text
stringlengths 254
1.16M
|
|---|
---
title: Cardiovascular outcomes with SGLT2 inhibitors versus DPP4 inhibitors and GLP-1
receptor agonists in patients with heart failure with reduced and preserved ejection
fraction
authors:
- Jimmy Gonzalez
- Benjamin A. Bates
- Soko Setoguchi
- Tobias Gerhard
- Chintan V. Dave
journal: Cardiovascular Diabetology
year: 2023
pmcid: PMC9999503
doi: 10.1186/s12933-023-01784-w
license: CC BY 4.0
---
# Cardiovascular outcomes with SGLT2 inhibitors versus DPP4 inhibitors and GLP-1 receptor agonists in patients with heart failure with reduced and preserved ejection fraction
## Abstract
### Background
No study has compared the cardiovascular outcomes for sodium–glucose cotransporter-2 inhibitors (SGLT2i) head-to-head against other glucose-lowering therapies, including dipeptidyl peptidase 4 inhibitor (DDP4i) or glucagon-like peptide-1 receptor agonist (GLP-1RA)—which also have cardiovascular benefits—in patients with heart failure with reduced (HFrEF) or preserved (HFpEF) ejection fraction.
### Methods
Medicare fee-for-service data (2013–2019) were used to create four pair-wise comparison cohorts of type 2 diabetes patients with: (1a) HFrEF initiating SGLT2i versus DPP4i; (1b) HFrEF initiating SGLT2i versus GLP-1RA; (2a) HFpEF initiating SGLT2i versus DPP4i; and (2b) HFpEF initiating SGLT2i versus GLP-1RA. The primary outcomes were [1] hospitalization for heart failure (HHF) and [2] myocardial infarction (MI) or stroke hospitalizations. Adjusted hazards ratios (HR) and $95\%$ CIs were estimated using inverse probability of treatment weighting.
### Results
Among HFrEF patients, initiation of SGLT2i versus DPP4i (cohort 1a; $$n = 13$$,882) was associated with a lower risk of HHF (adjusted Hazard Ratio [HR ($95\%$ confidence interval)], 0.67 (0.63, 0.72) and MI or stroke (HR: 0.86 [0.75, 0.99]), and initiation of SGLT2i versus GLP-1RA (cohort 1b; $$n = 6951$$) was associated with lower risk of HHF (HR: 0.86 [0.79, 0.93]), but not MI or stroke (HR: 1.02 [0.85, 1.22]). Among HFpEF patients, initiation of SGLT2i versus DPP4i (cohort 2a; $$n = 17$$,493) was associated with lower risk of HHF (HR: 0.65 [0.61, 0.69]) but not MI or stroke (HR: 0.90 [0.79, 1.02]), and initiation of SGLT2i versus GLP-1RA (cohort 2b; $$n = 9053$$) was associated with lower risk of HHF (0.89 [0.83, 0.96]), but not MI or stroke (HR: 0.97 [0.83, 1.14]). Results were robust across range of secondary outcomes (e.g., all-cause mortality) and sensitivity analyses.
### Conclusions
Bias from residual confounding cannot be ruled out. Use of SGLT2i was associated with reduced risk of HHF against DPP4i and GLP-1RA, reduced risk of MI or stroke against DPP4i within the HFrEF subgroup, and comparable risk of MI or stroke against GLP-1RA. Notably, the magnitude of cardiovascular benefit conferred by SGLT2i was similar among patients with HFrEF and HFpEF.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12933-023-01784-w.
## Introduction
The epidemiological trends for the incidence and prevalence of type 2 diabetes (T2D), heart failure (HF), and their co-occurrence have continued to worsen in the US and globally [1]. Diabetes is present in nearly half of HF patients, and the prevalence of HF is estimated to be $20\%$ among T2D patients [2]. Compared to T2D alone, the co-existence of T2D and heart failure augurs a clinical course characterized by greater insulin resistance, accelerated progression of T2DM, and an elevated risk of cardiovascular events and mortality [3, 4]. Recently, large cardiovascular outcome trials (CVOTs) have demonstrated the efficacy of a newer medication class: sodium–glucose cotransporter 2 inhibitors (SGLT2i) in reducing the incidence of hospitalizations for heart failure (HHF) and major adverse cardiovascular events (MACE)—comprised of myocardial infarction (MI), ischemic stroke, and cardiovascular death [5, 6]. In recent CVOTs that were initially conducted among patients with heart failure with reduced ejection fraction (HFrEF), and subsequently among patients with heart failure with preserved ejection fraction (HFpEF), SGLT2i reduced the incidence of HHF by approximately $30\%$ and improved heart-failure specific endpoints such as Kansas City Cardiomyopathy Questionnaire scores and N-terminal pro b-type natriuretic peptide levels [7]. However, these trials did not assess MACE endpoints such as MI or stroke hospitalizations [8–10], which are major contributors to cardiovascular morbidity and mortality among HFrEF and HFpEF patients [11, 12].
Similar to SGLT2i, glucagon-like peptide-1 receptor agonists (GLP-1RA) are a newer medication class with demonstrated benefits on MACE ($12\%$ risk reduction against placebo) and HHF ($9\%$ risk reduction against placebo) [13], leading to some speculation that they may also exert beneficial cardiovascular effects in patients with HF [14]. Consensus recommendations exist preferencing use of SGLT2i for T2D patients with heart failure and GLP-1RA for T2D patients with atherosclerotic cardiovascular disease [15, 16], Currently, no prospective or observational study has directly compared the magnitude of cardiovascular benefits conferred by SGLT2i head-to-head against any second-line glucose lowering therapies, including GLP-1RA, in patients with either HFrEF or HFpEF.
Accordingly, in a cohort of older adults, who have the highest prevalence of T2D and HF of any age group [17, 18], this study aimed to assess the cardiovascular effectiveness of SGLT2i compared to DPP4i and GLP-1RA among patients with HFrEF and HFpEF.
## Methods
The study was approved by the Rutgers University Institutional Review Board, and the appropriate data use agreements were in place.
## Data sources
Study subjects were drawn from Medicare insurance claims, a US federal program that provides healthcare to US citizens over 65 years of age. More specifically, we utilized a $50\%$ random sample of Medicare fee-for-service beneficiaries enrolled in Part D from March 2013—coinciding with the approval of SGLT2i in the US—to December 2019. Data elements of interest included patient demographics, medical and pharmacy monthly enrollment status, inpatient and outpatient medical service use (International Classification of Disease [ICD], Ninth and Tenth Revisions; Current Procedural Terminology codes, Fourth Edition), and outpatient pharmacy dispensing data (drug name and strength, units dispensed, and days’ supply).
## Study population and exposure definition
Within the database, a separate cohort was created for each pairwise comparison of SGLT2i versus an alternative non-gliflozin class. Cohort membership required patients to be new users of the study medications of interest (defined as no use of the medications in the 365-day washout period preceding medication initiation), be older than 65 years of age at cohort entry and have no evidence of gestational or type 1 diabetes (T1D), cancer, end-stage renal disease, or human immunodeficiency virus infection. With the sole exception of heart failure phenotype (see below), all baseline covariates including eligibility criteria and patient characteristics were assessed in the 365 days prior to the date of medication initiation.
The study cohort was further restricted to patients with the presence of HHF with ICD codes corresponding to HFrEF (ICD-9: 428.2× or ICD-10: I50.2×) or HFpEF (ICD-9: 428.3 × or ICD-10: I50.3×) in either the first or second position of the inpatient discharge diagnosis using all available lookback. The positive predictive value for this approach for identifying patients with HFrEF is $72\%$ and $90\%$ using ejection fraction [EF] thresholds of ≤ $40\%$ and ≤ $50\%$, respectively, and $92\%$ for HFpEF for an EF threshold of > $50\%$ [19]. Patients with evidence of both or neither HF subtypes were excluded from analyses.
The study was comprised of four pairwise comparison cohorts, which included patients with: (1a) HFrEF initiating SGLT2i versus DPP4i; (1b) HFrEF initiating SGLT2i versus GLP-1RA; (2a) HFpEF initiating SGLT2i or DPP4i; and (2b) HFpEF initiating SGLT2i or GLP-1RA. For SGLT2i versus DPP4i comparisons, patients using combination empagliflozin–linagliptin therapy were excluded from analysis. Further, individuals initiating SGLT2i and the comparator on the same day were also excluded from analyses. Patients meeting the inclusion and exclusion criteria could contribute to each cohort only once, but the same patient could be included in more than one cohort.
## Follow-up and study end points
Separately for each study outcome, patients began contributing to follow-up time on the day after cohort entry (i.e., medication initiation) up until the first occurrence of one of the following: end of pharmacy or health care eligibility, medication discontinuation defined as 60-day gap in treatment, medication switching (e.g., patients in SGLT2i arm initiating non-gliflozin therapy and vice versa), end of study data (December 2019), or the occurrence of the outcome.
The two primary outcomes of interest were [1] hospitalization for heart failure (HHF) (positive predictive value [PPV]: > $90\%$) [20], and [2] MI (PPV = $94\%$) or stroke (PPV = $85\%$) hospitalizations [21, 22]. Analysis for each of the two primary outcomes was conducted independently of the other.
## Baseline covariates and inverse probability of treatment weighting
To mitigate risk of confounding, we assessed and adjusted for > 30 baseline covariates that were assessed in the 12-month period prior to and including the index date. These covariates included patient sociodemographics (e.g., age at medication initiation, biological sex, and race, calendar year), complications of diabetes (e.g., diabetic neuropathy, nephropathy, retinopathy), oral and injectable glucose lowering therapies (e.g., metformin, sulfonylureas, insulin), diagnosis of cardiovascular conditions (e.g., myocardial infarction, stroke, HF), and cardiovascular medication use (e.g., dispensing of β-blockers, loop diuretics, statins). Frailty status was ascertained using the claims based frailty index, and using a threshold of ≥ 0.25 to define frailty [23].
Propensity scores were estimated using a logistic regression that modelled the probability of initiating SGLT2i (exposure) versus a non-gliflozin medication (control) conditional on the baseline covariates. These propensity scores were then used to estimate stabilized inverse probability of treatment weights (IPTW) to account for imbalances in patient characteristics [24].
## Statistical analysis
We assessed the performance of propensity scores based IPTW to control for confounding by examining the distribution of baseline covariates prior and after IPT weighting, and using a threshold of $10\%$ in standardized difference as a metric for a meaningful imbalance [25]. Using an as-treated approach, where patients were censored on treatment discontinuation or switching, we estimated the rates of the primary outcomes among patients using SGLT2i (exposure) or non-gliflozin medications (control) by calculating the number of events and incidence rates (IRs). Adjusted incidence-rate differences (RD) and hazard ratios (HR) along with their $95\%$ confidence intervals (CIs) were modelled through weighted Cox and Poisson regressions respectively.
Sensitivity and secondary analyses were conducted to assess the robustness of the study findings. First, we examined several secondary outcomes including a composite of the two primary outcomes (i.e., HF, MI or stroke hospitalizations), as well as individually examined MI hospitalizations, stroke hospitalizations, and all-cause mortality. Second, we conducted sensitivity analyses varying exposure-related censoring criteria, where instead of censoring patients at the time of treatment switching or discontinuation, we carried the index exposure forward to mimic an intention-to-treat approach with a maximum follow up truncated to 2 years.
Third, as our primary definitions to identify HF subtypes prioritize positive predictive values at the possible cost of lowered sensitivity (i.e., under-detection of patients with HF), we also employed alternative-more sensitive-HF definitions to identify HFrEF and HFpEF patients. More specifically, we allowed patients to be included in the study if they had presence of relevant HF codes in [1] any position of the inpatient discharge diagnosis, or [2] any inpatient or outpatient diagnoses fields. Fourth, we conducted sensitivity analyses where we excluded patients with a recent hospitalization (i.e., 30-days prior to the index date). Finally, to assess impact of the study estimates across calendar time, we also estimated stratified results before and after 2016. Other eligibility criteria (e.g., no evidence of T1D) were similar for all cohorts. For all cohorts, pairwise comparisons, and sensitivity analyses, the propensity scores were re-estimated, and stabilized inverse probability of treatment weights were re-calculated. All analyses were performed using SAS 9.4 (SAS Institute Inc, Cary, NC).
## Results
Within the data, we identified 240,252 patients who had evidence of heart failure, type 2 diabetes, and had used glucose lowering therapies (see Additional file 1: Appendix Fig. S1 for the CONSORT flow diagram). After the criteria of new use were applied, there were 23,959 and 95,564 remaining initiators of SGLT2i and DPP4i, respectively, and 27,340 and 35,111 eligible initiations for SGLT2i and GLP-1RA, respectively. After requirements for pertinent heart failure hospitalizations and other exclusion criteria (e.g., T1D) were applied, the four pair-wise comparison cohorts were comprised of: (Cohort 1a) 13,882 HFrEF patients initiating SGLT2i versus DPP4i; (1b) 6951 HFrEF patients initiating SGLT2i versus GLP-1RA; (2a) 17,493 HFpEF patients initiating SGLT2i versus DPP4i; and (2b) 9053 HFpEF patients initiating SGLT2i versus GLP-1RA (see Additional file 1: Appendix Tables S1–S4 for information on unadjusted and adjusted baseline characteristics for all four cohorts).
Prior to IPT weighting, SGLT2i users differed from their non-gliflozin counterparts with respect to pertinent baseline characteristics (defined as a standardized difference > $10\%$). Regardless of HF subtype, patients initiating DPP4i (compared to SGLT2i) were older and more likely to be diagnosed with MI, peripheral vascular disease and renal insufficiency, while those initiating GLP-1RA (compared to SGLT2i) were more likely to be female, have previously used insulin, and have evidence of microvascular complications of diabetes.
After IPT weighting, baseline characteristics in all four cohorts were well balanced with no standardized difference exceeding $10\%$ (Tables 1 and 2). The most commonly observed guideline-directed medical therapy for HFrEF was beta blockers, followed by ACE inhibitors or ARBs, while approximately one-third of the patients were on aldosterone antagonists. Reasons for censoring were similar among the four study cohorts, and discontinuation of index exposure was the most prevalent contributor to censoring (Additional file 1: Appendix Table S5).Table 1Baseline characteristics after IPT weighting among patients with heart failure with reduced ejection fractionSGLT2i versus DPP4iSDbSGLT2i versus GLP-1RASDbSGLT2i DPP4i SGLT2i GLP-1RA($$n = 2503$$)($$n = 2503$$)a($$n = 3214$$)($$n = 3214$$)aSociodemographics Age, mean (SD)73.3 (10.1)73.8 (11.1)3.369.9 (10.9)69.9 (10.6)0.0 Male1326 (54.8)1438 (57.2)4.91973 (61.1)1955 (61.0)0.2 Race, White1754 (72.5)1795 (71.4)2.32372 (73.5)2350 (73.3)0.3 Race, Black357 (14.8)385 (15.3)1.6454 (14.1)448 (14.0)0.2 Other Race309 (12.8)332 (13.2)1.4402 (12.5)407 (12.7)0.7 Calendar year [2013, 2014, 2015]831 (34.3)895 (35.6)2.8601 (18.6)575 (17.9)1.7 Calendar year [2016, 2017]753 (31.1)775 (30.9)0.6961 (29.8)960 (30.0)0.4 Calendar year [2018, 2019]837 (34.6)842 (33.5)2.21666 (51.6)1670 (52.1)1.0Diabetes-related factors Metformin1240 (51.2)1247 (49.6)3.11586 (49.1)1582 (49.4)0.5 Sulfonylureas1034 (42.7)1026 (40.8)3.91273 (39.4)1265 (39.5)0.1 DPP4i962 (29.8)946 (29.5)0.6 GLP-1RA119 (4.9)122 (4.9)0.1 Insulin850 (35.1)833 (33.2)4.21698 (52.6)1685 (52.6)0.1 Thiazolidinediones105 (4.3)110 (4.4)0.1145 (4.5)146 (4.5)0.2 Diabetes, ocular complications343 (14.2)377 (15.0)2.4624 (19.3)630 (19.7)0.8 Diabetes, renal complications905 (37.4)950 (37.8)0.81326 (41.1)1336 (41.7)1.2 Diabetes, neurological complications925 (38.2)928 (36.9)2.61444 (44.7)1430 (44.6)0.2Other factors Frailty status612 (25.3)554 (22.1)7.6446 (13.8)435 (13.6)0.7 Myocardial infarction177 (7.3)214 (8.5)4.4189 (5.9)181 (5.6)0.9 Stroke46 (1.9)44 (1.8)1.138 (1.2)38 (1.2)0.1 Peripheral vascular disease832 (34.4)833 (33.2)2.6951 (29.5)938 (29.3)0.5 Other ischemic heart disease2008 (83.0)2100 (83.6)1.52633 (81.6)2615 (81.6)0.0 Renal insufficiency1454 (60.1)1519 (60.5)0.81713 (53.1)1711 (53.4)0.6 ACE inhibitors1226 (50.6)1272 (50.6)0.11572 (48.7)1553 (48.5)0.5 ARBs674 (27.8)693 (27.6)0.6936 (29.0)926 (28.9)0.2 Beta blockers2163 (89.4)2245 (89.3)0.12887 (89.5)2868 (89.5)0.1 Calcium channel blockers528 (21.8)606 (24.1)5.4676 (21.0)675 (21.1)0.3 Non-dihydropyridine CCB226 (9.3)224 (8.9)1.5238 (7.4)236 (7.4)0.1 Thiazide diuretics538 (22.2)516 (20.5)4.1633 (19.6)636 (19.8)0.6 Loop diuretics2041 (84.3)2118 (84.3)0.12687 (83.3)2663 (83.1)0.4 Aldosterone antagonists745 (30.8)799 (31.8)2.21128 (34.9)1119 (34.9)0.1 Digoxin486 (20.1)497 (19.8)0.7559 (17.3)557 (17.4)0.2 Hydralazine/isosorbide698 (28.8)692 (27.5)2.9829 (25.7)835 (26.1)0.9 Other HF medicationsc141 (5.8)139 (5.5)1.2320 (9.9)318 (9.9)0.0 Statins1919 (79.3)2020 (80.4)2.72711 (84.0)2698 (84.2)0.5 Anticoagulants1018 (42.1)1069 (42.6)1.01361 (42.2)1346 (42.0)0.3 Antiplatelets833 (34.4)824 (32.8)3.41113 (34.5)1112 (34.7)0.4SGLT2i: sodium–glucose cotransporter-2 inhibitor; DPP4i: Dipeptidyl peptidase 4 inhibitor; SD: standardized difference; IPT: Inverse probability of treatment; ACE: Angiotensin converting enzyme; ARB: Angiotensin II receptor blockers; CCB: calcium channel blocker; StD: Standard Deviation; HF: Heart Failure; GLP-1RA: glucagon-like peptide 1 receptor agonistaFor ease of interpretation, the denominator of the DPP4i and GLP-1RA arms are weighted down to the SGLT2i groupbStandardized differences greater than $10\%$ imply a meaningful difference in the patient characteristic. After IPTW weighting, there were no differences that exceeded this thresholdcInclude angiotensin receptor-neprilysin Inhibitors and hyperpolarization-activated cyclic nucleotide-gated (HCN) channel blockersTable 2Baseline characteristics after IPT weighting among patients with heart failure with preserved ejection fractionSGLT2i versus DPP4iSDbSGLT2i versus GLP-1RASDbSGLT2i DPP4i SGLT2i GLP-1RA($$n = 2846$$)($$n = 2846$$)a($$n = 3578$$)($$n = 3578$$)aSociodemographics Age, mean (SD)75.7 (10.1)76.0 (11.0)2.071.3 (10.9)71.4 (10.5)0.6 Male1030 (36.9)1069 (37.4)0.91489 (41.4)1487 (41.6)0.5 Race, White2039 (73.2)2108 (73.7)1.22692 (74.8)2667 (74.7)0.4 Race, Black403 (14.5)396 (13.8)1.8461 (12.8)459 (12.8)0.1 Other Race345 (12.4)356 (12.5)0.3445 (12.4)446 (12.5)0.4 Calendar year [2013, 2014, 2015]894 (32.1)939 (32.8)1.6696 (19.4)669 (18.7)1.6 Calendar year [2016, 2017]871 (31.2)896 (31.3)0.21089 (30.3)1082 (30.3)0.1 Calendar year [2018, 2019]1023 (36.7)1026 (35.9)1.71813 (50.4)1820 (51.0)1.2Diabetes-related factors Metformin1375 (49.3)1378 (48.2)2.31735 (48.2)1714 (48.0)0.4 Sulfonylureas1094 (39.2)1119 (39.1)0.21356 (37.7)1328 (37.2)1.0 DPP4i1033 (28.7)1028 (28.8)0.2 GLP-1RA154 (5.5)160 (5.6)0.3 Insulin998 (35.8)1026 (35.9)0.22034 (56.5)2003 (56.1)0.9 Thiazolidinediones161 (5.8)165 (5.8)0.1217 (6.0)212 (5.9)0.3 Diabetes, ocular complications433 (15.5)447 (15.6)0.3736 (20.5)743 (20.8)0.8 Diabetes, renal complications1144 (41.0)1164 (40.7)0.71605 (44.6)1608 (45.0)0.8 Diabetes, neurological complications1209 (43.4)1197 (41.9)3.11851 (51.4)1824 (51.1)0.7Cardiovascular factors Frailty status800 (28.7)777 (27.1)3.4683 (19.0)676 (18.9)0.1 Myocardial infarction216 (7.8)241 (8.4)2.4202 (5.6)210 (5.9)1.1 Stroke54 (1.9)59 (2.1)0.953 (1.5)56 (1.6)0.8 Peripheral vascular disease995 (35.7)972 (34.0)3.61103 (30.6)1103 (30.9)0.5 Other ischemic heart disease1905 (68.3)1985 (69.4)2.32410 (67.0)2422 (67.8)1.8 Renal insufficiency1753 (62.9)1785 (62.4)1.02052 (57.0)2052 (57.5)0.9 ACE inhibitors1101 (39.5)1103 (38.6)1.91424 (39.6)1400 (39.2)0.8 ARBs754 (27.0)821 (28.7)3.71107 (30.8)1088 (30.5)0.6 Beta blockers2181 (78.3)2249 (78.6)1.02759 (76.7)2763 (77.3)1.6 Calcium channel blockers1040 (37.3)1097 (38.3)2.11269 (35.3)1268 (35.5)0.5 Non-dihydropyridine CCB488 (17.5)482 (16.9)1.7510 (14.2)507 (14.2)0.0 Thiazide diuretics701 (25.2)694 (24.3)2.0930 (25.8)916 (25.6)0.5 Loop diuretics2351 (84.3)2386 (83.4)2.53017 (83.8)2994 (83.8)0.0 Aldosterone antagonists550 (19.7)562 (19.7)0.1787 (21.9)773 (21.6)0.6 Digoxin315 (11.3)335 (11.7)1.2309 (8.6)313 (8.8)0.6 Hydralazine/isosorbide764 (27.4)800 (28.0)1.3928 (25.8)939 (26.3)1.1 Other HF medicationsc24 (0.8)25 (0.9)0.451 (1.4)51 (1.4)0.0 Statins2065 (74.1)2166 (75.7)3.72869 (79.7)2855 (79.9)0.5 Anticoagulants1187 (42.6)1170 (40.9)3.41402 (38.9)1408 (39.4)1.0 Antiplatelets695 (24.9)715 (25.0)0.1921 (25.6)924 (25.9)0.6SGLT2i: sodium–glucose cotransporter-2 inhibitor; DPP4i: Dipeptidyl peptidase 4 inhibitor; SD: standardized difference; IPT: Inverse probability of treatment; ACE: Angiotensin converting enzyme; ARB: Angiotensin II receptor blockers; CCB: calcium channel blocker; StD: Standard Deviation; HF: Heart Failure; GLP-1RA: glucagon-like peptide 1 receptor agonistaFor ease of interpretation, the denominator of the DPP4i and GLP-1RA arms are weighted down to the SGLT2i groupbStandardized differences are expressed in percentage points. Values greater than $10\%$ imply a meaningful difference in the patient characteristic. After IPTW weighting, there were no differences that exceeded this thresholdcInclude angiotensin receptor-neprilysin Inhibitors and hyperpolarization-activated cyclic nucleotide-gated (HCN) channel blockers
## HFrEF analysis
For the SGLT2i versus DPP4i comparison (Cohort 1a) and over a median follow up of 6.5–7.1 months, there were 704 (incidence rate per 100 person-year [IR] = 52.2) HHF events in the SGLT2i group compared to 5567 (IR = 80.9) events in the DPP4i group (Table 3), and 144 (IR = 8.8) versus 1240 (IR = 12.7) MI or stroke events for SGLT2i versus DPP4i users. After adjustment, SGLT2i users had a lower risk of both HHF: HR 0.67 ($95\%$ CI 0.63, 0.72), and MI or stroke: HR 0.86 ($95\%$ CI 0.75, 0.99). This corresponded to an adjusted rate difference [RD] per 100 person-year of − 20.9 ($95\%$ CI − 16.1, − 25.7) and − 1.6 ($95\%$ CI − 3.7, − 0.0) for the outcomes of HHF and MI or stroke, respectively. Table 3Risk of cardiovascular hospitalizations among patients initiating SGLT2i compared to non-gliflozin therapies, by heart failure subtypeHFrEFUnadjustedIPTW adjustedIPTW adjusted HR ($95\%$ CI)No. events (IR)aRD ($95\%$ CI)SGLT2iDPP4i($$n = 2503$$)($$n = 11$$,379)Follow up, months: median (IQR)6.5 (5.1, 12.9)7.1 (4.0, 14.3)Heart failure hospitalizations704 (52.2)5567 (80.9)− 20.9 (− 16.1, − 25.7)0.67 (0.63, 0.72)MI or Stroke hospitalizations144 (8.8)1240 (12.7)− 1.6 (− 3.7, − 0.0)0.86 (0.75, 0.99)SGLT2iGLP-1RA($$n = 3214$$)($$n = 3737$$)Follow up, median (IQR)6.6 (4.3, 14.7)7.3 (4.1, 15.1)Heart failure hospitalizations893 (49.4)1316 (60.6)− 5.5 (− 10.0, − 1.0)0.86 (0.79, 0.93)MI or Stroke hospitalizations180 (8.4)282 (10.2)0.6 (− 1.4, 2.5)1.02 (0.85, 1.22)HFpEFUnadjustedIPTW adjustedIPTW adjusted HR ($95\%$ CI)No. events (IR)aRD ($95\%$ CI)SGLT2iDPP4i($$n = 2846$$)($$n = 14$$,647)Follow up, median (IQR)6.1 (4.0, 11.2)6.6 (4.0, 12.5)Heart failure hospitalizations804 (51.2)7132 (80.5)− 23.0 (− 14.5, − 31.4)0.65 (0.61, 0.69)MI or Stroke hospitalizations167 (8.8)1624 (12.7)− 1.0 (− 3.0, 1.1)0.90 (0.79, 1.02)SGLT2iGLP-1RA($$n = 3578$$)($$n = 5475$$)Follow up, median (IQR)6.5 (4.1, 12.8)7.1 (4.1, 13.6)Heart failure hospitalizations1059 (53.2)2004 (60.2)− 4.7 (− 8.6, − 0.7)0.89 (0.83, 0.96)MI or Stroke hospitalizations213 (8.8)443 (10.3)− 0.4 (− 2.1, 1.3)0.97 (0.83, 1.14)HFrEF: Heart failure with reduced ejection fraction; HFpEF: Heart failure with preserved ejection fraction; SGLT2i: sodium–glucose cotransporter-2 inhibitors; DPP4i: dipeptidyl peptidase 4 inhibitors; GLP-1RA: Glucagon-like peptide-1 receptor agonists; CI: confidence intervals; IR: incidence rate; HR: hazard ratio; IQR: Interquartile range; RD: Rate differenceaRepresent the unadjusted number of events and incidence rates per 100 person-years of follow upbHazard ratios were adjusted for variables described in Tables 1 and 2 using stabilized inverse probability of treatment weighting Meanwhile, for the SGLT2i versus GLP-1RA comparison (Cohort 1b), over a median follow up of 6.6–7.3 months, there were 893 (IR = 49.4) versus 1316 (IR = 60.6) HHF events, and 180 (IR = 8.4) versus 282 (IR = 10.2) MI or stroke events. After adjustment, SGLT2i use was associated with a lower risk of HHF: HR 0.86 ($95\%$ CI 0.79, 0.93) and RD − 5.5 ($95\%$ CI − 10.0,-1.0), but not MI or stroke, HR 1.02 ($95\%$ CI 0.85, 1.22) and RD 0.6 ($95\%$ CI − 1.4, 2.5).
## HFpEF analysis
For the SGLT2i versus DPP4i comparison (Cohort 2a), over a median follow up of 6.1–6.6 months, there were 804 (IR = 51.2) versus 7132 (IR = 80.5) HHF events, and 167 (IR = 8.8) versus 1624 (IR = 12.7) MI or stroke events (Table 3). After adjustment, SGLT2i users exhibited significant reductions in risk of HHF: HR 0.65 ($95\%$ CI 0.61, 0.69) and RD − 23.0 ($95\%$ CI − 14.5, − 31.4), and numerical decreases in MI or stroke that did not reach statistical significance: HR 0.90 ($95\%$ CI 0.79, 1.02) and RD − 1.0 ($95\%$ CI − 3.0, 1.1).
For the SGLT2i versus GLP-1RA comparison (Cohort 2b), over a median follow up of 6.5–7.1 months, there were 1059 (IR = 53.2) versus 2004 (IR = 60.2) HHF events, and 213 (8.8) versus 443 (10.3) MI or stroke events. After adjustment, SGLT2i use was associated with a lower risk of HHF: HR 0.89 ($95\%$ CI 0.83, 0.96) and RD − 4.7 ($95\%$ CI − 8.6, − 0.7), but not MI or stroke, HR 0.97 ($95\%$ CI 0.83, 1.14) and RD − 0.4 (− 2.1, 1.3).
Notably, the magnitude of reduction in cardiovascular outcomes conferred by SGLT2i appeared to be similar for both the HFrEF and HFpEF cohorts with no evidence of interaction, and all p-values for heterogeneity were > 0.05 for all hazard ratios and rate differences.
## Sensitivity and secondary analysis
SGLT2i use was associated with a reduced risk for the [1] endpoint comprised of MI, stroke or HF hospitalizations, and [2] all-cause mortality against DPP4i regardless of HF subtype and was associated with a significant reduction in MI hospitalizations in the HFrEF but not the HFpEF cohort (Table 4). However, SGLT2i and GLP-1RA were similar in terms of all non-HF related endpoints. Findings for the sensitivity analysis using an intention-to-treat approach were consistent with the primary analyses though closer to null, and non-statistically significant for MI or stroke hospitalizations for any HF subtype. Table 4Adjusted risk of cardiovascular outcomes among patients initiating SGLT2i compared to non-gliflozin therapies, by heart failure subtype: Sensitivity and secondary analysisHFrEFSGLT2i versus DPP4iSGLT2i versus GLP-1RAOther secondary outcomes MI hospitalizations0.81 (0.69, 0.95)1.01 (0.83, 1.23) Stroke hospitalizations1.02 (0.74, 1.41)0.97 (0.63, 1.50) All-cause mortality0.39 (0.34, 0.46)0.86 (0.72, 1.03)MI, stroke or HF hospitalizations0.88 (0.82, 0.95)0.98 (0.90, 1.06)Intention to treat analyses HF hospitalizations0.79 (0.74, 0.83)0.89 (0.84, 0.95) MI or stroke hospitalizations1.09 (1.00, 1.20)0.99 (0.88, 1.12)HFpEFSGLT2i versus DPP4iSGLT2i versus GLP-1RAOther secondary outcomes MI hospitalizations0.88 (0.76, 1.02)0.96 (0.80, 1.14) Stroke hospitalizations0.90 (0.68, 1.21)0.94 (0.67, 1.33) All-cause mortality0.46 (0.40, 0.52)0.94 (0.80, 1.10)MI, stroke or HF hospitalizations0.91 (0.85, 0.97)0.96 (0.89, 1.03)Intention to treat analyses HF hospitalizations0.76 (0.72, 0.80)0.91 (0.86, 0.97) MI or stroke hospitalizations1.00 (0.92, 1.11)1.00 (0.90, 1.11)SGLT2i: sodium–glucose cotransporter-2 inhibitors; DPP4i: dipeptidyl peptidase 4 inhibitors; GLP-1RA: Glucagon-like peptide-1 receptor agonists; CI: confidence intervals; IR: incidence rate; HR: hazard ratioHazard ratios adjusted for variables described in Table 1 and 2 using stabilized inverse probability of treatment weightingSee Additional file 1: Appendix Table S6 for number of events and incidence rates for secondary outcomes and sensitivity analysis Study results were robust to alternative definitions for identifying heart failure phenotypes for HHF (Table 5). For the MI or stroke outcome, while the point estimates were similar to the primary analysis, only two alternative HF definitions for HFpEF yielded statistical significance against DPP4i whereas all others crossed the null. Outcomes were robust when stratified by calendar year—albeit underpowered, and accordingly MI or stroke hospitalization risk for Cohort 2a was not significantly different following 2016.Table 5Adjusted risk of cardiovascular outcomes among patients initiating SGLT2i compared to non-gliflozin therapies, using alternative definitions for HF subtypesHFrEFSGLT2i versus DPP4iSGLT2i versus GLP-1RAHF alternative definition 1a HF hospitalizations0.68 (0.64, 0.72)0.84 (0.79, 0.91) MI or stroke hospitalizations0.92 (0.83, 1.03)1.01 (0.87, 1.11)HF alternative definition 2a HF hospitalizations0.68 (0.64, 0.71)0.85 (0.80, 0.90) MI or stroke hospitalizations0.91 (0.82, 1.01)1.00 (0.87, 1.14)HF alternative definition 3a HF hospitalizations0.71 (0.66, 0.77)0.86 (0.79, 0.94) MI or stroke hospitalizations0.88 (0.75, 1.03)1.04 (0.86, 1.26)Year ≤ 2016 HF hospitalizations0.67 (0.59, 0.77)0.82 (0.71, 0.94) MI or stroke hospitalizations0.70 (0.53, 0.94)1.00 (0.75, 1.34)Year > 2016 HF hospitalizations0.61 (0.55, 0.67)0.84 (0.76, 0.92) MI or stroke hospitalizations0.93 (0.77, 1.11)0.93 (0.76, 1.22)HFpEFSGLT2i versus DPP4iSGLT2i versus GLP-1RAHF alternative definition 1a HF hospitalizations0.63 (0.60, 0.67)0.85 (0.80, 0.91) MI or stroke hospitalizations0.88 (0.79, 0.98)0.97 (0.85, 1.10)HF alternative definition 2a HF hospitalizations0.63 (0.60, 0.66)0.83 (0.79, 0.88) MI or stroke hospitalizations0.90 (0.81, 0.99)0.96 (0.85, 1.08)HF alternative definition 3a HF hospitalizations0.66 (0.62, 0.72)0.88 (0.81, 0.95) MI or stroke hospitalizations0.90 (0.78, 1.03)0.98 (0.83, 1.17)Year ≤ 2016 HF hospitalizations0.70 (0.63, 0.79)0.91 (0.83, 1.05) MI or stroke hospitalizations0.89 (0.71, 1.12)1.05 (0.81, 1.35)Year > 2016 HF hospitalizations0.59 (0.54, 0.64)0.84 (0.76, 0.92) MI or stroke hospitalizations0.89 (0.75, 1.05)0.93 (0.76, 1.13)SGLT2i: sodium–glucose cotransporter-2 inhibitors; DPP4i: dipeptidyl peptidase 4 inhibitors; GLP-1RA: Glucagon-like peptide-1 receptor agonists; CI: confidence intervals; IR: incidence rate; HR: hazard ratioHazard ratios adjusted for variables described in Table 1 using stabilized inverse probability of treatment weightingaFor the primary analysis, the cohort was restricted to patients with codes corresponding to HFrEF or HFpEF in either the first or second position in the discharge diagnosis. We conducted sensitivity analysis where we reconstructed the cohort to patients with the relevant diagnoses at any inpatient discharge diagnosis (alternative definition 1), or any inpatient or outpatient diagnosis (alternative definition 2). Finally, in the third sensitivity analysis, we excluded patients with a recent HF hospitalization within 30 days of index (alternative definition 3). For all analysis, no patient could have diagnosis of HFrEF and HFpEF at the same time, and IPTW was recalculated for each cohort
## Discussion
Initial CVOTs demonstrated the cardiovascular benefits of SGLT2i against placebo on the incidence of HHF and MACE; however, the proportion of patients with concomitant type 2 diabetes and heart failure varied substantially with no distinction made between HFrEF and HFpEF. Subsequent trials dedicated solely to HFrEF and HFpEF populations found similar reductions in HHF but did not assess incidence of MACE. In this population-based cohort study comprised of older adults co-diagnosed with T2D and HF, use of SGLT2i compared to DPP4i was associated with a 33–$35\%$ and 10–$14\%$ lower risk of HHF and MI or stroke respectively, and a 11–$14\%$ lower risk of HHF and a similar risk of MI or stroke against GLP-1RA. Notably, the magnitude of cardiovascular reduction attributable to SGLT2i was comparable in both HFrEF and HFpEF cohorts.
This investigation has pertinent clinical implications. As cardiovascular events remain the primary cause of excess mortality among patients with T2D and HF [26, 27], therapeutic strategies that inform and reduce the incidence of such events can be useful in guiding patient care. Moreover, in contrast to CVOTs which compared SGLT2i against placebo, our study represents the first comprehensive effort to evaluate SGLT2i against DPP4i and more importantly, GLP-1RA. Despite their relevance to clinical medicine, such head-to-head comparison data are unlikely to be generated from clinical trials. Finally, our study findings reinforce the effectiveness of SGLT2i in reducing HHF in HFpEF patients, a condition for which few viable treatment modalities exist [28].
Our finding of SGLT2i reducing HHF is in line with previous trials and observational studies, as well as clinical guidelines that advocate their use in patients with HF [29, 30]. Comparatively, the reduction in HHF attributable to SGLT2i was less pronounced against GLP-1RA relative to DPP4i. In contrast to DPP4i—which have a neutral effect on HHF, CVOTs have shown that GLP-1RA modestly reduce the incidence of HHF between 9 and 11 percent against placebo [13]. Lastly, these findings are also in line with a recent observational study that compared SGLT2i to GLP-1RA and found a $30\%$ reduction in HHF risk among individuals with established cardiovascular disease [31].
In contrast to their robust data for HHF, the evidence for SGLT2i is less consistent for MACE endpoints, with clinical trials indicating that such benefits are confined to patients with established cardiovascular disease. In this context, our finding of SGLT2i reducing risk of MI or stroke compared to DPP4i is consistent with earlier CVOTs where SGLT2i was evaluated against placebo. Notably, our study also demonstrated that the incidence of MI or stroke was comparable among patients initiating SGLT2i versus GLP-1RA—which have salutary effects on MACE endpoints [15, 32].
Despite their documented benefits on cardiovascular endpoints, there may exist some barriers associated with SGLT2i use among patients with HF. First, as all SGLT2i products are currently branded, high prescription drug costs may impose financial constraints among this population—which already has high levels of polypharmacy, and consequently medication-related costs [33]. Secondly, clinicians and patients may be hesitant to use these agents due to their unique adverse reaction profile that encompasses lower limb amputations, diabetic ketoacidosis and urogenital infections [34–37]; however, data from clinical trials suggests that such events do not seem to occur with greater frequency among patients with heart failure [38].
This study took several steps to mitigate concerns for confounding by restricting analysis to new users of study medications and adjusting for pertinent covariates. Patients were sourced from routine clinical care ensuring widespread generalizability of study findings to older adults. Moreover, study estimates were consistent across a range of sensitivity, secondary, and subgroup analyses. Finally, information on medication dispensing, rather than prescribing data, were available for *Medicare data* mitigating some concerns for exposure misclassification. However, study findings should be viewed in light of limitations. First, owing to the observational nature of the study, findings are susceptible to residual confounding. For instance, although we assessed and adjusted for several relevant confounders, information on important variables such as hemoglobin A1c, body weight or severity of HF were not directly available in Medicare data; however, prior studies have shown that balance on many of these unmeasured characteristics can be achieved with the use of claims-based proxies [39]. Second, study findings are most generalizable to older adults enrolled in Medicare fee-for-service plans. However, we would not expect the biological effects of SGLT2i to vary by insurance status. Third, our study lacked sufficient power to explore cardiovascular outcomes for individual SGLT2i. Further, given the time frame over which our study was conducted, we were unable to include more recently approved agents such as ertugliflozin or semaglutide. Finally, we were unable to study heart failure patients without diabetes as the use of SGLT2i among this population remained very low (< $0.7\%$) over the study period, which preceded the publication of the more recent SGLT2i trials dedicated to heart failure populations.
In conclusion, this population-based analyses found that the initiation of SGLT2i was associated with a reduced risk of HHF compared to DPP4i and GLP-1RA, reduced risk of MI or stroke compared to DPP4i, and comparable risk of MI or stroke compared to GLP-1RA. Notably, the cardiovascular benefit profile was similar in magnitude for SGLT2i across patients with HFrEF and HFpEF. These findings have important implications in prevention of cardiovascular morbidity and mortality among patients dually diagnosed with diabetes and heart failure.
## Supplementary Information
Additional file 1. Fig. S1: CONSORT flow diagram. Appendix Table S1: Baseline patient characteristics prior to after IPT weighting, SGLT2i v DPP4i among patients with heart failure with reduced ejection fraction. Appendix Table S2: Baseline patient characteristics prior to after IPT weighting, SGLT2i v DPP4i among patients with heart failure with preserved ejection fraction. Appendix Table S3: Baseline patient characteristics prior to after IPT weighting, SGLT2i v GLP-1RA among patients with heart failure with reduced ejection fraction. Appendix Table S4: Baseline patient characteristics prior to after IPT weighting, SGLT2i v GLP-1RA among patients with heart failure with preserved ejection fraction. Appendix Table S5: Follow up and reasons for censoring. Appendix Table S6: Risk of cardiovascular outcomes among patients initiating SGLT2i compared to other therapies, by heart failure subtype: Sensitivity and secondary analysis.
## References
1. 1.Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and chronic diseases and injuries in 188 countries, 1990–2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2015;386(9995):743–800.
2. Dunlay SM, Givertz MM, Aguilar D. **Type 2 diabetes mellitus and heart failure: a scientific statement From the American Heart Association and the Heart Failure Society of America: This statement does not represent an update of the 2017 ACC/AHA/HFSA heart failure guideline update**. *Circulation* (2019.0) **140** e294-e324. DOI: 10.1161/CIR.0000000000000691
3. Glovaci D, Fan W, Wong ND. **Epidemiology of diabetes mellitus and cardiovascular disease**. *Curr Cardiol Rep* (2019.0) **21** 1-8. DOI: 10.1007/s11886-019-1107-y
4. Aroor AR, Mandavia CH, Sowers JR. **Insulin resistance and heart failure: molecular mechanisms**. *Heart Fail Clin* (2012.0) **8** 609-617. DOI: 10.1016/j.hfc.2012.06.005
5. Neal B, Perkovic V, Mahaffey KW. **Canagliflozin and cardiovascular and renal events in type 2 diabetes**. *N Engl J Med* (2017.0) **377** 644-657. DOI: 10.1056/NEJMoa1611925
6. Wiviott SD, Raz I, Bonaca MP. **Dapagliflozin and cardiovascular outcomes in type 2 diabetes**. *N Engl J Med* (2019.0) **380** 347-357. DOI: 10.1056/NEJMoa1812389
7. Zou X, Shi Q, Vandvik PO. **Sodium–glucose cotransporter-2 inhibitors in patients with heart failure: a systematic review and meta-analysis**. *Ann Internal Med* (2022.0) **1** 1
8. Packer M, Anker SD, Butler J. **Effect of empagliflozin on the clinical stability of patients with heart failure and a reduced ejection fraction: the EMPEROR-reduced trial**. *Circulation* (2021.0) **143** 326-336. DOI: 10.1161/CIRCULATIONAHA.120.051783
9. Anker SD, Butler J, Filippatos G. **Empagliflozin in heart failure with a preserved ejection fraction**. *N Engl J Med* (2021.0) **385** 1451-1461. DOI: 10.1056/NEJMoa2107038
10. McMurray JJV, Solomon SD, Inzucchi SE. **Dapagliflozin in patients with heart failure and reduced ejection fraction**. *N Engl J Med* (2019.0) **381** 1995-2008. DOI: 10.1056/NEJMoa1911303
11. Hwang SJ, Melenovsky V, Borlaug BA. **Implications of coronary artery disease in heart failure with preserved ejection fraction**. *J Am Coll Cardiol* (2014.0) **63** 2817-2827. DOI: 10.1016/j.jacc.2014.03.034
12. Kamon D, Sugawara Y, Soeda T. **Predominant subtype of heart failure after acute myocardial infarction is heart failure with non-reduced ejection fraction**. *ESC heart failure* (2021.0) **8** 317-325. DOI: 10.1002/ehf2.13070
13. Kristensen SL, Rørth R, Jhund PS. **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* (2019.0) **7** 776-785. DOI: 10.1016/S2213-8587(19)30249-9
14. Khan MS, Fonarow GC, McGuire DK. **Glucagon-like peptide 1 receptor agonists and heart failure: the need for further evidence generation and practice guidelines optimization**. *Circulation* (2020.0) **142** 1205-1218. DOI: 10.1161/CIRCULATIONAHA.120.045888
15. 15.American Diabetes Association Professional Practice C, American Diabetes Association Professional Practice C, Draznin B, et al. 9. Pharmacologic approaches to glycemic treatment: standards of medical care in diabetes-2022. Diabetes Care. 2022;45(Supplement_1):S125–43.
16. Das SR, Everett BM, Birtcher KK. **2018 ACC expert consensus decision pathway on novel therapies for cardiovascular risk reduction in patients with type 2 diabetes and atherosclerotic cardiovascular disease: a report of the American College of Cardiology Task Force on Expert Consensus Decision Pathways**. *J Am Coll Cardiol* (2018.0) **72** 3200-3223. DOI: 10.1016/j.jacc.2018.09.020
17. Bullard KM, Cowie CC, Lessem SE. **Prevalence of diagnosed diabetes in adults by diabetes type—United States, 2016**. *MMWR Morb Mortal Wkly Rep* (2018.0) **67** 359-361. DOI: 10.15585/mmwr.mm6712a2
18. Bui AL, Horwich TB, Fonarow GC. **Epidemiology and risk profile of heart failure**. *Nat Rev Cardiol* (2011.0) **8** 30-41. DOI: 10.1038/nrcardio.2010.165
19. 19.Bates BA, Akhabue E, Nahass MM, et al. Validity of international classification of diseases (ICD)-10 Diagnosis codes for identification of acute heart failure hospitalization and heart failure with reduced versus preserved ejection fraction in a national medicare sample. Circ Cardiovasc Qual Outcomes. 2023:e009078.
20. Saczynski JS, Andrade SE, Harrold LR. **A systematic review of validated methods for identifying heart failure using administrative data**. *Pharmacoepidemiol Drug Saf* (2012.0) **21** 129-140. DOI: 10.1002/pds.2313
21. Kiyota Y, Schneeweiss S, Glynn RJ, Cannuscio CC, Avorn J, Solomon DH. **Accuracy of Medicare claims-based diagnosis of acute myocardial infarction: estimating positive predictive value on the basis of review of hospital records**. *Am Heart J* (2004.0) **148** 99-104. DOI: 10.1016/j.ahj.2004.02.013
22. Andrade SE, Harrold LR, Tjia J. **A systematic review of validated methods for identifying cerebrovascular accident or transient ischemic attack using administrative data**. *Pharmacoepidemiol Drug Saf* (2012.0) **21** 100-128. DOI: 10.1002/pds.2312
23. Kim DH, Schneeweiss S, Glynn RJ, Lipsitz LA, Rockwood K, Avorn J. **Measuring frailty in medicare data: development and validation of a claims-based frailty index**. *J Gerontol Ser A* (2018.0) **73** 980-987. DOI: 10.1093/gerona/glx229
24. Austin PC, Stuart EA. **Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies**. *Stat Med* (2015.0) **34** 3661-3679. DOI: 10.1002/sim.6607
25. Austin PC. **Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research**. *Commun Stat Simul Comput* (2009.0) **38** 1228-1234. DOI: 10.1080/03610910902859574
26. Kleinman JC, Donahue RP, Harris MI, Finucane FF, Madans JH, Brock DB. **Mortality among diabetics in a national sample**. *Am J Epidemiol* (1988.0) **128** 389-401. DOI: 10.1093/oxfordjournals.aje.a114979
27. 27.Chen J, Normand S-LT, Wang Y, Krumholz HM. National and regional trends in heart failure hospitalization and mortality rates for Medicare beneficiaries, 1998–2008. Jama. 2011;306(15):1669–78.
28. Heidenreich PA, Bozkurt B, Aguilar D. **AHA/ACC/HFSA guideline for the management of heart failure**. *J Am Coll Cardiol* (2022.0) **1** 1
29. Packer M, Butler J, Zannad F. **Effect of empagliflozin on worsening heart failure events in patients with heart failure and preserved ejection fraction: EMPEROR-preserved trial**. *Circulation* (2021.0) **144** 1284-1294. DOI: 10.1161/CIRCULATIONAHA.121.056824
30. 30.Patorno E, Goldfine AB, Schneeweiss S, et al. Cardiovascular outcomes associated with canagliflozin versus other non-gliflozin antidiabetic drugs: population based cohort study. Bmj. 2018;360:k119.
31. Patorno E, Htoo PT, Glynn RJ. **Sodium–glucose cotransporter-2 inhibitors versus glucagon-like peptide-1 receptor agonists and the risk for cardiovascular outcomes in routine care patients with diabetes across categories of cardiovascular disease**. *Ann Intern Med* (2021.0) **174** 1528-1541. DOI: 10.7326/M21-0893
32. Marso SP, Bain SC, Consoli A. **Semaglutide and cardiovascular outcomes in patients with type 2 diabetes**. *N Engl J Med* (2016.0) **375** 1834-1844. DOI: 10.1056/NEJMoa1607141
33. Faridi KF, Dayoub EJ, Ross JS, Dhruva SS, Ahmad T, Desai NR. **Medicare coverage and out-of-pocket costs of quadruple drug therapy for heart failure**. *J Am Coll Cardiol* (2022.0) **79** 2516-2525. DOI: 10.1016/j.jacc.2022.04.031
34. Dave CV, Schneeweiss S, Kim D, Fralick M, Tong A, Patorno E. **Sodium–glucose cotransporter-2 inhibitors and the risk for severe urinary tract infections: a population-based cohort study**. *Ann Intern Med* (2019.0) **171** 248-256. DOI: 10.7326/M18-3136
35. Fralick M, Schneeweiss S, Patorno E. **Risk of diabetic ketoacidosis after initiation of an SGLT2 inhibitor**. *N Engl J Med* (2017.0) **376** 2300-2302. DOI: 10.1056/NEJMc1701990
36. Dave CV, Schneeweiss S, Patorno E. **Association of sodium–glucose cotransporter 2 inhibitor treatment with risk of hospitalization for fournier gangrene among men**. *JAMA Intern Med* (2019.0) **179** 1587-1590. DOI: 10.1001/jamainternmed.2019.2813
37. Dave CV, Schneeweiss S, Patorno E. **Comparative risk of genital infections associated with sodium–glucose co-transporter-2 inhibitors**. *Diabetes Obes Metab* (2019.0) **21** 434-438. DOI: 10.1111/dom.13531
38. Butler J, Usman MS, Khan MS. **Efficacy and safety of SGLT2 inhibitors in heart failure: systematic review and meta-analysis**. *ESC Heart Fail* (2020.0) **7** 3298-3309. DOI: 10.1002/ehf2.13169
39. Patorno E, Gopalakrishnan C, Franklin JM. **Claims-based studies of oral glucose-lowering medications can achieve balance in critical clinical variables only observed in electronic health records**. *Diabetes Obes Metab* (2018.0) **20** 974-984. DOI: 10.1111/dom.13184
|
---
title: 'General versus central adiposity as risk factors for cardiovascular-related
outcomes in a high-risk population with type 2 diabetes: a post hoc analysis of
the REWIND trial'
authors:
- Edward Franek
- Prem Pais
- Jan Basile
- Claudia Nicolay
- Sohini Raha
- Ana Hickey
- Nadia N. Ahmad
- Manige Konig
- Hong Kan
- Hertzel C. Gerstein
journal: Cardiovascular Diabetology
year: 2023
pmcid: PMC9999507
doi: 10.1186/s12933-023-01757-z
license: CC BY 4.0
---
# General versus central adiposity as risk factors for cardiovascular-related outcomes in a high-risk population with type 2 diabetes: a post hoc analysis of the REWIND trial
## Abstract
### Background
In clinical practice, anthropometric measures other than BMI are rarely assessed yet may be more predictive of cardiovascular (CV) risk. We analyzed the placebo group of the REWIND CV Outcomes Trial to compare several anthropometric measures as baseline risk factors for cardiovascular disease (CVD)-related outcomes in participants with type 2 diabetes (T2D).
### Methods
Data from the REWIND trial placebo group ($$n = 4952$$) were analyzed. All participants had T2D, age ≥ 50 years, had either a previous CV event or CV risk factors, and a BMI of ≥ 23 kg/m2. Cox proportional hazard models were used to investigate if BMI, waist-to-hip ratio (WHR), and waist circumference (WC) were significant risk factors for major adverse CV events (MACE)-3, CVD-related mortality, all-cause mortality, and heart failure (HF) requiring hospitalization. Models were adjusted for age, sex, and additional baseline factors selected by LASSO method. Results are presented for one standard deviation increase of the respective anthropometric factor.
### Results
Participants in the placebo group experienced 663 MACE-3 events, 346 CVD-related deaths, 592 all-cause deaths, and 226 events of HF requiring hospitalization during the median follow-up of 5.4 years. WHR and WC, but not BMI, were identified as independent risk factors of MACE-3 (hazard ratio [HR] for WHR: 1.11 [$95\%$ CI 1.03 to 1.21]; $$p \leq 0.009$$; HR for WC: 1.12 [$95\%$ CI 1.02 to 1.22]; $$p \leq 0.012$$). WC adjusted for hip circumference (HC) showed the strongest association with MACE-3 compared to WHR, WC, or BMI unadjusted for each other (HR: 1.26 [$95\%$ CI 1.09 to 1.46]; $$p \leq 0.002$$). Results for CVD-related mortality and all-cause mortality were similar. WC and BMI were risk factors for HF requiring hospitalization, but not WHR or WC adjusted for HC (HR for WC: 1.34 [$95\%$ CI 1.16 to 1.54]; $p \leq 0.001$; HR for BMI: 1.33 [$95\%$ CI 1.17 to 1.50]; $p \leq 0.001$). No significant interaction with sex was observed.
### Conclusions
In this post hoc analysis of the REWIND placebo group, WHR, WC and/or WC adjusted for HC were risk factors for MACE-3, CVD-related mortality, and all-cause mortality; while BMI was only a risk factor for HF requiring hospitalization. These findings indicate the need for anthropometric measures that consider body fat distribution when assessing CV risk.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12933-023-01757-z.
## Background
Obesity, defined as excess adiposity that is detrimental to health, is a major risk factor for type 2 diabetes and other comorbidities [1]. Patients with type 2 diabetes and obesity have an increased risk for cardiovascular disease (CVD) [1]. Correspondingly, the American Diabetes Association recommends weight management strategies in addition to glycemic control for patients with type 2 diabetes [2].
Obesity can be assessed using different measures. In the absence of imaging modalities, which are typically not used in routine clinical practice, BMI, waist-to-hip ratio (WHR), and waist circumference (WC) are commonly used clinical measures. BMI can be readily calculated to estimate overall body fat, and WHR and WC can be measured during the office visit to estimate distribution of fat which may have varied pathophysiological effects. BMI measures weight to height squared ratio and is inclusive of total body fat and lean mass. WC and WHR measure central adiposity: WC examines the circumference of the abdomen at the level of the umbilicus, and WHR is a ratio of the circumference of the waist to that of the hips with higher ratios indicating more central adiposity [3].
Different measures of obesity have been associated with CVD and all-cause mortality. While some studies indicate that measures of central adiposity are superior to BMI when evaluating patients’ risk of cardiovascular (CV) events [4–6], the Emerging Risk Factors Collaboration showed no difference between BMI, WHR, and WC, in CVD risk prediction [7]. Similarly, while some studies suggest that WHR and WC are superior to BMI at predicting all-cause mortality risk [5], others indicate there is no difference between central and general adiposity measures [8–11]. Gender may also play a role in determining these relationships as WHR, but not BMI, was shown to independently predict major CV events (MACE) in female patients with coronary artery disease [12] and all-cause mortality in female patients with heart failure (HF) but not in male patients [13]. Superiority of BMI, WHR, or WC in predicting these events may also differ depending on the patient or population cohort.
The REWIND CV Outcomes Trial evaluated CVD-related events, including MACE-3, CVD-related mortality, all-cause mortality, and HF requiring hospitalization, over a median of 5.4 years [14, 15]. Patients had type 2 diabetes, were aged 50 years or older with CV risk factors or established CVD and had a baseline BMI of ≥ 23 kg/m2. The placebo group of the REWIND trial provides data on the CV outcomes of patients with type 2 diabetes being treated with the standard of care.
The aim of the current study was to evaluate and contrast measures of general and central adiposity as potential risk factors for MACE-3, CVD-related mortality, all-cause mortality, and HF requiring hospitalization in the placebo group of the REWIND CV Outcomes Trial.
## Study design and patients
Data from the placebo group of the REWIND trial were used for this analysis. Details of the REWIND trial are published elsewhere [14, 15]. In brief, the REWIND trial was a global, multi-center, randomized, double-blind, placebo-controlled clinical trial. Participants with type 2 diabetes were aged ≥ 50 years with established CVD, aged ≥ 55 years with subclinical CVD, or aged ≥ 60 years with two or more CV risk factors. Participants ($$n = 9901$$) were randomized 1:1 to receive once-weekly subcutaneous injections of dulaglutide 1.5 mg or placebo in addition to the standard of care for diabetes and CVD of the specific country during the trial period of August 2011 to August 2018. Median follow-up was 5.4 years. All participants provided written and informed consent and the trial was conducted in accordance with the International Conference on Harmonization Guidelines for Good Clinical Practice and the Declaration of Helsinki.
Weight measurements were taken at baseline and throughout the trial annually as well as at the final study visit. Height, waist circumference, and hip circumference were measured at baseline and every 24 months throughout the trial as well as at the final study visit. To calculate BMI, body weight and height were measured. Body weight was measured using a calibrated scale (mechanical or digital). BMI was calculated as weight in kilograms divided by the square of height in meters. WC and hip circumference (HC) measurements were obtained with the patient in the standing position. WC was measured immediately above the iliac crest and HC at the maximal circumference of the buttocks, both in centimeters. WHR was calculated by dividing WC by HC.
The current analysis examined obesity measures, measured at baseline, as potential risk factors for four outcomes: MACE-3 (non-fatal myocardial infarction, non-fatal stroke, or death from CV causes including unknown causes), CVD-related mortality, all-cause mortality, and HF requiring hospitalization or urgent care. Potential CV outcomes and all deaths were adjudicated by an independent clinical endpoint committee that was masked to treatment assignment. Further adjudication criteria are published elsewhere [15].
## Statistical analyses
Analyses were conducted on all patients in the REWIND placebo group. Baseline demographic and other characteristics are summarized as means and standard deviations (SD) (continuous variables) and/or as counts and proportions (categorical variables).
Regression models were used to evaluate the relationship between three baseline measures of obesity (BMI, WHR and WC) and incident outcomes as described below. To account for the possibility that both WC and HC contain some prognostic information that may be lost by estimating a fixed WHR for each participant, WC was also included in a model that adjusted for the HC.
Results for the obesity measures were analyzed as hazard ratios (HR; $95\%$ confidence intervals [CIs]) for one standard deviation (SD) increase. SD was 5.8 kg/m2 for BMI, 0.08 for WHR, 13.4 cm for WC, and 12.7 cm for HC.
Each obesity measure was assessed separately by first estimating its age and sex-adjusted hazard for each outcome with the Cox proportional hazards (CPH) regression model. If the respective obesity measure was a statistically significant risk factor for the corresponding outcome ($p \leq 0.05$) in this minimally adjusted model, the prespecified risk factors listed in Table 1 were added to this model, which was run using LASSO Cox regression to select significant variables [16]. This fully adjusted model was then scrutinized to determine whether the obesity measure continued to be a significant risk factor for the outcome. Table 1Baseline characteristics of the REWIND placebo group used as additional risk factors to adjust for obesity measuresPlacebo group ($$n = 4952$$)Age (years)66.2 (6.5)Female2283 (46.1)Race American Indian or Alaska Native543 (11.0) Asian218 (4.4) Black or African American346 (7.0) Native Hawaiian or other Pacific Islander22 (0.4) White3744 (75.6)BMI (kg/m2)32.3 (5.8)BMI categories (kg/m2) Normal (< 25 kg/m2)355 (7.2) Overweight (25- < 30 kg/m2)1566 (31.6) Class I (30- < 35 kg/m2)1640 (33.1) Class II-III (≥ 35 kg/m2)1391 (28.1)Waist circumference (cm) Female106.6 (13.0) Male110.8 (13.4)Hip circumference (cm) Female113.0 (13.6) Male108.5 (11.4)Waist-to-hip ratio Female0.95 (0.07) Male1.02 (0.07)Baseline HbA1c (%)$7.4\%$ (1.1)Baseline eGFR (mL/min/1.73 m2) < 3055 (1.1) 30–591063 (21.5) 60–892469 (49.9) ≥ 901238 (25.0)Prior CVDa1554 (31.4)History of myocardial infarction798 (16.1)History of myocardial ischemia by a stress test or with cardiac imaging447 (9.0)Ischemic stroke253 (5.1)Coronary, carotid or peripheral artery revascularization886 (17.9)Unstable angina306 (6.2)Hospitalization for unstable angina with ECG changes590 (11.9)Systolic blood pressure (mm Hg)137.3 (17.0)Diastolic blood pressure (mm Hg)78.5 (9.9)UACR (mg/mmol)1.9 (0.70–7.38)Total cholesterol (mmol/L)4.5 (1.16)LDL cholesterol (mmol/L)2.6 (0.98)HDL cholesterol (mmol/L)1.2 (0.36)Non-HDL cholesterol (mmol/L)3.3 (1.11)Triglycerides (mmol/L)1.60 (1.20–2.25)Current tobacco usage713 (14.4)Past tobacco usage2409 (48.6)Current alcohol consumption1736 (35.1)Antihypertensive agents4654 (94.0)Lipid lowering agents3485 (70.4)Anticoagulant agents2557 (51.6)Antithrombotic agents2909 (58.7)Data are presented as mean (SD), n (%), or median (IQR)BMI body mass index, CVD cardiovascular disease, ECG electrocardiogram, eGFR estimated glomerular filtration rate, HDL high-density lipoprotein, IQR interquartile range, LDL low-density lipoprotein, N number of patients in subgroup of population, n number of patients with stated characteristic, SD standard deviation, UACR urine albumin-to-creatinine ratioaPrior CVD was defined as myocardial infarction, ischemic stroke, unstable angina with electrocardiogram changes, myocardial ischemia on imaging or stress test, or coronary, carotid, or peripheral revascularization Two different combinations of three obesity measures (BMI, WHR, and WC or BMI, WC, and HC) were assessed together with multivariable CPH models, adjusted for age and sex. In contrast to the models described above, the three obesity measures were added to the pool of risk factors that underwent the variable selection process. As a result, final models could retain none to all three of the obesity measures. Analyses for the latter combination (BMI, WC, and HC) were repeated where HC was forced into the model to explore its impact on WC.
All final models were repeated with additional interaction factors for obesity measures and sex.
The proportional hazard assumption for the risk factors was checked visually as well as by testing whether their corresponding time dependent covariates were significant.
Collinearity of the four obesity measures (BMI, WHR, WC, and HC) was evaluated via calculating pairwise Pearson correlation coefficients and performing collinearity diagnostics following Belsley, Kuh, and Welsch [17]. WC was categorized into normal and obese, based on sex- and BMI-related thresholds [18], and cross-tabulated versus BMI categories.
Results from the multivariable CPH models are presented with HR and associated $95\%$ CIs as well as p-values. For continuous risk factors including obesity measures, HRs are given for one SD increase.
All analyses presented are exploratory in nature, and a p value < 0.05 was considered statistically significant. Analyses were performed using SAS© version 9.4., 2017 SAS Institute Inc., Cary, NC, USA.
## Baseline characteristics and demographics
There were 4952 participants in the REWIND placebo group. The average age was 66.2 years, $46.1\%$ were female, and $75.6\%$ were White (Table 1). At baseline, $31.4\%$ had prior established CVD. Mean weight was 88.9 kg and BMI was 32.3 kg/m2. Mean WC was 110.8 cm for men and 106.6 cm for women. Mean HC was 108.5 cm for men and 113.0 cm for women and WHR was 1.02 for men and 0.95 for women.
## Incidence of health outcomes
During follow-up in the placebo group, there were 663 MACE-3 events, 346 CVD-related deaths, 592 all-cause deaths, and 226 events of HF requiring hospitalization or urgent care.
## Association of obesity measures with health outcomes
The list of variables included in the Stepwise Variable Selection can be found in Table 1 alongside the respective baseline values. There was a high correlation between BMI, HC, and WC (pairwise correlation coefficients: 0.78–0.83). WHR had a modest correlation with WC (correlation coefficient: 0.43), and only a minor or no apparent correlation with HC (correlation coefficient: -0.23) and BMI (correlation coefficient: 0.06). The majority of participants in the normal and obese WC categories fell within the overweight ($25.7\%$ and $34.6\%$, respectively), obesity Class I ($37.5\%$ and $30.9\%$, respectively), and obesity Class II BMI categories ($32.6\%$ and $25.9\%$, respectively) (Additional file 1: Fig. S1).
Additional file 1: Fig. S2 shows the results for all obesity measures after adjustment for age and sex. After adjusting for additional variables identified as significant risk factors for the outcomes using the LASSO selection method and detailed in Additional file 1: Table S1, the relationship between obesity measures and outcomes varied by the outcome (Fig. 1).Fig. 1Association of BMI, WHR, WC, and WC adjusted for HC with (A) MACE-3, (B) CVD-related mortality, (C) all-cause mortality, and (D) HF requiring hospitalization or urgent care. Results are estimated from Cox proportional hazard regression models. Results are presented per 1 SD increase (WHR 0.08; BMI 5.8 kg/m2; WC 13.4 cm; HC 12.7 cm). All obesity measures were evaluated, after adjustment for age and sex (Step 1 of the statistical analysis approach). Those that were significant ($p \leq 0.05$) progressed to Step 2 (adjustment for age, sex, and selected baseline factors from the LASSO selection process), otherwise the process stopped after Step 1 (*). BMI body mass index, CI confidence interval, CVD cardiovascular disease, HC hip circumference, HF heart failure, HR hazard ratio; MACE = major adverse cardiovascular events, SD standard deviation; WC waist circumference, WHR waist-to-hip ratio For MACE-3, WHR was found to be a significant independent risk factor (HR = 1.11; $95\%$ CI 1.03 to 1.21; $$p \leq 0.009$$), as was WC (HR = 1.12; $95\%$ CI 1.02 to 1.22; $$p \leq 0.012$$) in the fully adjusted model. The analysis of WC adjusted for HC emerged as the strongest risk factor for MACE-3 (HR = 1.26; $95\%$ CI 1.09 to 1.46; $$p \leq 0.002$$). When either BMI, WHR, and WC or BMI, WC, and HC were included together, the resulting multivariable models did not identify any obesity measure (including WC adjusted for HC) as being significantly associated with MACE-3.
For CVD-related mortality, WHR was identified as a significant risk factor (HR = 1.19; $95\%$ CI 1.04 to 1.36; $$p \leq 0.010$$). WC was not significant when included alone ($$p \leq 0.057$$) but became significant when adjusted for HC, with the strongest association of the four measures for CVD-related mortality (HR = 1.33; $95\%$ CI 1.08 to 1.64; $$p \leq 0.007$$). In the model investigating BMI, WHR, and WC together, only WHR was included via the selection process, resulting in the same final model as the respective single model (HR = 1.19; $95\%$-CI 1.04 to 1.36; $$p \leq 0.010$$). In the model investigating BMI, WC, and HC together, none of the obesity measures were selected. However, when HC was forced into the model, WC was selected, resulting in the same final model as above (HR = 1.33; $95\%$ CI 1.08 to 1.64; $$p \leq 0.007$$).
For all-cause mortality, WC emerged as a significant risk factor (HR = 1.10; $95\%$ CI 1.00 to 1.20; $$p \leq 0.047$$) and remained significant when adjusted for HC, with the largest HR (HR = 1.17; $95\%$ CI 1.00 to 1.38; $$p \leq 0.049$$). In both models investigating a combination of three obesity measures, none of them were included via the selection process. This did not change in the latter model when HC was a forced factor.
BMI was a significant independent risk factor for HF requiring hospitalization (Fig. 1D; HR = 1.33; $95\%$ CI 1.17 to 1.50; $p \leq 0.001$). While WC alone was significant (HR = 1.34; $95\%$ CI 1.16 to 1.54; $p \leq 0.001$), it became nonsignificant after adjusting for HC ($$p \leq 0.077$$). In both models investigating a combination of three obesity measures, only BMI was included via the selection process, resulting in the same final model as the respective single model (HR = 1.33; $95\%$ CI 1.17 to 1.50; $p \leq 0.001$).
In all models no significant interaction with sex was observed.
## Discussion
This post hoc analysis of the placebo group of the REWIND CV Outcomes Trial showed that the anthropometric measures WHR and/or WC, but not BMI, were risk factors for MACE-3, CVD-related mortality, and all-cause mortality in patients with type 2 diabetes and CV risk factors or established CVD. BMI was a significant risk factor only for HF requiring hospitalization. WHR and/or WC were risk factors for all four outcomes, with varying strengths of associations when analyzed in a combination model with other obesity measures. WC adjusted for HC was one of the strongest risk factors for MACE-3, CVD-related mortality, and all-cause mortality, indicating that both WC and HC have independent information pertaining to CV risk which is not completely captured by WHR.
## General adiposity poorly reflects the risk of CV outcomes
While used routinely in clinical practice, increasing BMI is not a reliable universal risk factor for CV-related outcomes in patients with overweight or obesity. Data from the ORIGIN trial showed that obesity, categorized using BMI, had a U-shaped association with mortality and CV outcomes, and patients with overweight and moderate obesity (BMI 25–35 kg/m2) had the lowest mortality risk [19]. Similarly, a meta-analysis showed that the BMI category associated with the lowest risk of mortality in patient groups with varying CV risk was the overweight category (BMI 25-29.9 kg/m2) [20]. Our results indicate that in the REWIND placebo group, with an inclusion criterion of ≥ 23 kg/m2 and a mean baseline BMI of 32 kg/m2, BMI was not a significant independent risk factor for MACE-3, CVD-related mortality, or all-cause mortality. BMI was significant for HF requiring hospitalization. The relationship between BMI and HF has been documented previously [21, 22], including in patients with type 2 diabetes [23], and may be explained by the fact that fluid retention is a key contributor to the development of HF [24]. Additionally, it is recommended to use BMI with caution in patients with Asian ancestry, older adults, and muscular adults [25], further limiting its usefulness in the clinical setting. Overall, with the exception of HF, BMI may not be an accurate measure of patients’ risk of cardiovascular outcomes.
## Central adiposity measures as recommended risk factors for CV outcomes
Our results showed that either WHR or WC were risk factors for MACE-3, CVD-related mortality, and all-cause mortality. Given that different measures of obesity indicate general adiposity versus specific areas of fat depots, such as central fat, this may translate to different physiological effects and therefore varied associations with different outcomes. Although reports differ, most studies suggest that central obesity, measured by WHR or WC, is a risk factor for CVD [4], myocardial infarction [6], and CVD-related mortality [5]. Additionally, WC is a principal risk factor for a high metabolic syndrome score [26], and central obesity is associated with an increased risk of HF hospitalization or death in patients with type 1 diabetes [27]and type 2 diabetes [28]. In addition, multivariable Mendelian randomization analyses suggest that the risk of BMI on hospital admission rates is attenuated by WHR [29]. Higher central fat deposition increases the risk for CV events compared to subcutaneous fat deposition which is potentially caused by the impairment of CV mechanics by visceral adipose tissue; data which are captured by WC or WHR measures but not BMI or skinfold thickness [30]. Additionally, fat depots, particularly visceral and ectopic stores, are linked to increased levels of inflammatory mediators such as adipokines, which may drive decreased cardiac function in patients with central obesity anthropometric measures [31]. While some reports suggest that WC, WHR, and waist-to-height ratio can be used to predict all-cause mortality [5], most indicate that there are no differences in risk prediction by central and general obesity measures [8–11]. This divergence from the current results may be due to differences in study populations as the current study investigated patients aged ≥ 50 years with type 2 diabetes. Waist-to-height ratio was not explored in the current study, however, a previous study suggests it may be a better indicator than other central adiposity measurements for evaluating cardio-cerebrovascular events collectively [32]. Although sex has been to shown to play a role in some risk models [12, 13], we did not identify any interaction between sex and any of the obesity factors.
Given the strong evidence that central adiposity can inform patients’ risk of CV events, guidelines should include detail on collecting these measures in addition to weight and BMI. It is increasingly more widely acknowledged that central adiposity can not only contribute to CV risk but also to type 2 diabetes pathology [33]. The American Diabetes Association recommends assessing patients’ weight distribution to guide risk stratification and treatment plans [2] and the American Heart Association (AHA) and American College of Cardiology (ACC) have highlighted the importance of recording patients’ WC as well as BMI [25, 34, 35]. Patients should be individually treated according to both their BMI and WC category. Risk calculators such as the Framingham Risk Score and the ACC ASCVD calculator are valuable tools to assess CV risk and guide treatment strategies and should include weight, height, WC, and HC to fully inform on risk. Due to the heterogenous nature of obesity, one anthropometric measure does not suffice to inform patients’ risk of different CV outcomes.
## Strengths and limitations
This study had several strengths. The REWIND placebo group was a large cohort of patients with type 2 diabetes and CV risk factors or established CVD. The follow-up period was long (median 5.4 years). The REWIND study protocol did not prescribe interventions on body weight or weight change advice. The REWIND trial data provided detailed information such as general and central adiposity in addition to multiple risk factors which are not typically available in other settings.
This study also had limitations. This was a post hoc analysis that was not prespecified. Participants in the REWIND trial had a history of CVD or CV risk factors so results may not be generalizable to patients with no history or risk factors. Likewise, REWIND participants had type 2 diabetes which limits generalisability to other populations. The full spectrum of BMI was unlikely to be represented. Despite multivariable adjustments, some baseline differences may be unaccounted for which limits conclusions. For outcomes other than MACE, the power is low since there were much fewer events. No causal inference can be concluded from the observed associations.
## Conclusions
In a cohort of patients with type 2 diabetes with high risk for CVD, different general and central measures of obesity better reflected patients’ risk of CV events. There was no single obesity measure that was a risk factor for all outcomes (MACE-3, CVD-related or all-cause mortality, or HF requiring hospitalization), however WHR, WC and/or WC adjusted for HC were risk factors for most outcomes. Measuring BMI, WC, and HC collectively may be the most appropriate when assessing the risk of CV events in patients with type 2 diabetes and obesity.
## Supplementary Information
Additional file 1: Figure S1. Percentage of participants in the baseline Normal or Obese WC category in each BMI category. Figure S2. Association of BMI, WHR, WC, an WC adjusted for HC with (A) MACE-3, (B) CVD-related mortality, (C) all-cause mortality, and (D) HF requiring hospitalization or urgent care, minimally adjusted for age and sex (Step 1 of the statistical analysis approach). Table S1. Significant baseline characteristics used as additional risk factors to adjust for obesity measures.
## References
1. Bray GA, Heisel WE, Afshin A, Jensen MD, Dietz WH, Long M. **The science of obesity management: an Endocrine Society scientific statement**. *Endocr Rev* (2018) **39** 79-132. DOI: 10.1210/er.2017-00253
2. **8 obesity and weight management for the prevention and treatment of type 2 diabetes: standards of medical care in diabetes—2022**. *Diabetes Care* (2021) **45** 113-24. DOI: 10.2337/dc22-S008
3. Snijder MB, van Dam RM, Visser M, Seidell JC. **What aspects of body fat are particularly hazardous and how do we measure them?**. *Int J Epidemiol* (2006) **35** 83-92. DOI: 10.1093/ije/dyi253
4. Huxley R, Mendis S, Zheleznyakov E, Reddy S, Chan J. **Body mass index, waist circumference and waist:hip ratio as predictors of cardiovascular risk—a review of the literature**. *Eur J Clin Nutr* (2010) **64** 16-22. DOI: 10.1038/ejcn.2009.68
5. Schneider HJ, Friedrich N, Klotsche J, Pieper L, Nauck M, John U. **The predictive value of different measures of obesity for incident cardiovascular events and mortality**. *J Clin Endocrinol Metab* (2010) **95** 1777-1785. DOI: 10.1210/jc.2009-1584
6. Yusuf S, Hawken S, Ôunpuu S, Bautista L, Franzosi MG, Commerford P. **Obesity and the risk of myocardial infarction in 27 000 participants from 52 countries: a case-control study**. *Lancet* (2005) **366** 1640-1649. DOI: 10.1016/S0140-6736(05)67663-5
7. **Separate and combined associations of body-mass index and abdominal adiposity with cardiovascular disease: collaborative analysis of 58 prospective studies**. *Lancet* (2011) **377** 1085-1095. DOI: 10.1016/S0140-6736(11)60105-0
8. Carmienke S, Freitag MH, Pischon T, Schlattmann P, Fankhaenel T, Goebel H. **General and abdominal obesity parameters and their combination in relation to mortality: a systematic review and meta-regression analysis**. *Eur J Clin Nutr* (2013) **67** 573-585. DOI: 10.1038/ejcn.2013.61
9. Gnatiuc L, Alegre-Díaz J, Wade R, Ramirez-Reyes R, Tapia-Conyer R, Garcilazo-Ávila A. **General and abdominal adiposity and mortality in Mexico city**. *Ann Intern Med* (2019) **171** 397-405. DOI: 10.7326/M18-3502
10. 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** 2105-2120. DOI: 10.1056/NEJMoa0801891
11. Flegal KM, Graubard BI. **Estimates of excess deaths associated with body mass index and other anthropometric variables**. *Am J Clin Nutr* (2009) **89** 1213-1219. DOI: 10.3945/ajcn.2008.26698
12. Medina-Inojosa JR, Batsis JA, Supervia M, Somers VK, Thomas RJ, Jenkins S. **Relation of waist-hip ratio to long-term cardiovascular events in patients with coronary artery disease**. *Am J Cardiol* (2018) **121** 903-909. DOI: 10.1016/j.amjcard.2017.12.038
13. Streng KW, Voors AA, Hillege HL, Anker SD, Cleland JG, Dickstein K. **Waist-to-hip ratio and mortality in heart failure**. *Eur J Heart Fail* (2018) **20** 1269-1277. DOI: 10.1002/ejhf.1244
14. Gerstein HC, Colhoun HM, Dagenais GR, Diaz R, Lakshmanan M, Pais P. **Design and baseline characteristics of participants in the researching cardiovascular events with a weekly incretin in diabetes (REWIND) trial on the cardiovascular effects of dulaglutide**. *Diabetes Obes Metab* (2018) **20** 42-49. DOI: 10.1111/dom.13028
15. Gerstein HC, Colhoun HM, Dagenais GR, Diaz R, Lakshmanan M, Pais P. **Dulaglutide and cardiovascular outcomes in type 2 diabetes (REWIND): a double-blind, randomised placebo-controlled trial**. *Lancet* (2019) **394** 121-130. DOI: 10.1016/S0140-6736(19)31149-3
16. Tibshirani R. **Regression shrinkage and selection via the lasso**. *J R Stat Soc Ser B* (1996) **58** 267-288
17. Belsey DA, Kuh E, Welsch RE. *Regression diagnostics Identifying influential data and sources of collinearity* (1980)
18. Staiano AE, Bouchard C, Katzmarzyk PT. **BMI-specific waist circumference thresholds to discriminate elevated cardiometabolic risk in white and african american adults**. *Obes Facts* (2013) **6** 317-324. DOI: 10.1159/000354712
19. Doehner W, Gerstein HC, Ried J, Jung H, Asbrand C, Hess S. **Obesity and weight loss are inversely related to mortality and cardiovascular outcome in prediabetes and type 2 diabetes: data from the ORIGIN trial**. *Eur Heart J* (2020) **41** 2668-2677. DOI: 10.1093/eurheartj/ehaa293
20. Lamelas P, Schwalm JD, Leong D, Jolly S, Mehta S, Bangdiwala S. **Varying effects of body mass index and mortality in different risk groups**. *Am J Cardiol* (2018) **122** 1155-1160. DOI: 10.1016/j.amjcard.2018.06.038
21. Piepoli MF, Corrà U, Veglia F, Bonomi A, Salvioni E, Cattadori G. **Exercise tolerance can explain the obesity paradox in patients with systolic heart failure: data from the MECKI score research group**. *Eur J Heart Fail* (2016) **18** 545-553. DOI: 10.1002/ejhf.534
22. Kenchaiah S, Evans JC, Levy D, Wilson PWF, Benjamin EJ, Larson MG. **Obesity and the risk of heart failure**. *N Engl J Med* (2002) **347** 305-313. DOI: 10.1056/NEJMoa020245
23. Glogner S, Rosengren A, Olsson M, Gudbjörnsdottir S, Svensson A-M, Lind M. **The association between BMI and hospitalization for heart failure in 83 021 persons with Type 2 diabetes: a population-based study from the swedish national diabetes registry**. *Diabet Med* (2014) **31** 586-594. DOI: 10.1111/dme.12340
24. Chaney E, Shaw A. **Pathophysiology of Fluid Retention in Heart Failure**. *Contrib Nephrol* (2010) **164** 46-53. DOI: 10.1159/000313720
25. Arnett DK, Blumenthal RS, Albert MA, Buroker AB, Goldberger ZD, Hahn EJ. **2019 ACC/AHA guideline on the primary prevention of cardiovascular disease: a report of the american college of cardiology/american heart association task force on clinical practice guidelines**. *Circulation* (2019) **140** e596-646. PMID: 30879355
26. Lopez-Lopez JP, Cohen DD, Ney-Salazar D, Martinez D, Otero J, Gomez-Arbelaez D. **The prediction of Metabolic Syndrome alterations is improved by combining waist circumference and handgrip strength measurements compared to either alone**. *Cardiovasc Diabetol* (2021) **20** 68. DOI: 10.1186/s12933-021-01256-z
27. **The impact of central obesity on the risk of hospitalization or death due to heart failure in type 1 diabetes a 16-year cohort study**. *Cardiovasc Diabetol* (2021) **20** 153. DOI: 10.1186/s12933-021-01340-4
28. Ichikawa R, Daimon M, Miyazaki T, Kawata T, Miyazaki S, Maruyama M. **Influencing factors on cardiac structure and function beyond glycemic control in patients with type 2 diabetes mellitus**. *Cardiovasc Diabetol* (2013) **12** 38. DOI: 10.1186/1475-2840-12-38
29. Hazewinkel A-D, Richmond RC, Wade KH, Dixon P. **Mendelian randomization analysis of the causal impact of body mass index and waist-hip ratio on rates of hospital admission**. *Econ Hum Biol* (2022) **44** 101088. DOI: 10.1016/j.ehb.2021.101088
30. Selvaraj S, Martinez EE, Aguilar FG, Kim K-YA, Peng J, Sha J. **Association of central adiposity with adverse cardiac mechanics: findings from the hypertension genetic epidemiology network study**. *Circ Cardiovasc Imaging* (2016). DOI: 10.1161/CIRCIMAGING.115.004396
31. Piché M-E, Poirier P, Lemieux I, Després J-P. **Overview of epidemiology and contribution of obesity and body fat distribution to cardiovascular disease: an update**. *Prog Cardiovasc Dis* (2018) **61** 103-113. DOI: 10.1016/j.pcad.2018.06.004
32. Ke J-F, Wang J-W, Lu J-X, Zhang Z-H, Liu Y, Li L-X. **Waist-to-height ratio has a stronger association with cardiovascular risks than waist circumference, waist-hip ratio and body mass index in type 2 diabetes**. *Diabetes Res Clin Pract* (2022) **183** 109151. DOI: 10.1016/j.diabres.2021.109151
33. Lingvay I, Sumithran P, Cohen RV, le Roux CW. **Obesity management as a primary treatment goal for type 2 diabetes: time to reframe the conversation**. *Lancet* (2022) **399** 394-405. DOI: 10.1016/S0140-6736(21)01919-X
34. Bando Y, Kanehara H, Aoki K, Hisada A, Toya D, Tanaka N. **Obesity may attenuate the HbA1c-lowering effect of sitagliptin in Japanese type 2 diabetic patients**. *J Diabetes Investig* (2012) **3** 170-174. DOI: 10.1111/j.2040-1124.2011.00156.x
35. **10 cardiovascular disease and risk management: standards of medical care in diabetes 2022**. *Diabetes Care* (2021). DOI: 10.2337/dc22-S010
|
---
title: Bone marrow mesenchymal stem cell-derived exosomal microRNA-382 promotes osteogenesis
in osteoblast via regulation of SLIT2
authors:
- Hairong Su
- Yulan Yang
- Wanchun Lv
- Xiaoli Li
- Binxiu Zhao
journal: Journal of Orthopaedic Surgery and Research
year: 2023
pmcid: PMC9999516
doi: 10.1186/s13018-023-03667-y
license: CC BY 4.0
---
# Bone marrow mesenchymal stem cell-derived exosomal microRNA-382 promotes osteogenesis in osteoblast via regulation of SLIT2
## Abstract
### Background
Osteoporosis (OP) is a systemic skeletal disorder with increased bone fragility. Human bone marrow mesenchymal stem cells (hBMSCs) have multi-lineage differentiation ability, which may play important roles in osteoporosis. In this study, we aim to investigate the role of hBMSC-derived miR-382 in osteogenic differentiation.
