **0.01 < $p \leq 0.001$, ***0.001 < $p \leq 0.0001$, ****$p \leq 0.0001.$ When not indicated, differences are not significant ($p \leq 0.05$).Fig. 4Viral loads at upper respiratory tracts in COH04S1-, COH04S529-, and COH04S351-vaccinated hamsters following challenge with Omicron BA.1 or BA.2.12.1.SARS-CoV-2 genomic RNA (gRNA, a, b) and sub-genomic RNA (sgRNA, c, d) copies were quantified by qPCR in nasal turbinates of COH04S1, COH04S529, and COH04S351 vaccine and control groups at days 4 and 8 following challenge with Omicron subvariants BA.1 (a, c) and BA.2.12.1 (b, d). Lines indicate median RNA copies. One-way ANOVA followed by Tukey’s multiple comparison test was used following log transformation. When not indicated, differences are not significant ($p \leq 0.05$).
## Vaccine protection against Omicron BA.1 or BA.2.12.1 virus-induced lung pathology
Lung pathology in all three vaccine groups following virus challenge with BA.1 or BA.2.12.1 was significantly reduced when compared to controls (Fig. 5a–f, Supplementary Fig. 1), indicating efficacy of all three vaccines to protect hamsters from BA.1- and BA.2.12.1-induced lung injury. In addition, a subset of control hamsters had moderate bronchioalveolar hyperplasia (i.e., type II pneumocyte hyperplasia) at day 4 following BA.1 virus challenge, and all control animals had moderate to high-grade bronchioalveolar hyperplasia at day 8 following viral challenge with BA.1 or BA.2.12.1 (Fig. 5b, e). In contrast, bronchioalveolar hyperplasia was undetectable in animals of all three vaccine groups at day 4 and 8 following viral challenge with BA.1 or BA.2.12.1, with only one or two exceptions in the COH04S1 and COH04S529 vaccine groups that showed low-grade bronchioalveolar hyperplasia (Fig. 5b, e). Lung pathology in all three vaccine groups appeared mostly associated with inflammation, which was detectable at low levels in all vaccine groups at day 4 and 8 following virus challenge with BA.1 or BA.2.12.1, albeit at significantly reduced levels across all vaccine groups compared to controls (Fig. 5c, f), confirming efficacy of all three vaccines to protect against BA.1 or BA.2.12.1-induced lung pathology. Notably, in contrast to COH04S351-vaccinated animals, COH04S529-vaccinated animals had significantly reduced lung pathology and inflammation compared to COH04S1-vaccinated animals at day 4 following challenge with BA.1 (Fig. 5a, c), indicating improved protection by the BA.1-specific vaccine to control BA.1-induced lung pathology during early phase after viral challenge. No significant differences in lung pathology were observed between the three vaccine groups at day 8 following BA.1 virus challenge (Fig. 5a–c). Furthermore, no significant differences in lung pathology, hyperplasia, or inflammation were observed between the vaccine groups at day 4 or 8 following virus challenge with BA.2.12.1 (Fig. 5d–f), indicating comparable efficacy of all three vaccines to protect against BA.2.12.1-induced lung pathology. These results demonstrate that COH04S1 and Omicron BA.1 and Beta sequence-modified vaccine derivatives protect hamsters from lung pathology following virus challenge with Omicron subvariants BA.1 and BA.2.12.1.Fig. 5COH04S1 and Omicron BA.1- and Beta-modified vaccines protect hamsters from lung pathology following challenge with Omicron BA.1 or BA.2.12.1.Hematoxylin/eosin-stained lung sections of COH04S1-, COH04S529, and COH04S351-vaccinated hamsters and control animals at days 4 and 8 following challenge with SARS-CoV-2 BA.1 (a–c) or BA.2.12.1 (d–f) variants were evaluated by a board-certified pathologist and microscopic findings were graded based on severity on a scale from 1 to 5 (Supplementary Table 3). Cumulative pathology score of all histopathologic findings (a, d), grading of bronchioalveolar hyperplasia disease severity (b, e), and severity of lung inflammatory microscopic findings (c, f) are shown. Lines indicate median values. Two-way ANOVA followed by Tukey’s multiple comparison test was used following log transformation. * =0.05 < $p \leq 0.01$, **=0.01 < $p \leq 0.001$, ***=0.001 < $p \leq 0.0001$, ****=$p \leq 0.0001.$ When not indicated, differences are not significant ($p \leq 0.05$).
## Discussion
This report demonstrates that sMVA-based COVID-19 vaccine COH04S1 co-expressing Wuhan-Hu-1 S and N antigens and analogous Omicron BA.1 (COH04S529) and Beta (COH04S351) sequence-modified vaccines protect Syrian hamsters from SARS-CoV-2 Omicron subvariants BA.1 and BA.2.12.1. These results complement previous observations in Syrian hamster and non-human primate models demonstrating potent efficacy of COH04S1 and COH04S351 to protect against SARS-CoV-2 ancestral virus and Beta and Delta VOC22,23. These animal studies provide evidence that COH04S1 and variant-specific derivatives have potent capacity to confer cross-protective immunity against emerging SARS-CoV-2 VOC, including Omicron subvariants.
While repeated booster vaccination can counteract waning efficacy observed with first-generation Wuhan-Hu-1-based COVID-19 vaccines10–12, variant-specific vaccines have been developed to enhance vaccine efficacy against Omicron subvariants. Two Omicron BA.4/BA.5-specific mRNA booster vaccines have been recently authorized by the FDA20,25,26, while the EMA has authorized mRNA boosters with BA.1- and BA.4/BA.5-specific modifications. Studies in hamsters and non-human primates provide evidence that Wuhan-Hu-1-based COVID-19 vaccines can protect against Omicron B.1.1.529 and its BA.1 subvariant27–30. Yet, the impact of Omicron-specific sequence alterations to enhance COVID-19 vaccine efficacy is debatable25,26,30–32. Moreover, the advantage of the bivalent mRNA booster vaccine containing BA.4/BA.5- and Wuhan-Hu-1-based S antigens over a booster by the original Wuhan-Hu-1-based mRNA vaccine to stimulate cross-neutralizing responses against Omicron subvariants remains controversial20,33. Our observations in Syrian hamsters demonstrating efficacy of COH04S1 to protect against Omicron BA.1 and BA.2.12.1 supports that Wuhan-Hu-1-based COVID-19 vaccines have the capacity to confer potent cross-protective immunity against Omicron variants. Unlike the Beta-specific COH04S351 vaccine, the BA.1-specific COH04S529 conferred consistently improved efficacy compared to COH04S1 to protect against weight loss, lower respiratory tract infection, and lung pathology during early phase (at day 4) following homologous challenge with BA.1, while all three vaccines conferred comparable levels of protection against heterologous challenge with BA.2.12.1. These results suggest that Omicron-specific antigen modification can improve the efficacy against homologous Omicron strains, while it does not appear to be an effective strategy to enhance protection against heterologous Omicron strains. Updated Omicron-specific booster vaccines may not provide a significant advantage over first-generation Wuhan-Hu-1-based COVID-19 vaccines to confer long-term protection against evolving Omicron subvariants. However, heterologous-prime boost immunization strategies were not investigated, which could potentially be a more effective approach to further enhance cross-protective immunity conferred by the vaccines.
Whether the observed efficacy of COH04S1 or the variant-adapted vaccines to protect against Omicron BA.1 and BA.2.12.1 was mediated solely by S-specific responses or by a combination of S and N-specific responses remains unclear. The results that COH04S529 elicited more potent BA.1-specific antibody responses and provided greater protection against BA.1 virus challenge than COH04S1 is consistent with an important protective role of humoral immunity. Interestingly, despite the stimulation of low-to-undetectable BA.1-specific NAb responses by COH04S1 and low-to-undetectable BA.2.12.1-specific NAb responses by all three vaccines, all three vaccines provided potent protection against BA.1 and BA.2.12.1 virus challenge. While these findings may indicate that only low NAb titers are required to protect against Omicron variants in hamsters, these findings may also indicate that other responses besides NAb contribute to the efficacy of COH04S1 and its sequence-modified vaccine derivatives. This may include humoral responses promoting Fc-mediated effector functions as well as T cell responses to both S and N antigens. A recent study with Wuhan-Hu-1-based mRNA vaccines in hamsters shows that an antigen combination composed of S and N confers superior efficacy compared to S alone to protect against Omicron BA.134, supporting the use of a dual antigen design to confer protective immunity against Omicron variants. Other studies support potential benefits of N as a vaccine antigen35–37. Whether the efficacy observed with COH04S1 or the variant-specific vaccines against BA.1 and BA.2.12.1 extends to other Omicron subvariants remains to be addressed. The antigenic similarity of the S and N proteins between Omicron variants evolved from BA.2 suggests that the vaccines may have the capacity to similarly protect against more recently emerged Omicron variants.
The N antigen was included in COH04S1 primarily based on the rationale to broaden the induction of T cells, which are known to be less susceptible to antigen variation than NAb and therefore considered a critical second line of defense to provide long-term protective immunity against SARS-CoV-238–44. Our previous studies in mice and non-human primates and our Phase 1 clinical trial in healthy adults demonstrates potent capacity of COH04S1 to stimulate S and N-specific T cells21,22,24,38. Importantly, T cell responses to both the S and N antigens elicited in COH04S1-vaccinated healthy adults maintain potent cross-reactivity to Delta and Omicron BA.1 variants for up to six months post-vaccination38, whereas COH04S1-elicited NAb response, as shown for other COVID-19 vaccines45, decrease and confer reduced neutralizing activity against Delta and Omicron BA.1. While we did not assess T cell responses in the current study due to the limitations to evaluate cellular responses in the hamster model, these prior studies in animals and healthy adults suggest that T cells responses may have contributed to the efficacy of COH04S1 and its allied forms to protect against Omicron BA.1 and BA.2.12.1 in hamsters. T cell responses may become critically important for vaccine protection against SARS-CoV-2 when NAb responses are suboptimal, as suggested by our observations in this study. Furthermore, because of the higher conservation of the N protein compared to the S protein, N-specific T cells may have greater potential than S-specific T cells to confer cross-reactive immunity against SARS-CoV-2 VOC.
Additional studies are needed to identify immune correlates of protection that are associated with COH04S1 vaccine protection and to specifically address the contribution of vaccine-elicited S and N-specific responses to confer cross-protective immunity. Nonetheless, our current data in sum combined with our previous results in animal models demonstrates potent efficacy of COH04S1 and analogous variant-adapted vaccines to protect against multiple SARS-CoV-2 VOC, including Omicron subvariants. Importantly, COH04S1 is the most advanced clinically evaluated MVA-vectored COVID-19 vaccine and the only COVID-19 vaccine combining S and N that is actively investigated in Phase 2 clinical trials. COH04S1 may therefore be a valuable alternative to the currently available single-antigen FDA-approved COVID-19 vaccines based on adenoviral vectors or mRNA to improve cross-protective immunity against emerging SARS-CoV-2 VOC.
## Vaccine vectors
COH04S1 is a double-plaque purified virus isolate derived from the previously described sMVA-N/S vector (NCBI Accession# MW036243) with N and S antigen sequences inserted into the MVA deletion sites 2 (Del2) and 3 (Del3), respectively21,22,46. COH04S1 co-expresses full-length, unmodified S and N antigen sequences based on the Wuhan-Hu-1 reference strain (NCBI Accession# NC_045512). COH04S351 is a double-plaque purified virus isolate with analogous vaccine construction compared to the original COH04S1 vector and co-expresses modified S and N antigen sequences based on the B.1.351 Beta variant (Supplementary Table 1)2,47. COH04S529 is a non-plaque purified virus isolate with analogous vaccine construction compared to the original COH04S1 sMVA-N/S vaccine vector and co-expresses modified S and N antigen sequences based on the Omicron BA.1 variant (Supplementary Table 1). COH04S1, COH04S351, and COH04S529 were generated using the sMVA platform consisting of three bacterial artificial chromosome-cloned synthetic DNA fragments (sMVA fragments F1-F3) covering the MVA genome sequence published by Antoine and colleagues21,46. The antigen sequences were inserted into the sMVA fragments F1 and F3 by En passant mutagenesis in GS1783 E. coli cells21,48. The modified sMVA fragments F1 and F3 with inserted antigen sequences in Del2 or Del3, respectively, were co-transfected together with the unmodified sMVA fragment F2 into baby hamster kidney cells to initiate the reconstitution of recombinant sMVA virus in the presence of fowl pox virus as a helper virus21. Virus stocks of the vaccine vectors and sMVA control vector were produced using chicken embryo fibroblasts (CEF) and prepared by $36\%$ sucrose cushion ultracentrifugation and virus resuspension in 1 mM Tris-HCl (pH 9). Virus stocks were stored at −80 °C and titrated on CEF by immunostaining of viral plaques at 16–24 h post infection using polyclonal vaccinia antibody (9503-2057, Bio-Rad, dilution 1:2000)21. Viral stocks were validated for antigen insertion and expression by PCR, sequencing, and Immunoblot.
## Hamster study
In life portion of the hamster studies were carried out at Bioqual Inc. (Rockville, MD). Eighty 6–8 weeks old Syrian hamsters were randomly assigned to the groups, with 5 females and 5 males in each challenge group. Hamsters were intramuscularly vaccinated four weeks apart with 1 × 108 plaque forming units (PFU) of COH04S1, COH04S529, COH04S351 or sMVA virus stocks diluted in phosphate-buffered saline (PBS) or left unvaccinated. Blood samples were collected 2 weeks after the first vaccine dose, and two weeks after the second dose. At this latter time point animals were challenged intranasally (50 µl/nare) with 4.8 × 104 Median Tissue Culture Infectious Dose [TCID50] of SARS-CoV-2 BA.1 (BEI resources NR-56486 LOT#: 70049695, titered by BEI resources using Calu-3 cells) or with 5.16 × 104 TCID50 of SARS-CoV-2 BA.2.12.1 (BEI resources NR-56782 LOT#: 70052277, titered using Calu-3 cells). Body weight was recorded daily for 8 days. Hamsters were humanely euthanized for lung and turbinate samples collection at day 4 ($$n = 5$$/group) and day 8 ($$n = 5$$/group) post-challenge. All animal studies were conducted in compliance with local, state, and federal regulations and were approved by Bioqual (protocol 20-163) and City of Hope (protocol 20087) Institutional Animal Care and Use Committees.
## Binding antibody detection
SARS-CoV-2-specific binding antibodies in hamster serum samples were detected by indirect ELISA using purified ancestral-specific, Beta-specific, Omicron BA.1-, BA.2-, BA.2.12.1-, and BA.4-specific S proteins (Sino Biological 40589-V08B1, 40588-V07E9, 40589-V08H26, 40589-V08H28, 40589-V08H34, 40589-V08H32), and purified ancestral-specific, Beta-specific, Omicron BA.1-, BA.2-, and BA.4-specific N proteins (Sino Biological 40588-V08B, 40589-V08B7, 40588-V07E34, 40588-V07E35, 40588-V07E37). S and N mutations included in the antigens used for ELISA are indicated in Supplementary Table 2. 96-well plates were coated with 100 μl/well of protein at a concentration of 1 µg/ml in PBS and incubated overnight at 4oC. Plates were washed 5× with wash buffer ($0.1\%$ Tween-20/PBS), then blocked with 250 µl/well of blocking buffer ($0.5\%$ casein/154 mM NaCl/10 mM Tris-HCl [pH 7.6]) for 2 h at room temperature. After washing, threefold diluted heat-inactivated serum in blocking buffer was added to the plates and incubated 2 h at room temperature. After washing, anti-Hamster IgG(H + L) HRP secondary antibody (Southern Biotech 6061-05) was diluted 1:1000 in blocking buffer and added to the plates. After 1 h incubation, plates were washed and developed with 1 Step TMB-Ultra (Thermo Fisher 34029). The reaction was stopped with 1 M H2SO4 and plates were read on FilterMax F3 (Molecular Devices). Endpoint titers were calculated as the highest dilution to have an absorbance >0.100 nm.
## Neutralization assay
NAb were measured by PRNT using ancestral SARS-CoV-2 (USA-WA$\frac{1}{2020}$ strain), or BA.1, or BA.2.12.1 variants. Ancestral SARS-CoV-2 stock was generated using Vero-E6 cells infected with seed stock virus obtained from Kenneth Plante at UTMB (lot # TVP 23156). BA.1 stock (BIOQUAL Lot # 122121-700) was originally received from Emory (B.1.1.529 PP3P1 hCoV19/EHC_C19_2811C $\frac{12}{9}$/2021) and expanded in Calu-3 cells. BA.2.12.1 stock was obtained from BEI resources (NR-56782) and produced on Calu-3 cells. Vero E6 cells (ATCC, CRL-1586) were seeded in 24-well plates at 175,000 cells/well in DMEM/$10\%$ FBS/Gentamicin. Serial threefold serum dilutions were incubated in 96-well plates with 30 PFU of SARS-CoV-2 for 1 h at 37 °C. The serum/virus mixture was transferred to Vero-E6 cells and incubated for 1 h at 37 °C. After that, 1 ml of $0.5\%$ methylcellulose media was added to each well and plates were incubated at 37 °C for three days. Plates were washed, and cells were fixed with methanol. Crystal violet staining was performed, and plaques were recorded. IC50 titers were calculated as the serum dilution that gave a $50\%$ reduction in viral plaques in comparison to control wells.
## Genomic RNA quantification
SARS-CoV-2 gRNA copies per gram of tissue were quantified by qRT-PCR assay using primer/probe sequences binding to a conserved region of SARS-CoV-2 N gene. Viral RNA was extracted with RNA-STAT 60 (Tel-test B)/chloroform, precipitated and resuspended in RNAse-free water. SensiFAST Probe Lo‑ROX One‑Step Kit (Bioline BIO-78005) was used following manufacturer instructions. All samples were tested in triplicate. The control RNA was prepared to contain 106 to 107 copies per 3 µl. Eight 10-fold serial dilutions of control RNA were prepared using RNAse-free water. Duplicate samples of each dilution were prepared. Amplification was carried out using an Applied Biosystems 7500 Sequence detector. Amplification conditions were 48 °C for 30 min, 95 °C for 10 min followed by 40 cycles of 95 °C for 15 seconds, and 1 min at 55 °C. Copies of RNA/ml were calculated by extrapolation from the standard curve. Primer/probe sequences: 5′-GAC CCC AAA ATC AGC GAA AT-3′; 5′-TCT GGT TAC TGC CAG TTG AAT CTG-3′; and 5′-FAM-ACC CCG CAT TAC GTT TGG TGG ACC-BHQ1-3′.
## Subgenomic RNA quantification
SARS-CoV-2 sgRNA copies were assessed through quantification of N gene mRNA by qRT-PCR22,23. Briefly, SARS-CoV-2 RNA was extracted from tissues using TRIzol, precipitated and resuspended in RNAse-free water. For quantification, the standard curve of a plasmid containing a cDNA copy of the N gene mRNA target was used. Applied Biosystems 7500 Real-Time PCR System was used for amplification with the following program: 48 °C for 30 min, 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s, and 1 min at 55 °C. The number of copies of RNA per ml was calculated by extrapolation from the standard curve and multiplying by the reciprocal of 0.2 ml extraction volume. Primer/probe sequences: 5′-CGA TCT CTT GTA GAT CTG TTC TC-3′; 5′-GGT GAA CCA AGA CGC AGT AT-3′; 5′-FAM- TAA CCA GAA TGG AGA ACG CAG TGG G-BHQ-3′.
## Histopathology
Histopathological evaluation of hamster lung sections was performed by Experimental Pathology Laboratories, Inc. (Sterling, VA). At necropsy organs were collected and placed in $10\%$ neutral buffered formalin for histopathologic analysis. Tissues were processed through to paraffin blocks, sectioned once at approximately 5 microns thickness, and stained with hematoxylin/eosin. Board certified pathologists were blinded to the vaccine groups and controls were used as a comparator. Histopathological findings were assigned a severity score between 1 (minimal) and 5 (severe) (Supplementary Table 3).
## Statistical analysis
Statistical analyses were performed using Prism 8 (GraphPad, v8.3.0). One-way ANOVA and two-way ANOVA with Tukey’s multiple comparison test were used for statistical evaluation after logarithmic transformation. The significance level for each test is indicated in the figure legends.
## Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
## Supplementary information
Supplemental Information REPORTING SUMMARY The online version contains supplementary material available at 10.1038/s41541-023-00640-y.
## References
1. 1.Flores-Vega, V. R. et al. SARS-CoV-2: Evolution and emergence of new viral variants. Viruses14, 10.3390/v14040653 (2022).
2. Garcia-Beltran WF. **Multiple SARS-CoV-2 variants escape neutralization by vaccine-induced humoral immunity**. *Cell* (2021.0) **184** 2523. DOI: 10.1016/j.cell.2021.04.006
3. 3.Carreno, J. M. et al. Activity of convalescent and vaccine serum against SARS-CoV-2 Omicron. Nature, 10.1038/s41586-022-04399-5 (2021).
4. 4.Tegally, H. et al. Emergence of SARS-CoV-2 Omicron lineages BA.4 and BA.5 in South Africa. Nat. Med.10.1038/s41591-022-01911-2 (2022).
5. Wang Q. **Antibody evasion by SARS-CoV-2 Omicron subvariants BA.2.12.1, BA.4 and BA.5**. *Nature* (2022.0) **608** 603-608. DOI: 10.1038/s41586-022-05053-w
6. Iketani S. **Antibody evasion properties of SARS-CoV-2 Omicron sublineages**. *Nature* (2022.0) **604** 553-556. DOI: 10.1038/s41586-022-04594-4
7. Yu J. **Neutralization of the SARS-CoV-2 Omicron BA.1 and BA.2 Variants**. *N. Engl. J. Med.* (2022.0) **386** 1579-1580. DOI: 10.1056/NEJMc2201849
8. Hachmann NP. **Neutralization escape by SARS-CoV-2 Omicron subvariants BA.2.12.1, BA.4, and BA.5**. *N. Engl. J. Med.* (2022.0) **387** 86-88. DOI: 10.1056/NEJMc2206576
9. 9.Miller, J. et al. Substantial neutralization escape by SARS-CoV-2 Omicron variants BQ.1.1 and XBB.1. N. Engl. J. Med.10.1056/NEJMc2214314 (2023).
10. 10.Andrews, N. et al. Covid-19 vaccine effectiveness against the Omicron (B.1.1.529) variant. N. Engl. J. Med.10.1056/NEJMoa2119451 (2022).
11. 11.Regev-Yochay, G. et al. Efficacy of a fourth dose of Covid-19 mRNA vaccine against Omicron. N. Engl. J. Med.10.1056/NEJMc2202542 (2022).
12. Abu-Raddad LJ. **Effect of mRNA vaccine boosters against SARS-CoV-2 Omicron infection in Qatar**. *N. Engl. J. Med* (2022.0) **386** 1804-1816. DOI: 10.1056/NEJMoa2200797
13. Zhu N. **A novel coronavirus from patients with pneumonia in China, 2019**. *N. Engl. J. Med.* (2020.0) **382** 727-733. DOI: 10.1056/NEJMoa2001017
14. Polack FP. **Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine**. *N. Engl. J. Med.* (2020.0) **383** 2603-2615. DOI: 10.1056/NEJMoa2034577
15. Baden LR. **Efficacy and safety of the mRNA-1273 SARS-CoV-2 vaccine**. *N. Engl. J. Med.* (2021.0) **384** 403-416. DOI: 10.1056/NEJMoa2035389
16. Voysey M. **Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK**. *Lancet* (2021.0) **397** 99-111. DOI: 10.1016/S0140-6736(20)32661-1
17. 17.Sadoff, J. et al. Safety and efficacy of single-dose Ad26.COV2.S vaccine against Covid-19. N. Engl. J. Med.10.1056/NEJMoa2101544 (2021).
18. Zhou P. **A pneumonia outbreak associated with a new coronavirus of probable bat origin**. *Nature* (2020.0) **579** 270-273. DOI: 10.1038/s41586-020-2012-7
19. Choi A. **Safety and immunogenicity of SARS-CoV-2 variant mRNA vaccine boosters in healthy adults: an interim analysis**. *Nat. Med.* (2021.0) **27** 2025-2031. DOI: 10.1038/s41591-021-01527-y
20. 20.Chalkias, S. et al. A bivalent Omicron-containing booster vaccine against Covid-19. N. Engl. J. Med. 10.1056/NEJMoa2208343 (2022).
21. Chiuppesi F. **Development of a multi-antigenic SARS-CoV-2 vaccine candidate using a synthetic poxvirus platform**. *Nat. Commun.* (2020.0) **11** 6121. DOI: 10.1038/s41467-020-19819-1
22. Chiuppesi F. **Synthetic multiantigen MVA vaccine COH04S1 protects against SARS-CoV-2 in Syrian hamsters and non-human primates**. *NPJ Vaccines* (2022.0) **7** 7. DOI: 10.1038/s41541-022-00436-6
23. 23.Wussow, F. et al. COH04S1 and beta sequence modified vaccine protect hamsters from SARS-CoV-2 variants. iScience, 104457 (2022).
24. 24.Chiuppesi, F. et al. Safety and immunogenicity of a synthetic multiantigen modified vaccinia virus Ankara-based COVID-19 vaccine (COH04S1): an open-label and randomised, phase 1 trial. Lancet Microbe, 10.1016/S2666-5247(22)00027-1 (2022).
25. Callaway E. **New Omicron-specific vaccines offer similar protection to existing boosters**. *Nature* (2022.0) **609** 232-233. DOI: 10.1038/d41586-022-02806-5
26. Vogel G. **Omicron shots are coming-with lots of questions**. *Science* (2022.0) **377** 1029-1030. DOI: 10.1126/science.ade6580
27. Arunachalam PS. **Durable protection against the SARS-CoV-2 Omicron variant is induced by an adjuvanted subunit vaccine**. *Sci. Transl. Med.* (2022.0) **14** eabq4130. DOI: 10.1126/scitranslmed.abq4130
28. Chandrashekar A. **Vaccine protection against the SARS-CoV-2 Omicron variant in macaques**. *Cell* (2022.0) **185** 1549-1555.e1511. DOI: 10.1016/j.cell.2022.03.024
29. Halfmann PJ. **Efficacy of vaccination and previous infection against the Omicron BA.1 variant in Syrian hamsters**. *Cell Rep.* (2022.0) **39** 110688. DOI: 10.1016/j.celrep.2022.110688
30. van Doremalen N. **ChAdOx1 nCoV-19 (AZD1222) or nCoV-19-Beta (AZD2816) protect Syrian hamsters against Beta Delta and Omicron variants**. *Nat. Commun.* (2022.0) **13** 4610. DOI: 10.1038/s41467-022-32248-6
31. Hawman DW. **Replicating RNA platform enables rapid response to the SARS-CoV-2 Omicron variant and elicits enhanced protection in naive hamsters compared to ancestral vaccine**. *EBioMedicine* (2022.0) **83** 104196. DOI: 10.1016/j.ebiom.2022.104196
32. Gagne M. **mRNA-1273 or mRNA-Omicron boost in vaccinated macaques elicits similar B cell expansion, neutralizing responses, and protection from Omicron**. *Cell* (2022.0) **185** 1556-1571.e1518. DOI: 10.1016/j.cell.2022.03.038
33. 33.Wang, Q. et al. Antibody response to Omicron BA.4-BA.5 bivalent booster. N. Engl. J. Med.10.1056/NEJMc2213907 (2023).
34. Hajnik RL. **Dual spike and nucleocapsid mRNA vaccination confer protection against SARS-CoV-2 Omicron and Delta variants in preclinical models**. *Sci. Transl. Med.* (2022.0) **14** eabq1945. DOI: 10.1126/scitranslmed.abq1945
35. 35.Dangi, T., Class, J., Palacio, N., Richner, J. M. & Penaloza MacMaster, P. Combining spike- and nucleocapsid-based vaccines improves distal control of SARS-CoV-2. Cell Rep. 109664 (2021).
36. 36.Matchett, W. E. et al. Nucleocapsid vaccine elicits spike-independent SARS-CoV-2 protective immunity. J. Immunol.10.4049/jimmunol.2100421 (2021).
37. 37.Hong, S. H. et al. Immunization with RBD-P2 and N protects against SARS-CoV-2 in nonhuman primates. Sci. Adv.7, 10.1126/sciadv.abg7156 (2021).
38. Chiuppesi F. **Vaccine-induced spike- and nucleocapsid-specific cellular responses maintain potent cross-reactivity to SARS-CoV-2 Delta and Omicron variants**. *iScience* (2022.0) **25** 104745. DOI: 10.1016/j.isci.2022.104745
39. Keeton R. **T cell responses to SARS-CoV-2 spike cross-recognize Omicron**. *Nature* (2022.0) **603** 488-492. DOI: 10.1038/s41586-022-04460-3
40. Moss P. **The T cell immune response against SARS-CoV-2**. *Nat. Immunol.* (2022.0) **23** 186-193. DOI: 10.1038/s41590-021-01122-w
41. Tarke A. **SARS-CoV-2 vaccination induces immunological T cell memory able to cross-recognize variants from Alpha to Omicron**. *Cell* (2022.0) **185** 847-859.e811. DOI: 10.1016/j.cell.2022.01.015
42. Gao Y. **Ancestral SARS-CoV-2-specific T cells cross-recognize the Omicron variant**. *Nat. Med.* (2022.0) **28** 472-476. DOI: 10.1038/s41591-022-01700-x
43. Grifoni A. **Targets of T cell responses to SARS-CoV-2 coronavirus in humans with COVID-19 disease and unexposed individuals**. *Cell* (2020.0) **181** 1489-1501.e1415. DOI: 10.1016/j.cell.2020.05.015
44. 44.Peng, Y. et al. Broad and strong memory CD4(+) and CD8(+) T cells induced by SARS-CoV-2 in UK convalescent individuals following COVID-19. Nat. Immunol., 10.1038/s41590-020-0782-6 (2020).
45. 45.Geers, D. et al. SARS-CoV-2 variants of concern partially escape humoral but not T-cell responses in COVID-19 convalescent donors and vaccinees. Sci. Immunol.6, 10.1126/sciimmunol.abj1750 (2021).
46. Antoine G, Scheiflinger F, Dorner F, Falkner FG. **The complete genomic sequence of the modified vaccinia Ankara strain: comparison with other orthopoxviruses**. *Virology* (1998.0) **244** 365-396. DOI: 10.1006/viro.1998.9123
47. Zhou D. **Evidence of escape of SARS-CoV-2 variant B.1.351 from natural and vaccine-induced sera**. *Cell* (2021.0) **184** 2348-2361.e2346. DOI: 10.1016/j.cell.2021.02.037
48. Tischer BK, von Einem J, Kaufer B, Osterrieder N. **Two-step red-mediated recombination for versatile high-efficiency markerless DNA manipulation in Escherichia coli**. *Biotechniques* (2006.0) **40** 191-197. DOI: 10.2144/000112096
|
---
title: 'COVID-19 vaccine communication and advocacy strategy: a social marketing campaign
for increasing COVID-19 vaccine uptake in South Korea'
authors:
- Shin-Ae Hong
journal: Humanities & Social Sciences Communications
year: 2023
pmcid: PMC10018596
doi: 10.1057/s41599-023-01593-2
license: CC BY 4.0
---
# COVID-19 vaccine communication and advocacy strategy: a social marketing campaign for increasing COVID-19 vaccine uptake in South Korea
## Abstract
Research evidence suggests that communication is a powerful tool for influencing public opinion and attitudes toward various health-related issues, such as vaccine reluctance, provided it is well-designed and thoughtfully conducted. In particular, social marketing techniques that alter the target audience’s behaviors for the public good can substantially improve vaccine uptake if adopted as a communication strategy in immunization programs to counter public hesitancy. This study presents evidence from the Korean government’s current coronavirus disease 2019 (COVID-19) vaccination campaign, which successfully applied a social marketing approach. By the end of August 2022, South Korea had achieved high vaccine coverage, with $94.8\%$ of the population (12+) receiving a second dose, $71.3\%$ a third dose, and a fourth dose drive currently underway. There are five crucial factors to consider when preparing official communication for an immunization program: (i) a high degree of proactiveness, (ii) credibility, (iii) fighting misinformation, (iv) emphasizing social norms and prosocial behavior, and (v) coherence. Although using social marketing strategies may not be successful in all circumstances, the lessons learned and current implementation in Korea suggest their efficacy in fostering vaccine acceptance. This study offers valuable insights for government agencies and global public health practitioners to develop effective targeted campaign strategies that enhance the target population’s vaccination intention.
## Introduction
The unprecedented spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has resulted in the century’s first major public health and economic crisis.
Two years after the pandemic, the world experienced several deadly waves of the coronavirus disease-2019 (COVID-19). Governments worldwide have hurried to implement mass vaccine rollout programs, but public acceptance of the vaccine is crucial for long-term COVID-19 control and prevention. Vaccination is generally regarded as a critical and cost-effective protective intervention for public health (Polack et al., 2020; Anderson et al., 2020; WHO, 2021a). While vaccines are the epicenter of the global response to the pandemic, public health officials worldwide are encountering growing vaccine hesitancy, significantly impeding their efforts for reaching herd immunity. Vaccine hesitancy or refusal can be affected by numerous factors such as availability, confidence, cost, anxiety, convenience, and misinformation (Nguyen et al., 2021; Loomba et al., 2021). Additionally, poor public health communication can cause confusion, skepticism, and resistance in the population, which may negatively impact the implementation of immunization programs (Butler et al., 2015).
There is a need for locally contextualized research that closely examines effective immunization campaign practices. It is important to identify useful principles, approaches, and strategies for effective mass vaccination campaigns that increase uptake and address specific aspects influencing vaccine hesitancy. The author reviews a few lessons learned from public vaccination campaigns in South Korea (hereinafter Korea), which earned global acclaim for their effective and successful COVID-19 immunization programs. By August 2022, $94.8\%$ of Korean citizens over 12 years had already received their second dose, $71.3\%$ had received the third (booster) dose, and $16.2\%$ (individuals over 50 years) had received the fourth dose (Korean Ministry of Health and Welfare [KMHW], 2022a). These exemplary statistics are indicative of Korea’s high vaccination rate compared to the rest of the world. From the time vaccines became available for public use until August 2022, Korea administered 248.88 vaccine doses per 100 people. Meanwhile, China administered 240.75 vaccine doses per 100 people, Italy 236.93, Canada 232.32, Australia 225.29, the United Kingdom 224.60, France 221.56, Germany 221.44, Israel 196.12, and the United States 183.73 (Our World in Data, 2022). Before implementing the COVID-19 vaccination campaign, the Korean public largely showed strong hesitancy, having a “wait and see” attitude toward the COVID-19 vaccine; many were concerned about the adverse effects of vaccination (You, 2021; Views and News, 2021). However, the country overcame widespread vaccine hesitancy by applying social marketing as a behavioral change strategy to public health campaigns. Korea achieved high immunization coverage by addressing the causes of vaccine skepticism and positively appealing to the target audience’s rationale.
Although there is no catholicon communication method for public health communications in times of unparalleled pathogenic crisis, this study highlights the potential value of commercial marketing techniques in immunization promotion campaigns. Such techniques may bolster a campaign’s chances of success by fostering behavioral changes in the target population. This study introduces evidence from South Korea’s successful COVID-19 immunization campaign, incorporating core elements of commercial marketing techniques to influence people’s behavior. It also identifies five key communication attributes that the country’s public health authority considered for an optimum impact while implementing its campaign: (i) a high degree of proactiveness, (ii) credibility, (iii) fighting misinformation, (iv) emphasizing social norms and prosocial behavior, and (v) coherence. This finding offers new insights that can be employed at a national and global scale and is a notable contribution to the literature (Lefebvre, 2013; Thorpe et al., 2022; Jin et al., 2021; Murewanhema et al., 2022; Boyd and Buchwald, 2022; Hyland-wood et al., 2021).
## Theoretical perspective: social marketing as a public health intervention strategy
Social marketing has attracted significant attention from health researchers and practitioners as an effective and holistic intervention to increase vaccine uptake and respond to vaccine reluctance (Nowak et al., 2015; French and Gordon, 2020). The World Health Organization (WHO) has recommended social marketing strategies to build vaccine confidence and address low vaccination rates (WHO, 2020, 2022a).
Social marketing programs use commercial marketing-based principles to motivate individuals to adopt the suggested social behaviors to achieve common societal interests. Social marketing is defined by the International Social Marketing Association (iSMA) as an activity that “seeks to develop and integrate marketing concepts with other approaches to influence behaviors that benefit individuals and communities for the greater social good” (iSMA, 2020). Social marketing focuses on effecting change at multiple levels of society (person, society, and institution), culminating in transformation (Kemper and Ballantine, 2017). These models have been used widely in the public health field to enhance the health and well-being of people by increasing awareness of health issues and altering health behaviors (e.g., limiting cigarette use, advocating a healthy diet, using contraceptives, HIV prevention) (Kemper and Ballantine, 2017; Shams, 2018). With the release of COVID-19 vaccines, social marketing techniques are adopted in immunization campaigns to educate the general public about the vaccine’s safety and effectiveness, build positive social norms for vaccine uptake, and “nudge” people toward getting vaccines (Rhodes et al., 2021; Evans and French, 2021).
The core concept of the social marketing program is a willful, voluntary behavioral change for one’s own satisfaction and self-interest (Kotler and Armstrong, 2020). Customer buying behaviors are regarded as value exchanges in which both sides mutually benefit. As both players primarily act in their own interests, commercial marketers must first understand their targeted audience’s underlying needs, wants, and interests for their marketing to be successful. Customer orientation is a key theoretical cornerstone of social marketing programs. Following this assessment, a marketer delivers a product that meets customers’ needs and lowers the barriers that might hinder their purchasing behavior. Customers offset the disadvantages of alternatives and decide whether the marketer’s product or service is beneficial and valuable before voluntarily exchanging their resources for the goods offered (French, 2017).
## The marketing mix: The 4Ps for behavior change
The “4Ps” model—comprising product, price, place, and promotion—is a central element of the social marketing framework. First suggested by Kotler and Zaltman [1971], the model’s four key categories are beneficial for implementing marketing initiatives. The 4Ps help facilitate the development, communication, and promotion of a product to its target audience. Each dimension consists of a marketing variable that aims to make a product, service, or advocated behavior more appealing.
The premise is that a social marketer must produce the right product at the right price, distribute it to the right market in the right location, and market it to the right group (Lefebvre, 2013). The product comprises the offering’s qualities and functions, incorporating the benefits of utilizing the offering or engaging in a suggested set of actions. Critical to its desirability is how it corresponds to the consumer’s aspirations, needs, and interests and offers a solution to a problem. The price refers to the consumer’s cost or sacrifice in exchange for the product. It involves money, time, and physical or psychological effort invested in the exchange. The place is associated with the intermediary physical sales location, facilitating marketer–consumer exchange. The design of the place is crucial, as it provides sufficient incentives for consumers to engage with the product. They may include creating easy, convenient, or accessible outlets for people to engage in exchange. Promotion encompasses the marketer’s persuasive communication activities that emphasize product features/benefits, associated prices, and places to buy the offering. A promotional strategy usually comprises promotional activities via public relations, media and advertising, message delivery channels, and special events that influence change (MacDonald et al., 2013).
## The marketing mix for COVID-19 vaccine communication
The social marketing approach offers consumer behavioral insights into vaccines and immunizations. The approach helps develop programs and acquire knowledge about the wants, values, and needs of target individuals whose health behavior changes we aim to influence. Immunization program planning decisions can be made when policymakers understand the benefits and barriers of the inoculation program from the targeted audience’s perspective. This enables them to create a demand for vaccination services in the local community. Well-designed immunization campaigns can certainly boost vaccination coverage by enhancing public trust, alleviating anxiety and fear, and enabling people to connect better with the community and its goals (Lee et al., 2022; Shekhar, 2022).
As the Korean public health authority began the campaign, it projected “herd immunity” as the best way to end the pandemic and recommended that everyone be immunized against SARS-CoV-2 (KDCA, 2021a). This policy encouraged the public to embrace new behaviors (vaccine uptake) and avoid antisocial ones (vaccine delay or not getting vaccinated); however, this policy was not mandatory. It relied on voluntary cooperation because it was presumed that the general population would adopt a behavioral change if its benefits/rewards/social consequences suited their needs. Public health officials have applied the social marketing approach to health communication to positively stimulate vaccine-acceptance behavior in the population. They also tailored their communication strategies and distributed accurate vaccine information via diverse media channels. The 4Ps model is used to explain vaccine communication as follows.
## Product communication
“Product” refers to what is being sold—in this instance, the vaccine’s benefits. Vaccine intention can be predicted by the perceived benefits of vaccines (Lee et al., 2022). Recommended behaviors include adherence to the vaccination schedule or taking the vaccine; engaging in this behavior facilitates health benefits (Nowak et al., 2015; Wassler et al., 2022). Product communication is applied to vaccines by identifying the behavioral mechanisms that benefit people. These behavioral proposals aim to enhance individual and community health by stimulating the human immune system to produce antibodies and fight against COVID-19 infection. This reduces the incidence of contracting or transmitting the infection, lowers the mortality rate, and decreases the chance of developing severe COVID-19 symptoms among fully vaccinated populations (Dye, 2022; Sadarangani et al., 2021; Kerr et al., 2022).
Whenever faced with hesitancy or resistance, the Korean public health authority provided detailed information to build public trust in vaccines (products) (KDCA, n.d.-a). Specifically, they explicitly stated the need for vaccination uptake (behavior change), stressing its worth and the multifold benefits of COVID-19 immunization. Great values of COVID-19 vaccines include high efficacy; protection from developing severe symptoms, hospitalization, and death; prosocial behavior for the community; and reconnecting with social networks without concern of infection (KDCA, n.d.-a; KMHW, n.d.). Such a goal is attainable if each individual adopts protective behaviors and takes the COVID-19 vaccine (Statistic Korea, n.d.-b).
The KDCA presented scientific evidence-based communication about the benefits of vaccination during its daily public briefings, informing the public about clinical trial results that suggest high vaccine efficacy rates (Polack et al., 2020; Voysey et al., 2021). The KDCA addressed the benefits of booster shots to stay protected against subvariants by showing how people who had taken the third booster had reduced the risk of infection and hospitalizations compared with those who received only two doses (KMHW, 2022b). Communicating proven scientific data to the public helped individuals better understand the vaccine’s benefits and perceive immunization as a self-protective behavior.
Extending the values of immunization beyond individual health benefits, the Korean government has publicized the significant social benefits of vaccination to appeal to the public; as herd immunity is achieved, population immunity against pathogens can be attained. In this way, countries can protect the lives of their citizens and the most vulnerable groups and escape the tremendous social and economic burden of COVID-19. Thus, failing to curtail virus transmission greatly risks national security. Owing to the collectivistic nature of Korean society, where people tend to value the community and their common interests over individual needs, this was a convincing strategy (KDCA, 2021c; Hong, 2022). Consequently, the public made informed decisions. In October 2021, the nation’s herd immunity threshold was surpassed; 41.31 million Korean adults ($79.7\%$ of the total population) had completed their COVID-19 immunization (Our World in Data, 2022). A post-vaccination survey revealed that the public’s primary reasons for COVID-19 immunization were to protect their family members ($76.4\%$) and to help the nation achieve herd immunity ($63.9\%$) (Yonhap News, 2021).
During the rollout, the Korean government introduced incentives to support the behavior change (Hug). Benefits such as exemptions from quarantine rules, paid-vaccine holidays, free access to parks, free lodging in private resort facilities, and free meals in certain restaurants across the country were provided to vaccinated individuals (Korean Ministry of Culture, Sports, and Tourism [KMCST], 2021).
## Price communication
“Price” refers to the target group barriers to procuring the vaccine, such as financial cost, inaccessibility, inconvenience, and perceived low vaccine efficacy or safety. Immunization planners must consider interventions to reduce these hurdles, alter public perception, and increase the perceived value of vaccination (WHO, 2021b). For strategic communication, public health officials must highlight the drivers of vaccination (e.g., safety, efficacy, health, and social rewards) that outweigh the perceived risk or cost (unexpected side effects, undesirable social consequences, etc.). The Korean government encouraged the public to be immunized to achieve herd immunity, and the communicated goal was to protect the community from virus infection and return to pre-COVID-19 normalcy. Furthermore, it emphasized that adverse side effects are extremely infrequent and assured the public that they would be compensated in the event of an adverse reaction to the vaccine (KDCA, n.d.-a).