### Methods
The miRNA and mRNA expressions in peripheral blood monocytes between persons with high or low bone mineral density (BMD) were compared. Then we collected the hBMSC-secreted sEV and examined the dominant components. The over-expression of the miR-382 in MG63 cell and its progression of osteogenic differentiation were investigated by qRT-PCR, western blot and alizarin red staining. The interaction between miR-382 and SLIT2 was confirmed by dual-luciferase assay. The role of SLIT2 was also confirmed through up-regulation in MG63 cell, and the osteogenic differentiation-associated gene and protein were tested.
### Results
According to bioinformatic analysis, a series of differential expressed genes between persons with high or low BMD were compared. After internalization of hBMSC-sEV in MG63 cells, we observed that the ability of osteogenic differentiation was significantly enhanced. Similarly, after up-regulation of miR-382 in MG63 cells, osteogenic differentiation was also promoted. According to the dual-luciferase assay, the targeting function of miR-382 in SLIT2 was demonstrated. Moreover, the benefits of hBMSC-sEV in osteogenesis were abrogated through up-regulation of SLIT2.
### Conclusion
Our study provided evidence that miR-382-contained hBMSC-sEV held great promise in osteogenic differentiation in MG63 cells after internalization by targeting SLIT2, which can be served as molecular targets to develop effective therapy.
## Introduction
Osteoporosis (OP) is a systemic skeletal disorder, caused by imbalance of bone metabolism and featured with low bone mass and microarchitectural deterioration in bone tissue, which may result in increased bone fragility [1–3]. Statistics reveal that approximately over 200 million people suffered from OP worldwide and maintain gradual increase every year. In addition, the incidence is significantly higher in female patients than in male patients [4]. Accumulating evidence has illustrated that, as a common metabolic bone disorder, OP progression can be influenced by a variety of factors which may directly or indirectly lead to poor quality of life [5]. At present, therapeutic advance has been made in OP, including hormone replacement therapy, immunotherapy, selective estrogen receptor modulators, bisphosphonates and teriparatide, but limited efficacy and numerous adverse effects still exist. So it is urgent to explore novel treatment strategies for OP.
Human bone marrow mesenchymal stem cells (hBMSCs), identified as multi-potent stromal cells, have multi-lineage differentiation ability and immunosuppressive properties [6–8]. They can be derived from several sources, including the umbilical cord, bone marrow or fat tissue, making them ideal and regeneration as a promising candidate cell type [6, 9]. Accumulating evidence indicates that these features of hBMSCs are associated with extracellular vehicles secretion. Small extracellular vesicles (sEV) are extracellular vesicles generated by fusion with the cellular membrane of multi-vesicular bodies [10–12]. They are between 30 and 150 nm in diameter and contain abundant functional components such as proteins and microRNAs (miRNAs). Currently, hBMSC-derived sEV have been served as an effective strategy for a variety of disease with great characteristics [13]. Nonetheless, the curative effects and mechanisms of the action of hBMSC-sEV on osteogenesis are poorly understood, to the best of our knowledge, particularly in the regulation of osteoblast.
MicroRNA, a predominant component in hBMSC-sEV, is a type of noncoding RNA with a short, single-stranded that target mRNA sequences by binding the 3′-untranslated regions (3′-UTR), which causes degradation or translation blockade of mRNA [14–18]. Previous studies have shown that the miRNA could regulate and affect a variety of cell function, such as proliferation, invasion, inflammation responses and apoptosis [19–21]. A series of miRNAs have been revealed their association with osteogenesis, which participated in the process of bone remodeling, and presented a critical role in the progression of tendon homeostasis and osteoarthritis [22–24]. MiR-20a, miR-27, miR-29b, miR-196a, miR-210 and miR-335-5p have been proved the close relationship with the development of osteoporosis [25–27]. In our present study, we assume the potential role of miR-382 in the regulation of osteogenesis. It has been reported that the miR-382 family is crucial in a variety of cancers, especially in lung cancer, hepatocellular cancer and glioma, presenting significant suppressive function on cell proliferation and metastasis [28–30]. However, the miR-382-associated signaling network has not yet been explored in osteogenesis.
Interestingly, communication between osteoblasts and BMSCs has been identified to take place through small membrane-enclosed vesicular particles, namely exosomes, which is able to fuse with the surrounding cell membranes within circulatory pathways [31, 32]. Exosomes are considered as extracellular vesicles (EVs) and presented great biocompatibility and long-circulating ability [13, 33]. Notably, it has been demonstrated that MSC-derived exosomes exhibited lower immunogenicity and higher levels of regenerative bioactive molecules comparing to those with other cell origins, which emphasized the curative functions of exosomes as the delivery systems of exogenous medications. The current study also reveals that BMSC-derived exosomes may affect the biological properties of human osteoblasts through SATB2 [34]. In order to further explore its potential molecular biological mechanism, we identified by bioinformatics analysis that miR-382 bound to SLIT2.
In our present study, we intended to investigate the effect of hBMSC-sEV on the regulation of osteoblast. Moreover, we investigated the underlying molecular mechanism by using miRNA sequencing of hBMSC-sEV. We hypothesized the potential promoting function of miR-382 and found its targeting mRNA, SLIT2. Our results suggested that hBMSC-sEV can act as a nanotherapeutic agent via miR-382/SLIT2 to promote osteogenic differentiation of osteoblast, which provide a novel therapeutic strategy in the treatment of patients with OP.
## Transcriptome data retrieval and preparation
The microarray datasets (GSE62402 and GSE63446) were downloaded from GEO (http://www.ncbi.nlm.nih.gov/geo/) and were used as the training set. These datasets included simultaneously in PBMs from 5 high-hip-BMD subjects and 5 low-hip-BMD subjects. These datasets were produced by an Illumina Humanref-8 V2.0 Expression BeadChip platform. All microarrays from the two datasets were downloaded and normalized and log2 transformed.
## Differentially expressed genes analysis
To identify the differentially expressed genes(DEGs), we used R language (version 3.5.2, edgeR package) analysis among GEO databases. The thresholds were set as log fold change (FC) > 1.5 or 2 along with false discovery rate (FDR) < 0.05.
## Protein protein interaction (PPI) networks
The protein–protein interaction (PPI) network was conducted by the STRING database (https://string-db.org/, version 11.0) and visualized by Cytoscape software (version 3.7.2), and the CytoHubba plug-in in Cytoscape was used to screen the key genes in the PPI network based on the algorithm of maximum neighborhood component (MNC).
## Cell culture and exosome extraction
Human bone marrow mesenchymal stem cell were obtained according to procedures approved by the Ethics Committee at Maoming People's Hospital. hBMSCs were isolated from healthy volunteers and patients with OP. The hBMSCs were cultured under $10\%$ exosome-free FBS for 72 h. The cells, dead cells and debris were removed using several low-speed centrifugations. hBMSCs of passage 5 were used for in our experiments. Cells were cultured in a humidified incubator with $5\%$ CO2 at 37 °C and passaged with trypsin/EDTA after reaching the confluence. The human osteosarcoma cell line, MG63 cell, was purchased from ATCC company. MG63 cells were cultured in a humidified incubator with $5\%$ CO2 at 37 °C with RPMI 1604 complete medium.
## Isolation of sEV from hBMSC
We followed the MISEV 2018 guidelines to isolate and identify MSC-sEV. Briefly, after MSCs reached 80–$90\%$ confluence, the serum-free medium was added for 48 h to avoid contamination of vesicles from serum. The conditioned medium was collected and centrifuged 800 g for 30 min and additional 3000 g for 30 min to remove cells and debris. The supernatant was then subjected to a 0.1-mm-pore polyetherrsulfone membrane filter (Corning) filtration to eliminate cell debris and large vesicles, followed by a 100,000-Mw cutoff membrane concentration (CentriPlus-70, Millipore). The supernatant volume was reduced to less than 5 mL from approximately 250–500 mL. Using the 70Ti Rotor, the supernatant was then ultracentrifuged at 110,000 g for 2 h at 4 °C (Beckman Coulter). The resulting pellets were resuspended with PBS and ultracentrifuged with 100 Ti Rotor for 1 h at 110,000 g at 4 °C (Beckman Coulter). We used PBS buffer as a negative control in the experiments involving hBMSC-sEV.
## Electron microscopy
Transmission electron microscopy (TEM) was performed to detect the size and morphology of EV samples. Isolated EV samples were deposited onto formvar/silicone monoxide-coated 200 mesh copper grids (Electro-microscopy Sciences) for 2–3 min, followed by fixation with $4\%$ formalin and washed twice with water. The samples were contrasted with $2\%$ uranyl acetate (w/v). Then, the grids were visualized with transmission electron microscope (Tecnai G2 Spirit TEM, Zeiss, Oberkochen, Germany) at 120 kV.
## Nanoparticle tracking analysis (NTA)
We measured the EV particle size and concentration using nanoparticle tracking analysis (NTA) with ZetaView PMX 110 (Particle Metrix, Meerbusch, Germany), and corresponding software ZetaView 8.04.02. Isolated EV samples were appropriately diluted using 1X PBS buffer to measure the particle size and concentration. NTA measurement was recorded. The ZetaView system was calibrated using 110 nm polystyrene particles. Temperature was maintained around 23 °C and 30 °C. Size distribution data were analyzed by normalizing the concentration of particles of different diameters with bin widths of 1 nm and then taking the average of each measurement.
## Exosomes internalization assay
MSC-sEV were labeled with a red fluorescent dye (PKH26; Sigma) according to the manufacturer’s instructions. The labeled sEV were then added to MG63 cell and cocultured for 6 h. HUVECs were washed with PBS and fixed in $4\%$ paraformaldehyde for 15 min. Nuclei were stained with DAPI, and the signals were analyzed with a fluorescence microscope.
## Real-time quantitative RT-PCR assay
Total RNA was extracted from each sample with Trizol reagent. Reverse transcription of miRNA was performed using a tailing reverse kit. mRNA was reversed transcribed into first strand cDNA by Prime Script RT kit. The expression of miR-382, SLIT2, ALP, RUNX2 and OCN was detected with SYBR Premix Ex Taq™ (TaKaRa, China) by using Bio-Rad CFX96.
## Western blot analysis
The cells of each group were lysed in lysis buffer. Determine the quality of the harvested protein by used BCA kit. Then, 20 μg of total proteins was separated in SDS-PAGE gels and transferred to PVDF membrane. The membranes were blocked for 1 h at room temperature and incubated overnight at 4 °C with the relevant antibodies: anti-ALP antibody (1:1000, Abcam, USA), anti-OCN antibody (1:1000, Abcam, USA), anti-RUNX2 antibody (1:1000, Abcam, USA), anti-SLIT2 antibody (1:1000, Abcam, USA), anti-CD63 antibody (1:1000, Abcam, USA), anti-TSG101 antibody (1:1000, Abcam, USA), anti-HSP70 antibody (1:1000, Abcam, USA) and anti-GAPDH antibody (1:10,000, Abcam, USA). Membranes were rinsed and incubated for 1 h with secondary antibodies (Abcam, USA). After three times of washing, membranes were exposed with the ECL kit. Bands were analyzed using ImageJ software (version 1.6 NIH) to analyze the relative expression levels of the above markers.
## Cell proliferation assay
We seeded cells (5 × 103) into 96-well culture plates. CCK8 reagents were applied to the culture medium on days 0, 1, 2 and 3 day. We measured the absorbance at 490 nm in each well by a microplate reader after incubation for 1 h at 37 °C (Bio-Rad 680, Hercules, USA), and cell proliferation was represented by each individual well’s mean absorbance minus the blank value of each well.
## Dual-luciferase reporter assay
We first synthesized wild-type and mutant sequences of the SLIT2 3′UTR (untranslated region) containing the miR-382 binding site. These sequences were cloned into the dual-luciferase vector system pmirGLO (Promega, USA). For the dual-luciferase reporter assay, cells were transfected with the WT- or MUT-SLIT2 luciferase reporter plasmid system together with the indicated components. Cells were cultured for another 48 h and collected. A luciferase assay was then carried out using a dual-luciferase reporter assay system (Promega, USA). Experiments were performed according to the manufacturers’ instructions.
## Alizarin red S staining and quantitative analysis
After the in vitro treatment, the culture medium was abandoned and the cells were washed 2–3 times by PBS, fixed in $4\%$ paraformaldehyde for 15 min, washed with dd H2O and stained with alizarin red S staining solution (Beyotime) for 30 min. After being rinsed with dd H2O, the cells were observed under a microscope (IX50, Olympus, Japan) and photographed. For quantitative analysis, $10\%$ cetyl pyridine chlorophenol (Sigma) was used to dissolve the staining into 10 mm sodium phosphate (Aladdin). The absorbance value was measured at 540 nm using a microplate reader.
## Transfection of mimic and inhibitor
MG63 cell in 6-well plates (5 × 105 cells/well) was transfected with miR-382 mimic (50 nmol/L), or miR-382 inhibitor (100 nmol/L), or their corresponding negative controls. Lipo2000 transfection reagent was simultaneously added into the medium for efficient transfection. After 6 h, we replaced the culture medium in order to remove the transfection reagent. Detection was made 24 h after transfection. miR-382 mimic and miR-382 inhibitor, and their negative controls were purchased from Sangon Biotech.
## Statistical analysis
The data were presented as mean ± SD. Univariate factor analysis of variance (ANOVA) and Student’s t test were used for comparison between groups. The data analysis was carried out by SPSS 20.0 statistical software package. When $P \leq 0.05$, it was considered that the difference was statistically significant.
## Differentially expressed gene analysis and protein–protein interaction analysis.
In order to understand the pathogenesis of osteoporosis (OP), the microarray datasets were downloaded from GEO website, with 5 high bone density (BMD) from healthy subjects and 5 low hip BMD subjects from osteoporosis (OP) patients were included. R language analysis was used to identify the DEGs among these GEO datasets. A total of 54 DEGs (41 up-regulated genes and 13 down-regulated genes) of microRNAs (miRNA) in PBMs were identified from the OP group compared with the normal group, when the log Fold change > 2 or < − 2 along with false discovery rate (FDR) < 0.05, and the DEGs were visualized by the volcano plot and heatmap plot, respectively (Fig. 1A, B). A total of 301 DEGs (33 up-regulated genes and 268 down-regulated genes) of mRNA in PBMs were identified from the OP group compared with the normal group, when the log Fold change > 1.5 or or < − 1.5 along with false discovery rate (FDR) < 0.05, and the DEGs were visualized by the volcano plot and heatmap plot, respectively (Fig. 1A–D). Then GO functional enrichment and KEGG pathway analysis were performed on these DEGs using R software package, and the significant enrichment items of BPs, CCs, MFs were cell division, cytosol and protein binding (Fig. 1E). The KEGG pathway enrichment analysis showed that DEGs were enriched in the apoptosis signaling pathway, resistin as a regulator of inflammation signaling pathway and interleukin-11 signaling pathway (Fig. 1F). Furthermore, we used the STRING database to construct a PPI network for these DEGs with a threshold value of interaction score > 0.4, and the Cytoscape software is used to visualize the PPI network. A node represents a gene, and edges represent relationships between them. Cytoscape software was used to analyze the topology structure of the whole PPI network, and MCC algorithm was used to score the importance of each node in the network (Fig. 1G).Fig. 1Differentially expressed miRNA gene analysis and protein–protein interaction analysis. A Heatmap and B volcano plots of DEGs between human samples from OP patients and normal persons. C Heatmap and D volcano plots of DEGs between human samples from OP patients and normal persons. E Top 20 enrichment of GO enrichment analysis and Top 20 enrichment of KEGG enrichment analysis F. G PPI analysis of DEGs
## Characterization and bioinformatic analysis of hBMSC-sEV
To further explore the biological functions of sEV on osteoporosis, sEV were isolated from supernatant of hBMSCs using ultracentrifugation and characterized by western blot, transmission electron microscopy (TEM) and NTA. Through TEM analysis, hBMSC-sEV exhibited classic cup-shaped or sphere-shaped morphology. In addition, according to NTA, the distribution curve of the particle size of hBMSC-sEV was between 55 and 200 nm (Fig. 2A). Moreover, sEV-associated protein markers CD63 (transmembrane/lipid-bound protein) and TSG101 (cytosolic protein) were enriched in hBMSC-sEV and the negative protein marker HSP70 (an endoplasmic reticulum marker) was not found in hBMSC-sEV compared to hBMSC lysate, which proved the great purity of isolated sEV (Fig. 2B). As a crucial component of exosome cargo, miRNAs have been reported to play critical roles in mediating exosome functions. To identify which miRNA in hBMSC-sEV contributed to osteogenesis, we first performed miRNA sequencing of hBMSC-sEV. The differential expressed miRNAs were analyzed. Overall, we detected 80 mature miRNAs: 26 insignificance gens, 41 up-regulated genes and 13 down-regulated miRNAs (Fig. 2C, D). Then, we chose hsa-miR-382 and hsa-miR-411 for further analysis, as these miRNA are the differentially up-regulated miRNA in exosomes of high-BMD people and exosomes of low-BMD people, which is also the up-regulated miRNA from peripheral blood sequencing analysis of high-BMD people and exosomes of low-BMD people. Fig. 2Extraction and identification of exosomes from hBMSCs. A. Representative electron microscopy image of hBMSC-sEV from OP patients and normal persons. Bar: 100 nm. The EV particle size and concentration were measured by nanoparticle tracking analysis (NTA). B Western blotting analysis of markers (CD63, TSG101, HSP70) in hBMSC-sEV from OP patients and normal persons. C Heatmap and D volcano plots of DEGs analysis of miRNA between hBMSC-derived small EVs from OP patients and normal persons. Statistical evaluation was performed using one-way ANOVA. Mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$
## Active internalization of hBMSC-sEV promotes osteogenic differentiation in MG63 cell
To investigate whether hBMSC-sEV could be internalized by osteoblast,and promote osteogenic differentiation, we used MG63 cell as the cellular model. hBMSC-sEV were labeled with PKH26 and added to the medium of MG63 cells. sEV incorporation can be observed at 2 h after treatment, indicating the internalization of sEV into cells (Fig. 3A). Additionally, the results of qRT-PCR showed that internalization of hBMSC-sEV markedly elevated the expression of miR-382 and miR-411 in MG63 cell (Fig. 3B). Moreover, the internalization of hBMSC-sEV significantly promoted cell proliferation of MG63 cell (Fig. 3C). In terms of osteogenic differentiation, we observed that, according to the ALP staining and alizarin red S staining, the ability of osteogenic differentiation was also remarkably enhanced and the osteogenic differentiation-associated markers (ALP, OCN and RUNX2) were significantly rising compared with untreated group after treatment of hBMSC-sEV (Fig. 3D–F). According to the above results, we hypothesized that active internalization of hBMSC-sEV in osteoblasts may potentially promote the cell proliferation and osteogenic differentiation. Fig. 3Active internalization of hBMSC-sEV promotes osteogenic differentiation in MG63 cell. A Representative microphotographs of immunofluorescence staining of PKH26, B mRNA expression of miR-382 and miR-411, C cell proliferation, D ALP staining, E alizarin red S staining and F western blotting analysis of osteogenesis-associated protein in MG63 cells treated with PBS or hBMSC-sEV from OP patients and normal persons. Statistical evaluation was performed using two-way ANOVA. Mean ± SEM. * represents $P \leq 0.05$ compared with NC group. * $P \leq 0.05$, ***$P \leq 0.001$; # represents $P \leq 0.05$ compared with HG group. ### $P \leq 0.001$;
## Up-regulation of miR-382 promotes osteogenesis in MG63 cell
Due to the increasing level of miR-382 in MG63 cell after internalization of hBMSC-sEV, we further investigated the role of miR-382 in osteogenesis. To identify which targeting protein was affected by the miR-382, we combined the results of PPI analysis and chose the SLIT2 as a candidate. Notably, when we over-regulated the expression of miR-382 through transfection of miR-382 mimic or applied miR-382 inhibitor to suppress its downstream signaling, the miR-382 level was increased after transfection, while inhibitor significantly down-regulated its express. Notably, the expression of SLIT2 was suppressed with up-regulation of miR-382 and was increased with treatment of miR-382 inhibitor, which indicated the targeting function of miR-382 on SLIT2 (Fig. 4A). Similar results were also observed in the protein level (Fig. 4B). According to the starBase online, we predicted the target potential between miR-382 and SLIT2 because of the complementary sequence. Next, through analysis by dual-luciferase reporter gene experiment, we concluded that over-expression of miR-382 inhibited SLIT2 expression. Mechanically, single-stranded miR-382 target mRNA sequences of SLIT2 by binding the 3′-untranslated regions (3′-UTR), resulting in the degradation or translation blockade of SLIT2 expression (Fig. 4C). Additionally, we found that up-regulation of miR-382 remarkably promoted osteogenesis by ALP staining and Alizarin red staining (Fig. 4D–E). According to the representatives of staining, the ability of osteogenic differentiation in MG63 cell was significantly enhanced. Similarly, osteogenic differentiation-associated genes and proteins were gradually increased over time (0, 7, 14 and 21 days) (Fig. 4F, G). Taken together, these results suggest that miR-382 could effectively promoted osteogenesis of osteoblast. Fig. 4miR-382 promotes cell proliferation and osteogenesis in MG63 cell. A. qRT-PCR analysis of miR-382 level in MG63 cells after transfection with negative control, miR-382 mimic or miR-382 inhibitor. B Western blotting analysis of SLIT2 expression in MG63 cells after transfection with negative control, miR-382 mimic or miR-382 inhibitor. C Binding sites between miR-382 and SLIT2 gene predicted by TargetScan website (Left). The regulation of miR-382 on SLIT2 gene transcription is verified by dual-luciferase reporter gene assay (Right). D ALP staining, E alizarin red S staining, F qRT-PCR and G western blot analysis of miR-382, SLIT2 and osteogenesis-associated markers in MG63 cells after transfection with negative control, miR-382 mimic or miR-382 inhibitor. Statistical evaluation was performed using one-way ANOVA. Mean ± SEM. nsP > 0.05, *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ Statistical evaluation was performed using two-way ANOVA. Mean ± SEM. * represents $P \leq 0.05$ compared with NC group. * $P \leq 0.05$, ***$P \leq 0.001$; # represents $P \leq 0.05$ compared with HG group. ### $P \leq 0.001.$
## SLIT2 mediated the pro-osteogenic function of miR-382-contained hBMSC-sEV
To further investigate the role in hBMSC-sEV-induced osteogenesis, we up-regulated the expression of SLIT2 in MG63 cell, combined with treatment of hBMSC-sEV. Then we detected the expression of miR-382 and SLIT2 after plasmid transfection and hBMSC-sEV treatment. According to the qRT-PCR and western blot, the SLIT2 level was significantly increased after over-regulation, which inhibited the expression of miR-382 following internalization (Fig. 5A, B). In addition, the cell proliferation assay presented that the promotive function on cell proliferation of hBMSC-sEV was abrogated by the up-regulation of SLIT2. Similarly, ALP staining, alizarin red S staining and western blot showed that the promoting role of hBMSC-sEV in osteogenesis was also inhibited through over-expression of SLIT2 (Fig. 5C–E). Taken together, SLIT2 mediated the inhibitive function on osteogenesis, which was the target of hBMSC-sEV-derived miR-382.Fig. 5SLIT2 mediated the pro-osteogenic function of miR-382 in hBMSC-sEV. A CCK8 assays and B qRT-PCR of miR-382 and SLIT2 expression in MG63 cells treated with PBS or hBMSC-sEV after transfection of negative control or SLIT2 over-expression plasmid. C ALP staining, D alizarin red S staining and E western blot analysis of SLIT2 and osteogenesis-associated markers in MG63 cells treated with PBS or hBMSC-sEV following transfection of negative control or SLIT2 over-expression plasmid after osteogenic differentiation. Statistical evaluation was performed using two-way ANOVA. Mean ± SEM. * represents $P \leq 0.05$ compared with NC group. * $P \leq 0.05$, ***$P \leq 0.001$; # represents $P \leq 0.05$ compared with HG group. ### $P \leq 0.001$; &represents $P \leq 0.05$ compared with miR-182-5p group. && $P \leq 0.01$, &&&$P \leq 0.001$
## Discussion
Osteoporosis (OP) is a systemic skeletal disease with clinical manifestations of increased susceptibility to bone fragility and fracture, pathologically characterized by low bone density and degeneration of bone tissue microstructure [35–37]. Moreover, osteoblast differentiation is of great importance in skeletal development and osteogenic progression. Dominantly, the imbalance between bone-resorbing and bone-forming osteoblasts may cause bone destructive diseases. Thus, grasping osteogenic differentiation not only provides insights into the bone development, but could also offer therapeutic strategies for OP.
Currently, hBMSC-derived sEV have been served as an effective strategies for a variety of disease with great characteristics, including low immunogenicity, easy storage and high biosafety, and have striking advantages over whole-cell therapy [38, 39]. It has been reported that EVs could be taken up via a variety of endocytic pathways, including macropinocytosis, CME, caveolin-mediated endocytosis and clathrin- and caveolin-independent endocytosis. In particular, sEV are commonly studied as a nanotherapeutic agent for stroke and wound-healing treatment. However, the curative effects and mechanisms of the action of hBMSC-sEV on osteogenesis are poorly understood, particularly in the osteoblast-mediated osteogenic differentiation. In our present study, we observed that hBMSC-sEV could promote osteogenic differentiation of osteoblast through miR-382 after internalization by targeting SLIT2.
The miRNAs are small, approximately 20 nucleotides long, noncoding, single-stranded RNA molecules and act as posttranscriptional regulators of gene expression [40]. It has been reported that the miR-382 family is crucial in multiple cancer types [41]. For instance, the invasion and proliferation ability of pancreatic cancer is significantly inhibited by miR-382 by targeting STAT1/PD-L1 [42]. In squamous cell carcinoma, miR-382 from cancer-associated fibroblast-derived exosomes promoted the metastasis and invasion [43]. Additionally, a previous study revealed that expression of the miR-382 family is significantly higher in glioma patients, compared with healthy person, and is crucial in cancer formation and progression [44]. Moreover, it has been reported that a reduction in mature miR-382 is associated with the proliferation and invasion of glioblastoma cells [45]. However, the miR-382-associated signaling network has not yet been explored in osteogenesis. In our study, we confirmed that hBMSC-sEV-derived miR-382 can remarkably increase its expression in osteoblast and effectively promote cell proliferation and osteogenesis.
Moreover, we assumed that SLIT2 may be a potential downstream of miR-382. SLIT2 plays the anti-inflammatory role in human placenta and decrease LPS-induced endothelial inflammation. SLIT2 is the most commonly studied protein among SLIT family and plays diverse roles in the migration of various types of cells, neural formation, angiogenesis and cancer progression [46–48]. Regarding bone metabolism, previous study reported that SLIT2 were significantly expressed in pre-osteoclasts and/or mature osteoclasts. Furthermore, it has been shown that SLIT2 inhibits osteoclast differentiation, mainly by reducing the migration and fusion of pre-osteoclasts, which are mediated by the suppression of Cdc42 activity [49]. However, the relationship between miR-382 and SLIT2 is still unknown.
The starBase online, an open-source platform for studying the interaction of miRNA and messenger RNA (mRNA), predicted the target potential between miR-382 and SLIT2 because of the complementary sequence. Therefore, we assumed that miR-382 may mediated the expression of SLIT2. Next, through analysis by dual-luciferase reporter gene experiment, we concluded that over-expression of miR-382 inhibited SLIT2 expression. Mechanically, single-stranded miR-382 target mRNA sequences of SLIT2 by binding the 3′-untranslated regions (3′-UTR), resulting in the degradation or translation blockade of SLIT2 expression. Specifically, the important role of microRNA in the formation of osteoblasts has been highlighted. miRNAs have been demonstrated to regulate OP progression by inducing or suppressing osteogenic differentiation, suggesting the relationship between dysregulation of osteogenic differentiation and OP.
Accumulating evidence has noted that patient with OP presented higher rates of bone resorption as well as impaired osteogenesis simultaneously. Notably, osteoblasts exert critical roles in bone formation through regulation on osteogenic differentiation-related proteins such as OCN, ALP and RUNX2. Our results demonstrated that up-regulation of miR-382 in hBMSC-sEV, which can be packaged and secreted into the microenvironment and internalized by osteoblast, can effectively target the expression of SLIT2 and mediated the expression of osteogenesis-associated proteins. Thus, we concluded that a miR-382-SLIT2 axis could exert a critical role in regulating the osteogenic differentiation potentials in OP patients.
In summary, using a series of in vitro experiments, we found poorly expressed miR-382 and high SLIT2 expression in OP patients. Specifically, over-expression of miR-382 or inhibition of miR-382 in hBMSCs could promote the osteogenic differentiation potentials of osteoblast through hBMSC-sEV-mediated cellular communication. Our study provided evidence that miR-382-contained hBMSC-sEV held great promise in osteogenic differentiation in osteoblast by targeting SLIT2, which can be served as molecular targets to develop effective therapy.
## References
1. Liu Y, Wang C, Wang G. **Loureirin B suppresses RANKL-induced osteoclastogenesis and ovariectomized osteoporosis via attenuating NFATc1 and ROS activities**. *Theranostics* (2019) **9** 4648-4662. DOI: 10.7150/thno.35414
2. Lin C, Yu S, Jin R. **Circulating miR-338 cluster activities on osteoblast differentiation: potential diagnostic and therapeutic targets for postmenopausal osteoporosis**. *Theranostics* (2019) **9** 3780-3797. DOI: 10.7150/thno.34493
3. Sun Y, Cai M, Zhong J. **The long noncoding RNA lnc-ob1 facilitates bone formation by upregulating osterix in osteoblasts**. *Nat Metab* (2019) **1** 485-496. DOI: 10.1038/s42255-019-0053-8
4. Greenwood C, Clement J, Dicken A. **Age-related changes in femoral head trabecular microarchitecture**. *Aging Dis* (2018) **9** 976-987. DOI: 10.14336/AD.2018.0124
5. Duan X, Yang S, Zhang L. **V-ATPases and osteoclasts: ambiguous future of V-ATPases inhibitors in osteoporosis**. *Theranostics* (2018) **8** 5379-5399. DOI: 10.7150/thno.28391
6. Lu C, Gao S, Xu G. **Geraniin inhibits TNF-α-induced impairments of osteogenesis through NF-κB and p38 MAPK signalling pathways in bone marrow stem cells**. *Stroke Vasc Neurol* (2017) **2** 47-52. DOI: 10.1136/svn-2016-000046
7. Iaquinta MR, Lanzillotti C, Mazziotta C. **The role of microRNAs in the osteogenic and chondrogenic differentiation of mesenchymal stem cells and bone pathologies**. *Theranostics* (2021) **11** 6573-6591. DOI: 10.7150/thno.55664
8. Yu H, Yang X, Xiao X. **Human adipose mesenchymal stem cell-derived exosomes protect mice from DSS-induced inflammatory bowel disease by promoting intestinal-stem-cell and epithelial regeneration**. *Aging Dis* (2021) **12** 1423-1437. DOI: 10.14336/AD.2021.0601
9. Wang LQ, Lin ZZ, Zhang HX. **Timing and dose regimens of marrow mesenchymal stem cell transplantation affect the outcomes and neuroinflammatory response after ischemic stroke**. *CNS Neurosci Ther* (2014) **20** 317-326. DOI: 10.1111/cns.12216
10. Chen QT, Zhang ZY, Huang QL. **HK1 from hepatic stellate cell-derived extracellular vesicles promotes progression of hepatocellular carcinoma**. *Nat Metab* (2022) **4** 1306-1321. DOI: 10.1038/s42255-022-00642-5
11. Milbank E, Dragano NRV, González-García I. **Small extracellular vesicle-mediated targeting of hypothalamic AMPKα1 corrects obesity through BAT activation**. *Nat Metab* (2021) **3** 1415-1431. DOI: 10.1038/s42255-021-00467-8
12. Kuang H, Dou G, Cheng L. **Humoral regulation of iron metabolism by extracellular vesicles drives antibacterial response**. *Nat Metab* (2023) **5** 111-128. DOI: 10.1038/s42255-022-00723-5
13. Li Y, Tang Y, Yang GY. **Therapeutic application of exosomes in ischaemic stroke**. *Stroke Vasc Neurol* (2021) **6** 483-495. DOI: 10.1136/svn-2020-000419
14. Fabian MR, Sonenberg N. **The mechanics of miRNA-mediated gene silencing: a look under the hood of miRISC**. *Nat Struct Mol Biol* (2012) **19** 586-593. DOI: 10.1038/nsmb.2296
15. Zhang J, Sun P, Zhou C. **Regulatory microRNAs and vascular cognitive impairment and dementia**. *CNS Neurosci Ther* (2020) **26** 1207-1218. DOI: 10.1111/cns.13472
16. Sun Y, Wang T, Lv Y. **MALAT1 promotes platelet activity and thrombus formation through PI3k/Akt/GSK-3β signalling pathway**. *Stroke Vasc Neurol* (2022). DOI: 10.1136/svn-2022-001498
17. Ji Y, Luo Z, Gao H. **Hepatocyte-derived exosomes from early onset obese mice promote insulin sensitivity through miR-3075**. *Nat Metab* (2021) **3** 1163-1174. DOI: 10.1038/s42255-021-00444-1
18. Finger F, Ottens F, Springhorn A. **Author correction: Olfaction regulates organismal proteostasis and longevity via microRNA-dependent signalling**. *Nat Metab* (2020) **2** 126. DOI: 10.1038/s42255-019-0167-z
19. Du WW, Fang L, Li M. **MicroRNA miR-24 enhances tumor invasion and metastasis by targeting PTPN9 and PTPRF to promote EGF signaling**. *J Cell Sci* (2013) **126** 1440-1453. PMID: 23418360
20. Meng F, Liu J, Lu T. **SNHG1 knockdown upregulates miR-376a and downregulates FOXK1/Snail axis to prevent tumor growth and metastasis in HCC**. *Mol Ther Oncolytics* (2021) **21** 264-277. DOI: 10.1016/j.omto.2021.02.002
21. Ren H, Guo Z, Liu Y. **Stem cell-derived exosomal MicroRNA as therapy for vascular age-related diseases**. *Aging Dis* (2022) **13** 852-867. DOI: 10.14336/AD.2021.1110
22. Giordano L, Porta GD, Peretti GM. **Therapeutic potential of microRNA in tendon injuries**. *Br Med Bull* (2020) **133** 79-94. DOI: 10.1093/bmb/ldaa002
23. Oliviero A, Della Porta G, Peretti GM. **MicroRNA in osteoarthritis: physiopathology, diagnosis and therapeutic challenge**. *Br Med Bull* (2019) **130** 137-147. DOI: 10.1093/bmb/ldz015
24. Gargano G, Oliviero A, Oliva F. **Small interfering RNAs in tendon homeostasis**. *Br Med Bull* (2021) **138** 58-67. DOI: 10.1093/bmb/ldaa040
25. Wade SW, Strader C, Fitzpatrick LA. **Estimating prevalence of osteoporosis: examples from industrialized countries**. *Arch Osteoporos* (2014) **9** 182. DOI: 10.1007/s11657-014-0182-3
26. Li B. **MicroRNA regulation in osteogenic and adipogenic differentiation of bone mesenchymal stem cells and its application in bone regeneration**. *Curr Stem Cell Res Ther* (2018) **13** 26-30. PMID: 28578644
27. Zhu J, Liu B, Wang Z. **Exosomes from nicotine-stimulated macrophages accelerate atherosclerosis through miR-21-3p/PTEN-mediated VSMC migration and proliferation**. *Theranostics* (2019) **9** 6901-6919. DOI: 10.7150/thno.37357
28. Wang Y, Cong W, Wu G. **MiR-376a suppresses the proliferation and invasion of non-small-cell lung cancer by targeting c-Myc**. *Cell Biol Int* (2018) **42** 25-33. DOI: 10.1002/cbin.10828
29. Huang Q, Wang C, Hou Z. **Serum microRNA-376 family as diagnostic and prognostic markers in human gliomas**. *Cancer Biomark Sect A Dis Mark* (2017) **19** 137-144. DOI: 10.3233/CBM-160146
30. Choudhury Y, Tay FC, Lam DH. **Attenuated adenosine-to-inosine editing of microRNA-376a* promotes invasiveness of glioblastoma cells**. *J Clin Investig* (2012) **122** 4059-4076. DOI: 10.1172/JCI62925
31. He C, Zheng S, Luo Y. **Exosome theranostics: biology and translational medicine**. *Theranostics* (2018) **8** 237-255. DOI: 10.7150/thno.21945
32. Liao Z, Luo R, Li G. **Exosomes from mesenchymal stem cells modulate endoplasmic reticulum stress to protect against nucleus pulposus cell death and ameliorate intervertebral disc degeneration in vivo**. *Theranostics* (2019) **9** 4084-4100. DOI: 10.7150/thno.33638
33. Yang D, Zhang W, Zhang H. **Progress, opportunity, and perspective on exosome isolation—efforts for efficient exosome-based theranostics**. *Theranostics* (2020) **10** 3684-3707. DOI: 10.7150/thno.41580
34. Yu L, Xu Y, Qu H. **Decrease of MiR-31 induced by TNF-α inhibitor activates SATB2/RUNX2 pathway and promotes osteogenic differentiation in ethanol-induced osteonecrosis**. *J Cell Physiol* (2019) **234** 4314-4326. DOI: 10.1002/jcp.27210
35. Ramirez-Salazar EG, Carrillo-Patino S, Hidalgo-Bravo A. **Serum miRNAs miR-140-3p and miR-23b-3p as potential biomarkers for osteoporosis and osteoporotic fracture in postmenopausal Mexican-Mestizo women**. *Gene* (2018) **679** 19-27. DOI: 10.1016/j.gene.2018.08.074
36. Yuan L, Zhao N, Wang J. **Major vault protein (MVP) negatively regulates osteoclastogenesis via calcineurin-NFATc1 pathway inhibition**. *Theranostics* (2021) **11** 7247-7261. DOI: 10.7150/thno.58468
37. Bai H, Zhao Y, Wang C. **Enhanced osseointegration of three-dimensional supramolecular bioactive interface through osteoporotic microenvironment regulation**. *Theranostics* (2020) **10** 4779-4794. DOI: 10.7150/thno.43736
38. Luo Q, Xian P, Wang T. **Antioxidant activity of mesenchymal stem cell-derived extracellular vesicles restores hippocampal neurons following seizure damage**. *Theranostics* (2021) **11** 5986-6005. DOI: 10.7150/thno.58632
39. Vonk LA, van Dooremalen SFJ, Liv N. **Mesenchymal stromal/stem cell-derived extracellular vesicles promote human cartilage regeneration in vitro**. *Theranostics* (2018) **8** 906-920. DOI: 10.7150/thno.20746
40. Khanaghaei M, Tourkianvalashani F, Hekmatimoghaddam S. **Circulating miR-126 and miR-499 reflect progression of cardiovascular disease; correlations with uric acid and ejection fraction**. *Heart Int* (2016) **11** e1-e9. DOI: 10.5301/heartint.5000226
41. Lewinska A, Adamczyk-Grochala J, Deregowska A. **Sulforaphane-induced cell cycle arrest and senescence are accompanied by DNA hypomethylation and changes in microRNA profile in breast cancer cells**. *Theranostics* (2017) **7** 3461-3477. DOI: 10.7150/thno.20657
42. Zhang H, Zhu C, He Z. **LncRNA PSMB8-AS1 contributes to pancreatic cancer progression via modulating miR-382-3p/STAT1/PD-L1 axis**. *J Exp Clin Cancer Res* (2020) **39** 179. DOI: 10.1186/s13046-020-01687-8
43. Sun LP, Xu K, Cui J. **Cancer-associated fibroblast-derived exosomal miR-382-5p promotes the migration and invasion of oral squamous cell carcinoma**. *Oncol Rep* (2019) **42** 1319-1328. PMID: 31364748
44. Wang J, Chen C, Yan X. **The role of miR-382-5p in glioma cell proliferation, migration and invasion**. *Onco Targets Ther* (2019) **12** 4993-5002. DOI: 10.2147/OTT.S196322
45. Wang J, Li T, Wang B. **Circ-UBAP2 functions as sponges of miR-1205 and miR-382 to promote glioma progression by modulating STC1 expression**. *Cancer Med* (2021) **10** 1815-1828. DOI: 10.1002/cam4.3759
46. Jeon MJ, Lim S, You MH. **The role of Slit2 as a tumor suppressor in thyroid cancer**. *Mol Cell Endocrinol* (2019) **483** 87-96. DOI: 10.1016/j.mce.2019.01.010
47. Ballard MS, Hinck L. **A roundabout way to cancer**. *Adv Cancer Res* (2012) **114** 187-235. DOI: 10.1016/B978-0-12-386503-8.00005-3
48. Ypsilanti AR, Zagar Y, Chédotal A. **Moving away from the midline: new developments for Slit and Robo**. *Development* (2010) **137** 1939-1952. DOI: 10.1242/dev.044511
49. Park SJ, Lee JY, Lee SH. **SLIT2 inhibits osteoclastogenesis and bone resorption by suppression of Cdc42 activity**. *Biochem Biophys Res Commun* (2019) **514** 868-874. DOI: 10.1016/j.bbrc.2019.05.046
|
---
title: Anterior gradient 2 induces resistance to sorafenib via endoplasmic reticulum
stress regulation in hepatocellular carcinoma
authors:
- Hung-Wen Tsai
- Yi-Li Chen
- Chun-I Wang
- Ching‑Chuan Hsieh
- Yang-Hsiang Lin
- Pei-Ming Chu
- Yuh-Harn Wu
- Yi-Ching Huang
- Cheng-Yi Chen
journal: Cancer Cell International
year: 2023
pmcid: PMC9999520
doi: 10.1186/s12935-023-02879-w
license: CC BY 4.0
---
# Anterior gradient 2 induces resistance to sorafenib via endoplasmic reticulum stress regulation in hepatocellular carcinoma
## Abstract
### Background
Hepatocellular carcinoma (HCC) accounts for almost $80\%$ of all liver cancer cases and is the sixth most common cancer and the second most common cause of cancer-related death worldwide. The survival rate of sorafenib-treated advanced HCC patients is still unsatisfactory. Unfortunately, no useful biomarkers have been verified to predict sorafenib efficacy in HCC.
### Results
We assessed a sorafenib resistance-related microarray dataset and found that anterior gradient 2 (AGR2) is highly associated with overall and recurrence-free survival and with several clinical parameters in HCC. However, the mechanisms underlying the role of AGR2 in sorafenib resistance and HCC progression remain unknown. We found that sorafenib induces AGR2 secretion via posttranslational modification and that AGR2 plays a critical role in sorafenib-regulated cell viability and endoplasmic reticulum (ER) stress and induces apoptosis in sorafenib-sensitive cells. In sorafenib-sensitive cells, sorafenib downregulates intracellular AGR2 and conversely induces AGR2 secretion, which suppresses its regulation of ER stress and cell survival. In contrast, AGR2 is highly intracellularly expressed in sorafenib-resistant cells, which supports ER homeostasis and cell survival. We suggest that AGR2 regulates ER stress to influence HCC progression and sorafenib resistance.
### Conclusions
This is the first study to report that AGR2 can modulate ER homeostasis via the IRE1α-XBP1 cascade to regulate HCC progression and sorafenib resistance. Elucidation of the predictive value of AGR2 and its molecular and cellular mechanisms in sorafenib resistance could provide additional options for HCC treatment.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12935-023-02879-w.
## Introduction
Hepatocellular carcinoma (HCC) is the most common hepatic malignant tumor and the 2nd most common cause of cancer-related death worldwide [1]. Approximately $70\%$ of patients are ineligible for curative therapy when they are diagnosed. Sorafenib is the first-line systemic therapy for advanced HCC to prolong survival [2]. Sorafenib is a multiple tyrosine kinase inhibitor (TKI) that inhibits numerous cell surface tyrosine kinases, such as vascular endothelial growth factor receptor (VEGFR)-1, VEGFR-2, VEGFR-3, platelet-derived growth factor receptor (PDGFR)-β and downstream associated serine/threonine kinases involved in the mitogen-activated protein kinase (MAPK) cascade [3, 4]. In vitro, sorafenib blocks cell proliferation and induces cell apoptosis in HCC cell lines, and in vivo, sorafenib inhibits tumor growth and induces tumor cell apoptosis [5]. Sorafenib is effective in increasing the median survival time of HCC patients by approximately 3–5 months. However, several side effects are associated with sorafenib treatment and are usually followed by drug resistance [6]. Several mechanisms and pathways related to acquired resistance to sorafenib have been identified; these include the epithelial-mesenchymal transition (EMT) mechanism and activation of hypoxia-inducible pathways, the phosphatidylinositol-3-kinase (PI3K)/Akt pathway, and the Janus kinase-signal transducer and activator of transcription (JAK-STAT) pathway [6]. Moreover, JAK/STAT pathway-related molecules, such as phospho-STAT3 and its downstream proapoptotic proteins Mcl-1 and cyclin D1, exhibit dysregulated expression in HCC cell lines with sorafenib resistance [7]. Previously, Liovet et al. reported that HCC progression is enhanced after sorafenib treatment through paracrine secretion of hepatocyte growth factor by stromal cells stimulated by VEGFA [8]. Unfortunately, thus far, there are no useful markers to predict the efficiency of sorafenib targeted therapy in HCC.
In the present study, we retrieved information from two databases: the sorafenib-resistant dataset for Huh7 cells by Regan-Fendt et al. ( GSE94550, [9]) and the Roessler liver microarray dataset from Oncomine (GSE14520, [10]). To identify potential candidates for further study through intersection of the two datasets, we utilized a > 2-fold change as the criterion for the selection of sorafenib-modulated molecules in sorafenib-resistant Huh7 cells compared to parental cells (GSE94550) and a > 1.2-fold change as the criterion for choosing oncogenes related to survival in the bottom $25\%$ vs. the top $25\%$ of HCC patients from the Roessler Liver microarray (GSE14520). Following the above analysis, the selected genes were validated, more stringently filtered and applied to evaluate highly significant molecules. Interestingly, anterior gradient 2 (AGR2) was identified as a major gene highly correlated with survival rate and sorafenib resistance in liver cancer, but its associated mechanism and physiological significance have not been well elucidated. Therefore, AGR2 was selected for further study to investigate its molecular mechanism associated with sorafenib resistance and its physiological role in HCC.
AGR2 is a member of the protein disulfide isomerase (PDI) family of endoplasmic reticulum (ER) proteins that catalyze thiol-disulfide interchange and protein folding reactions [11, 12]. AGR2 was first identified in the estrogen receptor-expressing MCF-7 breast cancer cell line and was found to be regulated by estrogen both in vitro and in vivo [13–15]. AGR2 has been detected in different cancer types and is highly expressed in liver, breast, pancreas and bladder cancer tissues compared with healthy tissues [16]. Although AGR2 is an ER-resident protein, it has also been observed in the nucleus [17], cytoplasm [18], mitochondria [19], cell surface [20], extracellular matrix [21], urine [22] and blood [21]. Recently, Delom et al. showed that AGR2 localized in the extracellular matrix makes cancer cells more aggressive [23]. Moreover, AGR2 dysregulation has been implicated in certain disease processes, such as cancer progression and drug resistance [24]. Arumugam et al. reported that recombinant AGR2 can enhance pancreatic ductal adenocarcinoma cell migration, invasion and proliferation through C4.4a cell surface receptor-mediated signaling [25]. siRNA-mediated AGR2 knockdown induces cell death, inhibits cell growth and arrests cell cycle progression in breast cancer cells [26]. Similarly, AGR2 promotes cell growth and migration through the Akt signaling pathway in non-small cell lung cancer; moreover, the phospho-Akt level is reduced after depletion of AGR2 [27]. EGFR signaling is triggered by high AGR2 expression, which is defined as an early initiating factor and serves as a potential target for the curative treatment of neoplastic and chronic pancreatic disease [28]. Based on the above evidence, we suggest that AGR2 expression might be a marker to predict the presence of drug resistance in individual patients.
Previously, AGR2 was defined as a dominant factor in ER homeostasis [29]. Under ER stress, a series of adaptive mechanisms, such as unfolded protein response (UPR) signaling, are activated to cope with increased protein folding in the ER [30]. ER stress and UPR signaling activation result in the development and progression of several human diseases, including cancer [30]. Moreover, AGR2 has been demonstrated to be modulated by the UPR, likely through the PERK, ATF6, and IRE1α arms of the UPR, which can regulate the ER-associated degradation machinery (ERAD), resulting in induction of the cell’s ability to resolve ER stress [29]. Therefore, we suggest that AGR2 plays a critical role in the regulation of UPR signaling and ER stress.
In the present study, we found that AGR2 was highly correlated with overall and recurrence-free survival rates and with several clinical parameters in liver cancer. AGR2 was more highly expressed in sorafenib-resistant cells than in sorafenib-sensitive cells, and AGR2 was downregulated by sorafenib in both cell lines. Sorafenib-resistant cells were more tolerant to sorafenib and exhibited a lower apoptosis rate than sorafenib-sensitive cells. Sorafenib-induced cell death in HCC was found to be reversed and induced with recombinant AGR2 and AGR2-silencing constructs, respectively. The diverse regulatory mechanisms involved in sorafenib-induced cell death in both sorafenib-sensitive and sorafenib-resistant cells may be mediated by the modulation of ER stress and X-box binding protein (XBP) 1 status. This is the first report to uncover the molecular mechanism involved in sorafenib resistance, and further elucidation of the predictive role and molecular and cellular mechanisms of AGR2 related to sorafenib resistance may provide additional opportunities to establish complementary therapies for HCC.
## Database analysis
The two datasets, sorafenib-resistant dataset for Huh7 cells (GSE94550, [9]) and the Roessler Liver microarray dataset (GSE14520, [10]) were retrieved, and utilized > 2-fold change as the criterion for selection of sorafenib-modulated molecules in sorafenib-resistant Huh7 cells compared to parental cells (GSE94550) and > 1.2-fold as the criterion for choosing oncogenes related to survival in the bottom $25\%$ vs. the top $25\%$ of HCC patients from the Roessler Liver microarray dataset (GSE14520) to intersect the potential candidates.
## Cell culture and treatments
The HepG2, Huh7, J7 and Hep3B human hepatoma cell lines were routinely cultured in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen, Grand Island, NY) supplemented with $10\%$ fetal bovine serum (HyClone, Road Logan, UT), 100 U/ml penicillin and 100 mg/ml streptomycin at 37 °C with $5\%$ CO2. The hepatoma cells were stimulated with 0–10 μM sorafenib (Sigma–Aldrich, Burlington, MA).
## Quantitative RT-PCR
The cDNA template was prepared and Quantitative PCR reaction mixture contains 500 nM forward and reverse primers, and 1 × SYBR Green reaction mix (Applied Biosystems, Waltham, MA). SYBR Green fluorescence was determined by the ABI PRISM 7500 detection system (Applied Biosystems). AGR2 Forware-5′ GAGCCAAAAAGGACACAAAGGA 3′, Reverse-5′ TGAGTTGGTCACCCCAACCT 3′; 18SrRNA Forware-5'GCAGCTCACCTACCTGGAGAAATA3′, Reverse 5'TGCGTGTGTGGGTCTTTGAA3′.
## Immunohistochemistry
The use of archived formalin-fixed, paraffin-embedded tissue blocks was approved by the Institutional Review Board of National Cheng Kung University Hospital. Tissue slides from HCC patients were evaluated via immunohistochemistry and hematoxylin/eosin staining using a polyclonal antibody against AGR2 (GeneTex, Hsinchu, Taiwan) according to the avidin–biotin complex method, as described previously [31]. Immunoreactivity for AGR2 was visualized using DAB/nickel substrate (Vector Laboratories, Burlingame, CA).
## MTT assay
Cell viability was analyzed using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay(Sigma–Aldrich, Burlington, MA). Cells (5 × 103) were seeded on 96-well plates overnight. After treatment, 20 μl MTT reagent was added to each well for 3 h, and the absorbance at 570 nm was determined with a SpectraMax microplate reader.
## Apoptosis assay
At the end of treatment, the cells were washed and resuspended in binding buffer (Annexin V PE Apoptosis Detection kit; BD Biosciences, East Rutherford, NJ). After incubation with annexin V-PE and propidium iodine (PI), binding buffer was added, and the cells were analyzed via fluorescence-activated cell sorting (FACScan, BD Biosciences). Data analysis was performed using Cell Quest software.
## RT–PCR
The cDNA template was prepared and amplified via PCR for 30 cycles at 95 °C for 15 s, 54 °C for 15 s, and 72 °C for 30 s (AGR2) or for 33 cycles at 95 °C for 15 s, 57 °C for 15 s, and 72 °C for 30 s (XBP1). 18S rRNA was used as an internal control. PCR products were examined via $2\%$ agarose gel (Amresco, Solon, OH, USA) electrophoresis. The following primers were used: AGR2 Forward-5′ GAGCCAAAAAGGACACAAAGGA 3′, Reverse-5′ TGAGTTGGTCACCCCAACCT 3′; XBP1 Forward-5′ TTACGAGAGAAAACTCATGGCC 3′, Reverse-5′ GGGTCCAAGTTGTCCAGAATGC 3′. 18S rRNA Forward-5′ GCAGCTCACCTACCTGGAGAAATA 3′, Reverse-5′ TGCGTGTGTGGGTCTTTGAA 3′.
## Western blotting
Total protein was fractionated via 8–$12\%$ SDS–PAGE, transferred to an immobilon polyvinylidene difluoride membrane (Amersham Biosciences, Amersham, UK), and hybridized with specific primary antibodies against AGR2 (GeneTex, Hsinchu, Taiwan), ATF6 (GeneTex), p-IRE1α (GeneTex), IRE1α (Cell Signaling Technology, Danvers, MA), p-PERK (GeneTex), PERK (Cell Signaling Technology) and β-actin (GeneTex) overnight at 4 °C. Subsequently, the membrane was probed with the appropriate HRP-conjugated secondary antibody for 1 h at room temperature. Finally, immune complexes were visualized via the chemiluminescence method using an ECL detection kit (Merck, Darmstadt, Germany).
## Establishment of AGR2-silenced cells
Small hairpin (sh) RNA and small interfering (si) RNA targeting AGR2 were purchased from Academia Sinica and Thermo Fisher Scientific. HepG2 and Huh7 cells were transfected with siRNA and/or shRNA targeting AGR2 using Lipofectamine 2000 reagent (Invitrogen). After transfection, the expression of AGR2 was determined using RT–PCR and Western blotting.
## Establishment of sorafenib-resistant HCC cells
HepG2 and Huh7 hepatoma cells were cultured in medium containing increasing concentrations of sorafenib in the range of 0.5–7 μM over a period of 6 months. After successful establishment, sorafenib-resistant HepG2 (SR-HepG2) and Huh7 (SR-Huh7) cells were maintained in medium containing 7 μM sorafenib. The sorafenib-resistant cells were routinely cultured in DMEM medium containing sorafenib. For various experiments, the parental and sorafenib-resistant cells were seeded in the regular DMEM medium without sorafenib to attenuate the sorafenib effect in the routinely cultured sorafenib-resistant cells. After seeding 24 h, the parental and sorafenib-resistant cells were then treated with the indicated concentrations (5–10 μM) and time points (24–48 h) of sorafenib. At the end of treatment, the cell lysate and conditioned medium were collected.
## Preparation of conditioned medium (CM)
CM was collected and centrifuged at 2000 × g for 5 min to eliminate intact cells, concentrated in spin columns with a 3-kDa molecular weight cutoff (Amicon Ultra, Millipore), and stored at − 80 °C for subsequent experiments.
## Clinical HCC specimens
HCC tissues were obtained from the National Health Research Institute Biobank and the human biobank of National Cheng Kung University Hospital. All clinical specimen experiments were performed in accordance with the guidelines of the Institutional Review Board of National Cheng Kung University Hospital (IRB No: A-ER-109-533). Informed consent was waived for the use of the specimens from the National Health Research Institute Biobank and human biobank of National Cheng Kung University Hospital.
## Statistical analysis
Correlations between the AGR2 level (Roessler liver array, 39-△Ct) and clinicopathological indicators were assessed using a Wilcoxon rank sum test. Data are presented as the mean ± SD. Recurrence-free survival (RFS) and overall survival (OS) were calculated using the Kaplan–Meier method, and a log-rank test was used to assess differences between groups. A Cox proportional hazards regression model was used to measure the independence of different factors. Cox regression was performed via forward stepwise analysis, and only the prognostic variables that were significant in the univariate analysis were included in the model. All values are reported as the mean ± SD. Two-way ANOVA, Student’s t test, a chi-square test, or Fisher’s exact test was applied to evaluate experimental differences among groups. p values less than 0.05 were considered to indicate statistical significance.
## AGR2 is clinically relevant in HCC
We retrieved two datasets, the sorafenib-resistant dataset for Huh7 cells (GSE94550, [9]) and the Roessler Liver microarray dataset (GSE14520, [10]), and utilized > 2-fold change as the criterion for selection of sorafenib-modulated molecules in sorafenib-resistant Huh7 cells compared to parental cells (GSE94550) and > 1.2-fold as the criterion for choosing oncogenes related to survival in the bottom $25\%$ vs. the top $25\%$ of HCC patients from the Roessler Liver microarray dataset (GSE14520) to intersect the potential candidates for further study. According to the above analysis, the selected genes were validated and filtered more stringently and applied to evaluate highly significant molecules. After retrieving these datasets, 545 upregulated genes (sorafenib resistance vs. control >2-fold) and 609 downregulated genes (sorafenib resistance vs. control < 2-fold) were identified in the sorafenib-resistant hepatoma Huh7 cell dataset (GSE94550). In the Roessler Liver microarray dataset (GSE14520), we used > 1.2-fold as the criterion for choosing oncogenes related to survival in the bottom $25\%$ vs. the top $25\%$ of HCC patients, and only 12 dysregulated genes were identified (Fig. 1A). Finally, we identified 4 potential candidates in the two intersecting microarray datasets, namely, neurotensin (NTS), AGR2, alpha-fetoprotein (AFP) and meprin A, alpha (MEP1A) (Fig. 1A, B). Previously, the AGR2 protein was demonstrated to be a part of the PDI family of ER proteins that mediate the formation of disulfide bonds and catalyze protein folding [32]. Moreover, AGR2 is highly expressed in numerous cancer types, including liver cancer [16]. Through a clinical study, Hrstka et al. showed that AGR2 expression can be used as a marker to predict poor prognosis in breast cancer [15]. Hence, we suggest that AGR2 may be a potential candidate target in sorafenib-treated HCC. Interestingly, AGR2 is the major gene highly correlated with the survival rate and sorafenib resistance in liver cancer; however, the relationship between AGR2 and sorafenib treatment in HCC has not been demonstrated. Therefore, AGR2 was selected for further study to investigate its molecular mechanism associated with sorafenib resistance and its physiological role in HCC.Fig. 1Selection of potential candidate genes and analysis of clinical parameter correlations and survival. A Schematic diagram of the analytic protocol for the selection of sorafenib-mediated candidate genes. The 4 selected candidate genes were consistently observed in these two datasets. B *The* gene descriptions and fold changes in the GSE14520 and GSE94550 datasets are shown. C–D, J–K KM survival curves (OS and RFS) of 2 groups of patients with HCC grouped by AGR2 expression level [cutoff established on the basis of Roessler liver microarray scores] (C-D, GSE14520) and the cycle threshold value (Ct) obtained by quantitative PCR in our collected HCC specimens (J–K, median, 39-dCt). Patients with high AGR2 levels have worse overall survival and recurrence-free survival. E–I The AGR2 levels with several parameter correlations are shown. L, M Immunohistochemistry staining showing AGR2 expression in HCC specimens (200X). AGR2 is highly expressed in tumor tissues compared with normal liver tissue. d: delta The clinical significance of AGR2 expression was analyzed in the Roessler Liver database (GSE14520) and in our collected cohort (Fig. 1C–K). Patients with lower AGR2 expression (last $40\%$) had a better OS rate (log-rank $P \leq 0.05$; AGR2 high (top $40\%$): standard error, 2.738; $95\%$ CI 36.196–46.927; AGR2 low (last $40\%$): standard error, 2.432; $95\%$ CI 45.322–54.857) and RFS rate (log-rank $P \leq 0.05$; AGR2 high (top $40\%$): standard error, 2.766; $95\%$ CI 27.398–38.243; AGR2 low (last $40\%$): standard error, 2.645; $95\%$ CI 36.365–46.734) (Fig. 1C, D). The Roessler Liver microarray dataset (GSE14520) provides detailed clinical information; therefore, the correlation between AGR2 and various clinical parameters was statistically analyzed to define the role of sorafenib-regulated AGR2 in HCC progression (Tables 1–3). High AGR2 expression was significantly correlated with high AFP and ALT levels, a high predicted risk metastasis signature score (Fig. 1E, F, G), a large primary tumor size and more advanced pathological stages of HCC (Fig. 1H, I). A high AGR2 level in HCC was correlated with a high serum AFP level ($$P \leq 0.002$$) (Table 1). Univariate analysis showed that male sex ($$P \leq 0.009$$), tumor size ($$P \leq 0.045$$), CLIP score ($$P \leq 0.002$$), BCLC stage ($P \leq 0.001$), AJCC stage ($P \leq 0.001$), and high AGR2 level ($$P \leq 0.003$$) were significant predictors of worse RFS (Table 2). Multivariate analysis showed that male sex ($$P \leq 0.023$$, HR = 2.124, CI = 1.112–4.058), BCLC stage ($$P \leq 0.021$$, HR = 1.653, CI = 1.077–2.535), AJCC stage ($$P \leq 0.019$$, HR = 1.607, CI = 1.083–2.386) and a high AGR2 level ($$P \leq 0.010$$, HR = 1.572, CI = 1.112–2.224) were independently associated with RFS (Table 2). For OS, univariate analysis showed that cirrhosis ($$P \leq 0.023$$), AFP level ($$P \leq 0.011$$), tumor size ($$P \leq 0.001$$), CLIP score ($P \leq 0.001$), BCLC stage ($P \leq 0.001$), AJCC stage ($P \leq 0.001$), and a high AGR2 level ($$P \leq 0.002$$) were significant predictors of worse OS (Table 3). Multivariate analysis showed that cirrhosis ($$P \leq 0.034$$, HR = 4.563, CI = 1.122–18.555), BCLC stage ($P \leq 0.001$, HR = 2.928, CI = 1.908–4.494), and a high AGR2 level ($$P \leq 0.008$$, HR = 1.735, CI = 1.154–2.609) were independently associated with OS (Table 3).Table 1Association of AGR2 level (Roessler liver array) with clinicopathologic indicators of hepatocellular carcinomaFactorsGroupAGR2 (mean ± SE)PAge < 60 years4.3208 ± 0.14740.900 ≥ 60 years4.3935 ± 0.3107SexMale4.3334 ± 0.14220.926Female4.3525 ± 0.3850CirrhosisAbsent3.6259 ± 0.30761.162Present4.3963 ± 0.1415Serum AFP < 3003.8669 ± 0.14460.002*(ng/ml) ≥ 3004.8537 ± 0.2261Tumor size < 5 cm4.1165 ± 0.14620.055 ≥ 5 cm4.7315 ± 0.2591CLIP0–14.2594 ± 0.14290.420 ≥ 24.6445 ± 0.3413BCLC0–14.1908 ± 1.93420.094 ≥ 24.8530 ± 2.4513AJCC stageI4.1535 ± 1.82910.381 ≥ II4.4955 ± 2.2573*$P \leq 0.05.$ CLIP, Cancer of the Liver Italian Program score; BCLC, Barcelona Clinic Liver Cancer staging; AJCC, American Joint Committee on Cancer 2017; AFP, alpha-fetoproteinTable 2Prognostic significance of clinicopathologic indicators and AGR2 for recurrence-free survival in the Roessler liver arrayFactorRFS univariateRFS multivariateGroupHR$95\%$ CIPHR$95\%$ CIPAge < 60/ ≥ 60 years0.9520.628–1.4430.817SexFemale/Male2.3591.238–4.4930.009*2.1241.112–4.0580.023*Cirrhosis−/ + 2.0030.936–4.2870.074Serum AFP < 300/ ≥ 300 ng/ml1.3140.937–1.8420.113Tumor size < 5/ ≥ 5 cm1.4241.008–2.0120.045*NSCLIP0–1/ ≥ 21.8721.267–2.7660.002*NSBCLC0–1/ ≥ 22.4321.670–3.543 < 0.001*1.6531.077–2.5350.021*AJCC stageI/ ≥ II2.0631.405–3.029 < 0.001*1.6071.083–2.3860.019*AGR2Low/High1.6911.202–2.3780.003*1.5721.112–2.2240.010**$P \leq 0.05.$ RFS, recurrence-free survival; CLIP, Cancer of the Liver Italian Program score; BCLC, Barcelona Clinic Liver Cancer staging; AJCC, American Joint Committee on Cancer 2017; AFP, alpha-fetoproteinTable 3Prognostic significance of clinicopathologic indicators and AGR2 for overall survival in the Roessler liver arrayFactorRFS univariateRFS multivariateGroupHR$95\%$ CIPHR$95\%$ CIPAge < 60/ ≥ 60 years0.9900.972–1.0080.990SexFemale/Male1.8580.901–3.8330.094Cirrhosis−/ + 5.0931.255–20.6710.023*4.5631.122–18.5550.034*Serum AFP < 300/ ≥ 300 ng/ml1.6861.126–2.5270.011*Tumor size < 5/ ≥ 5 cm1.9601.309–2.9330.001*CLIP0–1/ ≥ 22.8111.832–4.313 < 0.001*BCLC0–1/ ≥ 23.1762.081–4.846 < 0.001*2.9281.908–4.494 < 0.001*AJCC stageI/ ≥ II2.2781.483–3.500 < 0.001*AGR2Low/high1.9191.282–2.8730.002*1.7351.154–2.6090.008**$P \leq 0.05.$ OS, overall survival; CLIP, Cancer of the Liver Italian Program score; BCLC, Barcelona Clinic Liver Cancer staging; AJCC, American Joint Committee on Cancer 2017; AFP, alpha-fetoprotein Moreover, our HCC specimen cohort was also analyzed. We utilized qRT–PCR to examine the levels of AGR2, and a median level of 15.3 (39-△Ct) was defined as the cutoff to divide the HCC specimens into high and low AGR2 expression groups. Similar results were observed; high AGR2 levels were related to significantly worse OS and RFS rates (Fig. 1J, K). Furthermore, the correlation between AGR2 expression and various clinical parameters in our collected cohort was analyzed (Tables 4–6). A higher AGR2 level in HCC was observed in female patients ($$P \leq 0.007$$) (Table 4). A high AGR2 level (39-△Ct ≥ 10.8) in HCC was significantly associated with worse OS ($$P \leq 0.016$$) (Fig. 1J) and RFS ($$P \leq 0.045$$) (Fig. 1K). Univariate analysis showed that cirrhosis ($$P \leq 0.022$$), AFP level ($$P \leq 0.033$$), vascular invasion ($$P \leq 0.003$$), AJCC stage ($$P \leq 0.003$$), and a high AGR2 level ($$P \leq 0.047$$) were significant predictors of worse RFS (Table 5). Multivariate analysis showed that cirrhosis ($$P \leq 0.023$$, HR = 1.600, CI = 1.067–2.401), vascular invasion ($$P \leq 0.005$$, HR = 1.825, CI = 1.198–2.780), and a high AGR2 level ($$P \leq 0.043$$, HR = 1.662, CI = 1.017–2.716) were independently associated with RFS (Table 5). For OS, univariate analysis showed that cirrhosis ($$P \leq 0.020$$), AFP level ($$P \leq 0.012$$), vascular invasion ($P \leq 0.001$), AJCC stage ($$P \leq 0.007$$), and a high AGR2 level ($$P \leq 0.018$$) were significant predictors of worse OS (Table 6). Multivariate analysis showed that cirrhosis ($$P \leq 0.017$$, HR = 1.655, CI = 1.093–2.505), vascular invasion ($$P \leq 0.001$$, HR = 2.200, CI = 1.392–3.476), and a high AGR2 level ($$P \leq 0.015$$, HR = 1.945, CI = 1.138–3.324) were independently associated with OS (Table 6).Table 4Association of AGR2 expression with clinicopathologic indicators of hepatocellular carcinomaFactorsGroupAGR2 (mean ± SE)PAge < 60 years15.8713 ± 0.70770.659 ≥ 60 years16.2616 ± 0.4603SexMale15.4998 ± 0.47550.007*Female17.6959 ± 0.5801Hepatitis viralAbsent15.9186 ± 0.60850.812infectionPresent16.2114 ± 0.4827CirrhosisAbsent16.1498 ± 0.46870.989Present16.0949 ± 0.6798Serum AFP < 20015.8232 ± 0.52540.403(ng/ml) ≥ 20016.5543 ± 0.5651TumorW14.4390 ± 1.53580.240differentiationM-P16.2353 ± 0.3985Tumor size < 5 cm15.7190 ± 0.57350.482 ≥ 5 cm16.3917 ± 0.5167VascularAbsent15.5317 ± 0.67400.198invasionPresent16.5114 ± 0.4624AJCC stageI15.3584 ± 0.86620.241 ≥ II16.3968 ± 0.4242*$P \leq 0.05.$ Tumor differentiation by WHO; AJCC, American Joint Committee on Cancer 2017; AFP, alpha-fetoproteinTable 5Prognostic significance of clinicopathologic indicators and AGR2 for recurrence-free survival in the clinical cohortFactorRFS univariateRFS multivariateGroupHR$95\%$ CIPHR$95\%$ CIPAge < 60/ ≥ 60 years1.079(0.713–1.632)0.720SexMale/female1.059(0.672–1.670)0.804Viral infection0.248No/B1.667(0.977–2.844)0.061No / C1.274(0.739–2.915)0.384No / B + C1.647(0.835–3.262)0.152Cirrhosis-/ + 1.602(1.069–2.401)0.022*1.600(1.067–2.401)0.023*Serum AFP < 200/ ≥ 200 ng/ml1.548(1.037–2.312)0.033*NSDifferentiationW/M-P3.053(0.964–9.670)0.058Tumor size < 5/ ≥ 5 cm1.161(0.764–1.765)0.484Vascular invasion−/ + 1.899(1.248–2.890)0.003*1.825(1.198–2.780)0.005*AJCC stageI/ ≥ II2.100(1.282–3.440)0.003*NSAGR2Low/high1.638(1.006–2.670)0.047*1.662(1.017–2.716)0.043**$P \leq 0.05.$ DFS, disease-free survival; Tumor differentiation according to WHO system; AFP, alpha-fetoprotein; AJCC, American Joint Committee on Cancer 2017Table 6Prognostic significance of clinicopathologic indicators and AGR2 for overall survival in the clinical cohortFactorOS univariateOS multivariateGroupHR$95\%$ CIPHR$95\%$ CIPAge < 60/ ≥ 60 years0.975(0.638–1.490)0.907SexMale/female1.407(0.906–2.186)0.936Viral infection0.281No/B1.546(0.890–2.685)0.122No / C1.014(0.573–1.794)0.962No / B + C1.433(0.704–2.916)0.322Cirrhosis−/ + 1.632(1.080–2.466)0.020*1.655(1.093–2.505)0.017*Serum AFP < 200/ ≥ 200 ng/ml1.692(1.122–2.551)0.012*NSDifferentiationW/M-P1.330(0.851–2.079)0.210Tumor size < 5/ ≥ 5 cm1.101(0.720–1.683)0.658Vascular invasion−/ + 2.302(1.458–3.635) < 0.001*2.200(1.392–3.476)0.001*AJCC stageI/ ≥ II2.076(1.224–3.521)0.007*NSAGR2Low/high1.894(1.114–3.221)0.018*1.945(1.138–3.324)0.015**$P \leq 0.05.$ OS, overall survival; Tumor differentiation according to WHO system; AFP, alpha-fetoprotein; AJCC, American Joint Committee on Cancer 2017 Moreover, immunohistochemical staining was utilized to examine the expression of AGR2 in 24 clinical HCC tissues, and the results indicated that AGR2 was highly expressed in tumor tissues compared to normal tissues (Fig. 1L, M). Overall, based on the evidence, we found that patients with lower AGR2 expression have a better OS rate; hence, we suggest that AGR2 might play an oncogenic role in HCC progression and thus might be a useful prognostic marker of HCC progression.
## Sorafenib decreases cell viability and increases cell apoptosis
First, to determine the effect of sorafenib on cell viability, HCC cell lines were treated with various doses of sorafenib (5–10 μM, 24–48 h). Cell viability was significantly decreased with sorafenib treatment in a dose-dependent manner in J7, Hep3B, HepG2 and Huh7 cells according to MTT assay results (Fig. 2A–D). Moreover, flow cytometry was utilized to determine whether sorafenib influences HCC cell apoptosis. HepG2 and Huh7 cells were stimulated with 5 and 10 μM sorafenib for 24 h. Cell apoptosis was slightly induced with 5 μM sorafenib; however, the increase in the apoptosis rate compared with that in the control reached approximately $25\%$ after 10 μM sorafenib stimulation in both HepG2 and Huh7 cells (Fig. 2E–H). Based on these results, we found that sorafenib can modulate HCC cell viability and apoptosis ability; subsequently, we evaluated whether AGR2 is involved in sorafenib-regulated phenotypes. Fig. 2Sorafenib decreases cell viability, increases cell apoptosis and induces AGR2 secretion in HCC. A–D The viability of J7, Hep3B, HepG2, and Huh7 HCC cells treated with 5 and 10 μM sorafenib for 24–48 h was examined using MTT assay. E–H *Cell apoptosis* was determined in HepG2 and Huh7 cells after stimulation with 5 and 10 μM sorafenib. The quantification of apoptotic cells is shown in F, H. Sorafenib decreases cell viability and increases cell apoptosis in HCC cells. I–R AGR2 RNA I–L and protein M–R levels, both intracellular M–P and extracellular Q, R, were examined via RT–PCR and Western blotting after 5 and 10 μM sorafenib treatment for 24–48 h. Sorafenib induces AGR2 secretion from the cytosol into conditioned medium rather than exerting transcriptional or translational regulation (lane 1, 4:untreatment; lane 2, 3, 5, 6:sorafenib treatment). Ponceau S was used as an internal control
## Sorafenib induces AGR2 secretion instead of transcriptional regulation
After we retrieved the GSE94550 dataset, AGR2 was found to be induced in sorafenib-resistant Huh7 cells compared to parental cells (Fig. 1A). To further examine whether sorafenib can regulate AGR2 expression in HCC, RT–PCR and Western blotting were applied in parental HCC cell lines treated with sorafenib. First, we found that the RNA level of AGR2 was not altered by sorafenib in J7, HepG2, Huh7 and Hep3B cells (Fig. 2I-L; Additional file 1: Figure S1A–D). However, the protein level of AGR2 in the lysates of these cells was unexpectedly decreased in a dose-dependent manner after sorafenib stimulation, as shown by Western blotting (lane 3 vs 1, lane 6 vs 4; Fig. 2M–P; Additional file 1: Figure S1E–H). Based on these contradictory results, sorafenib regulated the RNA and protein levels of AGR2 in parental HCC cells. Several reports have shown that AGR2 is an ER-resident protein that is also localized in the extracellular matrix, blood and urine [21, 22]. Therefore, we collected CM after 5 and 10 μM sorafenib treatments for 24–48 h. As expected, AGR2 was detected in CM from HepG2 and Hep3B cells treated with sorafenib (lane 2, 3 vs 1, lane 5, 6 vs 4; Fig. 2Q, R; Additional file 1: Figure S1I-J). Based on this evidence, we suggest that sorafenib regulates AGR2 through posttranslational modification, not transcriptional regulation, in parental HCC cells.
## AGR2 plays a role in cell viability and apoptosis
To analyze the roles of AGR2 in HCC progression, we established AGR2-silenced Hep3B, HepG2 and Huh7 cells (lane 2 vs 1; Fig. 3A(a), B(a), C(a); Additional file 2: Figure S2), and cell viability was determined using MTT assay. Cell viability was significantly decreased after AGR2 silencing, and the effect was more obvious when AGR2 silencing was combined with sorafenib treatment (Fig. 3A(b), B(b), C(b)). These findings suggest that AGR2 plays a role in promoting cancer progression. Additionally, we investigated whether AGR2 affects cell apoptosis. Flow cytometry analysis was utilized to demonstrate that sorafenib can induce significant cell apoptosis (approximately $10\%$) in Hep3B, HepG2 and Huh7 cells compared with control cells, which was more conspicuous after AGR2 silencing in the presence of sorafenib compared to the control (siNC) (Fig. 3D, E, F). Quantitative results are shown in panels 3D (b), E (b), and F (b). Apoptotic cells appeared among the siNC cells under untreated conditions, and we speculate that apoptosis might have been induced in these cells by the transfection process or the apoptosis assay procedures. However, the phenomenon of sorafenib-treated or AGR2-silenced cell apoptosis was not influenced by the basal apoptosis signal in untreated cells. The sorafenib-treated and AGR2-silenced cell apoptosis rates were normalized to the value in the untreated basal cells. However, we found that AGR2 can be secreted into CM and detected by Western blotting (CM, Fig. 2Q, R). Previously, Fessart et al. reported that extracellular AGR2 can be defined as an extracellular matrix pro-oncogenic regulator that makes cancer cells more aggressive [21]. Hence, we evaluated whether the addition of recombinant AGR2 protein regulates cell viability and apoptosis. Flow cytometry analysis indicated that 10 μM sorafenib can induce up to 30–$40\%$ cell apoptosis at 24 h in both HepG2 and Huh7 cells; however, the phenomenon was reversed with 60 ng/ml recombinant AGR2 (rAGR2) (Fig. 3G, H). Based on this evidence, we suggest that AGR2 secreted into CM in response to sorafenib stimulation plays an oncogenic role in HCC cancer progression. Fig. 3AGR2 is implicated in sorafenib-regulated cell viability and apoptosis. A–C The viability of Hep3B, HepG2, and Huh7 cells with silenced AGR2 was examined after 24 h in the presence or absence of sorafenib (5 μM) using MTT assay. The AGR2 levels (A(a), B(a), C(a)) and cell viability (A(b), B(b), C(b)) results are presented. AGR2 supports HCC cell viability. D–H HCC cell apoptosis was assessed after AGR2 silencing D–F or stimulation with 60 ng/ml recombinant AGR2 (rAGR2) G, H in the presence of 10 μM sorafenib. The quantification of apoptotic cells is shown (D(b), E(b), F(b)). AGR2 inhibition, using either siRNA transfection or recombinant protein treatment, decreases cell apoptosis. siNC: negative control siRNA, siRNA vector only; siAGR2: AGR2 siRNA
## Sorafenib induces ER stress in HCC
We determined the functions and regulatory mechanisms through which sorafenib influences AGR2 activity. Based on a literature search, AGR2 has been demonstrated to be upregulated upon ER stress, and ER stress-related molecules, such as protein kinase R (PKR)-like endoplasmic reticulum kinase (PERK), inositol-requiring enzyme 1 (IRE1) and activating transcription factor 6 (ATF6), are dysregulated in many cancer types [16]. Therefore, the relationship between sorafenib and ER-related factors was determined. Among these molecules, using Western blotting, we found that phospho-IRE1α (p-IRE1α) was upregulated by 10 μM sorafenib treatment in HepG2 and Huh7 cells (lane 3 vs 1; Fig. 4A; Additional file 3: Figure S3). The Bip protein level was determined using Western blotting after sorafenib treatments at numerous concentrations in J7 and Huh7 cells. However, Bip expression was not altered by sorafenib (lane 2, 3 vs 1; Additional file 4: Figure S4). Moreover, X-box binding protein 1 (XBP1) has been reported as a unique transcription factor that modulates ERAD gene expression and promotes protein folding [33]. In the UPR, IRE1α is activated via oligomerization and autophosphorylation, followed by the activation of its endoribonuclease to cleave and splice XBP1. The activated IRE1α endoribonuclease can remove 26 nucleotides from the intron of XBP1, converting XBP1 from preform XBP1 (XBP1 u: unspliced) to activated XBP1 (XBP1 s: spliced) [34]. Therefore, RT–PCR was used to detect the status of the IRE-1 downstream factor XBP1. Through RT–PCR analysis, we found that XBP1 was spliced from inactive XBP1 u to active XBP1 s after stimulation with sorafenib for 24 and 48 h in HepG2 and Huh7 cells (lane 3 vs 1, lane 6 vs 4; Fig. 4B, C; Additional file 5: Figure S5A, B). The data showed two bands: the upper band is full-length (unspliced, u) XBP1 (XBP1 u), and the lower band is spliced (s) XBP1 (XBP1 s). Therefore, we found that spliced XBP1 was increased and unspliced XBP1 was decreased after sorafenib treatment for 24 and 48 h. Subsequently, we sought to verify whether AGR2 plays a role in the conversion from inactive XBP1 u to active XBP1 s. As expected, the sorafenib-induced XBP1 s levels were more robust in HepG2 and Huh7 cells with silenced AGR2 (siAGR2) compared with vector control (siNC) under sorafenib treatment (lane 4 vs 2; Fig. 4D; Additional file 5: figure S5C, D). In contrast, we further verified whether added recombinant AGR2 (rAGR2) could modulate the splicing of inactive XBP1 u to activate XBP1 s. Similarly, the levels of XBP1 s induced by 10 μM sorafenib were reduced after stimulation with 60 ng/ml rAGR2 in HepG2 and Huh7 cells (lane 4 vs 3; Fig. 4E; Additional file 5: figure S5E, F). Moreover, we analyzed AGR2 and XBP1 expression in our HCC clinical specimen cohort using RT‒PCR. We analyzed 9 normal liver tissues and 5 HCC tumor tissues. AGR2 was slightly more highly expressed in HCC tumor tissues than in normal tissues, and the spliced XBP1 (XBP1 s) level was slightly more highly expressed in HCC tissues. We suggest that the correlation between AGR2 and XBP1 needs to be further demonstrated in more clinical specimens in the future (Normal: lanes 1–9, Tumor: lanes 1–5; Additional file 6: figure S6). Collectively, these findings indicate that sorafenib induces HCC ER stress via the IRE1α-XBP1 cascade through AGR2 regulation. Fig. 4Sorafenib regulates ER stress-related molecules. A–E The protein A and RNA B, C, D, E levels of ER stress-related molecules were examined by Western blotting A and RT–PCR B, C, D, E after treatment of HepG2 A, D, E (left), B and Huh7 A, D, E (right), C cells with 5 μM and 10 μM sorafenib with or without AGR2 silencing D and 60 ng/ml recombinant AGR2 (rAGR2) stimulation E. Sorafenib induces the dysregulation of several ER stress-related molecules. AGR2 is involved in sorafenib-induced XBP-1 splicing. C: cleavage; u: unspliced; s: spliced; siNC: vector control siRNA; siAGR2: AGR2 siRNA. 0: untreatment; 5, 10: sorafenib treatment
## AGR2 plays diverse roles
To determine whether AGR2 plays diverse roles in resistant sublines compared to sorafenib-sensitive HCC cells, HepG2 sorafenib-resistant (HepG2-SR) and Huh7 sorafenib-resistant (Huh7-SR) cells were established (Fig. 5A, B). We applied 7 µM sorafenib to the culture medium to generate sorafenib-resistant cell lines for the following experiments. Using MTT assay, we found that cell viability was decreased by approximately $50\%$ after sorafenib treatment (7 μM) in parental HepG2 and Huh7 cells (indicated as PCs); however, sorafenib only reduced cell viability by 10–$20\%$ in resistant cells (indicated as SR cells) (Fig. 5A, B), indicating that these SR cell lines are protected against sorafenib challenge. We found that AGR2 was related to resistance and upregulated by sorafenib in the GSE94550 dataset (Fig. 1A, B). Therefore, we verified whether this regulation was observed in our sorafenib-resistant cells. As expected, AGR2 was highly expressed intracellularly in sorafenib-resistant HepG2 and Huh7 cells compared with parental cells (lane 2 vs 1; Fig. 5C; Additional file 7: figure S7A). We also found that sorafenib reduced intracellular AGR2 levels in HepG2-SR and Huh7-SR cells (lane 3 vs 1; Fig. 5D; Additional file 7: figure S7B), and the tendency was similar to that of sorafenib-regulated AGR2 in HepG2-PCs and Huh7-PCs (Fig. 2M–P). Moreover, we found that sorafenib can induce AGR2 secretion in CM in HepG2-SR and Huh7-SR cells compared with controls. The effect was stronger in resistant cells than in parental cells (lane 4 vs 3; Fig. 5E; Additional file 7: figure S7C). Collectively, these results indicate that AGR2 induction was more robust in both the cell lysate and CM of sorafenib-resistant cells than in parental cells in the presence and absence of sorafenib. This finding might explain why SR cells are protected against sorafenib. Fig. 5AGR2 regulation in sorafenib-sensitive and sorafenib-resistant cells. A–E Cell viability A, B and AGR2 regulation C–E in both cell lysate C, D and CM (CM, E) from parental cells (PC) and sorafenib-resistant (SR) cells treated with sorafenib. Sorafenib-resistant cells show higher cell viability, and AGR2 is highly expressed in sorafenib-resistant cells compared to sorafenib-sensitive cells. 0: untreatment; 5, 10: sorafenib treatment
## Sorafenib-resistant cells modulate ER stress and reduce cell apoptosis
To determine whether sorafenib-resistant cells can resist the effect of sorafenib-induced cell apoptosis, flow cytometry analysis was utilized to demonstrate that cell apoptosis was induced after stimulation with various doses of sorafenib for 24 h in HepG2 (Fig. 6A, B) and Huh7 (Fig. 6C, D) parental cells. The ratio of cells undergoing sorafenib-induced apoptosis reached approximately $30\%$ among parental cells; however, the effect was not observed in sorafenib-resistant cells (Fig. 6A–D). Based on this evidence, we suggest that these resistant cells are protected against sorafenib toxicity, which prolongs cancer cell viability. Furthermore, we analyzed whether AGR2 plays a vital role in inducing HCC resistance to sorafenib. We silenced AGR2 in HepG2 SR and Huh7 SR cells and performed an apoptosis assay. The results indicated that sorafenib can induce cell apoptosis, but the effect was more robust after AGR2 knockdown (Fig. 6E–H). According to these findings, we speculate that AGR2 plays a critical role in inducing sorafenib resistance in HCC.Fig. 6AGR2 is involved in HCC resistance to sorafenib. A–D Apoptosis assay performed after stimulation of both HepG2 and Huh7 parental cell s (PC) and sorafenib-resistant (SR) cells with sorafenib at various concentrations (0–10 μM). The apoptosis rate was lower in sorafenib-resistant cells (indicated as SR cells) than in parental cells (indicated as PCs). Sorafenib-resistant cells have higher protection and resistance to sorafenib-induced toxicity. E–H Apoptosis assay performed in HepG2 SR and Huh7 SR cells with AGR2 silencing in the presence of sorafenib (10 μM). More apoptotic of HepG2 and Huh7 cells were observed in the AGR2-silenced SR group than in the vector control group. Sorafenib-induced apoptosis is more obvious with AGR2 silencing. siNC: negative control siRNA, siRNA vector only; siAGR2: AGR2 siRNA We found that the ER stress-related molecule p-IRE1α was induced by sorafenib treatment; subsequently, we determined whether p-IRE1α regulation occurred in sorafenib-resistant cells. Western blot analysis revealed that the levels of p-IRE1α were decreased after sorafenib stimulation in SR HepG2 and Huh7 cells and that the effect was inconsistent in PCs (lane 2 vs 1; Fig. 7A, B; Additional file 8: Figure S8A). Moreover, according to RT–PCR results, we found that the sorafenib-induced changes in the levels of XBP1 s, a downstream factor of IRE1α, were abolished or attenuated in HepG2-SR and Huh7-SR cells and that this was not observed in parental HepG2 and Huh7 cells (lane 6 vs 3; Fig. 7C; Additional file 8: Figure S8B). Furthermore, we evaluated whether the effect was mediated by AGR2 in sorafenib-resistant cells. The data indicated that XBP1 s expression was induced after silencing AGR2 expression and was further elevated after stimulation with sorafenib (lane 4 vs 2; Fig. 7D; Additional file 8: Figure S8C). We analyzed the cleaved (C) caspase3 and AGR2 levels in tumors from nude mice subcutaneously injected with Huh7 PCs and SR cell lines, both treated with sorafenib (defined in the figure as sora PC and sora SR, respectively). Immunohistochemistry results indicated that AGR2 was more highly expressed in sora-treated SR cells than in sora-treated PC cells; in contrast, c-casp3 was slightly more highly expressed in sora-treated PC cells than in sora-treated SR cells (Additional file 9: Figure S9). This suggests that AGR2 is more essential under cell stress conditions. Collectively, the contradictory evidence in sorafenib-sensitive and sorafenib-resistant cells may indicate that cells can resist sorafenib toxicity by modulating ER stress-related molecules to decrease cellular ER stress. Fig. 7AGR2 is involved in the sorafenib-regulated IRE1α-XBP1 cascade. A, B, C, D The RNA C, D and protein A, B levels of the ER stress-related molecules p-IRE1α, IRE1α A, B, and XBP1 (u and s, C, D) were examined via Western blotting A, B and RT–PCR C, D after AGR2 silencing D and treatment with sorafenib at concentrations ranging from 0 to 10 μM in both parental (PC) and sorafenib-resistant (SR) HepG2 (A, (C, D, left)) and Huh7 (B, (C, D, right)) cells. Sorafenib-induced p-IRE-1α, IRE-1α regulation and XBP-1 splicing in sorafenib-sensitive cells are attenuated in sorafenib-resistant cells. u: unspliced; s: spliced; siNC: negative control siRNA, siRNA vector only; siAGR2: AGR2 siRNA. 0: untreatment; 7, 10: sorafenib treatment Based on the evidence, we propose that sorafenib reduces cell viability and induces cell apoptosis via downregulation of AGR2 in the cell lysate and increased secretion in CM, which induces ER stress via upregulation of p-IRE1α and spliced XBP1 in HCC. However, the phenomenon of sorafenib-induced apoptosis was abolished in sorafenib-resistant cells, and this effect may occur through increased AGR2 expression in cell lysates and downregulation of p-IRE1α and spliced XBP1. Overall, AGR2 might modulate ER stress to protect cells from sorafenib toxicity and extend cell survival (Fig. 8).Fig. 8Proposed model of the roles of AGR2 in sorafenib-sensitive and sorafenib-resistant HCC. AGR2 plays different roles in sorafenib-sensitive (PC) and sorafenib-resistant (SR) cells. Sorafenib decreases AGR2 expression in the cell lysate and in turn induces secretion into CM. Sorafenib reduces cell viability and induces cell apoptosis. However, AGR2 is highly expressed in sorafenib-resistant cells, and the induction of AGR2 expression in CM was more robust in sorafenib-resistant cells than in parental cells, which in turn induces cell survival and suppresses cell apoptosis
## Discussion
In the present study, we found that AGR2 is significantly correlated with OS, RFS, and various clinical parameters, including AFP, ALT, the predicted risk metastasis signature score, tumor size, and pathological stage. The 160 clinical HCC specimens were divided into high and low groups according to the AGR2 level determined by qRT‒PCR, with a cut-off at 10.8 (39-ΔCt). The 120 patients with AGR2 levels ≥ 10.8 (39-△Ct) presented worse recurrence-free-survival ($$P \leq 0.045$$) (Table 5, Fig. 1K) and overall survival rates ($$P \leq 0.016$$) (Table 6, Fig. 1J) than the 40 patients with AGR2 levels ≤ 10.8 (39-△Ct). However, the AGR2 levels in HCC patients before and after sorafenib treatment remain unknown. Furthermore, we intend to investigate the AGR2 levels in patients treated with and without sorafenib to analyze the ratio of AGR2 levels under both conditions, which might provide a precise ratio for chemotherapy outcomes and aid in determining the clinical prognosis of patients with HCC. We utilized sorafenib-sensitive and sorafenib-resistant cells to determine the roles of AGR2 in HCC progression and drug resistance. We suggest that AGR2 plays diverse roles and has different mechanisms through which it influences HCC progression and sorafenib resistance in these two models. Functionally, AGR2 knockdown reduces cell viability and induces cell apoptosis with sorafenib treatment in parental HCC cell lines. However, the phenomenon of sorafenib-induced cell apoptosis in sorafenib-sensitive cells is almost abolished in sorafenib-resistant cells. Mechanistically, sorafenib modulates AGR2 through posttranslational modification instead of transcriptional regulation and activates the IRE1α-XBP1 cascade to induce death in parental cells, but this effect is not observed in sorafenib-resistant cells. This is the first report to uncover the role of AGR2 in sorafenib-resistant HCC and to explain how AGR2 induces HCC resistance to sorafenib and reduces cell apoptosis.