Barriers can be linked to various factors, such as values, knowledge, abilities, and economic standing. However, the most common and significant obstacle to vaccination is associated with risk perception, such as side effects. Although the COVID-19 vaccine was acknowledged as the most effective public health intervention for virus control, its adverse effects have been regarded as a major barrier to vaccine intention (Nguyen et al., 2021). As stated previously, safety and potential adverse events were the primary barriers among the Korean public regarding the COVID-19 vaccination. To reduce the psychological price/cost of vaccination, the KDCA provided advocative proof of the safety and adequacy of COVID-19 immunization. In their briefings, the KDCA shared their recommendation of the importance of undertaking the COVID-19 vaccine and offered evidence from trial data, suggesting vaccine-attributable severe side effects were extremely rare, mostly short-term injection site pain, tiredness, and mild headaches (KDCA, 2021c).
The KDCA’s public message also focused on fostering vaccine literacy (e.g., how mRNA/protein subunit vaccines work to produce viral proteins and develop immunity in the body, active ingredients, effectiveness against variant strains of COVID-19, up-to-date clinical trial outcomes, and side effects of vaccines (KDCA, n.d.-a). To tackle misinformation that creates negative public opinion, the Korean government widely employed fact-checking and debunking strategies against anti-vaccination propaganda on various social media. If invalidated misinformation is circulated and receives public attention, the public health authority promptly refutes the rumors via fact-checking and debunking.
The government aimed to reduce the target behavior’s costs/prices/barriers; it also increased the costs of the competing behavior by employing disincentive measures. They announced compulsory quarantine regulation for the contact between confirmed COVID-19 cases and inbound travelers, and a 14-day mandatory quarantine was applied for unvaccinated individuals (The Korea Herald, 2021). People had to prove their negative status with a negative PCR test before being released from isolation (Shove). On December 13, 2021, the Korean government introduced the “vaccine passport” to further increase barriers against non-compliant behavior (KDCA, n.d.-a). During this period, adults had to present a vaccine certificate or a negative PCR test to access public venues such as gyms, concert halls, cinemas, and hospitals (Smack). This regulation was abolished on March 1, 2022, because of strong public opposition. Nevertheless, the COVID-19 certificates helped the country control the COVID-19 pandemic and maintain behavior adoption after reaching herd immunity.
## Place communication
“Place” denotes where and when people can access the vaccination service (e.g., hospitals, clinics, mass vaccination centers, etc.). Given that herd immunity can only be achieved by substantial public buy-in, the majority of the community must embrace product communication (getting a vaccine). Hence, multiple communication tools have been employed via numerous media channels to communicate with the public, effectively manage information flow, and enhance risk communication.
Throughout the immunization process, the Korean government provided open and honest communication to the public and widely disseminated information regarding COVID-19 vaccination via diverse media channels. Such information included how to make reservations, where to get vaccinations, and access to separate queues for older adults during vaccination registration (KDCA, n.d.-b, n.d.-d). Regarding the booking process, the KDCA made multiple formats available to ensure quick and easy public access. The KDCA’s website was the primary booking method; it provided a clear, easy-to-understand format for people to follow.
In addition, telephone options and in-person reservations were made available for older adults who were not familiar with the internet. The campaign message also informed the public about the accessibility of the vaccine and vaccination locations in real-time. In May 2021, the KDCA announced the availability of mobile app services on Naver and KaKao Talk messenger. These services enabled the public to search for nearby vaccination centers and provided up-to-date information on vaccine data. The government carefully selected the physical sites to ensure easy and convenient venues. Large sports arenas, cinemas, conference halls, hospitals, and clinics were available, along with free vehicle pick-up and drop-off services for individuals in remote areas (KMHW, 2020b). The government used clear, multiple, easy-to-read signs to direct people to vaccination centers (especially older adults). In the centers, trained staff guided them through the vaccination services and addressed their concerns. They also provided special care and friendly guidance for older adults.
## Promotion communication
“Promotion” refers to communication strategies for clear, accurate, and coherent information on immunization plans (CDC, n.d.). This involves providing updated vaccine information and suggested actions through trusted media outlets. This communication strategy leveraged public service advertisements, spokespersons, media outreach, and communication materials to promote vaccinations.
There were three key messages in the Korean government’s COVID-19 vaccination promotion campaign: vaccine services were free, vaccines were safe and effective, and vaccines provided the best protection against the virus for individuals, their families, and the community (Statistic Korea, n.d.-a). The goal of the promotional message was to persuade the public that they would benefit more if vaccinated. In January, a month before the rollout, the KDCA stated that vaccines would be free for all Korean citizens and foreign residents. This promotion strategy removed financial barriers and facilitated uptake behavior.
For successful immunization against COVID-19, the Korean public health authority conducted an extensive public education campaign concerning the safety and efficacy of vaccines with rational arguments. In this process, they recruited credible healthcare professionals, religious and community leaders, and local celebrities who had established trusting relationships with the community. Their advocacy message about vaccine safety and efficacy could support the message of public health communication (Catholic Bishop’s Conference of Korea, 2021).
Finally, the government’s promotional message advertised the benefits of behavior change (vaccination) as the best protection for oneself, others, and the whole nation (Statistic Korea, n.d.-b). In collectivistic cultures such as Korea, prosocial messages appeal to the sense of social responsibility and are more effective in engaging the public emotionally to work toward a common goal.
To reach the entire nation, the government communicated through the media. They used numerous conventional and emerging social media outlets, including official media and YouTube, Facebook, Twitter, Naver, and Kakao. The following information channels play an important role in increasing public trust in the safety and effectiveness of vaccines by disseminating current vaccine research development throughout immunization campaigns: mass media (TV, radio, Internet, metro/bus posters, street banners and signs, electronic documents, etc.), public videos (hospital waiting areas, public facilities, local city offices, community service centers, etc.), service-based communication (doctors, nurses, and healthcare workers), special events, and advocacy via influencers (political leaders, public health officials, local community and religious leaders, local celebrities, etc.) ( KDCA, 2018). To further increase the urgency of vaccine uptake, the Korean government sent reminders via text messages, pamphlets, and other informational materials to capture public attention (nudge). Sending repeated messages, notifying people of the vaccination schedule, and prompting implementation reminded the public to get vaccinated.
## Integrating behavioral intervention into the marketing mix: value creation and exchange
When changing attitudes and behaviors is insufficient, behavioral science proposes the use of other behavioral intervention tools, such as “nudging” or the default option, in combination with persuasive communication to overcome barriers and influence behavior (Evans and French, 2021; French and Gordon, 2020; Dai et al., 2021). Within the proposed approach, Evans and French [2021] recommend using both incentivizing and disincentivizing elements in social marketing strategies to elicit a behavioral response and maintain behavior change. They suggest four types of behavioral interventions: Hug, Smack, Shove, and Nudge, to increase the target audience’s vaccine uptake against COVID-19. Hug is an active cognitive engagement and positive incentive for adaptation (e.g., offering rewards with financial incentives, vaccine badges, vaccine holidays, and access permits for facilities when you have a “vaccine passport”). Nudge is a passive cognitive engagement and positive incentive for adaptation (e.g., setting up default options for the entire populace, using small financial rewards, gift cards, or sending reminder text messages or emails for vaccination). Smack is an active cognitive engagement and punishment for non-adaptation (e.g., no access to shopping malls, penalty fines, or dismissal from work in high-risk industries without a vaccine certificate). Shove is passive cognitive engagement and punishment for non-adaptation (e.g., multiple polymerase chain reaction (PCR) test requirements before using public facilities if not vaccinated). Together with persuasive communication, Korea employed these schemes in its COVID-19 vaccination policy to encourage vaccine uptake.
## Methods
This study reviewed the relevant literature from immunization inception in January 2021 to August 2022. Four major databases were searched: Web of Science, PubMed, ScienceDirect, and Factiva, using the following keywords: COVID-19 vaccination and recommendation OR public health communication/campaign. The study also searched three major governmental agencies—the Korean Presidential Office Broadcast (KTV), Korea Disease Control and Prevention Agency (KDCA), and KMHW—following both website and non-website communications regarding COVID-19 vaccination recommendations. Any communication material used to inform people—official documents, daily briefings, news reports, and website information and announcements—was collected as primary source data for the study. Additionally, a review of local media reports on COVID-19 vaccination was conducted using the online database Bigkind through keyword search.
## Results
The present study found that the social marketing public health campaign targeting COVID-19 vaccination behavior effectively increased voluntary uptake of the COVID-19 vaccine in Korea. The intended behavioral outcomes were observed when social marketing techniques were adopted in the immunization campaign. In August 2022, Korea achieved high COVID-19 immunization coverage. About $94.8\%$ of Korean citizens over the age of 12 years had completed their second dose; $71.3\%$ had received the third (booster) dose; and $16.2\%$ (over the age of 50) had completed the fourth dose (KMHW, 2022a). The pre-campaign survey revealed that $67.7\%$ of Koreans had concerns about vaccine safety and side effects and preferred to wait and see how others responded to it (Dailymedi News, 2021). However, a high prevalence of vaccine hesitancy, delay, or refusal was significantly decreased and behavior adaptation rates increased when people were exposed to social marketing immunization campaign messages. This change indicates that social marketing communication techniques effectively responded to concerns in the public sector, successfully engaged with the target audience’s perceptions, and changed their attitude, leading to action.
## Korea’s COVID-19 vaccine communication strategy
Korea was among the first countries to be severely affected by COVID-19 due to its proximity to China. The first confirmed case was reported on January 20, 2020, followed by three waves of infection. The first peak was reached between February and March 2020, the second between August and September 2020, and the third between November 2020 and February 2021 (KMHW, 2020a). In February 2021, Korea announced a COVID-19 vaccination program, aiming to immunize $70\%$ of the adult population by September 2021 (KDCA, 2021a, 2021b). However, most Koreans were reluctant to take the vaccine because of uncertainties about adverse long- or short-term side effects, and lack of awareness. Understanding the public’s concern and general attitude toward COVID-19 vaccines, the Korean public health authority started a social marketing campaign to educate the public regarding vaccine safety and efficacy. The program gradually increased vaccine confidence among the general public and engagement behaviors were manifested. The country’s slow initial vaccination uptake progressively improved as the promotion campaign continued, and a high uptake rate was eventually achieved. By August 2022, $94.8\%$ of Korean citizens over the age of 12 had completed the second dose, $71.3\%$ had received the third (booster) dose, and $16.2\%$ (those aged over 50) had received the fourth dose (KMHW, 2020a).
## Campaign effect: increasing the COVID-19 vaccine acceptance
Social marketing is a proven behavioral change technique, widely adopted in public health campaigns for promoting changes in knowledge, norms, belief, attitude, and behavior of the general population (Lee et al., 2022; Melovic et al., 2020; Coffie et al., 2022; Evans and French, 2021; Osborne et al., 2021). In the context of COVID-19 vaccination, the current study found that the social marketing vaccination campaign was effective in increasing individuals’ intention to adopt immunization behaviors as it could significantly improve the targets’ knowledge, belief, attitude, and barriers to COVID-19 vaccination (Adane et al., 2022; Salali et al., 2022; Evans and French, 2021).
## Knowledge, belief, and attitude toward COVID-19 vaccines
The level of vaccine hesitancy significantly decreased in Korea due to the ongoing efforts of Korean public health officials to convince and educate the public about the safety and efficacy of COVID-19 vaccines. In the pre-vaccination period, a significant portion of the Korean population was hesitant to vaccinate (Dailymedi News, 2021; Views and News, 2021). Only a small portion of people indicated a positive intention to vaccinate (Heo, 2021; You, 2021; Lee and Yang, 2021). Importantly, these studies reveal that high rates of vaccine hesitancy and skepticism are associated with safety concerns, insufficient vaccine literacy (e.g., poor perception or low belief in the efficacy of vaccines), and lack of awareness about the threat of COVID-19.
To our knowledge, the Korean population’s overall perception, knowledge, and attitude toward COVID-19 vaccination have become significantly favorable as the vaccination campaign progressed. A previous study also suggested that there were substantial improvements in overall knowledge, perceptions, and attitudes toward COVID-19 vaccination among the Korean public during the vaccine rollout, indicated by a steady increase in vaccine acceptance (Choi et al., 2022). A survey by Choi et al. [ 2022] conducted during the first rollout in March 2022 indicates an improved perception toward vaccines. After months of public campaigns on COVID-19 vaccination, the majority of the Korean public ($75.5\%$) felt that the decision to vaccinate was important and worth it. Moreover, $74.3\%$ of people said they believed in the vaccine’s efficacy, indicating high vaccine intention (You, 2021). This increased vaccine confidence and their positive attitudes could lead them to have voluntary behavioral engagement toward vaccination.
## Perceived barriers toward COVID-19 vaccines
The study observed that the social marketing immunization campaign effectively reduced perceived barriers to vaccination. As noted, the Korean public was skeptical regarding vaccine safety and its side effects (You, 2021; Lee and Yang, 2021). However, the public’s perception of barriers was significantly reduced, and their risk perception toward vaccination diminished during the COVID-19 immunization campaign. On social media platforms, keywords such as “side effects” and “safety” (the most frequent terms on the top four platforms, Naver, Daum, Google, and Twitter, during pre-vaccination in Korea) received less attention after the public immunization campaign began. Negative keywords related to the side effects associated with COVID-19 vaccines (e.g., blood clots, severe allergic reactions, or death) did not appear either (Choi et al., 2022). This observation was supported by a national survey conducted in April 2021, suggesting that the Korean public’s perceived benefit of vaccines surpassed the perceived barriers in all age groups. This tendency was reported to be even greater among older adults (60+ years) (You, 2021).
## Five effective public health messages to boost vaccine intention
Overcoming the COVID-19 pandemic necessitates adequate behavioral changes. Communicating effectively and persuasively with the public is critical to elicit a behavioral response (Betsch, 2020). This section discusses how the Korean government framed campaign messages to maximize their message effectiveness, persuading the general population to adopt vaccination behavior and identifying the main characteristics of the campaign, such as high proactiveness, credibility, fighting misinformation, emphasizing social norms, and coherence of official communication.
(i) Proactiveness: Delivering critical vaccine information to the public during a pandemic could create confusion and concern (Lyu et al., 2022). Hence, public health communication must anticipate problems and share necessary information. If the target audience’s attention is not captured from inception, misinformation and fake news by anti-vaccine propagandists will impact the process (Loomba et al., 2021; Roetzel, 2019; Zheng, 2022). Given the enormous amount of negative information about vaccines on social media, a negative attitude toward vaccination is likely to develop, resulting in low immunization uptake rates (Olson, 2020; Kim et al., 2021). Thus, proactive communication interventions have been emphasized and implemented as effective communication strategies throughout Europe (Butler et al., 2015). Korea also adopted the “Be First” messaging principle, emphasizing proactive communication in responding to public health emergencies (KDCA, 2018). Striving to be the louder voice in the country as the official information channel regarding COVID-19 vaccines, the country could successfully overcome vaccine hesitancy. Consequently, by proactively hushing the misinformation and anti-scientific attitudes, the Korean public health authority decreased anxieties and skepticism about vaccine safety and efficacy (Choi et al., 2022), and effectively increased vaccination uptake (KDCA, n.d.-c).
(ii) Credibility: Research has shown that higher credibility will likely generate greater message compliance (De Meulenaer et al., 2018). Credibility is a central feature of effective and persuasive health communication. Perceiving a source as trustworthy and valid can significantly affect the beliefs and intentions of the target population to adopt the suggested behaviors. Vaccine hesitancy can considerably decrease when the public perceives the public health authority as trustworthy (Trent et al., 2022). Korea’s public healthcare system is well equipped to provide quality healthcare services; therefore, public trust in the system is already high. During the COVID-19 pandemic, the Korean public showed a significantly high level of trust in the KDCA, which led to the pandemic mitigation plans and national inoculation policies ($78\%$ in the first week of June 2021) (Korea Research, 2021). Vaccine hesitancy or refusal can increase when the public loses faith in the governmental authority. A higher degree of confidence in public health authorities rendered the Korean public less prone to fake news or conspiracy theories. People were more willing to listen to the immunization messages delivered by Chong En-Kyong, the KDCA’s key commissioner, and other well-known public health scholars and expert groups (Hong, 2022; Abu-Akel et al., 2021). We may infer that a high level of public confidence in government organizations increases the likelihood of message acceptance toward COVID-19 vaccines and decreases the perceived validity of fake news. Therefore, conspiracy theories may not significantly affect Koreans’ views and attitudes toward vaccination.
(iii) Fighting misinformation: Misinformation about COVID-19 vaccines can cause public vaccine hesitancy or refusal by arousing doubt, anxiety, and concerns about vaccines (Nguyen et al., 2021; Loomba et al., 2021; Springer and Özdemir, 2022). The Korean government used fact-checking and debunking strategies to tackle mis/disinformation and vaccine hesitancy, and discredit commonly held myths; they exposed logical flaws in the misinformation to correct any misconceptions (e.g., COVID-19 vaccines cause infertility and other diseases, they have a microchip, they cause death, they alter one’s DNA, etc.) ( KDCA, 2021d). Concurrently, the KDCA and Korea Communication Commission (KCC) introduced a reporting system on the website www.KCC.go.kr/vaccinejebo to prevent the spread of misinformation on vaccine efficacy and safety. All anti-vaccine messages were monitored using artificial intelligence (AI), and messages spreading misinformation, fear, and negative opinions regarding government intervention were reported on the website and deleted from the media platform (KCC, n.d.). To further reduce the effect of fake news on immunization, Twitter, Korea, and YouTube removed over 43,000 tweets and 1,000,000 erroneous COVID-19-related messages and videos that disseminate false claims about vaccines (WHO, 2022b). This strategy significantly reduced public exposure to misinformation, which could impact vaccine intention.
(iv) Emphasizing social norms and prosocial behavior: Perceived social norms and beliefs about how others conduct themselves impact a person’s behavior and intent (Schultz et al., 2007). In the context of COVID-19 vaccination, individuals are more inclined to follow the COVID-19 protocol and get vaccinated if they believe that more people are participating in these preventive actions (Rabb et al., 2022). Notably, a study on the Korean public’s awareness and attitude toward COVID-19 immunization corroborates this result, showing a shift in perception from negative to positive as more people get the inoculation (Choi et al., 2022). We can infer that people’s beliefs and intentions toward COVID-19 vaccines are affected by the higher vaccination intentions and behaviors of others.
In particular, people belonging to collectivistic cultures prioritize the interests of society over the individuals’ interests. Preferences are more likely to be affected by beliefs about other people’s intentions—social norms—within the community and their close social networks, such as family, friends, and neighbors (Salali et al., 2022; Cammett and Lieberman, 2020). During the vaccine rollout, the Korean public health authority effectively influenced people’s vaccination behavior and intention. They used messages such as “everyone gets vaccinated,” “taking the vaccine is the right thing to do,” “get vaccinated to return to normal life,” “achieve herd immunity with the vaccination,” and “protect your family elders with the COVID-19 vaccine,” to assert social norms and prosocial behavior (KDCA, 2021c; KDCA, n.d.-c). The Korean government appealed to its people’s collective responsibility and intrinsic motivation to avoid harm to their social ties, thereby aligning its population with the public health goal of achieving herd immunity within a year. These messages from public health authorities about the country’s new immunization goal gave the public the collective task of vaccination against COVID-19 as early as possible. This message significantly affected vaccine intention and uptake behavior among the Korean public (Hong, 2022).
(v) Coherence: Previous studies suggest that clear, coherent, and consistent public health messages about the safety and efficacy of vaccines can effectively decrease public confusion and anxiety about vaccination, and increase vaccination intention (Jin et al., 2021; Murewanhema et al., 2022; Hyland-Wood et al., 2021). In the context of COVID-19 immunization communication, Korean public authorities sought to increase message receptivity and public support for immunization, reducing belief in the infodemic about COVID-19 vaccines (KDCA, 2018).
Since the rollout began, the KDCA provided clear and specific instructions to the public, calling for action (e.g., “To register for the vaccine, visit the KDCA website”; or “call the 1339 help desk directly”) and particular events (e.g., “From February 6 to April 2021, healthcare professionals, older adults over 75 years of age, and people with chronic conditions have to be vaccinated”) (KDCA, 2021a). Communicating coherent and consistent pro-vaccine messages was used to prompt people to adopt healthy attitudes and vaccination behaviors (e.g., “COVID-19 vaccines provide the best protection against novel coronavirus,” “The benefits of getting a COVID-19 vaccine outweigh the risks”) (KDCA, 2021c). All public health figures and organizations in Korea (e.g., the Central Disaster and Safety Countermeasure Headquarters, KMHW, KDCA, and local governments) applied this coherent and consistent message strategy to their vaccination messages (KMHW, 2019). Korea was able to communicate more effectively with the public and boost their vaccination intentions by presenting a uniform and united front among all government agencies throughout the vaccination campaign.
## Conclusions and outlook
Government communication could play a substantial role in influencing public attitudes toward immunization. The public’s understanding of the pandemic, immunization, and vaccine intentions can be shaped by public health messages. This study analyzed the Korean government’s COVID-19 mass vaccination campaign strategies and interventions and offered insights contributing to global discussions on health communication strategies and approaches. The results indicate that a robust social marketing campaign can effectively customize messages according to the target population’s interests and values, persuade the public about the product (the need for vaccination), and overcome barriers to immunization acceptance. It enhances general confidence in the COVID-19 vaccine and, at least partly, overcomes hesitancy by increasing and reinforcing vaccine literacy, providing balanced information about the benefits and risks, and dispelling rumors and misconceptions. Furthermore, the study identified five key communication attributes—proactiveness, credibility, fighting misinformation, emphasizing social norms and prosocial behavior, and coherence—for official communication to improve communication interventions for maximum effect. These attributes can be applied in other countries’ vaccine messaging campaigns and national immunization programs. Although the study has explored the link between government communication and individuals’ intention to vaccinate through a social marketing perspective, it has limitations. This study relied on a document analysis design. Future studies must employ a quantitative design, such as a cross-sectional survey or regression analysis, to further examine the causal effects between variables. In addition, this study was conducted in Korea, and the results should not be generalized to other cultural settings. Consequently, future research must examine communication and messaging strategies from other geographic locations with more diverse population groups and sociocultural contexts. Repeated local studies using other approaches that facilitate vaccination will be essential for vaccine promotion. This is essential for public health policymakers to devise suitable communication intervention strategies to improve vaccination, booster uptake, and future immunization.