The regulation of ER stress-related molecules by sorafenib is also diverse in these two models. Collectively, our preliminary data highlight a novel regulatory mechanism of AGR2 that may serve as a critical determinant of cancer progression and drug resistance in HCC. Cancer cells often initiate ER stress via misfolded protein accumulation in the ER due to protein overexpression, nutrient deprivation or hypoxia. Cells can activate UPR signaling to trigger ER homeostasis and prolong cell survival [35]. AGR2 is an ER-resident protein that catalyzes thiol-disulfide interchange and protein folding reactions [11, 12]. Our results showed that dysregulated AGR2 expression and ER stress-related molecules appear in sorafenib-sensitive and sorafenib-resistant cells. Decreased AGR2 and increased p-IRE1α and spliced XBP1 levels induced by sorafenib were observed in the cell lysate of sorafenib-treated sensitive cells; in contrast, increased AGR2 in CM and decreased p-IRE1α and spliced XBP1 expression induced by sorafenib appeared in sorafenib-treated resistant cells. Blazanin and colleagues reported that v-Ha-Ras, an oncogenic protein, can induce ER stress partially through upregulation of both the mRNA and protein levels of total IRE1α and that phosphorylated IRE1α was also induced by ER stress [36, 37]. Moreover, keratinocytes treated with the ER stress inhibitor 4-phenyl butyric acid (4-PBA) exhibited increased levels of both total IRE1α and phosphorylated IRE1α [36]. Based on the evidence, we suggest that the mRNA and protein levels of IRE1α will be upregulated, followed by upregulation of phosphorylated IRE1α, upon ER stress. Our findings of sorafenib-induced ER stress via upregulation of both total and phosphorylated IRE1α are similar to the findings of other groups. According to these diverse effects, we speculate that the ER stress level differed in these two cell models. Sorafenib induced higher ER stress in parental HCC cells; however, sorafenib-induced ER stress was attenuated in sorafenib-resistant cells, indicating that ER stress might be a critical factor in determining whether cells can resist sorafenib. Therefore, we suggest that ER stress homeostasis is a therapeutic target to influence the status of sorafenib resistance in HCC.
In the present study, cell viability and cell apoptosis were altered in Hep3B, HepG2 and Huh7 parental cells upon sorafenib treatment after silencing AGR2. AGR2 contributes to ER homeostasis via UPR signaling, including the IRE1α-XBP1 cascade [29]. In the present study, we found that silencing AGR2 in the presence or absence of sorafenib increased spliced XBP1 levels in both sorafenib-sensitive and sorafenib-resistant cells. We further found that exogenous recombinant AGR2 can reduce sorafenib-induced XBP1 splicing. However, whether XBP1 is an upstream effector that modulates AGR2 expression during HCC progression and resistance to sorafenib is unclear. Some ER stress inhibitors, including tauroursodeoxycholic acid (TUDCA, 100 μM, [38]) and the IRE1α endonuclease inhibitor MKC-3946 (10 μM, [39]), can be used to inhibit ER stress to further delineate the relationship between AGR2 and ER stress in both sorafenib-sensitive and sorafenib-resistant cells. Collectively, our findings provide a new mechanism by which AGR2 might act as an upstream factor of XBP1 to modulate ER homeostasis and influence the cell death or survival status in sorafenib-sensitive and sorafenib-resistant HCC.
We preliminarily identified the functions and regulatory mechanisms of AGR2 in the response to sorafenib. AGR2 has been demonstrated to be upregulated upon ER stress, and ER stress-related molecules, such as PERK, IRE1 and ATF6, are dysregulated in many cancer types [16]. Herein, we demonstrated that AGR2 can modulate the IRE1α-XBP1 cascade to modulate ER homeostasis, switching HCC from the sorafenib-sensitive to the sorafenib-resistant type. However, in this study, other ER stress-related signaling molecules were shown to be affected by sorafenib; the levels of total ATF6, cleaved ATF6 and p-PERK were regulated by sorafenib, as shown by Western blotting, but the regulatory effects of sorafenib were slightly weaker than those of IRE1α. Hence, we speculate that ATF6 and PERK may constitute another potential pathway regulated by AGR2 that influences cancer cell progression and resistance to sorafenib. We used the cBioPortal software for Cancer Genomics developed by the Memorial Sloan Kettering Cancer Center (MSKCC) [40–42]. AGR2 expression has been demonstrated to be negatively correlated with ATF6 expression (Spearman: − 0.19, p value: 2.15e-4) based on the TCGA, Filehorse microarray dataset. Moreover, based on the TCGA, Pancancer microarray dataset, a negative correlation between AGR2 and ATF6 expression was also reported (Spearman: − 0.2, p value: 9.174e-5). Based on the online published microarray datasets, we suggest that AGR2 has a significant correlation with ER stress in HCC. According to the cBioPortal software information, several correlations between AGR2 and ER stress-related factors have been demonstrated in numerous cancer types, such as HCC and lung, breast and pancreatic cancers. The AGR2 mRNA levels were negatively correlated with XBP1 protein levels in HCC (Spearman: − 0.169, p value: 0.02) in the TCGA, Pancancer microarray dataset and in lung squamous cell carcinoma (Spearman: − 0.16, p value: 3.65e-3). Furthermore, a negative correlation between AGR2 and ATF6 was also reported in TCGA, lung squamous cell carcinoma (Spearman: − 0.104, p value: 0.02) and TCGA, pancreatic (Spearman: − 0.32, p value: 5.015e-3) cancer. Based on analyses of these published microarray datasets, AGR2 is highly correlated with ER stress-related molecules, such as XBP-1 and ATF6, and these results were similar to our findings, making our study more complete. In conclusion, we suggest that the AGR2-IRE1α-XBP1 cascade is an ER-related pathway that regulates HCC progression; hence, this signaling cascade might be a potential therapeutic target for curing sorafenib-resistant HCC in the future.
Previous research has shown that intracellular AGR2 (iAGR2) can promote cancer cell growth and survival and that extracellular AGR2 (eAGR2) can be defined as a microenvironment regulator that makes cancer cells more aggressive [21, 29]. We found that AGR2 can be detected in CM from both sorafenib-sensitive and sorafenib-resistant cells, and the induction ratio with sorafenib treatment was more robust in resistant cells than in sensitive cells. However, the roles of iAGR2 and eAGR2 in the presence of sorafenib are still unclear. Fessart et al. reported that extracellular AGR2 is an extracellular matrix pro-oncogenic regulator that makes cancer cells more aggressive [21]. AGR2 has been demonstrated to have numerous domains, contributing to diverse functions in cancer cells [32]. However, whether these domains are associated with cancer progression and drug resistance in the presence of sorafenib is not clear. The functions of AGR2 domains in both sorafenib-sensitive and sorafenib-resistant cells need to be elucidated in more detail using truncated AGR2 mutants. Two truncated AGR2 forms, including deletions of amino acids (AAs) 1-20 and AAs 172-175, which can be localized to the extracellular space [32], can be utilized in the future. Using these AGR2 mutants, we will be able to determine whether the subcellular location of AGR2 plays a critical role in regulating cancer progression and sorafenib resistance.
Sorafenib has been demonstrated to inhibit numerous receptor tyrosine kinases, such as VEGFR and PDGFR [4]. Moreover, extracellular AGR2 has been shown to bind directly to VEGF to enhance tumor angiogenesis and other activities [43]. We found that AGR2 can be secreted into CM after sorafenib stimulation of sorafenib-resistant HepG2 and Huh7 cells, but this was not observed in parental cells. However, the mechanism underlying AGR2-induced resistance to sorafenib in HCC has never been elucidated. Hence, it is necessary to analyze whether recombinant AGR2 can directly interact with recombinant VEGF. Previously, several VEGFR- and PDGFR-related signaling pathways, including the RAS, RAF, MEK, ERK, PI3K/Akt and JAK-STAT pathways, have been reported to be inhibited by sorafenib [6]. However, the signaling underlying AGR2-induced resistance to sorafenib in HCC has never been elucidated. Therefore, in the future, these pathways need to be examined in sorafenib-sensitive and sorafenib-resistant cells with AGR2 silencing and AGR2 overexpression in the presence of sorafenib. Elucidation of the predictive role and molecular and cellular mechanisms of AGR2 related to sorafenib resistance can provide additional opportunities to establish complementary therapies for HCC.
## Supplementary Information
Additional file 1: Figure S1. The RNA (A–D) and protein (E–J) levels both cell lysate (E–H) and conditioned medium (I, J, CM) of AGR2 stimulated with sorafenib (5–10 μM) in J7 (A, E), HepG2 (B, F, I), Huh7 (C, G) and Hep3B (D, H, J) cells using RT-PCR (A–D) and Western blotting (E–J) were quantified (normalized with untreatment [0] control).Additional file 2: Figure S2. The AGR2 levels with transfection of AGR2 siRNA (siAGR2) and vector control (siNC) in Hep3B (A), HepG2 (B) and Huh7 (C) cells were quantified (normalized with siNC control). NC: negative control. Additional file 3: Figure S3. The protein levels of cleaved (C) ATF6, phosphate (p)-IRE1α and p-PERK with sorafenib (5-10 μM) stimulation in HepG2 (A) and Huh7 (B) cells were quantified (normalized with untreatment [0] control).Additional file 4: Figure S4. J7 and Huh7 cells were stimulated with 5 and 10 μM sorafenib, followed by examination of Bip expression using Western blotting. 0: untreatment control. Additional file 5: Figure S5. The spliced XBP1 (XBP1s) levels with sorafenib (5–10 μM) stimulation (A, B), transfection of AGR2 siRNA (siAGR2) and vector control (siNC) (C, D), or recombinant (r) AGR2 stimulation (E, F) in HepG2 (A, C, E) and Huh7 (B, D, F) cells were quantified. NC: negative control. ( normalized with untreatment [0] control)Additional file 6: Figure S6. The levels of AGR2 and unspliced (u) XBP1 and spliced (s) XBP1 were determined by RT‒PCR in 9 normal tissues and 5 HCC tissues. 18S rRNA was used as an internal control. Additional file 7: Figure S7. The protein levels of AGR2 both cell lysate (A, B) and conditioned medium (C, CM) in parental cells (PC) or sorafenib-resistant (SR) HepG2 (A-C, left) and Huh7 (A-C, right) cells in the presence or absence of sorafenib (5–10 μM) detected by Western blotting were quantified. ( normalized with untreatment [0] control)Additional file 8: Figure S8. The protein level of phosphate (p)-IRE1α (A) and RNA level of XBP1 (XBP1s) (B-C) detected by Western blotting (A) and RT-PCR (B–C) with sorafenib (7–10 μM) stimulation (A, B), transfection of AGR2 siRNA (siAGR2) and vector control (siNC) (C) in parental cells (PC) and sorafenib-resistant (SR) HepG2 (A-C, left) and Huh7 (A–C, right) cells were quantified. Additional file 9: Figure S9. The levels of AGR2 and cleaved (c) caspase3 (CASP3) were determined by immunohistochemistry in vivo in nude mice injected with sorafenib (sora)-treated Huh7 parental cells (PCs) and sora-treated Huh7 resistant (SR) cells.
## References
1. Ahn JH, Kim SJ, Park WS, Cho SY, Ha JD, Kim SS, Kang SK, Jeong DG, Jung SK, Lee SH. **Synthesis and biological evaluation of rhodanine derivatives as PRL-3 inhibitors**. *Bioorg Med Chem Lett* (2006) **16** 2996-2999. DOI: 10.1016/j.bmcl.2006.02.060
2. Kudo M, Finn RS, Qin S, Han KH, Ikeda K, Piscaglia F, Baron A, Park JW, Han G, Jassem J. **Lenvatinib versus sorafenib in first-line treatment of patients with unresectable hepatocellular carcinoma: a randomised phase 3 non-inferiority trial**. *Lancet* (2018) **391** 1163-1173. DOI: 10.1016/S0140-6736(18)30207-1
3. Adnane L, Trail PA, Taylor I, Wilhelm SM. **Sorafenib (BAY 43–9006, Nexavar), a dual-action inhibitor that targets RAF/MEK/ERK pathway in tumor cells and tyrosine kinases VEGFR/PDGFR in tumor vasculature**. *Methods Enzymol* (2006) **407** 597-612. DOI: 10.1016/S0076-6879(05)07047-3
4. Wilhelm S, Carter C, Lynch M, Lowinger T, Dumas J, Smith RA, Schwartz B, Simantov R, Kelley S. **Discovery and development of sorafenib: a multikinase inhibitor for treating cancer**. *Nat Rev Drug Discov* (2006) **5** 835-844. DOI: 10.1038/nrd2130
5. Liu L, Cao Y, Chen C, Zhang X, McNabola A, Wilkie D, Wilhelm S, Lynch M, Carter C. **Sorafenib blocks the RAF/MEK/ERK pathway, inhibits tumor angiogenesis, and induces tumor cell apoptosis in hepatocellular carcinoma model PLC/PRF/5**. *Cancer Res* (2006) **66** 11851-11858. DOI: 10.1158/0008-5472.CAN-06-1377
6. Zhu YJ, Zheng B, Wang HY, Chen L. **New knowledge of the mechanisms of sorafenib resistance in liver cancer**. *Acta Pharmacol Sin* (2017) **38** 614-622. DOI: 10.1038/aps.2017.5
7. Chen KF, Tai WT, Hsu CY, Huang JW, Liu CY, Chen PJ, Kim I, Shiau CW. **Blockade of STAT3 activation by sorafenib derivatives through enhancing SHP-1 phosphatase activity**. *Eur J Med Chem* (2012) **55** 220-227. DOI: 10.1016/j.ejmech.2012.07.023
8. Llovet JM. **Focal gains of VEGFA: candidate predictors of sorafenib response in hepatocellular carcinoma**. *Cancer Cell* (2014) **25** 560-562. DOI: 10.1016/j.ccr.2014.04.019
9. Regan-Fendt K, Li D, Reyes R, Yu L, Wani NA, Hu P, Jacob ST, Ghoshal K, Payne PRO, Motiwala T. **Transcriptomics-based drug repurposing approach identifies novel drugs against Sorafenib-resistant hepatocellular carcinoma**. *Cancers (Basel)* (2020). DOI: 10.3390/cancers12102730
10. Roessler S, Jia HL, Budhu A, Forgues M, Ye QH, Lee JS, Thorgeirsson SS, Sun Z, Tang ZY, Qin LX. **A unique metastasis gene signature enables prediction of tumor relapse in early-stage hepatocellular carcinoma patients**. *Can Res* (2010) **70** 10202-10212. DOI: 10.1158/0008-5472.CAN-10-2607
11. Galligan JJ, Petersen DR. **The human protein disulfide isomerase gene family**. *Hum Genomics* (2012) **6** 6. DOI: 10.1186/1479-7364-6-6
12. Persson S, Rosenquist M, Knoblach B, Khosravi-Far R, Sommarin M, Michalak M. **Diversity of the protein disulfide isomerase family: identification of breast tumor induced Hag2 and Hag3 as novel members of the protein family**. *Mol Phylogenet Evol* (2005) **36** 734-740. DOI: 10.1016/j.ympev.2005.04.002
13. Thompson DA, Weigel RJ. **hAG-2, the human homologue of the Xenopus laevis cement gland gene XAG-2, is coexpressed with estrogen receptor in breast cancer cell lines**. *Biochem Biophys Res Commun* (1998) **251** 111-116. DOI: 10.1006/bbrc.1998.9440
14. Wilson CL, Sims AH, Howell A, Miller CJ, Clarke RB. **Effects of oestrogen on gene expression in epithelium and stroma of normal human breast tissue**. *Endocr Relat Cancer* (2006) **13** 617-628. DOI: 10.1677/erc.1.01165
15. Hrstka R, Nenutil R, Fourtouna A, Maslon MM, Naughton C, Langdon S, Murray E, Larionov A, Petrakova K, Muller P. **The pro-metastatic protein anterior gradient-2 predicts poor prognosis in tamoxifen-treated breast cancers**. *Oncogene* (2010) **29** 4838-4847. DOI: 10.1038/onc.2010.228
16. Obacz J, Takacova M, Brychtova V, Dobes P, Pastorekova S, Vojtesek B, Hrstka R. **The role of AGR2 and AGR3 in cancer: similar but not identical**. *Eur J Cell Biol* (2015) **94** 139-147. DOI: 10.1016/j.ejcb.2015.01.002
17. Li Z, Zhu Q, Hu L, Chen H, Wu Z, Li D. **Anterior gradient 2 is a binding stabilizer of hypoxia inducible factor-1alpha that enhances CoCl2 -induced doxorubicin resistance in breast cancer cells**. *Cancer Sci* (2015) **106** 1041-1049. DOI: 10.1111/cas.12714
18. Yu H, Zhao J, Lin L, Zhang Y, Zhong F, Liu Y, Yu Y, Shen H, Han M, He F. **Proteomic study explores AGR2 as pro-metastatic protein in HCC**. *Mol Biosyst* (2012) **8** 2710-2718. DOI: 10.1039/c2mb25160d
19. Liu Z, Hu Y, Gong Y, Zhang W, Liu C, Wang Q, Deng H. **Hydrogen peroxide mediated mitochondrial UNG1-PRDX3 interaction and UNG1 degradation**. *Free Radic Biol Med* (2016) **99** 54-62. DOI: 10.1016/j.freeradbiomed.2016.07.030
20. Dumartin L, Whiteman HJ, Weeks ME, Hariharan D, Dmitrovic B, Iacobuzio-Donahue CA, Brentnall TA, Bronner MP, Feakins RM, Timms JF. **AGR2 is a novel surface antigen that promotes the dissemination of pancreatic cancer cells through regulation of cathepsins B and D**. *Cancer Res* (2011) **71** 7091-7102. DOI: 10.1158/0008-5472.CAN-11-1367
21. Fessart D, Domblides C, Avril T, Eriksson LA, Begueret H, Pineau R, Malrieux C, Dugot-Senant N, Lucchesi C, Chevet E. **Secretion of protein disulphide isomerase AGR2 confers tumorigenic properties**. *Elife* (2016). DOI: 10.7554/eLife.13887
22. Bu H, Bormann S, Schafer G, Horninger W, Massoner P, Neeb A, Lakshmanan VK, Maddalo D, Nestl A, Sultmann H. **The anterior gradient 2 (AGR2) gene is overexpressed in prostate cancer and may be useful as a urine sediment marker for prostate cancer detection**. *Prostate* (2011) **71** 575-587. DOI: 10.1002/pros.21273
23. Delom F, Nazaraliyev A, Fessart D. **The role of protein disulphide isomerase AGR2 in the tumour niche**. *Biol Cell* (2018) **110** 271-282. DOI: 10.1111/boc.201800024
24. Brychtova V, Mohtar A, Vojtesek B, Hupp TR. **Mechanisms of anterior gradient-2 regulation and function in cancer**. *Semin Cancer Biol* (2015) **33** 16-24. DOI: 10.1016/j.semcancer.2015.04.005
25. Arumugam T, Deng D, Bover L, Wang H, Logsdon CD, Ramachandran V. **New blocking antibodies against novel AGR2-C44A pathway reduce growth and metastasis of pancreatic tumors and increase survival in mice**. *Mol Cancer Ther* (2015) **14** 941-951. DOI: 10.1158/1535-7163.MCT-14-0470
26. Vanderlaag KE, Hudak S, Bald L, Fayadat-Dilman L, Sathe M, Grein J, Janatpour MJ. **Anterior gradient-2 plays a critical role in breast cancer cell growth and survival by modulating cyclin D1, estrogen receptor-alpha and survivin**. *Breast Cancer Res* (2010) **12** R32. DOI: 10.1186/bcr2586
27. Xue X, Fei X, Hou W, Zhang Y, Liu L, Hu R. **miR-342-3p suppresses cell proliferation and migration by targeting AGR2 in non-small cell lung cancer**. *Cancer Lett* (2018) **412** 170-178. DOI: 10.1016/j.canlet.2017.10.024
28. Dumartin L, Alrawashdeh W, Trabulo SM, Radon TP, Steiger K, Feakins RM, di Magliano MP, Heeschen C, Esposito I, Lemoine NR. **ER stress protein AGR2 precedes and is involved in the regulation of pancreatic cancer initiation**. *Oncogene* (2017) **36** 3094-3103. DOI: 10.1038/onc.2016.459
29. Higa A, Mulot A, Delom F, Bouchecareilh M, Nguyen DT, Boismenu D, Wise MJ, Chevet E. **Role of pro-oncogenic protein disulfide isomerase (PDI) family member anterior gradient 2 (AGR2) in the control of endoplasmic reticulum homeostasis**. *J Biol Chem* (2011) **286** 44855-44868. DOI: 10.1074/jbc.M111.275529
30. Hetz C. **The unfolded protein response: controlling cell fate decisions under ER stress and beyond**. *Nat Rev Mol Cell Biol* (2012) **13** 89-102. DOI: 10.1038/nrm3270
31. Chen CY, Chung IH, Tsai MM, Tseng YH, Chi HC, Tsai CY, Lin YH, Wang YC, Chen CP, Wu TI. **Thyroid hormone enhanced human hepatoma cell motility involves brain-specific serine protease 4 activation via ERK signaling**. *Mol Cancer* (2014) **13** 162. DOI: 10.1186/1476-4598-13-162
32. Delom F, Mohtar MA, Hupp T, Fessart D. **The anterior gradient-2 interactome**. *Am J Physiol Cell Physiol* (2020) **318** C40-C47. DOI: 10.1152/ajpcell.00532.2018
33. Lee AH, Iwakoshi NN, Glimcher LH. **XBP-1 regulates a subset of endoplasmic reticulum resident chaperone genes in the unfolded protein response**. *Mol Cell Biol* (2003) **23** 7448-7459. DOI: 10.1128/MCB.23.21.7448-7459.2003
34. Yoshida H, Matsui T, Yamamoto A, Okada T, Mori K. **XBP1 mRNA is induced by ATF6 and spliced by IRE1 in response to ER stress to produce a highly active transcription factor**. *Cell* (2001) **107** 881-891. DOI: 10.1016/S0092-8674(01)00611-0
35. Hetz C, Chevet E, Oakes SA. **Proteostasis control by the unfolded protein response**. *Nat Cell Biol* (2015) **17** 829-838. DOI: 10.1038/ncb3184
36. Blazanin N, Son J, Craig-Lucas AB, John CL, Breech KJ, Podolsky MA, Glick AB. **ER stress and distinct outputs of the IRE1alpha RNase control proliferation and senescence in response to oncogenic Ras**. *Proc Natl Acad Sci U S A* (2017) **114** 9900-9905. DOI: 10.1073/pnas.1701757114
37. Yang L, Xue Z, He Y, Sun S, Chen H, Qi L. **A Phos-tag-based approach reveals the extent of physiological endoplasmic reticulum stress**. *PLoS ONE* (2010) **5** e11621. DOI: 10.1371/journal.pone.0011621
38. Ozcan U, Yilmaz E, Ozcan L, Furuhashi M, Vaillancourt E, Smith RO, Gorgun CZ, Hotamisligil GS. **Chemical chaperones reduce ER stress and restore glucose homeostasis in a mouse model of type 2 diabetes**. *Science* (2006) **313** 1137-1140. DOI: 10.1126/science.1128294
39. Mimura N, Fulciniti M, Gorgun G, Tai YT, Cirstea D, Santo L, Hu Y, Fabre C, Minami J, Ohguchi H. **Blockade of XBP1 splicing by inhibition of IRE1alpha is a promising therapeutic option in multiple myeloma**. *Blood* (2012) **119** 5772-5781. DOI: 10.1182/blood-2011-07-366633
40. Reimer N, Unberath P, Busch H, Borries M, Metzger P, Ustjanzew A, Renner C, Prokosch HU, Christoph J. **Challenges and experiences extending the cBioPortal for cancer genomics to a molecular tumor board platform**. *Stud Health Technol Inform* (2021) **287** 139-143. PMID: 34795098
41. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, Jacobsen A, Byrne CJ, Heuer ML, Larsson E. **The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data**. *Cancer Discov* (2012) **2** 401-404. DOI: 10.1158/2159-8290.CD-12-0095
42. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, Sun Y, Jacobsen A, Sinha R, Larsson E. **Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal**. *Sci Signal* (2013). DOI: 10.1126/scisignal.2004088
43. Guo H, Zhu Q, Yu X, Merugu SB, Mangukiya HB, Smith N, Li Z, Zhang B, Negi H, Rong R. **Tumor-secreted anterior gradient-2 binds to VEGF and FGF2 and enhances their activities by promoting their homodimerization**. *Oncogene* (2017) **36** 5098-5109. DOI: 10.1038/onc.2017.132
|
---
title: Multi-omics and experimental analysis unveil theragnostic value and immunological
roles of inner membrane mitochondrial protein (IMMT) in breast cancer
authors:
- Hung-Yu Lin
- Hsing-Ju Wu
- Pei-Yi Chu
journal: Journal of Translational Medicine
year: 2023
pmcid: PMC9999521
doi: 10.1186/s12967-023-04035-4
license: CC BY 4.0
---
# Multi-omics and experimental analysis unveil theragnostic value and immunological roles of inner membrane mitochondrial protein (IMMT) in breast cancer
## Abstract
### Background
The inner membrane mitochondrial protein (IMMT) is a central unit of the mitochondrial contact site and cristae organizing system (MICOS). While researchers continue to demonstrate the physiological function of IMMT in regulating mitochondrial dynamics and preserving mitochondrial structural integrity, the roles of IMMT in clinicopathology, the tumor immune microenvironment (TIME), and precision oncology in breast cancer (BC) remain unclear.
### Methods
Multi-omics analysis was used here to evaluate the diagnostic and prognostic value of IMMT. Web applications aimed at analyzing the whole tumor tissue, single cells, and spatial transcriptomics were used to examine the relationship of IMMT with TIME. Gene set enrichment analysis (GSEA) was employed to determine the primary biological impact of IMMT. Experimental verification using siRNA knockdown and clinical specimens of BC patients confirmed the mechanisms behind IMMT on BC cells and the clinical significance, respectively. Potent drugs were identified by accessing the data repositories of CRISPR-based drug screenings.
### Results
High IMMT expression served as an independent diagnostic biomarker, correlated with advanced clinical status, and indicated a poor relapse-free survival (RFS) rate for patients with BC. Although, the contents of Th1, Th2, MSC, macrophages, basophil, CD4 + T cell and B cell, and TMB levels counteracted the prognostic significance. Single-cell level and whole-tissue level analyses revealed that high IMMT was associated with an immunosuppressive TIME. GSEA identified IMMT perturbation as involved in cell cycle progression and mitochondrial antioxidant defenses. Experimental knockdown of IMMT impeded the migration and viability of BC cells, arrested the cell cycle, disturbed mitochondrial function, and increased the ROS level and lipid peroxidation. The clinical values of IMMT were amenable to ethnic Chinese BC patients, and can be extrapolated to some other cancer types. Furthermore, we discovered that pyridostatin acted as a potent drug candidate in BC cells harboring an elevated IMMT expression.
### Conclusion
This study combined a multi-omics survey with experimental verification to reveal the novel clinical significance of IMMT in BC, demonstrating its role in TIME, cancer cell growth and mitochondrial fitness, and identified pyridostatin as a promising drug candidate for the development of precision medicine.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12967-023-04035-4.
## Background
According to a status report on the global cancer burden provided by GLOBOCAN 2020, breast cancer (BC) is the most commonly diagnosed cancer among women, while it is the fourth leading cause of cancer deaths globally [1]. Based on the presence or absence of biomarkers for estrogen receptors, progesterone receptors and human epidermal growth factor 2 (HER2), BC can be categorized as luminal A/B, HER2-positive, or triple-negative breast cancer (TNBC) [2]. To date, surgery, radiation, and endocrine therapy remain the primary therapeutic strategies to manage BC [3]. Neoadjuvant therapies, including chemotherapy combined with targeted agents have been widely applied in BC patients with high recurrence and metastasis [4]. Unfortunately, a majority of patients develop resistance to these treatments and inevitably relapse [5]. Thus, identifying reliable and accurate biomarkers for the diagnosis, prognosis, and therapeutic targeting of BC is imperative.
The primary cause of death in BC patients is cancer metastasis, associated with metabolic reprogramming which cultivates a corrupted tumor microenvironment, thereby counteracting therapy-induced cell death [6]. Mitochondria are dynamic organelles which provide metabolic support and regulate cellular functions, including calcium homeostasis, redox status, and programmed cell death. Mounting evidence indicates that tumor cells modify mitochondrial dynamics to enhance proliferation and survival, thus the targeting of mitochondrial dynamics may be an effective strategy to suppress the metastatic ability of BC cells [7].
The predominant role of the mitochondrial contact site and cristae organizing system (MICOS) complex in structuring the inner mitochondrial membrane (IMM) and cristae junctions formation and in affecting mitochondrial dynamics and metabolism has recently become the focus of a growing number of investigations [8]. The MICOS complex is comprised of MIC10/MICOS10 (mitochondrial contact site and cristae organizing system subunit 10), MIC13/QIL1 (mitochondrial contact site and cristae organizing system subunit 13), MIC19/CHCHD3 (coiled-coil-helix-coiled-coil-helix domain containing 3), MIC25/CHCHD6 (coiled-coil-helix-coiled-coil-helix domain containing 6), MIC26/APOO (apolipoprotein O), MIC27/APOOL (apolipoprotein O like), and MIC60/IMMT (inner membrane mitochondrial protein, also known as mitofilin) [8]. IMMT is generally recognized as the core of MICOS [9, 10], while the downregulation, modification or destruction of IMMT may lead to mitochondrial dysfunction, and ultimately cell death. In addition, IMMT is a hallmark of various diseases, including cancer [10]. However, the role of IMMT in clinicopathology, the tumor immune microenvironment (TIME), and precision oncology remain unclear.
The integrated analysis we applied in this study aimed to investigate the theragnostic value of IMMT in BC. We first examined the differential expression, diagnostic efficacy, and prognostic value of IMMT. Second, the association between IMMT expression and TIME at the single-cell and whole-tissue levels were determined. Third, the possible mechanism underlying the tumor-promoting role of IMMT was identified by functional enrichment modules and confirmed by experimental verification. Fourth, we confirmed the clinicopathological significance of IMMT using real-world ethnic Chinese BC patients’ tissues and by pan-cancer bioinformatics. Finally, by accessing the Genomics of Drug Sensitivity in Cancer (GDSC) and Cancer Cell Line Encyclopedia (CCLE) cell data repositories, we identified pyridostatin as an effective drug candidate for cancer cells with high IMMT expression.
## Gene differential expression and prognostic significance
Data from TNMplot [11], UALCAN (The University of ALabama at Birmingham CANcer data analysis Portal) [12], Breast Cancer Gene-Expression Miner (bc-GenExMiner) v4.8 [13], and Gene Expression Omnibus (GEO) were used to analyze the gene expression levels in BC tumors and adjacent normal tissues. The Human Protein Atlas [14–16] was accessed to determine immunohistochemical (IHC) staining images. The Kaplan–Meier plotter [17] was applied to analyze the survival rates associated with the various clinical stages of BC, wherein the patient groups were divided by “Auto select best cut off”, thereby selecting the best-performing cutoff value.
## Single-cell and immune analyses
The Tumor Immune Single Cell Hub (TISCH) was utilized to conduct single-cell analyses [18] to determine which BC cell types may express IMMT. Next, independent datasets from the scTIME Portal [19], consisting of BC patients’ cells, were analyzed on GSE75688 and visualized in UMAP. A heatmap was used to illustrate the signature expression of Mitophagy in GSE75688. The cellular communication among various T cell subsets was analyzed by the LR network of the scTIME Portal. SpatialDB was used to analyze the spatial transcriptomics [20], whereby the gene expression in tissue sections can be visualized and quantified. The IMMT expression in the immune cells of BC tissue was determined based on the GSE114724 dataset. TISIDB was used to conduct the Spearman correlation test for IMMT expression with immune infiltration levels [21]. GEPIA2 was used to determine the correlation of IMMT with the immune cell signature [22]. Estimation of Stromal and Immune cells in Malignant Tumor tissues using *Expression data* (ESTIMATE) was used to evaluate the matrix content, immune cell infiltration levels, comprehensive score, and tumor purity [23].
## Genetic and enrichment analyses
The GSCALite platform was employed to analyze the single nucleotide variants (SNVs) [24]. The TIMER platform was employed to analyze the Pearson correlation coefficient of mutation with gene expression in TCGA samples [25–27]. The UCSC Xena portal and DriverDBv3 were used to evaluate the correlation of copy number with gene expression in TCGA samples [28, 29]. The LinkedOmics platform was used to identify IMMT co-expressed genes and conduct a gene set enrichment analysis (GSEA) to illustrate the Reactome pathway [30].
## Cell culture and transfection
The cell culture and transfection were conducted as previously described [31]. Briefly, MDA-MB-231 cells were grown in DMEM (HyClone) supplemented with $10\%$ fetal bovine serum (Thermo Fisher Scientific). The existence of mycoplasma was determined using the e-Myco™ plus Mycoplasma PCR Detection Kit (iNtODEWORLD, MA, USA). For the knockdown of IMMT, 20 nM of IMMT-targeting siRNA (4392420, Thermo Fisher Scientific) and its corresponding negative control RNA (4390843, Thermo Fisher Scientific) were introduced into cells using the Lipofectamine™ RNAiMAX Transfection Reagent (LMRNA015, Invitrogen, Carlsbad, CA, USA). The siRNA transfection efficiency was determined by western blot 72 h after transfection.
## Western blotting
Western blot was conducted as previously described [31]. Briefly, the RIPA lysis buffer was used to harvest cells. 30 μg of proteins were loaded and separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). After the proteins were transferred to the PVDF membranes, primary and secondary antibodies were used to probe the target protein, which would then be visualized using enhanced chemiluminescence (Millipore, Bedford, MA). The primary antibodies included anti-β-actin (A5441; Sigma-Aldrich) and anti-IMMT (PA3-870, Thermo Fisher Scientific, Waltham, MA, USA). The signal of β-actin served as a loading control. The blotting signal intensity was quantified by GelQuant. NET v1.8.2 (Accessed from http://biochemlabsolutions.com/).
## Wound healing assay
25,000 cells were seeded on a 24-well plate 1 day prior to the siRNA treatment. 72 h after transfection, a wound was scratched in each well using a 10 μl pipette tip. Cell debris was washed twice with 1×PBS and then cultivated in a fresh medium. Wound closure was monitored at 0 and 22 h on a Lionheart FX microscope (Agilent Technologies, CA, USA). The quantification of the wound area was determined by ImageJ.
## Cell viability analysis
The colorimetry-based Cell Counting Kit-8 (CCK-8) (96,992; Sigma-Aldrich) was used to evaluate cell viability. 10,000 per 100 μL of MDA-MB-231 cells were seeded into a 96-well plate. After 18 h, the attached cells were then transfected with si-ctrl or si-IMMT for 72 h. The CCK-8 solution was then added to the medium for 1 h at 37 °C. The colorimetric signal was acquired at a wavelength of 450 nm on a microplate (Hidex Sense microplate reader, Turku, Finland). The reference wavelength was set at 630 nm.
## Flow cytometry
Cell cycle analysis was conducted as described in a previous report [31]. Briefly, ice-cold $95\%$ ethanol was used to fix the cell. 10 μg/mL of propidium iodide was used to probe the DNA content. The signal representing the DNA content was then acquired by an LSR II flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA). To detect mitochondrial membrane potential (MMP), cells were incubated with 100 nM TMRM (Tetramethylrhodamine Methyl Ester Perchlorate) (T668; Thermo Fisher Scientific). After PBS washing, the fluorescence signal was examined on an LSR II flow cytometer. To measure reactive oxygen species (ROS), MitoSOX™ Red (M36008; Thermo Fisher Scientific) and CM-H2DCFDA (C6827, Thermo Fisher Scientific) were employed to detect mitochondrial ROS and intracellular general ROS, respectively. 5 μM of MitoSOX™ Red and CM-H2DCFDA were respectively applied to cells, which were incubated for 30 min at 37 °C, after which fluorescence analysis was performed using an LSR II flow cytometer.
## Analysis of oxygen consumption rate (OCR)
Cellular OCR was evaluated using a Seahorse XF24 analyzer (Seahorse Bioscience, Billerica, MA) as previously described [32], adhering to the manufacturer’s instructions with minor modifications. Briefly, 50,000 cells were plated in a Seahorse Flux Analyzer plate. After 18 h, the plate was pre-heated at 37 °C for 1 h. We documented three measurements each of basal OCR, proton-leak OCR, and maximal OCR. The proton-leak OCR was assessed using 1 μM oligomycin. Maximal OCR was driven by treating the cells with 300 nM FCCP. Finally, non-mitochondrial respiration was obtained by injection of 1 μM rotenone.
## Ethics statement and specimen collection for immunohistochemistry
The clinical studies were performed in compliance with the approved guidelines of the Institutional Review Board (IRB) of Show Chwan Memorial Hospital (Approval Number: 1091208; approval date: December 8, 2020). The immunohistochemistry (IHC) assays and scoring methods were carried out based on previously established standard procedures [33].
## Investigation of potent drug candidates according to IMMT expression
The CRISPR-screen of the Genomics of Drug Sensitivity in Cancer (GDSC) data repository was used to analyze IMMT expression-based drug sensitivity [34]. Data regarding sensitivity to pyridostatin was retrieved from the CCLE [35].
## Statistical analyses
Statistical comparisons were conducted as previously described [36]. Briefly, unpaired t-test and one-way analysis of variance (ANOVA) were employed to analyze quantitative data of two-group and three-or-more-group comparisons, respectively. Pearson correlation coefficient was utilized to determine the relationship between two variables. Kaplan–*Meier analysis* and log‐rank test were carried out to compare the survival rate. Receiver operating characteristics (ROC) was used to evaluate the diagnostic efficiency of IMMT. The interconnectivity between the high expression of MICOS genes and the survival rate was illustrated using the chord diagram in the circlize package of RStudio Cloud (https://rstudio.cloud/content/yours?sort=name_asc).
## Clinicopathological screening of IMMT-centric MICOS components in BC
We first examined the genetic variant profiles of MICOS subunits in BC samples of the TCGA repository (Fig. 1A). 19 out of 1,026 BC samples ($1.85\%$) showed an altered single nucleotide variant (SNV). IMMT showed the highest SNV counts (6 samples) compared to CHCHD6 (4 samples), CHCHD3 (3 samples), APOO (3 samples), APOOL (4 samples), and MICOS10 (0 samples) (Fig. 1B). The majority of the SNVs were missense mutations (Fig. 1B). In the TCGA datasets, BC tumor tissue showed elevated gene expression levels of IMMT, CHCHD6, CHCHD3, APOO, and MICOS10 as compared to normal tissue, while the APOOL level was decreased in tumor tissue as compared to normal tissue (Fig. 1C–H). We subsequently examined the protein level expression by accessing CPTAC datasets. The protein expressions of IMMT, CHCHD6, CHCHD3, and APOOL followed a similar trend to the gene expressions, except for APOO (Fig. 1I–M). The comparison of the different mutation status between the differential gene expressions revealed that the mutations of IMMT, CHCHD6, CHCHD3, APOO, and MICOS10 demonstrated no significant associations with gene expressions (Fig. 1N). The log-rank test was used to analyze the association between gene expressions and survival rates. As shown in Fig. 1O (all $p \leq 0.05$), high IMMT expression was associated with an unfavorable overall survival (OS) and disease-free survival (DFS); high CHCHD6 expression was associated with an unfavorable OS and distant metastasis free survival (DMFS); high APOOL and APOO expressions were associated with an unfavorable OS; while high MICOS10 expression was associated with favorable DMFS and DFS.Fig. 1Clinicopathological Screening of IMMT-centric MICOS Components in BC. A Schematic diagram depicting the MICOS subunits in mitochondria. B Single nucleotide variant (SNV) oncoplot of BC samples. The number and percentage of distinct SNVs, including missense mutation (green), nonsense mutation (red), splice site (orange), frame shift/deletion (light blue) and presence of multiple types of mutation in the same gene (Multi Hit, black) on each gene in 1,026 BC samples are indicated. Level of tumor mutation burden (TMB) in each sample is documented on the top of the oncoplot. C–H the gene expression levels of IMMT (C), CHCHD6 (D), CHCHD3 (E), APOOL (F), APOO (G), and MICOS10 (H) in normal tissues and BC tumors. I–M the protein expression levels of IMMT (I), CHCHD6 (D), CHCHD3 (E), APOOL (F), APOO (G), and MICOS10 (H) in normal tissues and BC tumors accessed from the CPTAC repository. * $p \leq 0.05$, ***$p \leq 0.001$ between the two groups. N heatmap comparing the differential gene expressions of IMMT, CHCHD6, CHCHD3, MICOS10, APOOL, and APOO between different mutation statuses. LOG FC, fold change expressed as log logarithm to the base 10. O Chord diagram illustrating the relationship between the hazard ratio in BC patients harboring high gene expressions of IMMT, CHCHD6, APOOL, APOO, and MICOS10 and survival rate. OS overall survival; DFS disease-free survival, DMFS distant metastasis free survival
## Differential expression and diagnostic value of IMMT in BC
Both the RNA-seq data and chip array data revealed that IMMT expression levels were increased in BC tumor tissue as compared to normal tissue (Fig. 2A, B). The specificity of RNA-seq data increased with a higher cut-off value in normal tissue, while the chip array data sustained a high specificity regardless of the cut-off point (Fig. 2C, D). The immunohistochemistry staining data of the Human Protein Atlas confirmed an overexpressed IMMT protein in BC tumor tissue compared to normal tissue (Fig. 2E–F). We constructed an ROC curve to evaluate the diagnostic efficiency of IMMT for BC based on GSE11925 datasets. We noted that the IMMT expression levels demonstrated a high diagnostic accuracy (AUC: 0.701, $95\%$ CI 0.65–0.74, $p \leq 0.0001$) (Fig. 2G). Spatial transcriptomic analysis of IMMT showed that a high level of IMMT was concurrently expressed with the cancerous section of BC tumor tissue (Fig. 2H). UCSC *Xena analysis* and TCGA-BRCA repository data demonstrated that the IMMT expression level closely correlated with the copy number (Fig. 2I, J). We then investigated the IMMT expression based on various clinical categories using different databases, including TCGA and The Sweden Cancerome Analysis Network-Breast (SCAN-B) [37]. BC tumors of all clinical stages showed higher IMMT levels than normal tissue (Fig. 3A). The TCGA repository data indicated that all BC subtypes, including luminal, HER2-enriched (HER2-E), and triple-negative breast cancer (TNBC) showed increased IMMT levels, while HER2-E had higher IMMT levels than luminal (Fig. 3B). Similarly, the SCAN-B repository data indicated that all BC subtypes had higher IMMT than normal, while HER2-E had higher IMMT than basal-like and luminal A (Fig. 3C). Grade 3 tumors showed increased IMMT compared to those of grades 1 or 2 (Fig. 3D). A TNBC tumor had increased IMMT as compared to a non-TNBC tumor (Fig. 3E). Nodal status positive (N+) showed higher IMMT levels compared to nodal status negative (N–) (Fig. 3F). IMMT levels were increased with an elevated Nottingham prognostic index (NPI) score (Fig. 3G). Additionally, we examined the IMMT expression based on the Ki67 staining levels with IHC, which is an independent prognostic biomarker for BC [38]. As shown in Fig. 3H, the Ki67-high tumor had a higher IMMT level than the Ki67-low tumor. Collectively, these results demonstrate that IMMT expression may serve as a diagnostic marker and that IMMT is linked to an advanced disease status. Fig. 2Differential expression and diagnostic value of IMMT. A and B violin plots showing the IMMT expression levels determined by RNA-seq (A) or gene chip (B) in normal and BC tumor tissues. *** $p \leq 0.0001$ between indicated groups. C and D the percent and specificity at various cut-off points in normal tissue from RNA-seq data (C) and gene chip data (D). E and F IHC staining for the detection of IMMT protein expression levels by antibody CAB022439 in normal and BC tissues (E). The number of BC patients ($$n = 11$$) determined with moderate, weak, or negative staining intensity. G ROC curves for normal and BC cohorts based on IMMT levels from the GSE119295 dataset. H Spatial IMMT expression in human BC tissue sections as determined by spatial transcriptomics. Black square enlarging the dark-staining tissue that indicates a tumor nest composed of hyperchromatic cancer cells. I and J the correlation of IMMT gene expression with its copy number. Heatmap analyzed o visualizing the expression-copy number relationship of GDC-TCGA Breast Cancer dataset (I). Pearson correlation coefficient (PCC) of tumors with IMMT copy number status of gain, loss, and none, and normal tissue based on TCGA breast cancer datasetsFig. 3IMMT expression according to clinical factors and pathological categories from multiple databases. A and B The Cancer Genomic Atlas (TCGA) repository-based IMMT expression levels. Box plot of IMMT expression according to clinical stages (A) and major subtypes (B). TPM, Transcripts Per Million. C–H SCAN-B (The Sweden Cancerome Analysis Network-Breast) repository-based IMMT expression levels. Violin plot of IMMT expression according to PAM subtypes (C), SBR (Scarff Bloom and Richardson) grade (D), IHC status of TNBC (triple-negative breast cancer) (E), nodal status (F), NPI (Nottingham prognostic index) (G) and IHC status of Ki67 (H). ** $p \leq 0.01$, ***$p \leq 0.001$ when compared with normal group. N−, NPI1, or Ki67 low groups. # $p \leq 0.05$, ###$p \leq 0.001$ when comparing indicated groups
## Prognostic value of IMMT
To gain more insight into prognosis, we examined the correlation between IMMT expression and survival rate. In this regard, recurrence-free survival (RFS) was considered as an appropriate indicator of the occurrence of metastasis. High IMMT expression in BC patients was found to be associated with decreased RFS time (Fig. 4A). More specifically, IMMT acted as a prognostic indicator in HER2-E, luminal A and luminal B, although not in basal type (Fig. 4B–E). A high IMMT expression indicated decreased RFS time in BC patients with grade 3, while no differences were noted in grades 1 and 2 (Fig. 4F–H). In terms of therapeutic scenarios, a high IMMT expression indicated decreased RFS time in BC patients undergoing all types of chemotherapy, adjuvant chemotherapy or tamoxifen, but not for neoadjuvant chemotherapy (Fig. 4I–L). In addition, IMMT can serve as an independent factor in N–, but not in N+ (Fig. 4M, N).Fig. 4Prognostic value of IMMT in BC. A–N Kaplan–Meier survival analysis of relapse-free survival (RFS) based on IMMT expression. “ Auto select best cut off” algorithm function was used to divide low/high groups in all BC patients (A), PAM50 subtype with basal (B), HER2-E (HER2-enriched) (C), luminal A (D) and luminal B (E), grade 1 (F), grade 2 (G), grade 3 (H), patients undergoing any chemotherapy (I), patients undergoing adjuvant chemotherapy (J), patients undergoing neoadjuvant chemotherapy (K), patients undergoing tamoxifen only (L), nodal status negative (M), and nodal status positive (N). O and P forest plot illustrating the HR of high IMMT, the $95\%$ CI when harboring various immune cell contents (O) and TMB extent (P) As the tumor-infiltrating immune cells (TIICs) are indispensable parts of the TIME and implicated in clinical outcome [39], we then looked into the prognostic value of IMMT in BC when considering the immune cell constituents. High IMMT did not statistically indicate decreased RFS time in BC tumors harboring decreased helper T cell type 1 (Th1), decreased Th2, enriched mesenchymal stem cell (MSC), decreased macrophage, decreased basophil, decreased CD4 + T cell, and decreased B cell (Fig. 4O). It is worth noting that high IMMT indicated decreased RFS time in BC tumors harboring a high tumor mutation burden (TMB), but not in those with a low TMB (Fig. 4P). Collectively, high IMMT present a prognostic prediction of an unfavorable outcome in BC patients, particularly in subtype HER-2E, luminal A and luminal B, tumor grade 3, treatment with adjuvant chemotherapy and tamoxifen, and in N−. However, the contents of Th1, Th2, MSC, macrophage, CD4 + T cell, B cell, and TMB can counteract the predictive efficacy.
## Single-cell level analysis
We subsequently attempted to localize IMMT at the single-cell level. A TISCH-based single-cell analysis of BC based on multiple GEO datasets confirmed the expression of IMMT in immune cells, malignant cells, and stromal cells (Fig. 5A). The single-cell RNA-seq based on GSE75688 then showed that IMMT was predominantly expressed in tumor cells, although was relatively low in T cells (Fig. 5B, C). Interestingly, the BC tumor was more abundant in gene signatures of mitophagy (Fig. 5D), in which IMMT acts as a key regulator [10]. Cellular communication retrieved from the scTIME Portal revealed complex interactions among various T cell subpopulations (Fig. 5E). The single-cell RNA-seq of GSE114724 revealed a positive correlation between a particular group of CD8 T cells (CD8-GZMK, ZNF683, PDCD1 and cycling-T) and IMMT expression (Fig. 5F–H).Fig. 5Analysis of IMMT at the single-cell level. A–H single-cell analysis. Heatmap showing the IMMT expression profile in immune cells, malignant cells, stromal cells and others of BC based on multiple GEO datasets (A). Uniform Manifold Approximation and Projection (UMAP) plots showing IMMT expression mapping on to different cell types in BC based on the GSE75688 dataset (B and C). Heatmap visualizing the gene signature MITOPHAGY of T cells and BC tumor (D). Communication network among immune cells (E). UMAP plots showing IMMT expression clusters (F–H)
## IMMT over-expression is associated with immunosuppressive TIME
Considering the vital role of TIME in tumor growth, spread, and escape from immune-mediated destruction [40], we analyzed the associations between IMMT and tumor-immune interaction. A Spearman correlation analysis of TISIDB revealed that IMMT expression negatively correlated with most tumor-infiltrating lymphocytes in BC (Fig. 6A), such as effector memory CD8 T cells (Tem_CD8) (Fig. 6B), CD56bright killer cells (Fig. 6C), natural killer (NK) cells (Fig. 6D), and natural killer T (NKT) cells (Fig. 6E). Moreover, we observed that IMMT has a positive association with T cell-suppressors, such as PD-L1 (Programmed Cell Death 1 Ligand [1] (Fig. 6F), PD-L2 (programmed cell death 1 ligand 2) (Fig. 6G), IDO1(indoleamine 2,3-dioxygenase [1] (Fig. 6H), and CTLA4 (cytotoxic t-lymphocyte associated protein [4] (Fig. 6I). A Pearson correlation analysis conducted using GEPIA2 demonstrated that IMMT had no significant correlation with the effector T cell signature (Fig. 6J), while showing a significant positive correlation with the exhausted T cell signature (Fig. 6K), resting Treg cell signature (Fig. 6L) and effector Treg cell signature (Fig. 6M). Furthermore, the ESTIMATE analysis [23] demonstrated that the cohort with high IMMT exhibited a reduced stromal score (Fig. 6N), immune score (Fig. 6O) and ESTIMATE score (Fig. 6P), while showing enhanced tumor purity (Fig. 6Q). Collectively, these results indicate that IMMT over-expression is associated with an immunosuppressive TIME.Fig. 6The role of IMMT in shaping TIME. A–M immunological analysis on immune infiltration and immunosuppressors. Heatmap showing the correlation between IMMT expression and lymphocytes infiltration across human cancers (A). Dot plots showing the correlation of IMMT with anti-tumor cell population such as effector memory CD8 T cell (Tem_CD8) (B), CD56bright killer cells (C), NK cells (D), NKT cells (E), with immunosuppressive molecules such as PD-L1 (F), PD-L2 (G), IDO1 (H), and CTLA4 (I), and with effector T cell signature (CX3CR1, FGFBP2 and FCGR3A) (J), with exhausted T cell signature (HAVCR2, TIGIT, LAG3, PDCD1, CXCL13 and LAYN) (K), with resting Treg cell signature (FOXP3 and IL2RA) (L), and with effector Treg cell signature (FOXP3, CTLA4, CCR8 and TNFRSF9) (M). Violin plots showing the stromal score (N), immune score (O), EXTIMATE score (P) and tumor purity (Q). * $p \leq 0.01$, **$p \leq 0.01$, and ***$p \leq 0.001$ between the low and high IMMT groups
## Involvement of IMMT in cell cycle progression and mitochondrial antioxidant defenses
We further utilized Linkedomics [30] to analyze the functional enrichment of IMMT. A total of 8,495 genes had significant positive associations with IMMT, while 11,660 genes had significant negative associations (Fig. 7A). A heatmap visualized the top 30 individual genes (Fig. 7B, C). *The* gene set enrichment analysis (GSEA) revealed that genes co-expressed with IMMT were involved in the activation of the following Reactome pathway terms: cell cycle, DNA repair, RHO GTPase effectors, transcriptional regulation by TP53 infectious disease, and mitochondrial protein import (Fig. 7D). Meanwhile, the inhibited terms included: GPCR ligand binding, O-glycosylation of proteins, muscle contraction, and extracellular matrix organization (Fig. 7D). Notably, the enriched Reactome pathway terms showed that cell cycle has the highest Normalized Enriched Score (NES), the most gene counts and the lowest false discovery rate (FDR) (Fig. 7D). Considering the predominant impact of IMMT on cell cycle activity, we examined its association with critical biomarkers of cancer progression. As shown in Fig. 7E–J, IMMT levels positively correlated with biomarkers linked to cancer proliferation (KI67, PCNA, and MCM2) and with key regulators of the cell cycle (CDK4, CDK2 and CDK1). Due to the pivotal role of IMMT in regulating the redox status [41], we investigated its association with mitochondrial antioxidant defenses in BC tumor samples. IMMT had a positive association with PGC-1β (proliferator-activated receptor-γ coactivator 1 beta) (Fig. 7K), PRDX1 (peroxiredoxin 1) (Fig. 7L), PRDX3 (peroxiredoxin 3) (Fig. 7M), HSPA9 (heat shock protein family A member 9) (Fig. 7N), HSPD1 (heat shock protein family D member 1) (Fig. 7O), and SOD2 (superoxide dismutase 2) (Fig. 7P). These results suggest that IMMT may play a role in BC cell cycle progression by regulating mitochondrial redox status. Fig. 7Involvement of IMMT in cell cycle progression and mitochondrial antioxidant defenses. Genetic and enrichment analyses of TCGA samples were conducted using the LinkedOmics functional modules. Volcano plot showing the PCC and p value of the IMMT-coexpressed genes (A). Heatmap showing top 30 positively (B) and negatively (C) correlated genes with IMMT. Bubble plot showing the normalized enrichment score, involved gene count and false discovery rate (FDR) of the Reactome pathway terms (D). Dot plots showing the correlation between IMMT and Ki67 (E), PCNA (F), and MCM2 (G), CDK4 (H), CDK2 (I) and CDK1 (J), PGC-1β (K), PRDX1(L), PRDX3(M), HSPA9(N), HSPD1(O), and SOD2(P) To verify the biological role of IMMT in BC, we conducted an siRNA-based knockdown in MDA-MB-231 cells (Fig. 8A–B). The wound healing assay revealed that cells undergoing IMMT-knockdown had reduced cell migration abilities and viability compared to cells subjected to negative control siRNA (Fig. 8C–E). In addition, knockdown of IMMT led to an increased proportion of sub-G1 phase, a decreased G1 phase, and an increased S phase (Fig. 8F, G). The respiration activity and mitochondrial membrane potential (MMP) were reduced by the knockdown of IMMT (Fig. 8H–J). Notably, the indicator of lipid peroxidation 4-Hydroxynonenal (4-HNE) was increased (Fig. 8K, L). Enhancements in mitochondrial ROS and general ROS were noted (Fig. 8M, N). Similarly, IMMT knockdown suppressed cell migration ability (Additional file 1: Fig. S1A, B) and induced lipid peroxidation (Additional file 1: Fig. S1C, D) in MCF-7 cells. Collectively, these results indicate that IMMT exerts positive roles in the regulation of mitochondrial fitness and intracellular oxidative stress, which may account for the motility and proliferative capacity of BC cells. Fig. 8Knockdown of IMMT arrests the cell cycle and induces oxidative stress in BC cells. A–G Experimental verification of the biological role of IMMT on cancer cell biology. Representative western blot of IMMT in MDA-MB-231 cells treated with 20 nM negative control siRNA (si-ctrl) and IMMT siRNA (si-IMMT) for 72 h. β-actin serving as loading control (A). Bar chart showing the quantitative result of western blot (B). Representative images of wound healing assay at 0 h and 22 h (C). Quantification results of the wound area determined by the migrated cells (D). Cell viability evaluated by CCK-8 assay (E). Representative histograms of each cell cycle phase (F). Bar chart representing percentages of cell cycle phases (G). H–N Assessments of mitochondrial function and oxidative stress. Basal OCR (oxygen consumption rate), proton leak-OCR, and maximal OCR were measured in basal assay medium, 1 μM oligomycin, and 300 nM FCCP, respectively (H). Quantitative results of OCR (I). Mitochondrial membrane potential (MMP) determined by TMRM staining on a flow cytometer (J). Representative western blot of lipid peroxidation assessment by probing 4-HNE abundance. β-actin as loading control (K). Quantitative bar chart of 4-HNE abundance (L). The percentage of mitochondrial ROS (mtROS) and intracellular general ROS detected by MitoSOXTMRed (M), and CM-H2DCFDA (N), respectively. *** $p \leq 0.001.$ ** $p \leq 0.01$ and **$p \leq 0.05$
## Validation of the clinicopathological significance of IMMT in ethnic Chinese patients and from a pan-cancer perspective
To verify the clinical value of IMMT, we collected tumor tissues and the clinical documents of BC patients ($$n = 461$$) from our hospital. IHC staining showed that the tumor tissues expressed higher IMMT levels than the tumor-adjacent normal tissue (Fig. 9A, B), and that grade 3 tumor tissue expressed higher IMMT levels than grade 2 (Fig. 9C, D). The Kaplan–*Meier analysis* revealed that patients with a high IMMT IHC score had decreased OS time (Fig. 9E). Similar to the observed bioinformatics results, grade 3 patients with high IMMT had a shorter overall survival time compared to low IMMT, even if there is no statistical significance (Additional file 1: Fig. S2). In this regard, our clinical findings correspond with the multi-omics data. Fig. 9Validation of the clinical significance of IMMT in ethnic Chinese BC specimens and from a bioinformatic pan-cancer perspective. A–E High IMMT expression as a pathological biomarker and a prognostic factor for poor outcome in ethnic Chinese patients with BC. Representative image and quantitative bar charts of tumor-adjacent normal tissue vs. tumor (A and B) and grade 1 vs. grade 2 vs. grade 3 tissues (C and D). Note that the non-neoplastic mammary gland duct cells and malignant mammary gland duct cells show mild and strong cytoplasmic expression of IMMT, respectively (100x) (A). The grade 1, grade 2, and grade 3 invasive ductal carcinoma cells show mild, moderate, and strong cytoplasmic expressions of IMMT, respectively (C). Kaplan–Meier curve of the overall survival (cutoff value high/low: $67\%$/$33\%$) (E). F–J Generalization value of IMMT across cancers. *** $p \leq 0.0001$, **$p \leq 0.01$, *$p \leq 0.05$ between tumor and normal (F). Kaplan–Meier curve representing the association between IMMT and RFS in cervical squamous cell carcinoma (CESC) (G), liver hepatocellular carcinoma (LIHC) (H), lung squamous cell carcinoma (LUSC) (I), and uterine corpus endometrial carcinoma (UCEC) (J) To clarify whether IMMT has a generalization value, we examined the expression profile across cancers and their corresponding normal tissues. We noted that multiple cancer types showed similar expression patterns to BC, including cervical squamous cell carcinoma (CESC), cholangiocarcinoma (CHOL), esophageal carcinoma (ESCA), head-neck squamous cell carcinoma (HNSC), liver hepatocellular carcinoma (LIHC), lung squamous cell carcinoma (LUSC), stomach adenocarcinoma (STAD), and uterine corpus endometrial carcinoma (UCEC) (Fig. 9F). High IMMT had significant associations with shorter RFS time in CESC (Fig. 9G), LIHC (Fig. 9H), LUSC (Fig. 9I), and UCEC (Fig. 9J).
## Pharmacogenetic analysis identified pyridostatin as a potent drug for high IMMT expression
To explore potentially effective pharmaceutical agents targeting BC, we conducted cross-association analyses between drug response and IMMT knockdown using single-guide RNA (sgRNA)-mediated CRSPR in BC cells. Among the 486 screened drugs, 6 drugs were identified to exert altered potency (Fig. 10A). BC cell lines with high sgIMMT efficiency exhibited increased sensitivity to AICAR (Fig. 10B). Meanwhile, BC cell lines with low sgIMMT efficiency exhibited increased sensitivity to MK-1775 (Fig. 10C), luminespib (Fig. 10D), ulixertinib (Fig. 10E), camptothecin (Fig. 10F), wee1 inhibitors (Fig. 10G), and pyridostatin (Fig. 10H). Moreover, CCLE datasets confirmed the relationship between pyridostatin and IMMT expression. The pyridostatin sensitivity correlated negatively with IMMT expression levels (Fig. 10I) and IMMT copy numbers (Fig. 10J). The analyses revealed that pyridostatin exerts a potentiation effect when cells exhibit a high expression of IMMT, indicating that it may be a viable candidate for the development of precision medicine. Fig. 10Pharmacogenetic analysis identified pyridostatin as a potent drug for high IMMT expression. A–J Identification of pyridostatin as a potential drug candidate. Scatter plot showing scores of predictivity and descriptivity for drugs acting on BC cells with various single guide IMMT (sgIMMT) efficacy. Dots in red and blue represent hits with a predictivity p-value of < 0.05 and a descriptivity p-value of 0.05 (A). Boxplots showing –log (IC50) M of AICAR (5-Aminoimidazole-4-carboxamide ribonucleotide) (B), MK-1775(C), luminespib (D), ulixertinib (E), camptothecin (F), wee1 inhibitor (G), and pyridostatin (H). * $p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ between indicated groups. Dot plots showing the correlation of pyridostatin sensitivity with IMMT expression (I) and with IMMT copy number (J)
## Discussion
The physiological function of IMMT in regulating mitochondrial dynamics and preserving mitochondrial structural integrity is an emerging focus of research; meanwhile, its clinicopathological value, association with TIME, and therapeutic implications in patients with BC have yet to be clarified. In this study, we combined multi-omics analysis, clinical validation and cellular experiments to reveal the novel role of IMMT. BC with high IMMT expression serves as an independent diagnostic biomarker, correlates with advanced clinical status, and predicts poor outcome. In addition, we reveal here that the contents of Th1, Th2, MSC, macrophages, basophil, CD4 + T cell and B cell, and TMB levels can counteract the prognostic significance. Our analyses at the single-cell level and of whole tissue samples demonstrate that elevated IMMT is associated with an immunosuppressive TIME. The potential mechanism of IMMT overexpression underlying BC progression may lie in the co-expressed genes implicated in cell cycle progression and mitochondrial antioxidant defenses. Genetic manipulation confirmed that IMMT plays a positive role in mitochondrial function and oxidative stress. Moreover, we demonstrate that the clinicopathological values of IMMT are amenable to ethnic Chinese BC patients and can be generalized to other cancer types, including CESC, LIHC, LUSC, and UCEC. Of particular note, we here identify pyridostatin as an effective drug candidate when BC cells are harboring elevated IMMT expression, thereby offering a promising therapeutic candidate for precision medicine. A schematic model of IMMT’s relevance in the clinical setting, immuno-oncology, redox biology, and precision medicine is illustrated in Fig. 11.Fig. 11Schematic model delineates IMMT’s relevance in diagnosis, prognosis, immuno-oncology, cell cycle progression, mitochondrial antioxidant defenses, oxidative stress, and potential precision treatments for BC Limited information regarding the clinicopathological significance of IMMT in breast cancer is currently available. Although, Suárez-Arroyo et al. have reported an upregulated IMMT protein level in BC cells compared to normal mammary epithelial cells [42]. Other cancer types, such as those in the liver, prostate, colon, and pancreas have also been reported to express altered IMMT levels [43]. Meanwhile, Sotgia et al. reported that high IMMT was associated with an unfavorable outcome in gastric cancer [44]. In this study, IMMT over-expression was noted in BC tissue, and correlated with a shorter RFS in the patient group. High IMMT is associated with several statuses indicating a poor prognosis, including advanced grade, TNBC subtype, nodal status positive, high NPI score, and high KI-67 level.
As disorganization of the mitochondrial structure may result in activation of the immune response via leaked mitochondrial genome [45, 46], IMMT may play a role in the immune response. Indeed, Ghosh et al. have demonstrated that IMMT knockdown in cancer cells causes a catastrophic collapse of mitochondrial integrity and the activation of mitochondrion-directed innate immunity [47]. The bioinformatic investigation in our study demonstrates that IMMT upregulation is associated with immunosuppressed TIME in BC. Nevertheless, it must be noted that a major caveat of the present study lies in the lack of experimental verification of the causal relationship between IMMT and immune signaling activation in BC. Further investigation is thus required to further elucidate the molecular activities involved.
IMMT has been reported as a requirement for tumor cell proliferation, including for osteosarcoma cells, prostate cancer cells, and BC cells [47]. IMMT-knockout cells exhibit dysregulated mitochondrial functions, leading to an arrested cell cycle, reduced cell proliferation, suppressed tumor growth, and increased apoptosis. Similarly, our study shows that the upregulation of IMMT is implicated in activating the cell cycle pathway and is positively associated with key biomarkers relevant to cell proliferation. We also found that IMMT-targeting siRNA hindered cell proliferation and migration.
Pyridostatin has been reported to exert an anti-tumor effect through its G-quadruplexes binding activity, and has been shown to exert a highly specific activity against BRCA$\frac{1}{2}$-deficient tumors [48]. Furthermore, pyridostatin activates the cytoplasmic STING signaling pathway in cancer cells. Our investigation indicates that pyridostatin could be a promising therapeutic drug candidate for BC harboring high IMMT expression levels. As such, pyridostatin is a suitable candidate for further therapeutic development in combination with genetic testing and immunotherapy.
In conclusion, our study demonstrates the novel diagnostic and prognostic significance of IMMT in BC and reveals its role in TIME and the cell cycle. Moreover, the identification of pyridostatin, based on IMMT expression, offers promise for the development of a precision medicine strategy.
## Supplementary Information
Additional file 1: Figure S1. IMMT knockdown inhibits migration and induces lipid peroxidation in MCF-7 cells. Representative images of wound healing assay at 0 h and 22 h (A). Quantification line plot of the wound area determined by the migrated cells (B). * $p \leq 0.05.$ Representative western blot of lipid peroxidation assessment by probing 4-HNE abundance. β-actin as loading control (C). Quantitative bar chart of 4-HNE abundance (D). Figure S2. Kaplan-*Meier analysis* of overall survival probability based on low/high IHC score of IMMT in BC patients with grade 3.
## References
1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F. **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
2. Eroles P, Bosch A, Perez-Fidalgo JA, Lluch A. **Molecular biology in breast cancer: intrinsic subtypes and signaling pathways**. *Cancer Treat Rev* (2012) **38** 698-707. DOI: 10.1016/j.ctrv.2011.11.005
3. Loibl S, Poortmans P, Morrow M, Denkert C, Curigliano G. **Breast cancer**. *Lancet* (2021) **397** 1750-1769. DOI: 10.1016/S0140-6736(20)32381-3
4. Waks AG, Winer EP. **Breast cancer treatment: a review**. *JAMA* (2019) **321** 288-300. DOI: 10.1001/jama.2018.19323
5. Garcia-Martinez L, Zhang Y, Nakata Y, Chan HL, Morey L. **Epigenetic mechanisms in breast cancer therapy and resistance**. *Nat Commun* (2021) **12** 1786. DOI: 10.1038/s41467-021-22024-3
6. El Ansari R, McIntyre A, Craze ML, Ellis IO, Rakha EA, Green AR. **Altered glutamine metabolism in breast cancer; subtype dependencies and alternative adaptations**. *Histopathology* (2018) **72** 183-190. DOI: 10.1111/his.13334
7. Rodrigues T, Ferraz LS. **Therapeutic potential of targeting mitochondrial dynamics in cancer**. *Biochem Pharmacol* (2020) **182** 114282. DOI: 10.1016/j.bcp.2020.114282
8. Kondadi AK, Anand R, Reichert AS. **Cristae membrane dynamics - a paradigm change**. *Trends Cell Biol* (2020) **30** 923-936. DOI: 10.1016/j.tcb.2020.08.008
9. John GB, Shang Y, Li L, Renken C, Mannella CA, Selker JM, Rangell L, Bennett MJ, Zha J. **The mitochondrial inner membrane protein mitofilin controls cristae morphology**. *Mol Biol Cell* (2005) **16** 1543-1554. DOI: 10.1091/mbc.e04-08-0697
10. Feng Y, Madungwe NB, Bopassa JC. **Mitochondrial inner membrane protein, Mic60/mitofilin in mammalian organ protection**. *J Cell Physiol* (2019) **234** 3383-3393. DOI: 10.1002/jcp.27314
11. Bartha A, Gyorffy B. **TNMplot com a web tool for the comparison of gene expression in normal tumor and metastatic tissues**. *Int J Mol Sci* (2021) **22** 2622. DOI: 10.3390/ijms22052622
12. Chandrashekar DS, Bashel B, Balasubramanya SAH, Creighton CJ, Ponce-Rodriguez I, Chakravarthi B, Varambally S. **UALCAN: a portal for facilitating tumor subgroup gene expression and survival analyses**. *Neoplasia* (2017) **19** 649-658. DOI: 10.1016/j.neo.2017.05.002
13. Jezequel P, Gouraud W, Ben Azzouz F, Guerin-Charbonnel C, Juin PP, Lasla H, Campone M. **bc-GenExMiner 4.5: new mining module computes breast cancer differential gene expression analyses**. *Database* (2021). DOI: 10.1093/database/baab007
14. Thul PJ, Akesson L, Wiking M, Mahdessian D, Geladaki A, Ait Blal H, Alm T, Asplund A, Bjork L, Breckels LM. **A subcellular map of the human proteome**. *Science* (2017) **356** 6340. DOI: 10.1126/science.aal3321
15. Uhlen M, Fagerberg L, Hallstrom BM, Lindskog C, Oksvold P, Mardinoglu A, Sivertsson A, Kampf C, Sjostedt E, Asplund A. **Proteomics tissue-based map of the human proteome**. *Science* (2015) **347** 1260419. DOI: 10.1126/science.1260419
16. Uhlen M, Zhang C, Lee S, Sjostedt E, Fagerberg L, Bidkhori G, Benfeitas R, Arif M, Liu Z, Edfors F. **A pathology atlas of the human cancer transcriptome**. *Science* (2017). DOI: 10.1126/science.aan2507
17. Gyorffy B. **Survival analysis across the entire transcriptome identifies biomarkers with the highest prognostic power in breast cancer**. *Comput Struct Biotechnol J* (2021) **19** 4101-4109. DOI: 10.1016/j.csbj.2021.07.014
18. Sun D, Wang J, Han Y, Dong X, Ge J, Zheng R, Shi X, Wang B, Li Z, Ren P. **TISCH: a comprehensive web resource enabling interactive single-cell transcriptome visualization of tumor microenvironment**. *Nucleic Acids Res* (2021) **49** D1420-D1430. DOI: 10.1093/nar/gkaa1020
19. Hong F, Meng Q, Zhang W, Zheng R, Li X, Cheng T, Hu D, Gao X. **Single-cell analysis of the pan-cancer immune microenvironment and scTIME portal**. *Cancer Immunol Res* (2021) **9** 939-951. DOI: 10.1158/2326-6066.CIR-20-1026
20. Fan Z, Chen R, Chen X. **SpatialDB: a database for spatially resolved transcriptomes**. *Nucleic Acids Res* (2020) **48** D233-D237. PMID: 31713629
21. Ru B, Wong CN, Tong Y, Zhong JY, Zhong SSW, Wu WC, Chu KC, Wong CY, Lau CY, Chen I. **TISIDB: an integrated repository portal for tumor-immune system interactions**. *Bioinformatics* (2019) **35** 4200-4202. DOI: 10.1093/bioinformatics/btz210
22. Tang Z, Kang B, Li C, Chen T, Zhang Z. **GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis**. *Nucleic Acids Res* (2019) **47** W556-W560. DOI: 10.1093/nar/gkz430
23. Yoshihara K, Shahmoradgoli M, Martinez E, Vegesna R, Kim H, Torres-Garcia W, Trevino V, Shen H, Laird PW, Levine DA. **Inferring tumour purity and stromal and immune cell admixture from expression data**. *Nat Commun* (2013) **4** 2612. DOI: 10.1038/ncomms3612
24. Liu CJ, Hu FF, Xia MX, Han L, Zhang Q, Guo AY. **GSCALite: a web server for gene set cancer analysis**. *Bioinformatics* (2018) **34** 3771-3772. DOI: 10.1093/bioinformatics/bty411
25. Li B, Severson E, Pignon JC, Zhao H, Li T, Novak J, Jiang P, Shen H, Aster JC, Rodig S. **Comprehensive analyses of tumor immunity: implications for cancer immunotherapy**. *Genome Biol* (2016) **17** 174. DOI: 10.1186/s13059-016-1028-7
26. Li T, Fan J, Wang B, Traugh N, Chen Q, Liu JS, Li B, Liu XS. **Timer: a web server for comprehensive analysis of tumor-infiltrating immune cells**. *Cancer Res* (2017) **77** e108-e110. DOI: 10.1158/0008-5472.CAN-17-0307
27. Li T, Fu J, Zeng Z, Cohen D, Li J, Chen Q, Li B, Liu XS. **TIMER2.0 for analysis of tumor-infiltrating immune cells**. *Nucleic Acids Res* (2020) **48** 509-514. DOI: 10.1093/nar/gkaa407
28. Goldman MJ, Craft B, Hastie M, Repecka K, McDade F, Kamath A, Banerjee A, Luo Y, Rogers D, Brooks AN. **Visualizing and interpreting cancer genomics data via the Xena platform**. *Nat Biotechnol* (2020) **38** 675-678. DOI: 10.1038/s41587-020-0546-8
29. Liu SH, Shen PC, Chen CY, Hsu AN, Cho YC, Lai YL, Chen FH, Li CY, Wang SC, Chen M. **DriverDBv3: a multi-omics database for cancer driver gene research**. *Nucleic Acids Res* (2020) **48** D863-D870. PMID: 31701128
30. Vasaikar SV, Straub P, Wang J, Zhang B. **LinkedOmics: analyzing multi-omics data within and across 32 cancer types**. *Nucleic Acids Res* (2018) **46** D956-D963. DOI: 10.1093/nar/gkx1090
31. Lin HY, Wu HJ, Chen SY, Hou MF, Lin CS, Chu PY. **Epigenetic therapy combination of UNC0638 and CI-994 suppresses breast cancer via epigenetic remodeling of BIRC5 and GADD45A**. *Biomed Pharmacother* (2022) **145** 112431. DOI: 10.1016/j.biopha.2021.112431
32. Lin HY, Weng SW, Shen FC, Chang YH, Lian WS, Hsieh CH, Chuang JH, Lin TK, Liou CW, Chang CS. **Abrogation of toll-like receptor 4 mitigates obesity-induced oxidative stress, proinflammation, and insulin resistance through metabolic reprogramming of mitochondria in adipose tissue**. *Antioxid Redox Signal* (2020) **33** 66-86. DOI: 10.1089/ars.2019.7737
33. Chu PY, Tzeng YT, Tsui KH, Chu CY, Li CJ. **Downregulation of ATP binding cassette subfamily a member 10 acts as a prognostic factor associated with immune infiltration in breast cancer**. *Aging* (2022) **14** 2252-2267. DOI: 10.18632/aging.203933
34. Jeong E, Lee Y, Kim Y, Lee J, Yoon S. **Analysis of cross-association between mrna expression and rnai efficacy for predictive target discovery in colon cancers**. *Cancers* (2020) **12** 3091. DOI: 10.3390/cancers12113091
35. Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, Wilson CJ, Lehar J, Kryukov GV, Sonkin D. **The cancer cell line encyclopedia enables predictive modelling of anticancer drug sensitivity**. *Nature* (2012) **483** 603-607. DOI: 10.1038/nature11003
36. Chang KT, Wu HJ, Liu CW, Li CY, Lin HY. **A novel role of arrhythmia-related gene KCNQ1 revealed by multi-omic analysis: theragnostic value and potential mechanisms in lung adenocarcinoma**. *Int J Mol Sci* (2022) **23** 2274. DOI: 10.3390/ijms23042279
37. Saal LH, Vallon-Christersson J, Hakkinen J, Hegardt C, Grabau D, Winter C, Brueffer C, Tang MH, Reutersward C, Schulz R. **The Sweden cancerome analysis network—breast (SCAN-B) Initiative: a large-scale multicenter infrastructure towards implementation of breast cancer genomic analyses in the clinical routine**. *Genome Med* (2015) **7** 20. DOI: 10.1186/s13073-015-0131-9
38. Inwald EC, Klinkhammer-Schalke M, Hofstadter F, Zeman F, Koller M, Gerstenhauer M, Ortmann O. **Ki-67 is a prognostic parameter in breast cancer patients: results of a large population-based cohort of a cancer registry**. *Breast Cancer Res Treat* (2013) **139** 539-552. DOI: 10.1007/s10549-013-2560-8
39. Zhang SC, Hu ZQ, Long JH, Zhu GM, Wang Y, Jia Y, Zhou J, Ouyang Y, Zeng Z. **Clinical Implications of tumor-infiltrating immune cells in breast cancer**. *J Cancer* (2019) **10** 6175-6184. DOI: 10.7150/jca.35901
40. Teijeira A, Garasa S, Etxeberria I, Gato-Canas M, Melero I, Delgoffe GM. **Metabolic consequences of T-cell costimulation in anticancer immunity**. *Cancer Immunol Res* (2019) **7** 1564-1569. DOI: 10.1158/2326-6066.CIR-19-0115
41. Van Laar VS, Otero PA, Hastings TG, Berman SB. **Potential role of Mic60/Mitofilin in Parkinson's disease**. *Front Neurosci* (2018) **12** 898. DOI: 10.3389/fnins.2018.00898
42. Suarez-Arroyo IJ, Feliz-Mosquea YR, Perez-Laspiur J, Arju R, Giashuddin S, Maldonado-Martinez G, Cubano LA, Schneider RJ, Martinez-Montemayor MM. **The proteome signature of the inflammatory breast cancer plasma membrane identifies novel molecular markers of disease**. *Am J Cancer Res* (2016) **6** 1720-1740. PMID: 27648361
43. Eramo MJ, Lisnyak V, Formosa LE, Ryan MT. **The 'mitochondrial contact site and cristae organising system' (MICOS) in health and human disease**. *J Biochem* (2020) **167** 243-255. DOI: 10.1093/jb/mvz111
44. Sotgia F, Lisanti MP. **Mitochondrial biomarkers predict tumor progression and poor overall survival in gastric cancers: companion diagnostics for personalized medicine**. *Oncotarget* (2017) **8** 67117-67128. DOI: 10.18632/oncotarget.19962
45. McArthur K, Whitehead LW, Heddleston JM, Li L, Padman BS, Oorschot V, Geoghegan ND, Chappaz S, Davidson S, San Chin H. **BAK/BAX macropores facilitate mitochondrial herniation and mtDNA efflux during apoptosis**. *Science* (2018). DOI: 10.1126/science.aao6047
46. Rigon M, Townley AR, Campanella M. **Mitochondria ensure immune surveillance by retro-communication with the nucleus**. *Cell Metab* (2021) **33** 853-855. DOI: 10.1016/j.cmet.2021.04.013
47. Ghosh JC, Perego M, Agarwal E, Bertolini I, Wang Y, Goldman AR, Tang HY, Kossenkov AV, Landis CJ, Languino LR. **Ghost mitochondria drive metastasis through adaptive GCN2/Akt therapeutic vulnerability**. *Proc Natl Acad Sci U S A* (2022). DOI: 10.1073/pnas.2115624119
48. Groelly FJ, Porru M, Zimmer J, Benainous H, De Visser Y, Kosova AA, Di Vito S, Serra V, Ryan A, Leonetti C. **Anti-tumoural activity of the G-quadruplex ligand pyridostatin against BRCA1/2-deficient tumours**. *EMBO Mol Med* (2022) **14** e14501. DOI: 10.15252/emmm.202114501
|
---
title: 'Associations of genetic markers of diabetes mellitus with carotid atherosclerosis:
a community-based case–control study'
authors:
- Tzu-Wei Wu
- Chao-Liang Chou
- Chun-Fang Cheng
- Shu-Xin Lu
- Yih-Jer Wu
- Li-Yu Wang
journal: Cardiovascular Diabetology
year: 2023
pmcid: PMC9999522
doi: 10.1186/s12933-023-01787-7
license: CC BY 4.0
---
# Associations of genetic markers of diabetes mellitus with carotid atherosclerosis: a community-based case–control study
## Abstract
### Background
Diabetes mellitus (DM) is a well-established determinant of atherosclerosis and cardiovascular diseases (CVD). Recently, genome-wide association studies (GWAS) identified several single nucleotide polymorphism (SNP) significantly correlated with DM. The study aimed to explore the relationships of the top significant DM SNPs with carotid atherosclerosis (CA).