## References
1. Abu-Akel A, Spitz A, West R. **The effect of spokesperson attribution on public health message sharing during the COVID-19 pandemic**. *PLoS ONE* (2021.0) **16** 1-15. DOI: 10.1371/journal.pone.0245100
2. Adane M, Ademas A, Kloos H. **Knowledge, attitudes, and perceptions of COVID-19 vaccine and refusal to receive COVID-19 vaccine among healthcare workers in northeastern Ethiopia**. *BMC Public Health* (2022.0) **22** 1-14. DOI: 10.1186/s12889-021-12362-8
3. Anderson RM, Vegvari C, Truscott J, Collyer BS. **Challenges in creating herd immunity to SARS-CoV-2 infection by mass vaccination**. *Lancet* (2020.0) **396** 1614-1616. DOI: 10.1016/S0140-6736(20)32318-7
4. Betsch C. **How behavioural science data helps mitigate the COVID-19 crisis**. *Nat Hum Behav* (2020.0) **4** 438. DOI: 10.1038/s41562-020-0866-1
5. Boyd AD, Buchwald D. **Factors that influence risk perceptions and successful COVID-19 vaccination communication campaigns with American Indians**. *Sci Commun* (2022.0) **44** 130-139. DOI: 10.1177/10755470211056990
6. Butler R, MacDonald NE. **Diagnosing the determinants of vaccine hesitancy in specific subgroups: the guide to tailoring immunization programmes (TIP)**. *Vaccine* (2015.0) **33** 4176-4179. DOI: 10.1016/j.vaccine.2015.04.038
7. Cammett M, Lieberman E, Edmond J. **Building solidarity: challenges, options, and implications for COVID-19 responses**. *COVID-19 rapid response impact initiative* (2020.0) 1-34
8. Catholic Bishop’s Conference of Korea (2021) 2021 Fund for ‘vaccine share movement’ delivered to Vatican. https://cbck.or.kr/Notice/20211081?gb=K1200
9. Centers for Disease Control and Prevention (n.d.) CDCyergy Lite: social marketing made simple. https://www.cdc.gov/healthcommunication/pdf/CDCynergyLite.pdf
10. Choi YJ, Lee J, Paek SY. **Public awareness and sentiment toward COVID-19 vaccination in South Korea: findings from big data analytics**. *Int J Environ Res Public Health* (2022.0) **19** 9914. DOI: 10.3390/ijerph19169914
11. Coffie IS, Nkukpornu A, Kankam WA, Ocloo CE. **Using social marketing to demystify the myths surrounding COVID-19 vaccination: the mediating role of important others**. *Soc Mark Q* (2022.0) **28** 169-183. DOI: 10.1177/15245004221097802
12. Dai H, Saccardo S, Han MA. **Behavioural nudges increase COVID-19 vaccinations**. *Nature* (2021.0) **597** 404-409. DOI: 10.1038/s41586-021-03843-2
13. Dailymedi news (2021) 2 Out of 3 Koreans said, I will not get COVID-19 vaccine right away. http://www.dailymedi.com/detail.php?number=865209
14. De Meulenaer S, Pelsmacker PD, Dens N. **Power distance, uncertainty avoidance, and the effects of source credibility on health risk message compliance**. *Health Commun* (2018.0) **33** 291-298. DOI: 10.1080/10410236.2016.1266573
15. Dye C. **The benefits of large scale COVID-19 vaccination**. *BMJ* (2022.0) **377** o867. DOI: 10.1136/bmj.o867
16. Evans WD, French J. **Demand creation for COVID-19 vaccination: overcoming vaccine hesitancy through social marketing**. *Vaccines* (2021.0) **9** 319. DOI: 10.3390/vaccines9040319
17. French J. *Social marketing and public health: theory and practice* (2017.0)
18. French J, Gordon R. *Strategic social marketing: for behavior and social change* (2020.0)
19. Heo JH (2021) Koreans’ happiness survey. National Assembly Future Institute. Korea Social Science Data Archive (KOSSDA) 10.22687/KOSSDA-A1-2021-0003-V1.0
20. Hong SA. **Six pivotal lessons learned in South Korea for whole-of-government approach to successful COVID-19 vaccine roll out in planetary health**. *OMICS J Integr Biol* (2022.0) **26** 567-579. DOI: 10.1089/omi.2022.0064
21. Hyland-Wood B, Gardner J, Leask J, Ecker UKH. **Toward effective government communication strategies in the era of COVID-19**. *Humanit Soc Sci Commun* (2021.0) **8** 1-11. DOI: 10.1057/s41599-020-00701-w
22. iSMA (2020) What is social marketing. https://isocialmarketing.org
23. Jin Q, Raza SH, Yousaf M, Zaman U, Siang JMLD. **Can communication strategies combat COVID-19 vaccine hesitancy with trade-off between public service messages and public skepticism? Experimental evidence from Pakistan**. *Vaccines* (2021.0) **9** 757. DOI: 10.3390/vaccines9070757
24. Kim H, Chang HI, Jang SJ. **Political psychology of vaccination: COVID-19 conspiracy theories, evaluation of government responses, and vaccination intention**. *Korean Party Stud Rev* (2021.0) **20** 99-130. DOI: 10.30992/KPSR.2021.09.20.3.99
25. Kemper JA, Ballantine PW. **Socio-technical transitions and institutional change: addressing obesity through macro-social marketing**. *J Macromark* (2017.0) **37** 381-392. DOI: 10.1177/0276146717715746
26. Kerr S, Vasileiou E, Robertson C, Sheikh A. **COVID-19 vaccine effectiveness against symptomatic SARS-CoV-2 infection and severe COVID-19 outcomes from Delta AY.4.2: cohort and test-negative study of 5.4 million individuals in Scotland**. *J Glob Health* (2022.0) **12** 05025. DOI: 10.7189/jogh.12.05025
27. Korea Communications Commission (n.d.) false claims on COVID-19 vaccine. https://www.kcc.go.kr/user.do?page=A09030700&dc=K09030700
28. Korea Disease Control and Prevention Agency (2018) Standard operating procedure for risk communication in case of public health emergencies https://www.kdca.go.kr/npt/biz/npp/portal/nppPblctDtaView.do?pblctDtaSeAt=9&pblctDtaSn=1247
29. Korea Disease Control and Prevention Agency (2021a) COVID-19 vaccination plan. https://ncv.kdca.go.kr/menu.es?mid=a10117010000
30. Korea Disease Control and Prevention Agency (2021b) Policy Briefing on COVID-19 vaccination, ‘Achieving Heard Immunity Within This Year. https://www.kdca.go.kr/gallery.es?mid=a20503030300&bid=0004&b_list=9&act=view&list_no=144955
31. Korea Disease Control and Prevention Agency (2021c) Be assured, it is safe to take COVID-19 vaccine. https://www.kdca.go.kr/gallery.es?mid=a20503030300&bid=0004&b_list=9&act=view&list_no=144959
32. Korea Disease Control and Prevention Agency (2021d) Claiming circulating online about President Moon’s faking his COVID-19 shot is not true https://ncv.kdca.go.kr/board.es?mid=a11802000000&bid=0030&act=view&list_no=298&tag=&nPage=5 Accessed 7 Sep 2022
33. Korea Disease Control and Prevention Agency (n.d.-a) Information on COVID-19 vaccination. https://ncv.kdca.go.kr/menu.es?mid=a12205000000
34. Korea Disease Control and Prevention Agency (n.d.-b) Vaccination for above 75 years of age and Nursing Home. https://ncv.kdca.go.kr/menu.es?mid=a12213000000
35. Korea Disease Control and Prevention Agency (n.d.-c) Explanation for news reports. https://kdca.go.kr/board/board.es?mid=a20501030000&bid=0015
36. Korea Disease Control and Prevention Agency (n.d.-d) Registration for COVID-19 vaccine https://ncvr.kdca.go.kr/cobk/index.html
37. Korea Research (2021) [COVID-19] 66th National Survey of Public Perceptions (3rd week of September) https://hrcopinion.co.kr/archives/24401#
38. Korean Ministry of Culture, Sport, and Tourism (2021) COVID-19 vaccination incentive Q&A http://www.korea.kr/news/policyNewsView.do?newsId=148887904
39. Korean Ministry of Health and WelfareInfectious disease disaster risk management standard manual2019SeoulGovernment of the Republic of Korea. *Infectious disease disaster risk management standard manual* (2019.0)
40. Korean Ministry of Health and Welfare (2020a) Coronavirus disease-19 http://ncov.mohw.go.kr/en/
41. Korean Ministry of Health and Welfare (2020b) Financial support of KRW39 billion for 69 COVID-19 designated hospitals. http://www.mohw.go.kr/react/al/sal0301vw.jsp?PAR_MENU_ID=04&MENU_ID=0403&CONT_SEQ=353539&page=1
42. Korean Ministry of Health and Welfare (2022a) COVID-19 Situation and vaccination briefing report. https://www.kdca.go.kr/board/board.es?mid=a20501010000&bid=0015&list_no=720586&cg_code=&act=view&nPage=2
43. Korean Ministry of Health and Welfare (2022b) Briefing on COVID-19 vaccination and situation update. http://ncov.mohw.go.kr/tcmBoardView.do?brdId=3&brdGubun=31&dataGubun=&ncvContSeq=6260&board_id=312
44. Korean Ministry of Health and Welfare (n.d.) Video news. http://www.mohw.go.kr/react/jb/sjb0707ls.jsp?PAR_MENU_ID=03&MENU
45. Kotler P, Armstrong G. *Principles of marketing* (2020.0)
46. Kotler P, Zaltman G. **Social marketing: an approach to planned social change**. *J Mark* (1971.0) **35** 3-12. DOI: 10.2307/1249783
47. Lee D, Rundle-Thiele S, Wut TM, Li G. **Increasing seasonal influenza vaccination among university students: a systematic review of programs using a social marketing perspective**. *Int J Environ Res Public Health* (2022.0) **19** 7138. DOI: 10.3390/ijerph19127138
48. Lee YH, Yang YR. **A study on perceptions of university students about the COVID-19 vaccine**. *J Health Care Life Sci* (2021.0) **9** 185-193. DOI: 10.22961/JHCLS.2021.9.1.185
49. Lefebvre RC. *Social marketing and social change: strategies and tools for health, well-being, and the environment* (2013.0)
50. Loomba S, de Figueiredo A, Piatek SJ, de Graaf K, Larson HJ. **Measuring the impact of COVID-19 vaccine misinformation on vaccination intent in the UK and USA**. *Nat Hum Behav* (2021.0) **5** 337-348. DOI: 10.1038/s41562-021-01056-1
51. Lyu H, Zheng Z, Luo J (2022) Misinformation versus Facts: understanding the influence of news regarding COVID-19 vaccines on vaccine uptake. Health Data Sci. 10.34133/2022/9858292
52. MacDonald L, Cairns G, Angus K, de Andrade M. **Promotional communications for influenza vaccination: a systematic review**. *J Health Commun* (2013.0) **18** 1523-1549. DOI: 10.1080/10810730.2013.840697
53. MacDonald NE. **Vaccine hesitancy: definition, scope and determinants**. *Vaccine* (2015.0) **33** 4161-4164. DOI: 10.1016/j.vaccine.2015.04.036
54. Melovic B, Jaksic Stojanovic A, Vulic TB, Dudic B, Benova E. **The impact of online media on parents’ attitudes toward vaccination of children—social marketing and public health**. *Int J Environ Res Public Health* (2020.0) **17** 5816. DOI: 10.3390/ijerph17165816
55. Murewanhema G, Musuka G, Mukwenha S, Chingombe I, Mapingure MP, Dzinamarira T. **Hesitancy, ignorance or uncertainty? The need for effective communication strategies as Zimbabwe’s uptake of COVID-19 vaccine booster doses remains poor**. *Public Health Res Pract* (2022.0) **3** 100244. DOI: 10.1016/j.puhip.2022.100244
56. Nguyen KH, Srivastav A, Razzaghi H, Williams W, Lindley MC, Jorgensen C, Abad N, Singleton JA. **COVID-19 Vaccination intent, perceptions, and reasons for not vaccinating among groups prioritized for early vaccination—United States, September and December 2020**. *Am J Transplant* (2021.0) **21** 1650-1656. DOI: 10.1111/ajt.16560
57. Nowak GJ, Gellin BG, MacDonald NE, Butler R. **Addressing vaccine hesitancy: the potential value of commercial and social marketing principles and practices**. *Vaccine* (2015.0) **33** 4204-4211. DOI: 10.1016/j.vaccine.2015.04.039
58. Osborne MT, Kenah E, Lancaster K, Tien J (2021) Catch the tweet to fight the flu: using Twitter to promote flu shots on a college campus. J Am Coll Health 1–15 10.1080/07448481.2021.1973480
59. Olson O, Berry C, Kumar N. **Addressing parental vaccine hesitancy toward childhood vaccines in the United States: a systemic literature review of communication interventions and strategies**. *Vaccines* (2020.0) **8** 590. DOI: 10.3390/vaccines8040590
60. Our World in Data (2022) Total number of COVID-19 vaccine doses administered by Country. https://ourworldindata.org/covid-vaccinations
61. Polack FP, Thomas SJ, Kitchin N. **Safety and efficacy of the BNT162b2 mRNA COVID-19 vaccine**. *N Engl J Med* (2020.0) **383** 2603-2615. DOI: 10.1056/NEJMoa2034577
62. Rabb N, Bowers K, Glick D, Wilson KH, Yokum D. **The influence of social norms varies with “others” groups: evidence from COVID-19 vaccination intentions**. *Psychol Cogn Sci* (2022.0) **119** e2118770119. DOI: 10.1073/pnas.2118770119
63. Rhodes A, Hoq M, Measey MA, Danchin M. **Intention to vaccinate against COVID-19 in Australia**. *Lancet Infect Dis* (2021.0) **21** E110. DOI: 10.1016/S1473-3099(20)30724-6
64. Roetzel PG. **Information overload in the information age: a review of the literature from business administration, business psychology, and related disciplines with a bibliometric approach and framework development**. *Bus Res* (2019.0) **12** 479-522. DOI: 10.1007/s40685-018-0069-z
65. Sadarangani M, Abu Raya B, Conway JM, Iyaniwura SA, Falcao RC, Colijn C, Coombs D, Gantt S. **Importance of COVID-19 vaccine efficacy in older age groups**. *Vaccine* (2021.0) **39** 2020-2023. DOI: 10.1016/j.vaccine.2021.03.020
66. Salali G, Uysal M, Bozyel G, Akpinar E, Aksu A. **Does social influence affect COVID-19 vaccination intention among the unvaccinated. ?**. *Evol Hum Sci* (2022.0) **4** 1-16
67. Schultz PW, Nolan JM, Cialdini RB, Goldstein NJ, Griskevicius V. **The constructive, destructive, and reconstructive power of social norms**. *Psychol Sci* (2007.0) **18** 429-434. DOI: 10.1111/j.1467-9280.2007.01917.x
68. Shams M, Haider M, Platter HN. **Social marketing for health: theoretical and conceptual considerations**. *Selected issues in global health communications* (2018.0) 1-90
69. Shekhar SK. **Social marketing plan to decrease the COVID-19 Vaccine hesitancy among senior citizens in rural India**. *Sustainability* (2022.0) **14** 7561. DOI: 10.3390/su14137561
70. Springer S, Özdemir V. **Disinformation as COVID-19’s twin pandemic: false equivalences, entrenched epistemologies, and causes-of-causes**. *OMICS J Integr Biol* (2022.0) **26** 82-87. DOI: 10.1089/omi.2021.0220
71. Statistic Korea (n.d.-a) Contents on COVID-19 vaccine promotion https://kosis.kr/covid/html/info_contents_list.do. Accessed 9 Oct 2022
72. Statistic Korea (n.d.-b) Please be immunized to reach herd immunity https://kosis.kr/covid/html/info_contents_list.do. Accessed 9 Oct 2022
73. The Korea Herald (2021) [KH Explains] ‘How to avoid 14-day quarantine in S. Korea if vaccinated abroad’. https://www.koreaherald.com/view.php?ud=20210616000780
74. Thorpe A, Fagerlin A, Drews FA, Butler J, Stevens V, Riddoch MS, Scherer LD. **Communications to promote interest and confidence in COVID-19 vaccines**. *AM J Health Promot* (2022.0) **36** 976-986. DOI: 10.1177/08901171221082904
75. Trent M, Seale H, Chughtai AA, Salmon D, MacIntyre R. **Trust in government, intention to vaccinate and COVID-19 vaccine hesitancy: a comparative survey of five large cities in the United States, United Kingdom, and Australia**. *Vaccine* (2022.0) **40** 2498-2505. DOI: 10.1016/j.vaccine.2021.06.048
76. Views & News (2021) [Korea Gallup] …. 72% of Koreans concern about side effects. https://www.viewsnnews.com/article?q=189377
77. Voysey M, Clemens SAC, Madhi SA. **Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: An interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK**. *Lancet* (2021.0) **397** 99-111. DOI: 10.1016/S0140-6736(20)32661-1
78. Wassler P, Chiappa GD, Nguyen THH, Fedeli G, Williams NL (2022) Increasing vaccination intention in pandemic times: a social marketing perspective. Ital J Mark 37–58 10.1007/s43039-022-00049-w
79. World Health Organization (2020) COVID-19 vaccines: safety surveillance manuel: COVID-19 vaccine safety communication https://www.who.int/docs/default-source/covid-19-vaccines-safety-surveillance-manual/covid19vaccines_manual_communication.pdf
80. World Health Organization (2021a) Vaccine efficacy, and effectiveness and protection. https://www.who.int/news-room/feature-stories/detail/vaccine-efficacy-effectiveness-and-protection
81. World Health Organization (2021b) Communicating with patients about COVID-19 vaccination. https://apps.who.int/iris/bitstream/handle/10665/340751/WHO-EURO-2021-2281-42036-57837-eng.pdf
82. World Health Organization (2022a) Strengthening COVID-19 vaccine demand and uptake in refugees and migrants. https://www.who.int/publications/i/item/WHO-2019-nCoV-immunization-demand_planning-refugees_and_migrants-2022.1
83. World Health Organization (2022b) Coronavirus (COVID-19) Dashboard. https://covid19.who.int/ Accessed 11 Jan 2023
84. Yonhap News (2021) Nearly 7 out of 10 Koreans willing to take vaccine: poll. https://www.koreatimes.co.kr/www/nation/2021/05/113_309690.html
85. You MS. *11th National survey of public perceptions concerning COVID-19 vaccination report. KBS* (2021.0)
86. Zheng H, Jiang S, Rosenthal S. **Linking online vaccine information seeking to vaccination intention in the context of the COVID-19 pandemic**. *Sci Commun* (2022.0) **44** 320-346. DOI: 10.1177/10755470221101067
|
---
title: A retrospective study on comparison of clinical characteristics and outcomes
of diabetic ketoacidosis patients with and without acute pancreatitis
authors:
- Adeel Ahmad Khan
- Fateen Ata
- Zohaib Yousaf
- Mohamad Safwan Aljafar
- Mohammed Najdat Seijari
- Ahmad Matarneh
- Bassel Dakkak
- Malik Halabiya
- Bassam Muthanna
- Abdul Majeed Maliyakkal
- Anand Kartha
journal: Scientific Reports
year: 2023
pmcid: PMC10018622
doi: 10.1038/s41598-023-31465-3
license: CC BY 4.0
---
# A retrospective study on comparison of clinical characteristics and outcomes of diabetic ketoacidosis patients with and without acute pancreatitis
## Abstract
The co-existence of diabetic ketoacidosis (DKA) with acute pancreatitis (AP) is associated with unfavorable clinical outcomes. However, diagnosing AP in DKA patients is challenging and often missed due to overlapping symptoms. The aim of this retrospective observational study was to compare the clinical characteristics and outcomes of patients with concomitant DKA and AP or DKA alone. Data of patients with DKA admitted between January 2015 to August 2021 to four hospitals in Qatar was extracted from the electronic health record (Cerner). American Diabetes *Association criteria* and *Atlanta criteria* were used for DKA and AP diagnosis, respectively. Independent T-test or Mann–Whitney U test was used to analyze continuous variables, whereas categorical variables were analyzed via Chi-square or Fischer exact tests as appropriate. Univariate and multivariate logistic regression models were generated to assess the correlations. A p-value of < 0.05 was considered statistically significant. Of 936 patients with DKA, 84 ($9.0\%$) had coexisting AP. AP was most common in the Asian race ($66\%$, $p \leq 0.001$). Patients with DKA and AP were older, had higher admission anion-gap, white cell count, hemoglobin (hb), neutrophil/lymphocyte ratio, urea, creatinine, maximum blood glucose during the episode, total cholesterol and triglyceride level (TGL) ($p \leq 0.05$). They had a lower admission venous pH and bicarbonate at 6 h. Patients in the DKA with AP group also had a longer length of stay (LOS), DKA duration and a higher rate of ICU admission (p-values ≤ 0.001). In-hospital mortality, 3-month all-cause readmission, 6-month and 12-month DKA recurrence did not differ between the two groups. Univariate logistic regression analysis showed age, Asian ethnicity, male gender, T2D, admission WBC count, hb, urea, creatinine, potassium, venous pH, bicarbonate, anion gap, total cholesterol, TGL and LDL level were significantly associated with the development of DKA with AP ($p \leq 0.05$). In multivariate logistic regression analysis, age and total cholesterol level were associated with concomitant DKA and AP ($p \leq 0.05$). Patients with concomitant DKA and AP have more severe derangement in markers of DKA severity, inflammation, kidney injury and metabolic profile, along with a longer DKA duration, LOS and requirement for ICU support compared to DKA patients without AP. This highlights the clinical significance of diagnosing the co-existence of DKA with AP, as the combination results in significantly worse clinical outcomes and greater healthcare utilization than in patients with only DKA.
## Introduction
Diabetic ketoacidosis (DKA) is a serious metabolic complication of diabetes mellitus (DM) characterized by the development of ketosis, hyperglycemia and a high anion gap metabolic acidosis1,2. It is more common in type 1 diabetes (T1D) patients than in type 2 diabetes (T2D). The incidence of DKA in T1D patients is reported to be $\frac{178.6}{10}$,000 patient-years, while in T2D, it is $\frac{20}{10}$,000 patient-years3. The disease is associated with a very high risk of short-term mortality, with a study reporting a mortality rate of $5.2\%$ over a 4-year follow-up after DKA admission4. Acute pancreatitis (AP) is another life-threatening condition with an estimated incidence of 33.74 cases per 100,000 patient years and 1.6 deaths per 100,000 patient-years5. The development of concomitant DKA and AP has been reported in the literature. Recognition of the co-existence of these two entities is associated with unfavorable clinical outcomes but is challenging to recognize because of the overlapping symptoms. Abdominal pain, the most common symptom of AP, is present in up to $46\%$ of cases of DKA and is strongly associated with more severe metabolic acidosis6. Interestingly, some case reports reported concomitant AP and DKA as even the initial manifestation of both T1D and T2D7,8.
Several studies have compared patients with coexisting DKA with AP to patients with AP alone. Yuan et al. reported a higher rate of acute kidney injury (AKI), length of stay (LOS), and severity of AP in patients of the DKA and AP group as compared to the AP group9. Wang et al. also reported a higher incidence of AKI, an increase in intensive care unit (ICU) admission and a higher APACHE II score in DKA with AP patients compared to AP only patients10. Similar results were described by Fu et al., who found statistically significant differences in ICU admission and LOS between the two groups11. However, data comparing patients with concomitant DKA and AP to patients with DKA alone is limited. Ma et al. compared the characteristics of DKA patients with and without pancreatitis. Their study showed a longer LOS and a higher ICU admission rate in patients with a combination of DKA and AP compared to DKA alone. However, the study only included 25 patients with coexisting AP and excluded patients with T1D12. Our study aimed to assess the differences in clinical characteristics and outcomes of patients with these two coexisting conditions compared to those with DKA alone. Early identification of this combination of illnesses can alert physicians to patients at high risk of developing worse clinical outcomes.
## Materials and methods
This is a retrospective, cross-sectional study and included consecutive patients with DKA diagnosis presenting to the emergency department (ED) of four hospitals of Hamad Medical Corporation (HMC), Doha, Qatar, between January 2015 and August 2021. HMC is Qatar's main tertiary healthcare provider and one of the leading healthcare systems in the Middle East.
## Inclusion criteria
All patients fulfilling the criteria of DKA were included in the study. We used American Diabetes Association (ADA) criteria for establishing the diagnosis of DKA, i.e., blood glucose greater than 250 mg/dL, high anion gap metabolic acidosis (pH < 7.3, bicarbonate < 18 mmol/L, anion gap > 10 mmol/L) and ketonemia/ketonuria13. Patients fulfilling the DKA diagnostic criteria were further assessed for the presence of concomitant pancreatitis using the Atlanta criteria, which includes the presence of two of the following three features: [1] Persistent upper abdominal pain, [2] Serum amylase or lipase level at least three times the upper limit of normal [3] Imaging evidence of AP14.
## Exclusion criteria
Patients aged less than 14 years, pregnant patients and patients with starvation or alcoholic ketosis were excluded from the analysis.
A total of 936 patients fulfilled the criteria of DKA and were included in the study (Fig. 1).Figure 1Flowchart of the process of inclusion of patients in the study.
Data were manually extracted from electronic medical records (EMR) by the research study's members. Demographic data included age, gender, ethnicity, weight, height and co-morbid conditions. All laboratory results at admission and sequential DKA related laboratory results during hospital stay (venous pH, bicarbonate, anion gap, lactate, electrolytes) were also recorded. Data related to outcome included length of stay (LOS), duration of DKA, a requirement for ICU admission, in-hospital mortality, 30-day all-cause readmission, 6-month DKA recurrence and 12-month DKA recurrence.
To ensure adequate data quality, it was collected manually from EMR and subsequently reviewed again for quality by the primary investigator. There was no missing data in laboratory variables used to establish DKA diagnosis and in variables to establish clinical outcomes of DKA. No formal method to deal with the missing data was used for other variables that had missing data.
## Statistical analysis
Continuous variables were described as either means (± standard deviation) or median (interquartile range), and an independent t-test or Mann–Whitney U test was used for comparison as appropriate. Categorical variables were described as percentages, and comparisons were performed using the chi-square and Fisher's tests. Factors associated with the development of coexisting DKA and AP were assessed using univariate and multivariate logistic regression analysis and reported as odds ratio (OR) and $95\%$ confidence intervals (CI). A p-value of < 0.05 was considered significant. Stata version 17 was used for analysis.
## Ethical declaration
The study is original and is not under consideration for publication in another journal. The study has been approved by the Medical Research Centre (MRC) at Hamad Medical Corporation, Qatar. All methods were performed in accordance with the relevant guidelines and regulations and according to the principles laid down in the Declaration of Helsinki. All the authors reviewed and approved the final manuscript.
## Informed consent
Due to the nature of this retrospective study and the preserved anonymity of patients, a waiver of informed consent was obtained from the Medical Research Centre (MRC) at Hamad Medical Corporation, Qatar.
## Baseline characteristics of the study population
A total of 936 patients fulfilling the DKA criteria were included in the analysis, of which 84 ($9.0\%$) patients had concomitant AP. DKA patients with coexisting AP were older (Mean (SD) age of 35.3 ± 14.5 vs. 42.7 ± 12.8 years, $p \leq 0.001$) as compared to those without AP. Both groups had a predominance of males (male to female $61.2\%$ vs. $38.8\%$ in the DKA group; $79.8\%$ vs. $20.2\%$ in the DKA and AP group; $p \leq 0.001$). Notably, most DKA patients belonged to the Arab ethnicity ($54\%$). However, coexisting AP was most common in the Asian race ($65.47\%$, $p \leq 0.001$). More patients in the DKA with AP group had T2D ($75\%$, $p \leq 0.001$). No statistically significant difference between mean BMI, duration of DM and co-morbid conditions was observed between the two groups (Table 1).Table 1Comparison of baseline characteristics between DKA versus DKA with AP patient groups. Baseline characteristicsUnitsDKA alone [852]DKA + AP [84]Significance (p-value)Age, mean ± SDYears35.3 ± 14.542.7 ± 12.8< 0.001GenderN (%) Male521 (61.2)67 (79.8) Female331 (38.8)17 (20.2)0.001EthnicitiesN (%) Arab483 (56.7)23 (27.4)< 0.001 Asian253 (29.7)55 (65.5) Africans90 (10.6)3 (3.6) Others26 [3]3 (3.6)BMI Median (IQR)kg/m223.83 (20.4–28)25.1 (21.6–27.8)0.14Duration of DM, mean ± SDYears4.5 ± 6.73 ± 5.60.06DM diagnosisN (%) Total T1D461 (54.6)21 [25]< 0.001 Total T2D383 (45.4)63 [75]Co-morbiditiesN (%) Dyslipidemia119 (13.9)16 (19.2)0.20 Coronary artery disease47 (5.5)7 (8.3)0.20 Heart failure10 (1.2)3 (3.6)0.10 Chronic liver Disease29 (3.4)4 (4.8)0.34 Hypertension183 (21.5)18 (21.4)0.99 Retinopathy69 (8.1)10 (11.9)0.23 Nephropathy52 (6.1)9 (10.7)0.10DKA Diabetic ketoacidosis, AP acute pancreatitis, BMI body mass index, SD standard deviation, IQR interquartile range, DM diabetes mellitus, T1D type 1 diabetes, T2D type 2 diabetes.
## Laboratory investigation results
In comparison to patients with DKA alone, DKA patients with AP had a statistically significant higher mean anion gap (22.9 ± 6.8 vs. 24.4 ± 7.7 mEq/L, p 0.044), median white cell count (WBC) (11.6 (IQR 8.2–17) vs. 15.1 (IQR 10.6–20.6) × 103/µL; $p \leq 0.001$) median neutrophil to lymphocyte ratio (NLR) 6 (IQR 2.5–11.2) vs. 7.6 (IQR 4.2–12.7); p 0.01), mean hemoglobin (14.2 ± 2.1 vs. 15 ± 2.4 g/dL; p 0.001), median serum urea level (5.7 (IQR 4–8.1) vs. 7.8 (IQR 4.9–12.4) mmol/L; $p \leq 0.001$), median serum creatinine 85 (IQR 64–120) vs. 109 (IQR 78–152) µmol/L; $p \leq 0.001$), median total cholesterol 4.3 (IQR 3.3–5.5) vs. 4.7 (3.5–7) mmol/L; p 0.01) and median serum triglyceride (TG) level 1.8 (IQR 1.2–2.5) vs. 2.2 (1.4–5.3) mmol/L; p 0.006) at admission as compared to the patients without AP. DKA patients with AP had lower mean potassium level (3.4 ± 0.5 vs. 3.2 ± 0.6 mmol/L; p 0.001) and lower mean sodium (131.5 ± 4.6 vs. 130 ± 6.6 mmol/L; p 0.048) during the DKA episode. These patients also had a lower mean venous pH at admission (7.15 ± 0.13 vs. 7.09 ± 0.15; p 0.001), median venous pH at 6-h 7.26 (IQR 7.2–7.32) vs. 7.24 (IQR 7.15–7.29); p 0.03) and mean bicarbonate at 6-h (15 ± 5 vs. 13.7 ± 5 mmol/L; p 0.007) in comparison to the DKA only group. There was no statistically significant difference between hbA1c, beta-hydroxybutyrate (BHB) and lactate levels at presentation between the two groups (Table 2).Table 2Comparison of laboratory parameters between DKA versus DKA with AP patient groups. VariableNormal rangeDKA alone [852]DKA + AP [84]Significance (p-value)Blood glucose at admission, Median (IQR)4–11.1 mmol/L23.3 (18.4–29.8)25 (17.9–30.4)0.64Highest glucose during hospital stay, mean ± SD4–11.1 mmol/L12.5 ± 64142.1 ± 69.10.02HbA1c at admission (mean ± SD)< $6.5\%$12.1 ± 2.811.8 ± 2.50.41White cell count at admission, median (IQR)4–10 × 103/µL11.6 (8.2–17)15.1 (10.6–20.6)< 0.001Hemoglobin at admission, mean ± SD12–15 g/dL14.2 ± 2.115 ± 2.4)0.001NLR at admission, median (IQR)NA6 (2.5–11.2)7.6 (4.2–12.7)0.01Urea at admission, median (IQR)2.5–7.8 mmol/L5.7 (4–8.1)7.8 (4.9–12.4)< 0.001Creatinine at admission, median (IQR)44–80 µmol/L85 (64–120)109 (78–152)< 0.001Lowest sodium, mean ± SD135–145 mmol/L131.5 ± 4.6130 ± 6.60.048Lowest potassium, mean ± SD3.5–5.3 mmol/L3.4 ± 0.53.2 ± 0.60.001BHB at admission, median (IQR)0.03–0.3 mmol/L5.8 (4.5–7.3)6 (4.6–7.8)0.53Lactate at admission, median (IQR)0.5–2.2 mmol/L1.7 (1.1–2.8)1.9 (1.2–2.7)0.94Serum pH at admission, mean ± SDNA7.15 ± 0.137.09 ± 0.150.001Serum pH at 6-h, median (IQR)NA7.26 (7.2–7.32)7.24 (7.15–7.29)0.03Bicarbonate at admission, mean ± SD22–29 mmol/L11.1 ± 4.39.8 ± 5.70.058Bicarbonate at 6-h, mean ± SD22–29 mmol/L15 ± 513.7 ± 50.007Anion Gap at admission, mean ± SD< 10 mEq/L22.9 ± 6.824.4 ± 7.70.044Total cholesterol at admission, median (IQR)< 5.2 mmol/L4.3 (3.3–5.5)4.7 (3.5–7)0.01TG at admission, median (IQR)< 1.7 mmol/L1.8 (1.2–2.5)2.2 (1.4–5.3)0.006Adjusted calcium, mean ± SD2.2–2.5 mmol/L2.39 ± 0.142.45 ± 0.150.8Bilirubin, median (IQR)0–21 µmol/L11.9 (7–20)13 (9–22)0.08AST, Median (IQR)0–33 IU/L23 (16–37)32 (19–82)0.003ALT, Median (IQR)0–32 IU/L22 (15–36)30.5 (19–70)< 0.001DKA Diabetic ketoacidosis, AP acute pancreatitis, SD standard deviation, IQR interquartile range, hbA1c glycated hemoglobin, NLR neutrophil to lymphocyte ratio, BHB beta-hydroxybutyrate, LDL low-density lipoprotein, TG triglyceride, AST aspartate aminotransferase, ALT alanine aminotransferase.
## Clinical outcomes
DKA patients with AP had a longer median length of hospital stay (2.4 (IQR 1.07–4.47) vs. 4.8 (IQR 2.6–7.2) days; $p \leq 0.001$), longer median DKA duration (17 (IQR 10–27) vs. 27 (IQR 13.3–41) hours; $p \leq 0.001$) and a higher rate of ICU admission (23.1 vs. $39.3\%$; p 0.001) in comparison to the patients without concomitant AP. In-hospital mortality, 3-month all-cause readmission, 6-month and 12-month DKA recurrence were not statistically significant between the two groups (Table 3).Table 3Comparison of clinical outcomes between DKA vs DKA with AP patient groups. VariableUnitsDKA aloneDKA + APSignificanceLength of stay, median (IQR)Days2.4 (1.07–4.47)4.8 (2.6–7.2)< 0.001DKA duration, median (IQR)Hours17 (10–27)27 (13.3–41)< 0.001COVID-19 infectionN (%)10 (1.17)1 (1.2)0.1In-hospital mortalityN (%)6 (0.7)2 (2.4)0.10Need for admission to ICUN (%)197 (23.1)33 (39.3)0.0013-month readmission (all-cause)N (%)131 (15.4)8 (9.5)0.156-month DKA recurrenceN (%)67 (7.9)5 (5.9)0.5312-month DKA recurrenceN (%)58 (6.8)3 (3.6)0.25DKA Diabetic ketoacidosis, AP acute pancreatitis, IQR interquartile range, ICU intensive care unit.
Univariate logistic regression analysis (Table 4) showed age (OR 1.03 (1.01–1.04), $p \leq 0.001$) Asian ethnicity (OR 6.52 (1.95–20.98), p 0.002), male gender (OR 2.5 (1.44–4.33), p 0.001), T2D (OR 3.61(2.16–6.02), $p \leq 0.001$), admission WBC count(OR 1.04 (1.02–1.07), p 0.001), hb (OR 1.2 (1.08–1.35), p 0.001), urea (OR 1.04 (1.02–1.07), $p \leq 0.001$), creatinine (OR 1.003 (1.001–1.004), $p \leq 0.001$), potassium (OR 1.4 (1.09–1.8), p 0.009), venous pH at admission (OR 0.05 (0.01–0.23), $p \leq 0.001$), bicarbonate at admission (OR 0.94 (0.89–0.98), p 0.01), anion gap at admission (OR 1.03 (1.0006–1.064), p 0.045), total cholesterol (OR 1.15 (1.06–1.25), p 0.001), LDL (OR 1.19 (1.008–1.40), p 0.03) and TG (OR 1.04 (1.005–1.09), p 0.02) to be the statistically significant factors ($p \leq 0.05$) associated with concomitant AP in DKA patients. Multivariate logistic regression analysis (Table 5) revealed only age and total cholesterol level to be associated with the co-existence of DKA and AP.Table 4Univariate analysis for factors associated with the development of concomitant DKA with AP.Characteristics (N)Odds ratioSignificance (p-value)Confidence intervalUpperLowerAge1.032< 0.0011.011.04Asian ethnicity6.520.0021.9520.98Male gender2.500.0011.444.33T2D3.61< 0.0012.166.02White cell count at admission1.040.0011.021.07Hemoglobin at admission1.200.0011.081.35Urea at admission1.04< 0.0011.021.07Creatinine at admission1.003< 0.0011.0011.004Potassium at admission1.400.0091.091.80pH vein at admission0.05< 0.0010.010.23Bicarbonate at admission0.940.010.890.98Anion gap at admission1.030.0451.00061.064Total cholesterol1.150.0011.061.25TG level1.040.021.0051.09LDL level1.190.031.0081.40DKA Diabetic ketoacidosis, AP acute pancreatitis, LDL low-density lipoprotein, TG triglyceride. Table 5Odds ratios for factors associated with DKA with AP (multivariate logistic regression analysis).Characteristics (N)Odds ratioSignificance (p-value)Confidence intervalUpperLowerAge1.040.021.0031.07Cholesterol1.20.011.051.4DKA Diabetic ketoacidosis, AP acute pancreatitis. Odds ratios adjusted for gender, admission WBC count, creatinine, venous pH, bicarbonate, anion gap and TG levels.
## Discussion
This retrospective study compared the differences in demographics, biochemical and clinical outcomes of patients with concomitant DKA and AP to those with DKA alone. Patients with DKA and AP were older, predominantly Asians and had T2D. These patients had higher anion gap, WBC count, hb, NLR, urea, creatinine, total cholesterol and TG while a lower venous pH at admission than those with DKA alone ($p \leq 0.05$). LOS, DKA duration and ICU admission rate was also higher in patients with DKA and AP than in DKA alone. Age, Asian ethnicity, male gender, T2D, admission WBC count, hb, urea, creatinine, potassium, venous pH, bicarbonate, anion gap, total cholesterol, TGL and LDL level were significant factors associated with the development of coexisting DKA and AP in univariate logistic regression analysis. In multivariate logistic regression analysis, age and total cholesterol level were associated with concomitant DKA and AP.
Studies assessing the prevalence of acute pancreatitis in DKA patients are minimal. A prospective study of 100 DKA patients reported an $11\%$ prevalence of AP in DKA patients15. Ma et al. found the prevalence of AP in DKA patients to be $15.53\%$ in a cohort of patients with T2D only12. In our cohort consisting of both T1D and T2D patients, $9\%$ of patients with DKA had evidence of concurrent AP. In particular, a statistically significant difference between T1D and T2D groups was observed as $4.1\%$ ($\frac{21}{482}$) T1D patients, and $13.5\%$ ($\frac{63}{486}$) T2D patients with DKA had coexisting AP. This highlights the need for particular attention to the presence of this combination in T2D patients.
Early recognition of coexisting DKA and AP is of particular significance as it carries a higher risk of unfavorable outcomes. Madsen et al. described a case of a 27-year-old patient who had a delay in the diagnosis of coexisting DKA with AP and died within 36 h of initial presentation16. Ma et al. reported worse biochemical markers in patients with DKA and AP than in DKA alone. Patients with DKA and AP had a lower pH, higher anion gap and evidence of haemoconcentration (high hemoglobin and hematocrit)12. Nair et al. also reported a lower pH, higher anion gap and higher blood glucose levels in DKA patients with AP compared to patients without AP15. In addition to the similar results concerning the parameters mentioned above, our study also found a higher WBC count, urea and creatinine in patients with DKA and AP. NLR is used widely as a marker of acute stress and increases rapidly following any pathological condition (within 6 h). Normal NLR is between 1–2, and higher values are associated with pathological states, including inflammation and infections17,18. King et al. reported an NLR cut-off value of 7.45 or above as predictive of an increased risk of the requirement of intensive care support and death19. On the other hand, Liu et al. reported a cut-off of 3.13 or above as predictive of ICU support in patients with COVID-19 infection aged 50 years or above20. To our knowledge, NLR in patients with concomitant DKA with AP has not been studied earlier. Our study found a higher NLR (p-value 0.01) in DKA with AP patients compared to the DKA only group indicating increased severity of illness in these patients. A careful assessment of this simple and readily available marker can alert physicians to the possibility of underlying coexisting AP in DKA patients.
An important finding in this study is the greater percentage of T2D ($75\%$) patients in the DKA and AP groups. Noel et al. reported 2.83 fold increased risk of AP in patients with T2D than those without T2D21. A meta-analysis conducted by Alexandra et al. also concluded an increased risk of local and systemic complications of AP in patients with DM22. This is most likely because patients with T2D have multiple risk factors, including obesity and deranged lipid profile (especially TG levels), that increase the risk of AP. Further studies are needed to understand the relationship between T2D and the risk of AP development.
An interesting aspect of the study was finding a higher percentage of concomitant DKA and AP in patients of Asian ethnicity. Patients of Asian ethnicity comprised $33\%$ of the study population, but $65.5\%$ of patients in DKA with AP group were Asians. The univariate analysis also revealed a statistically significant association between Asian ethnicity and the development of AP with DKA with an odds ratio of 6.52, revealing a higher risk of developing this coexisting condition in this cohort of patients. This could be related to a higher prevalence of metabolic risk factors in the Asian population. Ooi et al. reported lower bicarbonate and higher lactate levels in Asian T2D patients with DKA compared to the White ethnicity23. Patients belonging to the Asian race develop metabolic complications at a lower BMI than other racial groups due to a higher percentage of fat at a lower BMI24,25. Obesity has also been identified as a risk factor for increased severity and local and systemic complications in AP26. An interplay of increased metabolic risk due to high body fat and DM might explain the increased risk of AP in Asians. However, the exact mechanism for developing an increased risk of AP in patients of Asian ethnic group needs to be further assessed in more extensive studies.
## Clinical relevance
Recognition of patients with DKA with AP is clinically relevant due to its impact on patient outcomes and the utilization of healthcare resources. Each DKA admission can cost between $10,000 to $284,00027. A longer LOS further considerably increases healthcare costs28. Two previous studies comparing DKA patients with and without AP revealed a higher ICU admission rate and a longer LOS in patients with concomitant DKA and AP12,15. Our study also confirmed these findings with a longer duration of DKA, a longer LOS and a higher risk of ICU admission in DKA with AP cohort compared to DKA only patients. Knowledge regarding worse clinical outcomes in these patients can help physicians identify and effectively manage these high-risk patients.
## Strengths and limitations
Our study has several strengths. To our knowledge, of the studies comparing DKA patients with coexisting acute pancreatitis to DKA alone, our study has the largest number of patients with concomitant DKA and acute pancreatitis. Furthermore, the inclusion of patients from multiple ethnic backgrounds adds further to the study's strengths. In addition, our study included both T1D and T2D patients with DKA as opposed to a previous study in which patients with T1D were excluded12. We strictly followed ADA criteria for DKA diagnosis and *Atlanta criteria* for AP diagnosis during manual data collection to include patients in the study, thus excluding patients who were wrongly coded as either DKA or AP in the EMR. This robust method of data gathering adds to the authenticity of the study.
Our study has some limitations as well. An important limitation is its retrospective design, due to which adjustment for confounders cannot be performed. Another limitation is the lack of comparison of symptoms between DKA patients with and without pancreatitis. However, DKA and AP have a significant overlap in symptoms, so a comparison of symptoms between DKA versus DKA and AP groups is less meaningful. Severity scores for AP were not calculated and compared as the study involved comparing DKA patients with and without AP. Whether there is any difference in treatment modalities between these two groups of DKA patients was not studied. Larger prospective studies are required to understand the differences between DKA patients with and without AP to provide a higher level of evidence.