### Methods
We used a case–control design and randomly selected 309 cases and 439 controls with and without, respectively, carotid plaque (CP) from a community-based cohort. Eight recent GWAS on DM in East Asians reported hundreds of SNPs with genome-wide significance. The study used the top significant DM SNPs, with a p-value < 10–16, as the candidate genetic markers of CA. The independent effects of these DM SNPs on CA were assessed by multivariable logistic regression analyses to control the effects of conventional cardio-metabolic risk factors.
### Results
Multivariable analyses showed that, 9 SNPs, including rs4712524, rs1150777, rs10842993, rs2858980, rs9583907, rs1077476, rs7180016, rs4383154, and rs9937354, showed promising associations with the presence of carotid plaque (CP). Among them, rs9937354, rs10842993, rs7180016, and rs4383154 showed significantly independent effects. The means (SD) of the 9-locus genetic risk score (9-GRS) of CP-positive and -negative subjects were 9.19 (1.53) and 8.62 (1.63), respectively ($p \leq 0.001$). The corresponding values of 4-locus GRS (4-GRS) were 4.02 (0.81) and. 3.78 (0.92), respectively ($p \leq 0.001$). The multivariable-adjusted odds ratio of having CP for per 1.0 increase in 9-GRS and 4-GRS were 1.30 ($95\%$ CI 1.18–1.44; $$p \leq 4.7$$ × 10–7) and 1.47 ($95\%$ CI 1.74–9.40; $$p \leq 6.1$$ × 10–5), respectively. The means of multi-locus GRSs of DM patients were similar to those of CP-positive subjects and higher than those of CP-negative or DM-negative subjects.
### Conclusions
We identified 9 DM SNPs showing promising associations with CP. The multi-locus GRSs may be used as biomarkers for the identification and prediction of high-risks subjects for atherosclerosis and atherosclerotic diseases. Future studies on these specific SNPs and their associated genes may provide valuable information for the preventions of DM and atherosclerosis.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12933-023-01787-7.
## Background
Atherosclerosis is the gradual constriction of arteries by plaque formation within the artery walls which can reduce blood flow by > $50\%$ [1]. Atherosclerosis is a disease of slow progression and can be aggravated by other existing conditions such as hypertension [2]. The development of atherosclerosis lesions is characterized by life-long altered lipid accumulation, especially low-density lipoprotein (LDL), and chronic inflammation in the arterial wall [1, 3]. The blood vessel endothelium then responds to blood flow turbulence and activates the recruitment of circulating immune cells. Penetrating monocytes are differentiated into macrophages which take up lipid through phagocytosis and turn into foam cells in plaques [4]. Later stages of atherosclerosis are characterized by the formation of unstable plaques which rupture and result in a local clot [5], leading to more serious clinical cardiac events such as myocardial infarction and stroke [6]. Atherosclerosis has been a leading cause of morbidity and mortality in the world especially in developed countries while the incidence of this disease is also increasing in developing countries [7]. In Taiwan, five of the top ten causes of death are related to atherosclerosis [8].
Diabetes mellitus (DM) is a group of metabolic disorders symbolized by chronic hyperglycemia leading to symptoms such as polyuria, polydipsia, and polyphagia. DM results from the low level of insulin production and/or insulin resistance of the target tissues [9]. Based on the cause of metabolic abnormalities, diabetes is mainly grouped into 2 types. Type 1 diabetes is caused by the loss of insulin-producing pancreatic β-cells due to autoimmunity, and is mostly diagnosed in children and adolescents. Type 2 diabetes accounts for about $90\%$ of all diabetes, which begins with insulin resistance, however, the severity of the disease increase with a decline in β-cell function [10]. It is estimated that over 450 million people have diabetes worldwide and 4.2 million die because of it annually [11].
DM is a well-known risk determinant of vascular events [12]. There is a considerable increase in the literature related to DM and atherosclerosis in recent decades. Insulin resistance, which is an early preclinical stage of DM, induces several syndromes, including dyslipidemia, inflammation, hypertension, endothelial dysfunction, and vascular smooth muscle cell proliferation, which are correlated with the development of atherosclerosis [13]. Our previous community-based study also demonstrated significant relationships between DM and development and severity of carotid atherosclerosis [14]. To further explore the underlying mechanisms relating DM to atherosclerosis is scientific relevance. Recently, several genome-wide association studies (GWAS) in East Asians have identified hundreds of single nucleotide polymorphisms (SNPs) showing significant associations with DM [15–22]. It is reasonable to assume that these DM SNPs may account for the development of atherosclerosis. Therefore, we explored the relationships between the top 43 significant DM SNPs, with a p-value of < 10–16, and carotid atherosclerosis by a case–control study which enrolled 748 community-dwelling middle-aged adults and elders.
## Study subjects
The study used a case–control design to explore the relationships between DM SNPs and carotid atherosclerosis. The study performed stratified random sampling procedure to select study subjects from a community-based cohort, which enrolled middle-aged adults and elders from 3 townships in the northern coastal area of Taiwan [23]. From September 2010 to May 2013, a total of 1607 residents aged 40-to-74 years voluntarily provided informed consent and were enrolled. Twenty-seven subjects who lack good quality of recorded carotid ultrasound images and another 1 individual who lack blood pressure data were excluded. Another 40 subjects who had a positive history of physician-diagnosed myocardial infarction or had ever received a cardiac catheter or stent were excluded, leaving a total of 1539 subjects in the cohort.
Of the cohort members, 409 of them had detectable extracranial carotid plaques (CP). The study randomly selected 309 CP-positive individuals as the case group. The control group was a random sample of 439 individuals who had no detectable extracranial carotid plaque. The study complied with the 1975 Helsinki Declaration on ethics in medical research and were reviewed and approved by the institutional review boards of MacKay Medical College (No. P990001) and MacKay Memorial Hospital (No. 14MMHIS075).
## Measurements of anthropometric attributes and biochemical profiles
The measurements of anthropometric attributes and biochemical profiles have been described previously [23]. In brief, we used a digital system (BW-2200; NAGATA Scale Co. Ltd., Tainan, Taiwan) to measure the subject’s body weight and height. Waist circumference (WC) was measured at the level of mid-distance between the bottom of the rib cage and the top of the iliac crest. Hip circumference was the distance around the largest part of the subject’s hips. Blood pressure was measured three times, with an interval of 3 min, after 10 min of rest. The averages of repeated measurements of systolic blood pressure (SBP) and diastolic blood pressure (DBP) were used for analyses. The fasting blood levels of total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglycerides (FTG), and glucose (FPG) were determined by an autoanalyzer (Toshiba TBA c16000; Toshiba Medical System, Holliston, MA, USA) with commercial kits (Denka Seiken, Tokyo, Japan).
We also used a structured questionnaire to collect personal histories of common diseases in adults and health behaviors. In the study, hypertension was defined as subjects who had physician-diagnosed hypertension or a history of taking antihypertensive medications. Hyperlipidemia was defined as subjects having been diagnosed with high blood lipids by a physician or having a history of taking lipid-lowering medications. DM was defined as FPG ≥ 126 mg/dL or the use of insulin or other hypoglycemic agents. Cigarette smoking and alcohol drinking were defined as having smoked cigarettes or drank alcohol-containing beverages at least 4 days per week during the past month before enrollment.
## Determination of carotid plaque
The measurements of carotid atherosclerosis had been described previously [24]. In brief, we used high-resolution B-mode ultrasonography systems (GE Healthcare Vivid 7 and Vivid E9; General Electric Company, Milwaukee, USA) and followed the protocol recommended by the American Society of Echocardiography [25] to obtain the transverse and cross-sectional ultrasound images of the left and right carotid arteries. The thickness between the lumen-intima and media-adventitia interfaces was measured blindly by using automatic contouring software (GE Healthcare EchoPAC version 112.0.2; General Electric-Vingmed, Horten, Norway). In the study, a plaque was defined as a focal protrusion $50\%$ greater than the surrounding vessel wall, an intima-media thickness (IMT) ≥ 1.5 mm, or local thickening ≥ 0.5 mm [26].
## Candidate SNPs for genetic association study
There were 8 GWAS studies on DM in East Asians [18–22] enrolled in GWAS Catalog [27]. These GWAS studies identified more than 380 SNPs with genome-wide significance [27]. This study focused on the relationships between the top 95 significant DM SNPs (with a p-value < 10–16) and carotid atherosclerosis (Additional file 1: Table S1). These top significant DM SNPs or their closely linked SNPs were considered for the genetic association study.
We used the Ensemble Genome Browser [28] to retrieve the linkage disequilibrium (LD) data in the 1000 Human Genome Project Phase 3-Southern Han Chinese [29]. The cut-off LD (r2) value of linkage was set at 0.80. Among these top 95 significant DM SNPs, 22 of them are LD with more significant SNPs, leaving a total of 73 independent DM SNPs. Among these independent DM SNPs, rs3816157, rs2844623, rs610930, rs12549902, rs13266634, and rs2383208 are the designed SNPs of the plate. Besides, another 38 SNPs of the array plate are closely linked with DM SNPs (r2 > 0.80). Consequently, a total of 44 SNPs were regarded as the candidate SNPs.
In the study, we used a plate (Axiom® CHB 1 Array Plate; Affymetrix Ltd, Santa Clara, CA, USA) to determine the genotypes of these 44 DM SNPs. All genotyping was performed by the National Center for Genome Medicine, Academic Sinica, Taiwan. The frequency distributions of genotypes of these 44 DM SNPs in CP-positive and -negative individuals are shown in Additional file 1: Table S2. The call rates of all typed SNPs were greater than $95\%$ and the relative frequencies of the minor alleles of all typed SNPs were greater than $5\%$. Yet, SNP rs13342692 was excluded from association analysis for violation of the Hardy–Weinberg Equilibrium, leaving a total of 43 SNPs for association evaluation.
## Statistical analyses
In the study, we used the student’s t and the Pearson’s chi-square tests to compare whether there were significant differences in the anthropometric and biochemical measurements between CP-positive and -negative subjects. All anthropometric and clinical factors, which showed significant associations with carotid atherosclerosis in uni-variable analyses, were subject to multi-variable analyses. We used logistic regression model with stepwise selection method to obtain the best-fit model which includes conventional cardio-metabolic risk factors only. Then, each DM SNPs was separately added to the best-fit model to assess their independent effect on carotid atherosclerosis. The strength of association between each SNP and carotid atherosclerosis was manifested by multivariable-adjusted odds ratio (OR). To reduce the influence of false negativity, we used 0.10 as the pre-set inclusion criteria of promising SNPs. We further generated multi-locus genetic risk scores (GRSs) by summing the number of risk alleles or genotypes for each individual and then assessed the associations between GRSs and carotid atherosclerosis. All statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).
## Results
The anthropometric and clinical characteristics of CP-positive and -negative subjects are shown in Table 1. The CP-positive subjects had significantly higher means of age, BMI, waist circumference, waist-to-hip ratio, blood pressures, LDL-C, and blood glucose than the CP-negative subjects. As compared to CP-negative subjects, CP-positive subjects also had significantly higher proportions of the male sex, schooling year < 12 years, hypertension, DM, and cigarette smoking. The mean of HDL-C was significantly lower in the CP-positive subjects as compared to that of the CP-negative subjects. Table 1Comparisons of clinical characteristics between individuals with and without carotid plaque (CP)VariablesCP-negative ($$n = 439$$)CP-positive ($$n = 309$$)p-valuesMeanSDMeanSDAge at enrollment (years)52.68.658.09.3< 0.0001Body weight (kg)64.011.465.211.20.13Body height (cm)161.38.1160.78.60.34BMI (kg/m2)24.53.625.23.40.011Waist circumference (cm)80.89.483.59.20.0001Hip circumference (cm)95.29.195.36.90.84Waist-to-hip ratio (%)84.96.987.67.3< 0.0001SBP (mm Hg)127.219.0132.719.40.0001DBP (mm Hg)79.313.381.313.70.050Total cholesterol (mg/dL)207.936.9212.341.60.12LDL-C (mg/dL)123.833.2129.336.50.035HDL-C (mg/dL)56.116.353.514.50.025Fasting plasma triglyceride (mg/dL)117.983.0128.2103.40.15Fasting plasma glucose (mg/dL)98.326.0106.035.80.0015n%n%Male sex20646.916453.10.098Schooling years < 12 years28865.623174.80.0089Hypertension15735.817055< 0.0001Hyperlipidemia32173.123375.40.48Diabetes mellitus398.95919.1< 0.0001Cigarette smoking6514.871230.0043 Multivariable logistic regression analyses of the conventional cardio-metabolic risk factors showed that the best-fit model contained age, hypertension, DM, and cigarette smoking (Table 2). The multivariable-adjusted ORs of having CP per 5.0 years increases in age was 1.38 ($95\%$ CI 1.26–1.52) and were 1.66 ($95\%$ CI 1.19–2.26), 1.60 ($95\%$ CI 1.01–2.54,), and 2.34 ($95\%$ CI 1.51–3.64) for hypertension, DM, and cigarette smoking, respectively. Table 2Association analyses for carotid atherosclerosis with baseline clinical characteristicsVariableUnivariableMulti-variableOR ($95\%$ CI)OR ($95\%$ CI)Age (per 5.0 years)1.39*** (1.28–1.52)1.38*** (1.26–1.52)Male sex1.34+ (0.98–1.82)–BMI (per 5.0 kg/m2)1.24* (1.00–1.54)–Waist circumference (per 5.0 cm)1.12* (1.04–1.22)–Waist-to-hip ratio (per $5.0\%$)1.24** (1.10–1.38)–SBP (per 10.0 mm Hg)1.08* (1.00–1.18)–DBP (per 10.0 mm Hg)1.08 (0.96–1.21)–LDL-C (per 10.0 mg/dL)1.05* (1.00–1.10)–HDL-C (per 10.0 mg/dL)0.88* (0.79–0.97)–Glucose (per 10.0 mg/dL)1.06* (1.00–1.10)–Schooling years < 12 years (Y/N)1.08 (0.76–1.53)–Hypertension (Y/N)1.73** (1.27–2.37)1.66** (1.19–2.26)*Diabetes mellitus* (Y/N)2.42*** (1.57–3.74)1.60* (1.01–2.54)Cigarette smoking (Y/N)2.39*** (1.59–3.57)2.34*** (1.51–3.64)+0.05 < $p \leq 0.1$; *0.005 < $p \leq 0.05$; **0.0001 < $p \leq 0.005$; ***$p \leq 0.0001$ After adjustment for the effects of age, hypertension, and cigarette smoking, SNPs rs4712524, rs1150777, rs10842993, rs2858980, rs9583907, rs1077476, rs7180016, rs4383154, and rs9937354 showed promising associations with the presences of CP and were subject to further association analyses (Table 3). Model 1 showed that the multivariable-adjusted ORs of having CP was significantly elevated for rs9937354 G allele and were borderline significance for rs4712524 AA or AG genotypes, rs10842993 AG or GG genotypes, rs7180016 AA or GG genotypes, and rs4383154 AA or GG genotypes. The results of stepwise selection showed that rs10842993 AG or GG genotypes, rs7180016 AA or GG genotypes, rs4383154 AA or GG genotypes, and rs9937354 G allele were all correlated with significantly elevated ORs of having CP (Model 2).Table 3Association analyses for 9 promising SNPs with carotid atherosclerosisCandidate SNPTyped SNPRisk allele (genotype)Reference allele (genotype)MultivariableModel 1Model 2ORa ($95\%$ CI)ORb ($95\%$ CI)rs4712523rs4712524AG or AAGG1.51+ (0.91–2.49)–rs4711389rs1150777ACAA or CC1.31 (0.88–1.93)–rs3751236rs10842993AG or GGAA1.55+ (0.98–2.45)1.69* (1.08–2.66)rs7983505rs2858980AG1.23 (0.92–1.64)–rs9515905rs9583907CT1.23 (0.94–1.61)–rs4924455rs1077476TG1.20 (0.95–1.52)–rs8026714rs7180016AA or GGAG1.30+ (0.94–1.79)1.39* (1.02–1.91)rs117267808rs4383154AA or GGAG1.68+ (0.99–2.84)1.72* (1.03–2.88)rs1421085rs9937354GA1.37* (1.01–1.87)1.37* (1.01–1.86)aORs for each risk allele (genotype) were adjusted for age, sex, cigarette smoking, and hypertensionbORs were obtained from the best fit model contained all genetic markers and were adjusted for age, sex, cigarette smoking, and hypertension+0.05 < $p \leq 0.1$; *0.005 < $p \leq 0.05$; **0.0001 < $p \leq 0.005$; ***$p \leq 0.0001$ Table 4 shows that CP-positive individuals had significantly higher mean of 4-locus GRS (4-GRS) than CP-negative individuals (4.02 ± 0.81 vs. 3.78 ± 0.92, $$p \leq 1.4$$ × 10–4). As compared with individuals who had a 4-GRS of 3 or less, the multivariable-adjusted OR of having CP for a 4-GRS of 3, 4, and 5 were 2.02 ($95\%$ CI 0.87–4.74), 2.94 ($95\%$ CI 1.29–6.69), and 4.04 ($95\%$ CI 1.74–9.40), respectively. The multivariable-adjusted OR of having CP for per 1.0 increase in 4-GRS was 1.47 ($95\%$ CI 1.22–1.77; $$p \leq 6.1$$ × 10–5). Similar results were observed for 9-locus GRS (9-GRS). The multivariable-adjusted OR of having CP for per 1.0 increase in 9-GRS was 1.30 ($95\%$ CI 1.18–1.44; $$p \leq 4.7$$ × 10–7).Table 4Association analyses for carotid atherosclerosis with multi-locus genetic risk scores (GRS)VariableCP-negative ($$n = 439$$)CP-positive ($$n = 309$$)Multivariablen (%)n (%)ORa ($95\%$ CI)9-locus GRSbMean (SD)8.62 (1.63)9.19 (1.53)GRS 4–645 (10.3)12 (3.9)1.00 7–8160 (46.7)80 (25.9)1.91+ (0.92–3.97) 9–10179 (40.8)152 (49.2)3.37** (1.65–6.89) ≥ 1155 (12.5)65 (21.0)5.59*** (2.56–12.19)Per 1.0 risk score1.30*** (1.18–1.44)4-locus GRScMean (SD)3.78 (0.92)4.02 (0.81)GRS 0–235 (8.0)9 (2.9)1.00 3119 (27.1)68 (22.0)2.02 (0.87–4.74) 4187 (42.6)138 (44.7)2.94* (1.29–6.69) 598 (22.3)94 (30.4)4.04*** (1.74–9.40)Per 1.0 risk score1.47*** (1.22–1.77)aORs were adjusted for age, sex, cigarette smoking, and hypertensionbrs4712524-rs1150777-rs10842993-rs2858980-rs9583907-rs1077476-rs7180016-rs4383154-rs9937354crs10842993- rs7180016-rs4383154-rs9937354+0.05 < $p \leq 0.1$; *0.005 < $p \leq 0.05$; **0.0001 < $p \leq 0.005$; ***$p \leq 0.0001$
The means (SD) of 9-GRS and 4-GRS in DM patients were 8.91 (1.53) and 4.01 (0.82), respectively. The corresponding values in individuals who had no history of DM were 8.85 (0.73) and 3.86 (0.89), respectively. Test statistics showed that DM patients had non-significantly higher means of GRSs than individuals without DM history.
## Discussion
In the present study, we performed a case–control study that enrolled 309 CP-positive subjects and 439 CP-negative subjects from a community-based cohort. Multivariable analyses of anthropometric attributes and biochemical profiles showed that DM was one of the significantly independent predictors of the best-fit regression model for the presence of CP. Among 43 tested DM SNPs, 9 of them showed promising associations with carotid atherosclerosis after controlling the effects of age, cigarette smoking, and hypertension. Although not all these promising SNPs showed significantly independent effects by multivariable analyses, there was a significantly linear trend between their composite indicator 9-GRS and the risks of carotid atherosclerosis. Four SNPs, including rs9937354, rs10842993, rs7180016, and rs4383154 showed significantly independent effects with carotid atherosclerosis. Moreover, their composite variable 4-GRS also positively linearly correlated with significantly higher ORs of having CP. Additionally, DM patients had higher means of GRSs than individuals without DM history, yet the differences were not statistically significant.
DM has been sharing many common risk factors with increased CVD risk and recent studies have shown connections under different contents [12]. DM increases the risk of ischemic stroke in the general population and in some studies DM also rises perioperative neurological complications and mortality in patients [30]. Hoke et al. followed 1065 patients with neurological asymptomatic carotid atherosclerosis as evaluated by duplex sonography prospectively during a median of 11.8 years for cause-specific mortality [31]. Multivariable regression analysis showed that the risk for all-cause and cardiovascular death of DM patients remained significantly higher even after adjustment for various established cardiovascular risk factors. In a meta-analysis of 18 studies (17,106 patients), DM was also associated with a significantly increased risk of restenosis after carotid surgical revascularization [32]. Recently, specific noncoding RNAs including microRNAs are found to be strongly associated with both DM and CVD [33, 34].
Many pathological mechanisms, including dyslipidemia with an increased level of atherogenic LDL, hyperglycemia with advanced glycation end-product formation, inflammation, and oxidative stress, have been shown to connect DM with atherosclerosis [35]. Increased levels of apolipoprotein B and atherogenic LDL were found in patients with DM [36]. Atherogenic dyslipidemia in diabetes consists of elevated serum concentrations of TG-rich lipoproteins, a high prevalence of LDL, and low concentrations of HDL [37]. Not all LDL is atherogenic and circulating LDL is the major source of lipids to be accumulated in atherosclerotic plaques. Desialylation followed by multiple enzymatic and non-enzymatic modifications results in atherogenic LDL and increases blood atherogenicity [3]. High blood glucose in DM patients can induce glycation and glycoxidation of proteins which in turn induce adhesion molecule expression in endothelium and promote the entrance of monocytes and macrophages during plaque formation. These modified proteins also promote pro-inflammatory cytokine release. Excess glycation of extracellular matrix proteins promotes their interactions with macrophages, endothelial cells, and vascular smooth muscle cells, which results in pro-inflammatory effects [38]. Atherosclerosis is currently regarded as a chronic inflammatory condition. High glucose and hyperglycemia increase neutrophil extracellular trap, a specific type of inflammatory response, which could be involved in atherosclerotic lesion formation [39, 40]. Increased ROS production and decreased antioxidant activity are known to be associated with DM and atherosclerosis development [41, 42]. In this study, DM patients had a mean 4-GRS of 4.01, which was at a clearly elevated risk stratum for carotid atherosclerosis. Our study provided an unconventional link between DM and carotid atherosclerosis, further investigation of these linking SNPs may potentially discover novel underlying mechanisms between DM and atherogenesis.
In this study, we found that rs9937354, which locates in the 1st intron of the FTO gene, was significantly associated with the presence of CP. We retrieved the expression data in human cells by using the Ensemble Genome Browser and found that the expression levels of FTO in multiple tissues, including heart, aorta artery, pancreas, and thyroid, are significantly correlated with rs9937354 polymorphism [43]. FTO gene is known to be associated with obesity and type 2 DM across diverse ethnic backgrounds [44, 45]. The FTO protein is a member of Fe(II)- and 2-oxoglutarate-dependent dioxygenases superfamily and plays a role in the demethylation of nucleic acid [46]. FTO catalyzes the demethylation of m6A to regulate the processing, maturation, and translation of the mRNAs [47]. Hepatic FTO regulates glucose and lipid metabolism and its expression is regulated by metabolic signals such as nutrients and hormones [48]. Overexpression of FTO results in increased lipid accumulation in liver and muscle cells and reduces atherogenic dyslipidemia [49]. FTO is also found to inhibit macrophage lipid influx and accelerate cholesterol efflux, which delays foam cell formation and atherosclerosis development [50].
We also observed significant associations between the presence of CP and polymorphisms of rs7180016, rs4383154, and rs10842993. SNP rs7180016 is an intron variant of the genes encoding protein regulator of cytokinesis 1 (PRC1) and PRC1-antisense 1 (PRC1-AS1). It is closely linked with 2 splice region variants (rs2301825 and rs17636091) and 1 synonymous variant (rs2301826). SNP rs7180016 also closely links with 5 3’-UTR variants and 9 noncoding transcript exon variants [51]. The expressions of PRC1 and PRC1-AS1 genes in dozens of tissues, including adipose, fat, and pancreas, are significantly correlated with rs7180016 polymorphism [51]. Recently, Ndiaye et al. found that the decreased expression of PRC1 significantly influences insulin secretion from EndoC-βH1 cells, which is an immortalized human beta cell line [52]. PRC1 knockdown, by using siPRC1, significantly decreased the viability of EndoC-βH1 cells. Further network analyses showed that PRC1 was related to ‘concentration of D-glucose’, ‘quantity of insulin in blood’, ‘apoptosis of islets of Langerhans’, and ‘quantity of pancreatic cells’ [52]. More recently, Peiris et al. used an experimental approach combining *Drosophila* genetic and insulin assays with human islet genetics to explore the roles of 40 human DM genes. They identified 3 genes, including fascetto, CG9650, and optix, orthologs of human PRC1, and BCL11A, and SIX3 respectively, significantly related with in vivo insulin output [53]. These in vitro and in vivo evidences indicate that PRC1 may play key roles in glucose homeostasis. Further studies are necessary to validate the roles of PRC1 in the development of atherosclerosis.
SNP rs4383154 locates in the 1st intron of glycoprotein 2 (GP2) gene. SNP rs4383154 is completely linked with a synonymous variant rs73541251 and closely linked with a missense variant rs78193826 (r2 = 0.945 in Southern Han Chinese) [29]. The rs78193826 T allele results in an amino acid change from valine to methionine, which may affect protein structure and function of GP2. Additionally, the expression of GP2 gene in salivary gland, thyroid, and brain are influenced by rs4383154 polymorphism [54]. Although GP2 was identified decades ago, its’ function is still poorly understood. GP2 is predominately expressed in the pancreas and plays an antibacterial role in the gastrointestinal tract after being secreted from pancreatic acinar cells [55]. Recent meta-analysis studies showed that 3 nearby up-stream SNPs, including rs12597579, rs12597682, and rs57508503, were significantly correlated with body mass index in East Asians [56, 57]. However, these three SNPs are not closely linked with rs4383154 in Sothern Hans Chinese [29]. The role of GP2 in the pathogenesis of atherosclerotic diseases needs further exploration.
SNP rs10842993 is an intron variant of an uncategorized gene LOC105369709. It is a regulatory region variant and closely linked with 8 nearby regulatory region variants. Except for LOC105369709, Kelch-like protein 42 (KLHL42) gene is the closest gene to rs10842993. Additionally, the expressions of KLHL42 in multiple tissues, including arteries, blood, heart, and multiple immune cells, are significantly influenced by rs10842993 polymorphism [58]. The differential expressions of KLHL42 in multiple vascular and immune cells indicate that KLHL42 may play critical roles in immunological responses of cardiovascular system. Our speculation was supported by a recent bioinformatics study. Lu et al. analyzed the GSE100927 expression data, which contained 69 atherosclerotic carotid arteries and 35 normal carotid arteries, and identified KLHL42 as one of the down-regulated hub genes of atherosclerosis [59]. There is no other report correlates KLHL42 with atherosclerosis and vascular events. The pathogenesis of atherosclerosis in which KLHL42 involves is needed for exploration.
## Conclusions
The study enrolled study subjects from a community-based cohort and identified critical genetic variants linking DM with atherosclerosis. Our results shed light on the mechanisms between DM and atherosclerosis development. The multi-locus GRSs showed significant association with atherosclerosis and may be used as biomarkers for the identification and prediction of high-risks subjects for atherosclerosis and atherosclerotic diseases. Future studies on the functions of these specific SNPs and their associated genes may provide more informative evidences for the prevention of DM and atherosclerosis.
## Supplementary Information
Additional file 1: Table S1. Information of the top 95 significant DM genetic markers. Table S2. Genotype distributions of 44 DM SNPs in CP-positive and -negative subjects.
## References
1. Björkegren JLM, Lusis AJ. **Atherosclerosis: recent developments**. *Cell* (2022.0) **185** 1630-1645. DOI: 10.1016/j.cell.2022.04.004
2. Vasdev S, Gill V, Singal P. **Role of advanced glycation end products in hypertension and atherosclerosis: therapeutic implications**. *Cell Biochem Biophys* (2007.0) **49** 48-63. DOI: 10.1007/s12013-007-0039-0
3. Summerhill VI, Grechko AV, Yet SF, Sobenin IA, Orekhov AN. **The atherogenic role of circulating modified lipids in atherosclerosis**. *Int J Mol Sci* (2019.0) **20** 3561. DOI: 10.3390/ijms20143561
4. Cheng F, Torzewski M, Degreif A, Rossmann H, Canisius A, Lackner KJ. **Impact of glutathione peroxidase-1 deficiency on macrophage foam cell formation and proliferation: implications for atherogenesis**. *PLoS ONE* (2013.0) **8** e72063. DOI: 10.1371/journal.pone.0072063
5. Galis ZS, Sukhova GK, Lark MW, Libby P. **Increased expression of matrix metalloproteinases and matrix degrading activity in vulnerable regions of human atherosclerotic plaques**. *J Clin Invest* (1994.0) **94** 2493-2503. DOI: 10.1172/JCI117619
6. Robinson JG, Fox KM, Bullano MF, Grandy S. **Atherosclerosis profile and incidence of cardiovascular events: a population- based survey**. *BMC Cardiovasc Disord* (2009.0) **9** 46. DOI: 10.1186/1471-2261-9-46
7. Roth GA, Mensah GA, Johnson CO, Addolorato G, Ammirati E, Baddour LM. **Global burden of cardiovascular diseases and risk factors, 1990–2019: update from the GBD 2019 study**. *J Am Coll Cardiol* (2020.0) **76** 2982-3021. DOI: 10.1016/j.jacc.2020.11.010
8. Li YH, Ueng KC, Jeng JS, Charng MJ, Lin TH, Chien KL. **2017 Taiwan lipid guidelines for high risk patients**. *J Formos Med Assoc* (2017.0) **116** 217-248. DOI: 10.1016/j.jfma.2016.11.013
9. Kharroubi AT, Darwish HM. **Diabetes mellitus: the epidemic of the century**. *World J Diabetes* (2015.0) **6** 850-867. DOI: 10.4239/wjd.v6.i6.850
10. Stancáková A, Javorský M, Kuulasmaa T, Haffner SM, Kuusisto J, Laakso M. **Changes in insulin sensitivity and insulin release in relation to glycemia and glucose tolerance in 6,414 Finnish men**. *Diabetes* (2009.0) **58** 1212-1221. DOI: 10.2337/db08-1607
11. 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, 9th edition**. *Diabetes Res Clin Pract* (2019.0) **157** 107843. DOI: 10.1016/j.diabres.2019.107843
12. Cosentino F, Grant PJ, Aboyans V, Bailey CJ, Ceriello A, Delgado V. **2019 ESC Guidelines on diabetes, pre-diabetes, and cardiovascular diseases developed in collaboration with the EASD**. *Eur Heart J* (2020.0) **41** 255-323. DOI: 10.1093/eurheartj/ehz486
13. Di Pino A, DeFronzo RA. **Insulin resistance and atherosclerosis: implications for insulin-sensitizing agents**. *Endocr Rev* (2019.0) **40** 1447-1467. DOI: 10.1210/er.2018-00141
14. Wu TW, Chou CL, Cheng CF, Lu SX, Wang LY. **Prevalences of diabetes mellitus and carotid atherosclerosis and their relationships in middle-aged adults and elders: a community-based study**. *J Formos Med Assoc* (2022.0) **121** 1133-1140. DOI: 10.1016/j.jfma.2021.10.005
15. Yasuda K, Miyake K, Horikawa Y, Hara K, Osawa H, Furuta H. **Variants in**. *Nat Genet* (2008.0) **40** 1092-1097. DOI: 10.1038/ng.207
16. Takeuchi F, Serizawa M, Yamamoto K, Fujisawa T, Nakashima E, Ohnaka K. **Confirmation of multiple risk loci and genetic impacts by a genome-wide association study of type 2 diabetes in the Japanese population**. *Diabetes* (2009.0) **58** 1690-1699. DOI: 10.2337/db08-1494
17. Cui B, Zhu X, Xu M, Guo T, Zhu D, Chen G. **A genome-wide association study confirms previously reported loci for type 2 diabetes in Han Chinese**. *PLoS ONE* (2011.0) **6** e22353. DOI: 10.1371/journal.pone.0022353
18. Hara K, Fujita H, Johnson TA, Yamauchi T, Yasuda K, Horikoshi M. **Genome-wide association study identifies three novel loci for type 2 diabetes**. *Hum Mol Genet* (2014.0) **23** 239-246. DOI: 10.1093/hmg/ddt399
19. Imamura M, Takahashi A, Yamauchi T, Hara K, Yasuda K, Grarup N. **Genome-wide association studies in the Japanese population identify seven novel loci for type 2 diabetes**. *Nat Commun* (2016.0) **7** 10531. DOI: 10.1038/ncomms10531
20. Suzuki K, Akiyama M, Ishigaki K, Kanai M, Hosoe J, Shojima N. **Identification of 28 new susceptibility loci for type 2 diabetes in the Japanese population**. *Nat Genet* (2019.0) **51** 379-386. DOI: 10.1038/s41588-018-0332-4
21. Ishigaki K, Akiyama M, Kanai M, Takahashi A, Kawakami E, Sugishita H. **Large-scale genome-wide association study in a Japanese population identifies novel susceptibility loci across different diseases**. *Nat Genet* (2020.0) **52** 669-679. DOI: 10.1038/s41588-020-0640-3
22. Spracklen CN, Horikoshi M, Kim YJ, Lin K, Bragg F, Moon S. **Identification of type 2 diabetes loci in 433,540 East Asian individuals**. *Nature* (2020.0) **582** 240-245. DOI: 10.1038/s41586-020-2263-3
23. Wu TW, Hung CL, Liu CC, Wu YJ, Wang LY, Yeh HI. **Associations of cardiovascular risk factors with carotid intima-media thickness in middle-age adults and elders**. *J Atheroscler Thromb* (2017.0) **24** 677-686. DOI: 10.5551/jat.37895
24. Chou CL, Wu YJ, Hung CL, Liu CC, Wang SD, Wu TW. **Segment-specific prevalence of carotid artery plaque and stenosis in middle-aged adults and elders in Taiwan: a community-based study**. *J Formos Med Assoc* (2019.0) **118** 64-71. DOI: 10.1016/j.jfma.2018.01.009
25. Stein JH, Korcarz CE, Hurst RT, Lonn E, Kendall CB, Mohler ER. **Use of carotid ultrasound to identify subclinical vascular disease and evaluate cardiovascular disease risk: a consensus statement from the American Society of Echocardiography Carotid Intima-Media Thickness Task Force. Endorsed by the Society for Vascular Medicine**. *J Am Soc Echocardiogr* (2008.0) **21** 93-111. DOI: 10.1016/j.echo.2007.11.011
26. 26.Touboul PJ, Hennerici MG, Meairs S, Adams H, Amarenco P, Bornstein N, et al. Mannheim carotid intima-media thickness and plaque consensus (2004–2006–2011). An update on behalf of the advisory board of the 3rd, 4th and 5th watching the risk symposia, at the 13th, 15th and 20th European Stroke Conferences, Mannheim, Germany, 2004, Brussels, Belgium, 2006, Hamburg, Germany, 2011. Cerebrovasc Dis 2012;34:290–6.
27. Buniello A, MacArthur JAL, Cerezo M, Harris LW, Hayhurst J, Malangone C. **The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019**. *Nucleic Acids Res* (2019.0) **47** D1005-D1012. DOI: 10.1093/nar/gky1120
28. Cunningham F, Allen JE, Allen J, Alvarez-Jarreta J, Amode MR, Armean IM. **Ensembl 2022**. *Nucleic Acids Res* (2022.0) **50** D988-D995. DOI: 10.1093/nar/gkab1049
29. Yates A, Akanni W, Amode MR, Barrell D, Billis K, Carvalho-Silva D. **Ensembl 2016**. *Nucleic Acids Res* (2016.0) **44** D710-D716. DOI: 10.1093/nar/gkv1157
30. Dimic A, Markovic M, Vasic D, Dragas M, Zlatanovic P, Mitrovic A. **Impact of diabetes mellitus on early outcome of carotid endarterectomy**. *Vasa* (2019.0) **48** 148-156. DOI: 10.1024/0301-1526/a000737
31. Hoke M, Schillinger M, Minar E, Goliasch G, Binder CJ, Mayer FJ. **Carotid ultrasound investigation as a prognostic tool for patients with diabetes mellitus**. *Cardiovasc Diabetol* (2019.0) **18** 90. DOI: 10.1186/s12933-019-0895-0
32. Texakalidis P, Tzoumas A, Giannopoulos S, Jonnalagadda AK, Jabbour P, Rangel-Castilla L. **Risk factors for restenosis after carotid revascularization: a meta-analysis of hazard ratios**. *World Neurosurg* (2019.0) **125** 414-424. DOI: 10.1016/j.wneu.2019.02.065
33. Mahjoob G, Ahmadi Y, Fatima Rajani H, Khanbabaei N, Abolhasani S. **Circulating microRNAs as predictive biomarkers of coronary artery diseases in type 2 diabetes patients**. *J Clin Lab Anal* (2022.0) **36** e24380. DOI: 10.1002/jcla.24380
34. Tang N, Jiang S, Yang Y, Liu S, Ponnusamy M, Xin H, Yu T. **Noncoding RNAs as therapeutic targets in atherosclerosis with diabetes mellitus**. *Cardiovasc Ther* (2018.0) **36** e12436. DOI: 10.1111/1755-5922.12436
35. Poznyak A, Grechko AV, Poggio P, Myasoedova VA, Alfieri V, Orekhov AN. **The diabetes mellitus-atherosclerosis connection: the role of lipid and glucose metabolism and chronic inflammation**. *Int J Mol Sci* (2020.0) **21** 1835. DOI: 10.3390/ijms21051835
36. Albers JJ, Marcovina SM, Imperatore G, Snively BM, Stafford J, Fujimoto WY. **Prevalence and determinants of elevated apolipoprotein B and dense low-density lipoprotein in youths with type 1 and type 2 diabetes**. *J Clin Endocrinol Metab* (2008.0) **93** 735-742. DOI: 10.1210/jc.2007-2176
37. Hirano T. **Pathophysiology of diabetic dyslipidemia**. *J Atheroscler Thromb* (2018.0) **25** 771-782. DOI: 10.5551/jat.RV17023
38. Katakami N. **Mechanism of development of atherosclerosis and cardiovascular disease in diabetes mellitus**. *J Atheroscler Thromb* (2018.0) **25** 27-39. DOI: 10.5551/jat.RV17014
39. Menegazzo L, Ciciliot S, Poncina N, Mazzucato M, Persano M, Bonora B. **NETosis is induced by high glucose and associated with type 2 diabetes**. *Acta Diabetol* (2015.0) **52** 497-503. DOI: 10.1007/s00592-014-0676-x
40. Nahrendorf M, Swirski FK. **Immunology. Neutrophil-macrophage communication in inflammation and atherosclerosis**. *Science* (2015.0) **349** 237-238. DOI: 10.1126/science.aac7801
41. Chew P, Yuen DY, Stefanovic N, Pete J, Coughlan MT, Jandeleit-Dahm KA. **Antiatherosclerotic and renoprotective effects of ebselen in the diabetic apolipoprotein E/GPx1-double knockout mouse**. *Diabetes* (2010.0) **59** 3198-3207. DOI: 10.2337/db10-0195
42. Di Marco E, Jha JC, Sharma A, Wilkinson-Berka JL, Jandeleit-Dahm KA, de Haan JB. **Are reactive oxygen species still the basis for diabetic complications?**. *Clin Sci* (2015.0) **129** 199-216. DOI: 10.1042/CS20150093
43. 43.Website https://www.ensembl.org/Homo_sapiens/Variation/Mappings?db=core;r=16:53765435-53766435; v=rs9937354;vdb=variation;vf=730580740; Accessed on 15 Nov 2022.
44. Frayling TM, Timpson NJ, Weedon MN, Zeggini E, Freathy RM, Lindgren CM. **A common variant in the**. *Science* (2007.0) **316** 889-894. DOI: 10.1126/science.1141634
45. Yeo GS. **The role of the FTO (Fat Mass and Obesity Related) locus in regulating body size and composition**. *Mol Cell Endocrinol* (2014.0) **397** 34-41. DOI: 10.1016/j.mce.2014.09.012
46. Gerken T, Girard CA, Tung YC, Webby CJ, Saudek V, Hewitson KS. **The obesity-associated**. *Science* (2007.0) **318** 1469-1472. DOI: 10.1126/science.1151710
47. Fu Y, Jia G, Pang X, Wang RN, Wang X, Li CJ. **FTO-mediated formation of N6-hydroxymethyladenosine and N6-formyladenosine in mammalian RNA**. *Nat Commun* (2013.0) **4** 1798. DOI: 10.1038/ncomms2822
48. Mizuno TM. **Fat Mass and Obesity Associated (FTO) gene and hepatic glucose and lipid metabolism**. *Nutrients* (2018.0) **10** 1600. DOI: 10.3390/nu10111600
49. Kang H, Zhang Z, Yu L, Li Y, Liang M, Zhou L. **FTO reduces mitochondria and promotes hepatic fat accumulation through RNA demethylation**. *J Cell Biochem* (2018.0) **119** 5676-5685. DOI: 10.1002/jcb.26746
50. Mo C, Yang M, Han X, Li J, Gao G, Tai H. **Fat mass and obesity-associated protein attenuates lipid accumulation in macrophage foam cells and alleviates atherosclerosis in apolipoprotein E-deficient mice**. *J Hypertens* (2017.0) **35** 810-821. DOI: 10.1097/HJH.0000000000001255
51. 51.Website https://www.ensembl.org/Homo_sapiens/Variation/Explore?r=15:90982545-90983545;v=rs7180016;vdb=variation;vf=105439139; Accessed on 15 Nov 2022
52. Ndiaye FK, Ortalli A, Canouil M, Huyvaert M, Salazar-Cardozo C, Lecoeur C. **Expression and functional assessment of candidate type 2 diabetes susceptibility genes identify four new genes contributing to human insulin secretion**. *Mol Metab* (2017.0) **6** 459-470. DOI: 10.1016/j.molmet.2017.03.011
53. Peiris H, Park S, Louis S, Gu X, Lam JY, Asplund O. **Discovering human diabetes-risk gene function with genetics and physiological assays**. *Nat Commun* (2018.0) **9** 3855. DOI: 10.1038/s41467-018-06249-3
54. 54.Website https://www.ensembl.org/Homo_sapiens/Variation/Explore?r=16:20316844-20317844;v=rs78193826;vdb=variation;vf=733841797; Accessed on 15 Nov 2022.
55. Lin Y, Nakatochi M, Sasahira N, Ueno M, Egawa N, Adachi Y. **Glycoprotein 2 in health and disease: lifting the veil**. *Genes Environ* (2021.0) **43** 53. DOI: 10.1186/s41021-021-00229-8
56. Wen W, Cho YS, Zheng W, Dorajoo R, Kato N, Qi L. **Meta-analysis identifies common variants associated with body mass index in east Asians**. *Nat Genet* (2012.0) **44** 307-311. DOI: 10.1038/ng.1087
57. Sakaue S, Kanai M, Tanigawa Y, Karjalainen J, Kurki M, Koshiba S. **A cross-population atlas of genetic associations for 220 human phenotypes**. *Nat Genet* (2021.0) **53** 1415-1424. DOI: 10.1038/s41588-021-00931-x
58. 58.Website https://www.ensembl.org/Homo_sapiens/Variation/Explore?r=12:27811215-27812215;v=rs10842993;vdb=variation;vf=731358096; Accessed on 15 Nov 2022.
59. Lu Y, Zhang X, Hu W, Yang Q. **The identification of candidate biomarkers and pathways in atherosclerosis by integrated bioinformatics analysis**. *Comput Math Methods Med* (2021.0) **2021** 6276480. DOI: 10.1155/2021/6276480
|
---
title: Hypermethylation of ACADVL is involved in the high-intensity interval training-associated
reduction of cardiac fibrosis in heart failure patients
authors:
- Chih-Chin Hsu
- Jong-Shyan Wang
- Yu-Chiau Shyu
- Tieh-Cheng Fu
- Yu-Hsiang Juan
- Shin-Sheng Yuan
- Chao-Hung Wang
- Chi-Hsiao Yeh
- Po-Cheng Liao
- Hsin-Yi Wu
- Pang-Hung Hsu
journal: Journal of Translational Medicine
year: 2023
pmcid: PMC9999524
doi: 10.1186/s12967-023-04032-7
license: CC BY 4.0
---
# Hypermethylation of ACADVL is involved in the high-intensity interval training-associated reduction of cardiac fibrosis in heart failure patients
## Abstract
### Background
Emerging evidence suggests that DNA methylation can be affected by physical activities and is associated with cardiac fibrosis. This translational research examined the implications of DNA methylation associated with the high-intensity interval training (HIIT) effects on cardiac fibrosis in patients with heart failure (HF).
### Methods
Twelve HF patients were included and received cardiovascular magnetic resonance imaging with late gadolinium enhancement for cardiac fibrosis severity and a cardiopulmonary exercise test for peak oxygen consumption (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak). Afterwards, they underwent 36 sessions of HIIT at alternating $80\%$ and $40\%$ of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak for 30 min per session in 3–4 months. Human serum from 11 participants, as a means to link cell biology to clinical presentations, was used to investigate the exercise effects on cardiac fibrosis. Primary human cardiac fibroblasts (HCFs) were incubated in patient serum, and analyses of cell behaviour, proteomics ($$n = 6$$) and DNA methylation profiling ($$n = 3$$) were performed. All measurements were conducted after completing HIIT.
### Results
A significant increase ($$p \leq 0.009$$) in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak (pre- vs. post-HIIT = 19.0 ± 1.1 O2 ml/kg/min vs. 21.8 ± 1.1 O2 ml/kg/min) was observed after HIIT. The exercise strategy resulted in a significant decrease in left ventricle (LV) volume by $15\%$ to $40\%$ ($p \leq 0.05$) and a significant increase in LV ejection fraction by approximately $30\%$ ($$p \leq 0.010$$). LV myocardial fibrosis significantly decreased from 30.9 ± $1.2\%$ to 27.2 ± $0.8\%$ ($$p \leq 0.013$$) and from 33.4 ± $1.6\%$ to 30.1 ± $1.6\%$ ($$p \leq 0.021$$) in the middle and apical LV myocardium after HIIT, respectively. The mean single-cell migration speed was significantly ($$p \leq 0.044$$) greater for HCFs treated with patient serum before (2.15 ± 0.17 μm/min) than after (1.11 ± 0.12 μm/min) HIIT. Forty-three of 1222 identified proteins were significantly involved in HIIT-induced altered HCF activities. There was significant ($$p \leq 0.044$$) hypermethylation of the acyl-CoA dehydrogenase very long chain (ACADVL) gene with a 4.474-fold increase after HIIT, which could activate downstream caspase-mediated actin disassembly and the cell death pathway.
### Conclusions
Human investigation has shown that HIIT is associated with reduced cardiac fibrosis in HF patients. Hypermethylation of ACADVL after HIIT may contribute to impeding HCF activities. This exercise-associated epigenetic reprogramming may contribute to reduce cardiac fibrosis and promote cardiorespiratory fitness in HF patients.
Trial registration: NCT04038723. Registered 31 July 2019, https://clinicaltrials.gov/ct2/show/NCT04038723.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12967-023-04032-7.
## Introduction
Excessive deposition of extracellular matrix proteins derived from cardiac fibroblasts contributes to pathologic cardiac remodelling [1]. Cardiac fibrosis impairs the transverse connection between cardiomyocytes to give rise to abnormal cardiac mechanical and electrical functions [2]. In recent clinical studies, cardiac fibrosis has been identified as an independent predictive factor for major adverse cardiovascular events, including sudden cardiac death, myocardial infarction, heart failure (HF), or ventricular tachycardia [3, 4]. These fatal complications pave the way for therapies to attenuate cardiac fibrosis.
Physical exercise has been acknowledged as a nonpharmacological approach to reduce the health burden of cardiovascular disease [5, 6]. High-intensity interval training (HIIT) is characterized by alternating short periods of exercise at ≥ $80\%$ of one’s peak oxygen consumption (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak) interspersed with less intense exercise at 40–$50\%$ of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak to allow recovery [7, 8]. Several studies have reported that HIIT is associated with improved left ventricle (LV) geometry [7–9] and is beneficial for survival in HF patients [8]. Animal studies have revealed that chronic aerobic exercise reduces cardiac fibrosis [10, 11]. Although HIIT comes with many health benefits [7, 8, 10, 11], human investigation of the basic science behind the HIIT effects on cardiac fibrosis is still insufficient.
Cardiovascular magnetic resonance (CMR) imaging has been recognized as the gold standard in determining cardiac function owing to its high reproducibility and accuracy in assessing cardiac anatomy [12]. Gadolinium chelates are extracellular contrast agents with a delayed wash-out feature in fibrotic myocardium [13]. The extracellular volume (ECV), estimated by CMR imaging with late gadolinium enhancement (CMR-LGE), in patients with dilated cardiomyopathy was reported to be similar to the severity of cardiac fibrosis found at autopsy [14]. Therefore, the ECV fraction has been used to predict the severity of cardiac fibrosis [4] because it is sensitive to myocardial fibrosis [15].
Cardiac proteomic studies bridge the gap between transcription information and gene regulation at the cell and tissue levels [16]. Myokines, such as secreted protein acidic and rich in cysteine [17], and galectin-3 are implicated in the pathogenesis of cardiac fibrosis [18–20]. However, interactions between the above biomarkers and cardiac fibrosis are still debated [19, 20]. Emerging evidence suggests that DNA methylation [21] has been linked to cardiac fibrosis and can be affected by physical activities [22]. Exercise-induced attenuation of the migratory and proliferative capabilities of human cardiac fibroblasts (HCFs) has been proposed as a novel cardioprotective mechanism [23]. Therefore, we hypothesized that HIIT reduced cardiac fibrosis by modulating the cardiac proteomic profile through DNA methylation in HCFs. To verify this hypothesis, a translational study quantified cardiac fibrosis in HF patients by CMR-LGE imaging and defined cell biology features from a serum-treated HCF model before and after HIIT. The findings regarding HIIT-associated proteogenomic characteristics for cardiac fibrosis may provide additional insights into approaches for HF patients.
## Participants
The research was carried out according to the Declaration of Helsinki. The author’s institutional review board approved the study, and the Clinical Trial Registry number is NCT04038723. All participants provided their written informed consent after understanding the experimental procedure. HF patients, diagnosed according to the Framingham HF diagnostic criteria [24], who had stable clinical presentations ≥ 4 weeks and received individualized patient education under optimized guideline-based management [25], were initially surveyed. Individuals who were > 80 years old and < 20 years old, were unable to perform exercise due to other noncardiac diseases, were pregnant, would have future cardiac transplantation within 6 months, had uncompensated HF, and had an estimated glomerular filtration rate < 30 ml/min/1.73 m2 were not enrolled in the study. We also excluded individuals with absolute contraindications for exercise suggested by the American College of Sports Medicine (ACSM) [26].
## Experimental procedure
Baseline clinical information, including age, sex, body mass index (BMI), disease duration, comorbidities, medication, and LV ejection fraction (LVEF) obtained from 2D echocardiography [27], from each subject was recorded. Participants had blood sampling before the baseline CMR-LGE imaging study and then underwent a graded cardiopulmonary exercise test (CPET). The physical component score (PCS) and mental component score (MCS) of the Medical Outcomes Study Short Form-36 health survey (SF-36) for quality of life (QoL) were assessed before initiating each CPET. The follow-up CMR-LGE, CPET, and blood samplings were performed within 1 week after completing 36 sessions of HIIT. Haematocrit and b-type natriuretic peptide (BNP) were also measured before and after HIIT. After completing the above study, the remaining blood sample was centrifuged at 2500 rpm for 5 min at room temperature for serum preparation. A graphic depicting the experimental procedure is shown in Additional file 1.
## Exercise training
We followed the previous protocol [7] for the hospital-based HIIT program using a bicycle ergometer (Ergoselect 150P, ergoline GmbH, Germany). Briefly, participants exercised at alternating intensities of 3-min intervals of $80\%$ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak and 3-min intervals of $40\%$ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak for 30 min in each session. They were instructed to complete 36 sessions of exercise training with a frequency of 2 to 3 sessions per week.
## Graded cardiopulmonary exercise test
All participants underwent a graded CPET on a bicycle ergometer (Ergoselect 150P, ergoline GmbH, Bitz, Germany) within 1 week before HIIT. Minute ventilation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VE) and oxygen consumption (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2) were measured breath by breath using a computer-based system (CareFusion MasterScreen CPX, CPX International Inc., Germany). \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak was defined as described in the ACSM guidelines for graded CPETs [26]. The oxygen uptake efficiency sloe (OUES) during exercise was determined as described in our previous work [7]. A noninvasive continuous cardiac output (CO) monitoring system (NICOM, Cheetah Medical, Wilmington, DE, USA) was used to measure peak CO (COex) during CPET. The CPET procedure and determination of cardiorespiratory parameters are detailed in Additional file 2.
## Cardiovascular magnetic resonance imaging with late gadolinium enhancement
All participants were scheduled to have CMR-LGE examination just before each CPET. CMR-LGE examination involved a 3.0-Tesla Skyra scanner (Siemens Medical Systems, Erlangen, Germany) operating on the VD13 platform with a 32-channel phased-array receiver body coil. Short-axis (contiguous 8-mm-thick slices) and standard long-axis view (2-, 3- and 4-chamber views) cine images were obtained by steady-state free precession (SSFP) cine imaging with the following parameters: repetition time, 45 ms; echo time, 1.4 ms; matrix, 256 × 256; and field of view, 34 to 40 cm. LV geometry as well as functions, including LV end-diastolic volume (LVEDV), LV end-systolic volume (LVESV), resting CO (COrest), LVEF, LV mass, and left ventricle wall motion (LVWMS) were determined using SSFP cine imaging. The lower the LVWMS is, the better the LV contractility [28].
Quantitative parametric images of myocardial extracellular volume (ECV) fractions were acquired from longitudinal relaxation time (T1) mapping in short-axis slices before (pre) and after (post) contrast medium enhancement. The ECV was estimated by the following equation:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\large ECV=(1-hematocrit) \dfrac{(\dfrac{1}{{T1}_{myo\,post}}-\dfrac{1}{{T1}_{myo\,pre}})}{(\dfrac{1}{{T1}_{blood\,post}}-\dfrac{1}{({T1}_{blood\,pre}})}$$\end{document}ECV=(1-hematocrit)(1T1myopost-1T1myopre)(1T1bloodpost-1(T1bloodpre) The CMR-LGE system determines the T1 in each myocardial segment. Myocardial fibrosis was estimated with a modified Look-Locker inversion-recovery (MOLLI) sequence [15] acquired during the end-expiratory phase in the basal, middle and apical LV myocardial segments at short-axes before (T1myo pre) and approximately 15 to 20 min after (T1myo post) a 0.1 mmol/kg intravenous dose of gadolinium-DOTA (gadoterate meglumine, Dotarem, Guerbet S.A., France). The ECV value was further normalized by the blood T1 mapping image before (T1blood pre) and after (T1blood post) enhancement in the corresponding short-axis slices. The basal slice (Base), mid-cavity slice (Middle), and apical slice (Apex) of LV myocardial segments [29] were drawn along the epicardial and endocardial surfaces on matched pre- and post-contrast MOLLI images to identify the myocardium for ECV analysis.
## Cell migration assay
We used $10\%$ patient serum before and after HIIT, replacing $10\%$ foetal bovine serum (FBS), to treat HCFs isolated from adult ventricles (HCF-av cell, ScienCell Research Laboratories, Carlsbad, CA) for 5–10 passages to observe serum effects on cell behaviours. The HCFs in different media were prepared as described in Additional file 2 for time-lapse image studies. The migration speed was estimated from serial images according to the persistent random walk equation [30].
## Cell proliferation assay
Prepared live HCFs for 5–10 passages were stained with Hoechst 33342 (Thermo Fisher Scientific Inc., Waltham, MA) and were then separately treated with $10\%$ FBS, $10\%$ participant serum before HIIT, or $10\%$ participant serum after HIIT. Cell numbers at 0, 24, and 48 h after harvesting with the three different culture media were estimated. The relative cell count (RCC) was calculated as the cell number measured at each time point divided by that at 0 h (see Additional file 2 for the cell proliferation assay).
## Immunofluorescence staining
Procedures of HCFs (5–10 passages) prepared for immunofluorescent staining of mitochondria, β-actin, and actin-related protein-2 (Arp2) are detailed in Additional file 2.
## Proteomic analysis
HCFs (2.1 × 105 for 5–10 passages) were inoculated in a Petri dish with a 60-mm diameter (Sigma‒Aldrich) and then treated as described for the above cell behaviour assays. After 24 h of incubation in pre- and post-HIIT serum, cells were collected and homogenized in lysis buffer (8 M urea in 50 mM triethyl ammonium bicarbonate buffer, pH 8). The prepared sample was further analysed by nano-LC–ESI–MS on an Orbitrap LUMOS mass spectrometer (Thermo Fisher Scientific Inc.) for label-free quantification of the protein profile (see Additional file 2 for the proteomic study procedure).
## DNA methylation profiling
HCFs were cultured in pre- and post-HIIT serum from HF patients for DNA methylation profiling. Genomic DNA was isolated from the cells, and the detailed procedure can be seen in Additional file 2.
## Protein analysis before and after knockdown of the acyl-CoA dehydrogenase very long chain (ACADVL) gene
The ACADVL gene encodes for very long-chain acyl-CoA dehydrogenase (VLCAD), which functions within mitochondria and is essential for fatty acid oxidation. HCFs were prepared for western blot analysis of VLCAD, caspase-3 (CASP3), cytochrome c (Cyto C), lamin B1, β-actin, and Arp2 with the internal reference protein glyceraldehyde 3-phosphate dehydrogenase (GAPDH) before knockdown of the ACADVL gene. The above proteins were quantified again after knockdown of ACADVL. Detailed methods of the western blotting and knockdown procedure are provided in Additional file 2.
Knockdown of DNMT1 leads to generally decreased DNA methylation and activates cascades of genotoxic stress [31] in cells, resulting in signal transduction unrelated to cardiac fibrosis. Thus, we preferred to downregulate the ACADVL gene expression to simulate the HIIT-associated inhibition of human cardiac fibroblast activities.
## Bioinformatics analysis
The differences in proteomic and DNA methylation profiling before and after HIIT were estimated by ingenuity pathway analysis (IPA, Qiagen, Hilden, Germany), and signal transduction was analysed by the KEGG pathway database (see Additional file 2).
## Statistical analysis
Values are shown as the mean ± standard error of mean (SEM), and error bars for scatter dot plots represent one SEM. Since aerobic capacity and cardiac fibrosis are significant clinical outcomes related to the survival of HF patients [4, 8], power (1- β) analysis for paired sample t tests used to compare the difference in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak and ECV fractions before and after HIIT. Differences in physical PCS, MCS, and LVWMS were estimated by the chi-square test.
The nonparametric test was used in the study owing to the limited sample size. The Wilcoxon signed rank test was conducted to estimate within-group differences between data before and after HIIT, including exercise capacity function, CMR-LGE results (LV geometry, functions, and ECV fractions), and blood chemistry data. The Mann‒Whitney U test was used to estimate differences in selected protein amounts obtained from LC‒MS results and methylation levels between cells incubated in patient serum before and after HIIT. Relationships between the DNMT1 levels and health-related physical fitness and CMR-LGE findings were assessed by Spearman’s correlation analysis.
Relative protein expression (measurements/baseline) of VLCAD, Cyto C, CASP3, lamin B1, actin and Arp2 in HCFs between the original and knockdown of ACADVL was compared by the Mann‒Whitney U test. This test was also used to assess mitochondrial intensity in HCFs treated with patient serum before and after HIIT and in cells with and without ACADVL knockdown. Kruskall-Wallis test was conducted to assess cell migration speed in three different culture media and with different cell numbers at different times (baseline, 24 h and 48 h after inoculation). Multiple comparisons Dunn’s test was used to estimate differences of cell behaviours between each of the above sampling time. The relationships between normalized changes (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\Delta} {\text{Value}} =\dfrac{ {\text{Value}}_{{\text{post}} - {\text{HIIT}}}\,-\,{\text{Value}}_{{\text{pre}} - {\text{HIIT}}}}{{\text{Value}}_{{\text{pre}} - {\text{HIIT}}}}$$\end{document}ΔValue=Valuepost-HIIT-Valuepre-HIITValuepre-HIIT) in exercise performance and CMR-LGE measurements after HIIT were estimated by Spearman correlation and partial correlation analysis after controlling LV mass. All statistical assessments were considered significant at $p \leq 0.05.$
## Demographics
Twelve HF patients, with a mean age of 56.5 ± 3.9 years and stable disease, were enrolled in the study, and one participant smoked 20 cigarettes per day for 42 years (see Additional file 3 for the baseline demographics table). All participants completed 36 sessions of HIIT during a mean of 4.17 ± 0.31 months. We were not able to collect serum from one participant who refused to have venous blood sampling after completing HIIT. Overall, 11 of the 12 participants completed CMR-LGE imaging, BNP, and cell behaviour assays two times. After completing the above assessments, an adequate serum amount to harvest HCFs for proteomic analysis before and after HIIT was available in six HF patients. Only a limited serum amount for cell culture to elucidate the HIIT effects on DNA methylation profiles of HCFs was available in three HF patients.
## HIIT improved cardiorespiratory fitness
Participants showed a significant improvement in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak after completing 36 sessions of the HIIT (Table 1). The estimated statistical power for comparing the difference in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak before and after HIIT (Δ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak) was greater than 0.8 (https://homepage.univie.ac.at/robin.ristl/samplesize.php?test=pairedttest). The Δ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak showed significant positive correlations ($r = 0.642$, $$p \leq 0.045$$) with the difference in oxygen uptake efficiency slope (ΔOUES) but a significant inverse correlation (r = -0.637, $$p \leq 0.048$$) with the change in LVESV (ΔLVESV). The ΔCOex was significantly negatively correlated (r = -0.637, $$p \leq 0.048$$) with the change in total ECV (ΔECV) before and after HIIT. ΔCOex showed a significant negative partial correlation (r = − 0.702, $$p \leq 0.035$$) with the ΔECV fraction of the apical LV myocardium segment after adjusting for the LV mass. The detailed correlation findings can be found in Additional file 4.Table 1Effects of high-intensity interval training (HIIT) on aerobic capacity and quality of lifeAssessmentPre-HIITPost-HIITp valueAerobic Capacity\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak, O2 mL/min/kg19.0 ± 1.121.8 ± 1.10.009*COex, L/min10.1 ± 0.611.3 ± 0.70.036*OUES, O2 ml/min/log (L/min)664 ± 51710 ± 520.023*SF-36PCS51.6 ± 2.054.5 ± 1.40.158MCS46.1 ± 2.750.5 ± 1.90.182CMR ImageLV mass, g119 ± 18119 ± 120.721COrest, L/min4.3 ± 0.35.2 ± 0.30.182LVESV, mL82.7 ± 2051.4 ± 150.006*LVEDV, mL139 ± 22118 ± 180.033*LVEF, %45.9 ± 4.559.8 ± 3.70.010*LVWMS28.7 ± 1.422.3 ± 1.60.012†Cardiac StressBNP, pg/mL99.0 ± 2924.8 ± 5.80.003*Data are shown as the mean ± SEMBNP b-type natriuretic peptide, CMR cardiovascular magnetic resonance, CO cardiac output, COex CO during exercise, COrest resting CO, EF ejection fraction, EDV end-diastolic volume, ESV end-systolic volume, HIIT high-intensity interval training, LV left ventricle, MCS mental component score, PCS physical component score, OUES oxygen uptake efficient slope, QoL quality of life, SEM standard error of mean, SF-36 short form 36 questionnaire, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VE minute ventilation, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak peak oxygen consumption, WMS wall motion scoreSignificant difference between pre- and post-HIIT: Wilcoxon signed rank test (*), chi-square test (†)
## Reduced LV volume and improved LV contractility after HIIT
Chest roentgenograms typically showed a trend of decreased heart size after HIIT (see Additional file 5 for typical chest X-ray findings before and after HIIT). CMR images of significant decreases in LVESV and LVEDV supported the chest X-ray findings. HIIT was associated with significant improvement in contractility reflected in increased COex. CMR measurements of the significant increases in LVEF and decreases in LVWMS reinforced the increased contractility after HIIT. A striking decrease in BNP might suggest that HIIT could relieve cardiac stress. Details of the HIIT-associated physiological adaptation are shown in Table 1.
## HIIT reversed cardiac remodelling by reducing cardiac fibrosis
ECV fractions were derived from pre- and post-contrast CMR-LGE images of the LV in the short axis (Fig. 1A). The total, Base, Middle, and Apex of LV myocardial segments were used for estimation of the ECV. After determination of the T1 relaxation time (Fig. 1B) in pre- and post-contrast CMR-LGE images (Fig. 1C) in each myocardial segment, ECV fractions of 11 participants before and after HIIT were estimated. ECV fractions of the LV myocardium decreased significantly from 31.7 ± $1\%$ to 28.8 ± $1.0\%$ ($$p \leq 0.028$$) in total, from 30.9 ± $1.2\%$ to 27.2 ± $0.8\%$ ($$p \leq 0.013$$) at the Middle, and from 33.4 ± $1.6\%$ to 30.1 ± $1.6\%$ ($$p \leq 0.021$$) at the Apex, respectively. HIIT did not cause significant decreases in ECV fractions at the LV Base (pre-HIIT vs. post-HIIT = 31.2 ± $1.1\%$ vs. 29.6 ± $1.3\%$, $$p \leq 0.321$$) (Fig. 1D). The estimated statistical power for the difference in ECV fractions in each LV myocardial segment before and after HIIT was greater than 0.8 (https://homepage.univie.ac.at/robin.ristl/samplesize.php?test=pairedttest).Fig. 1Quantification of cardiac fibrosis in heart failure (HF) patients ($$n = 11$$) pre- and post-high-intensity interval trainings (HIIT). A. The region of interest (square) in short axis T1 mapping cardiovascular magnetic resonance with late-gadolinium enhanced (CMR-LGE) imaging of the left ventricle (LV) was used for myocardium extracellular volume (ECV) fraction studies to quantify the severity of LV fibrosis. B. The left panel shows the relaxation time (T1) of each region of interest (ROI) in a slice at middle cavity. The T1 was used to calculate the extracellular volume (ECV) fraction determined by the CMR-LGE system to estimate the myocardial fibrosis in the slice. The right panel shows the corresponding ROI. C. CMR-LGE imaging demonstrated ECV fractions at the basal, mid-cavity (Middle) and apical (Apex) myocardium segments pre- and post-HIIT. D. The mean ECV fractions of the total, Base, Middle, and Apex LV myocardium segments decreased significantly post-HIIT. Ant., Anterior; Post, Posterior; Sup., Superior; Inf., Inferior; RV, Right ventricle
## Impaired cardiac fibroblast activities after HIIT
Cell growth curves and movement were examined in HCFs ($$n = 11$$) incubated with $10\%$ FBS and serum from 11 participants before and after HIIT. Growth curves in HCFs treated with patient serum were significantly higher ($p \leq 0.01$) than those treated with FBS. The RCC was nonsignificantly lower for HCFs treated with patient serum after than before HIIT (see Additional file 6). At 24 and 48 h, the RCC in cells treated with patient serum before and after HIIT was ~ 1.2 times and ~ 1.4 times that in cells treated with FBS. The mean single-cell migration speed was significantly greater for HCFs treated with patient serum before than after HIIT and with FBS (2.15 ± 0.17 μm/min vs. 1.11 ± 0.12 μm/min, $$p \leq 0.044$$, and 1.13 ± 0.24 μm/min, $$p \leq 0.019$$) (see Additional files 6, 7).
## DNA (cytosine-5)-methyltransferase 1 (DNMT1) expression increased in cells harvested with serum after HIIT
The 1222 identified protein manifestations differed between the residual 6 paired samples at different exercise statuses. In total, 191 protein levels that differed between each pair of samples were included for further analysis. IPA revealed that 43 of the 191 proteins were involved in impaired cell movement, decreased cell proliferation, and increased cell death (see Additional file 8). Protein expression levels in cells before and after HIIT were different (Fig. 2A). The post-HIIT level of DNMT1, a remarkable epigenetic marker, was 3.993 times that before HIIT and was the protein with the greatest increase in the proteome profile analysis (Fig. 2B).Fig. 2Proteome profiling of primary human cardiac fibroblasts treated with heart failure patient ($$n = 6$$) serum pre- and post-HIIT. A. Heatmap showing the different protein presentations between cells treated with pre- and post-HIIT serum. B. Volcano plot showing the different protein presentations (blue transparent dots) between cells treated with serum pre- and post-HIIT. The labelled protein gene name (red transparent dots) selected by ingenuity pathway analysis (IPA) are significantly involved in cell movement, cell death and cell proliferation
## HIIT was associated with significant hypermethylation of the ACADVL gene
Genomic DNA ($$n = 3$$) was isolated, and the gene methylation levels on 9476 5’-cytosine-phosphate-guanine-3’ (CpG) islands after HIIT were quantified. A global DNA methylation study identified changes in DNA methylation profiles in 6977 genes (Fig. 3A). Among them, 3830 identified genes were hypermethylated, and 1192 out of the 3830 genes showed a significant increase in methylation ($p \leq 0.05$) after HIIT (Fig. 3B). IPA revealed that 119 genes were involved in the impairment of cell movement, proliferation, and vitality after HIIT (see Additional file 9). Among the significant hypermethylated genes, the most hypermethylated (fold change: 4.4742, $$p \leq 0.044$$) gene after HIIT was the ACADVL gene (Table 2). The hypermethylation site was located at transcription start site 200 (TSS200) in the CpG island between base pairs 1 and 680 (Fig. 3C).Fig. 3HIIT-associated methylation profiling of cardiac fibroblasts treated with heart failure patient ($$n = 3$$) serum pre- and post-HIIT. A. Volcano plot showing the differential gene methylation patterns in cells treated with heart failure patient serum pre- and post-HIIT. Genes with significant changes in methylation profiles after HIIT are shown in red dots. Others are shown in blue dots. B. The methylation severity of the significantly hypermethylated genes after HIIT in each chromosome is demonstrated. Lines between the chromosome and the heatmap indicate methylation hot spots in the chromosome. C. The acyl-CoA dehydrogenase very long chain (ACADVL) gene was found to be the most involved in the HIIT-associated DNA methylation ()Table 2Significant hypermethylated genes in primary human cardiac fibroblasts after high-intensity interval training (HIIT)Entrez GeneProtein NameGene NameFCPredicted Cellular BehaviourMovementDeathProliferation37Acyl-CoA Dehydrogenase Very Long ChainACADVL4.4742↓↑↓8851Cyclin-Dependent Kinase 5 Regulatory Subunit 1CDK5R14.2352↓↑↓9693Rap Guanine Nucleotide Exchange Factor 2RAPGEF23.2325↓↑3778Potassium Calcium-Activated Channel Subfamily M Alpha 1KCNMA12.0538↓↑27ABL Proto-Oncogene 2, Non-Receptor Tyrosine KinaseABL22.5627↓↑50,509Collagen Type V Alpha 3 ChainCOL5A31.8299↓↑56,606Solute Carrier Family 2 Member 9SLC2A92.0983↓↑3594Interleukin 12 Receptor Subunit Beta 1IL12RB11.7534↓↑4683NibrinNBN1.9575↓↑1387CREB Binding ProteinCREBBP1.6711↓↑9793Cytoskeleton-Associated Protein 5CKAP51.8684↓↑168,667BMP Binding Endothelial RegulatorBMPER1.9241↓↑8409Ubiquitously Expressed Prefoldin-Like ChaperoneUXT1.8092↓↑2353Fos Proto-Oncogene, AP-1 Transcription Factor SubunitFOS1.7576↓↑7087Intercellular Adhesion Molecule 5ICAM51.7978↓↓5872Ras-Related Protein Rab-13RAB131.7113↓↓85,458DIX Domain Containing 1DIXDC12.1024↓↓1809Dihydropyrimidinase Like 3DPYSL32.6682↓↓1326Mitogen-Activated Protein Kinase Kinase Kinase 8MAP3K84.4278↑↓10,818Fibroblast Growth Factor Receptor Substrate 2FRS21.9999↑↓343,472BarH Like Homeobox 2BARHL21.9680↓5195Peroxisomal Biogenesis Factor 14PEX141.6785↓9645Microtubule-Associated Monooxygenase, Calponin And LIM Domain Containing 2MICAL21.9794↓64,225Atlastin GTPase 2ATL21.9992↓23,336SyneminSYNM2.4273↓23,031Microtubule-Associated Serine/Threonine Kinase 3MAST31.7534↓64,083Golgi Phosphoprotein 3GOLPH32.2160↓9138Rho Guanine Nucleotide Exchange Factor 1ARHGEF13.2940↓10,540Dynactin Subunit 2DCTN22.1330↓10,434Lysophospholipase 1LYPLA12.5174↓6809Syntaxin 3STX32.8263↓4750NIMA-Related Kinase 1NEK12.0534↓10,961Endoplasmic Reticulum Protein 29ERP292.4136↓79,734Potassium Channel Tetramerization Domain Containing 17KCTD174.0509↓55,186Solute Carrier Family 25 Member 36SLC25A362.2247↓50,650Rho Guanine Nucleotide Exchange Factor 3ARHGEF31.8849↓4147Matrilin 2MATN22.8825↓7205Thyroid Hormone Receptor Interactor 6TRIP61.9351↓57,701NCK-Associated Protein 5 LikeNCKAP5L1.8715↓64,223MTOR-Associated Protein, LST8 HomologueMLST81.8991↓23,647ADP Ribosylation Factor Interacting Protein 2ARFIP22.4137↓11,019Lipoic Acid SynthetaseLIAS3.2797↑3169Forkhead Box A1FOXA12.9653↑175AspartylglucosaminidaseAGA2.7531↑23,600Alpha-Methylacyl-CoA RacemaseAMACR2.2991↑9643Mortality Factor 4 Like 2MORF4L22.2689↑58,524Doublesex and Mab-3 Related Transcription Factor 3DMRT32.0777↑3183Heterogeneous Nuclear Ribonucleoprotein CHNRNPC1.9362↑8567MAP Kinase Activating Death DomainMADD1.8934↑10,195ALG3 Alpha-1,3- MannosyltransferaseALG31.8191↑9646CTR9 Homologue, Paf1/RNA Polymerase II Complex ComponentCTR91.7912↑3382Islet Cell Autoantigen 1ICA11.6084↑4047Lanosterol SynthaseLSS0.5589↑8325Frizzled Class Receptor 8FZD82.1777↓55,553SRY-Box Transcription Factor 6SOX61.6703↓FC fold change estimated by ingenuity pathway analysis↑, increased; ↓, decreased
## Correlation between HIIT-induced DNMT1 levels and cardiorespiratory fitness and cardiac fibrosis
Each onefold increase in DNMT1 in HCFs after HIIT could contribute to an approximately 2–$3.5\%$ reduction in cardiac fibrosis, a $12.7\%$ decrease in LVESV, a $5.1\%$ decrease in LVEDV, a $10.1\%$ increase in LVEF, and a $14.9\%$ increase in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak. A significant correlation ($r = 0.723$, $$p \leq 0.018$$) was found between the DNMT1 level and the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak. The correlation coefficient and probability between the above measurements are detailed in Additional file 10.
## Knockdown of the ACADVL gene induced apoptosis and actin filament disassembly
Mitochondrial fluorescence intensities in cells incubated with post-HIIT serum (Fig. 4A) and in cells with knockdown of the ACADVL gene (Fig. 4D) decreased compared to those incubated with pre-HIIT serum ($$n = 6$$). Prominent decreases in actin filaments in cells harvested with post-HIIT serum (Fig. 4B) and after knockdown of the ACADVL gene were noted (Fig. 4E). Cell death- and cell movement-related proteins were estimated from proteomic profiling (Fig. 4C). A significant decrease ($$p \leq 0.004$$) in VLCAD was identified. A nonsignificant increase in Cyto C as well as CASP3 but a decrease in lamin B1 were also observed. A significant decrease ($$p \leq 0.041$$) in β-actin and a nonsignificant decrease in Arp2 were also identified. The western blot results ($$n = 6$$) showed decreases in VLCAD, lamin B1, β-actin and Arp2, but increases in Cyto C and CASP3 were observed after knockdown of the ACADVL gene (Fig. 4F).Fig. 4Primary human cardiac fibroblast responses to high-intensity interval training (HIIT) and knockdown of the acyl-CoA dehydrogenase very long chain (ACADVL) gene ($$n = 6$$). A. The red mitochondrial fluorescence, representing the mitochondrial amount, in cells treated with patient serum after (post) HIIT (grey dot) decreased significantly compared to that in cells treated with patient serum before (pre) HIIT (white dot). B. The expression of actin (green) and actin-related protein 2 (Arp2) (red) decreased in cells incubated in post-HIIT serum. Blue fluorescence indicates the nucleus. C. Proteome profiles involved in cell death and movement are presented. Very long-chain acyl-CoA dehydrogenase (VLCAD), cytochrome C (Cyto C), caspase-3 (CASP3) and lamin B1 are involved in apoptosis. Actin B and arp2 are involved in cell movement. D. A decrease in the mitochondrial amount (red) was noted in cells after knockdown of the ACADVL gene. E. Decreased actin (green) and Arp2 (red) fluorescence were identified in cells after knockdown of the ACADVL gene. F. Decreased VLCAD, Lamin B1, Actin B and Arp2, but increased Cyto C as well as CASP3 expression after knockdown of the ACADVL gene
## Discussion
We first reported a human investigation of the HIIT-associated reduction in cardiac fibrosis (~ $10\%$) using CMR-LGE imaging and provided a proteogenomic view of the HIIT effects on human cardiac fibrosis in HF patients with a cell model built for studying exercise-induced signal transduction. In a clinical study, HIIT successfully reversed pathological cardiac remodelling in HF patients by reducing LV fibrosis to improve LV contractility by approximately $30\%$. DNMT1 overexpression in HCFs after HIIT could result in DNA hypermethylation and altered cell behaviour. Among the hypermethylated genes identified in the study, the ACADVL gene has been reported to be associated with cardiac metabolism [32]. An approximate 4.5-fold increase in methylation of the ACADVL gene at TSS200 after HIIT could impair HCF activities, which may imply an in vivo reduction in cardiac fibrosis after HIIT.
The pathological cardiac remodelling process results in systolic dysfunction, ventricular dilatation, and clinical HF syndrome [33]. HIIT-induced anti-remodelling effects have been identified and are associated with reduced LV dimensions (−4 mm to −1 mm) in HF patients [7–9]. Our study has shown a similar trend of HIIT-associated physiological adaptations. Although the exercise effects on LV mass were not significant in the previous study [34] and the present work, LV mass was still adjusted during the interpretation of relationships between cardiac fibrosis and cardiac function because cardiac fibrosis could result in LV hypertrophy [33]. The effects of cardiac fibrosis on cardiac functions became dominant with or without controlling the LV mass factor in the study. Therefore, cardiac fibrosis is thought to be a critical issue in long-term care for HF patients.
Prolonged abnormal haemodynamics and/or neurohumoural activation, peripheral vasoconstriction, and an enlarged heart result in reduced lung compliance. This phenomenon contributes to reduced aerobic capacity [35] and the development of cardiac fibrosis [1]. Extensive studies have demonstrated that cardiac fibrosis exerts adverse effects on cardiac contractility [1] and increases the risk of HF [33]. Therefore, therapeutic strategies to reduce cardiac fibrosis have become a challenge in caring for HF patients. It has been reported that regular physical training contributes to the clinical improvement in cardiovascular health by ameliorating β-adrenergic receptor responsiveness [5]. However, human studies on the correlations between HIIT and cardiac fibrosis are still limited. Many animal studies have demonstrated that reactive interstitial fibrosis is a dynamic rather than a fixed static process and is reversible after adequate treatment [10, 11, 36].
Few animal studies have mentioned the effects of endurance exercise training on cardiac fibrosis, but the results are debated [10, 11]. Inhibition of the interplay between transforming growth factor-β1 (TGF-β1) and mitochondrial-associated redox signalling, which ameliorates the dysregulation of the profibrotic gene nuclear factor erythroid 2–related factor 2 (Nrf2), could reduce pressure overload-induced cardiac fibrosis in an animal study [36]. In another animal study, swimming activated adenosine-activated protein kinase (AMPK) to attenuate cardiac fibrosis by inhibiting NADPH oxidase [11]. Consensus from observations of the above laboratory animal works has shown that reduced cardiac fibrosis closely correlates with improved cardiac geometry and cardiac contractility.