## Conclusion
This study showed that patients with concomitant DKA and AP have more severe derangement in the laboratory markers of DKA severity, inflammation, kidney injury and metabolic profile, highlighting much higher severity of illness than patients with DKA alone. Furthermore, this combination also significantly impacts healthcare resources utilization as DKA patients with AP have a longer DKA duration, LOS and requirement for ICU support than DKA patients without AP. More awareness among physicians regarding this combination of life-threatening conditions can lead to early recognition and improved clinical outcomes in patients with DKA.
## References
1. Fazeli Farsani S, Brodovicz K, Soleymanlou N, Marquard J, Wissinger E, Maiese BA. **Incidence and prevalence of diabetic ketoacidosis (DKA) among adults with type 1 diabetes mellitus (T1D): a systematic literature review**. *BMJ Open* (2017) **7** e016587. DOI: 10.1136/bmjopen-2017-016587
2. Ooi E, Nash K, Rengarajan L, Melson E, Thomas L, Johnson A. **Clinical and biochemical profile of 786 sequential episodes of diabetic ketoacidosis in adults with type 1 and type 2 diabetes mellitus**. *BMJ Open Diabetes Res Care.* (2021) **9** e002451. DOI: 10.1136/bmjdrc-2021-002451
3. Davis TME, Davis W. **Incidence and associates of diabetic ketoacidosis in a community-based cohort: the Fremantle Diabetes Study Phase II**. *BMJ Open Diabetes Res. Care* (2020) **8** e000983. DOI: 10.1136/bmjdrc-2019-000983
4. Gibb FW, Teoh WL, Graham J, Lockman KA. **Risk of death following admission to a UK hospital with diabetic ketoacidosis**. *Diabetologia* (2016) **59** 2082-2087. DOI: 10.1007/s00125-016-4034-0
5. Xiao AY, Tan ML, Wu LM, Asrani VM, Windsor JA, Yadav D. **Global incidence and mortality of pancreatic diseases: A systematic review, meta-analysis, and meta-regression of population-based cohort studies**. *Lancet Gastroenterol. Hepatol.* (2016) **1** 45-55. DOI: 10.1016/S2468-1253(16)30004-8
6. Umpierrez G, Freire AX. **Abdominal pain in patients with hyperglycemic crises**. *J. Crit. Care.* (2002) **17** 63-67. DOI: 10.1053/jcrc.2002.33030
7. Kumar P, Sakwariya A, Sultania AR, Dabas R. **Hypertriglyceridemia-induced acute pancreatitis with diabetic ketoacidosis: A rare presentation of type 1 diabetes mellitus**. *J. Lab. Phys.* (2017) **9** 329-331
8. Kong MT, Nunes MP, Leong KF. **Diabetic ketoacidosis with acute severe hypertriglyceridaemia-induced pancreatitis as first presentation of type 2 diabetes**. *BMJ Case Rep.* (2021) **14** e239727. DOI: 10.1136/bcr-2020-239727
9. Yuan S, Liao J, Cai R, Xiong Y, Zhan H, Zheng Z. **Acute pancreatitis concomitant with diabetic ketoacidosis: A cohort from South China**. *J. Int. Med. Res.* (2020) **48** 300060520912128. DOI: 10.1177/0300060520912128
10. Wang Y, Attar BM, Hinami K, Jaiswal P, Yap JE, Jaiswal R. **Concurrent diabetic ketoacidosis in hypertriglyceridemia-induced pancreatitis: How does it affect the clinical course and severity scores?**. *Pancreas* (2017) **46** 1336-1340. DOI: 10.1097/MPA.0000000000000937
11. Fu Y, Liu X, Cui B, Wang C, Liu Z, Zhao B. **Clinical characteristics of concomitant diabetic ketoacidosis in type 2 diabetes patients with acute pancreatitis**. *Diabetes Metab. Syndr. Obes.* (2022) **15** 111-119. DOI: 10.2147/DMSO.S336619
12. Ma LP, Liu X, Cui BC, Liu Y, Wang C, Zhao B. **Diabetic ketoacidosis with acute pancreatitis in patients with type 2 diabetes in the emergency department: A retrospective study**. *Front. Med. (Lausanne).* (2022) **9** 813083. DOI: 10.3389/fmed.2022.813083
13. Kitabchi AE, Umpierrez GE, Miles JM, Fisher JN. **Hyperglycemic crises in adult patients with diabetes**. *Diabetes Care* (2009) **32** 1335-1343. DOI: 10.2337/dc09-9032
14. Oppenlander KM, Chadwick C, Do MR, Carman KD. **Acute pancreatitis: Rapid evidence review**. *Am. Fam. Phys.* (2022) **106** 44-50
15. Nair S, Yadav D, Pitchumoni CS. **Association of diabetic ketoacidosis and acute pancreatitis: Observations in 100 consecutive episodes of Dka**. *ACG.* (2000) **95** 2795-2800
16. Madsen KR. **Fatal hypertriglyceridemia, acute pancreatitis and diabetic ketoacidosis possibly induced by quetiapine**. *BMJ Case Rep.* (2014) **2014** bcr2013202039. DOI: 10.1136/bcr-2013-202039
17. Zahorec R. **Neutrophil-to-lymphocyte ratio, past, present and future perspectives**. *Bratisl Lek Listy.* (2021) **122** 474-488. PMID: 34161115
18. Abaza NM, El-Latif EMA, Gheita TA. **Clinical significance of neutrophil/lymphocyte ratio in patients with granulomatosis with polyangiitis**. *Reumatol. Clín. (Engl. Ed.).* (2019) **15** 363-367. DOI: 10.1016/j.reumae.2017.11.009
19. King AH, Mehkri O, Rajendram P, Wang X, Vachharajani V, Duggal A. **A high neutrophil-lymphocyte ratio is associated with increased morbidity and mortality in patients with coronavirus disease 2019**. *Crit. Care Explor.* (2021) **3** e0444. DOI: 10.1097/CCE.0000000000000444
20. Liu J, Liu Y, Xiang P, Pu L, Xiong H, Li C. **Neutrophil-to-lymphocyte ratio predicts critical illness patients with 2019 coronavirus disease in the early stage**. *J. Transl. Med.* (2020) **18** 206. DOI: 10.1186/s12967-020-02374-0
21. Noel RA, Braun DK, Patterson RE, Bloomgren GL. **Increased risk of acute pancreatitis and biliary disease observed in patients with type 2 diabetes: A retrospective cohort study**. *Diabetes Care* (2009) **32** 834-838. DOI: 10.2337/dc08-1755
22. Mikó A, Farkas N, Garami A, Szabó I, Vincze Á, Veres G. **Preexisting diabetes elevates risk of local and systemic complications in acute pancreatitis: Systematic review and meta-analysis**. *Pancreas* (2018) **47** 917-923. DOI: 10.1097/MPA.0000000000001122
23. Ooi E, Nash K, Rengarajan L, Melson E, Thomas L, Johnson A. **Clinical and biochemical profile of 786 sequential episodes of diabetic ketoacidosis in adults with type 1 and type 2 diabetes mellitus**. *BMJ Open Diabetes Res. Care* (2021) **9** e002451. DOI: 10.1136/bmjdrc-2021-002451
24. Yi SS, Kwon SC, Wyatt L, Islam N, Trinh-Shevrin C. **Weighing in on the hidden Asian American obesity epidemic**. *Prev. Med.* (2015) **73** 6-9. DOI: 10.1016/j.ypmed.2015.01.007
25. Mui P, Hill SE, Thorpe RJ. **Overweight and obesity differences across ethnically diverse subgroups of Asian American men**. *Am. J. Mens Health.* (2018) **12** 1958-1965. DOI: 10.1177/1557988318793259
26. Martínez J, Sánchez-Payá J, Palazón JM, Suazo-Barahona J, Robles-Díaz G, Pérez-Mateo M. **Is obesity a risk factor in acute pancreatitis? A meta-analysis**. *Pancreatology* (2004) **4** 42-48. DOI: 10.1159/000077025
27. Lyerla R, Johnson-Rabbett B, Shakally A, Magar R, Alameddine H, Fish L. **Recurrent DKA results in high societal costs—A retrospective study identifying social predictors of recurrence for potential future intervention**. *Clin. Diabetes Endocrinol.* (2021) **7** 13. DOI: 10.1186/s40842-021-00127-6
28. Cyganska M. **The impact factors on the hospital high length of stay outliers**. *Proc. Econ. Finance.* (2016) **39** 251-255. DOI: 10.1016/S2212-5671(16)30320-3
|
---
title: 'Future of myocardial infarction mortality in Iran: a scenario-based study'
authors:
- Gisoo Alizadeh
- Kamal Gholipour
- Maryam Kazemi Shishavan
- Reza Dehnavieh
- Salime Goharinejad
- Morteza Arab-Zozani
- Mohammad Farough Khosravi
- Rahim Khodayari-Zarnaq
journal: Journal of Health, Population, and Nutrition
year: 2023
pmcid: PMC10018627
doi: 10.1186/s41043-023-00356-8
license: CC BY 4.0
---
# Future of myocardial infarction mortality in Iran: a scenario-based study
## Abstract
This study defines futures myocardial infarction landscapes and proposes a few policy options to reduce the burden of cardiovascular diseases using the scenario development method. We identified the effective drivers of myocardial infarction by reviewing the literature and completed the returned list with “experts” opinions. The results were classified using the STEEP (Social, Technological, Environmental, Economic, and Political) framework. We plotted the critical uncertainties in a two-dimensional ranking of “effect” and “uncertainty” levels. Eleven drivers with uncertainty and high potential impact were selected and categorized into three groups: Political Development, Access to health services, and Self-Care. Scenarios were developed, and 3 scenarios (optimistic, pessimistic, and possible) were selected based on scoring. For each scenario, policy options were formulated. Utilizing the capacity of Non-Governmental Organizations and charities and strengthening restrictive and punitive legislation was chosen as policy options for addressing possible scenarios. Building infrastructure and improving prevention services, designing and regenerating curative infrastructure were selected as optimal strategies for addressing issues related to the optimistic scenario. Strengthening restrictive and punitive legislation related to community health and population empowerment were proposed as critical policy options for health improvement regarding the pessimistic scenario. Increasing people’s participation, strengthening infrastructure and punitive policies can be effective in Myocardial infarction mortality prevention policies in Iran.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s41043-023-00356-8.
## Introduction
Futures study is a scientific study in which scientists use a wide range of methods to analyze the past and the present and collect data to predict futures in the various fields of science and technology [1]. Recently, adopting futures studies for planning, decision-making, and policymaking has gained popularity in Iran [2]. There are various methods for conducting futures studies, the most famous of which are time series analysis, modeling, simulation, and scenario development [3]. Scenario development is commonly used in economics, studies of society, culture, and politics. Initially, this method is known as the primary tool used in information analysis.
Scientists use futures studies in health-related studies to explore various diseases. Admittedly, its benefits are indicated by a scoping review published in 2015. In this review, while authors were outlining the application of scenario development method in health-related studies, they marked the benefits of using this technique for health care planning and strategic decision-making in public health [4, 5]. The advantages of adopting the scenario development method are also evident in “The bioeconomy to 2030” project for the Organization for Economic Cooperation and Development countries. After depicting the developed scenarios in this project, final interests were supported using various policy options [6]. Also, in the study that it was published in 2007 to study the future of IS Organization in 2020. Four scenarios have been proposed based on differing assumptions about two drivers: the advances in the reliability of international telecommunications and the value placed on computerization in businesses and society [7]. Non-communicable diseases are important issues in the field of health that future studies can be used to improve prevention, treatment, and rehabilitation services. Due to the importance of non-communicable diseases, it is one of the fields that can be used in futures studies [8].
Non-communicable diseases are essential health issues, and futures studies can be used to improve prevention, treatment and rehabilitation [8]. This is because many non-communicable diseases, such as cardiovascular diseases, usually burden the working-age population and the elderly. Consequently, their overall consequences wreak their toll on almost everybody in the community and getting ready for the possible scenarios might have an underestimated benefits [9]. In this regard, indicated enough by bloom et al. between 2011 and 2030, diabetes, cancer, chronic lung disease, and cardiovascular disease will be accounted for about $30 trillion in economic loss [10], and the results of the modeling study showed that the burden of CVD would increase steeply in Iran over 2005–2025 [11]. Regretfully, about $80\%$ of CVD deaths occur in low- and middle-income countries where they are more exposed to risk factors such as tobacco, which lead to CVDs and other non-communicable diseases [12]. However, the burden of non-communicable diseases is not restricted to just low- and middle-income countries; a study in Ireland shows that the prevalence of being overweight and obese will increase by $89\%$ and $85\%$ in men and women by 2030, respectively, leading to an increase in the prevalence of cardiovascular disease, cancer and diabetes type 2 [13]. Another study in China reported that 2.9 to 5.7 million deaths could be prevented by reducing risk factors for cardiovascular diseases until 2030 [13]. Overall, the evidence on the necessity of implementing effective interventions aiming to reduce cardiovascular disease attributed risk factors in forthcoming years is abundant, which urges and justifies assessing Iran’s possible future cardiovascular-related scenarios and developing appropriate policy options [14, 15]. In order to reduce the burden of cardiovascular diseases, setting strategies for treating and preventing non-communicable diseases has become the main focus of the World Health Organization’s (WHO) Joint Prevention Programs [16]. Furthermore, align with The United Nations(UN), WHO calls for a $25\%$ reduction in mortality from non-communicable diseases in 30–70 years by 2025 [17]. This objective is expected to be achieved by the countries by setting nation-specific goals and action plans because different countries have different contexts and priorities and often have different healthcare provision and management systems [18]. Fortunately, *Iran is* among the first developing countries that pursued this objective and developed and published a document claiming 13 objectives for achieving the principal goal named “National Document for the Prevention and Control of Non-communicable Diseases and Related Risk Factors” [19]. The challenges of implementing the package of essential disease interventions in the Iranian health system were examined in 2015. The study reported that the reconstruction of the health system and encouraging cooperation between different governmental and non-governmental sectors and promoting health education could promote non-communicable disease intervention delivery [20]. Therefore, this study intended to use the scenario method to develop futures cardiovascular disease-related scenarios to illustrate possible situations, aiming to help policymakers make prompt or provident decisions to prevent increased myocardial infarction risk and provide country-specific policy options related to different scenarios.
## Methods
A comprehensive search was conducted to find relevant studies in MEDLINE, Scopus, Web of Science, and Google Scholar between 2002 and 2018. Besides, a manual search of these articles’ reference lists and bibliographies was executed to capture additional articles for consideration. The following keywords were used in all databases: “Key Factors,” “Strategy,” “Driver,” “Driving Force,” “Restraining Force,” “cardiovascular diseases,” “Heart Attack,” “Noncommunicable Diseases,” “Myocardial Infarctions,” “Control,” and “Prevention.” Fifteen experts from the fields of management policy and health economics [7], community and preventive medicine [2], epidemiology [3], and cardiology [3] participated in completing and promoting the list of identified drivers. The STEEP framework (Social, Economical, Political, Environmental and Technological) was used to analyze the items of the final list.
In the next step, we selected 18 pre-identified and well-known experts from all over the country and invited them to participate in the study via email, Skype calls, or in person. The purposeful sampling method was used, and experts with the most knowledge and experience in the prevention of CVD were selected. Then, using the final list of drivers, we prepared a questionnaire (Q1), sent them to the experts, and asked them to grade the drivers regarding their degree of importance and the degree of uncertainty in the range of 1 to 10. Admittedly, in this grading system, 1 indicates the driver with the lowest impact or the most uncertain one that might happen in futures and 10 indicates the highest impact that the driver could have regarding the cardiovascular event or the most certain driver that could influence the event. ( Additional file 1). After that, we plotted all of them in a two-dimensional matrix to extract the set of drivers with the highest impact and certainty (policy intervention), using their mean scores of impacts and uncertainty (Fig. 1).Fig. 1The identification of key uncertainties [21] In order to draw a cross-impact matrix, we made a second questionnaire (Q2) which entailed three groups of drivers categorized upon their similarities. Then, we asked the experts to assess the drivers’ one-to-one relationships as positive or negative influences relative to each other and determine the magnitude of the influence by giving scores ranging from − 3 to + 3. Finally, the Scenario Wizard software 4.31 was used to perform a cross-impact matrix balance analysis, in which the inconsistency coefficient was set on 2. Also, total impact score was used.
In the next step, we listed the most believable scenarios that were returned running Scenario Wizard software and illustrated a causal network determining the interaction of the drivers with the final situation in the scenario. After that, based on the illustrations, we narrated a most relatable story for each scenario and reviewed them and edited them several times. In order to determine the validity of the scenarios, we developed a third questionnaire (Q3) via which the experts were asked to grade the strength of the scenarios answering the following questions: “Is Believable?”, “ Is it challenging?”, “ Is it internally consistent?”, “ Is it relevant?”, and “is it well structured?”. The grading scale for determining the strength of each scenario was in the range of 1–10. The mean score ≥ 5 was considered acceptable in terms of the validity of the scenario.
In the next step, utilizing a fourth questionnaire (Q4), the experts were asked to select three of the all valid scenarios as “the most optimistic scenario,” “the most pessimistic scenario,” and “the most probable scenario.” Experts could assign “the most probable scenario” to the same scenario that might be determined as the most optimistic or the most pessimistic scenario. Additionally, they were asked to suggest and list strategic approaches to address the alleged scenarios if any of them might occur in the futures. After that, we used the analytical hierarchy process (AHP) to rank preferences of the experts for alternative approaches and weight their preferences according to four attributes listed as “effectiveness” in terms of increasing the overall population health, “political feasibility,” “economically beneficial,” and the “required budget” for implementing the proposed approach. Expert Choice 11.5 was used for AHP analysis.
## Results
The initial search resulted in 5332 articles in four databases. Then, after cleaning for duplicates and articles without full text, 274 original articles from the remaining 1852 were screened using titles and abstracts. The papers that met all the eligibility criteria were selected for deliberate review. After reading full texts, 64 articles were included as the prime source for identifying the drivers. Preliminary, 98 key drivers were identified from included studies; then, after eliminating duplicates and similar factors, 38 items were listed as key drivers, which then promoted to a 53-item list adding the suggested drivers by the experts (Fig. 2).Fig. 2Drivers identified in STEEP framework The dissemination of the factors determined by their average scores for impactfulness and uncertainty is demonstrated in Fig. 3. According to the plot, 11 items were determined as most certain and the most impactful drivers: 1. improved diet quality, 2. increased social support, 3. increase in diplomatic sanctions, 4. the inadequacy of skilled physicians, 5. improved health services accessibility, 6. a decline in the numbers of job opportunities, 7. improved political stability, 8. increased accessibility to recreational facilities, 9. increased polypharmacy, 10. economic depression, and 11. increased workplace pollutant exposures. Fig. 3The identification of key uncertainties The cross-impact matrix entailed all the eleven selected drivers classified into three groups based on their conceptual and semantic similarities, i.e., “political evolutions” (items 1 to 4), “health services accessibility” (items 5 to 7), and “self-care” (items 8 to 11) (Fig. 4).Fig. 4Factors with high uncertainty and high impact potential The cross-impact matrix balance analysis on one-to-one relationships of the data returned six scenarios, and subsequently, each received a careful and deliberate narration. From all the validation questionnaires (Q3) sent to the experts, 16 ($88.9\%$) were sent back. Table 1 summarizes the final validation results of the narrations in which the narrations entitled “Bird in a cage” and “Hope and Fear” received the highest validity scores. Table 1Consistent and believable scenariosCriteriaScenario 1*Scenario 2Scenario 3Scenario 4Scenario 5*Scenario 6*Believable5.685.96.035.807.407.59challenging6.836.86.786.957.397.49Internal compatibility7.247.076.966.817.527.5Relevant6.356.266.336.186.816.76Design7.966.857.747.187.717.21Total6.816.576.766.587.367.31Acceptable criteria > 5 > 5 > 5 > 5 > 5 > 5*Selective scenarios were underlined
The results of the fourth questionnaire (Q4), which was designed to determine “the most optimistic scenario,” “the most pessimistic scenario,” and “the most probable scenario,” revealed that the experts agreed upon the scenario (Table 2).Table 2Policy options for probabilistic, optimistic, and pessimistic scenarios based on priorityScenarioPolicy options based on priority“A gentle path” OptimisticImproving the infrastructure for providing prevention servicesDesign and regenerate infrastructure and programs to strengthen preventive prevention to reduce risk factors for cardiovascular disease in the communityUsing the power of the private and non-governmental sector to provide cardiovascular disease prevention care on the outskirts of cities and slumsUtilizing health marketing methods and strategies in the prevention of cardiovascular disease“Hope and Fear” PessimisticStrengthen health-related restrictive and punitive legislationEmpowering the community in order to improve the business situation and healthImproving the infrastructure for providing prevention servicesModeling social prescription in the prevention of cardiovascular disease“Bird in a cage” PossibleThe use of NGOs and charities in the prevention of cardiovascular diseaseStrengthen health-related restrictive and punitive legislationEmpowering the community in order to improve the business situation and healthImproving the infrastructure for providing prevention services
## Discussion
In this study, we examined cardiovascular precipitating drivers, within which 53 divers identified as influential in the occurrence of cardiovascular diseases. Overall, childhood and development circumstances, the quality of places where people reside, and social support seem to have immense importance among the attributing drivers. Hence, the prevalence, incidence, and quality of life of those affected with cardiovascular diseases are impacted by all the functions of health systems, and the exposure to the CVD’s risk factors changes with the governmental and sociocultural circumstances, all policies regarding promoting physical activity, developing recreational facilities, transportation, preserving and expanding urban green spaces, residential and workplace environments, and tobacco control influences the burden of the cardiovascular diseases in the communities[22].
Therefore, the interventions targeting the living conditions of the community members and urban designs aiming to reduce risk factors of CVDs should not be overlooked in designing and implementing preventing and controlling measures and making policies [23–29].
The application of futures studies in health policy researches enables policymakers to have a potentially holistic perspective view over future health outcomes. In this regard, Goharinezhad et al. stated that despite the aging population of Iran, the government do not prioritize building infrastructure for future health care coverage; therefore, the different scenarios illustrating futures of the quality of health services for the elderly and health outcomes would give them a chance to act wisely and give a better chance to the optimistic scenario to manifest [30]. Result of study in England was shown that coronary heart disease mortality reductions of up to $45\%$, accompanied by significant reductions in area deprivation mortality disparities, would be possible by implementing optimal preventive policies [31].
Various studies aimed to depict the futures of health conditions in Iran and risk factor reduction policies identified similar uncertainties for raising the probability of unfolding an optimistic scenario. Some of these uncertainties are political ideation overarching decision-making in the country, overall economy, environmental circumstances such as pollution or transportation, availability, and accessibility of quality health care and sufficient self-management within the population [29, 30, 32]. These uncertainties also composed the substantial basis of our scenarios, indicating that policymakers should pay deliberate attention to challenges that these items could bring about in the futures.
Like most upper-middle-income countries, Iran struggles with the burden of non-communicable diseases and has committed to achieving a $25\%$ reduction in premature mortality from NCDs by 2025 (current WHO target) [33]. Having an unstable economy, the health system and people need the philanthropic and targeted activities of the NGOs in addition to the governmental health-related actions. Therefore, a large part of the policies and action plans of these organizations should be aligned with the health-related objectives and risk reduction behaviors such as increasing health literacy and physical activity opportunities, adopting and implementing policies for ensuring the availability and accessibility of healthy and balanced diet and promoting other preventive services [34, 35]. Admittedly, the government should allocate adequate resources for research, education, and promoting preventive measures to increase the effectiveness of interventions targeting CVDs, collaborate with other sectors, and encourage population participation in health care decision-making. Supposedly, with the optimistic scenario in hand, such goals are achievable; however, futures *Iran is* far from fulfilling WHO’s 2025 targets in the pessimistic and probable scenarios.
The project FRESHER, whose objective was to represent alternative futures where the detection of emerging health scenarios is used to test futures research policies to tackle the burden of NCDs, deified four scenarios which overall were narrated based on various sociocultural, technological, economic, and environmental factors. The authors discussed that positive depictions could only be realized in a society where the resources are spread equitably, sustainable growth does not depend only on limited resources, and health care is more person-centered and community-based care, people age in a fruitful process, and in this urbanized societies, people comply with healthy lifestyles and engage in life-long learning opportunities and have sufficient health literacy [36]. As is stated, our scenarios have also emphasized socio-demographic trends, economic factors, and technology advancements in depicting a future with decreased CVD incidents. However, in the circumstances with an unstable economy, allegedly, any political changes could deviate each optimistic or pessimistic scenario from its original narrative; moreover, costly technology-based health services, which still imposes significant concerns to low- and middle-income countries, would paradoxically jeopardize equitable health care delivery. Zahedi et al. conducted a study aiming to forecast the role of Non-Governmental Organizations (NGOs) in the futures of the health system’s accountability in Iran. Similar to the scenarios portrayed in our study, empowering the communities to take hold of their health, encouraging community participation, and empowering the NGO’s to expand their activities are the main principles that would increase accountable health care delivery in futures Iran [34].
## Limitations
In futures studies, the determination of the uncertainties and composing scenarios depend on the logic and intuitions of those who participate as experts. Although we tried to use a comprehensive roster of experts, given that futures studies are new in our country and have the experts relatively unfamiliar with its methodology, we had difficulty finding experts to assess and speculate Iran’s futures of CVDs in the framework of futures studies. Furthermore, due to limitations that the COVID-19 pandemic wreaked, some of the supposed panels of experts were canceled, and the experts’ opinions were asked via email. Likewise, we could not assess the objectivity of the suggested policies and wrote our report based on the array of opinions collected from the experts.
## Conclusion
The results of this study have shed light on how futures of cardiovascular diseases in Iran might be in the upcoming years and by depicting influential drivers and connecting them to the CVD outcomes. Furthermore, these findings demonstrate that futures of CVDs in Iran strongly depend on social and economic factors that any reforms or alteration of policies or regulations could substantially change regarding the alleged scenarios. These insights would help policymakers to intentionally watch the impacting trends and attributing factors predisposing populations to burden CVDs, among which fostering healthy behaviors that are directly influenced by economic and political circumstances.
The optimistic scenario illustrates improving social and economic circumstances. In this scenario, increased quality care delivery and popularity of self-management within the communities decreases the burden of CVD events. Priorities are set on rectifying infrastructures for increasing prevention services and focusing on decreasing population attributable risk factors. This scenario seems to be promising concerning the fulfillment of the National non-communicable disease document’s objectives.
The pessimistic and the probable scenarios do not foresee any improvement in social and economic circumstances in futures and predict a rise in cardiovascular events. These scenarios encourage policymakers to adopt restrictive health policies and give greater scope to NGOs and philanthropists to promote preventive health measures in the communities. If these scenarios take place in futures, the burden of CVDs would be extensive. Its prevention depends on the government and the nation, requiring them to align their goals with increasing health literacy, promoting environmental conditions, and complying with a healthy lifestyle.
## Supplementary Information
Additional file 1. Experts’ characteristics
## References
1. Traxler J, Connor S, Hayes S, Jandrić P. **Futures studies, mobilities, and the postdigital condition: contention or complement**. *Postdigital Sci Educ* (2022.0) **4** 494-518. DOI: 10.1007/s42438-021-00245-5
2. Paya A, Shoraka H-RB. **Futures studies in Iran: learning through trial and error**. *Futures* (2010.0) **42** 484-495. DOI: 10.1016/j.futures.2009.11.033
3. Puglisi M. **The study of the futures: an overview of futures studies methodologies**. *Interdepend Between Agric Urban Confl Sustain Use Soil Water Bari CIHEAM Options Méditerr Série A Sémin Méditer* (2001.0) **44** 439-463
4. Zimmet P. **Globalization, coca-colonization and the chronic disease epidemic: Can the Doomsday scenario be averted?**. *J Intern Med* (2001.0) **249** 17-26. DOI: 10.1046/j.1365-2796.2001.00625.x
5. Vollmar HC, Ostermann T, Redaèlli M. **Using the scenario method in the context of health and health care–a scoping review**. *BMC Med Res Methodol* (2015.0) **15** 1-10. DOI: 10.1186/s12874-015-0083-1
6. Oborne M. **The bioeconomy to 2030: designing a policy agenda**. *Organ Econ Cooper Dev OECD Obs* (2010.0) **278** 35
7. Gray P, Hovav AZ. **The IS organization of the future: four scenarios for 2020**. *Inf Syst Manag* (2007.0) **24** 113-120. DOI: 10.1080/10580530701220967
8. Alizadeh G, Gholipour K, Dehnavieh R, JafarAbadi MA, Azmin M, Khanijahani A. **A Scenario-based modelling study of the prevention of myocardial infarction in Iran**. *Iran Red Crescent Med J* (2020.0) **22** 13-14
9. Ghaffar A, Reddy KS, Singhi M. **Burden of non-communicable diseases in South Asia**. *BMJ* (2004.0) **328** 807-810. DOI: 10.1136/bmj.328.7443.807
10. 10.Bloom DE, Cafiero E, Jané-Llopis E, Abrahams-Gessel S, Bloom LR, Fathima S, et al. The global economic burden of noncommunicable diseases. Program on the Global Demography of Aging; 2012.
11. Sadeghi M, Haghdoost AA, Bahrampour A, Dehghani M. **Modeling the burden of cardiovascular diseases in Iran from 2005 to 2025: the impact of demographic changes**. *Iran J Public Health* (2017.0) **46** 506. PMID: 28540267
12. Jagannathan R, Patel SA, Ali MK, Narayan KV. **Global updates on cardiovascular disease mortality trends and attribution of traditional risk factors**. *Curr DiabRep* (2019.0) **19** 1-12
13. Keaver L, Webber L, Dee A, Shiely F, Marsh T, Balanda K. **Application of the UK foresight obesity model in Ireland: the health and economic consequences of projected obesity trends in Ireland**. *PLoS ONE* (2013.0) **8** e79827. DOI: 10.1371/journal.pone.0079827
14. Fallahzadeh H, Saadati H, Keyghobadi N. **Estimating the prevalence and trends of obesity in Iran populations from 2000 to 2011: a meta-analysis study**. *SSU_Journals* (2017.0) **25** 681-689
15. Webber L, Kilpi F, Marsh T, Rtveladze K, Brown M, McPherson K. **High rates of obesity and non-communicable diseases predicted across Latin America**. *PLoS ONE* (2012.0) **7** e39589. DOI: 10.1371/journal.pone.0039589
16. Masoudkabir F, Sarrafzadegan N, Gotay C, Ignaszewski A, Krahn AD, Davis MK. **Cardiovascular disease and cancer: evidence for shared disease pathways and pharmacologic prevention**. *Atherosclerosis* (2017.0) **263** 343-351. DOI: 10.1016/j.atherosclerosis.2017.06.001
17. Sacco RL, Roth GA, Reddy KS, Arnett DK, Bonita R, Gaziano TA. **The heart of 25 by 25: achieving the goal of reducing global and regional premature deaths from cardiovascular diseases and stroke: a modeling study from the American Heart Association and World Heart Federation**. *Circulation* (2016.0) **133** e674-e690. DOI: 10.1161/CIR.0000000000000395
18. Peykari N, Hashemi H, Dinarvand R, Haji-Aghajani M, Malekzadeh R, Sadrolsadat A. **National action plan for non-communicable diseases prevention and control in Iran; a response to emerging epidemic**. *J Diabetes Metab Disord* (2017.0) **16** 1-7. DOI: 10.1186/s40200-017-0288-4
19. Ghazizadeh-Hashemi S, Larijani B. *National action plan for prevention and control of non communicable diseases and the related risk factors in the Islamic Republic of Iran, 2015–2025* (2015.0) 47-65
20. Etemad K, Heidari A, Panahi M, Lotfi M, Fallah F, Sadeghi S. **A challenges in implementing package of essential noncommunicable diseases interventions in iran’s healthcare system**. *J Health Res Community* (2016.0) **2** 32-43
21. Bierbooms JJ, Bongers I, van Oers HA. **A scenario analysis of the future residential requirements for people with mental health problems in Eindhoven**. *BMC Med Inform Decis Mak* (2011.0) **11** 1-12. DOI: 10.1186/1472-6947-11-1
22. Alizadeh G, Gholipour K, Azami-Aghdash S, Dehnavieh R, JafarAbadi MA, Azmin M. **Social, economic, technological, and environmental factors affecting cardiovascular diseases: a systematic review and thematic analysis**. *Int J Prev Med* (2022.0) **13** 78. PMID: 35706860
23. Pou SA, Tumas N, Soria DS, Ortiz P, del Pilar DM. **Large-scale societal factors and noncommunicable diseases: urbanization, poverty and mortality spatial patterns in Argentina**. *Appl Geogr* (2017.0) **86** 32-40. DOI: 10.1016/j.apgeog.2017.06.022
24. da Armstrong AC, Ladeia AMT, Marques J, da Armstrong DMFO, da Silva AML, da Morais JC. **Urbanization is associated with increased trends in cardiovascular mortality among indigenous populations: the PAI study**. *Arq bras de cardiol* (2018.0) **110** 240-245
25. Armstrong AC, Ladeia AMT, Marques J, Armstrong D, Silva A, Correia LCL. **Urbanization is associated to increased trends in cardiovascular mortality among indigenous populations**. *Circulation* (2016.0) **134** A20588-A
26. 26.Sørensen M, Pershagen G. Transportation noise linked to cardiovascular disease independent from air pollution. European Heart Journal. 2019.
27. Bauman A, Bull F. *Environmental correlates of physical activity and walking in adults and children: a review of reviews* (2007.0)
28. Alsheikh-Ali AA, Omar MI, Raal FJ, Rashed W, Hamoui O, Kane A. **Cardiovascular risk factor burden in Africa and the middle east: the Africa middle east cardiovascular epidemiological (ACE) study**. *PLoS ONE* (2014.0) **9** e102830. DOI: 10.1371/journal.pone.0102830
29. Nazari R. **Future study of Iran’s elderly sports using cross-impact matrix**. *SSU_Journals* (2018.0) **26** 1095-1109
30. Goharinezhad S, Maleki M, Baradaran HR, Ravaghi H. **A qualitative study of the current situation of elderly care in Iran: What can we do for the future?**. *Glob Health Action* (2016.0) **9** 32156. DOI: 10.3402/gha.v9.32156
31. Scholes S, Bajekal M, Norman P, O’Flaherty M, Hawkins N, Kivimäki M. **Quantifying policy options for reducing future coronary heart disease mortality in England: a modelling study**. *PLoS ONE* (2013.0) **8** e69935. DOI: 10.1371/journal.pone.0069935
32. 32.Alizadeh A, Shammaee A, Nazemi A, Ghadiri R. Biotechnology foresight: IRAN 2025.
33. 33.Organization WH. Global NCD target: reduce premature deaths from NCDs. World Health Organization; 2016.
34. Alizadeh G, Gholipour K, Khosravi MF, Khodayari-Zarnaq R. **Preventive community-based strategies of cardiovascular diseases in Iran: a multi-method study**. *Soc Work Public Health* (2020.0) **35** 177-186. DOI: 10.1080/19371918.2020.1764432
35. 35.Organization WH. Global action plan on physical activity 2018–2030: more active people for a healthier world: World Health Organization; 2019.
36. 36.Kreibich R, Oertel B, Wölk M, editors. Futures studies and future-oriented technology analysis principles, methodology and research questions. In: 1st Berlin symposium on internet and society; 2011.
|
---
title: 'Married women pre-marital HIV testing status in Ethiopia: Individual and community
level factor analysis'
authors:
- Molla Yigzaw Birhanu
- Daniel Bekele Ketema
- Melaku Desta
- Samuel Derbie Habtegiorgis
- Belayneh Mengist
- Alehegn Aderaw Alamneh
- Ayenew Negesse Abeje
- Eniyew Tegegne
- Aytenew Geremew Mengist
- Migbar Dessalegn
- Getamesay Molla Bekele
- Selamawit Shita Jemberie
journal: Frontiers in Medicine
year: 2023
pmcid: PMC10018750
doi: 10.3389/fmed.2023.913040
license: CC BY 4.0
---
# Married women pre-marital HIV testing status in Ethiopia: Individual and community level factor analysis
## Abstract
### Introduction
Marriage between serodiscordant individuals accounts for 65–$85\%$ of new infections. Pre-marital Human Immune Virus (HIV) testing opens the door for HIV infection prevention and control. There are no studies that have evaluated the coverage and factors influencing pre-marital HIV testing at the community level in Ethiopia.
### Methods
This study was conducted using 10,008 samples of data extracted from Ethiopian demographic and health surveys (EDHS), 2016. To identify individual and community level factors a multi-level binary logistic regression model was used. Among fitted models, “full” model was taken as the best model. To declare the presence or absence of significant association with pre-marital HIV testing, a p-value < 0.05 with confidence interval (CI) was used.
### Results
In Ethiopia, $21.4\%$ ($95\%$ CI: 20.6, $22.2\%$) of study participants had pre-marital HIV testing. Age 35–49 years (AOR = 0.25; $95\%$ CI: 0.09, 0.66), educated (AOR = 1.76; $95\%$ CI: 1.17, 2.79), rich (AOR = 1.95; $95\%$ CI: 1.13, 3.55), having media exposure (AOR = 1.54; $95\%$ CI: 1.30, 4.71), and high community level literacy (AOR = 0.38; $95\%$ CI: 0.22, 0.66) were factors significantly associated with pre-marital HIV testing.
### Conclusion
The low coverage of pre-marital HIV testing in *Ethiopia is* insufficient to have a significant influence on the HIV/Acquired Immune Deficiency Syndrome (AIDS) epidemic. Information dissemination to create awareness about human rights and public health implications of pre-marital HIV testing áre necessary while it is made mandatory.
## Background
The Human Immune Virus/Acquired Immune Deficiency Syndrome (HIV/AIDS) is a fatal viral disease that has spread worldwide [1, 2]. Women and girls account for approximately 750,000 of the 1.5 million new infections, and HIV affects 20 million women and girls out a total population of 36.0 million adults. Aside from the aforementioned, 4,200 adolescent girls and young women between the aged of 15 and 24 years are infected with HIV each week worldwide [3]. However, approximately $16\%$ (6 million people) still require HIV testing services [4]. In Africa, HIV screening rates range from 33.5 to $82.3\%$ (5–8). Sub-Saharan Africa, including Ethiopia, is home to more than three-quarters of HIV/AIDS patients, and one women are disproportionately affected [9].
Human Immune Virus testing is a public health program that focuses on screening to reduce the spread of HIV/AIDS in order to combat its impact on community and national economic output [10, 11], because HIV test is regarded as a critical entry point for HIV detection, care and treatment, prevention, and support services [6, 7]. It protects people who have been infected by an HIV seropositive partner as well as their infants from HIV infection [9]. Goal 3 of the Sustainable Development Goals (SDGs) focuses on “Good health and wellbeing,” with one of the key priorities being to end the HIV/AIDS epidemic by 2030. This study adds to the growing consensus that HIV/AIDS prevention and control remain critical agenda items [12]. According to the prior studies, HIV testing is the most cost-effective measure for the prevention and control of HIV transmission in Africa [13, 14]. The Ethiopian government has embraced pre-marital voluntary HIV counseling and testing as a key component of the country’s HIV/AIDS prevention and control efforts, providing prospective couples with the opportunity to know their HIV status before marriage [15].
A clinical trial investigation found that HIV sero-discordant couples account for around 65–$85\%$ of new infections acquired via a married/cohabiting partner [16]. HIV positive people in sero-discordant marriages endanger HIV negative spouses [17]. In the same spirit, a study conducted in Zambia and Rwanda discovered that sero-discordant couples are responsible for an estimated $50\%$ of new heterosexual HIV infections [18]. Urban residency [19], secondary and post-secondary education [20], women aged 25–34 years and 35 years and older [21], being rich in wealth index, divorced/widowed in marital status [22], drinking alcohol [23], and being unemployed [24] were the identified factors having a significant association with pre-marital HIV testing at individual level.
Knowing the pre-marital HIV testing status is critical not only for the HIV-uninfected individual, but also for the HIV-infected individual, in order to start antiretroviral prophylaxis before the immune system deteriorates and to exercise the right to marry and find a family. As a result, this multi-level analysis of determinants of pre-marital HIV testing status and associated factors among married women in Ethiopia was conducted.
## Study setting, period, and design
This study was conducted in Ethiopia using secondary data, extracted from the Ethiopian demographic and health survey (EDHS) 2016. Ethiopia is organized as a Federal Democratic Republic having nine regional states and two city administrations. It has a total of 1,100,000 km2, and its regional states are divided into zones, which are further subdivided into districts, which are further subdivided into kebeles, the lowest administrative divisions [25]. Ethiopia is the second most populous country in Africa, after Nigeria, with a population of about 112 million people (56,010, 000 females and 56, 069, 000 males in 2019) [26]. Ethiopian culture is diverse and generally organized along ethnolinguistic lines. There are over 80 ethnic groups in the country that speak different local languages (such as Amharic, Oromo, Tigrinya, and others), and *English is* the most commonly spoken foreign language and is taught in secondary schools and universities. According to the 2007 population and housing census (PHC), Ethiopia had 84,915 enumeration areas/clusters, 67,730 of which were rural clusters and 17,185 of which were urban clusters, with a total of 15,411, 559 households enumerated. An Enumeration Area (EA) is a geographical area with an average of 181 households. These EAs were served as a sampling frame for the 2016 EDHS survey, which was conducted in across nine regions and two administrative councils of Ethiopia from October 18, 2016 to June 27, 2016 [27]. This study was conducted from January 18, 2016 to June 27, 2016.
## Eligibility criteria
All married women who were registered in the EDHS, 2016 were included in this study and married women having incomplete registration were excluded.
## Data sources and sampling procedures
The data for this analysis were derived from EDHS 2016 and obtained from the measure DHS website at http://www.dhsprogram.com. With measure DHS’s permission, the data sets were downloaded in Stata format. The study participants were drawn using a stratified, two-stage cluster design, with EAs as primary sampling units and households as secondary sampling units. Each region was divided into urban and rural clusters, with a total of 21 strata. At each of the lower administrative levels, the proportional allocation was achieved by sorting the sampling frame within each sampling stratum before sample selection, according to administrative units at different levels, and by using a probability proportional to size selection at the first stage of sampling. In the first step, 645 EA (202 EAs in urban areas and 443 EAs in rural regions) were chosen with a probability proportionate to EA size from a total list of 84,915 EAs established for the 2007 PHC [28].
## Population
The source population contained of all married reproductive age women living in Ethiopia, the study population consisted of married reproductive age women found in the selected cluster, and the study units comprised of married reproductive age women in the reproductive age group found in the selected household.
## Dependent
Pre-marital HIV testing (yes/no) status of married women in the reproductive age group was the outcome/dependent variable of the study.
## Individual-level variables
Women’s ages, Educational achievement of mothers, wealth in the family, women’s occupational status, residence, media exposure.
## Community-level variables
We relied primarily on EA to convey aggregate-level statistics at the community level because the DHS did not collect them. As a result of aggregating individual level variables at the community (cluster) level, aggregate community-level variables were formed, and the aggregate variables were classed as low or high based on the distribution of the proportion values obtained for each cluster. Hence, media exposure, literacy, poverty, behavior, and residence were the community-level variables.
## Media use
If a respondent used any newspaper/magazine, radio, television, or internet, regardless of frequency levels, “nearly every day,” “at least once a week,” and “less than once a week” were recoded as “Yes,” while the response level “not at all” was recoded as “No.”
## Community-level media use
Was classified as high if the proportion was 50–$100\%$, and low if the proportion of women using media in the community was less than $50\%$.
## Community-level literacy
Community-level literacy was classified as high if the proportion of women with primary, secondary, and higher education was 50–100 percent, and low if the proportion was less than 50 percent.
## Community-level poverty
The proportion of women from the two lowest wealth quintiles in a specific community was classified as high if it was 50–100 percent, and low if it was less than 50 percent.
## Patient and public involvement
The study did not include any patients. Through the publication of study reports and open access journal articles, study findings are made available to participants and the general public. The study webpages included contact information for the research team in case anyone wanted to directly request publications.
## Data management and analysis
A multivariable multi-level analysis model was used to determine the fixed and random effects of covariates related with pre-marital HIV testing among married women in Ethiopia. We came up with four distinct models. The first model, which was an empty or unconditional model with no explanatory variables, was used to deconstruct the variance in the enumeration region (cluster). It was also employed in the fitting of a multi-level statistical application. The second model included variables at the individual level, while the third model included variables at the community level. Finally, the fourth model (Full model) took variables at both the individual and community levels into account. The statistical significance level was chosen at $p \leq 0.05.$ The Akaike Information Criterions (AIC) were used to choose the best model from among the four options. As a result, the lower value of the Akaike Information Criterion suggests a better model fit, which was the full model. The adjusted odds ratio (AOR) was used to report the results of fixed effects (measures of association), along with $95\%$ confidence intervals (CI) and p-value < 0.05. STATA version 17.0 software was used for data management and analysis.