In the present human clinical study, the baseline ECV fractions of HF patients in the study (30-$35\%$) were similar to the value (37±$6\%$) with nonischaemic cardiomyopathy [37]. The HIIT-associated significant reduction in cardiac fibrosis by approximately $10\%$, especially at the middle and apical LV myocardium segments in HF patients, was similar to that in previous animal studies [10, 11]. In-depth analysis has shown that the decrease in cardiac fibrosis was closely correlated with the improvement in cardiac output during exercise. The observed anti-remodelling effects may further improve cardiac contractility to provide long-term benefits for cardiorespiratory fitness in HF patients.
The human plasma proteome encompasses proteins from all tissues, making it a medium to study the integrative biology of cardiorespiratory fitness. Indeed, the identified circulating proteins span many of the organ systems, including the nervous, musculoskeletal, pulmonary, haematologic, and circulatory systems [38, 39]. However, these valuable observations cannot specify the proteome profile of a certain cardiac cell type. Most exercise-related cardiac proteomic findings are derived from animal reports and are characterized by metabolic turnover, upregulation of antioxidant systems, induction of tissue regeneration and activation of specific kinases [40]. Proteomic profiling of myocardial tissue specimens from HF patients during LV assistive device (LVAD) implantation was reported. Reversal of cardiac remodelling after the procedure was associated with downregulation of α-1-antichymotrypsin and specific atrophic changes in protein expression profiles predominantly involved in cytoskeleton integrity and mitochondrial energy metabolism [41]. However, these previous observations are still unable to clarify cardiac fibrosis-related proteomics after LVAD implantation. HF patient serum is a good candidate to link cell biology to clinical presentations. Cardiac fibroblast behaviour and proteomic profiling in our cell model demonstrated that HCF activities decreased after HIIT, which reinforces our clinical evidence of reduced cardiac fibrosis after this exercise training strategy.
In response to pathological stresses such as myocardial infarction or pressure overload, epigenetic machinery is activated to promote cardiac fibroblast proliferation, leading to cardiac fibrosis [42]. Emerging evidence suggests that cardiac remodelling-associated lncRNAs are related to the pathophysiology after acute ischaemic events [43] or chronic HF [36]. Overexpression of DNMT1 causing the activation of animal cardiac fibroblasts to result in cardiac fibrosis has been reported [44]. However, the gene methylation level after exercise training varied across studies for different sampled tissues. Exercise-associated DNA methylation is known to be involved in skeletal muscle adaptations to physical activities [22, 45, 46]. Although global and genome-wide methylation increases following chronic exercise training [22], the epigenetic modulations causing the chronic exercise effects on cardiac fibrosis are still not well established in HF patients. Our cell model was developed to describe the effects of chronic exercise on cardiac fibrosis at the molecular level in HF patients. The present work identified increased DNMT1 levels in HCFs after chronic HIIT, which was highly informative of HIIT-associated gene regulation for cardiac fibrosis [42], and the increased DNMT1 was significantly associated with the improvement of aerobic capacity.
The ACADVL gene is located on chromosome band 17p13.1 and encodes VLCAD [47], which is distributed in the inner mitochondrial membrane and is a key enzyme for energy metabolism in mitochondria [48]. Hypermethylation of the ACADVL gene results in VLCAD deficiency in the mitochondria that provokes marked mitochondrial swelling [49]. Thereafter, cytochrome c is released from mitochondria and activates downstream caspase-mediated disassembly of actin filaments [50] as well as degradation of lamin B1 [51]. Proteomic profiling in HCFs incubated in post-HIIT serum and cells that underwent knockdown of the ACADVL gene showed that VLCAD deficiency induced apoptosis responses and actin filament disassembly (Fig. 5).Fig. 5Proposed mechanism for the high-intensity interval training (HIIT) effects on cardiac fibroblast. HIIT induces an increased DNA methyltransferase 1 (DNMT1) level in cardiac fibroblasts and results in hypermethylation () on the acyl-CoA dehydrogenase very long chain (ACADVL) gene. Silencing of the gene impairs mitochondrial function by downregulating very long-chain acyl-CoA dehydrogenase (VLCAD) expression and facilitates the release of cytochrome C (Cyto C) into the cytoplasm. This regulation activates caspase cascade-associated actin filament disassembly and possibly permeabilizes (broken arrow) the nuclear envelope by decreasing lamin B1 (LMNB) to reduce cardiac fibrosis. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\normalsize \tt V$$\end{document}VO2peak, peak oxygen consumption; Arp2, actin-related protein 2
## Conclusion
The present human investigation has shown that the HIIT-associated reduction in cardiac fibrosis contributed to the improvement in cardiac contractility based on clinical CMR-LGE imaging in HF patients. In addition, the cell model, built to reflect the exercise effects on HCFs, confirmed that HIIT was associated with hypermethylation of the ACADVL gene and that silencing of the gene impeded HCF activities. The impaired HCF behaviour after HIIT could imply in vivo HIIT effects on cardiac fibrosis (Fig. 6). Therefore, HIIT-associated epigenetic reprogramming could benefit cardiac morphology and function, further promoting cardiorespiratory fitness and providing a potential therapeutic target for HF patients. Fig.6High-intensity interval training-associated DNA methylation () may result in silencing of the ACADVL gene to impede cell movement and proliferation. The epigenetic programming could lead to reduced cardiac fibrosis. Reversal of the pathological cardiac remodelling provides benefits for the cardiac morphology as well as contractility and further promotes cardiorespiratory fitness in HF patients
## Limitations
The pandemic prohibited HF patient inclusion in the last 3 years. In addition, HF patients in the study were allowed daily water intake of 1000 to 1200 ml, and we obtained limited serum samples for experiments, which prevented further cell biology investigation to specify the epigenetic regulation in HCFs during HIIT. The lack of baseline and follow-up information for HF patients without HIIT (controls) was a major disadvantage of our present work. Both the limited samples and a lack of controls may affect our interpretation of our observations. However, promising changes in aerobic capacity as well as ECV fractions and the proteogenomic profile observed in the study provide a competing explanation for the HIIT effects on cardiac fibrosis in HF patients. The increased DNMT1 expression in HCFs in the study is one exercise-induced gene regulation effect that varies in different cardiovascular tissue cells [5, 6, 22] and is not the sole cause of the exercise-induced reduction in cardiac fibrosis. Overexpression of DNMT1 in HCFs may alter signal transduction that is not specific to exercise-induced physiological adaptations. Therefore, more translational human investigations are still required to support the above point of view.
## Supplementary Information
Additional file 1. Experimental designs. Additional file 2. Supplementary methods. Additional file 3. Baseline demographics of enrolled cardiac patients with heart failure. Additional file 4. Correlations of improved exercise performance after high-intensity interval training (HIIT) to changes in left ventricular (LV) geometry, function, and fibrosis severity. Additional file 5. Typical posterior-anterior (P-A) standing views of chest roentgenograms before and after high-intensity interval training (HIIT).Additional file 6. Cell behaviors in fetal bovine serum ($$n = 11$$), and HF patient serum ($$n = 11$$) before and after high-intensity interval training (HIIT).Additional file 7. Cell migration in pre- and post-high-intensity interval training (HIIT) serum. Additional file 8. Proteomic profiles of cardiac fibroblasts treated with serum post- and pre-high-intensity interval training (HIIT) serum. Additional file 9. *Hypermethylated* genes in cardiac fibroblasts after high-intensity interval training (HIIT).Additional file 10. The Spearman’s correlations between different measurements.
## References
1. Fan D, Takawale A, Lee J, Kassiri Z. **Cardiac fibroblasts, fibrosis and extracellular matrix remodeling in heart disease**. *Fibrogenesis Tissue Repair* (2012) **5** 1-13. DOI: 10.1186/1755-1536-5-15
2. Zhang P, Su J, Mende U. **Cross talk between cardiac myocytes and fibroblasts: from multiscale investigative approaches to mechanisms and functional consequences**. *Am J Physiol Heart Circ Physiol* (2012) **303** H1385-H1396. DOI: 10.1152/ajpheart.01167.2011
3. Mewton N, Liu CY, Croisille P, Bluemke D, Lima JA. **Assessment of myocardial fibrosis with cardiovascular magnetic resonance**. *J Am Coll Cardiol* (2011) **57** 891-903. DOI: 10.1016/j.jacc.2010.11.013
4. Gulati A, Jabbour A, Ismail TF. **Association of fibrosis with mortality and sudden cardiac death in patients with nonischemic dilated cardiomyopathy**. *JAMA* (2013) **309** 896-908. DOI: 10.1001/jama.2013.1363
5. Santulli G, Ciccarelli M, Trimarco B, Iaccarino G. **Physical activity ameliorates cardiovascular health in elderly subjects: the functional role of the β adrenergic system**. *Front Physiol* (2013) **4** 209. DOI: 10.3389/fphys.2013.00209
6. Zheng J, Cheng J, Zheng S, Zhang L, Guo X, Zhang J, Xiao X. **Physical exercise and its protective effects on diabetic cardiomyopathy: What is the evidence?**. *Front Endocrinol (Lausanne)* (2018) **9** 729. DOI: 10.3389/fendo.2018.00729
7. Fu TC, Wang CH, Lin PS, Hsu CC, Cherng WJ, Huang SC, Liu MH, Chiang CL, Wang JS. **Aerobic interval training improves oxygen uptake efficiency by enhancing cerebral and muscular hemodynamics in patients with heart failure**. *Int J Cardiol* (2013) **167** 41-50. DOI: 10.1016/j.ijcard.2011.11.086
8. Hsu CC, Fu TC, Yuan SS, Wang CH, Liu MH, Shyu YC, Cherng WJ, Wang JS. **High-intensity interval training is associated improved long-term survival in heart failure patients**. *J Clin Med* (2019) **8** 409. DOI: 10.3390/jcm8030409
9. Wisløff U, Støylen A, Loennechen JP, Bruvold M, Rognmo O, Haram PM, Tjønna AE, Helgerud J, Slørdahl SA, Lee SJ. **Superior cardiovascular effect of aerobic interval training versus moderate continuous training in heart failure patients: a randomized study**. *Circulation* (2007) **115** 3086-3094. DOI: 10.1161/CIRCULATIONAHA.106.675041
10. Novoa U, Arauna D, Moran M, Nuñez M, Zagmutt S, Saldivia S, Valdes C, Villaseñor J, Zambrano CG, Gonzalez DR. **High-intensity exercise reduces cardiac fibrosis and hypertrophy but does not restore the nitroso-redox imbalance in diabetic cardiomyopathy**. *Oxid Med Cell Longev* (2017) **2017** 7921363. DOI: 10.1155/2017/7921363
11. Ma X, Fu Y, Xiao H, Song Y, Chen R, Shen J, An X, Shen Q, Li Z, Zhang Y. **Cardiac fibrosis alleviated by exercise training is AMPK-dependent**. *PLoS ONE* (2015) **10** e0129971. DOI: 10.1371/journal.pone.0129971
12. Muzzarelli S, Ordovas K, Higgins CB. **Cardiovascular MRI for the assessment of heart failure: focus on clinical management and prognosis**. *J Magn Reson Imaging* (2011) **33** 275-286. DOI: 10.1002/jmri.22433
13. Kim RJ, Chen E-L. **Lima JoAC, Judd RM: Myocardial Gd-DTPA Kinetics Determine MRI Contrast Enhancement and Reflect the Extent and Severity of Myocardial Injury After Acute Reperfused Infarction**. *Circulation* (1996) **94** 3318-3326. DOI: 10.1161/01.CIR.94.12.3318
14. McCrohon JA, Moon JC, Prasad SK, McKenna WJ, Lorenz CH, Coats AJ, Pennell DJ. **Differentiation of heart failure related to dilated cardiomyopathy and coronary artery disease using gadolinium-enhanced cardiovascular magnetic resonance**. *Circulation* (2003) **108** 54-59. DOI: 10.1161/01.CIR.0000078641.19365.4C
15. Kellman P, Wilson JR, Xue H, Ugander M, Arai AE. **Extracellular volume fraction mapping in the myocardium, part 1: evaluation of an automated method**. *J Cardiovasc Magn Reson* (2012) **14** 63. DOI: 10.1186/1532-429X-14-63
16. Adamo L, Yu J, Rocha-Resende C, Javaheri A, Head RD, Mann DL. **Proteomic signatures of heart failure in relation to left ventricular ejection fraction**. *J Am Coll Cardiol* (2020) **76** 1982-1994. DOI: 10.1016/j.jacc.2020.08.061
17. Berezin AE, Berezin AA, Lichtenauer M. **Myokines and heart failure: challenging role in adverse cardiac remodeling, myopathy, and clinical outcomes**. *Dis Markers* (2021) **2021** 6644631. DOI: 10.1155/2021/6644631
18. Zaborska B, Sygitowicz G, Smarż K, Pilichowska-Paszkiet E, Budaj A. **Galectin-3 is related to right ventricular dysfunction in heart failure patients with reduced ejection fraction and may affect exercise capacity**. *Sci Rep* (2020) **10** 16682. DOI: 10.1038/s41598-020-73634-8
19. AbouEzzeddine OF, Haines P, Stevens S, Nativi-Nicolau J, Felker GM, Borlaug BA, Chen HH, Tracy RP, Braunwald E, Redfield MM. **Galectin-3 in heart failure with preserved ejection fraction. A RELAX trial substudy (Phosphodiesterase-5 Inhibition to Improve Clinical Status and Exercise Capacity in Diastolic Heart Failure)**. *JACC Heart Fail* (2015) **3** 245-252. DOI: 10.1016/j.jchf.2014.10.009
20. Fiuzat M, Schulte PJ, Felker M, Ahmad T, Neely M, Adams KF, Donahue MP, Kraus WE, Piña IL, Whellan DJ, O'Connor CM. **Relationship between galectin-3 levels and mineralocorticoid receptor antagonist use in heart failure: analysis from HF-ACTION**. *J Card Fail* (2014) **20** 38-44. DOI: 10.1016/j.cardfail.2013.11.011
21. Papait R, Serio S, Condorelli G. **Role of the epigenome in heart failure**. *Physiol Rev* (2020) **100** 1753-1777. DOI: 10.1152/physrev.00037.2019
22. Voisin S, Eynon N, Yan X, Bishop DJ. **Exercise training and DNA methylation in humans**. *Acta Physiol (Oxf)* (2015) **213** 39-59. DOI: 10.1111/apha.12414
23. Guo Y, Chen J, Qiu H. **Novel mechanisms of exercise-induced cardioprotective factors in myocardial infarction**. *Front Physiol* (2020) **11** 199. DOI: 10.3389/fphys.2020.00199
24. McKee PA, Castelli WP, McNamara PM, Kannel WB. **The natural history of congestive heart failure: the Framingham study**. *N Engl J Med* (1971) **285** 1441-1446. DOI: 10.1056/NEJM197112232852601
25. Mao CT, Liu MH, Hsu KH, Fu TC, Wang JS, Huang YY, Yang NI, Wang CH. **Effect of multidisciplinary disease management for hospitalized heart failure under a national health insurance programme**. *J Cardiovasc Med (Hagerstown)* (2015) **16** 616-624. DOI: 10.2459/JCM.0000000000000089
26. Pescatello LS, Arena R, Riebe D, Thompson PD. *ACSM’s guidelines for exercise testing and prescription* (2014)
27. Lang RM, Badano LP, Mor-Avi V, Afilalo J, Armstrong A, Ernande L, Flachskampf FA, Foster E, Goldstein SA, Kuznetsova T. **Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging**. *J Am Soc Echocardiogr* (2015) **28** 1-39.e14. DOI: 10.1016/j.echo.2014.10.003
28. Lebeau R, Serri K, Morice MC, Hovasse T, Unterseeh T, Piéchaud JF, Garot J. **Assessment of left ventricular ejection fraction using the wall motion score index in cardiac magnetic resonance imaging**. *Arch Cardiovasc Dis* (2012) **105** 91-98. DOI: 10.1016/j.acvd.2012.01.002
29. Cerqueira MD, Weissman NJ, Dilsizian V, Jacobs AK, Kaul S, Laskey WK, Pennell DJ, Rumberger JA, Ryan T, Verani MS. **Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart. A statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association**. *Circulation.* (2002) **105** 539-542. DOI: 10.1161/hc0402.102975
30. Lo CM, Buxton DB, Chua GC, Dembo M, Adelstein RS, Wang YL. **Nonmuscle myosin IIb is involved in the guidance of fibroblast migration**. *Mol Biol Cell* (2004) **15** 982-989. DOI: 10.1091/mbc.e03-06-0359
31. Unterberger A, Andrews SD, Weaver ICG, Szyf M. **DNA methyltransferase 1 knockdown activates a replication stress checkpoint**. *Mol Cell Biol* (2006) **26** 7575-7586. DOI: 10.1128/MCB.01887-05
32. Previs MJ, O'Leary TS, Morley MP, Palmer BM, LeWinter M, Yob JM, Pagani FD, Petucci C, Kim MS, Margulies KB. **Defects in the proteome and metabolome in human hypertrophic cardiomyopathy**. *Circ Heart Fail* (2022) **15** e009521. DOI: 10.1161/CIRCHEARTFAILURE.121.009521
33. Hill JA, Olson EN. **Cardiac plasticity**. *N Engl J Med* (2008) **358** 1370-1380. DOI: 10.1056/NEJMra072139
34. Edelmann F, Gelbrich G, Düngen HD, Fröhling S, Wachter R, Stahrenberg R, Binder L, Töpper A, Lashki DJ, Schwarz S. **Exercise training improves exercise capacity and diastolic function in patients with heart failure with preserved ejection fraction: results of the Ex-DHF (Exercise training in Diastolic Heart Failure) pilot study**. *J Am Coll Cardiol* (2011) **58** 1780-1791. DOI: 10.1016/j.jacc.2011.06.054
35. Cross TJ, Kim C-H, Johnson BD, Lalande S. **The interactions between respiratory and cardiovascular systems in systolic heart failure**. *J Appl Physiol* (2019) **128** 214-224. DOI: 10.1152/japplphysiol.00113.2019
36. Goh KY, He L, Song J, Jinno M, Rogers AJ, Sethu P, Halade GV, Rajasekaran NS, Liu X, Prabhu SD. **Mitoquinone ameliorates pressure overload-induced cardiac fibrosis and left ventricular dysfunction in mice**. *Redox Biol* (2019) **21** 101100. DOI: 10.1016/j.redox.2019.101100
37. Ugander M, Oki AJ, Hsu L-Y, Kellman P, Greiser A, Aletras AH, Sibley CT, Chen MY, Bandettini WP, Arai AE. **Extracellular volume imaging by magnetic resonance imaging provides insights into overt and sub-clinical myocardial pathology**. *Eur Heart J* (2012) **33** 1268-1278. DOI: 10.1093/eurheartj/ehr481
38. Adams V, Besler C, Fischer T, Riwanto M, Noack F, Hollriegel R, Oberbach A, Jehmlich N, Volker U, Winzer EB. **Exercise training in patients with chronic heart failure promotes restoration of high-density lipoprotein functional properties**. *Circ Res* (2013) **113** 1345-1355. DOI: 10.1161/CIRCRESAHA.113.301684
39. Robbins JM, Peterson B, Schranner D, Tahir UA, Rienmüller T, Deng S, Keyes MJ, Katz DH, Beltran PMJ, Barber JL. **Human plasma proteomic profiles indicative of cardiorespiratory fitness**. *Nat Metab* (2021) **3** 786-797. DOI: 10.1038/s42255-021-00400-z
40. Ferreira R, Moreira-Goncalves D, Azevedo AL, Duarte JA, Amado F, Vitorino R. **Unraveling the exercise-related proteome signature in heart**. *Basic Res Cardiol* (2015) **110** 454. DOI: 10.1007/s00395-014-0454-5
41. de Weger RA, Schipper MEI, Siera-de Koning E, van der Weide P, van Oosterhout MFM, Quadir R, Steenbergen-Nakken H, Lahpor JR, de Jonge N, Bovenschen N. **Proteomic profiling of the human failing heart after left ventricular assist device support**. *J Heart Lung Transplant* (2011) **30** 497-506. DOI: 10.1016/j.healun.2010.11.011
42. Felisbino MB, McKinsey TA. **Epigenetics in cardiac fibrosis: emphasis on inflammation and fibroblast activation**. *JACC Basic Transl Sci* (2018) **3** 704-715. DOI: 10.1016/j.jacbts.2018.05.003
43. Correia CCM, Rodrigues LF, de Avila Pelozin BR, Oliveira EM, Fernandes T. **Long Non-Coding RNAs in cardiovascular diseases: Potential function as biomarkers and therapeutic targets of exercise training**. *Noncoding RNA* (2021) **7** 65. DOI: 10.3390/ncrna7040065
44. Xu SS, Ding JF, Shi P, Shi KH, Tao H. **DNMT1-induced miR-152-3p suppression facilitates cardiac fibroblast activation in cardiac fibrosis**. *Cardiovasc Toxicol* (2021) **21** 984-999. DOI: 10.1007/s12012-021-09690-x
45. Brown WM. **Exercise-associated DNA methylation change in skeletal muscle and the importance of imprinted genes: a bioinformatics meta-analysis**. *Br J Sports Med* (2015) **49** 1567-1578. DOI: 10.1136/bjsports-2014-094073
46. Kanzleiter T, Jahnert M, Schulze G, Selbig J, Hallahan N, Schwenk RW, Schurmann A. **Exercise training alters DNA methylation patterns in genes related to muscle growth and differentiation in mice**. *Am J Physiol Endocrinol Metab* (2015) **308** E912-920. DOI: 10.1152/ajpendo.00289.2014
47. Shchelochkov O, Wong L-J, Shaibani A, Shinawi M. **Atypical presentation of VLCAD deficiency associated with a novel ACADVL splicing mutation**. *Muscle Nerve* (2009) **39** 374-382. DOI: 10.1002/mus.21157
48. Seminotti B, Leipnitz G, Karunanidhi A, Kochersperger C, Roginskaya VY, Basu S, Wang Y, Wipf P, Van Houten B, Mohsen AW, Vockley J. **Mitochondrial energetics is impaired in very long-chain acyl-CoA dehydrogenase deficiency and can be rescued by treatment with mitochondria-targeted electron scavengers**. *Hum Mol Genet* (2019) **28** 928-941. DOI: 10.1093/hmg/ddy403
49. Ribas GS, Vargas CR. **Evidence that oxidative disbalance and mitochondrial dysfunction are involved in the pathophysiology of fatty acid oxidation disorders**. *Cell Mol Neurobiol* (2022) **42** 521-532. DOI: 10.1007/s10571-020-00955-7
50. Desouza M, Gunning PW, Stehn JR. **The actin cytoskeleton as a sensor and mediator of apoptosis**. *BioArchitecture* (2012) **2** 75-87. DOI: 10.4161/bioa.20975
51. Freund A, Laberge R-M, Demaria M, Campisi J. **Lamin B1 loss is a senescence-associated biomarker**. *Mol Biol Cell* (2012) **23** 2066-2075. DOI: 10.1091/mbc.e11-10-0884
|
---
title: ISL1 controls pancreatic alpha cell fate and beta cell maturation
authors:
- Romana Bohuslavova
- Valeria Fabriciova
- Laura Lebrón-Mora
- Jessica Malfatti
- Ondrej Smolik
- Lukas Valihrach
- Sarka Benesova
- Daniel Zucha
- Zuzana Berkova
- Frantisek Saudek
- Sylvia M Evans
- Gabriela Pavlinkova
journal: Cell & Bioscience
year: 2023
pmcid: PMC9999528
doi: 10.1186/s13578-023-01003-9
license: CC BY 4.0
---
# ISL1 controls pancreatic alpha cell fate and beta cell maturation
## Abstract
### Background
Glucose homeostasis is dependent on functional pancreatic α and ß cells. The mechanisms underlying the generation and maturation of these endocrine cells remain unclear.
### Results
We unravel the molecular mode of action of ISL1 in controlling α cell fate and the formation of functional ß cells in the pancreas. By combining transgenic mouse models, transcriptomic and epigenomic profiling, we uncover that elimination of Isl1 results in a diabetic phenotype with a complete loss of α cells, disrupted pancreatic islet architecture, downregulation of key ß-cell regulators and maturation markers of ß cells, and an enrichment in an intermediate endocrine progenitor transcriptomic profile.
### Conclusions
Mechanistically, apart from the altered transcriptome of pancreatic endocrine cells, Isl1 elimination results in altered silencing H3K27me3 histone modifications in the promoter regions of genes that are essential for endocrine cell differentiation. Our results thus show that ISL1 transcriptionally and epigenetically controls α cell fate competence, and ß cell maturation, suggesting that ISL1 is a critical component for generating functional α and ß cells.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s13578-023-01003-9.
## Background
Understanding the mechanisms that regulate generation and maintenance of pancreatic endocrine cells is critical for developing treatments for diabetes. Pancreatic endocrine hormone-secreting α (glucagon), β (insulin), δ (somatostatin), ε (ghrelin), and PP (pancreatic polypeptide) cells form the islets of Langerhans, which are essential for regulating glucose homeostasis.
The early organogenesis of the pancreas undergoes two main transitions (reviewed in [1–3]): the primary transition between E9.5-E12.5 in the mouse, when glucagon-expressing cells are formed preferentially, and the secondary transition at ~ E13.5-E15.5, when all types of endocrine cells are produced. It is not clear whether glucagon-expressing cells generated during the primary transition persist into adulthood [4]. During the secondary transition, endocrine progenitors have a higher propensity to differentiate into α cells at earlier stages, whereas at later stages they preferentially form β cells [5]. All hormone+ islet cell types in mice originate from Neurogenin 3 (NGN3) expressing endocrine progenitors [4–6]. NGN3+ endocrine progenitors proceed to differentiate via the activation of complex gene regulatory networks through multiple intermediate cell stages (reviewed in [7]).
Recent studies have shown that endocrine differentiation into distinct islet lineages is also regulated epigenetically [5, 8–12]. Epigenetic modifiers facilitate histone and nuclear DNA modifications that induce chromatin opening, and recruitment of additional transcription factors and other regulatory proteins that subsequently activate transcriptional programs of lineage specification and differentiation. For example, pancreatic β cells deficient in DNA methyltransferase 1 (Dnmt1) are reprogrammed to α cells via DNA hypomethylation of aristaless related homeobox (Arx) enhancers [13]. Conditional deletion of Dnmt3a mediated by Ins-Cre results in β-to-α-cell transdifferentiation in adult mice [14]. Deletion of Kdm6b, a histone demethylase for histone 3 lysine 27 trimethyl (H3K27me3), in endocrine progenitors results in abnormalities in the formation of islets, reduced overall endocrine mass, and a diabetic phenotype [11]. The presence of both repressive H3K27me3 and activating histone 3 lysine 4 trimethyl (H3K4me3) epigenetic marks of histone modifications in the promoter regions of regulatory genes represent a dynamic regulation of cell fate differentiation during endocrine pancreatic development [9].
ISL1, a LIM-homeodomain transcription factor, has a crucial role in neuronal, cardiac, and sensory development [15–18]. Apart from the direct transcriptional regulation of multiple downstream targets, emerging evidence indicates that ISL1 is an important factor in the epigenetic control of embryonic development, forming regulatory complexes involved in histone modifications [19, 20], regulating KDM6B demethylase activity [21], chromatin looping [22], and acting as a pioneer factor [23]. In the pancreas, ISL1 is expressed in all developing endocrine cells [5, 24, 25], and in adult α-, β-, PP-, and δ-islet cells [26, 27]. In adult β-cells, ISL1 interacts with neurogenic differentiation factor 1 (NEUROD1) to maintain insulin gene transcription activity [28], and furthermore, ISL1 alters the status of histone H3K4 and H3K27 methylation on the insulin promoter based on glucose concentrations [29]. During pancreas development, the first ISL1 expressing cells are detected in the dorsal pancreatic epithelium at E9.0 [24]. Germline knockout of the *Isl1* gene results in a complete loss of differentiating endocrine cells without affecting the expression domain of pancreatic and duodenal homeobox 1 (PDX1) in the dorsal pancreatic epithelium at E9.5 [24]. Embryonic arrest at E9.5 and lethality of Isl1-null mice [30] preclude an in-depth investigation of mechanisms by which ISL1 contributes to pancreatic endocrine development. Delayed conditional deletion of Isl1 in Pdx1lateCre;Isl1f/f during the secondary transition results in a severe hyperglycemia phenotype due to a significant reduction in insulin+, glucagon+, and somatostatin+ endocrine cells [31]. Although the cellular phenotype has been described previously [31], we still lack mechanistic insight into the critical role of ISL1 in endocrine pancreatic development and function. In particular, no investigation has been done of pancreatic endocrine development during the primary transition in the context of ISL1 deletion, nor ISL1 function during the transition from early undifferentiated cell types to mature adult-like cell states. Thus, despite the established critical role of ISL1 in endocrine pancreatic development and function, much remains to be discovered to understand its molecular mode of action.
We created an early conditional deletion of Isl1 (Isl1CKO) that disrupts the primary transition of pancreas endocrine development. We aimed to determine the role of ISL1 in α- and β- cell differentiation and maturation using transcriptome-wide gene expression profiling. Critically, our molecular analyses of Isl1CKO endocrine cells show a shift in the transcriptomic signature towards intermediate progenitor states, loss of α-cell differentiation, and changes in molecular programs driving the formation of mature β cells. Additionally, elimination of Isl1 affected repressive H3K27me3 and activating H3K4me3 modification patterns in promoter regions of differentially expressed genes in pancreatic endocrine cells. Thus, these findings provide mechanistic insight into the role of ISL1 in transcriptional and epigenetic regulatory networks orchestrating the proper development and maturation of pancreatic islet α and β cells.
## Elimination of Isl1 results in a diabetic phenotype with impaired molecular characteristics of islets of Langerhans
To determine the functional requirements of ISL1 for the development of endocrine cells in the pancreas, we generated novel Isl1 conditional knockout mice by crossing Neurod1Cre/+ mice [32] with floxed Isl1 (Isl1f/f) [15]. Neurod1Cre activity was detected at the initiation of the first transformation of endocrine cell formation in pancreas development at E9.5 and corresponded to the expression pattern of NEUROD1 (Fig. 1a). Accordingly, ISL1 expression was apparently reduced in the dorsal pancreas of Isl1CKO compared to control littermates at E10.5 (Fig. 1b, c). Noticeably, ISL1-positive mesenchymal cells surrounded the PDX1+ pancreas, as previously described [24]. Analyses of ISL1 expression confirmed efficient Neurod1Cre-mediated Isl1 elimination, as $87\%$ of NEUROD1+ progenitors in the dorsal pancreatic epithelium of Isl1CKO did not express ISL1 compared to controls at E10.5 (Fig. 1d–f). Virtually no expression of ISL1 was detected later in endocrine cells of the developing pancreas (Additional file 1: Fig. S1). Thus, the use of the Neurod1Cre driver line led to the efficient and earlier elimination of ISL1 than observed previously in the Pdx1lateCre;Isl1f/f model with reported elimination of ISL1 protein at E13.5 [31, 33].Fig. 1Efficient deletion of ISL1 in the Isl1CKO pancreas during the primary transition. a Representative whole-mount immunolabeling of the dorsal pancreas shows that NEUROD1 matches Neurod1Cre expression visualized by anti-CRE antibody in the PDX1+ pancreatic domain (arrowheads indicate cells co-expressing CRE and NEUROD1). b, c A reduced number of ISL1+ cells shown in the Isl1CKO dorsal pancreatic epithelium delineated by the expression of PDX1 (whole mounts). d–f Higher-magnification images show significantly reduced number of cells expressing ISL1 in NEUROD1+ area in the Isl1CKO dorsal pancreas compared to littermate controls. Data are presented as mean ± SD ($$n = 5$$ pancreases per genotype), unpaired t-test (****$P \leq 0.0001$). Scale bars: 50 μm Isl1CKO mice demonstrated a severe diabetic phenotype with significant neonatal hyperglycemia that worsened with age (Fig. 2a; Additional file 1: Fig. S2), consistent with earlier findings using delayed conditional deletion of Isl1 in Pdx1lateCre;Isl1f/f [31]. None of the Isl1CKO mice survived past 3 months of age. Blood glucose concentration was higher and more variable in neonatal Isl1CKO than in heterozygous and control littermates (Fig. 2a). Accordingly, total pancreatic insulin content was significantly reduced already at P0 and more than 25 times lower in adult pancreas of Isl1CKO mice fed ad libitum (3.509 ± 0.678 ng/mg, $$n = 19$$) compared to controls (88.29 ± 6.539 ng/mg, $$n = 12$$; Fig. 2b). Both female and male Isl1CKO mice fed ad libitum showed increased blood glucose levels during postnatal development before weaning compared to controls and Isl1 heterozygous mutants (Additional file 1: Fig. S2a). Of the 40 Isl1CKO mice measured, approximately $50\%$ had unmeasurable fasting blood glucose levels (> 35 mmol/L). Glucose tolerance tests (GTTs) confirmed that both male and female heterozygous Neurod1Cre/+; Isl1f/+ mice were comparable to controls (Additional file 1: Fig. S2c), but we were unable to perform GTTs by intraperitoneal injection of glucose (2 g/kg body weight) in Isl1CKO. Although we selected Isl1CKO mice with low hyperglycemia, the mutants died after administration of exogenous glucose, indicating a severe inability to maintain an insulin secretion response when challenged with glucose. Fig. 2Physiological changes and molecular abnormalities in islets of Langerhans associated with diabetic phenotype of Isl1CKO. a Blood glucose levels of Isl1CKO ($$n = 59$$ pups), heterozygous (HET, NeuroD1Cre/Isl1f/+; $$n = 77$$ pups), and controls ($$n = 138$$ pups) fed ad libitum after birth at P0-P3. Violin plots indicate median (middle line), 25th, and 75th percentile (dotted lines). Data were analyzed by one-way ANOVA with Tukey’s multiple comparisons test (****$P \leq 0.0001$). b Total pancreatic insulin content per pancreatic tissue (ng/mg) at P0 ($$n = 6$$ pancreases per genotype), P15 ($$n = 13$$ pancreases per control, 14 pancreases per Isl1CKO), and in the adult mice ($$n = 19$$ pancreases per control, $$n = 16$$ pancreases per Isl1CKO) fed ad libitum. Data are presented as mean ± SEM, Student’s t test (*$$P \leq 0.017$$, ****$P \leq 0.0001$). c-f Representative sections from the control and Isl1CKO pancreas immunostained for glucagon (GCG), insulin (INS), PDX1 (marker of β cells), or alpha amylase (marker for exocrine cells) at P0 demonstrate reduced endocrine tissue and abnormalities in the formation of pancreatic islets in Isl1CKO with lower production of INS, missing GCG+ cells, and reduced expression of PDX1. g, h Immunolabeling for proliferating cell nuclear antigen Ki67 (red) shows proliferating GCG+ and INS+ cells in the islets of the Isl1CKO and control pancreas at P0. i Total number of INS+ and GCG+ cells in 80-μm sections of the pancreas ($$n = 5$$ pancreases per genotype), and the percentage of INS+ cells expressing Ki67 ($$n = 5$$ mice per genotype) and phosphorylated histone H3 (pHH3; $$n = 4$$ mice per genotype). See also Additional file 1: Fig. S3. Data are presented as mean ± SD, unpaired t test (****$P \leq 0.0001$, ***$P \leq 0.001$, ns = not significant). j-o Immunostaining of α- and β-cell markers (PDX1, INS, GCG, PAX6, and NKX6.1) in the pancreatic sections at P9 shows abnormalities in β cells, including reduced production of INS, variable expression levels of PDX1 with some cells expressing INS but not PDX1 (arrowheads in k indicate INS+ cells without PDX1 expression). GCG producing α cells are lost in the Isl1CKO islets of Langerhans. Arrows in m indicate the unusual GCG+ cells with a missing expression of PAX6, a marker of both β and α cells. p, q At P35, compare to the control pancreatic islets, the expression of INS and PDX1 is reduced, and GCG+ producing cells are missing in the Isl1CKO islets. r, s Representative confocal microscopy images of double staining with anti-PDX1 and anti-ISL1 antibodies show that the expression of ISL1 is not detected in nuclei of endocrine cells of pancreatic islets of Isl1CKO in contrast to the control. Arrowheads indicate nuclei with ISL1 and PDX1 co-expression in the control pancreatic islet. Note reduced PDX1 expression in Isl1CKO. Scale bars: 50 μm. HS, Hoechst nuclear staining Next, we investigated the formation and structure of islets of Langerhans using anti-insulin and glucagon as markers for β and α cells, respectively. At P0, the islets of Langerhans of Isl1CKOs contained only β cells expressing insulin and PDX1, a marker of differentiated β cells (Fig. 2c–f). Interestingly, immunostaining for proliferating cell nuclear antigen Ki67 revealed an increased number of β cells positive for Ki67 in the Isl1CKO pancreas at P0 (Fig. 2g–i). As Ki67 marks all cells engaged in the cell cycle, we also used the marker of cellular mitosis, phosphorylated histone H3 (pHH3) [34]. Quantification of the percentage of β (insulin+) cells that were pHH3 positive indicated similar mitotic activity in the Isl1CKO and control pancreases (Fig. 2i, Additional file 1: Fig. S3). A high Ki67 index may reflect a lengthened cell cycle or cell-cycle arrest of Isl1CKO β cells.
At P9, when a mature functional glucose-stimulated-insulin-secretion phenotype of β-cells is acquired [35], instead of β cells co-expressing insulin and PDX1 as shown in controls (Fig. 2j), expression of PDX1 was undetectable in many insulin+ cells in Isl1CKO (arrowheads in Fig. 2k). In contrast to PDX1, the expression of paired box 6 (PAX6), which is expressed in both glucagon and insulin-expressing cells in controls, seemed unaffected in Isl1CKO β cells (Fig. 2l, m). Note the unusual glucagon+ cells without the expression of PAX6 (arrows in Fig. 2m), indicating abnormalities in the differentiation of these cells. Immunolabeling for the β-cell programing factor NK 6 homeobox1 (NKX6.1) demonstrated that many NKX6.1+ cells did not express insulin compared to controls (Fig. 2n, o). Profound diminished PDX1 levels were found in adult Isl1CKO pancreas (P35), correlating with a reduced number of insulin-producing cells in the islets of Langerhans and severe diabetic phenotype of Isl1CKO mice (Fig. 2p, q). Diminished PDX1 expression reflected essentially complete elimination of ISL1 in Isl1CKO pancreatic islets (Fig. 2r, s).
We applied light sheet fluorescence microscopy (LSFM) to uncover the formation, and spatial distribution of the islets of Langerhans in the 3D tissue microenvironment of the pancreas (Fig. 3a–d, Additional files 2, 3, 4, 5: Videos S1-S4). Endocrine cells were visualized by combinations of genetic labeling using tdTomato reporter expression and immunolabeling of α cells with anti-GLP1 (glucagon like peptide 1) and β-cells with anti-insulin (Fig. 3a, b). In the second preparation, tdTomato+ endocrine cells were co-labeled with anti-GLP1 (α cells) and anti-TUBB3 depicted innervation in the pancreas (Fig. 3c, d). Endocrine cells were scattered, forming smaller cell clumps without α cells at the periphery in the Isl1CKO pancreas compared to the characteristic islet structure of the control pancreas. In line with abnormalities in islet formation, the total number of isolated islets from the adult Isl1CKO pancreas was severely reduced compared to controls (11 ± 10 islets/Isl1CKO pancreas, $$n = 8$$ vs. 273 ± 77 islets/control pancreas, $$n = 6$$, $P \leq 0.0001$). Taken together, these changes indicated a loss of α cells, abnormalities in PDX1 expression in β cells, production of insulin, and formation of the islets of Langerhans in Isl1CKO.Fig. 3Distribution and formation of islets of Langerhans in the microenvironment of the pancreas. Microdissected pancreases of tdTomato reporter control-Ai14 and Isl1CKO-Ai14 mice were cleared (CUBIC protocol), immunolabeled, imaged, and reconstructed in 3D using light-sheet fluorescence microscopy (LFSM; see Additional files 2, 3, 4, 5: Videos S1–S4). LFSM images depict the distribution and formation of islets in the anatomical microenvironment of the pancreas at P9, showing tdTomato+ endocrine cell population and INS producing β cells together with anti-GLP1 labeled α cells in (a, b) or with neuronal fibers labeled by anti-TUBB3 in (c, d). Scale bars: 100 μm
## Elimination of Isl1 alters the α-cell differentiation program during the primary transition of pancreas development
The primary transition of mouse pancreas development is initiated by expression of the transcription factor PDX1 that specifies multipotent pancreatic progenitors, and evagination of a dorsal pancreatic bud from the foregut endoderm around E9.0 [1] (Fig. 4a). The primary transition of endocrine cell formation mainly generates glucagon-expressing α cells. To investigate changes associated with Isl1 deletion during the primary transition, we evaluated the presence of glucagon-expressing cells in the dorsal pancreatic bud, as their appearance represents the first sign of endocrine cell differentiation [4]. Glucagon-producing cells at the periphery of the dorsal pancreatic bud of Isl1CKO did not express ISL1 (Fig. 4b, c). Overall, the generation of glucagon+ cells and formation of glucagon+ clusters during the first transition was significantly reduced by $50\%$ in the dorsal pancreas of Isl1CKO compared to littermate controls (Fig. 4d–h). Additionally, a transient population of developing endocrine cells co-producing glucagon and insulin was diminished in Isl1CKO, indicating changes in the composition of glucagon-producing subpopulations (Fig. 4i, j). Defects in developmental programs during early pancreatic organogenesis were further confirmed by qPCR at E12.5 (Fig. 4k). Expression of key genes encoding differentiation regulators of the α-cell lineage, such as Arx1, MafB, Peg10, and Pou3f4, were significantly reduced in developing Isl1CKO pancreas at E12.5. Interestingly, expression of the E26 transformation-specific transcription factor, Fev, was increased. A pancreas lineage study suggested that Fev+ cells represent an intermediate endocrine progenitor state following Ngn3 expression [36]. Consistent with the Fev+ endocrine progenitor expression profile [36, 37], we found no changes in Ngn3 levels in the developing pancreas of Isl1CKO (Fig. 4k). These results suggested that elimination of Isl1 abolished α-cell lineage development during the primary transition. Fig. 4Aberrant α-cell lineage development during primary transition in the Isl1CKO pancreas. a Schematic presentation of the pancreas formation, which begins with the independent budding of the dorsal and ventral buds at the posterior region of the foregut. These two buds, surrounded by the mesenchyme, eventually fuse after rotation of the gut to form the pancreatic endoderm. b, c Representative higher-magnification images of whole-mount immunolabeling show a loss of ISL1 and glucagon (GCG) expressing cells in the dorsal pancreas of Isl1CKO at E10.5. d-g Whole-mount immunolabeling shows the formation of GCG+ clusters in the dorsal (DP) and ventral pancreatic buds (VP) at E11.5. Higher-magnification images show DP with GCG+ clusters. h Quantification of the GCG+ area in the PDX1+ domain shows the reduced size of GCG+ clusters in the dorsal pancreas of Isl1CKO ($$n = 8$$ pancreases) compared to controls ($$n = 9$$ pancreases). Data are presented as mean ± SD, Unpaired t-test, **$$P \leq 0.0069$$). i, j Representative images of whole-mount immunolabeling of the pancreas show endocrine cells co-expressing GCG and insulin (INS) in the control (arrowheads) but not in the Isl1CKO pancreas (asterisks indicate autofluorescent red blood cells). k Quantitative RT-PCR analyses of mRNA levels of Gcg and selected transcription factors in the E12.5 pancreas. Data are presented as mean ± SEM ($$n = 8$$ pancreases per genotype), Unpaired t-test (****$P \leq 0.0001$, ***$P \leq 0.001$, **$P \leq 0.01$, ns = not significant). Scale bars: 50 μm
## The α-cell population disappears during the secondary transition and ß-cell proliferation is reduced in Isl1CKO pancreas development
The trend of increased expression of Fev continued in the developing Isl1CKO pancreas at E14.5, during the secondary transition (Fig. 5a). The secondary transition represents morphogenetic events resulting in a branched structure of the epithelium containing endocrine progenitors that differentiate into different endocrine cells [3]. The early elimination of Isl1 in Isl1CKO resulted in a faster onset of changes in endocrine development than observed previously after Isl1 deletion using a delayed Pdx1lateCre;Isl1f/f model [31]. We confirmed there was a significant reduction of insulin, somatostatin, and pancreatic polypeptide mRNA in the developing Isl1CKO pancreas as early as E14.5 (Fig. 5a). We found essentially no glucagon mRNA expression in the E14.5 pancreas of Isl1CKO (Fig. 5a), indicating that the use of the Neurod1Cre driver line led to an earlier elimination of glucagon production than observed previously [31]. In line with changes in mRNA expression, a near complete loss of glucagon+ cells was found in Isl1CKO pancreas at E15.5 (Fig. 5b, c). ß-cell proliferation was already significantly diminished in Isl1CKO compared to littermate controls at E17.5 (Fig. 5d–i), in contrast to the previously reported reduced proliferation of ß cells in Pdx1lateCre;Isl1f/f at P6 [31]. The percentage of ß cells expressing the proliferation marker Ki67 was decreased nearly threefold in Isl1CKO pancreas compared to control littermates at E17.5. Similarly, ß cells in the Isl1CKO pancreas proliferated less as shown by pHH3 staining, the marker of cellular mitosis. These data indicate that ISL1 is important for the induction of the expansion of the endocrine ß-cell population during pancreas development. Notably, expression of PDX1 was comparable between control and Isl1CKO endocrine cells at E17.5.Fig. 5Attenuated expression of insulin and a loss of glucagon producing cells during the secondary transition of pancreas development in Isl1CKO. a Quantitative RT-PCR analyses show reduced mRNA levels of endocrine hormones and increased expression of the transcription factor Fev in the E14.5 pancreas of Isl1CKO compared to controls. Data are presented as mean ± SEM ($$n = 8$$ pancreases per control, $$n = 7$$ pancreases per Isl1CKO), Unpaired t-test (****$P \leq 0.0001$, ***$P \leq 0.001$, **$P \leq 0.01$). b-g Representative sections from the control and Isl1CKO pancreas immunostained for glucagon (GCG), insulin (INS), PDX1 (differentiation marker of β cells), or alpha amylase (marker for exocrine cells) demonstrate loss of endocrine α cells (glucagon, GCG) in Isl1CKO. h, i Representative sections immunostained for proliferating cell nuclear antigen Ki67 in endocrine α (GCG) and β cells (INS). j Relative quantification of GCG+ and INS+ cells per α-amylase+ area (marker of exocrine tissue; $$n = 5$$ pancreases per genotype), and k the percentage of INS+ cells expressing Ki67 ($$n = 5$$ pancreases per genotype) and phosphorylated histone H3 (pHH3; $$n = 3$$ pancreases per genotype) per total number of INS+ cells is decreased in the Isl1CKO pancreas compared to littermate controls at E17.5. See also Additional file 1: Fig. S3. Data are presented as mean ± SD. Unpaired t-test (**$P \leq 0.01$, *$P \leq 0.05$, ns = not significant). Scale bars: 50 μm
## Elimination of Isl1 results in a shift of transcriptomic signatures characterizing endocrine cell populations in the E14.5 pancreas
To gain insight into molecular mechanisms underlying ISL1 function, we performed RNA-seq analyses of pancreatic endocrine cells during the secondary transition at E14.5, as E14.5 endocrine progenitors have a higher propensity to form α cells [5]. We opted to use Bulk-RNA sequencing to obtain sequencing depth and high-quality data [38]. Pancreases were dissected, dissociated into single cells, and 100 fluorescent tdTomato+ cells were isolated via fluorescence-activated cell sorting (FACS) per each biological replicate. We used a Rosa26-tdTomato reporter mouse line to genetically label Neurod1Cre, expressing endocrine cells in the Isl1CKO (genotype: Isl1f/f; Neurod1Cre; TomatoAi14; $$n = 5$$) and control pancreas (genotype: Isl1f/+; Neurod1Cre; TomatoAi14; $$n = 6$$) (experimental design in Fig. 6a). Compared to controls, 292 protein-coding genes were differentially expressed in Isl1CKO pancreatic endocrine cells (adjusted P-value, Padj < 0.05, fold change > 1.5 and < 0.5 cut-off values; Fig. 6b and Additional file 6: Dataset S1a). Functional profiling revealed that the top gene clusters for down-regulated genes were related to highly enriched biological pathways and specific gene ontology (GO) term categories associated with synthesis, secretion, and inactivation of the incretion, GLP1, glucose-dependent insulinotropic polypeptide (GIP), glucagon receptor binding, and peptide hormone secretion (Additional file 1: Fig. S4a). We found a significant reduction of hallmark genes associated with the α-cell lineage in Isl1CKO endocrine cells, including Peg10, Pou6f2, and Gcg [5, 12, 38].Fig. 6Isl1 elimination induces transcriptomic and epigenetic changes altering the composition of the endocrine cell population in the developing pancreas. a Overview of study design. b Volcano plot showing differentially expressed genes between FACS-sorted pancreatic endocrine cells from Isl1CKO ($$n = 5$$) and control E14.5 embryos ($$n = 6$$) identified by RNA sequencing (Padj < 0.05 displayed as –log10 and log2 fold change − 1 and 0.585). Thresholds are indicated by dotted lines. Complete list of differentially expressed genes in Supplementary Dataset 1a. c A simplified schematic overview of the different pancreatic cell types represented in single cell RNA-seq [39] used as a reference for the cell type deconvolution of our bulk cell RNA-seq data. d The deconvolved cell type proportions in E14.5 endocrine population from our bulk RNA-seq data. The proportion of major cell types is shown as an average and per individual samples. Data are presented as mean ± SD, Unpaired t-test (****$P \leq 0.0001$, ***$$P \leq 0.0007$$). See also Additional file 1: Fig. S4. e Heatmaps represent the top 30 enriched genes in endocrine progenitors, and in α and β cells in Isl1CKO and control samples based on the transcriptome analyses (*indicates genes with ISL1 binding sites at their promoter regions identified by the ISL1 CUT&Tag-seq, see page 13). Complete list of genes in Supplementary Dataset 1b and 1c, and S2. f Heatmap representation of expression profiles of differentially expressed genes that have differences in H3K4me3 or H3K27me3 enrichment in their promoter regions (± 3 kb from the TSS) between control and Isl1CKO cells. Scale bar for RNA-seq data represents average values calculated from the ratio of rlog of the sample to average rlog of the row (all samples per gene). Scale bars for CUT&Taq-seq data represent average counts per million (cpm; > 25 cpm) per each group. g Genome track view of representative gene loci showing H3K4me3 (green), H3K27me3 (magenta), and negative antibody control IgG (grey) normalized read peaks. EPs, endocrine progenitors Using single-cell transcriptomic profiling data of cells in embryonic mouse pancreas [39] as a reference for cell type deconvolution, we estimated the proportion of cell types in our bulk RNA-seq samples [40]. Single cell transcriptomic signatures of five major cell types in the developing pancreas, e.g. endocrine progenitors, epithelial tip, trunk epithelium, ß and α cells, were used for the deconvolution of our bulk expression data (Fig. 6c, d, and Additional file 1: Fig. S4b, c). A significantly higher proportion of endocrine progenitors marked by high Fev expression was found in Isl1CKOs with enriched genes defining intermediate states of endocrine differentiation, such as Fev, Chgb, Chga, Vim, and Cldn4 (Fig. 6d, e; Additional file 6: Dataset S1b) [5, 36, 39]. Concurrent with a deficiency in the generation of α cells in Isl1CKOs, the proportion of α cells was substantially decreased (Fig. 6d) together with a reduced expression of genes associated with the α-cell lineage, including Gcg, Peg10, Pou6f2, Meis2, Ripply3, and Zcchc18 (Fig. 6e; Additional file 6: Dataset S1b) [5, 12, 38]. Interestingly, no α cells of a late differentiation stage (cluster 8) were found in Isl1CKO compared to control cells (Additional file 1: Fig. S4c), indicating differentiation arrest in the α-cell lineage. Motif enrichment transcription factor target analysis [41] showed that ISL1 and FEV were the top ranking transcription factors associated with differentially expressed genes in Isl1CKO together with transcription factors essential for endocrine specification, regulatory factor X6 (RFX6) [42], and paternally-expressed gene 3, PEG3 (also known as PW1) [43], suggesting a shift in the transcriptomic signature of Isl1CKO endocrine cells towards earlier progenitor states (Additional file 6: Dataset S1c).
## H3K4me3 and H3K27me3 patterns are changed at promoter regions in Isl1CKO
Since our data demonstrated that the deletion of Isl1 resulted in a loss of α cells and enrichment of endocrine progenitor populations, we then asked whether Isl1 elimination affected the epigenetic basis of endocrine cell identity in the pancreas. We chose to investigate histone H3 modifications, H3K4me3, a marker of active transcription and H3K27me3, a marker of transcriptional silencing, because of their roles in the regulation of transcription and cell-fate determination [9, 44]. Bivalently H3K4me3 and H3K27me3 marked promoters are associated with transcriptionally inactive genes or genes expressed at very low levels [45]. The presence of both marks keeps genes at a poised state enabling them to be rapidly activated, particularly during embryogenesis and differentiation [44]. We performed chromatin profiling of H3K4me3 and H3K27me3 modifications on FACS-sorted tdTomato+ cells from the E14.5 pancreas Cleavage Under Targets and Tagmentation sequencing (CUT&Tag-seq) [46] (experimental design in Fig. 6a). Genome-wide H3K4me3 or H3K27me3 marks were distributed similarly in control and Isl1CKO endocrine cells with the majority of H3K4me3 deposition at gene promoter regions, while H3K27me3 peaks were distributed along the promoter, gene body and intergenic regions (Additional file 1: Fig. S4d, e). To correlate H3K4me3 and H3K27me3 modifications with gene expression changes, we focused only on histone modifications at the promoter regions of differentially expressed genes identified in our RNA-seq (Padj < 0.05 and Fold change $50\%$). We compared the normalized read counts of H3K4me3 or H3K27me3 to identify a variable H3 methylation pattern at promoter regions between control and Isl1CKO cells (Fig. 6f, Additional file 1: Fig. S4f). In Isl1CKO endocrine cells, downregulated regulatory genes, such as Mafa, Ripply3, Sorcs1, and Mnx1, were marked by a suppressive H3K27me3 methylation signature in contrast to control cells (Fig. 6g). These downregulated genes were associated with the α-and β-cell lineages (Fig. 6e). Our data confirmed that histone methylation patterns were affected in the absence of Isl1, contributing to abnormalities in the regulation of endocrine differentiation in the embryonic pancreas.
## Elimination of Isl1 results in downregulation of maturation markers of β cells in the Isl1CKO pancreas at P9
Next, we analyzed the transcriptome of pancreatic endocrine cells at P9, when a mature functional glucose-stimulated-insulin-secretion phenotype of β cells is acquired [35]. We identified 1049 protein-coding genes differentially expressed in endocrine cells, comparing Isl1CKO to controls (Fig. 7a; Additional file 6: Dataset S1d). GO analysis of the differentially downregulated genes showed a sustained downregulation of gene classes involved in insulin secretion, peptide secretion, and transport, including reduced levels of all major hormone transcripts Gcg, Ins1, Ins2, Ppy, and Sst (Fig. 7b, c; Additional file 6: Dataset S1e). The deficiency in endocrine hormone production corresponds with observed abnormalities in the islets of Langerhans and severe postnatal diabetic phenotype of Isl1CKO (Fig. 2). Although differentiated insulin-producing ß cells were found in Isl1CKO, we next wanted to elucidate the maturation state of these cells. Some of the transcription factors that have been reported to regulate mature ß-cell function, such as Mafa, Pparg, Bcl6, Pdx1 [38], together with mature ß-cell markers Ucn3, Slc2a2, Trpm5, and G6pc2 [35, 47] were downregulated in Isl1CKO (Fig. 7c, Additional file 6: Dataset S1d), suggesting abnormalities in the transition between immature and mature ß cells. PDX1 is critical for inducing and maintaining ß-cell maturation and identity by regulating target genes, such as Slc2a2, Mafa, and Ins1 [48, 49]. The immature state of the ß-cell population in Isl1CKO was further indicated by the increase of Wif1, Dkk3, Tgfbr3, and Tab3 associated with Wnt and TGF-ß signaling pathways that are normally downregulated in mature ß cells [38]. Furthermore, the most enriched GO categories for the upregulated transcripts in Isl1CKO endocrine cells were involved in biological processes related to development (Fig. 7b, Additional file 6: Dataset S1e). Consistent with an immature transcriptomic signature, Isl1CKO endocrine cells displayed an enrichment for genes defining intermediate endocrine progenitors, such as Fev, Tle1, Evpl, Sez6l, Gc, and Vim (Fig. 7d, Additional file 6: Dataset S1f), suggesting abnormalities in the progression of endocrine cell differentiation. Fig. 7ISL1 modulates expression profiles of endocrine cells in the P9 pancreas. a Volcano plot shows differentially expressed genes (Padj < 0.05 (displayed as –log10) and log2 fold change − 1 and 0.585) identified by RNA sequencing in 100 FACS-sorted endocrine cells from the P9 pancreas of Isl1CKO ($$n = 6$$) in comparison to the control ($$n = 5$$). Complete list of genes in Supplementary Dataset 1d. b The most enriched Gene Ontology (GO) biological processes for downregulated and upregulated genes identified by RNA-seq. Complete list of genes for the GO terms in Supplementary Dataset 1e. c Heatmap of insulin secretion genes from the GO analysis shows expression levels for each sample. d Heatmap shows the top upregulated genes in P9 endocrine cells of Isl1CKO that were found enriched in E14.5 endocrine progenitors of Isl1CKO based on deconvolution analyses (Supplementary Dataset 1f). e Venn diagram representing the overlap of genes with differential expression (RNA-seq) and genes with H3K4me3 and/or H3K27me3 marks (CUT&Tag-seq) in their promoter regions (± 3 kb from the TSS) in control and Isl1CKO endocrine cells at P9. f Overlap between H3K4me3 and H3K27me3 modifications in the promoter regions of differentially expressed genes in control and Isl1CKO endocrine cells. g Heatmap shows the selection of differentially expressed genes identified by RNA-seq (average expression level per control and Isl1CKO) correlating to differences in H3K4me3 and H3K27me3 modification patterns at their promoter regions (± 3 kb from the TSS) between control and Isl1CKO pancreatic endocrine cells at P9. h Genome track view of representative gene loci showing H3K4me3 (green), H3K27me3 (magenta), and negative antibody control IgG (grey) normalized read peaks based on CUT&Tag-seq data. The presence of both marks at the promoter region, representing a poised state, is shown in control cells for Fev, Chga, Chgb, and *Abat* genes associated with endocrine progenitors. In contrast, a poised state is shown at promoter regions of genes Pyy (α cells); Celsr1, and Glp1r (markers of ß cells) in Isl1CKO but not control cells. A silencing H3K27me3 mark is located at the promoter region of a key ß-cell maturation marker Ucn3 only in Isl1CKO cells To determine whether the absence of Isl1 had an impact on the chromatin architecture of postnatal endocrine cells, we performed CUT&Tag-seq of H3K4me3 and H3K27me3 marks in FACS-sorted tdTomato+ cells from the P9 pancreas (Additional file 1: Fig. S5). To correlate changes in gene expression with changes in H3K4me3 and H3K27me3 patterns in Isl1CKO endocrine cells, we focused on H3 methylation states in the promoter regions of differentially expressed genes at P9 (Fig. 7e). $62\%$ of differentially expressed genes were found to have either or both H3K4me3 and H3K27me3 marks at their promoter regions, and $64\%$ of those genes acquired distinct H3 modification patterns between control and Isl1CKO (Fig. 7e, f). Next, we correlated distinct H3K4me3 and H3K27me3 modifications at the promoter regions with changes in expression of signature genes associated with endocrine progenitors, α and ß cells (Fig. 7g). We identified transcription activator H3K4me3 marks at the promoter regions of endocrine progenitor genes, which had upregulated expression in Isl1CKO cells in contrast to these genes in endocrine cells from control mice which exhibited a suppressive histone methylation pattern associated with either loss of H3K4me3 or acquisition of H3K27me3 marks. Notably, the promoter of the signature gene Fev, characterizing the endocrine progenitor lineage [5, 36, 37], was transcriptionally poised in endocrine cells from the control pancreas by acquiring a suppressive methylation mark H3K27me3 (Fig. 7g, h). Consistently, a silencing methylation signature of H3K27me3 was acquired by control endocrine cells at promoter regions of Chgb and Chga. Although Chga and Chgb are often utilized as markers of differentiated endocrine lineages, these genes were highly expressed in Isl1CKO endocrine cells compared to controls, presumably reflecting an immature endocrine lineage expression profile [36]. In correlation with epigenetic repression by H3K27me3, these genes exhibited diminished expression in control cells compared to Isl1CKO. Thus, these observations suggested that control cells are becoming terminally differentiated compared to those of Isl1CKO. In Isl1CKOs, we found a bivalent state at promoter regions of several downregulated ß- and α-cell genes, including Celsr1, Glp1r, and Pyy (Fig. 7g, h). Interestingly, a silencing methylation signature of H3K27me3 was uniquely acquired at the promoter region of a key ß-cell maturation marker Ucn3 in Isl1CKO cells, indicating the contribution of epigenetic repression to the regulation of ß-cell maturation.
## ISL1 directly targets regulatory elements of critical genes involved in endocrine development
To understand the molecular modes of action of ISL1 during endocrine development, we then performed ISL1 CUT&Tag-seq to map the genome-wide binding of ISL1 in FACS-sorted tdTomato+ cells of the E14.5 pancreas (Additional file 7: Dataset S2). In peak calling, 31,445 ISL1-occupied loci were identified. ISL1 binding sites were found at both promoter and non-promoter regions, with approximately $29\%$ of ISL1-loci were annotated within 3 kb from the transcription start site (TSS) of a gene, $36\%$ of ISL1-loci detected in introns and exons, and $35\%$ in distal intergenic regions (Fig. 8a, Additional file 1: Fig. S6a). Comparing the distribution of H3K27me3 loci and ISL1 loci at gene promoters between control and Isl1CKO mutant, we found that approximately $6.5\%$ (2031 loci) and $5\%$ (1569 loci) of ISL1-binding sites were associated with differential H3K27me3 depositions at promoter regions between control and Isl1CKO, respectively (Fig. 8b). Identified ISL1-loci were annotated to 13,577 genes. *Many* genes with ISL1 bound at their promoter regions represent critical regulators in development, such as transcription factors, signaling molecules, epigenetic modifiers, and members of the SWI/SNF chromatin remodeling complex. These ISL1 targets included key regulators in pancreatic endocrine development, such as Pax6, Mafb, Nkx6.2, Insm1, Sox9, Arx, Fev, Nkx6.1, Foxa2, Neurod1, Rfx3, and Rfx6 [1]. Among the most enriched transcription factor motifs at the sites occupied by ISL1 were consensus binding sites for the E2F family, KLF14, and NKX6.1 (Fig. 8c). For example, NKX6.1 is essential for both early and late stages of pancreatic development with a critical role in the formation of β cells [50–52], KLF14 is an important regulator of metabolic diseases, such as diabetes and obesity[53, 54], and E2F3 activates β-cell proliferation [55]. High percentage of ISL1-bound regions were enriched for transcription factor PDX1 motif, β-cell-fate determining transcription factor [49], and NKX2.2 and RFX6 essential for endocrine cell development and β-cell-fate [56–58]. To investigate ISL1 activity further, we compared genes with differential histone H3K27me3 patterns between control and Isl1CKO, genes with ISL1 binding sites, and differentially expressed genes identified from RNA-seq (Fig. 8d). $73\%$ genes deregulated in Isl1CKO endocrine cells were bound by ISL1. $53\%$ downregulated genes were associated with Isl1CKO-distinctive H3K27me3 marks and were bound by ISL1, indicating an intriguing association between Isl1 deficiency and H3K27me3 repressive epigenetic state. The most enriched GO Biological processes for ISL1-bound genes with Isl1CKO-distinctive H3K27me3 depositions were related to cell fate determination, embryonal development, and morphogenesis (Additional file 1: Fig. S6b). In contrast, only $13\%$ of ISL1-bound upregulated genes (14 genes) exhibited distinctive H3K27me3 modifications. NKX6.1, a homeobox-containing transcription factor, was found to be one of the most enriched motifs with a high percentage of ISL1 peaks and H3K27me3 depositions (Fig. 8e). NKX6.1, the ISL1 target gene, plays a crucial role in regulating the chronological development of different endocrine cell types [50], and initiates and maintains β cell-specific gene expression programs while repressing programs of alternative endocrine lineages [52]. These data indicate that ISL1 and NKX6.1 may interact during pancreas endocrine cell development. Isl1CKO-distinctive H3K27me3 signatures were found in 48 downregulated ISL1 target genes, which are key regulators of endocrine development, including Mafa, Ripply3, Sorcs1, Phox2b, Eya1, and Mnx1 (Fig. 8f). In contrast, distinctive silencing H3K27me3 marks observed in control endocrine cell samples were detected in 13 upregulated ISL1 targets, including Vim and an inhibitor of the PBX1 homeodomain transcription factor, Pbxip1 [59] (Fig. 8f). It is worth noting that some endocrine progenitor signature genes that were upregulated in ISL1CKO contained ISL1 binding sites at their promoter regions, among them Fev, Apoe, Chgb, Chga, B2m, Gprc5a, Wif1, and Ttr (genes are marked by asterisks in Fig. 6e; Additional file 7: Dataset S2). These data indicate that Isl1 deficiency was associated with the remodeling of epigenetic H3K27me3 marks and altered gene expression levels of ISL1 target genes, contributing to the dysregulation of endocrine development in Isl1CKO mice. Fig. 8ISL1 binds regulatory elements of critical developmental genes and reprograms the H3K27me3 landscape in endocrine cells. a Pie chart showing genomic distribution of ISL1 loci. b Venn diagram indicating overlap of ISL1-loci, and of distinctive H3K27me3 loci from pairwise comparisons control and Isl1CKO endocrine cells at E14.5 in promoter regions (Additional file 7: Dataset S2). c Sequence logos of the significantly enriched motifs against ISL1 peaks (2 kb areas) from Homer FindMotifsGenome analysis. Percent of target sites in ISL1 peaks is indicated. d Venn diagrams illustrating intersection of ISL1-bound genes, genes with differential H3K27me3 depositions from pairwise comparisons of control and Isl1CKO endocrine cells at E14.5, and differentially expressed genes at E14.5 identified from RNA-seq data. e Sequence logos of the significantly enriched motifs against ISL1 peaks containing H3K27me3 peaks (2 kb areas) from Homer FindMotifsGenome analysis. f Genome track view of representative gene loci showing ISL1 and H3K27me3 normalized read peaks based on CUT&Tag-seq data from E14.5 endocrine cells (an arrow indicates a TSS). The presence of both marks at the promoter region indicates overlapping occupancy of silencing methylation signature of H3K27me3 and ISL1. g Venn diagrams indicating overlap of ISL1-bound genes and genes with distinctive H3K27me3 depositions from pairwise comparisons P9 control and Isl1CKO endocrine cells, and differentially expressed genes identified from RNA-seq data at P9 (Additional file 8: Dataset S3). h and i Genome track view of representative gene loci showing ISL1 and H3K27me3 normalized read peaks based on CUT&Tag-seq data. The presence of both marks at the promoter region indicates overlapping occupancy of silencing methylation signature of H3K27me3 and ISL1 in upregulated (h) and downregulated (i) genes. Samples used for comparison: ISL1-CUT&Tag-seq E14.5 endocrine cells, and H3K27me3-CUT&Tag-seq P9 endocrine cells Next, we wanted to investigate the extent of shared ISL1-bound genes between embryonic E14.5 endocrine cells and functionally matured P9 endocrine cells. To accomplish this, we compared the set of ISL1-bound genes with genes that displayed differential H3K27me3 patterns between P9 control and Isl1CKO, as well as differentially expressed genes identified from RNA-seq data in P9 endocrine cells (Fig. 8g, Additional file 8: Dataset S3). The intersection of the RNAseq data with CUT&*Tag data* revealed that $73\%$ (480 genes) of upregulated and $51\%$ (196 genes) of downregulated genes were bound by ISL1, respectively. *Downregulated* genes with ISL1 bound at their promoter regions included genes associated with insulin secretion and β-cell function (Fig. 7b), such as Abcc8, Acvr1c, Cdk16, Kcnj11, Nnat, and Mafa [60, 61]. ISL1-bound promoters of upregulated genes associated with endocrine progenitor state included Fev, Chga, Chgb, Cldn4, Pbxip1, Gpc4, Tle1, and Wif1 (Additional file 8: Dataset S3). $25\%$ (99 genes) of downregulated genes had ISL1 binding sites and contained Isl1CKO-specific-H3K27me3 silencing marks, while $7\%$ (45 genes) upregulated genes had ISL1 binding sites and contained control-specific-H3K27me3 silencing marks (Fig. 8g–i). This analysis revealed a significant enrichment of H3K27me3 silencing marks in ISL1-bound genes that were downregulated in P9 Isl1CKO endocrine cells, highlighting the critical role of ISL1 in regulating functional glucose-stimulated-insulin-secretion phenotype of β cells and the diabetic phenotype of Isl1CKO.
## Discussion
Cell replacement or in vivo differentiation of cells in the islets of Langerhans necessitate understanding of molecular programs driving pancreatic endocrine differentiation and maturation. Although signaling and transcription factor networks regulating different stages of pancreatic development and differentiation of islet cell types have been well studied, the unique role of transcription factors in programming the epigenome during lineage differentiation is largely unexplored. This study sought to determine how ISL1 regulatory networks control α- and ß-cell differentiation and how Isl1-deficiency cause endocrine progenitors to fail to produce functional α- and ß-cells, resulting in the severe diabetic phenotype associated with Isl1 deletion [31]. Here we show that ISL1 drives α cell differentiation and controls the acquisition of the β-cell mature endocrine phenotype. Using RNA sequencing together with CUT&Tag DNA sequencing, we uncovered changes in the epigenetic landscape of H3K4me3 and H3K27me3 modifications at promoter regions correlating with differential gene expression in Isl1CKO. Additionally, we explored multifaceted roles of ISL1 in epigenetic and transcriptional regulations by genome-wide profiling of ISL1 binding. These results indicated that the absence of Isl1 resulted in changes in transcriptional networks and epigenetic remodeling that may contribute to the regulation of gene expression and abnormalities in endocrine cell differentiation in the pancreas of Isl1CKO (see graphic summary in Fig. 9).Fig. 9Schematics of changes induced by the elimination of Isl1 in pancreas development There are two main contributions of our research to the understanding of ISL1 function. First, we are first to provide the evidence that ISL1 regulates an advancement towards the late stage of α-cell differentiation. Second, we are the first to report that ISL1 controls the maturation of β-cells. In terms of our first main contribution, transcriptome analyses of Isl1CKO endocrine cells at E14.5 revealed a shift towards the endocrine progenitor population at the expense of α-cell differentiation during the secondary transition of pancreas development of Isl1CKO. We found increased levels of Fev in Isl1CKO endocrine cells, associated with a transitory intermediate endocrine progenitor population before branching into α and ß differentiated endocrine cell types [25, 36, 37, 39]. For example, FEV+ endocrine progenitors co-expressing Peg10 tend to differentiate into α cells [36]. Our lineage reconstruction clustering analysis showed that α cell development was arrested in the early stage of α-cell differentiation, as we did not find any late α cells associated with high expression of Mafb, Meis2, and Scgn [37, 39] and expression of α-cell fate promoting transcription factors Peg10 and Pou6f2 [36, 40] was reduced in Isl1CKO endocrine cells. Motif enrichment analyses identified ISL1 and FEV as the top-ranking transcription factors associated with differentially expressed genes in Isl1CKOs at E14.5, suggesting possible functional interactions of ISL1 and FEV. Fev is an ISL1 direct target gene, as shown by our ISL1 CUT&Tag-seq. We uncovered that ISL1 directly binds promoters of key regulatory genes for pancreatic endocrine development, including Pax6, Mafb, Nkx6.2, Insm1, Sox9, Arx, Fev, Nkx6.1, Foxa2, Neurod1, Rfx3, and Rfx6 [1, 3, 5]. Additionally, motif analyses revealed the enrichment of motifs for NKX6.1 [50, 52], RFX6 [42, 56, 58], PDX1 [10, 49], and NKX2.2 [14, 57], critical regulators of endocrine cell development, at the sites occupied by ISL1, suggesting their cooperation with ISL1. *Downregulated* genes in Isl1CKO endocrine cells had enhanced depositions of silencing H3K27me3 in ISL1-bound genes. This suggests that the absence of ISL1 can alter the epigenetic landscape of endocrine cells during differentiation. For example, the ISL1 binding sites and H3K27me3 depositions were associated with downregulated genes defining the α- and β-cell lineages, among them Mafa, Meis2, Ripply3[62], Eya1, Phox2b, and Mnx1. Distinct regulatory patterns were observed in the upregulated genes associated with intermediate states of endocrine differentiation, including Chgb, Chga, Vim, Pbxip1, and Fev. Specifically, these genes exhibited ISL1 binding sites and a repressive H3K27me3 mark at their promoter regions in control endocrine cells. Our results further suggested that changes in distinct histone H3K4me3 and H3K27me3 methylation patterns at promoter regions of transcriptional regulators, such as Mafa, Ripply3, Phox2b, and Mnx1 contributed to regulation of gene expression in Isl1CKO. Thus, we have shown that deletion of Isl1 at the onset of endocrine cell formation results in a loss of multiple markers of α-cell identity characterized by the absence of glucagon expressing cells, and in a shift in the transcriptomic signature towards intermediate FEV+ progenitor states.
In terms of our second main contribution, ISL1 controls acquisition of the β-cell mature endocrine phenotype. At P9, the characteristic core-mantle organization of the islets with the β-cell core and α cells mainly at the islet periphery was disrupted in Isl1CKO together with reduced insulin and PDX1 levels in β cells. Our transcriptome profiling of Isl1CKO endocrine cells at P9 demonstrated compromised regulation of insulin secretion, including downregulation of transcription factors Pdx1 and Mafa. Both Pdx1 and Mafa are direct transcriptional targets of ISL1 in adult β cells [31, 62]. Although Mafa has been suggested as a possible factor contributing to the development of compromised β cells [31], it cannot explain neonatal diabetic phenotype of Isl1-deficient mice. Glucose-regulated insulin secretion, β-cell mass, and islet cell architecture are compromised only in adult mice with pancreas-specific Mafa deletion [60], in contrast to the neonatal diabetic phenotype in a β-cell-specific Pdx1 deletion mutant [63]. Notably, pancreas-specific knockout Mafa mice are glucose intolerant but have normal fasting glucose levels [60], whereas loss of PDX1 from pancreatic β cells causes overt hyperglycemia in different models [48, 49, 63]. Based on our results, we postulate that ISL1 regulates PDX1 expression during pancreatic β-cell maturation. PDX1 is critical to induce and maintain ß-cell maturation and identity [48, 49]. Compared to endocrine cells from the control pancreas, expression of ß-cell maturation markers, such as Ucn3, Trpm5, and G6pc2 [35, 47], was reduced in Isl1CKO, suggesting that ISL1 plays an important role in the maturation of ß cells. Additionally, we found an enrichment of endocrine progenitor signature genes, such as Fev, Chga/b, Vim, Cldn4, and Sez6l in Isl1CKO cells, representing an intermediate transition state endocrine progenitor population [25, 36, 37]. This transcriptional profile indicated that Isl1CKO cells might be “trapped” in a transition endocrine progenitor state. Consistent with the immature endocrine transcriptomic signature, loss of Isl1 was associated with significantly reduced levels of all major hormone transcripts Gcg, Ins1, Ins2, Ppy, and Sst. In addition, a silencing H3K27me3 modification was uniquely acquired at the promoter region of Ucn3 in Isl1CKO cells. Consistent with the idea that pancreatic endocrine cells display a mature functional glucose-stimulated-insulin-secretion phenotype at P9 [35], signature genes of endocrine progenitors, such as Fev, Chga, and Chgb [36], which have ISL1 binding sites at their promoter regions, were marked by a suppressive H3K27me3 mark in endocrine cells from the control P9 pancreas in contrast to an active state (H3K27me3− H3K4me3+) in Isl1CKO cells.
The limitation of this study is that bulk-cell sequencing approaches provide an average of molecular differences from multiple cells. Although we applied the deconvolution method to our bulk-cell RNA-seq data to estimate the cell proportion and transcriptomic signatures of major cell types in the developing pancreas, future single-cell RNA-seq analyses are needed to fully establish molecular differences linked to specific cell states, cell-to-cell variability, and uncover the pathways of cell lineage differentiation affected by Isl1 deletion.
## Conclusions
Significantly advancing previous research [31], our study has revealed the molecular bases of two different regulatory roles of ISL1 during the development of pancreatic endocrine cells. First, ISL1 controls an α-cell lineage fate. Our lineage reconstruction clustering analysis of Isl1CKO endocrine cells showed that α cell development was arrested in the early stage of α-cell differentiation, which was associated with a loss of multiple markers of α-cell identity that was characterized, in turn, by the absence of glucagon expressing cells during the secondary transition at E14.5. Second, ISL1 regulates the acquisition of the β-cell mature phenotype. At P9, when a mature functional glucose-stimulated-insulin-secretion phenotype of β cells is acquired [35], transcriptome profiling of Isl1CKO endocrine cells identified downregulation of key ß-cell regulators and mature ß-cell markers. Most importantly this was particularly the case of Pdx1, which is necessary for establishing and maintaining mature β cells [49, 63]. Additionally, an enrichment of endocrine progenitor signature genes towards FEV+ intermediate progenitor states indicates a shift in the transcriptomic signature in Isl1CKO. Such findings suggest abnormalities in the transition between immature and mature ß cells and correlate with the progressive decline of ß-cell function and diabetic phenotype of Isl1CKO. Moreover, our study provides the first insights into the function of ISL1 directly or indirectly orchestrating chromatin remodeling in correlation with gene expression changes during pancreatic endocrine development. Altogether, these results represent compelling evidence that ISL1 transcriptionally and epigenetically controls pancreas endocrine development, affecting α-cell lineage fate decisions and maturation processes of ß cells. Future studies are needed to determine how ISL1 modulates the epigenetic landscape and how Isl1 deficiency affects transcriptomic signatures of major cell types in the developing pancreas.
## Experimental model
Animal experiments were conducted according to protocols approved by the Animal Care and Use Committee of the Institute of Molecular Genetics, Czech Academy of Sciences. All experiments were performed with littermates (males and females) cross-bred from two transgenic mouse lines: floxed Isl1 [Isl1f/f; Isl1tm2Sev, Stock No: 028501 Jackson Laboratory, [15]], and Neurod1-Cre [Tg(Neurod1-cre)1Able, Stock No: 028364 Jackson Laboratory, [32]]. Lines were maintained on C57BL/6 background. Neurod1-Cre mice do not have any detectable phenotype. Breeding scheme: Female mice Isl1f/f were crossed with Isl1f/+; Neurod1-Cre males, in which, Neurod1-cre knock-in allele was inherited paternally to minimize the potential influence of maternal genotype on the developing embryos. Isl1f/+ or Isl1f/f mice were used as the controls. The reporter tdTomato line (Ai14, B6.Cg-Gt(ROSA)26Sortm14(CAG−tdTomato)Hze, Stock No: 7914 Jackson Laboratory) was used. Genotyping was performed by PCR on tail DNA (Additional file 1: Table S1). Mice were kept under standard experimental conditions with a constant temperature (23–24 °C) and fed on soy-free feed (LASvendi, Germany). The females were housed individually during the gestation period and the litter size was recorded. Blood glucose levels were measured in animals by glucometer (COUNTOUR TS, Bayer); blood glucose levels maintained above 13.9 mmol/L are classified as diabetic. For total pancreatic insulin content, pancreases were excised, weighed, minced, and homogenized in acid–ethanol. A hormone concentration in extracts was measured by ELISA using Mouse Insulin ELISA kit (Mercodia, Sweden).
## Reverse transcription-quantitative Real-Time Polymerase Chain Reaction
RT-qPCR was performed as described previously [64]. Briefly, total RNA was isolated from the whole pancreas at E12.5 and E14.5 or ($$n = 8$$ samples/group) by Trizol RNA extraction. Following RT, quantitative real-time PCR (qPCR) was performed with the initial AmpliTaq activation at 95 °C for 10 min, followed by 40 cycles at 95 °C for 15 s and 60 °C for 30 s, as described. The *Hprt1* gene was selected as the best reference gene for our analyses from a panel of 12 control genes (TATAA Biocenter AB, Sweden). The relative expression of a target gene was calculated based on qPCR efficiencies and the quantification cycle (Cq) difference (Δ) of an experimental sample versus a control. Primers were designed using Primer Blast tool (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). Primers were selected according to the following parameters: length between 18 and 24 bases, melting temperature (Tm) between 58° and 60 °C, G + C content between 40 and $60\%$ (optimal $50\%$) and efficiency above $80\%$. Primer sequences are presented in Additional file 1: Table S2.
## Immunohistochemistry and morphological evaluations
For vibratome sections, dissected tissues were fixed in $4\%$ PFA, embedded in $4\%$ agarose gel and sectioned at 80 µm on a Leica VT1000S vibratome. All company names and catalog numbers of primary and secondary antibodies, and their dilutions used in this study, are in Additional file 1: Table S3. The nuclei were counterstained with Hoechst 33342. Image acquisition was completed using the Zeiss LSM 880 NLO scanning confocal microscope, with ZEN lite software. The number of glucagon (GCG) and insulin (INS) expressing cells, and Ki67+ cells were counted in one vibratome section of Isl1CKO and control embryos or mice ($$n = 5$$ pancreases per genotype and for each age) with the largest pancreatic footprint per individual using the Cell Counter plugin of Image J (NIH). The number of INS and pHH3 expressing cells were counted in vibratome sections of Isl1CKO and control pancreases at E17.5 ($$n = 3$$ pancreases per genotype) and P0 ($$n = 4$$ pancreases per genotype). The number of NEUROD1+/ISL1+ cells at E10.5 were counted in the whole mount of the dorsal pancreas ($$n = 5$$ pancreases per genotype) using the Cell Counter plugin of Image J (NIH). For the evaluation of glucagon delaminating cells at E11.5 were quantified using the thresholding tool Image J (NIH) and expressed as a percentage of the total GCG+ area to PDX1+ area ($$n = 9$$ control pancreases; $$n = 8$$ Isl1CKO pancreases).
## Light-sheet fluorescent microscopy (LFSM) and analysis of images
The pancreas was microdissected from control-Ai14 and Isl1CKO-Ai14 mice (postnatal day P0). We used an advanced CUBIC protocol [65] for tissue clearing to enable efficient imaging by light-sheet microscopy. Briefly, the microdissected tissue was fixed in $4\%$ PFA for 1 h, washed with PBS, and incubated in a clearing solution Cubic 1 for 5 days at 37 ℃. Before immunolabeling, samples were washed in PBT ($0.5\%$ Triton-X in PBS) 4 × for 30 min. In addition to tdTomato expression, cleared samples were immunolabeled using different combinations of antibodies (anti-INS, anti-GLP1, and anti-TUBB3). Samples were stored before imaging in Cubic 2 at room temperature. Zeiss Lightsheet Z.1 microscope with illumination objective Lightsheet Z.1 5x/0.1 and detection objective Dry objective Lightsheet Z.1 5x/0.16 was used for imaging at the Light Microscopy Core Facility of the Institute of Molecular Genetics of the Czech Academy of Sciences. IMARIS software v8.1.1 (Bitplane AG, CA, USA) was used for image processing.
## Isolation of pancreatic endocrine cells
The pancreases from P9-P10 pups were first perfused by 1 mg/ml collagenase in Hank’s balanced salt solution, dissected, and incubated at 37 °C for 10–12 min to release endocrine cells/islets. Digested pancreatic tissue was washed 3 × by $1\%$ FBS in Hank’s solution. *To* generate single cells, the tissue was further dissociated by trypsinization as described [66]. Briefly, tissue was dissociated using $0.05\%$ trypsin/0.53 mM EDTA at 37 °C for 5 min. Digestion was stopped by the FACS buffer ($2\%$ FBS and 10 mM EGTA in PBS [66]), and cells were then 1 × washed by FACS buffer. The pancreases microdissected from E14.5 embryos were directly trypsinized and prepared for FACS as described above. Finally, cell suspensions were filtered through 40 µm nylon mesh and immediately tdTomato+ cells were sorted using a flow cytometer (BD FACSAria™ Fusion), through a 100 µm nozzle in 20 psi, operated with BD FACSDiva™ Software (Additional file 1: Fig. S7). For RNA sequencing, 100 sorted cells were collected into individual wells of 96-well plate containing 5 µl of lysis buffer of NEB Next single-cell low input RNA library prep kit for Illumina (New England Biolabs #E6420). Plates were frozen immediately on dry ice and stored at − 80 °C. The total time from euthanasia to cell collection was ∼3 h. For the epigenetic study, on average, 4700 cells/sample at E14.5 and 14,700 cells/sample at P9 were sorted. Cell sorting was performed in the Imaging Methods Core Facility at BIOCEV.
## RNA sequencing and analyses
RNA-seq libraries were prepared from 100 FACS-sorted cells/sample obtained from the pancreases of reporter Isl1CKO-Ai14 mutant ($$n = 5$$ samples) and reporter control-Ai14 ($$n = 6$$) from E14.5 embryos; and Isl1CKO-Ai14 mutant ($$n = 6$$) and reporter control-Ai14 ($$n = 5$$) from P9 mice. Each sample contained 100 tdTomato+ endocrine cells. Following the manufacturer's instructions, the NEB Next single-cell low input RNA library prep kit for Illumina was used for cDNA synthesis, amplification, and library generation [67] at the Gene Core Facility (Institute of Biotechnology CAS, Czechia). Fragment Analyzer assessed the quality of cDNA libraries. The libraries were sequenced on an Illumina NextSeq 500 next-generation sequencer. NextSeq $\frac{500}{550}$ High Output kit 75 cycles (Illumina #200,024,906) were processed at the Genomics and Bioinformatics Core Facility (Institute of Molecular Genetics CAS, Czechia). RNA-Seq reads in FASTQ files were mapped to the mouse genome using STAR [version 2.7.0c [68]] GRCm38 primary assembly and annotation version M8. The raw data of RNA sequencing were processed with a standard pipeline. Using cutadapt v1.18 [69], the number of reads (minimum, 32 million; maximum, 73 million) was trimmed by Illumina sequencing adaptor and of bases with reading quality lower than 20, subsequently reads shorter than 20 bp were filtered out TrimmomaticPE version 0.36 [70].
Ribosomal RNA and reads mapping to UniVec database were filtered out using bowtie v1.2.2. with parameters -S -n 1 and SortMeRNA [71]. A count table was generated by Rsubread v2.0.1 package using default parameters without counting multi mapping reads. The raw RNA-seq data were deposited at GEO: (https://www.ncbi.nlm.nih.gov/geo/).
DESeq2 [v1.26.0 [72]] default parameters were used to normalize data and compare the different groups. Genes were then filtered using the criteria of an adjusted P-value Padj < 0.05, and a base mean ≥ 50, and Fold change > 1.5 for upregulated genes and < 0.5 for downregulated genes for both E14.5 and P9 data to identify differentially expressed genes between Isl1CKO and control endocrine cells. The enrichment of the functional categories and functional annotation clustering of the differentially expressed genes was performed using g: Profiler [73] using version e104_eg51_p15_3922dba with g: SCS multiple testing correction methods applying a significance threshold of 0.05. Transcription factor (TF) enrichment analysis (TFEA) [41] was used to identify the enrichment of TF target genes in our set of differentially expressed genes. The top seven enriched TFs are listed (Additional file 6: Dataset S1c).
## Deconvolution of endocrine cell subtypes
Deconvolution was performed using the CibersortX algorithm at cibersortx.stanford.edu [74]. Single-cell transcriptomic profiling dataset of cells in the embryonic pancreas [39] was used as a reference, including count matrix and metadata labels. Particularly, only cells with pancreatic epithelial or endocrine cell fate were used, corresponding to the annotation of five broader cell types—α cells, β cells, endocrine progenitors, trunk epithelium and tip epithelium [39]. The reference matrix was built out of the 2589 cells and gene list of 18,565 gene features, as deposited by [39]. Each cell population counted > 250 cells. The units of the reference matrix were UMI counts. Calculation of the scRNA-seq signature matrix was done in default mode (quantile normalization disabled, minimal expression of 0.75, replicates of 5, sampling of 0.5). Imputation of cell fractions and group-mode expression were used in default settings, with S-mode batch correction enabled, quantile normalization disabled and $$n = 100$$ permutations for significance analysis. Sample mixture file was submitted with unfiltered gene list 27,124 features for Isl1CKO and in UMI counts.
## Cut&Tag sequencing and analyses
Bench top CUT&Tag version 3 was performed as previously described [46, 75], with minor modifications. Specifically, nuclei from freshly FACS-sorted tdTomato+ pancreatic endocrine cells were captured by Concanavalin A-coated magnetic beads to facilitate subsequent washing steps and the reaction was carried out in 0.2 ml PCR tubes. CUT&Tag validated primary antibodies anti-H3K4me3 (Active Motif, #39,159, 1:100), Anti-H3K27me3 (Active Motif, #39155, 1:100), anti-ISL1 (Developmental Hybridoma Bank, #39.4D5, 1:50), normal rabbit IgG negative control (EpiCypher, #13–0042, 0.5 µg/reaction), anti-mouse secondary antibody (EpiCypher, #13–0048, 0.5 µg/reaction), and anti-rabbit secondary antibody (EpiCypher, #13-0047, 0.5 µg/reaction). Binding of pAG-Tn5 (EpiCypher, 15-1017, 2.5µL/reaction) was at RT for 60 min followed by tagmentation. To stop tagmentation and solubilize DNA fragments, 1.67 μL 0.5 M EDTA, 0.5 μL $10\%$ SDS and 0.42 μL 20 mg/mL Proteinase K (20 mg/mL) was added to each sample. Samples were incubated for 1 h at 55 ºC to digest (and reverse cross-links), followed by DNA precipitation and purification. DNA was dissolved in 22 μL of 1 mM Tris–HCl pH 8, 0.1 mM EDTA buffer and utilized as template for library generation with Universal i5 Primer and Uniquely Barcoded i7 Primers for Illumina. DNA libraries were sequenced on the MiSeq, Illumina using MiSeq Reagent Kit v3, which allows extend read lengths up to 2 × 75 bp at the OMICS Genomics facility (BIOCEV). CUT&Tag-seq H3K27me3 data are from two independent biological replicates and CUT&Tag-seq H3K4me3 data are from one biological sample, each sample was pooled together from five to nine pancreases of E14.5 embryos (〜 4700 cells/sample) and from two to three pancreases of P9 mice per genotype (〜 14,700 cells/sample). CUT&Tag-seq ISL1 data are from two independent biological replicates, each biological sample was pooled together from eleven E14.5 pancreases (〜 6100 cells/sample).
Data analyses were performed following CUT&Tag Data processing tutorial [76]. Paired-end sequencing data were mapped using bowtie2 (version 2.2.5) [77] to mouse genome GRCm38 primary assembly. PCR duplicates were not removed. After filtering and conversion to bedgraph format, peak calling was performed with usage of relevant IgG controls and stringent mode with tool SEACR (version 1.3) [78]. Peaks were annotated using CHIPseeker (version 1.30.3) [79] and annotation version M8 of mouse genome. Enriched peak detection was performed using EdgeR (version 3.36.0) [80] with filtering criteria for H3K4me3 modification > 25 CPM (count per million). Enrichments in H3K27me3 peaks between control and Isl1CKO were identified in the pairwise comparisons as a peak difference equal to or greater than threefold. Comparative analyses of ISL1 and H3K27me CUT&Tag-seq data were done as follows. After filtering of ISL1 CUT&*Tag data* and conversion to bam format, bam files from replicates were merged with samtools (version 1.6). Peak calling was performed with usage of relevant IgG controls and applying the –keep-dup all –nomodel –extsize 200 settings to the callpeak command in MACS2 (version 1.3) [78]. Peaks were annotated using CHIPseeker (version 1.30.3) [79] and annotation version M8 of mouse genome. Normalization of samples was performed using EdgeR (version 3.36.0) [80]. Sum of normalized counts of all peaks in a gene was performed for detecting epigenetic enrichment at the gene level. A gene was considered as enriched for H3K27me3 between control and Isl1CKO when the total levels was equal to or greater than fivefold. Motif enrichment analysis was performed using the findMotifsGenome.pl function in HOMER (version 4.11). Bed files with ISL1 peak summits were used as input to look for motif enrichment in a 2 kb area.
## Experimental design and statistical analyses
All comparisons were made between animals with the same genetic background, typically littermates, and we used male and female mice. The number of samples (n) for each comparison can be found in the individual method descriptions and are given in the corresponding figure legends. Phenotyping and data analysis were performed blind to the genotype of the mice. All values are presented either as the mean ± standard deviation (SD) or standard error of the mean (SEM). For statistical analysis, GraphPad Prism software was used. To assess differences in the mean, one-way or two-way ANOVA with Bonferroni's multiple comparison test, and unpaired two-tailed t-tests were employed. Significance was determined as $P \leq 0.05$ (*), $P \leq 0.01$ (**), $P \leq 0.001$ (***) or $P \leq 0.0001$ (****). Complete results of the statistical analyses are included in the figure legends.
## Supplementary Information
Additional file 1: Fig. S1. Efficient deletion of ISL1 in the Isl1CKO developing pancreas. Representative whole-mount immunolabeling of the pancreas of tdTomato reporter control-Ai14 and Isl1CKO-Ai14 embryos during the primary transitions (E10.5 and E11.5) and the secondary transition (E13.5) shows ISL1 expression in the Neurod1Cre positive domain visualized by tdTomato expression. The pancreatic epithelium is delineated by the expression of PDX1. Scale bars: 50 μm. Fig. S2. Diabetic phenotype of Isl1CKO. a The average blood glucose levels over time (from 1 week to 5 weeks of age) in females and males mice fed ad libitum. The 5 weeks of age female mice had blood glucose unmeasurable (above 35 mmol/l), the 5 week of age males show a high variability with 7 mice with blood glucose unmeasurable, and 3 with 0.8, 7.9 and 31.9 mmol/l. Data are presented as mean ± SD, analyzed by VA (****$P \leq 0.0001$). b The average body weight of adult female and male mice. Data are presented as mean ± SEM, Student’s t test. c Glucose tolerance test plotted using glucose vs time in heterozygous (Neurod1Cre/Isl1flox+/−), and control mice. Data are presented as mean ± SD, analyzed by twoway Anova with Bonferroni post-hoc analysis for glucose vs time. d Blood glucose concentration in adults fed ad libitum, 6-8 weeks of age. Only measurable levels of glucose (< 35 mmol/l) are shown for Isl1CKO. Data are presented as mean ± SEM, Student’s t test (****$P \leq 0.0001$). e The weight of pancreas of adult mice. Total pancreas weights of 244 ± 78 mg ($$n = 22$$) in adult Isl1CKO mutants compared to those of controls (211 ± 57 mg, $$n = 21$$). Data are presented as mean ± SEM, Student’s t test. Fig. S3. Immunolabeling for the marker of cellular mitosis, phosphorylated histone H3 (pHH3). Representative sections from the control and Isl1CKO pancreas immunostained for insulin (INS) and PDX1 (marker of β cells) at P0 and E17.5. Scale bars: 50 μm. Fig. S4. H3K4me3 and H3K27me3 CUT&Tag-seq analyses of E14.5 pancreatic endocrine cells. a The functional enrichment analysis of the differentially expressed genes in Isl1CKO was performed using g: Profiler (Gene Ontology: MF, molecular function; BP, biological processes). b A UMAP overview of nine cell clusters of different pancreatic cell types used as a reference for the cell type deconvolution analysis (van Gurp et al., 2019). c The deconvolved cell cluster proportions in E14.5 endocrine population from our bulk RNA-seq data of Isl1CKO and control. d The UCSC browser view of whole genome showing H3K4me3 and H3K27me3 peaks in control and Isl1CKO endocrine cells at E14.5 based on CUT&Tag-seq analyses. The mapped read counts distributed across all chromosomes are in comparable read depth for control and Isl1CKO samples. e Bar plot showing percentage of H3K4me3 and H3K27me3 peaks at promoter regions (± 3 kb from TSS), gene body regions, and intergenic regions. f Pie chart illustrating the proportion of differentially expressed genes that differentially exhibited one or both H3K4me3 and H3K27me3 marks at their promoter regions from pairwise comparison of Isl1CKO and control pancreatic endocrine cells at E14.5. Fig. S5. H3K4me3 and H3K27me3 CUT&Tag-seq analyses of P9 pancreatic endocrine cells. a The UCSC browser view of whole genome showing H3K4me3 and H3K27me3 peaks in control and Isl1CKO endocrine cells at P9 based on CUT&Tag-seq analyses. The mapped read counts distributed across all chromosomes are in comparable read depth for control and Isl1CKO samples. b Bar plot showing percentage of H3K4me3 and H3K27me3 peaks at promoter regions (± 3 kb from TSS), gene body regions, and intergenic regions of endocrine cells of the P9 pancreas. Fig. S6. ISL1 binding CUT&Tag-seq analyses of E14.5 pancreatic endocrine cells. a Pie chart showing genomic distribution of ISL1 loci in E14.5 pancreatic endocrine cells. b The most enriched Gene Ontology (GO) biological processes for genes bound by ISL1 and with Isl1CKO-specific H3K27me3 depositions. Fig. S7. Gating strategy used to isolate tdTomato+ cells. Representative example to show gating to purify live and individual tdTomato+ cells for RNA-seq and CUT&Tag-seq. Table S1. Primer sequences for genotyping. Table S2. Primer sequences for RT-qPCR. Table S3. List of antibodies. Additional file 2: Video S1. P9 control tdTomato GLP1 INS. Microdissected pancreas of tdTomato reporter control-Ai14 mice was cleared (CUBIC protocol), immunolabeled, imaged, and reconstructed in 3D using light-sheet fluorescence microscopy (LFSM). Video shows the distribution and formation of islets in the anatomical microenvironment of the pancreas at P9; tdTomato+ endocrine cell population (magenta), β cells with expression of insulin (white), and α cells expressing glucagon-like peptide-1 (green).Additional file 3: Video S2. P9 ISL1CKO tdTomato GLP1 INS. Microdissected pancreas of tdTomato reporter Isl1CKO-Ai14 mice were cleared (CUBIC protocol), immunolabeled, imaged, and reconstructed in 3D using light-sheet fluorescence microscopy (LFSM). Video shows the distribution and formation of islets in the anatomical microenvironment of the pancreas; tdTomato+ endocrine cell population (magenta), β cells with expression of insulin (white), and α cells expressing glucagon-like (green)Additional file 4: Video S3. P9 control tdTomato GLP1 Tubulin. Microdissected pancreas of tdTomato reporter control-Ai14 mice were cleared (CUBIC protocol), immunolabeled, imaged, and reconstructed in 3D using light-sheet fluorescence microscopy (LFSM). LFSM video shows the distribution and formation of islets in the anatomical microenvironment of the pancreas at P9; tdTomato+ endocrine cells (magenta), α cells expressing glucagon-like peptide-1 (green), and neuronal fibers labeled by antitubulin (white fibers).Additional file 5: Video S4. P9 Isl1CKO tdTomato GLP1 Tubulin. Microdissected pancreas of tdTomato reporter Isl1CKO-Ai14 mice were cleared (CUBIC protocol), immunolabeled, imaged, and reconstructed in 3D using light-sheet fluorescence microscopy (LFSM). LFSM video shows the distribution and formation of islets in the anatomical microenvironment of the pancreas at P9; tdTomato+ endocrine cells (magenta), α cells expressing glucagon-like peptide-1 (green) and neuronal fibers labeled by anti-tubulin (white fibers).Additional file 6: Dataset S1a. Differentially expressed genes based on RNA sequencing of pancreatic endocrine cells at E14.5. S1b. Deconvolution analyses. S1c. Transcription factor enrichment analysis for differentially expressed genes at E14.5. S1d. Differentially expressed genes based on RNA sequencing of P9 pancreatic endocrine cells. S1e. GO terms enrichment for differentially expressed genes based on RNA-seq at P9. S1f. *Upregulated* genes at P9 defining the endocrine progenitor state. Additional file 7: Dataset S2. ISL1 and H3K27me3 CUT&Tag-seq data analyses: pancreatic endocrine cells at E14.5.Additional file 8: Dataset S3. ISL1 and H3K27me3 CUT&Tag-seq data analyses: pancreatic endocrine cells at P9.
## References
1. Bastidas-Ponce A, Scheibner K, Lickert H, Bakhti M. **Cellular and molecular mechanisms coordinating pancreas development**. *Development* (2017.0) **144** 2873-2888. DOI: 10.1242/dev.140756
2. Pan FC, Wright C. **Pancreas organogenesis: from bud to plexus to gland**. *Dev Dyn* (2011.0) **240** 530-565. DOI: 10.1002/dvdy.22584
3. Bakhti M, Bottcher A, Lickert H. **Modelling the endocrine pancreas in health and disease**. *Nat Rev Endocrinol* (2019.0) **15** 155-171. DOI: 10.1038/s41574-018-0132-z
4. Johansson KA, Dursun U, Jordan N, Gu G, Beermann F, Gradwohl G, Grapin-Botton A. **Temporal control of neurogenin3 activity in pancreas progenitors reveals competence windows for the generation of different endocrine cell types**. *Dev Cell* (2007.0) **12** 457-465. DOI: 10.1016/j.devcel.2007.02.010
5. Scavuzzo MA, Hill MC, Chmielowiec J, Yang D, Teaw J, Sheng K, Kong Y, Bettini M, Zong C, Martin JF, Borowiak M. **Endocrine lineage biases arise in temporally distinct endocrine progenitors during pancreatic morphogenesis**. *Nat Commun* (2018.0) **9** 3356. DOI: 10.1038/s41467-018-05740-1
6. Gu G, Dubauskaite J, Melton DA. **Direct evidence for the pancreatic lineage: NGN3+ cells are islet progenitors and are distinct from duct progenitors**. *Development* (2002.0) **129** 2447-2457. DOI: 10.1242/dev.129.10.2447
7. Yu XX, Xu CR. **Understanding generation and regeneration of pancreatic beta cells from a single-cell perspective**. *Development* (2020.0). DOI: 10.1242/dev.179051
8. Liu J, Banerjee A, Herring CA, Attalla J, Hu R, Xu Y, Shao Q, Simmons AJ, Dadi PK, Wang S, Jacobson DA, Liu B, Hodges E, Lau KS, Gu G. **Neurog3-independent methylation is the earliest detectable mark distinguishing pancreatic progenitor identity**. *Dev Cell* (2019.0) **48** 49-63 e47. DOI: 10.1016/j.devcel.2018.11.048
9. Bramswig NC, Everett LJ, Schug J, Dorrell C, Liu C, Luo Y, Streeter PR, Naji A, Grompe M, Kaestner KH. **Epigenomic plasticity enables human pancreatic alpha to beta cell reprogramming**. *J Clin Invest* (2013.0) **123** 1275-1284. DOI: 10.1172/JCI66514
10. Spaeth JM, Liu JH, Peters D, Guo M, Osipovich AB, Mohammadi F, Roy N, Bhushan A, Magnuson MA, Hebrok M, Wright CVE, Stein R. **The Pdx1-Bound Swi/Snf Chromatin Remodeling Complex Regulates Pancreatic Progenitor Cell Proliferation and Mature Islet beta-Cell Function**. *Diabetes* (2019.0) **68** 1806-1818. DOI: 10.2337/db19-0349
11. Yu XX, Qiu WL, Yang L, Li LC, Zhang YW, Xu CR. **Dynamics of chromatin marks and the role of JMJD3 during pancreatic endocrine cell fate commitment**. *Development* (2018.0). DOI: 10.1242/dev.163162
12. Ackermann AM, Wang Z, Schug J, Naji A, Kaestner KH. **Integration of ATAC-seq and RNA-seq identifies human alpha cell and beta cell signature genes**. *Mol Metab* (2016.0) **5** 233-244. DOI: 10.1016/j.molmet.2016.01.002
13. Dhawan S, Georgia S, Tschen SI, Fan G, Bhushan A. **Pancreatic beta cell identity is maintained by DNA methylation-mediated repression of Arx**. *Dev Cell* (2011.0) **20** 419-429. DOI: 10.1016/j.devcel.2011.03.012
14. Papizan JB, Singer RA, Tschen SI, Dhawan S, Friel JM, Hipkens SB, Magnuson MA, Bhushan A, Sussel L. **Nkx2.2 repressor complex regulates islet beta-cell specification and prevents beta-to-alpha-cell reprogramming**. *Genes Dev.* (2011.0) **25** 2291-2305. DOI: 10.1101/gad.173039.111
15. Sun Y, Dykes IM, Liang X, Eng SR, Evans SM, Turner EE. **A central role for Islet1 in sensory neuron development linking sensory and spinal gene regulatory programs**. *Nat Neurosci* (2008.0) **11** 1283-1293. DOI: 10.1038/nn.2209
16. Cai CL, Liang X, Shi Y, Chu PH, Pfaff SL, Chen J, Evans S. **Isl1 identifies a cardiac progenitor population that proliferates prior to differentiation and contributes a majority of cells to the heart**. *Dev Cell* (2003.0) **5** 877-889. DOI: 10.1016/S1534-5807(03)00363-0
17. Whitney IE, Raven MA, Ciobanu DC, Poche RA, Ding Q, Elshatory Y, Gan L, Williams RW, Reese BE. **Genetic modulation of horizontal cell number in the mouse retina**. *Proc Natl Acad Sci U S A* (2011.0) **108** 9697-9702. DOI: 10.1073/pnas.1103253108
18. Filova I, Pysanenko K, Tavakoli M, Vochyanova S, Dvorakova M, Bohuslavova R, Smolik O, Fabriciova V, Hrabalova P, Benesova S, Valihrach L, Cerny J, Yamoah EN, Syka J, Fritzsch B, Pavlinkova G. **ISL1 is necessary for auditory neuron development and contributes toward tonotopic organization**. *Proc Natl Acad Sci U S A* (2022.0) **119** e2207433119. DOI: 10.1073/pnas.2207433119
19. Yang Z, Zhang Q, Lu Q, Jia Z, Chen P, Ma K, Wang W, Zhou C. **ISL-1 promotes pancreatic islet cell proliferation by forming an ISL-1/Set7/9/PDX-1 complex**. *Cell Cycle* (2015.0) **14** 3820-3829. DOI: 10.1080/15384101.2015.1069926
20. Yu Z, Kong J, Pan B, Sun H, Lv T, Zhu J, Huang G, Tian J. **Islet-1 may function as an assistant factor for histone acetylation and regulation of cardiac development-related transcription factor Mef2c expression**. *PLoS ONE* (2013.0) **8** e77690. DOI: 10.1371/journal.pone.0077690
21. Wang Y, Li Y, Guo C, Lu Q, Wang W, Jia Z, Chen P, Ma K, Reinberg D, Zhou C. **ISL1 and JMJD3 synergistically control cardiac differentiation of embryonic stem cells**. *Nucleic Acids Res* (2016.0) **44** 6741-6755. DOI: 10.1093/nar/gkw301
22. Caputo L, Witzel HR, Kolovos P, Cheedipudi S, Looso M, Mylona A, van IJcken WFJ, Laugwitz KL, Evans SM, Braun T, Soler E, Grosveld F, Dobreva G. **The Isl1/Ldb1 Complex Orchestrates Genome-wide Chromatin Organization to Instruct Differentiation of Multipotent Cardiac Progenitors**. *Cell Stem Cell* (2015.0) **17** 287-299. DOI: 10.1016/j.stem.2015.08.007
23. Gao R, Liang X, Cheedipudi S, Cordero J, Jiang X, Zhang Q, Caputo L, Gunther S, Kuenne C, Ren Y, Bhattacharya S, Yuan X, Barreto G, Chen Y, Braun T, Evans SM, Sun Y, Dobreva G. **Pioneering function of Isl1 in the epigenetic control of cardiomyocyte cell fate**. *Cell Res* (2019.0) **29** 486-501. DOI: 10.1038/s41422-019-0168-1
24. Ahlgren U, Pfaff SL, Jessell TM, Edlund T, Edlund H. **Independent requirement for ISL1 in formation of pancreatic mesenchyme and islet cells**. *Nature* (1997.0) **385** 257-260. DOI: 10.1038/385257a0
25. Krentz NAJ, Lee MYY, Xu EE, Sproul SLJ, Maslova A, Sasaki S, Lynn FC. **Single-cell transcriptome profiling of mouse and hESC-Derived Pancreatic Progenitors**. *Stem Cell Reports* (2018.0) **11** 1551-1564. DOI: 10.1016/j.stemcr.2018.11.008
26. Muraro MJ, Dharmadhikari G, Grun D, Groen N, Dielen T, Jansen E, van Gurp L, Engelse MA, Carlotti F, de Koning EJ, van Oudenaarden A. **A single-cell transcriptome atlas of the human pancreas**. *Cell Syst* (2016.0) **3** 385-394383. DOI: 10.1016/j.cels.2016.09.002
27. Xin Y, Kim J, Ni M, Wei Y, Okamoto H, Lee J, Adler C, Cavino K, Murphy AJ, Yancopoulos GD, Lin HC, Gromada J. **Use of the Fluidigm C1 platform for RNA sequencing of single mouse pancreatic islet cells**. *Proc Natl Acad Sci U S A* (2016.0) **113** 3293-3298. DOI: 10.1073/pnas.1602306113
28. Zhang H, Wang WP, Guo T, Yang JC, Chen P, Ma KT, Guan YF, Zhou CY. **The LIM-homeodomain protein ISL1 activates insulin gene promoter directly through synergy with BETA2**. *J Mol Biol* (2009.0) **392** 566-577. DOI: 10.1016/j.jmb.2009.07.036
29. Wang W, Shi Q, Guo T, Yang Z, Jia Z, Chen P, Zhou C. **PDX1 and ISL1 differentially coordinate with epigenetic modifications to regulate insulin gene expression in varied glucose concentrations**. *Mol Cell Endocrinol* (2016.0) **428** 38-48. DOI: 10.1016/j.mce.2016.03.019
30. Pfaff SL, Mendelsohn M, Stewart CL, Edlund T, Jessell TM. **Requirement for LIM homeobox gene Isl1 in motor neuron generation reveals a motor neuron-dependent step in interneuron differentiation**. *Cell* (1996.0) **84** 309-320. DOI: 10.1016/S0092-8674(00)80985-X
31. Du A, Hunter CS, Murray J, Noble D, Cai CL, Evans SM, Stein R, May CL. **Islet-1 is required for the maturation, proliferation, and survival of the endocrine pancreas**. *Diabetes* (2009.0) **58** 2059-2069. DOI: 10.2337/db08-0987
32. Li HJ, Kapoor A, Giel-Moloney M, Rindi G, Leiter AB. **Notch signaling differentially regulates the cell fate of early endocrine precursor cells and their maturing descendants in the mouse pancreas and intestine**. *Dev Biol* (2012.0) **371** 156-169. DOI: 10.1016/j.ydbio.2012.08.023
33. Herrera PL. **Adult insulin- and glucagon-producing cells differentiate from two independent cell lineages**. *Development* (2000.0) **127** 2317-2322. DOI: 10.1242/dev.127.11.2317
34. Fernandez-Ruiz R, García-Alamán A, Esteban Y, Mir-Coll J, Serra-Navarro B, Fontcuberta-PiSunyer M, Broca C, Armanet M, Wojtusciszyn A, Kram V, Young MF, Vidal J, Gomis R, Gasa R. **Wisp1 is a circulating factor that stimulates proliferation of adult mouse and human beta cells**. *Nat Commun* (2020.0) **11** 5982. DOI: 10.1038/s41467-020-19657-1
35. Blum B, Hrvatin S, Schuetz C, Bonal C, Rezania A, Melton DA. **Functional beta-cell maturation is marked by an increased glucose threshold and by expression of urocortin 3**. *Nat Biotechnol* (2012.0) **30** 261-264. DOI: 10.1038/nbt.2141
36. Byrnes LE, Wong DM, Subramaniam M, Meyer NP, Gilchrist CL, Knox SM, Tward AD, Ye CJ, Sneddon JB. **Lineage dynamics of murine pancreatic development at single-cell resolution**. *Nat Commun* (2018.0) **9** 3922. DOI: 10.1038/s41467-018-06176-3
37. Bastidas-Ponce A, Tritschler S, Dony L, Scheibner K, Tarquis-Medina M, Salinno C, Schirge S, Burtscher I, Bottcher A, Theis FJ, Lickert H, Bakhti M. **Comprehensive single cell mRNA profiling reveals a detailed roadmap for pancreatic endocrinogenesis**. *Development* (2019.0). DOI: 10.1242/dev.173849
38. Qiu WL, Zhang YW, Feng Y, Li LC, Yang L, Xu CR. **Deciphering Pancreatic Islet beta cell and alpha cell maturation pathways and characteristic features at the single-cell level**. *Cell Metab* (2017.0) **25** 1194-12051194. DOI: 10.1016/j.cmet.2017.04.003
39. van Gurp L, Muraro MJ, Dielen T, Seneby L, Dharmadhikari G, Gradwohl G, van Oudenaarden A, de Koning EJP. **A transcriptomic roadmap to alpha- and beta-cell differentiation in the embryonic pancreas**. *Development* (2019.0). DOI: 10.1242/dev.173716
40. Baron M, Veres A, Wolock SL, Faust AL, Gaujoux R, Vetere A, Ryu JH, Wagner BK, Shen-Orr SS, Klein AM, Melton DA, Yanai I. **A single-cell transcriptomic map of the human and mouse pancreas reveals inter- and intra-cell population structure**. *Cell Syst* (2016.0) **3** 346-360 e344. DOI: 10.1016/j.cels.2016.08.011
41. Keenan AB, Torre D, Lachmann A, Leong AK, Wojciechowicz ML, Utti V, Jagodnik KM, Kropiwnicki E, Wang Z, Ma'ayan A. **ChEA3: transcription factor enrichment analysis by orthogonal omics integration**. *Nucleic Acids Res* (2019.0) **47** W212-W224. DOI: 10.1093/nar/gkz446
42. Taleb N, Polychronakos C. **RFX6 is needed for the development and maintenance of the beta-cell phenotype**. *Islets* (2011.0) **3** 291-293. DOI: 10.4161/isl.3.5.15944
43. Sojoodi M, Stradiot L, Tanaka K, Heremans Y, Leuckx G, Besson V, Staels W, Van de Casteele M, Marazzi G, Sassoon D, Heimberg H, Bonfanti P. **The zinc finger transcription factor PW1/PEG3 restrains murine beta cell cycling**. *Diabetologia* (2016.0) **59** 1474-1479. DOI: 10.1007/s00125-016-3954-z
44. Blanco E, Gonzalez-Ramirez M, Alcaine-Colet A, Aranda S, Di Croce L. **The bivalent genome: characterization, structure, and regulation**. *Trends Genet* (2020.0) **36** 118-131. DOI: 10.1016/j.tig.2019.11.004
45. Mikkelsen TS, Ku M, Jaffe DB, Issac B, Lieberman E, Giannoukos G, Alvarez P, Brockman W, Kim TK, Koche RP, Lee W, Mendenhall E, O'Donovan A, Presser A, Russ C, Xie X, Meissner A, Wernig M, Jaenisch R, Nusbaum C, Lander ES, Bernstein BE. **Genome-wide maps of chromatin state in pluripotent and lineage-committed cells**. *Nature* (2007.0) **448** 553-560. DOI: 10.1038/nature06008
46. Kaya-Okur HS, Wu SJ, Codomo CA, Pledger ES, Bryson TD, Henikoff JG, Ahmad K, Henikoff S. **CUT&Tag for efficient epigenomic profiling of small samples and single cells**. *Nat Commun* (2019.0) **10** 1930. DOI: 10.1038/s41467-019-09982-5
47. Balboa D, Barsby T, Lithovius V, Saarimäki-Vire J, Omar-Hmeadi M, Dyachok O, Montaser H, Lund P-E, Yang M, Ibrahim H, Näätänen A, Chandra V, Vihinen H, Jokitalo E, Kvist J, Ustinov J, Nieminen AI, Kuuluvainen E, Hietakangas V, Katajisto P, Lau J, Carlsson P-O, Barg S, Tengholm A, Otonkoski T. **Functional, metabolic and transcriptional maturation of human pancreatic islets derived from stem cells**. *Nat Biotechnol* (2022.0). DOI: 10.1038/s41587-022-01219-z
48. Bastidas-Ponce A, Roscioni SS, Burtscher I, Bader E, Sterr M, Bakhti M, Lickert H. **Foxa2 and Pdx1 cooperatively regulate postnatal maturation of pancreatic beta-cells**. *Mol Metab* (2017.0) **6** 524-534. DOI: 10.1016/j.molmet.2017.03.007
49. Gao T, McKenna B, Li C, Reichert M, Nguyen J, Singh T, Yang C, Pannikar A, Doliba N, Zhang T, Stoffers DA, Edlund H, Matschinsky F, Stein R, Stanger BZ. **Pdx1 maintains beta cell identity and function by repressing an alpha cell program**. *Cell Metab* (2014.0) **19** 259-271. DOI: 10.1016/j.cmet.2013.12.002
50. Aigha II, Abdelalim EM. **NKX6.1 transcription factor: a crucial regulator of pancreatic beta cell development, identity, and proliferation**. *Stem Cell Res Ther* (2020.0) **11** 459. DOI: 10.1186/s13287-020-01977-0
51. Schwitzgebel VM, Scheel DW, Conners JR, Kalamaras J, Lee JE, Anderson DJ, Sussel L, Johnson JD, German MS. **Expression of neurogenin3 reveals an islet cell precursor population in the pancreas**. *Development* (2000.0) **127** 3533-3542. DOI: 10.1242/dev.127.16.3533
52. Schaffer AE, Taylor BL, Benthuysen JR, Liu J, Thorel F, Yuan W, Jiao Y, Kaestner KH, Herrera PL, Magnuson MA, May CL, Sander M. **Nkx6.1 controls a gene regulatory network required for establishing and maintaining pancreatic Beta cell identity**. *PLoS Genet* (2013.0) **9** e1003274. DOI: 10.1371/journal.pgen.1003274
53. Civelek M, Lusis AJ. **Conducting the metabolic syndrome orchestra**. *Nat Genet* (2011.0) **43** 506-508. DOI: 10.1038/ng.842
54. Wang L, Tong X, Gu F, Zhang L, Chen W, Cheng X, Xie L, Chang Y, Zhang H. **The KLF14 transcription factor regulates hepatic gluconeogenesis in mice**. *J Biol Chem* (2017.0) **292** 21631-21642. DOI: 10.1074/jbc.RA117.000184
55. Rady B, Chen Y, Vaca P, Wang Q, Wang Y, Salmon P, Oberholzer J. **Overexpression of E2F3 promotes proliferation of functional human beta cells without induction of apoptosis**. *Cell Cycle* (2013.0) **12** 2691-2702. DOI: 10.4161/cc.25834
56. Smith SB, Qu HQ, Taleb N, Kishimoto NY, Scheel DW, Lu Y, Patch AM, Grabs R, Wang J, Lynn FC, Miyatsuka T, Mitchell J, Seerke R, Desir J, Vanden Eijnden S, Abramowicz M, Kacet N, Weill J, Renard ME, Gentile M, Hansen I, Dewar K, Hattersley AT, Wang R, Wilson ME, Johnson JD, Polychronakos C, German MS. **Rfx6 directs islet formation and insulin production in mice and humans**. *Nature* (2010.0) **463** 775-780. DOI: 10.1038/nature08748
57. Churchill AJ, Gutierrez GD, Singer RA, Lorberbaum DS, Fischer KA, Sussel L. **Genetic evidence that Nkx2.2 acts primarily downstream of Neurog3 in pancreatic endocrine lineage development**. *Elife* (2017.0). DOI: 10.7554/eLife.20010
58. Piccand J, Strasser P, Hodson DJ, Meunier A, Ye T, Keime C, Birling MC, Rutter GA, Gradwohl G. **Rfx6 maintains the functional identity of adult pancreatic beta cells**. *Cell Rep* (2014.0) **9** 2219-2232. DOI: 10.1016/j.celrep.2014.11.033
59. Kim SK, Selleri L, Lee JS, Zhang AY, Gu X, Jacobs Y, Cleary ML. **Pbx1 inactivation disrupts pancreas development and in Ipf1-deficient mice promotes diabetes mellitus**. *Nat Genet* (2002.0) **30** 430-435. DOI: 10.1038/ng860
60. Hang Y, Yamamoto T, Benninger RK, Brissova M, Guo M, Bush W, Piston DW, Powers AC, Magnuson M, Thurmond DC, Stein R. **The MafA transcription factor becomes essential to islet beta-cells soon after birth**. *Diabetes* (2014.0) **63** 1994-2005. DOI: 10.2337/db13-1001
61. Vedovato N, Cliff E, Proks P, Poovazhagi V, Flanagan SE, Ellard S, Hattersley AT, Ashcroft FM. **Neonatal diabetes caused by a homozygous KCNJ11 mutation demonstrates that tiny changes in ATP sensitivity markedly affect diabetes risk**. *Diabetologia* (2016.0) **59** 1430-1436. DOI: 10.1007/s00125-016-3964-x
62. Ediger BN, Du A, Liu J, Hunter CS, Walp ER, Schug J, Kaestner KH, Stein R, Stoffers DA, May CL. **Islet-1 Is essential for pancreatic beta-cell function**. *Diabetes* (2014.0) **63** 4206-4217. DOI: 10.2337/db14-0096
63. Gannon M, Ables ET, Crawford L, Lowe D, Offield MF, Magnuson MA, Wright CV. **pdx-1 function is specifically required in embryonic beta cells to generate appropriate numbers of endocrine cell types and maintain glucose homeostasis**. *Dev Biol* (2008.0) **314** 406-417. DOI: 10.1016/j.ydbio.2007.10.038
64. Cerychova R, Bohuslavova R, Papousek F, Sedmera D, Abaffy P, Benes V, Kolar F, Pavlinkova G. **Adverse effects of Hif1a mutation and maternal diabetes on the offspring heart**. *Cardiovasc Diabetol* (2018.0) **17** 68. DOI: 10.1186/s12933-018-0713-0
65. Susaki EA, Tainaka K, Perrin D, Yukinaga H, Kuno A, Ueda HR. **Advanced CUBIC protocols for whole-brain and whole-body clearing and imaging**. *Nat Protoc* (2015.0) **10** 1709-1727. DOI: 10.1038/nprot.2015.085
66. Sugiyama T, Rodriguez RT, McLean GW, Kim SK. **Conserved markers of fetal pancreatic epithelium permit prospective isolation of islet progenitor cells by FACS**. *Proc Natl Acad Sci U S A* (2007.0) **104** 175-180. DOI: 10.1073/pnas.0609490104
67. Mamanova L, Miao Z, Jinat A, Ellis P, Shirley L, Teichmann SA. **High-throughput full-length single-cell RNA-seq automation**. *Nat Protoc* (2021.0) **16** 2886-2915. DOI: 10.1038/s41596-021-00523-3
68. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR. **STAR: ultrafast universal RNA-seq aligner**. *Bioinformatics* (2013.0) **29** 15-21. DOI: 10.1093/bioinformatics/bts635
69. Martin M. **Cutadapt removes adapter sequences from high-throughput sequencing reads**. *EMBnet J* (2011.0) **17** 3. DOI: 10.14806/ej.17.1.200
70. 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
71. Kopylova E, Noe L, Touzet H. **SortMeRNA: fast and accurate filtering of ribosomal RNAs in metatranscriptomic data**. *Bioinformatics* (2012.0) **28** 3211-3217. DOI: 10.1093/bioinformatics/bts611
72. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol* (2014.0) **15** 550. DOI: 10.1186/s13059-014-0550-8
73. Raudvere U, Kolberg L, Kuzmin I, Arak T, Adler P, Peterson H, Vilo J. **g:Profiler: a web server for functional enrichment analysis and conversions of gene lists (2019 update)**. *Nucleic Acids Res* (2019.0) **47** W191-W198. DOI: 10.1093/nar/gkz369
74. Newman AM, Steen CB, Liu CL, Gentles AJ, Chaudhuri AA, Scherer F, Khodadoust MS, Esfahani MS, Luca BA, Steiner D, Diehn M, Alizadeh AA. **Determining cell type abundance and expression from bulk tissues with digital cytometry**. *Nat Biotechnol* (2019.0) **37** 773-782. DOI: 10.1038/s41587-019-0114-2
75. Kaya-Okur HS, Janssens DH, Henikoff JG, Ahmad K, Henikoff S. **Efficient low-cost chromatin profiling with CUT&Tag**. *Nat Protoc* (2020.0) **15** 3264-3283. DOI: 10.1038/s41596-020-0373-x
76. 76.Henikoff S, Henikoff JG, Kaya-Okur HS, Ahmad K. Efficient chromatin accessibility mapping in situ by nucleosome-tethered tagmentation. eLife. 2020;9:e63274. 10.7554/eLife.63274.
77. Langmead B, Salzberg SL. **Fast gapped-read alignment with Bowtie 2**. *Nat Methods* (2012.0) **9** 357-359. DOI: 10.1038/nmeth.1923
78. Meers MP, Tenenbaum D, Henikoff S. **Peak calling by sparse enrichment analysis for CUT&RUN chromatin profiling**. *Epigenetics Chromatin* (2019.0) **12** 42. DOI: 10.1186/s13072-019-0287-4
79. Yu G, Wang LG, He QY. **ChIPseeker: an R/Bioconductor package for ChIP peak annotation, comparison and visualization**. *Bioinformatics* (2015.0) **31** 2382-2383. DOI: 10.1093/bioinformatics/btv145
80. Robinson MD, McCarthy DJ, Smyth GK. **edgeR: a Bioconductor package for differential expression analysis of digital gene expression data**. *Bioinformatics* (2010.0) **26** 139-140. DOI: 10.1093/bioinformatics/btp616
|
---
title: Alpha-2-macroglobulin is involved in the occurrence of early-onset pre-eclampsia
via its negative impact on uterine spiral artery remodeling and placental angiogenesis
authors:
- Jingyun Wang
- Ping Zhang
- Mengyuan Liu
- Zhengrui Huang
- Xiaofeng Yang
- Yuzhen Ding
- Jia Liu
- Xin Cheng
- Shujie Xu
- Meiyao He
- Fengxiang Zhang
- Guang Wang
- Ruiman Li
- Xuesong Yang
journal: BMC Medicine
year: 2023
pmcid: PMC9999529
doi: 10.1186/s12916-023-02807-9
license: CC BY 4.0
---
# Alpha-2-macroglobulin is involved in the occurrence of early-onset pre-eclampsia via its negative impact on uterine spiral artery remodeling and placental angiogenesis
## Abstract
### Background
Pre-eclampsia (PE) is one of the leading causes of maternal and fetal morbidity/mortality during pregnancy, and alpha-2-macroglobulin (A2M) is associated with inflammatory signaling; however, the pathophysiological mechanism by which A2M is involved in PE development is not yet understood.
### Methods
Human placenta samples, serum, and corresponding clinical data of the participants were collected to study the pathophysiologic mechanism underlying PE. Pregnant Sprague–Dawley rats were intravenously injected with an adenovirus vector carrying A2M via the tail vein on gestational day (GD) 8.5. Human umbilical artery smooth muscle cells (HUASMCs), human umbilical vein endothelial cells (HUVECs), and HTR-8/SVneo cells were transfected with A2M-expressing adenovirus vectors.
### Results
In this study, we demonstrated that A2M levels were significantly increased in PE patient serum, uterine spiral arteries, and feto-placental vasculature. The A2M-overexpression rat model closely mimicked the characteristics of PE (i.e., hypertension in mid-to-late gestation, histological and ultrastructural signs of renal damage, proteinuria, and fetal growth restriction). Compared to the normal group, A2M overexpression significantly enhanced uterine artery vascular resistance and impaired uterine spiral artery remodeling in both pregnant women with early-onset PE and in pregnant rats. We found that A2M overexpression was positively associated with HUASMC proliferation and negatively correlated with cell apoptosis. In addition, the results demonstrated that transforming growth factor beta 1 (TGFβ1) signaling regulated the effects of A2M on vascular muscle cell proliferation described above. Meanwhile, A2M overexpression regressed rat placental vascularization and reduced the expression of angiogenesis-related genes. In addition, A2M overexpression reduced HUVEC migration, filopodia number/length, and tube formation. Furthermore, HIF-1α expression was positively related to A2M, and the secretion of sFLT-1 and PIGF of placental origin was closely related to PE during pregnancy or A2M overexpression in rats.
### Conclusions
Our data showed that gestational A2M overexpression can be considered a contributing factor leading to PE, causing detective uterine spiral artery remodeling and aberrant placental vascularization.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12916-023-02807-9.
## Background
As a multifactorial disorder, PE is a serious blood pressure condition associated with excessive proteinuria or organ dysfunction in pregnant women after the 20th week of pregnancy; its pathological mechanism is poorly understood to date [1, 2]. Notably, PE is the dominant factor causing maternal and fetal morbidity and mortality worldwide [3, 4], and subsequent risks of PE are particularly severe, endangering the lives of both mothers and their babies [4]. Persistent hypoxia in the placenta is a major factor in the pathogenesis of PE. Hypoxia-induced oxidative stress in PE can cause an imbalance between proangiogenic factors (such as vascular endothelial growth factor (VEGF) and placental growth factor [PlGF]) and antiangiogenic factors (such as soluble fms-like tyrosine kinase-1 [sFlt1]), thereby impacting on placental vascular function [5]. In addition, the renin–angiotensin–aldosterone system (RAAS) plays a key role in the high blood pressure of early-onset PE, i.e., there is an increased sensitivity to circulating angiotensin II (ANG II) [6]. Furthermore, compared with normal pregnancies, the serum levels and activities of most components of the RAAS in PE are decreased, especially the levels of AT1R autoantibodies (AT1R-AA), Ang I, Ang II, and Ang-[1-7] [7, 8]. These biologically active molecules enhance vasoconstriction of the resistance arteries in the placenta, leading to increased hypoxia in PE.
Growing evidence indicates that the development of PE may be most likely due to the inadequate remodeling of the uterine spiral artery during the process of vascular remodeling because of dysfunctional trophoblasts [9]. Proper uterine spiral artery remodeling is accomplished when the disruption of the maternal uterine muscular elastic wall facilitates the invasion of extravillous trophoblasts (EVTs), i.e., when epithelial and vascular smooth muscle cells (VSMCs) are replaced by infiltrating EVTs [10]. Transforming growth factor β1 (TGFβ1) derived from uterine natural killer (uNK) cells regulates vascular smooth muscle cell apoptosis and migration, which ensures proper remodeling of the spiral artery [11]. If this process does not proceed correctly, inadequate remodeling of the uterine spiral artery restricts the blood supply to the placenta and subsequently leads to placental hypoxia, which will increase the possibility of PE occurrence [12].
Defective placental angiogenesis is thought to be involved in the pathogenesis of PE [13]. For example, suppressing placental angiogenesis with suramin (a VEGF inhibitor), pregnancy-associated plasma protein-A2 (PAPP-A2), or induction of oxidative stress during pregnancy might lead to maternal hypertension, placental dysfunction, and fetal growth retardation, i.e., the diagnostic index of PE [14–16]. Furthermore, the activate autophagy by protein kinase Cβ (PKCβ) downregulation leads to impaired placental angiogenesis and ultimately induces PE-like symptoms in mice [17]. Taken together, these data suggest the importance of improper placental angiogenesis in the pathogenesis of PE.
A2M is a serum panprotease inhibitor that plays a role in a unique “trapping” mechanism [18–20]. In addition, A2M exerts anti-infective and anti-inflammatory effects through trapping and inhibiting proteases released by neutrophils [21]. A2M is mainly synthesized in the liver and expressed in the brain, heart, and reproductive tract, and in these tissues, A2M is involved in many physiological functions and pathological changes [18, 22, 23]. A variety of important angiogenetic factors are inactivated by binding to A2M, such as basic fibroblast growth factor (bFGF), VEGF, and PlGF [22]. In particular, A2M can coordinately regulate the uterine vasculature during pregnancy [22], which prompted us to investigate the biological functions of A2M in the context of PE. We hypothesize that the increased A2M expression might play a negative role in uterine spiral artery remodeling and placental angiogenesis during pregnancy, thereby contributing to the development of PE.
## Human tissue collection
Placental tissues were obtained from 53 healthy pregnancies and 52 pregnant women with early-onset PE (diagnosed before 34 gestational weeks). Serum samples were collected from pregnant women in early pregnancy (22 healthy pregnancies in the normal group and 18 early-onset PE patients in the PE group), middle pregnancy (23 healthy pregnancies and 12 early-onset PE patients), late pregnancy (35 healthy pregnancies and 30 early-onset PE patients), and a week after delivery (18 healthy pregnancies and 21 early-onset PE patients). These pregnant women, including healthy pregnancies and early-onset PE pregnancies, were hospitalized in the Department of Gynecology and Obstetrics of Overseas Hospital, Jinan University, China, from 1 January 2018 to 23 May 2021. The PE diagnosis was based on the criteria issued by the International Society for the Study of Hypertension in Pregnancy (ISSHP) in 2018 [24]. The inclusion and exclusion criteria for tissue collection are listed in Table 1.Table 1Inclusion and exclusion criteriaInclusion criteria• Asian• Age: 19–40 years• Singleton pregnancy• Within the third trimester of pregnancy• Meeting the diagnostic criteria for early-onset pre-eclampsia• Free of chronic diseases (kidney disease diabetes, hypertension, or other chronic diseases), autoimmune disorders, infections, or hepatitis in preconception• Obtain patient informed consentExclusion criteria• Multiple gestations• Fetal congenital malformation• Fetal chromosomal disorders• History of chronic diseases• Complicated with serious internal or surgery-related disease The placental villi and decidual tissues were immediately collected after delivery, washed in ice-cold phosphate-buffered saline (PBS) 2 or 3 times to remove the blood, and fixed with liquid nitrogen or $4\%$ paraformaldehyde for further study. The placental villous tissues in the first trimester of pregnancy were obtained from cases of uncomplicated pregnancies and fixed with $4\%$ paraformaldehyde. In addition, peripheral venous blood was collected from pregnant women in the normal group and the early-onset PE group at three sampled times of gestation: 11–13+6 weeks of gestation, 14–27+6 weeks of gestation, and the first day of the latest hospital admission (usually occurring within a week before delivery). Postnatal blood was collected from the third day after delivery, and umbilical cord blood was collected immediately after delivery. Maternal and umbilical cord blood were collected into EDTA vacuum blood collection tubes and centrifuged (3000×g, 4 °C, 15 min), and the supernatant (i.e., plasma) was extracted and stored at −80 °C for further study.
This study was approved by the Ethics Committee of Overseas Hospital, Jinan University, China (approval number: KY-2021-054) and conducted in accordance with the Declaration of Helsinki. Signed informed consent was obtained from all study participants.
## Animal model
SPF Sprague–Dawley rats (6~8 weeks age, 180~200 g weight) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (SCXK 2012-0001, Beijing, China). Based on a previous report [25], we established an A2M-overexpression rat model via tail vein injection as previously described [26]. Briefly, on gestational day (GD) 8.5, rats (excluding non-pregnant rats) were injected with adenoviruses expressing A2M (OBiO Technology Co., Shanghai, China), and the sequencing results are shown in Additional file 1: Supplementary Result 1. We injected an adenoviral dose of approximately 1–2×109 pfu per animal, and the adenoviruses were dissolved in phosphate-buffered saline (PBS) to a total volume of 400 μl. The treated rats were sacrificed on GD19.5 (corresponding to the third trimester) for further study. The following parameters were assessed: blood pressure, blood flow, and proteinuria. The primary outcome of this study will be hypertension with blood pressure measurement. Secondary outcomes constitute blood flow, proteinuria, and histological analyses to measure the morphology and cell function of the spiral artery and placental vascular. For the rat samples, the pregnant rats were first euthanized to collect placentas and fetuses. According to Resource Equation Approach [27], a total of 20 rats were studied, the rats were randomly divided into two groups ($$n = 10$$ in each group): the control and A2M-overexpression groups (note: the rats used in this experiment came from at least three different modelling batches). Random numbers were generated using the standard = RAND() function in Microsoft Excel. Each rat was euthanized by cervical dislocation after the experiment. Experiments involving animals were performed in accordance with the ARRIVE guidelines. All experimental processes involving animal treatments were conducted in accordance with the procedures of the Ethics Committee for Animal Experimentation, Jinan University (approval number: 20210302-46).
## Blinding
For each animal, at least seven different investigators were involved as follows: J.W. was the only person aware of the group allocation based on the randomization table. P.Z. administered intravenous tail vein injections with the assistance of J.W.. Then, G.W. performed the data analysis with the support of X.C.. Finally, P.Z., G.W., R.L., and X.Y. (also unaware of the group allocation) were responsible for the outcome assessment.
## Measuring blood pressure and Doppler ultrasound evaluation as well as proteinuria
As previously described [28, 29], the blood pressure of conscious pregnant rats was measured using an automated computerized tail-cuff system after five consecutive training periods (Visitech BP2000, Visitech Systems, Inc., USA). The blood flow of the rat uterine artery was measured using a high-resolution ultrasound device (Esaote MyLab30 Gold, Esaote, Genova, Italy) to obtain two-dimensional images. Twenty-four-hour urine samples were collected on GD7.5 and GD19.5 for urine protein analysis.
## Histological analysis
Hematoxylin and eosin (HE) and periodic acid Schiff (PAS) staining were performed as follows. The tissues were fixed in $4\%$ paraformaldehyde and subsequently embedded in paraffin. Then, 4-μm-thick cross-sections were processed and stained with HE or PAS for morphological analysis.
Immunohistochemical and immunofluorescent staining were performed as follows. Human or rat tissue was fixed in $4\%$ paraformaldehyde, dehydrated, embedded in paraffin wax, and serially sectioned at a thickness of 4 μm. The sections were incubated with primary antibodies overnight at 4 °C. Subsequently, the sections were stained with fluorescent secondary antibodies. The nuclei were stained with DAPI (Invitrogen). The sections were imaged using a fluorescence microscope (Olympus BX53, Tokyo, Japan). A minimum of 5 random images from 3 samples were analyzed per group. Immunohistochemical statistical analysis was conducted with the Fromowitz comprehensive scoring method [30]. The details of the antibodies are listed in Additional file 1: Supplementary Table 1.
## Western blotting analysis
Proteins from human and rat tissues, HUASMCs, HTR-8/SVneo cells, and HUVECs in Western blotting experiments were analyzed with at least three replicates as previously described [31]. The details of the antibodies are listed in Additional file 1: Supplementary Table 2.
## Enzyme-linked immunosorbent assay (ELISA)
Whole blood samples were collected from human and rat maternal or umbilical cord blood. The substances to be tested in the sera were measured by UV spectrophotometry using detection kits according to the manufacturer’s instructions (Mbbiology Biological, Jiangsu, China). The kit details are provided in Additional file 1: Supplementary Table 3.
## Transmission electron microscopic analysis
Biopsies from rat kidneys (1 mm3) were fixed for 2–4 h at 4 °C, washed, and stored overnight at 37°C. The fixed samples were then prepared for ultrathin sectioning. After uranium–lead double staining, the samples were incubated at room temperature overnight, and images were collected and analyzed under a transmission electron microscope (HT7700, Hitachi, Japan).
## Flow cytometry
An Annexin V-FITC apoptosis kit (88-8005-72, Thermo Fisher, USA) was used to determine the apoptosis rates of HUASMC and HTR-8/SVneo cells by flow cytometry analysis. The cells were analyzed using a FACS flow cytometer (Becton-Dickinson, San Jose, CA, USA). The acquired data were analyzed using FCS-Express software version 3.0 (De Novo).
Cell cycle analysis was also performed by flow cytometry. Briefly, the cells were collected and washed in PBS, followed by fixation in ethanol ($70\%$). After overnight incubation at −20 °C, the cells were stained with PI and subjected to flow cytometry. Then, the distribution of cells in the G1, S, and G2/M phases of the cell cycle was determined.
## Wound healing assay
A total of 5 × 105 HTR-8/SVneo cells or HUVECs administered vehicle or A2M-expressing adenovirus vectors over 48 h were seeded in 6-well plates and grown to reach confluent monolayers. Then, a 2-μl pipette tip was used to create the scratches. The images of migrated cells were recorded at 0–36 h. The percentage of wound closure was analyzed.
## Transwell invasion/migration assay
Transfected cells (1×105 HTR-8/SVneo cells or HUVECs in 100 μl serum-free medium) were seeded into transwell inserts (8 -μm pores; #3422, Costar, Cambridge, MA, USA), and the rest of the protocol was previously described [32] (note: the cells were serum-starved for 24 h before being harvested from the plates).
## Tube formation assay
HUVECs (5 × 104) were plated on the Matrigel-coated wells of 24-well plates and incubated for 6 h. HUVEC tubes were evaluated under an inverted fluorescence microscope (Nikon TE300, Japan). The length of the formed tubes was analyzed. The experiments were performed in triplicate.
## Data analysis
In all of our experimental studies, each experiment was performed at least in triplicate, and blinded outcome assessment was implemented. The mean coefficients of variation (CVs) for triplicate values were calculated, and a grand mean CV was then determined based on these values. Statistical analysis was performed using the SPSS 23.0 statistical package program. Construction of statistical charts was performed using the GraphPad Prism 8 software package (GraphPad Software, CA, USA). T tests were used to analyze the normally distributed continuous variables, and Mann–Whitney U tests were used to analyze the skewed variables (data were normally distributed). All values are presented as the mean ± SD or median (interquartile range). The rates were compared using Fisher’s exact and Pearson’s chi-square tests, which were employed to establish whether there was any difference between the control and experimental data. $P \leq 0.05$ was considered significant [31, 33, 34]. All animal experiments statistics data and the exact value of n are presented in Additional file 1: Supplementary Table 4.
## A2M was predominantly expressed in the vascular smooth muscle of the spiral artery and feto-placental endothelium, and its expression was significantly elevated in PE patients
In this study, the data of 53 healthy pregnant women and 52 pregnant women with early-onset PE were statistically analyzed to assess their clinical and laboratory characteristics and adverse pregnancy outcomes (Table 2). The results are consistent with the characteristics of PE (Additional file 1: Supplementary Result 2). Using ELISA and immunofluorescent staining, we demonstrated that maternal serum A2M levels in the second and third trimesters were significantly elevated in PE patients compared to healthy subjects, but there was no significant difference in the postnatal period (Fig. 1a). Immunofluorescent staining of A2M and α-SMA showed that A2M was predominantly expressed in the smooth muscles of the spiral artery in the human decidua basalis (Fig. 1b, d), and there was significantly higher expression of α-SMA and A2M in the un-remodeled spiral arteries than in the remodeled spiral arteries (Fig. 1c). In addition, we discovered the more un-remodeled spiral arteries in PE decidua basalis than in normal pregnancy (Fig. 1d, d1). Furthermore, immunofluorescence double staining of the cross-sections of the spiral arteries demonstrated significantly higher expression of α-SMA and A2M in the PE group compared to the normal group (Fig. 1d, d2–d3), which was confirmed by Western blotting data from the human decidua basalis (Fig. 1e, e1–e2).Table 2Clinical and laboratory characteristics of the study populationVariablesNormaln=53PEn=52P-valueGestational age at diagnosis in weeks—30.3±0.5—Maternal age (years)30.3±0.530.9±0.70.562Weight (kg)66.2±1.269.5±1.80.142Height (cm)160.8±0.8155.8±3.20.129BMI (kg/m2)25.6±0.427.6±0.40.019**Systolic blood pressure (mmHg)116.6±1.4163.5±2.1<0.001***Diastolic blood pressure (mmHg)73.5±1.1102.0±1.4<0.001***MAP (mmHg)87.8±1.0123.1±1.5<0.001***Gestational weeks at delivery (week)39.3±0.137.0±0.4<0.001***Proteinuria (g/24 h)—1.0±0.1—Platelet count (×10^9/L)210.0±6.9193.1±7.70.105Birthweight (kg)3.2±0.52.4±0.1<0.001***Umbilical cord length (cm)50.6±1.045.88±1.30.004**Amniotic fluid volume (ml)547.7±39.9467.9±20.50.079Placental weight (g)555.2±8.7466.9±14.0<0.001***Cesarean section rate (n, %)16 (30.2)42 (80.8)<0.001***Maternal adverse outcomes (n, %)1 (1.9)15 (35.7)<0.001*** Placental abruption (n, %)0 (0.0)3 (7.1)0.076 PPH (n, %)1 (1.9)2 (4.8)0.547 Eclampsia (n, %)0 (0.0)0 (0.0)— Incomplete uterine rupture (n, %)0 (0.0)2 (4.8)0.149 Cardiac insufficiency (n, %)0 (0.0)1 (1.9)0.310 Fetal/neonatal adverse outcomes (n, %)4 (7.5)30 (57.7)<0.001*** Preterm birth (<37 weeks) (n, %)0 (0.0)16 (30.8)<0.001*** LBW (n, %)0 (0.0)15 (28.8)<0.001*** Myocardial damage (n, %)1 (1.9)13 (25.0)<0.001*** PFO (n, %)1 (1.9)8 (15.4)0.014* Hypoalbuminemia (n, %)0 (0.0)16 (30.8)<0.001*** *Neonatal anemia* (n, %)0 (0.0)9 (17.3)0.002** *Neonatal hyperbilirubinemia* (n, %)3 (5.7)15 (28.0)0.002**Differences between the groups were compared with the Mann–Whitney U test or chi-square test or Fisher’s exact test; *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ PE, pre-eclampsia; weight, the maternal weight in the first day of the latest hospital admission (usually occur within a week before delivery); BMI, body mass index; MAP, mean arterial pressure; PPH, postpartum hemorrhage; LBW, low-birth-weight infant; PFO, patent foramen ovale. Values are mean ± S.E.MFig. 1Assessing A2M expression in the maternal portion and fetal portion of the placenta. a Determination of the A2M levels in maternal serum obtained from the first, second, and third trimesters of pregnancy and the third day after delivery in the normal and PE groups by ELISA. b, c Representative immunofluorescence staining of A2M and α-SMA in the cross-sections of the spiral artery from the decidua basalis of the first trimester (counterstained with DAPI) (b), and c is the quantitative analysis of the positive expression area of α-SMA and A2M in the total vessel. d, d1–d3 Representative immunofluorescent staining of α-SMA and A2M in the cross-sections of the spiral artery from the decidua basalis of the third trimester (counterstained with DAPI) (d), d1 is the analysis of the ratio of un-remodeled blood vessels per field, and d2–d3 is the quantitative analysis of the positive expression area of α-SMA (d2) and A2M (d3) in the total vessel. e, e1–e2 Western blotting data showing the expression of α-SMA and A2M in the decidua basalis of the third trimester (e), and e1–e2 are the quantitative analysis of the expression of α-SMA (e1) and A2M (e2) in the normal and PE groups. f Determination of the A2M levels in umbilical cord serum in the normal and PE groups by ELISA. g, g1 Western blotting data showing the level of A2M in the villous chorion of the third trimester (g), and g1 is the quantitative analysis of A2M expression in the normal and PE groups. h A2M immunohistochemical analysis of cross-sections of first or third trimester villous chorion. i Representative HE and CD31 immunohistochemical staining on cross-sections of villous chorion of the third trimester from the normal and PE groups. j, j1–j2 VEGF and VEGFR2 immunohistochemical staining on cross-sections of the villous chorion of the third trimester from the normal and PE groups (j), and j1–j2 show the quantitative analysis of VEGF (j1) and VEGFR2 (j2) expression in the two groups. Scale bars = 100 μm in b and d; 20 μm in h–j. *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ Using ELISA and Western blotting, we demonstrated that the A2M levels in umbilical cord serum and villous chorion were significantly elevated in PE patients compared to healthy subjects (Fig. 1f, g, g1), and A2M was specifically expressed in the vascular endothelium of the villous chorion in human placenta, as revealed by immunohistochemistry staining of A2M (Fig. 1h). In addition, fewer blood vessels and smaller vascular capacitances were observed in PE patient placentas than in normal group placentas (Fig. 1i), while we discovered the significantly decreased expression of VEGF and its receptor VEGFR2 in the PE group (Fig. 1j, j1–j2).
## The A2M-overexpression rat model closely mimicked the phenotypes observed in PE patients
The successful establishment of the A2M-overexpression rat model (Fig. 2a) was verified by the data from ELISA (Fig. 2b) and Western blotting (Fig. 2c, d). There were significantly increased A2M levels in the maternal serum and A2M-overexpression rats at gestational day 19.5 (Fig. 2b) and enhanced expression of A2M protein in the A2M-overexpression rat decidua basalis (Fig. 2c, c1) and feto-placenta (Fig. 2d, d1) compared to those in the control (vehicle) rats. Fig. 2Assessment of blood pressure and liver and kidney function indices in the A2M-overexpression rat model. a Schematic illustration of the establishment of a humanized A2M-overexpression pregnant rat model. b Determination of the A2M levels in the sera of pregnant rats during different stages of gestation (GD7.5 and GD19.5) in the control and A2M-overexpression groups by ELISA. c, d, c1–dI Western blotting data showing the level of A2M in the decidua basalis (c) and feto-placenta (d), and c1–d1 are the quantitative analysis of A2M expression in control and A2M-overexpression rats. e–h The trends in the variation in systolic (e, g) or diastolic (f, h) pressure from the control and A2M-overexpressing groups in both non-pregnant (e, f) and pregnant (g, h) rats. i, j Representative TEM images of the kidneys from both groups. k–o Representative HE staining (k–l) and PAS staining (m, n) of cross-sections of rat kidneys from the control and A2M-overexpression groups, and o shows the quantitative analysis of the area of Bowman’s space from both groups. p Urine protein (mg/24 h) levels at GD7.5 and GD19.5 from the control and A2M-overexpression rats. q–u Determination of the levels of BUN (q), CREA (r), UA (s), ALT (t), and AST (u) in rat sera from the control and A2M-overexpression groups. Scale bars = 5 μm in i, j; 50 μm in k–n. Abbreviations: TEM, transmission electron microscope; Enc, endothelial cell; Podo, podocyte; cap, capillary; BUN, blood urea nitrogen; CREA, creatinine; UA, uric acid; ALT, alanine aminotransferase; AST, aspartate aminotransferase. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ To evaluate whether manifestations of the A2M-overexpression rat model were consistent with the clinical manifestations of PE, we first dynamically measured the blood pressure of rats administered either vehicle or A2M-expressing adenovirus vectors. The results showed no changes in blood pressure in non-pregnant rats after A2M administration (Fig. 2e, f), but both the systolic and diastolic pressures were significantly but gradually enhanced in A2M-administered rats within 1 week of A2M administration (Fig. 2g, h). TEM images of the rat glomerulus clearly indicated that A2M overexpression caused ultrastructural damage to the glomerulus, such as edema, collapsed vascular lumen, and endothelial hyperplasia (Fig. 2i, j). HE and PAS staining showed that A2M overexpression indeed caused an increase in inflammatory cell infiltration and even morphological damage in the glomerulus, such as a narrower Bowman’s capsule, compared to the control (Fig. 2k–o). The 24-h urine protein measure was performed using a BCA protein assay, and the results showed a significant increase in the A2M-overexpression rat compared to the control rats at gestational day 19.5 (Fig. 2p). The ELISA results indicated significant increases in the BUN and CREA level but no significant changes in the ALT, AST and UA levels in the A2M-administered rat serum compared to control rat serum (Fig. 2q–u). In addition, maternal A2M overexpression led to fetal growth restriction (e.g., fewer and smaller fetuses, and placenta size) (Additional file 1: Fig. S1).
## A2M overexpression led to elevated vascular resistance and defective uterine spiral artery remodeling
Because A2M is highly expressed in the vascular smooth muscles of pregnant women with early-onset PE, we reasonably hypothesized that PE was caused by notably defective uterine spiral artery remodeling [35], which further led to increased placental vascular resistance [36] (Fig. 3a). Therefore, we assessed a series of vascular resistance indexes [37] in human pregnant women (Fig. 3b–g). The results showed that umbilical cord PI (Fig. 3b) and RI (Fig. 3c) were significantly increased in the second and third trimesters of pregnancy in women with PE compared to healthy women, and there were significant increases in the left uterine artery pulsatility index (Lt ut-PI) (Fig. 3d), the right uterine artery pulsatility index (Rt ut-PI) (Fig. 3e), the left uterine artery resistive index (Lt ut-RI) (Fig. 3f), and the right uterine artery resistive index (Rt ut-RI) (Fig. 3g) in the pregnant women with PE compared to the healthy pregnant women. Moreover, ultrasound evaluation of pregnant rats (Fig. 3h) indicated that A2M overexpression significantly enhanced Lt ut-PI (Fig. 3h1) and Lt ut-RI (Fig. 3h2) compared to the control. Immunohistochemistry displayed enhanced expression of α-SMA in the spiral artery and more un-remodeled spiral arteries in A2M-overexpression rats (Fig. 3i, i1–i2). Similar results were obtained for α-SMA expression by Western blotting (Fig. 3j).Fig. 3Assessing the index of spiral artery remodeling in pregnant women and A2M-overexpression rats. a Schematic illustration of PE establishment due to failure of spiral artery remodeling. b, c Determination of the values of umbilical artery PI (b) and RI (c) in the normal and PE groups during the second and third trimesters of pregnancy. d–g Determination of the levels of Lt ut-PI (d), Rt ut-PI (e), Lt ut-RI (f), and Rt ut-RI (g) in the normal and PE groups during the first, secondary, and third trimesters of pregnancy. h, h1–h2 Representative ultrasonography of rat uterine spiral artery from control and A2M-overexpression rats (h), and h1–h2 show the quantitative analysis of Lt ut-PI (h1) and Lt ut-RI (h2) from the above groups. i, i1–i2 α-SMA immunohistochemical analysis of cross-sections of spiral arteries from the control and A2M-overexpression groups (i). i1 is the quantitative analysis of α-SMA expression, and i2 is the ratio of un-remodeled spiral arteries per field in the two groups. j, j1 Western blotting data showing the level of α-SMA in the decidua basalis (j) and j1 is the quantitative analysis of α-SMA expression in the control and A2M-overexpressing groups. Scale bars = 100 μm in i. Abbreviations: PI, pulsatility index; RI, resistive index; Lt ut, left uterine artery; Rt ut, right uterine artery. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$
## A2M overexpression-induced defective uterine spiral artery remodeling was partially derived from aberrant cell proliferation and apoptosis
We investigated the effect of A2M expression on the proliferation and apoptosis of uterine spiral artery smooth muscle cells by manipulating A2M expression in HUASMCs. The CCK8 assay (Fig. 4a) and Western blotting (Fig. 4b, b1–b2) indicated that cell proliferation was dramatically increased in the A2M overexpression group and decreased in the A2M-downregulated group. Flow cytometry analysis demonstrated that knockdown of A2M expression in HUASMCs did not affect cell cycle progression or apoptosis, but A2M overexpression caused cell cycle arrest at the S phase (Fig. 4c–f) and suppression of cell apoptosis (Fig. 4g–j). Experiments on human samples reinforced the observation as follows. Double immunofluorescence staining of α-SMA and PCNA showed increased numbers of PCNA-positive cells in the decidua basalis of human patients with PE (Fig. 4k, k1), and similar results were obtained for PCNA expression by Western blotting (Fig. 4l). In addition, Western blotting analysis showed that FAS expression was decreased in the decidua basalis of PE patients (Fig. 4m).Fig. 4Assessment of proliferation and apoptosis in human umbilical artery smooth muscle cells (HUASMCs) following manipulation of A2M expression and human decidua basalis. a, b, b1–b2 Determination of HUASMC viability in the control, A2M-overexpression and A2M-downregulated groups after 12-, 24-, 48-, 72-, and 96-h incubation by CCK-8 assay (a). Western blotting data showing the expression of A2M and α-SMA in HUASMCs (b), and b1–b2 show the quantitative analysis of A2M (b1) and α-SMA (b2) expression among the control, A2M-overexpression, and A2M-downregulated groups. c–f Flow cytometry data showing the analysis of the DNA contents in HUASMCs transfected with negative control (control) (c), A2M-silencing vector (A2Msi) (d), or A2M-overexpressing vector (A2M) (e), and f shows the quantitative analysis of the proportion of cells in each phase of the cell cycle in the three groups. g–j Apoptosis of HUASMCs transfected with negative control (control) (g), A2M-silencing vectors (h), or A2M-overexpression vector (i) was determined by flow cytometry using the annexin V-FITC/PI apoptosis assay, and j shows the quantitative analysis of the cell apoptosis rates in the three groups. k, k1 Representative double immunofluorescence staining of α-SMA and PCNA on the cross-sections of the spiral arteries (counterstained with DAPI) of the third trimester from the normal and PE groups, and k1 shows the quantitative analysis of PCNA expression in both groups. l, m Western blotting data showing the expression of PCNA (l) and FAS (m) and the quantitative analysis of the human decidua basalis of the third trimester from the normal and PE groups. Scale bars = 50 μm in k. *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ To determine the effects of EVTs on uterine spiral artery remodeling [38, 39] in the context of A2M overexpression, wound healing and transwell invasion assays were used, and the results demonstrated that A2M overexpression significantly suppressed the migratory and invasive abilities of HTR-8/SVneo cells (Additional file 1: Fig. S2). We also found that trophoblast cell proliferation was decreased and cell apoptosis was enhanced in the context of A2M overexpression (Additional file 1: Fig. S3).
## TGFβ1 played an important role in the overproliferation of vascular smooth muscle cells in the context of A2M overexpression
Western blotting showed that A2M overexpression enhanced the expression of PCNA and p-Smad$\frac{2}{3}$, while A2M downregulation suppressed the expression of both genes in HUASMCs (Fig. 5a, a1–a4); furthermore, the addition of TGFβ1 significantly increased A2M, PCNA, and p-Smad$\frac{2}{3}$ expression (Fig. 5b, b1–b4). Western blotting data demonstrated that TGFβ1 expression was elevated in the decidua basalis of patients with PE in comparison to normal pregnant women (Fig. 5c).Fig. 5Assessment of proliferation and TGFβ signaling in smooth muscle cells following the manipulation of A2M expression and human decidua basalis. a, a1–a4 Western blotting data showing the expression of A2M, p-Smad$\frac{2}{3}$, TGFβ1, and PCNA in HUASMCs transfected with negative control, A2M-overexpresion, or A2M-silencing vectors (a), and a1–a4 shows the quantitative analysis of A2M (a1), p-Smad$\frac{2}{3}$ (a2), TGFβ1 (a3), and PCNA (a4) in the control, A2M-overexpression and A2M-downregulated groups. b, b1–b4 Western blotting data showing the expression of A2M, p-Smad$\frac{2}{3}$, TGFβ1, and PCNA in HUASMCs transfected with negative control, treat with TGFβ1 and A2M-silenced + treated with TGFβ1 (b), and b1–b4 show the quantitative analysis of A2M (b1), p-Smad$\frac{2}{3}$ (b2), TGFβ1 (b3), and PCNA (b4) expressions among the control, treatment with TGFβ1 and A2M-silenced + treated with TGFβ1 groups. c Western blotting data showing the expression of TGFβ1 and the quantitative analysis of TGFβ1 expression in human decidua basalis of the third trimester from the normal and PE groups. d Schematic illustration of A2M overexpression leading to cell proliferation through the TGFβ signaling pathway. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$
## A2M overexpression during pregnancy dramatically restricted feto-placental angiogenesis
PAS staining showed that the ratios of labyrinth size to whole placenta size were smaller in A2M-overexpression rats than in control rats (Fig. 6a, a1). HE staining showed that the area of blood sinusoids was decreased in the labyrinth layer of A2M-overexpression rats compared to the control (Fig. 6a, a2). Immunofluorescence staining revealed the elevated A2M and reduced Caveolin1 expression in the labyrinth of the A2M-overexpression rats compared to control rats (Fig. 6b, b1–b2). These findings were confirmed by Western blotting data from the rat labyrinth (Fig. 6c, d, c1–d1). In addition, Western blotting showed that VEGF and CD31 expression in A2M-overexpression placental labyrinth was significantly higher than that in the control (Fig. 6e, f, e1–f1).Fig. 6Assessment of feto-placental angiogenesis in the A2M-overexpression rat model. a, a1–a2 Representative PAS and HE staining (a) on the cross-sections of rat placenta from the control and A2M-overexpression groups, and a1–a2 are the quantitative analysis of the area of the placental labyrinthine zone (a1) and the area of placental blood sinusoid (a2) from the control and A2M-overexpression groups. b, b1–b2 Representative immunofluorescent staining of A2M and Caveolin1 in the cross-sections of rat placenta from the control and A2M-overexpression groups (b), and b1–b2 are the quantitative analysis of positive areas of A2M and Caveolin1 expression (%). c–f, c1–f1 Western blotting data showing the expression of A2M (c), Caveolin1 (d), VEGF (e), and CD31 (f) in the placental labyrinthine zone from the control and A2M-overexpression groups. c1–f1 shows the quantitative analysis of A2M (c1), Caveolin1 (d1), VEGF (e1), and CD31 (f1) expression from the control and A2M-overexpression groups. Scale bars = 2 mm in the upper panel of a; 100 μm in the lower panel of a; 50 μm in b. *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ Next, wound healing and transwell migration assays were used to assess the migration and invasion capabilities of HUVECs. The results showed that both the wound-closure rate (Fig. 7a, a1) and the number of HUVECs that migrated to the lower side of the membrane (Fig. 7b, b1) significantly decreased in A2M overexpression groups compared to control groups. Filopodia affect cell migration, as presented in Fig. 7c, F-actin staining in HUVECs indicated the reduced length and numbers of filopodia in A2M overexpression HUVECs compared to control (Fig. 1c, c1–c2). Furthermore, A2M overexpression in HUVECs significantly decreased ZO-1 expression (Fig. 7d, d1), and significantly suppressed HUVEC tube formation (Fig. 7e, e1), implying the inhibition of epithelial cell migration capability. Fig. 7Assessing cell migration and tube formation of HUVECs following manipulation of A2M expression. a, a1 Representative images of wound healing assays of HUVECs at 24 h from negative control (control) or A2M-overexpression groups (a), and a1 shows the quantitative analysis of relative cell migration in both groups. b, b1 Representative images of transwell migration assays of HUVECs transfected with either negative control or A2M-overexpression vectors after 24 h of incubation (b), and b1 shows the quantitative analysis of the numbers of migrated cells in both groups. c, c1–c2 Representative fluorescent staining and high magnification images of F-actin on the negative control or A2M-overexpression HUVECs (c), and c1–c2 show the quantitative analysis of the numbers (c1) and length (c2) of filopodia of each cell in both groups. d, d1 Representative immunofluorescent staining of ZO-1 on the negative control or A2M-overexpression HUVECs (d), and d1 shows the quantitative analysis of relative ZO-1 fluorescence intensity (% of control) in both groups. e, e1 Representative images of tube formation assays of HUVECs after 6- and 24-h incubation from negative control or A2M-overexpression groups, and e1 shows the quantitative analysis of total tube length (% of control) at 6-h incubation in both groups. Scale bars = 200 μm in a and e; 50 μm in b and d; 20 μm in c. **$P \leq 0.01$, ***$P \leq 0.001$
## Placental ischemia/hypoxia derived from defective spiral artery remodeling and aberrant placental angiogenesis promoted sFLT-1 and suppressed PIGF secretion
Western blotting demonstrated high HIF-1α expression in A2M-overexpression HUASMCs but no change in HIF-1α expression when A2M was downregulated (Fig. 8b–b1). These results suggested that defective spiral artery remodeling as well as aberrant placental angiogenesis causes ischemia, which further induces the high expression of HIF-1α. Notably, we found that the sFLT-1 level was significantly higher and the PIGF level was significantly reduced in the PE group compared to that in the normal group during the different stages of pregnancy (Fig. 8c, d). There was a negative correlation between PIGF and A2M (R = −0.768, $P \leq 0.001$) and a positive correlation between sFLT-1 and A2M ($R = 0.659$, $P \leq 0.001$) in maternal plasma of the pre-eclampsia women (Additional file 1: Fig. S4). In the A2M-overexpression rat model, sFLT-1 levels increased (Fig. 8e), and PIGF levels decreased (Fig. 8f), significantly in the sera of A2M-overexpression rats compared to the control rats at 19.5 days of gestation. Fig. 8Measurement of HIF-1α and sFIt-1/PIGF levels in maternal serum and A2M-manipulated HUASMCs. a Schematic illustration of inappropriate spiral artery remodeling and abnormal placental angiogenesis in the presence of A2M overexpression. b, b1 Western blotting data showing the expression of HIF-1α in the control, A2M-overexpression, and A2M-downregulated groups (b), and b1 shows the quantitative analysis of HIF-1α expression in the three groups. c, d ELISA data showing the sFLT-1 (c) and PIGF (d) levels in human maternal serum obtained from the first, second, and third trimesters of pregnancy in the normal and PE groups. e, f ELISA data showing the sFLT-1 (e) and PIGF (f) levels in maternal rat serum at GD7.5 and GD19.5 in the control and A2M-overexpression groups. * $P \leq 0.05$, ***$P \leq 0.001$
## Discussion
A previous study demonstrated that mild systemic inflammation actually occurred in healthy pregnant women, and the status worsened in PE [40]. In the present study, we reported that an increased level of A2M was involved in the occurrence of early-onset pre-eclampsia via its negative impact on uterine spiral artery remodeling and placental angiogenesis (Fig. 9). Because A2M can reduce endogenous/exogenous inflammatory injury, it has been used in a variety of managements of orthopedic pains, such as subacromial bursitis, lateral epicondylitis, and *Achilles tendonitis* [41]. A2M is also used to ascertain inflammation status in degenerative, immune, digestive, and urinary system diseases [42–45]. In this study, we found an imbalance between pro-inflammatory and anti-inflammatory cytokines in pregnancy with PE (Additional file 1: Fig. S5), implying the possibility that A2M is involved in developing PE during pregnancy. As a unique proteinase inhibitor, A2M does not completely remove pro-inflammatory cytokines in PE, while elevated A2M can be responsible for the PE-like phenotype, suggesting that A2M may play a double-edged sword role in the development of PE [46].Fig. 9A proposed model that shows the underlying assumption for the role of A2M in the development of PE. The proposed mechanism by which A2M is involved in the occurrence of PE via its negative impact on uterine spiral artery remodeling and placental angiogenesis Both uterine spiral arterial modification and placental angiogenesis are crucial events that provide sufficient blood supply to fully perfuse the placenta and thus provide the demands of the growing fetus during pregnancy [47, 48]. Therefore, if the modifications are interrupted for any reason, the consequence would be the inadequate modification of spiral arteries and aberrant placental angiogenesis, which could greatly increase the risk of the PE scenario [49]. Hence, in this study, we investigated the possible effects of A2M overexpression on both events in developing PE during pregnancy.
Our clinical pregnancy cohort study showed that A2M was predominantly expressed in the spiral artery from the third trimester of pregnancy, which is similar to the expression of α-SMA; the A2M levels in the sera of pregnant women with early-onset PE were significantly increased in the secondary and third trimesters of pregnancy, and similar results were observed in the human uterine decidua basalis (Fig. 1a–e), which indicated the occurrence of inadequate uterine spiral artery remodeling. In other words, A2M overexpression might be closely associated with the pathophysiology of PE by negatively affecting uterine spiral artery remodeling. We speculated that A2M inhibited related proteases, cytokines, and growth factors [50], which would impact the survival and physiological functions of smooth muscle cells during spiral artery remodeling. In addition, we found that the overexpression of A2M in the placental vascular bed dramatically restricted placental angiogenesis (Fig. 1f–j); thus, abnormal vascularization of placental villi is obviously a key factor that cannot be neglected. It should be noted that A2M was reported to be involved in atherosclerosis by facilitating the lipogenesis of vascular smooth muscle cells [51], suggesting a possible pathological mechanism in PE, i.e., a similar vascular lesion in pregnancy.
Overexpression of A2M was established in non-pregnant and pregnant rats (Fig. 2) to investigate the possibility of A2M involvement in PE. The different blood pressure responses to A2M overexpression between non-pregnant and pregnant rats notably indicated that the increased blood pressure induced by A2M overexpression was associated with pregnancy (Fig. 2e–h). Furthermore, the following phenotypes were observed: morphological changes in rat kidneys and proteinuria in the rats that undoubtedly contributed to the onset of PE in A2M-overexpression rats (Fig. 2i–u) and intrauterine growth restriction and poor placentation (Additional file 1: Fig. S1) that were extremely coincident with the diagnosis indexes of PE [52, 53]. Additionally, A2M overexpression also greatly suppressed placental vascularization (Fig. 6). All of these data prompted us to further explore the causal relationship between A2M overexpression and PE.
Regarding uterine spiral artery remodeling [54], we found that A2M overexpression promoted HUASMC cell proliferation and inhibited HUASMC cell apoptosis (Fig. 4a–j), implying that the normal replacement of uterine spiral endothelial cells and smooth muscle cells was restricted in the context of A2M overexpression. In other words, A2M overexpression might prevent the cascade regulation from the normally progressive breakdown of the endothelial and smooth muscle cells in uterine spiral arteries, thereby increasing the risk of developing PE. On the other hand, the features of trophoblast cells overexpressed by the A2M gene were assessed because the migration and invasion of trophoblast cells play very important roles in uterine spiral artery remodeling [55, 56]. The experimental results revealed that both migratory and invasive abilities, as well as apoptosis and proliferation of trophoblast cells, were dramatically affected by elevated A2M expression (Additional file 1: Fig. S2-S3), suggesting that EVTs of fetal origin could be influenced by overloaded A2M through an unclear mechanism. To further address the abovementioned phenotypes, we assessed TGFβ signaling following A2M gene manipulation since the TGFβ superfamily is known to regulate vascular endothelial and smooth muscle cell responses during vessel remodeling under both physiological and pathological conditions [57]. As expected, our experimental evidence (Fig. 5a, b) clearly showed a causal relationship between A2M gene expression and TGFβ signaling activation, i.e., TGFβ signaling might be responsible for the A2M overexpression-induced cell responses of endothelial and smooth muscles in the uterine spiral artery described above, which then results in an inappropriate vascular remodeling. This finding is well established and confirmed by the fact that TGFβ1 was highly expressed in the uterine spiral artery smooth muscles of pregnant women with early-onset PE (Fig. 5c).