## Ethical consideration
We used population-based secondary records from the public domain/online in this investigation. Measure DHS has given the authors permission to use the data. The DHS program adheres to criteria that protect respondents’ privacy. Data accessed from measure DHS database at http://dhsprogram.com/data/availabledatasets.cfm. We obeyed the terms and conditions of data sharing policy; data kept confidential, used for the current study only.
## Socio-demographic characteristics
As a weighted sample, 10,008 married reproductive-aged women were included in this study. About 3,007 ($30.05\%$) of the study participants in the study were between the ages of 25 and 29. In terms of income, approximately 3,740 ($37.73\%$) of the study participants were the poorest, and 4,927 ($49.24\%$) were uneducated. Furthermore, 60,609 ($60.55\%$) of them were not working in their occupation. Approximately 7,625 ($76.19\%$) of survey participants had no media exposure (Table 1).
**TABLE 1**
| Variables | Categories | Weighted frequency | Percentage |
| --- | --- | --- | --- |
| Age | 15–19 | 347 | 3.47 |
| | 20–24 | 2036 | 20.34 |
| | 25–29 | 3007 | 30.05 |
| | 30–34 | 2234 | 22.32 |
| | 35–39 | 1591 | 15.9 |
| | 40–44 | 605 | 6.05 |
| | 45–49 | 188 | 1.88 |
| Household wealth quintile | Poorest | 3776 | 37.73 |
| | Poorer | 1918 | 19.16 |
| | Middle | 1587 | 15.86 |
| | Richer | 1401 | 14.0 |
| | Richest | 1326 | 13.25 |
| Husband’s/partner level of education | No education | 4928 | 49.24 |
| | Primary | 3220 | 32.17 |
| | Secondary | 1015 | 10.14 |
| | Higher | 845 | 8.44 |
| Maternal level of education | No education | 6490 | 64.85 |
| | Primary | 2486 | 24.84 |
| | Secondary | 671 | 6.7 |
| | Higher | 361 | 3.61 |
| Occupation status of women | Not working | 6060 | 60.55 |
| | Working | 3948 | 39.45 |
| Media exposure | Yes | 2383 | 23.81 |
| | No | 7625 | 76.19 |
## From 2011 to 2016, community-level characteristics of married reproductive-age women in Ethiopia
The study included 645 clusters, 269 ($41.74\%$) of which had significant community poverty and 504 ($78.19\%$) of which had no media exposure. Furthermore, more than 436 ($67.60\%$) of them had a high level of community literacy, and 442 ($68.54\%$) of them lived in rural areas (Table 2).
**TABLE 2**
| Variables | Categories | Weighted frequency | Percentage |
| --- | --- | --- | --- |
| Poverty | Low | 374 | 58.26 |
| | High | 268 | 41.74 |
| Media exposure | Yes | 140 | 21.81 |
| | No | 502 | 78.19 |
| Literacy | Low | 208 | 32.4 |
| | High | 434 | 67.6 |
| Residency | Rural | 202 | 31.46 |
| | Urban | 440 | 68.54 |
## Pre-marital HIV tested
Pre-marital HIV testing was performed on 2,142 ($21.40\%$) of the study participants. When we examined at the experience of pre-marital HIV testing based on women’s educational status, we found that 783 (36.55) and 700 ($32.68\%$) of the women who underwent pre-marital HIV testing had primary education and 32.68 percent had no education, respectively (Figure 1). In terms of community media exposure, around $55\%$ of pre-marital HIV tests had community level media exposure (Table 3).
**FIGURE 1:** *Pre-martial HIV testing over the educational status of married reproductive aged women in Ethiopia.* TABLE_PLACEHOLDER:TABLE 3
## Model selection and comparisons
There was high heterogeneity between clusters for pre-marital HIV testing, accounting for $57.89\%$ of the overall variation. Pre-marital HIV testing varies due to married women living in different clusters. As a result, we decides to model the data using a nested structure. To account for the presence of nested data structures, a multi-level binary logistic regression model was considered. Table 4 displays the outcomes of the four models developed and compared using AIC.
**TABLE 4**
| Comparison criteria | Model I | Model II | Model III | Model IV |
| --- | --- | --- | --- | --- |
| AIC | 8257.524 | 641.4457 | 7401.721 | 518.5607 |
Based on the data in Table 4, model 4, is the final model that includes both individual and community level variables with random effects, is most suited for interpretation. The statistically substantial difference in model fit ($p \leq 0.001$) and lower values in Model 4 provide additional statistical evidence for the suitability of the multi-level binary logistic regression model. According to the full model shown below the table, community-level variables (high community literacy, presence of community media exposure), and individual-level variables (being between the ages of 35 and 49, being educated, and being rich in wealth index) were significantly associated with pre-marital HIV testing at P-value < 0.05 as cutoff of point.
## Factors associated with pre-marital HIV testing
The chance of experiencing pre-marital HIV testing was 1.76 (AOR = 1.76; 95 percent CI: 1.17, 2.79) times more likely in educated married reproductive age women than in uneducated married reproductive aged women. Rich women are 1.95 (AOR = 1.95; 95 percent CI: 1.13, 3.55) times more likely than low-income women to get tested for HIV before marriage. women who have been exposed to the media are 1.54 (AOR = 1.54; $95\%$ CI: 1.30, 4.71) times more likely to have HIV test prior to marriage. Furthermore, communities with high level of literacy are 0.62 (AOR = 0.38; 95 percent CI: 0.22, 0.66) times less likely to have HIV tested before marriage (Table 5).
**TABLE 5**
| Variable | Categories | AOR (95% CI) | AOR (95% CI).1 | AOR (95% CI).2 | AOR (95% CI).3 |
| --- | --- | --- | --- | --- | --- |
| | | Model I | Model II | Model III | Model IV |
| Age | 15–19 | | 1.00 | | 1.00 |
| | 20–24 | | 1.33 (0.96, 1.82) | | 1.45 (0.86, 1.84) |
| | 25–29 | | 0.9 (0.67, 1.35) | | 0.70 (0.26, 1.85) |
| | 30–34 | | 0.78 (0.78, 1.01) | | 0.60 (0.23, 0.55) |
| | 35–49 | | 0.36 (0.26, 0.51)* | | 0.25 (0.09, 0.66)* |
| Occupation | Not working | | 1.00 | | 1.00 |
| | Working | | 1.14 (0.98, 1.34 | | 1.20 (0.80, 1.80) |
| Chat chewing | No | | 1.00 | | 1.00 |
| | Yes | | 1.55 (1.21, 1.99)* | | 1.82 (0.94, 3.24) |
| Smoking | No | | 1.00 | | 1.00 |
| | Yes | | 0.82 (0.35, 1.90) | | 2.54 (2.12, 3.98) |
| Wealth index | Poor | | 1.00 | | 1.00 |
| | Rich | | 0.98 (0.34, 1.34) | | 1.95 (1.13, 3.55)* |
| Residence | Urban | | 1.00 | 1.00 | 1.00 |
| | Rural | | 0.17 (0.13, 0.23)* | 3.54 (2.91, 4.00)* | 0.69 (0.40, 1.21) |
| Community media exposure | No | | 1.0 | 1.00 | 1.00 |
| | Yes | | 1.74 (1.49, 2.02)* | 0.69 (5.03, 9.51)* | 1.54 (1.30, 2.58)* |
| Community-poverty | Low | | 1.00 | 1.00 | 1.00 |
| | High | | 1.93 (1.60, 2.31)* | 0.39 (2.85, 5.22)* | 0.76 (0.42, 1.40) |
| Community-literacy | Low | | 1.00 | 1.00 | 1.00 |
| | High | | 2.65 (2.26, 3.11)* | 3.4 (1.29, 9.50)* | 0.38 (0.22, 0.66)* |
| Community educational status | Not educated | ICC = 0.5789 | | 1.00 | 1.00 |
| | Educated | | | 2.39 (2.15, 4.23) | 1.76 (1.17, 2.79)* |
## Discussion
This study looked at both individual and community-level factors associated with pre-marital HIV testing among married reproductive-aged women in Ethiopia who had married within the previous 5 years of the survey. About $21.4\%$ ($95\%$ CI: 20.6, $22.2\%$) of women married within the 5 years of the survey in Ethiopia had had pre-marital HIV testing. This figure was lower than that found in Yining, china where $33.74\%$ of them underwent pre-marital HIV testing [29]. This is owing to the fact that in China, women have more media exposure than in Ethiopia, this high media exposure and less literacy rate in China encouraged people to get pre-marital HIV testing more than Ethiopian. That is why pre-marital HIV testing practice in Ethiopians was lower than in China.
The odds of experiencing pre-marital HIV testing among study participants aged 35–49 years was reduced by $75\%$ (AOR: 0.25, $95\%$ CI: 0.09, 0.66) compared to participants aged 15–19 years. It is consistent with the findings of an Iranian study on the acceptability of pre-marital HIV testing [30, 31]. This could be due to the fact that as women get older, they become less influenced by others and less sexually exposed than women between the ages of 15 and 19. As a result, they were less eager to undergo pre-marital HIV testing. Hence, health professionals should promote premarital HIV testing equally among reproductive age women in order to include all reproductive age groups, particularly those aged 35 to 49.
The chance of experiencing pre-marital HIV testing among educated married women was 1.79 (AOR: 1.79, $95\%$ CI: 1.17, 2.79) times more likely to have pre-marital HIV testing than those who were not educated. Prior research backs up this finding (32–34). This could be attributed to educated women having greater access to knowledge about pre-marital HIV testing, which allows them to know as it prevents HIV transmission from a positive to a negative spouse and mother to children, ultimately leading them to have a more positive attitude and experience pre-marital HIV testing. As a result, the minister of health should take the women’s educational backgrounds into account when promoting perimarital HIV testing among reproductive-age women in order to create awareness equity. Policymakers and programmers should prioritize uneducated women over educated women when it comes to creating work awareness about the importance of perimarital HIV testing.
The odds of pre-marital HIV testing was 1.95 (AOR: 1.95, 95 percent CI: 1.13, 3.55) times more likely among rich women as compared with poor women. This is supported by a study conducted in Cameroon which was entitled with “effect of sociodemographic and health seeking behaviors” (35–37). This could be attributed to rich women being less concerned about the cost of testing far from their home in order to avoid the stigma that is understood when one of the two is positive. As a result, wealthy women were more willing to have pre-marital HIV testing. As the government, in collaboration with non-governmental organizations, has made HIV testing payment free, it should cover the transportation costs for those who travel from afar for perimarital HIV testing to alleviate transportation issues. In collaboration with religious leaders, healthcare professionals should promote the importance of perimarital HIV testing and the associated expenses covered by the government and other stakeholders for the community at large through meeting and social media.
The odds of pre-marital HIV testing was 1.54 (AOR: 1.54, 95 percent CI: 2.30, 4.71) times more likely among women who were lived in a community with high media exposure as compared with women who were lived in a community with having low media exposure. This finding is consistent with the previous studies (38–40). This is because these women have a better chance of accessing information, which leads to pre-marital HIV testing by increasing awareness and, ultimately, acceptability of pre-marital HIV testing because media is one source of information. The government should make a concerted effort to increase the availability and accessibility of media exposure (such as local FM radio, newspapers, and television) to communities with limited media exposure, as this is very important for information dissemination to create awareness regarding perimarital HIV testing among reproductive age women.
The odds of pre-marital HIV testing for women lived in communities having high literacy was lowered by $62\%$ (AOR: 0.38, $95\%$ CI: 0.22, 0.66) as compared to those women lived in low literacy. This is supported by the study conducted entitled “Low Health Literacy Is associated with HIV test Acceptance” (41–43). This is because women who lived in high health literacy communities appear to be less willing to undergo pre-marital HIV testing due to their limited ability to absorb advice and information provided by healthcare workers to undertake pre-marital HIV testing when compared to women who lived in low health literacy communities. In addition to the foregoing, the minister of health incorporation of minister of education should improve the availability and accessibility of education for those communities having high literacy rates.
## Strength and limitation of the study
This study has provided a wealth of information, particularly on community-level factors influencing premarital HIV testing, which are critical for the prevention and control of HIV transmission between couples and from mother to fetus. This study relied on secondary data, which may have underestimated or overestimated the study participants’ pre-marital HIV testing status. Those scientific societies that will have the opportunity to read this research should consider this point.
## Conclusion and recommendations
The low coverage of pre-marital HIV testing among married reproductive age women in *Ethiopia is* insufficient to have any significant impact on the HIV/AIDS epidemic. Being between the ages of 35 and 49, and high levels of community literacy were characteristics that hampered pre-marital HIV testing, while individual factors including higher levels of education, wealth index and media exposure were factors that increased the likelihood of accepting pre-marital HIV testing. The use of different media to educate and raise awareness among all reproductive age groups may be encouraged. Governments and HIV program implementers may consider establishing guidelines and regulations for mandatory pre-marital HIV testing among all reproductive age groups, even while they recognize the human rights and privacy implications of mandatory pre-marital HIV testing and the fact that HIV can be acquired and transmitted to women post-marital as well.
## Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/supplementary material.
## Ethics statement
The studies involving human participants were reviewed and approved by the Major DHS. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
MB: conceptualization. MB, GB, AM, DK, and ANA: data curation. MB, DK, GB, AM, ANA, and SJ: formal analysis and methodology. MB, GB, AM, DK, and ANA: software. MB, GB, AM, SH, ANA, and SJ: supervision. MB, GB, AM, and SJ: validation. MB, DK, and SJ: visualization. MB, SJ, DK, GB, AM, and SH: writing original draft. MB, GB, AM, DK, SH, ANA, and SJ: writing review and editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. 1.PEPFAR.
Ethiopia country/regional operational plan (COP/ROP), strategic direction summary.
Geneva: UNAIDS (2021).. (2021)
2. Michelo C.. (2007)
3. 3.Joint United Nations Programme on Hiv and Aids [UNAIDS].
Global HIV & AIDS statistics–Fact sheet. (2020). Available online at: https://www.unaids.org/en/resources/fact-sheet (accessed March 20, 2022).. (2020)
4. 4.PEPFAR, Global AIDS.
Global HIV/AIDS overview. HIV.gov. (2020).. (2020)
5. Salima N, Leah E, Stephen L. **HIV testing among women of reproductive age exposed to intimate partner violence in Uganda.**. (2018) **11** 275-87. DOI: 10.2174/1874944501811010275
6. Muyunda B, Mee P, Todd J, Musonda P, Michelo C. **Estimating levels of HIV testing coverage and use in prevention of mother-to-child transmission among women of reproductive age in Zambia.**. (2018) **76**. DOI: 10.1186/s13690-018-0325-x
7. Muyunda B, Musonda P, Mee P, Todd J, Michelo C. **Educational attainment as a predictor of HIV testing uptake among women of child-bearing age: analysis of 2014 demographic and health survey in Zambia.**. (2018) **6**. DOI: 10.3389/fpubh.2018.00192
8. Nall A, Chenneville T, Rodriguez L, O’Brien J. **Factors affecting HIV testing among youth in Kenya.**. (2019) **16**. DOI: 10.3390/ijerph16081450
9. Painter T. **Voluntary counseling and testing for couples: a high-leverage intervention for HIV/AIDS prevention in sub-Saharan Africa.**. (2001) **53** 1397-411. DOI: 10.1016/s0277-9536(00)00427-5
10. Ojo K, Delaney M. **Economic and demograhic consequences of AIDS in Namibia: rapid assessment of the costs.**. (1997) **12** 315-26. DOI: 10.1002/(SICI)1099-1751(199710/12)12:4<315::AID-HPM492>3.0.CO;2-A
11. Beck E, Miners A, Tolley K. **The cost of HIV treatment and care.**. (2001) **19** 13-39. DOI: 10.2165/00019053-200119010-00002
12. 12.United Nations [UN].
Sustainable development goals: 17 goals to transform our world.
New York, NY: United Nations (2015).. (2015)
13. Van de Perre P. **HIV voluntary counselling and testing in community health services.**. (2000) **356** 86-7. DOI: 10.1016/S0140-6736(00)02462-4
14. Godfrey-Faussett P, Maher D, Mukadi Y, Nunn P, Perriëns J, Raviglione M. **How human immunodeficiency virus voluntary testing can contribute to tuberculosis control.**. (2002) **80** 939-45. PMID: 12571721
15. 15.Federal Ministry of Health Ethiopia.
Guidelines for HIV counseling and testing in Ethiopia.
Ethiopia: Federal HIV/AIDS Prevention and Control Office (2007).. (2007)
16. Campbell M, Mullins J, Hughes J, Celum C, Wong K, Raugi D. **Partners in prevention HSV/HIV transmission study team 2011. Viral linkage in HIV-1 seroconverters and their partners in an HIV-1 prevention clinical trial.**. (2011) **6**. DOI: 10.1371/journal.pone.0016986
17. Matovu J. **Preventing HIV transmission in married and cohabiting HIV-discordant couples in sub-Saharan Africa through combination prevention.**. (2010) **8** 430-40. DOI: 10.2174/157016210793499303
18. Dunkle K, Stephenson R, Karita E, Chomba E, Kayitenkore K, Vwalika C. **New heterosexually transmitted HIV infections in married or cohabiting couples in urban Zambia and Rwanda: an analysis of survey and clinical data.**. (2008) **371** 2183-91. DOI: 10.1016/S0140-6736(08)60953-8
19. Misiri H, Muula A. **Attitudes towards premarital testing on human immunodeficiency virus infection among Malawians.**. (2004) **45** 84-7. PMID: 14968460
20. Ayiga N, Nambooze H, Nalugo S, Kaye D, Katamba A. **The impact of HIV/AIDS stigma on HIV counseling and testing in a high HIV prevalence population in Uganda.**. (2013) **13** 278-86. DOI: 10.4314/ahs.v13i2.12
21. Fanta W, Worku A. **Determinants for refusal of HIV testing among women attending for antenatal care in Gambella Region, Ethiopia.**. (2012) **9**. DOI: 10.1186/1742-4755-9-8
22. Desta W, Sinishaw M, Bizuneh K. **Factors affecting utilization of voluntary HIV counseling and testing services among teachers in Awi zone, Northwest Ethiopia.**. (2017) **2017**. DOI: 10.1155/2017/9034282
23. Bekele Y, Fekadu G. **Factors associated with HIV testing among young females; further analysis of the 2016 Ethiopian demographic and health survey data.**. (2020) **15**. DOI: 10.1371/journal.pone.0228783
24. Akoku D, Tihnje M, Tarh E, Tarkang E, Mbu R. **Predictors of willingness to accept pre-marital HIV testing and intention to sero-sort marital partners; risks and consequences: findings from a population-based study in Cameroon.**. (2018) **13**. DOI: 10.1371/journal.pone.0208890
25. 25.Central Statistical Agency [CSA].
Population projections for Ethiopia.
Addis ABaba: Central Statistical Agency (2013).. (2013)
26. 26.United Nations [UN].
The world population prospects: 2015 revision.
New York, NY: United Nations (2015).. (2015)
27. 27.Central Statistical Agency [CSA], ICF.
Ethiopia demographic and health survey 2016.
Addis Ababa: CSA (2016).. (2016)
28. Ng M, Fleming T, Robinson M, Thomson B, Graetz N, Margono C. **Global, regional, and national prevalence of overweight and obesity in children and adults during 1980–2013: a systematic analysis for the global burden of disease study 2013.**. (2014) **384** 766-81. DOI: 10.1016/S0140-6736(14)60460-8
29. Mao Y, Zheng X, Re Z, Pan C, Guli R, Song J. **An epidemiological study on sexual transmission of human immunodeficiency virus among pre-marital group in Yining city, Xinjiang.**. (2004) **25** 322-4. PMID: 15231201
30. Ayatollahi J, Sarab M, Sharifi M, Shahcheraghi S. **Acceptability of HIV/AIDS testing among pre-marital couples in Iran (2012).**. (2014) **55**. DOI: 10.4103/0300-1652.137188
31. Lekas H, Schrimshaw E, Siegel K. **Pathways to HIV testing among adults aged fifty and older with HIV/AIDS.**. (2005) **17** 674-87. DOI: 10.1080/09540120412331336670
32. Arulogun O, Adefioye O. **Attitude towards mandatory pre-marital HIV testing among unmarried youths in Ibadan northwest local government area, Nigeria.**. (2010) **14** 83-94. PMID: 20695141
33. Shahcheraghi S, Ayatollahi J, BagheriNasabSarab M, Akhondi R, Bafghi A. **Premarital couple’s opinions about prevention and treatment of aids in Yazd, Iran.**. (2015) **2** 557-63
34. Beitz J. **Education for health promotion and disease prevention: convince them, don’t confuse them.**. (1998) **44(Suppl. 3A)** 71S-6S. PMID: 9626000
35. Li X, Deng L, Yang H, Wang H. **Effect of socioeconomic status on the healthcare-seeking behavior of migrant workers in China.**. (2020) **15**. DOI: 10.1371/journal.pone.0237867
36. Habte D, Deyessa N, Davey G. **Assessment of the utilization of pre-marital HIV testing services and Shabbir Ismael determinants of VCT in Addis Ababa, 2003.**. (2006) **20** 18-23. DOI: 10.4314/ejhd.v20i1.10014
37. Godlonton S, Thornton R.. (2015)
38. Kim J, Jung M. **Associations between media use and health information-seeking behavior on vaccinations in South Korea.**. (2017) **17**. DOI: 10.1186/s12889-017-4721-x
39. Hogue M, Doran E, Henry DA. **A prompt to the web: the media and health information seeking behaviour.**. (2012) **7**. DOI: 10.1371/journal.pone.0034314
40. Soroya S, Farooq A, Mahmood K, Isoaho J, Zara S. **From information seeking to information avoidance: understanding the health information behavior during a global health crisis.**. (2021) **58**. DOI: 10.1016/j.ipm.2020.102440
41. Barragán M, Hicks G, Williams M, Franco-Paredes C, Duffus W, Del Rio C. **Low health literacy is associated with HIV test acceptance.**. (2005) **20** 422-5. DOI: 10.1111/j.1525-1497.2005.40128.x
42. Merchant R, Liu T, Clark M, Carey M. **Facilitating HIV/AIDS and HIV testing literacy for emergency department patients: a randomized, controlled, trial.**. (2018) **18**. DOI: 10.1186/s12873-018-0172-7
43. Hepburn M. **Health literacy, conceptual analysis for disease prevention.**. (2012) **4** 228-38
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.