On the other hand, A2M overexpression affected placental vascularization, such as the coarctation of the placental labyrinth and the reduced expression of specific endothelial markers (Caveolin1, CD31) and VEGF (Fig. 6). Angiogenesis is involved in cell proliferation, migration, adhesion, and tube formation [58]. In this study, A2M overexpression clearly inhibited HUVEC migration, probably by inhibiting the formation of endothelial filopodia and cell–cell junctions, as well as tube formation (Fig. 7a–e). These results revealed that A2M overexpression brought about the obstacle in placental vascularization, which would provoke or aggravate the occurrence of PE.
The initiating event in early-onset PE is usually considered to be placental ischemia and hypoxia, which in turn cause the release of various factors of placental origin and eventually affect maternal blood pressure during pregnancy [59–61]. A2M-induced excessive proliferation of smooth muscle cells aggravated inappropriate vascular remodeling, as well as aberrant placental vascularization, which may be crucial factors for placental ischemia and hypoxia. Notably, ischemia and anoxia were proven to be closely associated with the level of A2M gene expression (Fig. 8b). Similarly, we observed a significant increase in sFLT-1 levels and a decrease in PIGF levels in PE patient serum in the second and/or third trimesters of pregnancy (Fig. 8c, d), and sFLT-1 and PIGF levels were closely associated with the level of A2M in maternal plasma of the pre-eclampsia women (Additional file 1: Fig. S4). Furthermore, a similar trend of change in the levels of sFLT-1 and PIGF in the sera of pregnant rats was clearly observed in A2M-overexpression rats (Fig. 8e, f). These results suggest that A2M overexpression participates in inducing the release of these factors of placental origin in the context of exacerbated placental ischemia/hypoxia.
The renin–angiotensin–aldosterone system (RAAS) plays a vitally important role in maintaining adequate uteroplacental circulation in normal pregnancy, as well as the development of PE [62, 63]. In addition, angiotensin II type 1 receptor agonistic autoantibody (AT1-AAs) was first discovered in women with PE and can activate the angiotensin II type 1 receptor in response to placental ischemia, thereby increasing vasoconstriction of the placental vasculature [64]. In this study, many components of the RAAS and AT1-AAs showed abnormal expressions in PE (Additional file 1: Fig. S6-S8). Therefore, dysregulation of the RAAS might act as a subsequent mechanism of PE in the context of A2M overexpression.
## Conclusions
In summary, our data clearly showed how A2M was involved in the pathophysiology of PE. Briefly, A2M is predominantly expressed in the vascular smooth muscle of the spiral artery and feto-placental vasculature, and its level is elevated under the influence of activated TGFβ1 in the context of PE; the high expression in turn causes the excessive proliferation and reduced apoptosis of vascular smooth muscle cells in the uterine spiral artery as well as insufficient trophoblast migration and invasion (i.e., inappropriate uterine spiral artery remodeling). Meanwhile, the overexpressed A2M in placental villi also greatly restricts placental angiogenesis. The two comprehensive results exacerbate placental ischemia/hypoxia and alter the release of sFLT-1 and PIGF from the placenta, which lead to the occurrence of maternal hypertension, proteinuria, and fetal growth restriction, i.e., PE. In addition, this work has several limitations, such as the lack of precise molecular mechanisms underlying A2M involvement in PE progression, small numbers of PE and control samples, and the lack of prospective studies. Finally, it is important to design more integrated experiments to completely address the pathophysiological mechanism underlying PE. If so, we can expect that A2M will become a potential biomarker and therapeutic target for PE in the future.
## Supplementary Information
Additional file 1: Table S1. Antibodies for immunohistochemistry. Table S2. Antibodies for Western blotting. Table S3. Details of Elisa kits. Table S4. Animal experiments statistics data. Supplementary Result 1. A2M sequencing result. Supplementary Result 2. The clinical and laboratory characteristics and adverse pregnancy outcomes of pregnant women enrolled in this study. Figure S1. Assessment of placental and fetal development in the A2M-overexpression rat model. Figure S2. Determining HTR-8/SVneo cell migration and cell viability following A2M upregulation. Figure S3. Determining HTR-8/SVneo cell proliferation and apoptosis following A2M upregulation. Figure S4. Correlation between PlGF, sFLT-1 and A2M levels in maternal plasma of the preeclampsia women. Figure S5. Determining the serum and placental levels of human inflammatory cytokins and NF-κB. Figure S6. Determining key components of the RAAS system in human serum. Figure S7. Determining key components of the RAAS system in rat serum in the presence of high levels of A2M. Figure S8. Schematic illustration of the changes in key components of the RAAS system in the presence of high A2M levels. Additional file 2. Western blot source data.
## References
1. Chappell LC, Cluver CA, Kingdom J, Tong S. **Pre-eclampsia**. *Lancet* (2021.0) **398** 341-354. DOI: 10.1016/S0140-6736(20)32335-7
2. Magee LA, Nicolaides KH, von Dadelszen P. **Preeclampsia**. *N Engl J Med* (2022.0) **386** 1817-1832. DOI: 10.1056/NEJMra2109523
3. Ghulmiyyah L, Sibai B. **Maternal mortality from preeclampsia/eclampsia**. *Semin Perinatol* (2012.0) **36** 56-59. DOI: 10.1053/j.semperi.2011.09.011
4. Duley L. **Maternal mortality associated with hypertensive disorders of pregnancy in Africa, Asia, Latin America and the Caribbean**. *Br J Obstet Gynaecol* (1992.0) **99** 547-553. DOI: 10.1111/j.1471-0528.1992.tb13818.x
5. Stepan H, Hund M, Andraczek T. **Combining biomarkers to predict pregnancy complications and redefine preeclampsia: the angiogenic-placental syndrome**. *Hypertension* (2020.0) **75** 918-926. DOI: 10.1161/HYPERTENSIONAHA.119.13763
6. Gathiram P, Moodley J. **The role of the renin-angiotensin-aldosterone system in preeclampsia: a review**. *Curr Hypertens Rep* (2020.0) **22** 89. DOI: 10.1007/s11906-020-01098-2
7. Menkhorst E, Zhou W, Santos L, Zhang JG, St-Pierre Y, Young MJ, Dimitriadis E. **Galectin-7 dysregulates renin-angiotensin-aldosterone and NADPH oxide synthase pathways in preeclampsia**. *Pregnancy Hypertens* (2022.0) **30** 130-136. DOI: 10.1016/j.preghy.2022.09.008
8. Leal CRV, Costa LB, Ferreira GC, Ferreira AM, Reis FM, Simoes ESAC. **Renin-angiotensin system in normal pregnancy and in preeclampsia: a comprehensive review**. *Pregnancy Hypertens* (2022.0) **28** 15-20. DOI: 10.1016/j.preghy.2022.01.011
9. Peng M, Yu L, Ding YL, Zhou CJ. **Trophoblast cells invaing the placenta bed and change of spiral arteries and microvessels in pre-eclampsia**. *Zhong Nan Da Xue Bao Yi Xue Ban* (2008.0) **33** 121-129
10. Bulmer JN, Innes BA, Levey J, Robson SC, Lash GE. **The role of vascular smooth muscle cell apoptosis and migration during uterine spiral artery remodeling in normal human pregnancy**. *FASEB J* (2012.0) **26** 2975-2985. DOI: 10.1096/fj.12-203679
11. Liu W, Luo M, Zou L, Liu X, Wang R, Tao H, Wu D, Zhang W, Luo Q, Zhao Y. **uNK cell-derived TGF-β1 regulates the long noncoding RNA MEG3 to control vascular smooth muscle cell migration and apoptosis in spiral artery remodeling**. *J Cell Biochem* (2019.0) **120** 15997-16007. DOI: 10.1002/jcb.28878
12. Li H, Ohta H, Tahara Y, Nakamura S, Taguchi K, Nakagawa M, Oishi Y, Goto Y-I, Wada K, Kaga M. **Artificial oxygen carriers rescue placental hypoxia and improve fetal development in the rat pre-eclampsia model**. *Sci Rep* (2015.0) **5** 1-9
13. Levine RJ, Lam C, Qian C, Yu KF, Maynard SE, Sachs BP, Sibai BM, Epstein FH, Romero R, Thadhani R. **Soluble endoglin and other circulating antiangiogenic factors in preeclampsia**. *N Engl J Med* (2006.0) **355** 992-1005. DOI: 10.1056/NEJMoa055352
14. Carlström M, Wentzel P, Skøtt O, Persson AEG, Eriksson UJ. **Angiogenesis inhibition causes hypertension and placental dysfunction in a rat model of preeclampsia**. *J Hypertens* (2009.0) **27** 829-837. DOI: 10.1097/HJH.0b013e328324f8ce
15. Nezu M, Souma T, Yu L, Sekine H, Takahashi N, Wei AZ-S, Ito S, Fukamizu A, Zsengeller ZK, Nakamura T. **Nrf2 inactivation enhances placental angiogenesis in a preeclampsia mouse model and improves maternal and fetal outcomes**. *Sci Signal* (2017.0) **10** eaam5711. DOI: 10.1126/scisignal.aam5711
16. Chen X, Chen K, Feng Y, Ren C, Li W, Xiao J, Fan L, Beejadhursing R, Xi L, Chen S. **The potential role of pregnancy-associated plasma protein-A2 in angiogenesis and development of preeclampsia**. *Hypertens Res* (2019.0) **42** 970-980. DOI: 10.1038/s41440-019-0224-8
17. Zhao H, Gong L, Wu S, Jing T, Xiao X, Cui Y, Xu H, Lu H, Tang Y, Zhang J. **The inhibition of protein kinase c β contributes to the pathogenesis of preeclampsia by activating autophagy**. *EBioMedicine* (2020.0) **56** 102813. DOI: 10.1016/j.ebiom.2020.102813
18. Blacker D, Wilcox MA, Laird NM, Rodes L, Horvath SM, Go RC, Perry R, Watson B, Bassett SS, McInnis MG. **Alpha-2 macroglobulin is genetically associated with Alzheimer disease**. *Nat Genet* (1998.0) **19** 357-360. DOI: 10.1038/1243
19. Borth W. **α2 Macroglobulin, a multifunctional binding protein with targeting characteristics**. *FASEB J* (1992.0) **6** 3345-3353. DOI: 10.1096/fasebj.6.15.1281457
20. Strickland DK, Ashcom J, Williams S, Burgess W, Migliorini M, Argraves WS. **Sequence identity between the alpha 2-macroglobulin receptor and low density lipoprotein receptor-related protein suggests that this molecule is a multifunctional receptor**. *J Biol Chem* (1990.0) **265** 17401-17404. DOI: 10.1016/S0021-9258(18)38172-9
21. Chen X, Kong X, Zhang Z, Chen W, Chen J, Li H, Cao W, Ge Y, Fang S. **Alpha-2-macroglobulin as a radioprotective agent: a review**. *Chin J Cancer Res* (2014.0) **26** 611. PMID: 25400428
22. Tayade C, Esadeg S, Fang Y, Croy B. **Functions of alpha 2 macroglobulins in pregnancy**. *Mol Cell Endocrinol* (2005.0) **245** 60-66. DOI: 10.1016/j.mce.2005.10.004
23. Umans L, Serneels L, Overbergh L, Lorent K, Van Leuven F, Van den Berghe H. **Targeted inactivation of the mouse α2-macroglobulin gene**. *J Biol Chem* (1995.0) **270** 19778-19785. DOI: 10.1074/jbc.270.34.19778
24. Brown MA, Magee LA, Kenny LC, Karumanchi SA, McCarthy FP, Saito S, Hall DR, Warren CE, Adoyi G, Ishaku S. **Hypertensive disorders of pregnancy: ISSHP classification, diagnosis, and management recommendations for international practice**. *Hypertension* (2018.0) **72** 24-43. DOI: 10.1161/HYPERTENSIONAHA.117.10803
25. Bewig B, Schmidt W. **Accelerated titering of adenoviruses**. *BioTechniques* (2000.0) **28** 870-873. DOI: 10.2144/00285bm08
26. Karumanchi SA, Stillman IE. **In vivo rat model of preeclampsia**. *Methods Mol Med* (2006.0) **122** 393-399. PMID: 16511996
27. Festing MF, Altman DG. **Guidelines for the design and statistical analysis of experiments using laboratory animals**. *ILAR J* (2002.0) **43** 244-258. DOI: 10.1093/ilar.43.4.244
28. Douglas G, Bendall JK, Crabtree MJ, Tatham AL, Carter EE, Hale AB, Channon KM. **Endothelial-specific Nox2 overexpression increases vascular superoxide and macrophage recruitment in ApoE(-)/(-) mice**. *Cardiovasc Res* (2012.0) **94** 20-29. DOI: 10.1093/cvr/cvs026
29. Freitas LA, Mota GL, Silva HVR, Silva LDM. **Two-dimensional sonographic and Doppler changes in the uteri of bitches according to breed, estrus cycle phase, parity, and fertility**. *Theriogenology* (2017.0) **95** 171-177. DOI: 10.1016/j.theriogenology.2017.03.012
30. Fromowitz FB, Viola MV, Chao S, Oravez S, Mishriki Y, Finkel G, Grimson R, Lundy J. **ras p21 expression in the progression of breast cancer**. *Hum Pathol* (1987.0) **18** 1268-1275. DOI: 10.1016/S0046-8177(87)80412-4
31. Xu JJ, Wang G, Luo X, Wang LJ, Bao Y, Yang X. **Role of nuclear factor-kappaB pathway in the transition of mouse secondary follicles to antral follicles**. *J Cell Physiol* (2019.0) **234** 22565-22580. DOI: 10.1002/jcp.28822
32. Pijuan J, Barcelo C, Moreno DF, Maiques O, Siso P, Marti RM, Macia A, Panosa A. **In vitro cell migration, invasion, and adhesion assays: from cell imaging to data analysis**. *Front Cell Dev Biol* (2019.0) **7** 107. DOI: 10.3389/fcell.2019.00107
33. Wang G, Liang J, Gao LR, Si ZP, Zhang XT, Liang G, Yan Y, Li K, Cheng X, Bao Y. **Baicalin administration attenuates hyperglycemia-induced malformation of cardiovascular system**. *Cell Death Dis* (2018.0) **9** 234. DOI: 10.1038/s41419-018-0318-2
34. Song JH, Wang CJ, Long DL, Li ZG, You LS, Brand-Saberi B, Wang G, Yang XS. **Dysbacteriosis-induced LPS elevation disturbs the development of muscle progenitor cells by interfering with retinoic acid signaling**. *FASEB J* (2020.0) **34** 6837-6853. DOI: 10.1096/fj.201902965R
35. Albrecht ED, Pepe GJ. **Regulation of uterine spiral artery remodeling: a review**. *Reprod Sci* (2020.0) **27** 1932-1942. DOI: 10.1007/s43032-020-00212-8
36. Myatt L. **Control of vascular resistance in the human placenta**. *Placenta* (1992.0) **13** 329-341. DOI: 10.1016/0143-4004(92)90057-Z
37. Özkan MB, Ozyazici E, Emiroglu B, Özkara E. **Can we measure the spiral and uterine artery blood flow by real-time sonography and Doppler indices to predict spontaneous miscarriage in a normal-risk population?**. *Aust J Ultrasound Med* (2015.0) **18** 60-66. DOI: 10.1002/j.2205-0140.2015.tb00043.x
38. Moser G, Weiss G, Sundl M, Gauster M, Siwetz M, Lang-Olip I, Huppertz B. **Extravillous trophoblasts invade more than uterine arteries: evidence for the invasion of uterine veins**. *Histochem Cell Biol* (2017.0) **147** 353-366. DOI: 10.1007/s00418-016-1509-5
39. McNally R, Alqudah A, Obradovic D, McClements L. **Elucidating the pathogenesis of pre-eclampsia using in vitro models of spiral uterine artery remodelling**. *Curr Hypertens Rep* (2017.0) **19** 1-13. DOI: 10.1007/s11906-017-0786-2
40. Redman CW, Sargent IL. **Immunology of pre-eclampsia**. *Am J Reprod Immunol* (2010.0) **63** 534-543. DOI: 10.1111/j.1600-0897.2010.00831.x
41. Cuéllar JM, Cuéllar VG, Scuderi GJ. **α2-Macroglobulin: autologous protease inhibition technology**. *Phys Med Rehabil Clin* (2016.0) **27** 909-918. DOI: 10.1016/j.pmr.2016.06.008
42. Varma V, Varma S, An Y, Hohman T, Seddighi S, Casanova R, Beri A, Dammer E, Seyfried N, Pletnikova O. **Alpha-2 macroglobulin in Alzheimer’s disease: a marker of neuronal injury through the RCAN1 pathway**. *Mol Psychiatry* (2017.0) **22** 13-23. DOI: 10.1038/mp.2016.206
43. 43.Naseraldeen N, Michelis R, Barhoum M, Chezar J, Tadmor T, Aviv A, Shvidel L, Litmanovich A, Shehadeh M, Stemer G, Shaoul E, Braester A. The Role of Alpha 2 Macroglobulin in IgG-Aggregation and Chronic Activation of the Complement System in Patients With Chronic Lymphocytic Leukemia. Front Immunol. 2021;11:603569. 10.3389/fimmu.2020.603569. PMID: 33643290; PMCID: PMC7905172.
44. Solerte SB, Adamo S, Viola C, Schianca GPC, Crippa A, Ferrari E. **Acute-phase protein reactants pattern and α2-macroglobulin in diabetes mellitus. Pathophysiological aspects in diabetic microangiopathy**. *Ricerca in clinica e in laboratorio* (1984.0) **14** 575-579. DOI: 10.1007/BF02904891
45. Stevenson FT, Greene S, Kaysen GA. **Serum α2-macroglobulin and α1-inhibitor 3 concentrations are increased in hypoalbuminemia by post-transcriptional mechanisms**. *Kidney Int* (1998.0) **53** 67-75. DOI: 10.1046/j.1523-1755.1998.00734.x
46. Lagrange J, Lecompte T, Knopp T, Lacolley P, Regnault V. **Alpha-2-macroglobulin in hemostasis and thrombosis: an underestimated old double-edged sword**. *J Thromb Haemost* (2022.0) **20** 806-815. DOI: 10.1111/jth.15647
47. Soares MJ, Chakraborty D, Kubota K, Renaud SJ, Rumi MK. **Adaptive mechanisms controlling uterine spiral artery remodeling during the establishment of pregnancy**. *Int J Dev Biol* (2014.0) **58** 247. DOI: 10.1387/ijdb.140083ms
48. Young BC, Levine RJ, Karumanchi SA. **Pathogenesis of preeclampsia**. *Ann Rev Pathol* (2010.0) **5** 173-192. DOI: 10.1146/annurev-pathol-121808-102149
49. Croy BA, Burke SD, Barrette VF, Zhang J, Hatta K, Smith GN, Bianco J, Yamada AT, Adams MA. **Identification of the primary outcomes that result from deficient spiral arterial modification in pregnant mice**. *Pregnancy Hypertens* (2011.0) **1** 87-94. DOI: 10.1016/j.preghy.2010.10.002
50. Harwood SL, Lyngso J, Zarantonello A, Kjoge K, Nielsen PK, Andersen GR, Pedersen JS, Enghild JJ. **Structural investigations of human A2M identify a hollow native conformation that underlies its distinctive protease-trapping mechanism**. *Mol Cell Proteomics* (2021.0) **20** 100090. DOI: 10.1016/j.mcpro.2021.100090
51. Liu Y, Yu F, Han Y, Li Q, Cao Z, Xiang X, Jiang S, Wang X, Lu J, Lai R. **SUMO-specific protease 3 is a key regulator for hepatic lipid metabolism in non-alcoholic fatty liver disease**. *Sci Rep* (2016.0) **6** 1-11. PMID: 28442746
52. Fox R, Kitt J, Leeson P, Aye CY, Lewandowski AJ. **Preeclampsia: risk factors, diagnosis, management, and the cardiovascular impact on the offspring**. *J Clin Med* (2019.0) **8** 1625. DOI: 10.3390/jcm8101625
53. LaMarca B, Amaral LM, Harmon AC, Cornelius DC, Faulkner JL, Cunningham MW. **Placental ischemia and resultant phenotype in animal models of preeclampsia**. *Curr Hypertens Rep* (2016.0) **18** 1-9. DOI: 10.1007/s11906-016-0633-x
54. Ma YL, Yu X, Zhang LM, Liu J, Shao X, Li YX, Wang YL. **Uterine decidual niche modulates the progressive dedifferentiation of spiral artery vascular smooth muscle cells during human pregnancy**. *Biol Reprod* (2021.0) **104** 624-637. DOI: 10.1093/biolre/ioaa208
55. Saito S, Nakashima A. **A review of the mechanism for poor placentation in early-onset preeclampsia: the role of autophagy in trophoblast invasion and vascular remodeling**. *J Reprod Immunol* (2014.0) **101** 80-88. DOI: 10.1016/j.jri.2013.06.002
56. Lyall F, Robson SC, Bulmer JN. **Spiral artery remodeling and trophoblast invasion in preeclampsia and fetal growth restriction: relationship to clinical outcome**. *Hypertension* (2013.0) **62** 1046-1054. DOI: 10.1161/HYPERTENSIONAHA.113.01892
57. Bobik A. **Transforming growth factor-βs and vascular disorders**. *Arterioscler Thromb Vasc Biol* (2006.0) **26** 1712-1720. DOI: 10.1161/01.ATV.0000225287.20034.2c
58. Yao MD, Jiang Q, Ma Y, Zhu Y, Zhang QY, Shi ZH, Zhao C, Yan B. **Targeting circular RNA-MET for anti-angiogenesis treatment via inhibiting endothelial tip cell specialization**. *Mol Ther* (2022.0) **30** 1252-1264. DOI: 10.1016/j.ymthe.2022.01.012
59. Gilbert JS, Ryan MJ, LaMarca BB, Sedeek M, Murphy SR, Granger JP. **Pathophysiology of hypertension during preeclampsia: linking placental ischemia with endothelial dysfunction**. *Am J Phys Heart Circ Phys* (2008.0) **294** H541-H550
60. Granger JP, Alexander BT, Llinas MT, Bennett WA, Khalil RA. **Pathophysiology of preeclampsia: linking placental ischemia/hypoxia with microvascular dysfunction**. *Microcirculation* (2002.0) **9** 147-160. DOI: 10.1038/sj.mn.7800137
61. Myatt L, Webster RP. **Vascular biology of preeclampsia**. *J Thromb Haemost* (2009.0) **7** 375-384. DOI: 10.1111/j.1538-7836.2008.03259.x
62. Tamanna S, Lumbers ER, Morosin SK, Delforce SJ, Pringle KG. **ACE2: a key modulator of the renin-angiotensin system and pregnancy**. *Am J Physiol-Reg I* (2021.0) **321** R833-R843
63. Verdonk K, Visser W, Van Den Meiracker AH, Danser AJ. **The renin–angiotensin–aldosterone system in pre-eclampsia: the delicate balance between good and bad**. *Clin Sci* (2014.0) **126** 537-544. DOI: 10.1042/CS20130455
64. Wallukat G, Homuth V, Fischer T, Lindschau C, Horstkamp B, Jüpner A, Baur E, Nissen E, Vetter K, Neichel D. **Patients with preeclampsia develop agonistic autoantibodies against the angiotensin AT 1 receptor**. *J Clin Invest* (1999.0) **103** 945-952. DOI: 10.1172/JCI4106
|
---
title: A novel rhein-huprine hybrid ameliorates disease-modifying properties in preclinical
mice model of Alzheimer’s disease exacerbated with high fat diet
authors:
- Triana Espinosa-Jiménez
- Amanda Cano
- Elena Sánchez-López
- Jordi Olloquequi
- Jaume Folch
- Mònica Bulló
- Ester Verdaguer
- Carme Auladell
- Caterina Pont
- Diego Muñoz-Torrero
- Antoni Parcerisas
- Antoni Camins
- Miren Ettcheto
journal: Cell & Bioscience
year: 2023
pmcid: PMC9999531
doi: 10.1186/s13578-023-01000-y
license: CC BY 4.0
---
# A novel rhein-huprine hybrid ameliorates disease-modifying properties in preclinical mice model of Alzheimer’s disease exacerbated with high fat diet
## Abstract
### Background
Alzheimer’s disease (AD) is characterized by a polyetiological origin. Despite the global burden of AD and the advances made in AD drug research and development, the cure of the disease remains elusive, since any developed drug has demonstrated effectiveness to cure AD. Strikingly, an increasing number of studies indicate a linkage between AD and type 2 diabetes mellitus (T2DM), as both diseases share some common pathophysiological features. In fact, β-secretase (BACE1) and acetylcholinesterase (AChE), two enzymes involved in both conditions, have been considered promising targets for both pathologies. In this regard, due to the multifactorial origin of these diseases, current research efforts are focusing on the development of multi-target drugs as a very promising option to derive effective treatments for both conditions.
In the present study, we evaluated the effect of rhein-huprine hybrid (RHE-HUP), a synthesized BACE1 and AChE inhibitor, both considered key factors not only in AD but also in metabolic pathologies. Thus, the aim of this study is to evaluate the effects of this compound in APP/PS1 female mice, a well-established familial AD mouse model, challenged by high-fat diet (HFD) consumption to concomitantly simulate a T2DM-like condition.
### Results
Intraperitoneal treatment with RHE-HUP in APP/PS1 mice for 4 weeks reduced the main hallmarks of AD, including Tau hyperphosphorylation, Aβ42 peptide levels and plaque formation. Moreover, we found a decreased inflammatory response together with an increase in different synaptic proteins, such as drebrin 1 (DBN1) or synaptophysin, and in neurotrophic factors, especially in BDNF levels, correlated with a recovery in the number of dendritic spines, which resulted in memory improvement. Notably, the improvement observed in this model can be attributed directly to a protein regulation at central level, since no peripheral modification of those alterations induced by HFD consumption was observed.
### Conclusions
Our results suggest that RHE-HUP could be a new candidate for the treatment of AD, even for individuals with high risk due to peripheral metabolic disturbances, given its multi-target profile which allows for the improvement of some of the most important hallmarks of the disease.
## Background
Alzheimer’s disease (AD) is defined as a chronic neurodegenerative disease that involves a progressive and irreversible memory loss, followed by a state of total dementia, as well as behavioral disturbances [1, 2]. This neurodegenerative disorder considered the most common form of dementia worldwide [3], displays a high prevalence and increasing incidence, especially among elderly people. In fact, about 33.9 million people worldwide are suffering from AD, and it is expected to triple over the next 40 years [4, 5].
AD is mainly characterized by the presence of abundant extracellular amyloid-beta peptide deposits (Aβ) and intracellular hyperphosphorylated Tau protein (p-Tau), that accumulate to form senile plaques and neurofibrillary tangles (NFTs) respectively, both contributing to neuronal loss [6, 7]. Aβ plaques are produced by the proteolytic cleavages of the amyloid precursor protein (APP) by the beta-secretase 1 (BACE1) enzyme activity and subsequently by γ-secretase, resulting in Aβ peptides of different length, including 38, 40 and 42 amino acids (aa). Specifically, those Aβ composed by 42 aa readily tend to aggregate, resulting in Aβ plaque formation [8, 9]. Phosphorylation is the major modification of Tau protein and it has been described as a critical step in the formation of NFTs [10]. Evidence suggests that Aβ plaques could be involved in the induction of aberrant Tau phosphorylation, thus supporting a causal crosslink between these two pathogenic processes [11–13]. In addition, the aggregation of Aβ into oligomers and fibrils in the brain is also modified by factors such as acetylcholinesterase (AChE), which precipitates the formation of toxic aggregates by accelerating Aβ deposition and increasing its neurotoxicity, contributing to neuroinflammation, oxidative stress and synaptic dysfunction [14, 15]. Additionally, the role of AChE in AD goes much further, since numerous studies have shown the existence of a cholinergic deficit in AD patients due to the modification in the activity of AChE and the decrease in acetylcholine levels [16, 17]. In fact, some of the compounds used as anti-AD drugs like donepezil, galantamine and rivastigmine are AChE inhibitors [18]. However, none of them have been able to totally stop the progression of pathology. For this reason, new approaches to its etiology are being studied nowadays [19]. In addition, it has been described that elevated AChE concentrations could also trigger the systemic inflammation, key in T2DM and AD, representing an interesting therapeutic target for both diseases, which support previous studies that described the possible relationship between AD and metabolic alterations [20–22], stressing AD as a multifactorial disease. In fact, obesity, type 2 diabetes mellitus (T2DM) and metabolic syndrome, all associated with insulin resistance, are recognized risk factors for cognitive disturbances [23–25] and type 3 diabetes has been proposed as a term to describe the complex interlink between insulin resistance and AD [26–28].
Hence, the regulation of metabolic alterations could be an effective strategy to reduce cognitive decline and dementia [29]. In this way, some studies have shown the role of BACE1 in AD progression, not only as a key regulator of the formation of the Aβ peptide but also its function in metabolic regulation [30, 31]. In fact, it has been demonstrated that subtle neuronal expression of human BACE1 resulted in AD phenotypes alongside systemic T2DM-like symptoms, suggesting that BACE1 inhibitors could be used for the treatment of T2DM-associated pathologies [32].
Taken together, evidence suggests that AD is a complex disorder that arises from multiple molecular alterations, therefore, the design of drugs with multiple biological targets could be key for an effective treatment [33]. A recent developed multi-target RHE-HUP hybrid compounds [34] combine the pharmacophores of rhein, a natural product structurally related to some hydroxyanthraquinones with tau anti-aggregating activity, and huprine Y, a strong AChE inhibitor. RHE-HUP displays a strong in vitro activity against its primary targets (tau aggregation and AChE) and a not less strong BACE1 inhibitory activity. Studies conducted in vivo [35] have demonstrated that RHE-HUP reduced Aβ levels, Tau phosphorylation and memory impairment in an APPswe/PS-1dE9 double transgenic mouse model. However, the effect of RHE-HUP on metabolic dysregulation associated to AD has not been evaluated yet. For this reason, the aim of our study was to evaluate the efficacy of this new compound in the progression of AD when it is comorbid with metabolic alterations generated by the chronic consumption of a high-fat diet (HFD).
## Animals and treatment
6 month old female APPswe/PS1dE9 (APP/PS1) double transgenic mice and wild-type (WT) littermates with the same genetic background (C57BL/6) were used. This animal model was chosen according to previous studies reporting that female mice develop higher progressive memory impairment and AD-like neuropathology compared to male mice [36, 37]. These transgenic mice express a Swedish (K594M/N595L) mutation of a chimeric mouse/human APP (mo/huAPP695swe), together with the human exon-9-deleted variant of PS1 (PS1-dE9). In all cases, animals were obtained from established breeding couples in the animal facility (Animal facility from the Faculty of Pharmacy and Food Sciences of the University of Barcelona; approval number C-0032). After the weaning, at 21 days old, and throughout their growth, animals were fed with conventional chow (control diet, CT; ENVIGO, Madison, Wt 53744–4220) or with a palmitic acid-enriched diet containing $60\%$ of fat mainly from hydrogenated coconut oil (HFD) (Research Diets Inc., NB, US). RHE-HUP hybrid (+)-(7R,11R)-N-{9-[(3-chloro-6,7,10,11-tetrahydro-9-methyl-7,11-methanocycloocta[b]quinolin-12-yl)amino]nonyl}-9,10-dihydro-4,5-dihydroxy-9,10-dioxoanthracene-2-carboxamide was prepared as previously reported [38]. When animals were 5 months old, they were treated intraperitoneally (i.p.), either with saline solution or with RHE-HUP at a dose of 2.0 mg/Kg and diluted in bidistilled water with $3\%$ DMSO, three times per week during 4 weeks (Fig. 1). Thus, the study included three experimental groups: WT CT SALINE, APP/PS1 HFD SALINE and APP/PS1 HFD RHE-HUP.Fig. 1Graphical representation of experimental design. 6 month-old female APP/PS1 and WT littermates were used. After the weaning, animals were fed either control or HFD. When animals were 5 months old, they were treated intraperitoneally (i.p.), either with saline solution or with RHE-HUP at a dose of 2.0 mg/Kg. Then, animals were subjected to two different behavioral tests: MWM and NORT. After that, GTT and ITT were performed and animals were sacrificed by cervical dislocation in order to obtain tissue samples and to perform Golgi Staining Kit, or by intracardially perfusion for immunochemistry/ThS All animals were kept under stable conditions of humidity and temperature, standard light-dark cycle (12 h light/dark cycle) and food and water ad libitum following the ethical guidelines defined by the European Committee (European Communities Council Directive $\frac{2010}{63}$/EU). Manipulation protocols were previously approved by the ethics committee from the University of Barcelona, and, at all times, it was made sure that animal numbers, their stress, and pain were kept under a necessary minimum following the appropriate animal manipulation ethical methodologies. All the experiments were performed in accordance with the European Community Council Directive $\frac{86}{609}$/EEC and the procedures were established by the Department d’Agricultura, Ramaderia i Pesca of the Generalitat de Catalunya.
## Glucose and insulin tolerance tests
Mice were fasted for 6 h and the tests were performed in a room preheated to + 28 ℃. For the glucose tolerance test (GTT), glucose was administered at a dose of 1 g/Kg i.p. For the insulin tolerance test (ITT), a dose of 0.75 IU/Kg was used. Samples from the tail vein were extracted in consecutive periods. Glucose was measured using an Accu-check® Aviva glucometer at 5, 15, 30, 60 and 120 min after glucose administration and at 15, 30, 45, 60 and 90 min after the insulin administration. To those animals in which blood glucose levels dropped under a concentration of 20 mg/dl in the ITT, a dosage of 1 g/Kg of glucose was administered i.p. 13 animals per group were used.
## Behavioral tests assessments
Morris water maze (MWM)Hippocampal spatial memory and learning memory were assessed by the Morris Water Maze (MWM) test, which was performed as previously reported [39]. Acquired data was analyzed using SMART V3.0 (Panlab Harvard Apparatus, Germany) video tracking system. 13 animals per group were utilized. Novel object recognition test (NORT) NORT was used to assess the hippocampal-dependent recognition memory. 13 animals per group were evaluated in a room with a circular open-field arena of 40 cm in diameter surrounded by black curtains and constant illumination (30 lx) as it has been previously detailed [40]. Data were analyzed by discrimination index (DI) which was calculated using the following equation:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ DI\, = \,\frac{B\,exploration\,time\, - \,A\,exploration\,time}{{Total\,exploration\,time}} $$\end{document}DI=Bexplorationtime-AexplorationtimeTotalexplorationtime All spaces were properly cleaned with $96\%$ ethanol between animals, in order to eliminate odor or other cues. Data was measured and represented in seconds.
## Immunoblot analysis
At 6 months, 4–5 animals of each group were sacrificed by cervical dislocation and the liver and hippocampus were dissected and kept at − 80 °C until use. To perform hippocampi and liver extractions, tissues were homogenized in lysis buffer (Tris HCl 1 M pH 7.4, NaCl 5 M, EDTA 0.5 M pH 8, Triton, distilled H20) containing protease and phosphatase inhibitor cocktails (Complete Mini, EDTA-free; Protease Inhibitor cocktail tablets). Total protein concentration was determined using the Pierce™ BCA Protein Assay Kit (Thermo ScientificTM). Samples containing 10 µg of protein were analyzed by Western Blot as previously described [41]. Measurements were expressed in arbitrary units and all results were normalized with the corresponding loading control (Glyceraldehyde-3-phosphate dehydrogenase; GAPDH). The used antibodies are detailed in Table 1.Table 1Primary and secondary antibodies for Western BlottingProteinAntibodyADAM10ab124695 (abcam)AppSIG-39152 (Convance)App C terminal fragmentSIG-39152 (Convance)DBN1ABN 207 (Merck Millipore)GAPDHMAB374 (Merck Millipore)GSK3β#9315 (Cell Signaling Technology)P-GSK3β (TYR216)ab75745 (abcam)IDEab32216 (abcam)IRS24502S (Cell Signaling)Neurexinab34245 (abcam)PTP1BGTX55767 (Genetex)sAPPβSIG-39138-0 (Covance)SynaptophisinM0776 (Dako)TauGTX112981 (Genetex)P-Tau(ser396)44752G (Invitrogen)P-Tau(ser404)44-758G(Invitrogen)TLR4Sc-293072 (Santa Cruz Biotechnology)Β-actinA5441 (Sigma)2nd-ary Goat anti-Rabbit31460 (Invitrogen)2nd-ary Goat anti-Mouse31430 (Invitrogen)
## Enzyme-linked immunosorbent assay (ELISA)
BDNF (Cusabio, China; CSB-E04505m) and amyloid β1-42 (ThermoFisher Scientific; kit KHB3441) levels in the cerebral cortex homogenate were detected by ELISA according to manufacturer’s instruction. In both cases, 7 animals per group were analyzed and absorbances were read in a Varioskan LUX Multimode Microplate Reader (Thermo Fisher Scientific). Amyloid β1-42 data is expressed in pg/μg protein and BDNF levels are expressed in pg/mg protein.
## β-secretase activity assay kit
Hippocampal tissue from 7 animals were homogenized according to the manufacturer protocol (Abcam; Kit ab282921), and 35 µL of each sample were placed into a 96 well black plate. BACE1 Positive Control and EDANS Standard Curve were also added to the plate. Following the addition of the Reaction Mix, the plate was measured at Ex/Em = $\frac{345}{500}$ nm in a kinetic mode for 60 min at 37 °C. Data was treated as specified in the manufacturer’s instructions.
## Immunofluorescence and thioflavin-S staining
15 animals were previously anesthetized by i.p. injection of ketamine (100 mg/Kg) and xylazine (10 mg/Kg). When they were in the no-pain sleep phase, they were intracardially perfused with $4\%$ paraformaldehyde (PFA) diluted in 0.1 M phosphate buffer (PB). After perfusion, brains were removed and stored in $4\%$ PFA at 4 °C overnight (O/N). The next day, the solution was replaced by $4\%$ PFA + $30\%$ sucrose. Coronal sections of 20 μm were obtained by a cryostat (Leica Microsystems, Wetzlar, Germany) and they were kept in a cryoprotectant solution and stored at − 20 °C until use. To perform the experiments, the free-floating technique was used. Briefly, free-floating sections were rinsed in 0.1 M phosphate-buffered saline (PBS) pH 7.35, and after that in PBS-T (PBS 0.1 M, $0.2\%$ Triton X-100). Then they were incubated in a blocking solution ($10\%$ fetal bovine serum (FBS), $1\%$ Triton X-100, PBS 0.1 M + $0.2\%$ gelatin) for 1–2 h at room temperature. Later, sections were washed with PBS-T and incubated O/N at 4 °C with the corresponding primary antibody (Table 2). Brain slices were washed with PBS-T and incubated with the corresponding secondary antibody (Table 2) for 2 h at room temperature. Thioflavin-S (ThS) protocol was carried out as previously described [42]. Finally, sections were treated with 0.1 μg/mL Hoechst (Sigma-Aldrich, St Louis, MO, United States), used for cell nuclei staining, for 8 min in the dark at room temperature and washed with 0.1 M PBS. All reagents, containers and materials exposed to Hoechst were properly handled and processed to avoid any cytotoxic contamination. Ultimately, all the samples were mounted in Superfrost® microscope slides using *Fluoromount medium* (EMS) and were left to dry O/N. Image acquisition was obtained using an epifluorescence microscope (BX61 Laboratory Microscope, Melville, NY OlympusAmerica Inc.) and quantified by ImageJ. 5 animals per group were analyzed. Table 2Primary and secondary antibodies for ImmunofluorescenceProteinAntibodyGFAPZ0334 (Dako)IBA1O19-19741 (Wako)2nd-ary Alexa Fluor 488 (Goat-AntiMouse)A11001 (Life Technologies)2nd-ary Alexa Fluor 594 (Goat-Anti Rabbit)A11080 (Life Technologies)
## Hippocampal dendritic spine density analysis
To carry out the spine density analysis, 5 mice in each group were sacrificed by cervical dislocation. Brains were isolated and processed following the instructions of the GolgiStainTM Kit purchased from FD Neurotechnologies, Inc. (FD Rapid GolgiStainTM Kit; Cat #PK401). Images were obtained with a Leica Thunder Microscope (Leica Thunder Imager; Leica Microsystems). The quantification was carried out in 2 different zones, dentate gyrus (DG) and CA1, and 5 neurons per zone and animal were selected. DG was quantified in the secondary branches of the final fragment of the dendrites. In the DG, when analyzing the terminal fragment, 20 µm of dendrite were always left uncounted, and the counting was performed in the following 30 µm. In secondary branches, 20 µm from the ramification were left uncounted and the following 30 µm were analyzed. In CA1, two zones of the neuron were distinguished: CA1 basal and CA1 apical. In CA1 basal, the final part of the dendrite was selected, and again 20 µm of dendrite were always left uncounted, and the counting were performed in the following 30 µm. In CA1 apical, the secondary branches were selected, leaving 20 µm uncounted and analyzing the next 30 µm. Spine density was expressed as the number of spines per 30 μm of dendrite. 5 animals per group were analyzed.
## Statistical analysis
All results are presented as mean ± standard deviation (SD). Normality test was performed, when data followed a parametric distribution and more than two groups were compared, significant differences were determined by one-way analysis of variances (ANOVA), followed by Tukey’s post hoc test for comparison among groups. When only two groups were compared, Student’s t test was performed. However, when data followed a non-parametric distribution, Mann–Whitney and Kruskal–Wallis tests were performed to compare two or more than three groups, respectively. All analyses were obtained using Graph Pad Prism software for Mac version 6.01; Graph Pad Software, Inc.
## RHE-HUP does not reverse the body weight increase and glucose pathway alterations induced by HFD at peripheral level
As it has been widely described, the consumption of HFD is related to the increase in body weight, as well as to hyperglycemia and insulin resistance in mice [43, 44]. As expected, animals following a HFD showed a significant 6 month increased body weight compared with WT CT SALINE group ($p \leq 0.0001$) (Fig. 2a). The RHE-HUP treatment did not attenuate the weight gain induced by the HFD. Regarding glucose and insulin metabolism, HFD feeding showed a significant effect in both GTT (WT CT SALINE vs APP/PS1 HFD SALINE $p \leq 0.001$; WT CT SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.001$) and ITT (WT CT SALINE vs APP/PS1 HFD SALINE $p \leq 0.001$; WT CT SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.001$), regardless of the treatment (Fig. 2b–e). Because the insulin receptor substrate protein 2 (IRS2) is a key target in the hormonal control of metabolism, we measured the hepatic IRS2 protein level. A significant decrease in APP/PS1 HFD SALINE compared with WT CT SALINE ($p \leq 0.01$) was detected. However, no significant reduction was observed after the RHE-HUP treatment (Fig. 2f) suggesting that RHE-HUP does not regulate metabolic alterations observed after HFD consumption. Fig. 2a. Analysis and representation of changes in body weight ($$n = 13$$ animals per group). b. GTT and d. ITT experiment profiles ($$n = 13$$ animals per group). Area under curve (AUC) data were calculated from the time point 0 until the end of the experiment for both c. GTT and e. ITT. f. Semi-quantification of IRS2 levels in the liver where two representative samples out of four or five per group are shown ($$n = 4$$–5). All results were represented as mean ± SD. Statistical analysis was conducted through one-way ANOVA and Tukey post-test, except in the case of the analysis of weights, where the Kruskal–Wallis test was performed. In all cases, ** $p \leq 0.01$, *** $p \leq 0.001$ and **** $p \leq 0.0001$
## RHE-HUP treatment improves brain insulin signaling and attenuates Tau hyperphosphorylation
Alterations in the insulin signaling pathway have been observed in brains of AD patients [45, 46], in which IRS2 represents an important component. Our results demonstrated that the hippocampal levels of IRS2 were significantly decreased in the group APP/PS1 HFD SALINE compared with the control group ($p \leq 0.05$). Surprisingly, a recovery in IRS2 was observed after RHE-HUP treatment ($p \leq 0.05$) (Fig. 3). Since the increase in IRS2 levels has been related with an attenuation in Tau hyperphosphorylation [47], we evaluated the glycogen synthase kinase-3β (GSK3β), a main Tau kinase converging between AD and insulin resistance. Our results displayed a non-significant upward trend in the group APP/PS1 HFD SALINE when compared with WT CT SALINE. By contrast, those animals treated with RHE-HUP showed a significant decrease of GSK3β phosphorylation levels in tyrosine 216 when compared to the APP/PS1 HFD SALINE mice ($p \leq 0.05$) (Fig. 3). Regarding Tau phosphorylation in the hippocampus, our results showed a significant increase in P-Tau levels at serine 404 and serine 396 in APP/PS1 HFD SALINE mice when comparing with WT CT SALINE (P-Tauser404 $p \leq 0.05$; P-Tauser396 $p \leq 0.001$) and this effect was significantly reduced after RHE-HUP treatment (P-Tauser404 $p \leq 0.01$; P-Tauser396 $p \leq 0.05$). Our data did not show any significant changes in total Tau protein levels (Fig. 3).Fig. 3Semi-quantification of hippocampal insulin signaling pathway related proteins and Tau. Two representative samples out of four or five per group are shown ($$n = 4$$–5). All results were represented as mean ± SD. Groups were compared against each other using one-way ANOVA and Tukey post-test, except in the case of Tau protein, where Kruskal–Wallis was performed. In all cases, * $p \leq 0.05$, ** $p \leq 0.01$ and *** p $p \leq 0.01$
## RHE-HUP reduces Aβ plaques by regulating APP processing and Aβ degradation in APP/PS1 mice fed with HFD
To assess the state of Aβ burden in the hippocampus and cortex, ThS was used for detection of senile plaques. Our results demonstrated a significant decrease in the number of plaques after treatment in both regions, as shown in the images (Fig. 4a–c) and in the graphic representation ($p \leq 0.05$) (Fig. 4d–e). This result was corroborated with the significant reduction of Aβ (1–42) levels ($p \leq 0.05$) observed in the cortex after RHE-HUP administration (Fig. 4f). To elucidate the mechanisms by which RHE-HUP induced Aβ reduction, the analysis of APP processing and Aβ degradation was performed. Regarding the first one, full-length APP was analyzed. As expected, non-treated transgenic mice showed a significant increase in this protein level ($p \leq 0.05$) whereas these levels were reduced in those animals treated with RHE-HUP ($p \leq 0.05$) (Fig. 4h). In this line, BACE1 activity also showed a significant increase in APP/PS1 HFD SALINE when compared with WT CT SALINE ($p \leq 0.01$) and decreased after treatment ($p \leq 0.05$) (Fig. 4g).Fig. 4a–c. Illustrative images of Aβ plaques in the hippocampus and cortex. Scale bar: 200 µm. Graphic representation of Aβ plaques quantification in d. hippocampus and e. cortex ($$n = 5$$ independent samples per group, with at least 5 slices analyzed per sample). In hippocampus analysis, Mann–Whitney test was performed, * $p \leq 0.05.$ In cortex analysis, t-test was performed, where * $p \leq 0.05.$ f. Measurement of the levels of Aβ42 peptide in the cortex ($$n = 7$$). Statistical analysis was performed by T-test, where * $p \leq 0.05.$ g. Determination of β-secretase activity in the hippocampus ($$n = 7$$). Data were analyzed by one-way ANOVA and Tukey’s post-test, where * $p \leq 0.05$ and ** $p \leq 0.01$ h. APP processing related protein levels. Two representative samples out of four or five per group are shown ($$n = 4$$–5). All results were represented as mean ± SD. Groups were compared against each other using one-way ANOVA and Tukey post-test, * $p \leq 0.05$, ** $p \leq 0.01$ and **** $p \leq 0.0001$ APP-C-terminal fragment (APP-CTF) was significantly increased in non-treated transgenic mice compared to control group whereas soluble amyloid precursor protein β fragment (sAPPβ) did not show differences in WT vs APP/PS1 HFD. However, both proteins were reduced after treatment (APP-CTF: WT CT SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.05$; APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.0001$; sAPPβ: WT CT SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.05$; APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.05$).
The insulin-degrading enzyme (IDE) is one of the main proteases involved not only in the degradation of insulin but also in that of Aβ peptide [48]. Our results showed a significant reduction in the hippocampus of APP/PS1 HFD SALINE mice compared to WT CT SALINE, levels which were recovered after RHE-HUP treatment (WT CT SALINE vs APP/PS1 HFD SALINE $p \leq 0.01$; APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.05$).Similarly, ADAM10, a neuroprotective protein involved in the non-amyloidogenic pathway, experimented a significant reduction in APP/PS1 HFD SALINE ($p \leq 0.05$) when compared with WT CT SALINE mice, levels that were recovered after RHE-HUP treatment, reaching values similar to those of controls ($p \leq 0.05$) (Fig. 4h).
## RHE-HUP treatment decreases glial reactivity in APP/PS1 HFD mice
Increasing evidence correlates neuroinflammation with the development of AD [49, 50]. In our study, the evaluation of astrocytes and microglial reactive profile was studied in the dentate gyrus of the hippocampus by detecting glial fibrillary acidic protein (GFAP) and ionized calcium-binding adapter molecule 1 (IBA1), astrocyte and microglial markers, respectively (Fig. 5a–f). Our results showed a glial activation in those transgenic animals fed with HFD compared to WT and a clear reduction of this reactivity after the RHE-HUP treatment. These results were corroborated by the fluorescence intensity quantification data. A significant increase in astrogliosis and microglial activation in transgenic mice fed with HFD in comparison to the WT CT SALINE groups was found ($p \leq 0.001$). By contrast, this increase was significantly attenuated when these animals were treated with RHE-HUP (GFAP: APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.05$; IBA1: APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.01$) (Fig. 5 g–h).Fig. 5Evaluation of inflammatory responses. Representative images for the detection of astrocytes a–c, and microglia d–f, co-stained with Hoechst for the detection of cellular nucleus (blue). Scale bar: 200 µm. Graphic representation of fluorescence intensity quantification for GFAP g and IBA1 h. In both cases, statistical analysis was performed through one-way ANOVA ($$n = 5$$) and Tukey’s post hoc test, * $p \leq 0.05.$ ** $p \leq 0.01$ and *** $p \leq 0.001.$ i. protein levels for TLR4 and PTP1B where two representative samples out of four or five per group are shown ($$n = 4$$–5). All results were represented as mean ± SD. Groups were compared against each other using one-way ANOVA and Tukey post-test, * $p \leq 0.05$ and ** $p \leq 0.01$ Toll-like receptor 4 (TLR4) and protein tyrosine phosphatase (PTP1B), both related with neuroinflammation, were analyzed in the hippocampus. In agreement with glial profile, our results showed a similar pattern where concentrations of both proteins were significantly increased in the APP/PS1 HFD SALINE group compared to WT CT SALINE (TLR4: $p \leq 0.01$; PTP1B: $p \leq 0.05$), returning to baseline levels after treatment with RHE-HUP ($p \leq 0.01$, in both cases) (Fig. 5i).
## RHE-HUP increases dendritic spines density and synaptic biomarkers in APP/PS1 HFD mice
The reduction in the number of dendritic spines together with alterations in cognition has been widely demonstrated in AD patients, suggesting that they could play a key pathogenic role [51, 52]. Optical microscope images of the hippocampus are shown in Fig. 6a–c, accompanied by a representative magnification image of dendritic spines of each experimental group (Fig. 6d–f). A significant decrease in the number of dendritic spines was observed in APP/PS1 HFD SALINE when comparing with the control group (Fig. 6g–j), while in those animals treated with RHE-HUP, this reduction was reverted reaching levels similar to the control regardless of the studied area in the hippocampus (DG TERMINAL: WT CT SALINE vs APP/PS1 HFD SALINE $p \leq 0.001$; APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.05.$ DG RAMIFICATION: WT CT SALINE vs APP/PS1 HFD SALINE $p \leq 0.0001$; APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.01.$ CA1 BASAL: WT CT SALINE vs APP/PS1 HFD SALINE $p \leq 0.0001$; APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.001.$ CA1 APICAL: WT CT SALINE vs APP/PS1 HFD SALINE $p \leq 0.001$; APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.01.$).Fig. 6Optical microscope images of the hippocampus a–c and representative magnification images of dendritic spines of each experimental group d–e. g–j. Quantification of dendritic spines of each 30 µm of dendrite in different areas of the hippocampus ($$n = 5$$). Groups were compared against each other using one-way ANOVA and Tukey post-test, * $p \leq 0.05$, ** $p \leq 0.01$, *** $p \leq 0.001$ and **** $p \leq 0.0001$ k. Representative images of synaptic proteins levels were determined, where two representative samples out of four or five per group are shown ($$n = 4$$–5). Graphs barts represent mean ± SD. Data were analyzed by one-way ANOVA and Tukey’s post-test, * $p \leq 0.05$, ** $p \leq 0.01$ and *** $p \leq 0.001.$ l. Quantification of BDNF protein levels in the cortex ($$n = 7$$). Data were analyzed by one-way ANOVA and Tukey’s post-test, ** $p \leq 0.01$ Different synaptic proteins involved in memory process and plasticity, such as drebrin 1 (DBN1), synaptophysin and neurexin, were measured by Western Blot. Our results showed a significant decrease in DBN1 protein levels in the APP HFD SALINE group when they were compared with the control group ($p \leq 0.001$), while DBN1 levels were rescued after RHE-HUP administration ($p \leq 0.01$). A similar pattern was observed for the other synaptic proteins studied, but in the case of synaptophysin the values did not reach statistical significance, and only a positive trend was observed (Synaptophysin: APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $P \leq 0.05$; Neurexin: WT CT SALINE vs APP/PS1 HFD SALINE $p \leq 0.01$; APP/PS1 HFD SALINE vs APP/PS1 HFD RHE-HUP $p \leq 0.05$) (Fig. 6k).
Moreover, one protein that deserves special mention is BDNF plays a critical role not only in the growth and development of the nervous system, but also as a modulator of synaptic plasticity, suggesting that its regulation could play a key role in the preservation of cognitive function [53]. In this line and, in accordance with the results shown above, the analysis of BDNF levels in the cortex demonstrated a significant decrease in APP/PS1 HFD SALINE in comparison with WT CT SALINE ($p \leq 0.01$). Nevertheless, the treatment with RHE-HUP resulted in an increase of BDNF ($p \leq 0.01$) (Fig. 6l).
## The treatment with RHE-HUP improves the cognitive process in APP/PS1 HFD mice
It has been described that one of the most important features of APP/PS1 mice is cognitive decline in terms of memory and spatial memory [54, 55]. To demonstrate the efficacy of RHE-HUP treatment in the recovery of cognitive decline, MWM and NORT tests were performed. Regarding MWM, APP/PS1 HFD SALINE mice showed an obviously more erratic trajectory, being unable to find the platform compared with WT CT SALINE mice. However, after RHE-HUP treatment, the trajectory of APP/PS1 HFD RHE-HUP tended to return to normality (Fig. 7a–c). In Fig. 7d, the escape latency of all groups throughout the training period is shown. The training performed by the different groups demonstrated an improvement of the learning ability in those animals treated with RHE-HUP in comparison to those treated with saline. In the same line, the results obtained on the test day showed a significant increase in escape latency in the APP/PS1 HFD SALINE when they were compared with control group ($p \leq 0.05$), effect which was reverted in those animals treated with the drug ($p \leq 0.05$) (Fig. 7e). Moreover, other parameters studied in the same test, such as the number of entries on the platform or the mean distance traveled to reach it, showed the same tendency toward improvement of cognitive function after RHE-HUP administration. Regarding the number of entries, the time of crossing through the target platform was significantly reduced in non-treated animals ($p \leq 0.01$), whereas after treatment that number was recovered, reaching similar values to WT CT ($p \leq 0.01$) (Fig. 7f). In the case of the mean distance traveled to find the platform, non-treated animals swam a longer distance compared to the control group ($p \leq 0.05$), while after treatment, they reached the platform more easily ($p \leq 0.05$) (Fig. 7g). In agreement, in the NORT APP/PS1 HFD SALINE mice presented a decreased DI compared to the control group ($p \leq 0.001$), whereas the DI was recovered after treatment ($p \leq 0.001$), clearly indicating that RHE-HUP rescued mice from the memory deficit observed in this pathological model (Fig. 7h).Fig. 7a–c. Representative swim paths on the memory test. Learning curves of MWM during the spatial acquisition phase d and escape latency e, entries in platform f and mean distance traveled g on test day ($$n = 13$$). One-way ANOVA and Tukey’s post- test were performed, except in the case of the analysis of entries in the platform where Krushal-Wallis was conducted. In all cases, * $p \leq 0.05$ and ** $p \leq 0.01.$ h. NORT, Discrimination Index (DI) expressed in seconds ($$n = 13$$). Statistical analysis was performed by one-way ANOVA and Tukey post-test, *** $p \leq 0.001$
## Discussion
AD is nowadays recognized as a multifactorial and heterogeneous disease in which metabolic alterations play an important role [56–58]. Previous work has shown that RHE-HUP improves the main hallmarks of AD in APP/PS1 mice [35]. However, the effect of RHE-HUP in an AD familial model of mice with a metabolic syndrome-like was not evaluated, yet. Our results demonstrated that RHE-HUP significantly reduces neuroinflammation, Aβ deposition and Tau phosphorylation, considered some of the main underlying disease mechanisms. Additionally, RHE-HUP treatment succeeded in increasing the levels of BDNF and other synapse-related proteins in the brain, which resulted in an increase in the number of dendritic spines, improving memory and learning. However, these changes were not associated with modifications in the metabolic peripheral parameters.
HFD consumption leads to metabolic alterations, including insulin resistance and T2DM [59, 60], both conditions frequently associated with the development of dementia [41, 61]. T2DM is a complex disorder that begins with a state of insulin resistance, leading to hyperinsulinemia and hyperglycemia, which is known to cause different alterations in the brain. Our study confirmed that HFD induces an increase in body weight, hyperglycemia and insulin resistance in APP/PS1 mice accompanied by the downregulation of IRS2 protein levels in the liver, a protein involved in insulin signaling regulation. However, the treatment with RHE-HUP did not reverse these effects, leading us to the conclusion that the observed benefits provided by RHE-HUP may not be due to a peripheral metabolic regulation, rather to a central effect. One of the possible answers could be that this molecule was designed to hit multiple targets involved in the pathogenesis of AD, i.e., to reach biological targets located at the central nervous system. Indeed, studies performed in parallel artificial membrane permeability assays for blood-brain barrier (PAMPA-BBB) clearly demonstrated that this compound was able to enter the brain [34]. This fact was supported by the results obtained in previous studies where a reduction of Aβ levels and Tau phosphorylation leading to a memory amelioration, was observed after chronic administration of RHE-HUP to APPswe/PS-1dE9 mice [35]. Moreover, ex vivo [62–64] and in vivo biodistribution [65] studies with other hybrid compounds, closely related to RHE-HUP in terms of chemical structure and physicochemical properties, have demonstrated that this type of compounds readily enters the brain, some of them with more favorable brain/plasma ratio than the most prescribed anti-Alzheimer drug donepezil [65]. Very likely, this could be also the case for RHE-HUP, which might account for its preferential central vs peripheral effects observed in this work using the familial AD mouse model, challenged by high-fat diet.
Brain insulin, apart from controlling energy metabolism, is also involved in other multiple functions including synaptogenesis, synaptic remodeling, and neurotransmitter level modulation. Thus, unbalanced insulin signaling, and metabolism may lead to cognitive decline and AD [66]. IRS2, a major component of the insulin/insulin-like growth factor-1 signaling pathway and a key factor in T2DM, also has a role in synaptic plasticity, learning and memory. A study carried out by Tanokashira and colleagues found that young adult C57BL/6 J mice lacking IRS2 displayed hippocampus-associated behavioral alterations due to IRS2 deficiency-induced impairments of brain energy metabolism [67]. Our results agree with these data, since a IRS2 reduction was observed in the APP/PS1 HFD SALINE group recovering its levels after the RHE-HUP treatment. It has been also described that IRS2 signaling promotes the dephosphorylation of Tau, suggesting that failure on this pathway could lead to an hyperphosphorylation of Tau protein, considered one of the main early mechanisms of AD. Therefore, Tau phosphorylation might be a direct consequence of reduced insulin–IGF signaling during aging [47, 68]. Likewise, one of the main kinases responsible for Tau phosphorylation is GSK3β [69]. The phosphorylation of this kinase in Tyr216 leads to its own activation which results not only in the increase in Tau phosphorylation levels [70], but also contributes to neuronal death independently of Tau [71]. In agreement with this, the present study demonstrated that RHE-HUP administration significantly reduced Tau phosphorylation, by IRS2 and p-GSK3β regulation, which could explain the restoration of dendritic spine number and the resulting behavioral improvement observed in A PP/PS1 HFD mice after the treatment.
In addition to hyperphosphorylated Tau, another well-known hallmark of AD is the accumulation of β-amyloid deposits. Several studies have interconnected both processes defining Aβ plaques as the main triggers of Tau hyperphosphorylation and Tau tangle formation, as a result of an imbalance between Aβ production and Aβ clearance [14, 72]. In agreement with these previous data, we observed a significant reduction in the number of hippocampal and cortical Aβ plaques induced by RHE-HUP due to BACE1 inhibition. In turn, this correlated with the reduction of the levels of Aβ42, the most hydrophobic and aggregation-prone form of this peptide and, the predominant one in senile plaques [73, 74]. This event also explained the reduction in hyperphosphorylated Tau observed in this group.
As described by Pérez-Areales and coworkers, RHE-HUP seems to inhibit AChE [38], a prime target in AD, since the cholinergic deficit has been widely observed in AD patients and is directly responsible for the cognitive decline [75, 76]. However, the importance of this enzyme in the disease goes much further, since it has been described that it might bind to Aβ and promote its deposition [77], turning the combination of AChE + Aβ into more toxic to cells than Aβ alone [78].
Taking all this into account and according to our findings, the effect of RHE-HUP on decreasing the Aβ production and subsequent accumulation might be attributed to four main factors: (i) the inhibition of AChE, avoiding the interaction with Aβ and the consequent formation of the toxic aggregates; (ii) the inhibition of the amyloidogenic pathway by decreasing hippocampal BACE1 activity; (iii) the direct reduction of APP protein levels and (iv) the activation of the non-amyloidogenic pathway by increasing ADAM10 levels [79–83]. In addition, our results show that RHE-HUP treatment increased IDE levels in the hippocampus, an enzyme that not only participates in Aβ elimination, but also plays a key role in insulin degradation, all together contributing to a reduction in Aβ deposition and cognitive improvement [84].
The glial activation in the brain is also an important pathological feature of neurodegenerative diseases, including AD [85–87]. Although early in the disease neuroinflammation may represent a protective response, an excessive reaction can cause or contribute to the pathology [88]. Several reports have described that the presence of Aβ and Tau hyperphosphorylation activate microglia and astrocytes [89–91], demonstrating that microglia can play dual roles in Aβ pathogenesis. Microglia may help to eliminate Aβ aggregation, and it may facilitate Aβ accumulation through the release of neurotoxic proteases and pro-inflammatory factors, which contribute to the neuroinflammation [92–96]. Thus, it generates a vicious circle in which Aβ plaques potentiate the release of inflammatory molecules and, at the same time, these molecules stimulate the formation and accumulation of Aβ [97, 98]. Moreover, it is well-known that the chronic consumption of HFD increases stress in different pathways including neuroinflammation [99], contributing to the development of cognitive impairment. In this line, Wieckowska-Gacek et al. demonstrated that 4-months-old APPswe transgenic mice fed with western diet exhibited such brain neuroinflammation and accelerated amyloid pathology comparable to that induced by the administration of pro-inflammatory lipopolysaccharide (LPS). Hence, it highlighted the role that diet can play in neuroinflammation and, consequently, in AD [100]. In this sense, the observed decrease in the activation of microglia and astrocytes after RHE-HUP treatment might be due to the reduction in Tau phosphorylation and in Aβ deposition, but also to the improvement in the insulin signaling pathway at the central level observed upon treatment. Toll-like receptors play a pivotal role in brain injury and neurodegeneration, and, in CNS, they are mainly expressed in glial cells [101]. Specifically, the activation of TLR4 triggers the downstream stimulation of the nuclear factor kappa-light-chain-enhancer of activated B cells (NFK-β) and the induction of genes that encode inflammation-associated molecules and cytokines, such as IL-6 and TNF-α [102, 103]. Furthermore, it has been demonstrated that TLR4 deficiency protects against ethanol-induced glial activation, induction of inflammatory mediators, and apoptosis [101]. For this reason, the attenuation of the neuroinflammation observed after the RHE-HUP treatment could be related with the decrease of TLR4 levels, in agreement with previous studies which demonstrated that the treatment with resveratrol attenuated the increase in protein levels and the downstream activation of the pathway [104, 105].
In the same way, PTP1B also demonstrated a significant decrease in the RHE-HUP treated mice. Several studies have reported that the inhibition of PTP1B favors the inactivation of unfolded protein response (UPR) and neuroinflammation, thereby protecting against cognitive decline [106]. For this reason, PTP inhibitors have been suggested as a promising therapeutic modulation of microglial activation in neuroinflammatory diseases, including AD [107]. In addition, PTP1B not only has been related to this group of pathologies, but also represents a convergent point between AD and T2DM. In fact, preclinical studies have demonstrated that mice lacking PTP1B were resistant to weight gain and remained sensitive to insulin after HFD consumption [108, 109] suggesting that PTP1B downregulation could be key in order to improve the features observed in AD pathogenesis by the regulation of insulin signaling pathway and neuroinflammatory processes [110].
Moreover, in a pathological environment the released cytokines and chemokines contribute to an excessive pruning of synaptic terminals causing synaptic dysfunction and neuronal loss [111]. In fact, another important pathway in which PTP1B is involved is the BDNF/TrkB pathway [112]: PTP1B down-regulates neuronal BDNF-TrkB pathway, whereas the PTP1B inhibition stimulates BDNF signaling [113, 114]. Considering that preclinical studies suggest that the increase in BDNF levels is a suitable strategy to enhance the cognitive process [115], the decrease in PTP1B levels induced by RHE-HUP treatment observed in our results and the consequent increase in BDNF levels could explain the recovery in dendritic spines number caused by the treatment. In addition, dendritic spines loss is also related with Aβ and Tau pathology, since a study performed by Bittner et al., demonstrated that mice coexpressing mutant APP, PS1 and Tau, presented a strong loss of dendritic spines with accumulation of hyperphosphorylated Tau protein as well as soluble Aβ [116]. Therefore, the reduction already discussed in Aβ accumulation and Tau hyperphosphorylation caused by the treatment might also be contributing to the recovery of dendritic spines. These results were also accompanied by an increase in DBN1 levels. DBN1 is typically located in postsynaptic regions of excitatory synapses, and it is responsible for controlling spine function and morphology [117, 118]. Its preservation has been related to neuroprotection, and, by contrast, its reduction in the hippocampus has been linked to cognitive deficits [119, 120]. Thus, our data confirm that the increase in DBN1 could be associated with the improvement observed in cognitive functioning. In the same way, synaptophysin and neurexin showed a similar profile. Synaptophysin is a glycoprotein present in synaptic vesicles which is related to synaptic plasticity. Thus, a decrease in its levels has been related to cognitive impairment [121]. At the same time, neurexin downregulation has also been associated with cognitive impairments since it has been found to be active in synapse maturation and adaptation of synaptic strength [122]. In addition, it has been demonstrated that Aβ42 oligomers bind to neurexin, and this interaction leads to a decrease in its expression, inducing synapse pathology [123]. This would explain the increase in neurexin protein levels produced by the decrease in Aβ42 levels observed after treatment with RHE-HUP.
Recent postmortem studies in people with AD have shown that the number of dendritic spines is lower in patients with clinically evident AD compared to controls, and similar between control subjects and subjects that are cognitively normal but present the underlying biological features of AD. Thus, these observations provide cellular evidence supporting the hypothesis that dendritic spine plasticity provides a mechanism of cognitive resilience that protects people with an early stage of dementia from developing AD [124, 125]. In fact, numerous preclinical studies have related the loss of dendritic spines with hippocampus-dependent learning and memory ability impairments [126–128]. In the present study, RHE-HUP treatment induced the recovery in the number of dendritic spines, which was accompanied by an improvement in hippocampal-dependent recognition memory assessed by NORT, as well as spatial and learning memory evaluated by MWM.
In conclusion, the present study demonstrates that the multi-target compound RHE-HUP restores the number of dendritic spines and enhances cognition in APP/PS1 mice, whose pathology is exacerbated with HFD consumption, by regulation of brain insulin signaling and neuroinflammation, which contributes to the reduction of hyperphosphorylated Tau and Aβ levels (Fig. 8). However, we did not observe peripheral metabolic regulation induced by the drug administration, suggesting that the improvement observed in our model is exclusively due to a regulation at central level. These results support RHE-HUP as a new promising molecule for the treatment of AD, also in those individuals with metabolic disturbances. Fig. 8Schematic representation of the effects of RHE-HUP treatment in APP/PS1 mice fed with HFD. The figure shows the pathological mechanisms targeted by RHE-HUP that could explain the improvement in cognition observed in this double pathological model
## References
1. Forstl H, Burns A, Levy R, Cairns N, Luthert P, Lantos P. **Neuropathological correlates of behavioural disturbance in confirmed Alzheimer’s disease**. *Br J Psychiatry* (1993) **163** 364-368. DOI: 10.1192/bjp.163.3.364
2. Jahn H. **Memory loss in Alzheimer’s disease**. *Dialogues Clin Neurosci* (2022) **15** 445-454. DOI: 10.31887/DCNS.2013.15.4/hjahn
3. Scheltens P, Blennow K, Breteler MMB, de Strooper B, Frisoni GB, Salloway S. **Alzheimer’s disease**. *Lancet* (2016) **388** 505-517. DOI: 10.1016/S0140-6736(15)01124-1
4. Barnes DE, Yaffe K. **The projected effect of risk factor reduction on Alzheimer’s disease prevalence**. *Lancet Neurol* (2011) **10** 819-828. DOI: 10.1016/S1474-4422(11)70072-2
5. Gauthier S, Panisset M, Nalbantoglu J, Poirier J. **Alzheimer’s disease: current knowledge, management and research**. *CMAJ* (1997) **157** 8
6. Wegmann S, Biernat J, Mandelkow E. **A current view on Tau protein phosphorylation in Alzheimer’s disease**. *Curr Opin Neurobiol* (2021) **69** 131-138. DOI: 10.1016/j.conb.2021.03.003
7. Rahman MM, Lendel C. **Extracellular protein components of amyloid plaques and their roles in Alzheimer’s disease pathology**. *Mol Neurodegener* (2021) **16** 1-30. DOI: 10.1186/s13024-021-00465-0
8. Wilson CA, Doms RW, Lee VMY. **Intracellular APP processing and Aβ production in Alzheimer disease**. *J Neuropathol Exp Neurol* (1999) **58** 787-794. DOI: 10.1097/00005072-199908000-00001
9. Guo JP, Arai T, Miklossy J, McGeer PL. **Aβ and tau form soluble complexes that may promote self aggregation of both into the insoluble forms in Alzheimer’s diseases**. *Proc Natl Acad Sci USA* (2006) **103** 1953-1958. DOI: 10.1073/pnas.0509386103
10. Takashima A. **GSK-3 is essential in the pathogenesis of Alzheimer’s disease**. *J Alzheimer’s Dis* (2006) **9** 309-317. DOI: 10.3233/JAD-2006-9S335
11. Stoothoff WH, Johnson GVW. **Tau phosphorylation: physiological and pathological consequences**. *Biochim Biophys Acta* (2005) **1739** 280-297. DOI: 10.1016/j.bbadis.2004.06.017
12. Reddy PH. **Abnormal tau, mitochondrial dysfunction, impaired axonal transport of mitochondria, and synaptic deprivation in Alzheimer’s disease**. *Brain Res* (2011) **1415** 136-148. DOI: 10.1016/j.brainres.2011.07.052
13. Vaquer-Alicea J, Diamond MI, Joachimiak LA. **Tau strains shape disease**. *Acta Neuropathol* (2021) **142** 57-71. DOI: 10.1007/s00401-021-02301-7
14. Alvarez G, Muñoz-Montaño JR, Satrústegui J, Avila J, Bogónez E, Díaz-Nido J. **Regulation of tau phosphorylation and protection against β-amyloid-induced neurodegeneration by lithium possible implications for Alzheimer’s disease**. *Bipolar Disord* (2002) **4** 153-165. DOI: 10.1034/j.1399-5618.2002.01150.x
15. Inestrosa NC, Sagal JP, Colombres M. **Acetylcholinesterase interaction with Alzheimer amyloid beta**. *Subcell Biochem* (2005) **38** 299-317. DOI: 10.1007/0-387-23226-5_15
16. García-Ayllón MS, Small DH, Avila J, Sáez-Valero J. **Revisiting the role of acetylcholinesterase in Alzheimers disease: cross-talk with β-tau and p-amyloid**. *Front Mol Neurosci* (2011). DOI: 10.3389/fnmol.2011.00022
17. Mesulam M. **The cholinergic lesion of Alzheimer’s disease: pivotal factor or side show?**. *Learn Mem* (2004) **11** 43-49. DOI: 10.1101/lm.69204
18. Bachurin SO, Bovina EV, Ustyugov AA. **Drugs in clinical trials for Alzheimer’s disease: the major trends**. *Med Res Rev* (2017) **37** 1186-1225. DOI: 10.1002/med.21434
19. de de Paula JRV, Guimarães FM, Diniz BS, Forlenza OV. **Neurobiological pathways to Alzheimer’s disease: amyloid-beta, TAU protein or both?**. *Dement Neuropsychol* (2009) **3** 188-194. DOI: 10.1590/S1980-57642009DN30300003
20. González A, Calfío C, Churruca M, Maccioni RB. **Glucose metabolism and AD: evidence for a potential diabetes type 3**. *Alzheimer Res Ther* (2022) **14** 1-11. DOI: 10.1186/s13195-022-00996-8
21. Rao AA, Sridhar GR, Das UN. **Elevated butyrylcholinesterase and acetylcholinesterase may predict the development of type 2 diabetes mellitus and Alzheimer’s disease**. *Med Hypotheses* (2007) **69** 1272-1276. DOI: 10.1016/j.mehy.2007.03.032
22. Vieira MNN, Lima-Filho RAS, de Felice FG. **Connecting Alzheimer’s disease to diabetes: underlying mechanisms and potential therapeutic targets**. *Neuropharmacology* (2018) **136** 160-171. DOI: 10.1016/j.neuropharm.2017.11.014
23. Awad N, Gagnon M, Messier C. **The relationship between impaired glucose tolerance, type 2 diabetes, and cognitive function**. *J Clin Exp Neuropsychol* (2010) **26** 1044-1080. DOI: 10.1080/13803390490514875
24. Nooyens ACJ, Baan CA, Spijkerman AMW, Monique Verschuren WM. **Type 2 diabetes and cognitive decline in middle-aged men and womenthe doetinchem cohort study**. *Diabetes Care* (2010) **33** 1964-1969. DOI: 10.2337/dc09-2038
25. Luchsinger JA. **Type 2 diabetes and cognitive impairment: linking mechanisms**. *J Alzheimers Dis* (2012) **30** S185-S198. DOI: 10.3233/JAD-2012-111433
26. de La Monte SM, Wands JR. **Alzheimer’s disease is type 3 diabetes-evidence reviewed**. *J Diabetes Sci Technol* (2008) **2** 1101-1113. DOI: 10.1177/193229680800200619
27. de La Monte SM, Tong M, Lester-Coll N, Plater M, Wands JR. **Therapeutic rescue of neurodegeneration in experimental type 3 diabetes: relevance to Alzheimer’s disease**. *J Alzheimers Dis* (2006) **10** 89-109. DOI: 10.3233/JAD-2006-10113
28. Kroner Z. **The relationship between Alzheimer’s disease and diabetes: type 3 diabetes**. *Altern Med Rev* (2009) **14** 373-379. PMID: 20030463
29. Wolf PA, Beiser A, Elias MF, Au R, Vasan RS, Seshadri S. **Relation of obesity to cognitive function: importance of central obesity and synergistic influence of concomitant hypertension the framingham heart study**. *Curr Alzheimer Res* (2007) **4** 111-116. DOI: 10.2174/156720507780362263
30. Taylor HA, Przemylska L, Clavane EM, Meakin PJ. **BACE1: more than just a β-secretase**. *Obes Rev* (2022) **23** e13430. DOI: 10.1111/obr.13430
31. Dekeryte R, Franklin Z, Hull C, Croce L, Kamli-Salino S, Helk O. **The BACE1 inhibitor LY2886721 improves diabetic phenotypes of BACE1 knock-in mice**. *Biochim Biophys Acta* (2021) **1867** 166149. DOI: 10.1016/j.bbadis.2021.166149
32. Plucińska K, Dekeryte R, Koss D, Shearer K, Mody N, Whitfield PD. **Neuronal human BACE1 knockin induces systemic diabetes in mice**. *Diabetologia* (2016) **59** 1513-1523. DOI: 10.1007/s00125-016-3960-1
33. Muñoz-Torrero D. *Multitarget anti-Alzheimer hybrid compounds do they work in vivo? design of hybrid molecules for drug development* (2017)
34. Viayna E, Sola I, Bartolini M, de Simone A, Tapia-Rojas C, Serrano FG. **Synthesis and multitarget biological profiling of a novel family of rhein derivatives as disease-modifying anti-Alzheimer agents**. *J Med Chem* (2014) **57** 2549-2567. DOI: 10.1021/jm401824w
35. Serrano FG, Tapia-Rojas C, Carvajal FJ, Cisternas P, Viayna E, Sola I. **Rhein-huprine derivatives reduce cognitive impairment, synaptic failure and amyloid pathology in AβPPswe/PS-1 mice of different ages**. *Curr Alzheimers Res* (2016) **13** 1017-1029. DOI: 10.2174/1567205012666151027141542
36. Jiao SS, le Bu X, Liu YH, Zhu C, Wang QH, Shen LL. **Sex dimorphism profile of Alzheimer’s disease-type pathologies in an APP/PS1 mouse model**. *Neurotox Res* (2015) **29** 256-266. DOI: 10.1007/s12640-015-9589-x
37. Wang J, Tanila H, Puoliväli J, Kadish I, van Groen T. **Gender differences in the amount and deposition of amyloidβ in APPswe and PS1 double transgenic mice**. *Neurobiol Dis* (2003) **14** 318-327. DOI: 10.1016/j.nbd.2003.08.009
38. Pérez-Areales FJ, Betari N, Viayna A, Pont C, Espargaró A, Bartolini M. **Design, synthesis and multitarget biological profiling of second-generation anti-Alzheimer rhein–huprine hybrids**. *Future Med Chem* (2017) **9** 965-981. DOI: 10.4155/fmc-2017-0049
39. Ettcheto M, Sánchez-López E, Gómez-Mínguez Y, Cabrera H, Busquets O, Beas-Zarate C. **Peripheral and central effects of memantine in a mixed preclinical mice model of obesity and familial Alzheimer’s disease**. *Mol Neurobiol* (2018) **55** 7327-7339. DOI: 10.1007/s12035-018-0868-4
40. Ettcheto M, Petrov D, Pedros I, Alva N, Carbonell T, Beas-Zarate C. **Evaluation of neuropathological effects of a high-fat diet in a presymptomatic Alzheimer’s disease stage in APP/PS1 mice**. *J Alzheimers Dis* (2016) **54** 233-251. DOI: 10.3233/JAD-160150
41. Petrov D, Pedrós I, Artiach G, Sureda FX, Barroso E, Pallàs M. **High-fat diet-induced deregulation of hippocampal insulin signaling and mitochondrial homeostasis deficiences contribute to Alzheimer disease pathology in rodents**. *Biochim Biophys Acta* (2015) **1852** 1687-1699. DOI: 10.1016/j.bbadis.2015.05.004
42. Busquets O, Ettcheto M, Pallàs M, Beas-Zarate C, Verdaguer E, Auladell C. **Long-term exposition to a high fat diet favors the appearance of β-amyloid depositions in the brain of C57BL/6J mice a potential model of sporadic Alzheimer’s disease**. *Mech Ageing Dev* (2017) **162** 38-45. DOI: 10.1016/j.mad.2016.11.002
43. Yang Y, Smith DL, Keating KD, Allison DB, Nagy TR. **Variations in body weight, food intake and body composition after long-term high-fat diet feeding in C57BL/6J mice**. *Obesity* (2014) **22** 2147-2155. DOI: 10.1002/oby.20811
44. Avtanski D, Pavlov VA, Tracey KJ, Poretsky L. **Characterization of inflammation and insulin resistance in high-fat diet-induced male C57BL/6J mouse model of obesity**. *Animal Model Exp Med* (2019) **2** 252-258. DOI: 10.1002/ame2.12084
45. Gabbouj S, Ryhänen S, Marttinen M, Wittrahm R, Takalo M, Kemppainen S. **Altered insulin signaling in Alzheimer’s disease brain-special emphasis on pi3k-akt pathway**. *Front Neurosci* (2019). DOI: 10.3389/fnins.2019.00629
46. Bedse G, di Domenico F, Serviddio G, Cassano T. **Aberrant insulin signaling in Alzheimer’s disease: current knowledge**. *Front Neurosci* (2015) **9** 204. PMID: 26136647
47. Schubert M, Brazil DP, Burks DJ, Kushner JA, Ye J, Flint CL. **Insulin receptor substrate-2 deficiency impairs brain growth and promotes Tau phosphorylation**. *J Neurosci* (2003) **23** 7084-7092. DOI: 10.1523/JNEUROSCI.23-18-07084.2003
48. Kurochkin IV, Goto S. **Alzheimer’s β-amyloid peptide specifically interacts with and is degraded by insulin degrading enzyme**. *FEBS Lett* (1994) **345** 33-37. DOI: 10.1016/0014-5793(94)00387-4
49. Hoozemans JJM, Rozemuller AJM, van Haastert ES, Eikelenboom P, van Gool WA. **Neuroinflammation in Alzheimer’s disease wanes with age**. *J Neuroinflamm* (2011) **8** 1-8. DOI: 10.1186/1742-2094-8-171
50. Morales I, Guzmán-Martínez L, Cerda-Troncoso C, Farías GA, Maccioni RB. **Neuroinflammation in the pathogenesis of Alzheimer’s disease a rational framework for the search of novel therapeutic approaches**. *Front Cell Neurosci* (2014) **8** 112. PMID: 24795567
51. Boros BD, Greathouse KM, Gentry EG, Curtis KA, Birchall EL, Gearing M. **Dendritic spines provide cognitive resilience against Alzheimer’s disease**. *Ann Neurol* (2017) **82** 602-614. DOI: 10.1002/ana.25049
52. Knobloch M, Mansuy IM. **Dendritic spine loss and synaptic alterations in Alzheimer’s disease**. *Mol Neurobio* (2008) **37** 73-82. DOI: 10.1007/s12035-008-8018-z
53. Li Y, Li F, Qin D, Chen H, Wang J, Wang J. **The role of brain derived neurotrophic factor in central nervous system**. *Front Aging Neurosci* (2022) **14** 1065
54. Puoliväli J, Wang J, Heikkinen T, Heikkilä M, Tapiola T, van Groen T. **Hippocampal Aβ42 levels correlate with spatial memory deficit in APP and PS1 double transgenic mice**. *Neurobiol Dis* (2002) **9** 339-347. DOI: 10.1006/nbdi.2002.0481
55. Su D, Zhao Y, Xu H, Wang B, Chen X, Chen J. **Isoflurane exposure during mid-adulthood attenuates age-related spatial memory impairment in APP/PS1 transgenic mice**. *PLoS ONE* (2012) **7** e50172. DOI: 10.1371/journal.pone.0050172
56. Gong CX, Liu F, Iqbal K. **Multifactorial hypothesis and multi-targets for Alzheimer’s disease**. *J Alzheimers Dis* (2018) **64** S107-S117. DOI: 10.3233/JAD-179921
57. Carmo Carreiras M, Mendes E, Jesus Perry M, Paula Francisco A, Marco-Contelles J. **The multifactorial nature of Alzheimer’s disease for developing potential therapeutics**. *Curr Top Med Chem* (2013) **13** 1745-1770. DOI: 10.2174/15680266113139990135
58. Iqbal K, Grundke-Iqbal I. **Alzheimer’s disease, a multifactorial disorder seeking multitherapies**. *Alzheimers Dement* (2010) **6** 420-424. DOI: 10.1016/j.jalz.2010.04.006
59. An Y, Xu W, Li H, Lei H, Zhang L, Hao F. **High-fat diet induces dynamic metabolic alterations in multiple biological matrices of rats**. *J Proteome Res* (2013) **12** 3755-3768. DOI: 10.1021/pr400398b
60. Nesci V, Russo E, Arcidiacono B, Citraro R, Tallarico M, Constanti A. **Metabolic alterations predispose to seizure development in high-fat diet-treated mice: the role of metformin**. *Mol Neurobiol* (2020) **57** 4778-4789. DOI: 10.1007/s12035-020-02062-6
61. Kang S, Lee YH, Lee JE. **Metabolism-centric overview of the pathogenesis of Alzheimer’s disease**. *Yonsei Med J* (2017) **58** 479-488. DOI: 10.3349/ymj.2017.58.3.479
62. Galdeano C, Viayna E, Sola I, Formosa X, Camps P, Badia A. **Huprine-tacrine heterodimers as anti-amyloidogenic compounds of potential interest against Alzheimer’s and prion diseases**. *J Med Chem* (2012) **55** 661-669. DOI: 10.1021/jm200840c
63. Sola I, Aso E, Frattini D, López-González I, Espargaró A, Sabaté R. **Novel levetiracetam derivatives that are effective against the Alzheimer-like phenotype in mice: synthesis, in vitro, ex vivo, and in vivo efficacy studies**. *J Med Chem* (2015) **58** 6018-6032. DOI: 10.1021/acs.jmedchem.5b00624
64. Sola I, Viayna E, Gómez T, Galdeano C, Cassina M, Camps P. **Multigram synthesis and in vivo efficacy studies of a novel multitarget anti-Alzheimer’s compound**. *Molecules* (2015) **20** 4492-4515. DOI: 10.3390/molecules20034492
65. Viayna E, Coquelle N, Cieslikiewicz-Bouet M, Cisternas P, Oliva CA, Sánchez-López E. **Discovery of a potent dual inhibitor of acetylcholinesterase and butyrylcholinesterase with antioxidant activity that alleviates Alzheimer-like pathology in old APP/PS1 mice**. *J Med Chem* (2020) **64** 812-839. DOI: 10.1021/acs.jmedchem.0c01775
66. Craft S, Cholerton B, Baker LD. **Insulin and Alzheimer’s disease: untangling the web**. *J Alzheimers Dis* (2013) **33** S263-S275. DOI: 10.3233/JAD-2012-129042
67. Tanokashira D, Wang W, Maruyama M, Kuroiwa C, White MF, Taguchi A. **Irs2 deficiency alters hippocampus-associated behaviors during young adulthood**. *Biochem Biophys Res Commun* (2021) **559** 148-154. DOI: 10.1016/j.bbrc.2021.04.101
68. Ochiai T, Sano T, Nagayama T, Kubota N, Kadowaki T, Wakabayashi T. **Differential involvement of insulin receptor substrate (IRS)-1 and IRS-2 in brain insulin signaling is associated with the effects on amyloid pathology in a mouse model of Alzheimer’s disease**. *Neurobiol Dis* (2021) **159** 105510. DOI: 10.1016/j.nbd.2021.105510
69. Anderton BH, Betts J, Blackstock WP, Brion JP, Chapman S, Connell J. **Sites of phosphorylation in tau and factors affecting their regulation**. *Biochem Soc Symp* (2001) **67** 73-80. DOI: 10.1042/bss0670073
70. Das TK, Jana P, Chakrabarti SK, Abdul Hamid MRW. **Curcumin downregulates GSK3 and Cdk5 in scopolamine-induced alzheimer’s disease rats abrogating Aβ 40/42 and Tau hyperphosphorylation**. *J Alzheimers Dis Rep* (2019) **3** 257-267. DOI: 10.3233/ADR-190135
71. Toral-Rios D, Pichardo-Rojas PS, Alonso-Vanegas M, Campos-Peña V. **GSK3β and Tau protein in Alzheimer’s disease and epilepsy**. *Front Cell Neurosci* (2020) **14** 19. DOI: 10.3389/fncel.2020.00019
72. Hardy J, Selkoe DJ. **The amyloid hypothesis of Alzheimer’s disease: progress and problems on the road to therapeutics**. *Science* (2002) **297** 353-356. DOI: 10.1126/science.1072994
73. Klein AM, Kowall NW, Ferrante RJ. **Neurotoxicity and oxidative damage of beta amyloid 1–42 versus beta amyloid 1–40 in the mouse cerebral cortex**. *Ann N Y Acad Sci* (1999) **893** 314-320. DOI: 10.1111/j.1749-6632.1999.tb07845.x
74. Maccioni RB, Muñoz JP, Barbeito L. **The molecular bases of Alzheimer’s disease and other neurodegenerative disorders**. *Arch Med Res* (2001) **32** 367-381. DOI: 10.1016/S0188-4409(01)00316-2
75. Minger SL, Esiri MM, McDonald B, Keene J, Carter J, Hope T. **Cholinergic deficits contribute to behavioral disturbance in patients with dementia**. *Neurology* (2000) **55** 1460-1467. DOI: 10.1212/WNL.55.10.1460
76. Mega MS. **The cholinergic deficit in Alzheimer’s disease: impact on cognition, behavior and function**. *Int J Neuropsychopharmacol* (2000) **3** S3-12. DOI: 10.1017/S1461145700001942
77. Geula C, Greenberg BD, Mesulam MM. **Cholinesterase activity in the plaques, tangles and angiopathy of Alzheimer’s disease does not emanate from amyloid**. *Brain Res* (1994) **644** 327-330. DOI: 10.1016/0006-8993(94)91697-7
78. Alvarez A, Opazo C, Alarcón R, Garrido J, Inestrosa NC. **Acetylcholinesterase promotes the aggregation of amyloid-β-peptide fragments by forming a complex with the growing fibrils**. *J Mol Biol* (1997) **272** 348-361. DOI: 10.1006/jmbi.1997.1245
79. Bandyopadhyay S, Goldstein LE, Lahiri DK, Rogers JT. **Role of the APP non-amyloidogenic signaling pathway and targeting β-secretase as an alternative drug target for treatment of Alzheimer’s disease**. *Curr Med Chem* (2007) **14** 2848-2864. DOI: 10.2174/092986707782360060
80. Coimbra JRM, Marques DFF, Baptista SJ, Pereira CMF, Moreira PI, Dinis TCP. **Highlights in BACE1 inhibitors for Alzheimer’s disease treatment**. *Front Chem* (2018) **6** 178. DOI: 10.3389/fchem.2018.00178
81. Espeseth AS, Xu M, Huang Q, Coburn CA, Jones KLG, Ferrer M. **Compounds that bind APP and inhibit Aβ processing in vitro suggest a novel approach to Alzheimer disease therapeutics**. *J Biol Chem* (2005) **280** 17792-17797. DOI: 10.1074/jbc.M414331200
82. Farkhondeh T, Forouzanfar F, Roshanravan B, Samarghandian S. **Curcumin effect on non-amyloidogenic pathway for preventing Alzheimer’s disease**. *Biointerface Res Appl Chem* (2019) **9** 4085-4089. DOI: 10.33263/BRIAC94.085089
83. Guo T, Hobbs DW. **Development of BACE1 inhibitors for Alzheimer’s disease**. *Curr Med Chem* (2006) **13** 1811-1829. DOI: 10.2174/092986706777452489
84. Pivovarova O, Höhn A, Grune T, Pfeiffer AFH, Rudovich N. **Insulin-degrading enzyme: new therapeutic target for diabetes and Alzheimer’s disease?**. *Ann Med* (2016) **48** 614-624. DOI: 10.1080/07853890.2016.1197416
85. Zhang F, Jiang L. **Neuroinflammation in Alzheimer’s disease**. *Neuropsychiatr Dis Treat* (2015) **11** 243. DOI: 10.2147/NDT.S75546
86. Hong H, Kim BS, Im HI. **Pathophysiological role of neuroinflammation in neurodegenerative diseases and psychiatric disorders**. *Int Neurourol J* (2016) **20** S2. DOI: 10.5213/inj.1632604.302
87. Chen WW, Zhang X, Huang WJ. **Role of neuroinflammation in neurodegenerative diseases (review)**. *Mol Med Rep* (2016) **13** 3391-3396. DOI: 10.3892/mmr.2016.4948
88. Agostinho P, Cunha RA, Oliveira C. **Neuroinflammation, oxidative stress and the pathogenesis of Alzheimers disease**. *Curr Pharm Des* (2012) **16** 2766-2778. DOI: 10.2174/138161210793176572
89. Verri M, Pastoris O, Dossena M, Aquilani R, Guerriero F, Cuzzoni G. **Mitochondrial alterations, oxidative stress and neuroinflammation in Alzheimer’s disease**. *Int J Immunopathol Pharmacol* (2012) **25** 345-353. DOI: 10.1177/039463201202500204
90. Metcalfe MJ, Figueiredo-Pereira ME. **Relationship between Tau pathology and neuroinflammation in Alzheimer’s disease**. *Mt Sinai J Med* (2010) **77** 50-58. DOI: 10.1002/msj.20163
91. Ismail R, Parbo P, Madsen LS, Hansen AK, Hansen KV, Schaldemose JL. **The relationships between neuroinflammation, beta-amyloid and tau deposition in Alzheimer’s disease: a longitudinal PET study**. *J Neuroinflamm* (2020) **17** 1-11. DOI: 10.1186/s12974-020-01820-6
92. Hu J, Akama KT, Krafft GA, Chromy BA, van Eldik LJ. **Amyloid-β peptide activates cultured astrocytes: morphological alterations, cytokine induction and nitric oxide release**. *Brain Res* (1998) **785** 195-206. DOI: 10.1016/S0006-8993(97)01318-8
93. Szczepanik AM, Ringheim GE. **IL-10 and glucocorticoids inhibit Aβ(1–42)- and lipopolysaccharide-induced pro-inflammatory cytokine and chemokine induction in the central nervous system**. *J Alzheimers Dis* (2003) **5** 105-117. DOI: 10.3233/JAD-2003-5205
94. Grathwohl SA, Kälin RE, Bolmont T, Prokop S, Winkelmann G, Kaeser SA. **Formation and maintenance of Alzheimer’s disease β-amyloid plaques in the absence of microglia**. *Nat Neurosci* (2009) **12** 1361-1363. DOI: 10.1038/nn.2432
95. LeBlanc AC, Xue R, Gambetti P. **Amyloid precursor protein metabolism in primary cell cultures of neurons, astrocytes, and microglia**. *J Neurochem* (1996) **66** 2300-2310. DOI: 10.1046/j.1471-4159.1996.66062300.x
96. Majumdar A, Capetillo-Zarate E, Cruz D, Gouras GK, Maxfield FR. **Degradation of Alzheimer’s amyloid fibrils by microglia requires delivery of CIC-7 to lysosomes**. *Mol Biol Cell* (2011) **22** 1664-1676. DOI: 10.1091/mbc.e10-09-0745
97. Cai Z, Zhao B, Ratka A. **Oxidative stress and β-amyloid protein in Alzheimer’s disease**. *NeuroMol Med* (2011) **13** 223-250. DOI: 10.1007/s12017-011-8155-9
98. Yana MH, Wang X, Zhu X. **Mitochondrial defects and oxidative stress in Alzheimer disease and Parkinson disease**. *Free Radic Biol Med* (2013) **62** 90-101. DOI: 10.1016/j.freeradbiomed.2012.11.014
99. Reilly AM, Tsai AP, Lin PB, Ericsson AC, Oblak AL, Ren H. **Metabolic defects caused by high-fat diet modify disease risk through inflammatory and amyloidogenic pathways in a mouse model of Alzheimer’s disease**. *Nutrients* (2020) **12** 2977. DOI: 10.3390/nu12102977
100. Wiȩckowska-Gacek A, Mietelska-Porowska A, Chutorański D, Wydrych M, Długosz J, Wojda U. **Western diet induces impairment of liver-brain axis accelerating**. *Front Aging Neurosci* (2021) **13** 136. DOI: 10.3389/fnagi.2021.654509
101. Alfonso-Loeches S, Pascual-Lucas M, Blanco AM, Sanchez-Vera I, Guerri C. **Pivotal role of TLR4 receptors in alcohol-induced neuroinflammation and brain damage**. *J Neurosci* (2010) **30** 8285-8295. DOI: 10.1523/JNEUROSCI.0976-10.2010
102. Rahimifard M, Maqbool F, Moeini-Nodeh S, Niaz K, Abdollahi M, Braidy N. **Targeting the TLR4 signaling pathway by polyphenols: a novel therapeutic strategy for neuroinflammation**. *Ageing Res Rev* (2017) **36** 11-19. DOI: 10.1016/j.arr.2017.02.004
103. Yang Y, Wu J, Wang J. **Original article a database and functional annotation of NF-κB target genes**. *Int J Clin Exp Med* (2016) **9** 7986-7995
104. Azimi H, Khakshur AA, Abdollahi M, Rahimi R. **Potential new pharmacological agents derived from medicinal plants for the treatment of pancreatic cancer**. *Pancreas* (2015) **44** 11-15. DOI: 10.1097/MPA.0000000000000175
105. Zhang Z, Chen N, Liu JB, Wu JB, Zhang J, Zhang Y. **Protective effect of resveratrol against acute lung injury induced by lipopolysaccharide via inhibiting the myd88-dependent Toll-like receptor 4 signaling pathway**. *Mol Med Rep* (2014) **10** 101-106. DOI: 10.3892/mmr.2014.2226
106. Liu B, Ou G, Chen Y, Zhang J. **Inhibition of protein tyrosine phosphatase 1B protects against sevoflurane-induced neurotoxicity mediated by ER stress in developing brain**. *Brain Res Bull* (2019) **146** 28-39. DOI: 10.1016/j.brainresbull.2018.12.006
107. Song GJ, Kim J, Kim JH, Song S, Park H, Zhang ZY. **Comparative analysis of protein tyrosine phosphatases regulating microglial activation**. *Exp Neurobiol* (2016) **25** 252. DOI: 10.5607/en.2016.25.5.252
108. Elchebly M, Payette P, Michaliszyn E, Cromlish W, Collins S, Loy AL. **Increased insulin sensitivity and obesity resistance in mice lacking the protein tyrosine phosphatase-1B gene**. *Science* (1999) **283** 1544-1548. DOI: 10.1126/science.283.5407.1544
109. Klaman LD, Boss O, Peroni OD, Kim JK, Martino JL, Zabolotny JM. **Increased energy expenditure, decreased adiposity, and tissue-specific insulin sensitivity in protein-tyrosine phosphatase 1B-deficient mice**. *Mol Cell Biol* (2000) **20** 5479-5489. DOI: 10.1128/MCB.20.15.5479-5489.2000
110. Nakandakari SCBR, Muñoz VR, Kuga GK, Gaspar RC, Sant’Ana MR, Pavan ICB. **Short-term high-fat diet modulates several inflammatory, ER stress, and apoptosis markers in the hippocampus of young mice**. *Brain Behav Immun* (2019) **79** 284-293. DOI: 10.1016/j.bbi.2019.02.016
111. DiSabato DJ, Quan N, Godbout JP. **Neuroinflammation: the devil is in the details**. *J Neurochem* (2016) **139** 136-153. DOI: 10.1111/jnc.13607
112. Vieira MNN, Lyra Silva NME, Ferreira ST, de Felice FG. **Protein tyrosine phosphatase 1B (PTP1B): a potential target for Alzheimer’s therapy?**. *Front Aging Neurosci* (2017). DOI: 10.3389/fnagi.2017.00007
113. Krishnan N, Krishnan K, Connors CR, Choy MS, Page R, Peti W. **PTP1B inhibition suggests a therapeutic strategy for Rett syndrome**. *J Clin Invest* (2015) **125** 3163-3177. DOI: 10.1172/JCI80323
114. Ozek C, Kanoski SE, Zhang ZY, Grill HJ, Bence KK. **Protein-tyrosine phosphatase 1B (PTP1B) is a novel regulator of central brain-derived neurotrophic factor and tropomyosin receptor kinase B (TrkB) signaling**. *J Biol Chem* (2014) **289** 31682-31692. DOI: 10.1074/jbc.M114.603621
115. Miranda M, Morici JF, Zanoni MB, Bekinschtein P. **Brain-derived neurotrophic factor: a key molecule for memory in the healthy and the pathological brain**. *Front Cell Neurosci* (2019). DOI: 10.3389/fncel.2019.00363
116. Bittner T, Fuhrmann M, Burgold S, Ochs SM, Hoffmann N, Mitteregger G. **Multiple events lead to dendritic spine loss in triple transgenic Alzheimer’s disease mice**. *PLoS ONE* (2010) **5** e15477. DOI: 10.1371/journal.pone.0015477
117. Hayashi K, Ishikawa R, Ye LH, He XL, Takata K, Kohama K. **Modulatory role of drebrin on the cytoskeleton within dendritic spines in the rat cerebral cortex**. *J Neurosci* (1996) **16** 7161-7170. DOI: 10.1523/JNEUROSCI.16-22-07161.1996
118. Sekino Y, Kojima N, Shirao T. **Role of actin cytoskeleton in dendritic spine morphogenesis**. *Neurochem Int* (2007) **51** 92-104. DOI: 10.1016/j.neuint.2007.04.029
119. Counts SE, He B, Nadeem M, Wuu J, Scheff SW, Mufson EJ. **Hippocampal drebrin loss in mild cognitive impairment**. *Neurodegener Dis* (2012) **10** 216-219. DOI: 10.1159/000333122
120. Harigaya Y, Shoji M, Shirao T, Hirai S. **Disappearance of actin-binding protein, drebrin, from hippocampal synapses in Alzheimer’s disease**. *J Neurosci Res* (1996) **43** 87-92. DOI: 10.1002/jnr.490430111
121. Zhang X, Shen X, Dong J, Liu WC, Song M, Sun Y. **Inhibition of reactive astrocytes with fluorocitrate ameliorates learning and memory impairment through upregulating CRTC1 and synaptophysin in ischemic stroke rats**. *Cell Mol Neurobiol* (2019) **39** 1151-1163. DOI: 10.1007/s10571-019-00709-0
122. Knight D, Xie W, Boulianne GL. **Neurexins and neuroligins: recent insights from invertebrates**. *Mol Neurobiol* (2011) **44** 426. DOI: 10.1007/s12035-011-8213-1
123. Naito Y, Tanabe Y, Lee AK, Hamel E, Takahashi H. **Amyloid-β oligomers interact with neurexin and diminish neurexin-mediated excitatory presynaptic organization**. *Sci Rep* (2017) **7** 1-13. DOI: 10.1038/srep42548
124. Walker CK, Herskowitz JH. **Dendritic spines: mediators of cognitive resilience in aging and Alzheimer’s disease**. *Neuroscientist* (2020) **27** 487-505. DOI: 10.1177/1073858420945964
125. Herms J, Dorostkar MM, Landsteiner K. **Dendritic spine pathology in neurodegenerative diseases**. *Annu Rev Pathol* (2016) **11** 221-250. DOI: 10.1146/annurev-pathol-012615-044216
126. Cao Z, Yang X, Zhang H, Wang H, Huang W, Xu F. **Aluminum chloride induces neuroinflammation, loss of neuronal dendritic spine and cognition impairment in developing rat**. *Chemosphere* (2016) **151** 289-295. DOI: 10.1016/j.chemosphere.2016.02.092
127. Chabrier MA, Cheng D, Castello NA, Green KN, LaFerla FM. **Synergistic effects of amyloid-beta and wild-type human tau on dendritic spine loss in a floxed double transgenic model of Alzheimer’s disease**. *Neurobiol Dis* (2014) **64** 107-117. DOI: 10.1016/j.nbd.2014.01.007
128. Kandimalla R, Manczak M, Yin X, Wang R, Reddy PH. **Hippocampal phosphorylated Tau induced cognitive decline, dendritic spine loss and mitochondrial abnormalities in a mouse model of Alzheimer’s disease**. *Hum Mol Genet* (2018) **27** 30-40. DOI: 10.1093/hmg/ddx381
|
---
title: 'Diabetic retinopathy screenings in West Virginia: an assessment of teleophthalmology
implementation'
authors:
- Travis Schofield
- Ami Patel
- Joel Palko
- Ghassan Ghorayeb
- L. Carol Laxson
journal: BMC Ophthalmology
year: 2023
pmcid: PMC9999538
doi: 10.1186/s12886-023-02833-4
license: CC BY 4.0
---
# Diabetic retinopathy screenings in West Virginia: an assessment of teleophthalmology implementation
## Abstract
### Background
The prevalence of diabetes in the state of West Virginia (WV) is amongst the highest in the United States, making diabetic retinopathy (DR) and diabetic macular edema (DME) a major epidemiological concern within the state. Several challenges exist regarding access to eye care specialists for DR screening in this rural population. A statewide teleophthalmology program has been implemented. We analyzed real-world data acquired via these systems to explore the concordance between image findings and subsequent comprehensive eye exams and explore the impact of age on image gradeability and patient distance from the West Virginia University (WVU) Eye Institute on follow-up.
### Methods
Nonmydriatic fundus images of diabetic eyes acquired at primary care clinics throughout WV were reviewed by retina specialists at the WVU Eye Institute. Analysis included the concordance between image interpretations and dilated examination findings, hemoglobin A1c (HbA1c) levels and DR presence, image gradeability and patient age, and distance from the WVU Eye Institute and follow-up compliance.
### Results
From the 5,512 fundus images attempted, we found that 4,267 ($77.41\%$) were deemed gradable. Out of the 289 patients whose image results suggested DR, 152 patients ($52.6\%$) followed up with comprehensive eye exams—finding 101 of these patients to truly have DR/DME and allowing us to determine a positive predictive value of $66.4\%$. Patients within the HbA1c range of 9.1-$14.0\%$ demonstrated significantly greater prevalence of DR/DME ($p \leq 0.01$). We also found a statistically significant decrease in image gradeability with increased age. When considering distance from the WVU Eye Institute, it was found that patients who resided within 25 miles demonstrated significantly greater compliance to follow-up ($60\%$ versus $43\%$, $p \leq 0.01$).
### Conclusions
The statewide implementation of a telemedicine program intended to tackle the growing burden of DR in WV appears to successfully bring concerning patient cases to the forefront of provider attention. Teleophthalmology addresses the unique rural challenges of WV, but there is suboptimal compliance to essential follow-up with comprehensive eye exams. Obstacles remain to be addressed if these systems are to effectively improve outcomes in DR/DME patients and diabetic patients at risk of developing these sight-threatening pathologies.
## Background
Among working age adults, diabetic retinopathy (DR) is the most frequent cause of blindness [1]. Progression to eye pathology can be rapid, with nearly $100\%$ of type I diabetes patients and > $60\%$ of type II diabetes patients presenting with DR within the first two decades of diagnosis [2]. It has been nationally estimated that $28.5\%$ and $4.4\%$ of diabetic patients in the U.S. have DR and vision-threatening retinopathy, respectively [3]. While this is certainly a national concern with about 34.1 million American adults being diagnosed with diabetes, West Virginia (WV) has the highest prevalence of diabetes ($16.2\%$ as of 2018) [4, 5]. The state also faces unique challenges given its predominantly rural setting (over $37\%$ of its population designated as rural in comparison to $14\%$ of the total U.S. population) [6, 7]. The rural challenges of WV are compounded by the state’s notably high rates of poverty, unemployment, and low education [8, 9]. In hopes of circumventing some of these challenges, clinicians have turned to novel approaches like telemedicine in order to provide WV’s diabetic population with improved care. Teleophthalmology is one such approach and serves as the foundation for the investigations of this study.
Primary care offices may be more accessible to patients than those of specialists, especially in rural locations. Trained nurses and staff at these locations use cameras to acquire fundus photographs that can be uploaded for review by off-site specialists. Although there are limitations to the single-field, nonmydriatic fundus photography implemented at these primary care sites, these tools have allowed for detection of eye pathology in a variety of settings [10, 11], and it has been proven to be a sensitive screening tool for retina pathology, such as DR and diabetic macular edema (DME) [12]. Hence, teleophthalmology systems have been emplaced within the West Virginia University (WVU) Hospitals system. Using the U.S.-Food-and-Drug-Administration-approved Intelligent Retinal Imaging Systems (IRIS), primary care offices throughout the state have incorporated teleophthalmology into their clinical practice. Utilizing data acquired via teleophthalmology, ophthalmologists of the WVU Eye Institute have been enabled to provide guidance across the state based on their assessments of images acquired at these remote locations.
The aim of this study is to assess the success and shortcomings of WV’s teleophthalmology implementation by analyzing data regarding image gradeability and concordance between photographic screenings and subsequent comprehensive eye exams in clinic. While studies have shown that teleophthalmology is effective in assessing retina pathology and guiding appropriate referral decisions [12], we utilize this opportunity to assess the use of this technology specifically within WVU Medicine and its affiliates. Different screening modalities have been explored in the literature (nonmydriatic versus mydriatic screening [13, 14], varying fields of view [15], artificial intelligence systems [11], and smartphone-based retinal photography [13, 16]). Given that nonmydriatic, 45-degree photography was utilized in screening our population, we were interested in comparing our findings to that which has been observed and reported through other telehealth programs [17, 18]. With $20.9\%$ of WV’s population being over 65 years, we were also interested if age would play a role in the gradeability of images obtained during screening [19].
While diabetic retinopathy can be vision-threatening, proper management of diabetes and ophthalmic interventions like pan-retinal photocoagulation (PRP) and intravitreal anti-vascular endothelial growth factor (anti-VEGF) agents have shown to be effective and have become the current standard of care in managing diabetic retinopathy at various stages of its progression [20]. Hemoglobin A1c (HbA1c) severity and the presence of hyperreflective spots on spectral domain optical coherence tomography (an indicator of diabetic retinopathy progression) has shown to be linear with any HbA1c over $5.4\%$ demonstrating a high likelihood of presenting with hyperreflective spots [21]. Therefore, we have utilized the opportunity of this retrospective chart review to investigate this correlation and to determine how it might be reflected in the process and outcomes of this screening modality.
Expansion and improved accuracy in screening modalities holds substantial promise as the burden of diabetes continues to increase across the country. However, the success of these screening programs in facilitating appropriate care for patients under suspicion for vision-threatening diabetic retinopathy heavily relies on patient compliance to their providers’ recommendations. Numerous factors can affect patient compliance to care plans for diabetic retinopathy, including age, education, duration of their diabetes, practical understanding of their condition, and understanding/communication of the purpose behind teleretinal screenings [22, 23]. Given the rural setting of WV, we also sought to explore how the geographic boundaries might impact patient follow-up, which is essential to the ultimate success of these screening programs [22].
## Methods
This retrospective medical chart review consisted of collecting data regarding diabetic patients 18 years and older who have participated in the teleophthalmology program offered throughout the state of WV between January 2017 and June 2019. The WVU institutional review board approved the study protocol. The Volk Pictor (Volk Optical, Inc., Mentor, OH, USA) nonmydriatic cameras used by trained nurses and staff acquired 45-degree fundus images from patients at various primary care and endocrinology clinic settings. In these settings, patients waited in rooms with the lights turned off to maximize pupillary dilation sans mydriatic drop administration. Staff would use the handheld fundus cameras to take photographs that were then uploaded and subsequently reviewed by retina specialists. Both eyes were photographed when possible with hopes of acquiring at least one viable image per eye. The number of attempts made was contingent on the judgment of the trained staff acquiring the images and the tolerance demonstrated by the patients being screened for repeated attempts.
Images were graded by a retina specialist at the WVU Eye Institute. These specialists included three WVU board-certified retina faculty and one vitreoretinal fellow—all patients were assigned to have their set of acquired images evaluated by one of these four specialists. Images were noted as gradable or ungradable, and the extent of DR (absent, mild, moderate, severe, or proliferative) and/or DME (absent, mild, moderate, or severe) was described in accordance to the International Classification of DR scale [24]. Care plan recommendations and suspicion of other pathologies were also noted. The results with their accompanying care plan recommendations were uploaded to the Epic electronic medical record (EMR) for the use of primary care physicians (PCPs) in their advising of diabetic patients in accordance to the American Academy of Ophthalmology’s guidelines for DR follow-up (Fig. 1). Referral recommendations were made in accordance to those proposed by the International Council of Ophthalmology (ICO) and American Diabetes Association (ADA) [25]—albeit with the decision to recommend referral for suspected DR of any severity. Recommendations could also be made on the basis of other ocular pathologies that were remarked by reviewing ophthalmologists (e.g., age-related macular degeneration, choroidal nevi, colobomas, hypertensive retinopathy, glaucomatous optic nerves). For the purpose of this study, we exclusively followed patients whose screening findings indicated suspicion for diabetic retinopathy of any severity in at least one eye. Fig. 1Teleophthalmology flow chart
## Data collection
Lists of photography instances were generated, and these lists were used to investigate all photography orders recorded in the Epic EMR utilized by WVU Hospitals between January 2017 and June 2019. Photography orders that were unfulfilled (due to premature order placement by clinicians, for instance) were excluded from the study. Patient information was de-identified, and spreadsheets in Microsoft Excel were created to collect and organize the data.
Each valid photography order was investigated in the following fashion. First, the IRIS results adjoined to patients’ charts for the photography order in question would be accessed. The gradeability and presence of pathology would be recorded (specifically noting DR as mild, moderate, severe, or proliferative and DME as mild, moderate, or severe). If the screening results indicated suspicion for pathology, further investigation was conducted. Date of birth, the time that had passed since their diabetes diagnosis, their diabetes classification (Type 1 or Type 2), and HbA1c within 3-months of their photography date were all collected. Patient receipt of their results (either through record of PCP communications or indications that patients had read their results via the patient-accessible WVU MyChart system) was recorded, and whether or not an appointment was subsequently set and maintained (within 12 months of the photography order date or prior to a future repeat screening with their PCP) was also noted. Using patients’ home addresses, distances from the WVU Eye Institute to patients’ hometowns were recorded using Google Maps driving estimates. The results of patients’ dilated eye exams were recorded (noting severity as mild, moderate, severe, or proliferative for DR and absent or present for DME). Where feasible, these data were acquired from offices outside of WVU Medicine by either viewing documentation that had already been uploaded to the Epic EMR by patients or their providers or by contacting these offices directly where references in PCP notes indicated completion of ophthalmic follow-up outside WVU Medicine and permission had been granted.
## Statistical analysis
Using the data review functions of Microsoft Excel, summations and calculations were performed with the data acquired from the 2,756 patients who were studied using teleophthalmology within our selected timeframe. The totals and percentages of each attribute of interest were calculated—gradeability and the totals and proportions of DR/DME severities in PCP screenings and subsequent dilated eye exams. Pearson’s chi-squared tests were performed to compare the gradeability data found within different age ranges (18–49 years, 50–64 years, and ≥ 65 years) and the prevalence of DR within different HbA1c ranges (5.4–$6.4\%$, 6.5–$9.0\%$, and 9.1–$14.0\%$). This method was also used to investigate the relationship between patient distance from the WVU Eye Institute and compliance to follow-up with dilated eye exams.
## Results
Through WVU Medicine’s teleophthalmology screenings, 2,756 patients received screenings between January 2017 and June 2019 (first order date: $\frac{01}{12}$/2017, last order date: $\frac{06}{28}$/2019). From these 2,756 patients, 2,327 photography results ($84.43\%$) were deemed to possess at least one gradable eye by retina specialists at the WVU Eye Institute. Both eyes were deemed gradable in 1,940 patients ($70.39\%$). Two hundred, eighty-nine patients ($12.4\%$ of the patients with at least one gradable eye) had results noting some form of DR or DME. These patient cases were explored further, and it was found that 152 of these patients followed up in clinic within 12 months of their screening or prior to receiving another nonmydriatic screening with their PCP (124 within WVU Medicine, 28 with outside/external ophthalmologists). DR/DME was confirmed in 101 of these patients (Fig. 2). The confirmation of true positives with dilated eye exams enabled a calculation of the screening method’s positive predictive value. The positive predictive value was calculated to be $66.4\%$ for the teleophthalmology screening’s capacity to detect true DR/DME pathology. Other pathology notes and specifiers were also recorded, and 114 instances of age-related macular degeneration, 43 instances of hypertensive retinopathy, 60 instances of glaucomatous optic nerves, 17 instances of choroidal nevi, 3 instances of dot and blot hemorrhages, and 1 case of chorioretinal scar versus coloboma were noted throughout the data collection of screening results and subsequent dilated eye exam findings. Fig. 2PCP Screening to clinic pipeline with patient outcomes The gradeability of the screening photographs varied by age in that patients aged 65 years and older were found to have statistically significantly fewer gradable eyes than patients younger than 65 years ($63.9\%$ versus $72.7\%$, respectively, $p \leq 0.00001$), and the mean age was found to be 57.97 years (σ = 12.66) (Table 1). Breaking the < 65 years of age group down further reveals a statistically significant difference between the age ranges of 18–49 and 50–64 ($75.6\%$ versus $70.9\%$, respectively, $p \leq 0.02$), 18–49 and ≥ 65 ($75.6\%$ versus $63.9\%$, respectively, $p \leq 0.000001$), and 50–64 and ≥ 65 ($70.9\%$ versus $63.9\%$, respectively, $p \leq 0.01$) (Fig. 3).Table 1Gradeability and age analysisDate Range$\frac{01}{2017}$–$\frac{06}{2019}$Total number of patients with photography2,756Total number of patients with at least one gradable eye2,327 ($84.43\%$)Mean Age57.97 years (σ = 12.66)Age RangesTotal PatientsGradable Eyes (OU)18–49 years783592 ($75.6\%$)50–64 years1,256890 ($70.9\%$) ≥ 65 years717458 ($63.9\%$)Fig. 3Increasing ungradable fundus images with age With 2,756 patients screened, 5,512 eyes were attempted to be screened. However, 4,267 eyes ($77.41\%$) were deemed gradable, and 1,245 eyes ($22.59\%$) were deemed ungradable. No suspicion for DR was raised in 3,813 eyes ($89.36\%$), and 4,119 eyes ($96.53\%$) did not raise suspicion for DME. Some severity of DR was described in 451 eyes ($10.6\%$), and 146 eyes ($3.42\%$) showed some severity of DME. The majority of DR cases, 234 eyes ($51.9\%$), were described as mild. Moderate DR was described in 161 eyes ($35.7\%$), and 38 eyes ($8.4\%$) were described to demonstrate severe DR. PDR was noted in 18 eyes ($4.0\%$). Mild DME was described for 55 eyes ($37.7\%$), moderate DME was described for 49 eyes ($33.6\%$), and severe DME was described for 42 eyes ($28.8\%$) (Fig. 4).Fig. 4Eye Findings as per PCP screening Patient cases with DR/DME pathology were further investigated. Some note of an appointment being set was indicated in the EMR for 170 patients. One hundred, nine patients had record of an appointment being set within three months of their screening, 12 patients had record of an appointment being set within six months of their screening, 17 patients had record of an appointment being set within 12 months of their screening, and 16 patients were noted to have had an appointment set beyond 12 months but prior to their next screening with their PCPs. It was found that some form of PCP follow-up occurred in 272 cases, or $94.1\%$ of the patient cases in which DR/DME pathology was noted. PCP follow-up was deemed to have occurred if there existed some recorded form of communication between the PCP and the patient in the EMR (e-mail, phone conversation, WVU MyChart messages, et cetera) or if it was indicated that the patient had viewed their results on WVU MyChart. Compliance with follow-up varied depending on patient hometown distance from the WVU Eye Institute. It was found that patients who resided within 25 miles demonstrated statistically significantly greater compliance to follow-up with a dilated eye exam than those residing farther than 25 miles away ($60\%$ versus $43\%$, respectively, $p \leq 0.01$) (Table 2). As mentioned previously, 28 of the 152 patients who followed up in clinic were found to have records available regarding their follow-up appointments for a dilated eye exam with an ophthalmologist outside WVU Medicine. Outside appointments made up $3\%$ of the follow-up visits for those residing within 25 miles and $16\%$ of the follow-up visits for those residing beyond 25 miles. Table 2Patient distance and follow-up complianceThis patient data includes all patients with DR/DME noted in photography results (including those who complied to follow-up at external/outside offices where information was available)
Data collected from the follow-up exams revealed 187 eyes ($61.5\%$) with DR and 67 eyes ($22\%$) with DME. The majority of eyes with confirmed DR had mild DR (82 eyes, or $43.9\%$). Fifty-four eyes ($28.9\%$) were diagnosed with moderate DR, 21 eyes ($11.2\%$) were diagnosed with severe DR, and 30 eyes ($16.0\%$) were diagnosed with PDR. The presence of DME was found in 67 eyes ($22\%$) (Fig. 5).Fig. 5Eye findings as per comprehensive eye exams in clinic Regarding patient diabetes status of those under suspicion for DR/DME pathology in their initial screenings, $91\%$ were diagnosed with type 2 diabetes, and $9\%$ were diagnosed with type 1 diabetes. The mean duration of diabetes (determined by calculating the time transpired between the date of the patient’s photography and the earliest mention of a diabetes diagnosis or a historical account of such a diagnosis predating the EMR) was 6.8 years (σ = 5.3). The mean HbA1c was calculated to be $8.9\%$ (σ = 2.2) (Table 3). When the prevalence of DR/DME pathology (confirmed in clinic via dilated eye exam) was compared among patients falling within three ranges of HbA1c levels via Pearson’s chi-squared tests, no statistically significant difference was found when comparing the 5.4–$6.4\%$ range to the 6.5–$9.0\%$ range ($$p \leq 0.39$$). However, a statistically significant difference was appreciated in comparison between the 5.4–$6.4\%$ range and the 9.1–$14.0\%$ range ($p \leq 0.01$) and between the 6.5–$9.0\%$ range and the 9.1–$14.0\%$ range ($p \leq 0.01$) (Fig. 6).Table 3Patient results and concordance of findingsTotal number of patients with DR/DME via teleophthalmology289Mean HbA1C$8.9\%$σ = 2.2Type I diabetes prevalence$9\%$Type II diabetes prevalence$90\%$Mean duration of diabetesa6.8 yearsσ = 5.3 yearsTotal number of patients with DR/DME(according to comprehensive eye exam)101 ($34.9\%$)Positive Predictive Value$66.4\%$aTime between documented diagnosis and photography dateFig. 6DR/DME Prevalence Among HbA1c Ranges
## Discussion
The success of an implemented teleophthalmology screening program is contingent on both the screening process and subsequent patient compliance to recommendations based on the screening results. Through this retrospective chart review concerning WV’s teleophthalmology program, we have come to identify several areas of interest for improving our understanding of the screening process and its outcomes in this population.
The first essential step of these programs is the acquisition of the fundus images. Accurate assessment is reliant on the successful attainment of gradable fundus photographs. Our investigation revealed that $84.43\%$ of the 2,756 patients screened had at least one gradable eye, but both eyes were gradable in only $70.39\%$ of cases. The proportion of images acquired and deemed gradable in our cohort was comparable to previous studies. For instance, Tarabishy et al. found that $95.1\%$ of the 1,175 patients from which they acquired 45-degree images with a nonmydriatic camera to have gradable eyes [26]. Benjamin et al., however, acquired 1,377 45-degree images via nonmydriatic fundus photography and found that $67.4\%$ were gradable [27]. Several factors may play a role in this variability of gradeability outcomes among studies. How photographers are trained, what equipment is utilized, and whether or not dilation is an option are all factors that need to be considered. Additionally, individual variation in photographers’ thresholds for the number of attempts they make in acquiring images and the number of attempts a patient may be willing to tolerate during their PCP visits may also influence these results. Time and resources at PCP offices are also likely to vary. While the equipment, guidelines, and absence of dilation were consistent among our screenings, these other factors are more elusive and may very well have impacted the outcomes we observed. Additional standardization and documentation of the imaging protocol would provide insight into the limitations of the current screening methodology.
There is also the question of image grader variability. The concern of variability for deeming images gradable or ungradable could also be extrapolated to the image interpretations and the use of modifiers describing the extent of DR/DME observed. A limitation of this study was the use of single graders to evaluate the images—precluding the calculation of an inter-observer correlation. While we could not explore this aspect further in our own work, previous telemedicine investigations like those conducted by Liu et al. in 2019 have provided reassurance—concluding that, while it is recommended that uniform standards be established to improve consensus on image gradeability, it is unlikely that there exists much variability among ophthalmologists when assessing diabetic retinopathy through these screening methods [28].
When patients’ screening results were found to raise suspicion for DR/DME, their cases were explored further to analyze concordance with subsequent dilated eye exam findings. However, in order to confirm the diagnosis, patients first needed to comply with the recommendation to see a specialist. We found that out of the 289 patients who had DR/DME pathology noted on their screening results, only 152 ($52.6\%$) complied with subsequent appointments for a dilated eye exam. This seems to be a problem appreciated among other teleophthalmology programs as well. The investigations reported by Bresnick et al. in 2020 noted the effectiveness of these screening modalities in identifying patients in need of further examination and possibly specialized care. However, they recognized that about half of their patients failed to keep their first ophthalmology appointments and have, hence, initiated the implementation of a tracking/recall system to ensure that these at-risk patients do not miss this potentially crucial step in their vision care [29].
We attempted to explore this aspect by investigating follow-up by PCPs. While we found that $94.1\%$ of patients with noted DR/DME pathology on screening received some form of notice regarding their results, the notification method and content varied widely. Some patients received phone calls, e-mails, or WVU MyChart messages from their providers with explanations of their results and the appropriate next steps in their care. Others merely looked over the results themselves once uploaded and made viewable on WVU MyChart—possibly without any further explanation of what the results mean for their care. Some patients set appointments but failed to adhere. Some never made appointments, and other charts contained notes suggesting that an outside ophthalmologist or optometrist was planned to be seen with regards to their vision care in general. The variability in this crucial step of teleophthalmology may have contributed to the lack of compliance we observed. Furthermore, we observed that follow-up recommendations varied by PCP. While the ICO/ADA guidelines suggest referral for any cases in which photographs cannot be adequately obtained or assessed [25], we noticed a discrepancy in how PCPs managed these results. As mentioned previously, some communication of the results was sparse—some patients only seeing that their results were deemed ungradable in WVU MyChart. Other PCPs directly contacted their patients in some way, but sometimes repeat screening was chosen over referring patients to specialist care. While we chose to focus this study on the follow-up of positive screenings, this is indubitably concerning and warrants intervention for improved adherence to protocol and limiting the number of potential DR suspects who may be missing opportunities for diagnosis confirmation and subsequent care. When patients with suspected DR/DME did comply with follow-up, we found that 101 patients truly had DR/DME of various severities. While our study was limited in that we lacked the true and false negative data to explore sensitivity and specificity like previous studies [26] (we did not have patients who screened negative report to clinic for dilated eye exams for confirmation), we were able to determine that the positive predictive value of our screening was $66.4\%$.
One variable we anticipated having an impact on the gradeability results was patient age. As mentioned previously, a notably high proportion of WV’s population is aged 65 years and older [19]. The state of WV also demonstrates the greatest prevalence of diabetes [5]. With these details in mind, it was suspected that age could influence the outcomes of this study. Through our investigations, we found that there was a statistically significant difference in image gradeability between patients aged 65 years and older and those aged younger than 65 years ($63.9\%$ versus $72.7\%$, respectively, $p \leq 0.00001$). We suspect that this may be related to other ophthalmic changes commonly associated with the aging eye. For instance, refractive status and cataract development could impact the clarity of the images obtained. Nonmydriatic cameras were utilized in our screenings—attempts to maximize pupillary dilation only being achieved by having patients wait in a dark room prior to screening. Given that pupillary diameter is known to decrease with age, this could have contributed to the significant difference in gradeability we observed among patient screenings in this population [30]. Further research is required to determine if dilation in more elderly populations would substantially lower the rate of ungradable images. Interestingly, stratifying this data further into three age ranges further elucidates a negative correlation between age and image gradeability. We found a statistically significant difference between the age ranges of 18–49 and 50–64 ($75.6\%$ versus $70.9\%$, respectively, $p \leq 0.02$), 18–49 and ≥ 65 ($75.6\%$ versus $63.9\%$, respectively, $p \leq 0.000001$), and 50–64 and ≥ 65 ($70.9\%$ versus $63.9\%$, respectively, $p \leq 0.01$).
Other relevant details were explored for patients with pathology noted on screening in order to compare to previously observed trends. For instance, HbA1c severity has been shown to correlate with indicators of diabetic retinopathy severity and has served as a useful biomarker of chronic hyperglycemia, and blood glucose control has been shown to improve outcomes for retinopathy [20, 21, 31]. With this in mind, we divided patients with suspected DR/DME pathology into three HbA1c categories: 5.4–$6.4\%$ to represent the prediabetes range (diabetic patients with presumably better glycemic control), 6.5–$9.0\%$ to represent a mid-range, and 9.1–$14.0\%$ to represent the most severe cases. While we did not find a statistically significant difference between the 5.4–$6.4\%$ and 6.5–$9.0\%$ ranges ($$p \leq 0.39$$), both of these ranges demonstrated a statistically significant difference when compared to the 9.1–$14.0\%$ range ($p \leq 0.01$, $p \leq 0.01$). Our mean HbA1c was $8.9\%$ (σ = 2.2). These findings together seem to align with previous studies [27].
The false negative data was unfortunately not available for our study. Since we retrospectively studied a real-world application of teleophthalmology in which it was not recommended for patients to pursue ophthalmic follow-up for negative screenings [25], we were unable to confirm the true and false negatives. Previous studies, however, reveal data not dissimilar to those found in our study—demonstrating a greater proportion of absent or mild DR than more severe cases [32]. Nevertheless, this does not make it possible to extrapolate true and false negative rates. Furthermore, our prevalence of DR/DME by screening is notably lower than expected when compared to pooled prevalence data reported in the literature. Globally, the prevalence of DR has been estimated to be $22.27\%$ [33]—some studies estimating as high as $34.6\%$ [34]. Our screening raised suspicion for DR/DME in only $12.4\%$ of patients (with at least one gradable eye), and capturing an accurate prevalence of DR in our population is challenged further as only a subset of these patients maintained follow-up to confirm their diagnoses. However, variation in this prevalence data appears to be commonly reported among individual studies and population subgroups [33, 34]. These variations may be explained by aspects as technical as the differences in screening modalities or as fundamental as the patient demographics. Variables expected to influence the prevalence data include major risk factors, such as duration of diabetes and HbA1c [34]. Our mean HbA1c and mean age, however, suggest these factors are less likely to be contributing to the lower-than-expected DR prevalence we observed since they bear semblance to those of other studies [27]. According to the findings reported by Sato et al., our mean duration of diabetes also suggests there was ample time for expected progression to PDR [35].
It has also been reported that there is a significant difference in DR prevalence among different races—with a significantly higher DR prevalence in blacks and Hispanics ($36.7\%$ and $37.4\%$, respectively) compared to whites ($24.8\%$) [36]. There also appears to be intra-ethnic variation. For instance, Yau et al. reported a significantly higher prevalence of DR in a U.S. Caucasian population compared to an Australian Caucasian population ($35\%$ versus $15.3\%$) [34]. Genetic and environmental risk factors may all play a role in disease progression and management, and these variations may render it difficult to assess whether the DR population of WV is sufficiently being addressed. However, they may also suggest that some variation is to be expected with the unique genetic and environmental makeup of a population. Unlike past studies with greater representation of ethnic minorities [18, 27, 28], $93.1\%$ of WV’s population is white [8]. Additionally, the state’s rural setting and notably high rates of poverty, unemployment, and low education could impact the screening and subsequent follow-up on which this prevalence data relies [8, 9]. Interestingly, a study reported an unexpected lack of association between low socioeconomic status and higher grades of DR, which could be relevant to the socioeconomic impact in our WV population [37]. Ultimately, it is difficult to pinpoint whether our lower-than-expected prevalence is due to false negative screenings, ungradable images of patients with DR, or selection bias of our retrospective cohort.
While we are unable to explore the true and false negatives and this is undoubtedly a valuable component in understanding the fundamentals of teleophthalmology, our findings seem to align with past findings while offering the value of context in the subsequent follow-up phase. Furthermore, we had adjusted the ICO/ADA guidelines in hopes of minimizing false negatives. While current recommendations do not necessarily require referral to a specialist for cases of suspected mild non-proliferative DR [25], this program recommended referrals for all cases of suspected DR on screening. Not only are these recommendations comparable to those followed in previous teleophthalmology studies in other settings [27], but these recommendations granted some advantages relevant to patient care. For instance, the limited view and gradeability of our images may warrant concern for potentially missing false negative moderate-severe cases of DR that require more immediate referrals as per the ICO/ADA guidelines [25]. Given our awareness of the technological limitations and our later appreciation of the gradeability and limited view concerns, it was important that even suspected mild cases of DR be investigated further to limit missed cases of moderate-severe DR. As our prevalence data revealed, a larger proportion of patients who followed up with ophthalmologists had confirmed cases of moderate, severe, and proliferative DR. We also found that out of the population with exclusively mild DR suspected on at least one screening image, $48\%$ completed follow-up and $5\%$ of these patients were noted to have moderate, severe, or proliferative DR and/or the presence of DME—providing perhaps some support for these recommendations in the given context. Still, $36\%$ of these patients with suspected mild DR had at least one eye with mild DR exclusively, and $59\%$ had bilateral absence of DR/DME. However, with only $56\%$ of suspects for moderate or severe non-proliferative DR, PDR, and/or DME following up ($52.6\%$ of DR suspects for all severities), it is apparent that the referral process and patient adherence are important areas in need of improvement for this program and should be key points to consider for other hospital systems hoping to adopt similar programs.
It is also paramount to consider that the technology involved in teleophthalmology is constantly evolving. Automation based on artificial intelligence has been proving its effectiveness in recent studies [11]. Likewise, upgrades to imaging technologies has also been promising. We utilized handheld, nonmydriatic cameras to take 45-degree images, but newer systems could grant specialists improved field of view and resolution. Ultrawide field technology, for instance, has shown notable success [38]. According to the findings of Silva et al., ultrawide field imaging technology has been shown to reduce the number of ungradable eyes by $81\%$ [15, 32]. Improved field of view is also important for accurate disease interpretation. In addition to retinopathy, these improvements have implications for the use of telemedicine in addressing other pathology as well. In our study, we noted an abundance of other ocular pathologies, including age-related macular degeneration, hypertensive retinopathy, glaucomatous optic nerves, and choroidal nevi. Improvements in imaging would certainly benefit the identification of other pathologies as well.
Our teleophthalmology program hopes to make upgrades to the imaging technologies we utilize in order to improve the outcomes we observed in our current study and hopefully draw comparisons in the future. However, several important obstacles remain. Regardless of the improvements we achieve in our screening methods, the outcomes could fall short if patient compliance to follow-up does not improve. We suspected that the unique rural setting of West Virginia could play a role in this, and we found a statistically significant difference among patients who resided within 25 miles of the WVU Eye Institute and those who resided beyond 25 miles ($60\%$ versus $43\%$, respectively, $p \leq 0.01$). A potential limitation of this study entails possible follow-up with external providers throughout the state. Some patients’ providers had uploaded documentation regarding outside care. For others, we managed to find documentation mentioning ophthalmologists outside WVU Medicine. With permission and when feasible, we acquired documentation regarding follow-up visits from these outside offices. Unfortunately, there were likely still many patient follow-up appointments that were missed. Nevertheless, access to specialized care is a challenge for patients, and patient compliance to follow-up appointments may not be an uncommon issue amongst telemedicine programs. For instance, Peavey et al. found poorer follow-up among socioeconomically disadvantaged patients with milder DR severities in a predominantly rural population [39]. As mentioned previously, Bresnick et al. noted similar drops in compliance with plans to implement systems to hasten the delivery of results, improve engagement with patients when explaining their results and the implications they have for their vision, and reducing the window between result delivery and referral placement [29].
While we found numerous studies conducted in the context of urban settings [10, 16, 18, 23, 27, 28] and found common ground with socioeconomic obstacles [8, 9], we find our investigations of this statewide program that involves a rural setting to be unique and possibly useful to other programs. Teleophthalmology programs aiming to connect diabetic patients with specialist care through more accessible, feasible, efficient, and cost-effective screening approaches have the opportunity to improve outcomes for an ever-growing population of patients that is at risk of sight-threatening pathology. However, there are numerous obstacles to consider—one being the inherent geographic concern that is especially relevant to rural areas. New or current programs operating under similar circumstances might find a basis for comparison in our findings to set expectations and begin the process of addressing the next steps that follow the screening—ascertaining that correctly identified patients adhere to follow-up. Preferably, this can be accomplished within the hospital system or with external providers who can ensure the pipeline from screening to appropriate care is not broken. Working on methods to improve access to specialist care in WV and optimize/standardize the process of scheduling appointments for those identified by our screening to need dilated eye exams (standardizing PCP protocols/education and tracking these referrals and subsequent adherence) will be an important challenge to address as we seek to maximize the benefit of this teleophthalmology system and the quality of care it promotes.
## Conclusions
The success of teleophthalmology is contingent on a variety of factors. Many of these factors, such as age and distance from specialist care, were explored in this real-world application of teleophthalmology. These factors may be especially impactful in a rural setting, but they may also be applicable to teleophthalmology programs in other settings. While the implementation of telehealth technologies has facilitated the expansion of effective screening, follow-up confirmation of suspected diagnoses and appropriate initiation of treatment may remain hindered. This was especially suggested by the negative relationship we noted between distance from specialists and follow-up compliance among our patient population. The screening methods and statewide implementation of the program thus far among participating PCP sites has enabled extensive screening and identification of pathology. Improvements in equipment may also be promising for enhancing the accuracy of these screening approaches and possibly improving our image gradeability concern for the aging population. However, in order to improve outcomes in DR/DME patients and diabetic patients at risk of developing these sight-threatening pathologies, providers and program developers should be cognizant of the limitations brought about by geography and lack of convenient access to specialist care.
## References
1. Brechner RJ, Cowie CC, Howie LJ, Herman WH, Will JC, Harris MI. **Ophthalmic examination among adults with diagnosed diabetes mellitus**. *JAMA* (1993.0) **270** 1714-1718. DOI: 10.1001/jama.1993.03510140074032
2. Fong DS, Aiello L, Gardner TW, King GL, Blankenship G, Cavallerano JD, Ferris FL, Klein R. **American Diabetes Association. Retinopathy in diabetes**. *Diabetes Care* (2004.0) **27** S84-S87. DOI: 10.2337/diacare.27.2007.S84
3. Zhang X, Saaddine JB, Chou CF. **Prevalence of diabetic retinopathy in the United States, 2005–2008**. *JAMA* (2010.0) **304** 649-656. DOI: 10.1001/jama.2010.1111
4. 4.Centers for Disease Control and Prevention. National Diabetes Statistics Report. 2020. https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf. Accessed 9 Dec 2022.
5. 5.Centers for Disease Control and Prevention. Surveillance - United States diabetes surveillance system. 2022. https://gis.cdc.gov/grasp/diabetes/diabetesatlas-surveillance.html#. Accessed 9 Dec 2022.
6. 6.U.S. Department of Agriculture. Economic Research Service. State Fact Sheets: West Virginia. 2020. https://data.ers.usda.gov/reports.aspx?StateFIPS=54&StateName=West%20Virginia&ID=17854.
7. 7.U.S. Department of Agriculture. Economic Research Service. Rural America at a Glance: 2021 Edition. https://www.ers.usda.gov/webdocs/publications/102576/eib-230.pdf?v=4409#:~:text=The%2046%20million%20U.S.%20residents,percent%20of%20the%20U.S.%20population.
8. 8.U.S. Census Bureau. QuickFacts West Virginia; United States. July 1st, 2021. https://www.census.gov/quickfacts/fact/table/WV,US/PST045221.
9. 9.U.S. Bureau of Labor StatisticsUnemployment Rates for States2019. *Unemployment Rates for States* (2019.0)
10. Owsley C, McGwin G, Lee DJ, Lam BL, Friedman DS, Gower EW, Haller JA, Hark LA, Saaddine J. **Diabetes eye screening in urban settings serving minority populations: detection of diabetic retinopathy and other ocular findings using telemedicine**. *JAMA Ophthalmol* (2015.0) **133** 174-81. DOI: 10.1001/jamaophthalmol.2014.4652
11. Liu J, Gibson E, Ramchal S, Shankar V, Piggott K, Sychev Y, Li AS, Rao PK, Margolis TP, Fondahn E, Bhaskaranand M, Solanki K, Rajagopal R. **Diabetic retinopathy screening with automated retinal image analysis in a primary care setting improves adherence to ophthalmic care**. *Ophthalmol Retina* (2021.0) **5** 71-77. DOI: 10.1016/j.oret.2020.06.016
12. Shi L, Wu H, Dong J. **Telemedicine for detecting diabetic retinopathy: a systematic review and meta-analysis**. *Br J Ophthalmol* (2015.0) **99** 823-831. DOI: 10.1136/bjophthalmol-2014-305631
13. Han YS. **Comparison of Telemedicine Screening of Diabetic Retinopathy by Mydriatic Smartphone-Based vs Nonmydriatic Tabletop Camera-Based Fundus Imaging**. *J Vitreoretin Dis.* (2021.0) **5** 199-207. DOI: 10.1177/2474126420958304
14. Klein R, Klein BE, Neider MW, Hubbard LD, Meuer SM, Brothers RJ. **Diabetic retinopathy as detected using ophthalmoscopy, a nonmydriatic camera and a standard fundus camera**. *Ophthalmology* (1985.0) **92** 485-491. DOI: 10.1016/S0161-6420(85)34003-4
15. Silva PS, Horton MB, Clary D. **Identification of diabetic retinopathy and ungradable image rate with ultrawide field imaging in a national teleophthalmology program**. *Ophthalmology* (2016.0) **123** 1360-1367. DOI: 10.1016/j.ophtha.2016.01.043
16. Queiroz MS, de Carvalho JX, Bortoto SF, de Matos MR, das Graças Dias Cavalcante C, Andrade EAS, Correa-Giannella ML, Malerbi FK. **Diabetic retinopathy screening in urban primary care setting with a handheld smartphone-based retinal camera**. *Acta Diabetol* (2020.0) **57** 1493-1499. DOI: 10.1007/s00592-020-01585-7
17. Gómez-Ulla F, Fernandez MI, Gonzalez F, Rey P, Rodriguez M, Rodriguez-Cid MJ, Casanueva FF, Tome MA, Garcia-Tobio J, Gude F. **Digital retinal images and teleophthalmology for detecting and grading diabetic retinopathy**. *Diabetes Care* (2002.0) **25** 1384-1389. DOI: 10.2337/diacare.25.8.1384
18. Chin EK, Ventura BV, See KY, Seibles J, Park SS. **Nonmydriatic fundus photography for teleophthalmology diabetic retinopathy screening in rural and urban clinics**. *Telemed J E Health* (2014.0) **20** 102-108. DOI: 10.1089/tmj.2013.0042
19. 19.U.S. Census Bureau. Vintage 2020 Population Estimates. https://www.census.gov/programs-surveys/popest/technical-documentation/research/evaluation-estimates/2020-evaluation-estimates.html. Accessed 9 Dec 2022.
20. 20.American Academy of Ophthalmology. Diabetic retinopathy PPP 2019. https://www.aao.org/education/preferred-practice-pattern/diabetic-retinopathy-ppp. Accessed 9 Dec 2022.
21. Wong BS, Sharanjeet-Kaur S, Ngah NF, Sawri RR. **The Correlation between Hemoglobin A1c (HbA1c) and Hyperreflective Dots (HRD) in Diabetic Patients**. *Int J Environ Res Public Health.* (2020.0) **17** 3154. DOI: 10.3390/ijerph17093154
22. Schoenfeld ER, Greene JM, Wu SY, Leske MC. **Patterns of adherence to diabetes vision care guidelines: baseline findings from the Diabetic Retinopathy Awareness Program**. *Ophthalmology* (2001.0) **108** 563-571. DOI: 10.1016/S0161-6420(00)00600-X
23. Ogunyemi O, George S, Patty L, Teklehaimanot S, Baker R. **Teleretinal screening for diabetic retinopathy in six Los Angeles urban safety-net clinics: final study results**. *AMIA Annu Symp Proc* (2013.0) **2013** 1082-1088. PMID: 24551394
24. Wilkinson CP, Ferris FL, Klein RE. **Proposed international clinical diabetic retinopathy and diabetic macular edema disease severity scales**. *Ophthalmology* (2003.0) **110** 1677-1682. DOI: 10.1016/S0161-6420(03)00475-5
25. Wong TY, Sun J, Kawasaki R. **Guidelines on diabetic eye care: the international council of ophthalmology recommendations for screening, follow-up, referral, and treatment based on resource settings**. *Ophthalmology* (2018.0) **125** 1608-1622. DOI: 10.1016/j.ophtha.2018.04.007
26. Tarabishy AB, Campbell JP, Misra-Hebert A, Seballos RJ, Lang RS, Singh RP. **Non-mydriatic single-field fundus photography for the screening of retinal diseases in an executive health clinic**. *Ophthalmic Surg Lasers Imaging* (2011.0) **42** 102-6. DOI: 10.3928/15428877-20110316-01
27. Benjamin JE, Sun J, Cohen D, Matz J, Barbera A, Henderer J, Cheng L, Grachevskaya J, Shah R, Zhang Y. **A 15 month experience with a primary care-based telemedicine screening program for diabetic retinopathy**. *BMC Ophthalmol* (2021.0) **21** 70. DOI: 10.1186/s12886-021-01828-3
28. Liu Y, Rajamanickam VP, Parikh RS. **Diabetic retinopathy assessment variability among eye care providers in an urban teleophthalmology program**. *Telemed J E Health* (2019.0) **25** 301-308. DOI: 10.1089/tmj.2018.0019
29. Bresnick G, Cuadros JA, Khan M, Fleischmann S, Wolff G, Limon A, Chang J, Jiang L, Cuadros P, Pedersen ER. **Adherence to ophthalmology referral, treatment and follow-up after diabetic retinopathy screening in the primary care setting**. *BMJ Open Diabetes Res Care* (2020.0) **8** e001154. DOI: 10.1136/bmjdrc-2019-001154
30. Guillon M, Dumbleton K, Theodoratos P, Gobbe M, Wooley CB, Moody K. **The Effects of Age, Refractive Status, and Luminance on Pupil Size**. *Optom Vis Sci* (2016.0) **93** 1093-1100. DOI: 10.1097/OPX.0000000000000893
31. Aiello LP, Cahill MT, Wong JS. **Systemic considerations in the management of diabetic retinopathy**. *Am J Ophthalmol* (2001.0) **132** 760-776. DOI: 10.1016/S0002-9394(01)01124-2
32. Silva PS, Cavallerano JD, Tolson AM, Rodriguez J, Rodriguez S, Ajlan R, Tolls D, Patel B, Sehizadeh M, Thakore K. **Real-Time Ultrawide field image evaluation of retinopathy in a diabetes telemedicine program**. *Diabetes Care* (2015.0) **38** 1643-1649. DOI: 10.2337/dc15-0161
33. Teo ZL, Tham YC, Yu M. **Global prevalence of diabetic retinopathy and projection of burden through 2045: systematic review and meta-analysis**. *Ophthalmology* (2021.0) **128** 1580-1591. DOI: 10.1016/j.ophtha.2021.04.027
34. Yau JWY, Rogers SL, Kawasaki R. **Global prevalence and major risk factors of diabetic retinopathy**. *Diabetes Care* (2012.0) **35** 556-564. DOI: 10.2337/dc11-1909
35. Sato Y, Lee Z, Hayashi Y. **Subclassification of preproliferative diabetic retinopathy and glycemic control: relationship between mean hemoglobin A1C value and development of proliferative diabetic retinopathy**. *Jpn J Ophthalmol* (2001.0) **45** 523-527. DOI: 10.1016/S0021-5155(01)00380-X
36. Wong TY, Klein R, Islam FMA. **Diabetic retinopathy in a multi-ethnic cohort in the United States**. *Am J Ophthalmol* (2006.0) **141** 446-455. DOI: 10.1016/j.ajo.2005.08.063
37. Haffner SM, Hazuda HP, Stern MP, Patterson JK, Van Heuven WA, Fong D. **Effects of socioeconomic status on hyperglycemia and retinopathy levels in Mexican Americans with NIDDM**. *Diabetes Care* (1989.0) **12** 128-134. DOI: 10.2337/diacare.12.2.128
38. Patel SN, Shi A, Wibbelsman TD, Klufas MA. **Ultra-widefield retinal imaging: an update on recent advances**. *Ther Adv Ophthalmol* (2020.0) **12** 2515841419899495. PMID: 32010879
39. Peavey JJ, D’Amico SL, Kim BY, Higgins ST, Friedman DS, Brady CJ. **Impact of socioeconomic disadvantage and diabetic retinopathy severity on poor ophthalmic follow-up in a rural Vermont and New York population**. *Clin Ophthalmol* (2020.0) **14** 2397-2403. DOI: 10.2147/OPTH.S258270
|
---
title: 'Weight loss and outcomes in subjects with progressive pulmonary fibrosis:
data from the INBUILD trial'
authors:
- Michael Kreuter
- Elisabeth Bendstrup
- Stéphane Jouneau
- Toby M. Maher
- Yoshikazu Inoue
- Corinna Miede
- Dirk Lievens
- Bruno Crestani
journal: Respiratory Research
year: 2023
pmcid: PMC9999543
doi: 10.1186/s12931-023-02371-z
license: CC BY 4.0
---
# Weight loss and outcomes in subjects with progressive pulmonary fibrosis: data from the INBUILD trial
## Abstract
### Background
Lower body mass index (BMI) and weight loss have been associated with worse outcomes in some studies in patients with pulmonary fibrosis. We analyzed outcomes in subgroups by BMI at baseline and associations between weight change and outcomes in subjects with progressive pulmonary fibrosis (PPF) in the INBUILD trial.
### Methods
Subjects with PPF other than idiopathic pulmonary fibrosis were randomized to receive nintedanib or placebo. In subgroups by BMI at baseline (< 25, ≥ 25 to < 30, ≥ 30 kg/m2), we analyzed the rate of decline in FVC (mL/year) over 52 weeks and time-to-event endpoints indicating disease progression over the whole trial. We used a joint modelling approach to assess associations between change in weight and the time-to-event endpoints.
### Results
Among 662 subjects, $28.4\%$, $36.6\%$ and $35.0\%$ had BMI < 25, ≥ 25 to < 30 and ≥ 30 kg/m2, respectively. The rate of decline in FVC over 52 weeks was numerically greater in subjects with baseline BMI < 25 than ≥ 25 to < 30 or ≥ 30 kg/m2 (nintedanib: − 123.4, − 83.3, − 46.9 mL/year, respectively; placebo: − 229.5; − 176.9; − 171.2 mL/year, respectively). No heterogeneity was detected in the effect of nintedanib on reducing the rate of FVC decline among these subgroups (interaction $$p \leq 0.83$$). In the placebo group, in subjects with baseline BMI < 25, ≥ 25 to < 30 and ≥ 30 kg/m2, respectively, $24.5\%$, $21.4\%$ and $14.0\%$ of subjects had an acute exacerbation or died, and $60.2\%$, $54.5\%$ and $50.4\%$ of subjects had ILD progression (absolute decline in FVC % predicted ≥ $10\%$) or died over the whole trial. The proportions of subjects with these events were similar or lower in subjects who received nintedanib versus placebo across the subgroups. Based on a joint modelling approach, over the whole trial, a 4 kg weight decrease corresponded to a 1.38-fold ($95\%$ CI 1.13, 1.68) increase in the risk of acute exacerbation or death. No association was detected between weight loss and the risk of ILD progression or the risk of ILD progression or death.
### Conclusions
In patients with PPF, lower BMI at baseline and weight loss may be associated with worse outcomes and measures to prevent weight loss may be required.
Trial registration: https://clinicaltrials.gov/ct2/show/NCT02999178.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12931-023-02371-z.
The online version contains supplementary material available at 10.1186/s12931-023-02371-z.
## Plain language summary
Patients with worsening fibrosis (scarring) of the lungs may lose weight. This study suggests that the course of disease may be worse in patients who lose weight. Measures to prevent weight loss may be needed in these patients.
## Background
Patients with various forms of fibrosing interstitial lung disease (ILD) may develop progressive fibrosing ILD, more recently termed progressive pulmonary fibrosis (PPF), characterized by decline in lung function, increasing fibrosis on radiology, worsening symptoms and quality of life, and high mortality [1]. Decline in lung function in patients with pulmonary fibrosis is associated with mortality [2–4].
Many patients with pulmonary fibrosis experience weight loss [5–7]. Published data on the association between weight and outcomes in patients with pulmonary fibrosis are conflicting. While some studies have suggested that lower body mass index (BMI) is associated with worse outcomes [7–14], others have found no significant association [2, 5, 15–17]. Weight loss has been associated with worse outcomes in patients with pulmonary fibrosis [5, 7, 9, 10, 13, 18–21], although among overweight and obese patients, intentional weight loss may improve lung function [22, 23].
The randomized placebo-controlled INBUILD trial of nintedanib was conducted in subjects with progressive fibrosing ILDs other than idiopathic pulmonary fibrosis (IPF). The results showed that nintedanib slowed decline in lung function, with an adverse event profile characterized mainly by gastrointestinal events [24–26]. We analyzed outcomes in the INBUILD trial in subgroups by BMI at baseline and assessed associations between weight change and time-to-event outcomes using a joint modelling approach.
## Methods
The design of the INBUILD trial has been described and the protocol is publicly available [24]. Briefly, subjects had an ILD other than IPF with reticular abnormality with traction bronchiectasis (with or without honeycombing) of > $10\%$ extent on high-resolution computed tomography (HRCT), forced vital capacity (FVC) ≥ $45\%$ predicted and diffusing capacity of the lung for carbon monoxide (DLco) ≥ 30– < $80\%$ predicted. Subjects met criteria for ILD progression within the prior 24 months, based on worsening of FVC, abnormalities on HRCT, or symptoms, despite management deemed appropriate in clinical practice. Subjects were randomized to receive nintedanib 150 mg bid or placebo, stratified by pattern on HRCT (usual interstitial pneumonia [UIP]-like fibrotic pattern or other fibrotic patterns [24]). Treatment interruptions (for ≤ 4 weeks for adverse events considered related to trial medication or ≤ 8 weeks for other adverse events) and dose reductions to 100 mg bid were used to manage adverse events. The trial consisted of two parts. Part A comprised 52 weeks of treatment. Part B was a variable period during which subjects continued to receive blinded treatment until all the subjects had completed the trial. The final database lock took place after all subjects had completed the follow-up visit or had entered the open-label extension study, INBUILD-ON (NCT03820726); the data available at this point are referred to as data from the whole trial.
## Analyses in subgroups by BMI at baseline
In these post-hoc analyses, we analyzed the rate of decline in FVC (mL/year) over 52 weeks in subgroups by BMI at baseline (< 25, ≥ 25 to < 30, ≥ 30 kg/m2). In the same subgroups, we analyzed two time-to-event endpoints: time to acute exacerbation of ILD (defined in [24]) or death and time to ILD progression (absolute decline in FVC % predicted ≥ $10\%$) or death. Exploratory interaction p-values were calculated to evaluate potential heterogeneity in the treatment effect of nintedanib versus placebo across the subgroups, a recommended approach for the reporting of subgroup analyses of clinical trials [27, 28]. The analyses were not adjusted for multiple testing. Adverse events are presented descriptively.
## Joint modelling
We used joint models for longitudinal and time-to-event data. These comprise two sub-models for the respective processes and an association structure to connect them. We assessed the association between change in weight (kg) and three time-to-event endpoints (time to acute exacerbation of ILD or death, time to ILD progression, time to ILD progression or death) over 52 weeks and over the whole trial. In the longitudinal sub-model, a normal mixed effects model of weight was used, with HRCT pattern (UIP-like fibrotic pattern or other fibrotic patterns) and weight at baseline as predictor variables. Separate mean slopes for subjects in the nintedanib and placebo groups were assumed. Trajectories were modelled by a linear trend with an unstructured variance–covariance matrix assumed. Weight was assessed at baseline, at weeks 2, 4, 6, 12, 24, 36 and 52 and every 16 weeks thereafter. Only values obtained before an event or censoring timepoint were considered. In the time-to-event sub‑model, a piecewise exponential model with five knots was used to model the baseline hazard. Weight was used as the endogenous time-dependent covariate and treatment as a predictor variable. The sub-model was stratified by HRCT pattern. Subjects were censored once they experienced an event and were not considered at risk of further events.
The shared parameter in each of the joint models was the estimated slope of weight (i.e., the annual rate of change in weight), which assumed that the rate of change in weight affected the risk of an event. Joint models were fitted for each time-to-event endpoint. We present the risk of the first event in terms of 1-unit, 4-unit and 10-unit decreases in weight (kg). The joint model approach was implemented with the SAS macro %JM [29]. Analyses were performed in subjects who had ≥ 1 post-baseline weight measurement and for whom data on the respective time-to-event endpoint were available.
## Characteristics of subgroups by weight loss
We present descriptive analyses of the baseline characteristics of subjects with weight loss ≤ $5\%$ and > $5\%$ over the whole trial, based on change in weight from baseline to any visit.
## Baseline BMI and outcomes
Among 662 subjects with available data, mean (SD) BMI at baseline was 28.3 (5.3) kg/m2; $28.4\%$, $36.6\%$ and $35.0\%$ of subjects had a BMI of < 25, ≥ 25 to < 30 and ≥ 30 kg/m2, respectively. Compared with subjects with a baseline BMI ≥ 25 to < 30 or ≥ 30 kg/m2, a numerically greater proportion of subjects with BMI ≤ 25 kg/m2 were Asian, a greater proportion had autoimmune disease-related ILDs, and a greater proportion had a UIP-like fibrotic pattern on HRCT (Additional file 1: Table S1). The mean time since diagnosis of ILD, FVC % predicted and DLco % predicted were similar across the subgroups by baseline BMI (Additional file 1: Table S1).
In both the nintedanib and placebo groups, the rate of decline in FVC over 52 weeks was numerically greater in subjects with baseline BMI < 25 than ≥ 25 to < 30 or ≥ 30 kg/m2 (Fig. 1). The exploratory interaction p-value did not indicate heterogeneity in the effect of nintedanib on reducing the rate of FVC decline among the subgroups by baseline BMI ($$p \leq 0.83$$). The median follow-up for the time-to-event endpoints over the whole trial was ≈19 months. In the placebo group, the proportions of subjects who had an acute exacerbation or died, or who had ILD progression or died, was greater in subjects with baseline BMI < 25 or ≥ 25 to < 30 than ≥ 30 kg/m2 (Table 1). The proportions of subjects with these events were similar or lower in the nintedanib than the placebo group across the subgroups by BMI, with no heterogeneity detected among the subgroups (Table 1).Fig. 1Rate of decline in FVC (mL/year) over 52 weeks by baseline BMI in the INBUILD trialTable 1Outcomes over the whole INBUILD trial in subgroups by BMI at baselineBMI < 25 kg/m2BMI ≥ 25 to < 30 kg/m2BMI ≥ 30 kg/m2Nintedanib($$n = 90$$)Placebo($$n = 98$$)Nintedanib($$n = 130$$)Placebo($$n = 112$$)Nintedanib($$n = 111$$)Placebo($$n = 121$$)Acute exacerbation or death, n (%) with event12 (13.3)24 (24.5)18 (13.8)24 (21.4)16 (14.4)17 (14.0) Hazard ratio ($95\%$ CI)0.52 (0.26, 1.04)0.66 (0.36, 1.21)0.92 (0.47, 1.83) Treatment-by-subgroup interactionp = 0.52Progression of ILDa or death, n (%) with event44 (48.9)59 (60.2)48 (36.9)61 (54.5)42 (37.8)61 (50.4) Hazard ratio ($95\%$ CI)0.74 (0.50, 1.10)0.65 (0.44, 0.94)0.64 (0.43, 0.95) Treatment-by-subgroup interactionp = 0.85Analyzed as time to first eventBMI body mass index, ILD interstitial lung diseaseaAbsolute decline in forced vital capacity % predicted ≥ $10\%$ Adverse events and dose adjustments are shown in Table 2. The adverse event profile of nintedanib was similar across subgroups by BMI, with gastrointestinal adverse events the most common events. In the nintedanib group, adverse events of diarrhea and weight decrease were more frequent in subjects with baseline BMI < 25 than ≥ 25 to < 30 or ≥ 30 kg/m2. Decreased appetite was more frequent in subjects with baseline BMI < 25 than ≥ 25 to < 30 or ≥ 30 kg/m2 in both treatment groups. In the nintedanib group, adverse events leading to dose reduction were more frequent in subjects with baseline BMI < 25 than ≥ 25 to < 30 or ≥ 30 kg/m2. Adverse events led to treatment discontinuation more frequently in subjects with baseline BMI < 25 than ≥ 25 to < 30 or ≥ 30 kg/m2 in both treatment groups. The most frequent adverse event leading to discontinuation of nintedanib was diarrhea, which occurred at rates of 4.3, 6.1 and 3.9 events per 100 patient-years in subjects with baseline BMI < 25, ≥ 25 to < 30 and ≥ 30 kg/m2, respectively. One subject in the nintedanib group (with baseline BMI < 25 kg/m2) and one subject in the placebo group (with baseline BMI ≥ 25 to < 30 kg/m2) discontinued treatment due to weight loss. Table 2Adverse events over the whole INBUILD trial in subgroups by BMI at baselineBMI < 25 kg/m2BMI ≥ 25 to < 30 kg/m2BMI ≥ 30 kg/m2Nintedanib ($$n = 90$$)Placebo ($$n = 98$$)Nintedanib ($$n = 130$$)Placebo ($$n = 112$$)Nintedanib ($$n = 111$$)Placebo ($$n = 121$$)n (%)Rate per 100patient-yearsn (%)Rate per 100patient-yearsn (%)Rate per 100patient-yearsn (%)Rate per 100patient-yearsn (%)Rate per 100patient-yearsn (%)Rate per 100patient-yearsAny adverse event(s)89 (98.9)1048.395 (96.9)435.8127 (97.7)589.3104 (92.9)288.4109 (98.2)716.1109 (90.1)281.7Most frequent adverse eventsa Diarrhea71 (78.9)199.225 (25.5)23.392 (70.8)130.929 (25.9)23.377 (69.4)112.831 (25.6)22.5 Nausea27 (30.0)30.79 (9.2)6.939 (30.0)30.511 (9.8)7.433 (29.7)30.513 (10.7)8.2 Vomiting19 (21.1)18.23 (3.1)2.221 (16.2)14.95 (4.5)3.324 (21.6)19.48 (6.6)4.8 Decreased appetite21 (23.3)22.19 (9.2)6.818 (13.8)11.97 (6.3)4.615 (13.5)10.87 (5.8)4.2 Nasopharyngitis19 (21.1)19.923 (23.5)20.221 (16.2)14.314 (12.5)9.514 (12.6)9.711 (9.1)6.9 Dyspnea14 (15.6)13.011 (11.2)8.419 (14.6)12.627 (24.1)19.419 (17.1)13.419 (15.7)12.0 Bronchitis12 (13.3)11.219 (19.4)15.521 (16.2)14.420 (17.9)14.315 (13.5)10.725 (20.7)16.4 Weight decrease19 (21.1)18.87 (7.1)5.218 (13.8)12.08 (7.1)5.212 (10.8)8.53 (2.5)1.7 ALT increased14 (15.6)13.92 (2.0)1.518 (13.8)11.95 (4.5)3.217 (15.3)12.06 (5.0)3.6 Cough8 (8.9)7.29 (9.2)6.915 (11.5)9.812 (10.7)8.117 (15.3)12.130 (24.8)20.8 Progression of ILDb9 (10.0)8.127 (27.6)22.312 (9.2)7.317 (15.2)11.17 (6.3)4.612 (9.9)7.2 AST increased15 (16.7)15.22 (2.0)1.513 (10.0)8.54 (3.6)2.615 (13.5)10.37 (5.8)4.2Adverse event(s) leading to treatment discontinuation27 (30.0)24.018 (18.4)13.227 (20.8)16.615 (13.4)9.519 (17.1)12.415 (12.4)8.8Adverse event(s) leading to dose reduction44 (48.9)63.84 (4.1)3.046 (35.4)40.02 (1.8)1.334 (30.6)28.29 (7.4)5.4Serious adverse event(s)c42 (46.7)46.457 (58.2)56.059 (45.4)44.257 (50.9)45.446 (41.4)36.750 (41.3)35.0Fatal adverse event(s)4 (4.4)3.412 (12.2)8.710 (7.7)6.016 (14.3)10.17 (6.3)4.58 (6.6)4.6Based on adverse events reported between the first trial drug intake and 28 days after the last trial drug intakeALT alanine aminotransferase, AST aspartate aminotransferase, BMI body mass index, MedDRA Medical Dictionary for Regulatory ActivitiesaAdverse events were coded based on preferred terms in the MedDRA version 22.0. Adverse events with an incidence rate of >10 events per 100 patient-years in either treatment group in the overall population are shownbCorresponded to the MedDRA preferred term “interstitial lung disease”cEvent that resulted in death, was life-threatening, resulted in hospitalization or prolonged hospitalization, resulted in persistent or clinically significant disability or incapacity, was a congenital anomaly or birth defect, or was deemed serious for any other reason
## Weight loss and outcomes
Subjects in the nintedanib group had a significantly greater decrease in weight than subjects in the placebo group over 52 weeks (Additional file 1: Table S2) and over the whole trial (Table 3). The baseline characteristics of the subgroups of subjects by weight loss ≤ $5\%$ and > $5\%$ over the whole trial were similar (Additional file 1: Table S3).Table 3Association between change in weight (slope) and risk of outcomes over the whole INBUILD trialAcute exacerbation or deathILD progressionILD progression or deathN included Placebo330327327 Nintedanib331326326Longitudinal sub-modela Estimated change in weight (kg) ($95\%$ CI) with placebo − 1.60 (− 2.08, − 1.12) − 0.76 (− 1.34, − 0.19) − 0.93 (− 1.50, − 0.37) Estimated change in weight ($95\%$ CI) difference for nintedanib vs placebo − 1.51 (− 2.19, − 0.84) − 2.07 (− 2.84, − 1.31) − 2.00 (− 2.76, − 1.24) p-value < 0.001 < 0.001 < 0.001Time-to-event sub-modelbn (%) with event Placebo64 (19.4)156 (47.7)177 (54.1) Nintedanib46 (13.9)109 (33.4)129 (39.6) Hazard ratio ($95\%$ CI) for nintedanib vs placebo0.60 (0.41, 0.89)0.68 (0.52, 0.89)0.64 (0.50, 0.81) p-value0.0110.004 < 0.001Association between change in weight (slope) and risk of event, hazard ratio ($95\%$ CI) Per 1 kg decrease1.08 (1.03, 1.14)0.96 (0.91, 1.01)1.00 (0.96, 1.04) Per 4 kg decrease1.38 (1.13, 1.68)0.83 (0.68, 1.02)1.01 (0.87, 1.18) Per 10 kg decrease2.23 (1.36, 3.65)0.64 (0.39, 1.05)1.03 (0.70, 1.53) p-value0.0020.080.88UIP usual interstitial pneumonia, ILD interstitial lung diseaseaRandom effects normal linear model of weight (kg) with HRCT pattern (UIP-like pattern or other fibrotic patterns) and weight at baseline as predictor variables, a separate slope assumed for the nintedanib and placebo groups, and trajectories modelled by a linear trend with an unstructured variance–covariance matrix assumedbProportional hazard model with a piecewise exponential baseline hazard, stratified by HRCT pattern, with treatment as a predictor variable and the endogenous time-dependent covariate of weight (kg) as estimated slope of the longitudinal response Over the whole trial, $19.4\%$ of subjects in the placebo group and $13.9\%$ of subjects in the nintedanib group had an acute exacerbation or died. There was a significant association between weight decrease and time to acute exacerbation or death over 52 weeks (Additional file 1: Table S2) and over the whole trial (Table 3). Based on the estimated slope, a 4 kg weight decrease corresponded to a 1.38-fold ($95\%$ CI 1.13, 1.68) increase in the risk of acute exacerbation or death (Fig. 2).Fig. 2Association between change in weight (slope) and risk of outcomes over the whole INBUILD trial Over the whole trial, $47.7\%$ of subjects in the placebo group and $33.4\%$ of subjects in the nintedanib group experienced ILD progression, and $54.1\%$ of subjects in the placebo group and $39.6\%$ of subjects in the nintedanib group experienced ILD progression or death. No association was detected between weight decrease and the risk of ILD progression or the risk of ILD progression or death (Table 3, Fig. 2 and Additional file 1: Table S2).
## Discussion
These analyses of data from the INBUILD trial suggest that there may be associations between baseline BMI or weight loss and clinically relevant outcomes in patients with PPF. The rate of FVC decline, and the risk of ILD progression or death, were numerically greater in subjects with baseline BMI < 25 kg/m2 than in those with higher BMI. Weight loss during the trial was associated with a significantly increased risk of acute exacerbation or death. As observed in clinical trials in patients with other ILDs [10, 30], nintedanib had a consistent effect on slowing the progression of ILD across the subgroups by baseline BMI.
Our finding that the rate of decline in FVC was greatest in subjects with baseline BMI < 25 kg/m2 is consistent with observations in subjects with IPF in the INPULSIS trials [10] and other studies in subjects with IPF and systemic sclerosis-associated ILD [13, 14]. In an analysis including data from trials of pirfenidone in patients with IPF, the annualized decline in FVC % predicted was greater in patients with baseline BMI < 25 kg/m2 than BMI ≥ 25 to < 30 or ≥ 30 kg/m2 in the placebo groups, but this was not observed in patients who received pirfenidone [13]. Our finding that the risk of ILD progression or death was numerically greater in subjects with baseline BMI < 25 kg/m2 is consistent with observations from previous studies showing higher mortality in patients with pulmonary fibrosis who have lower BMI [7–9, 12]. The reasons why low BMI is associated with worse outcomes in patients with ILDs are not understood, but may be related to loss of muscle mass [31, 32] or to increased levels of pro-inflammatory cytokines, such as tumor necrosis factor, which have been associated with weight loss in animal studies [33].
Various approaches can be used to investigate associations between a longitudinal measure such as weight and the risk of an outcome. In these analyses, we used a joint modelling approach, as this enables longitudinal markers and time-to-event endpoints to be analyzed simultaneously, overcoming issues of bias and measurement error that occur when repeated measurements and outcomes are analyzed separately, when analyses are based on post-baseline subgroups, or when analyses do not consider longitudinal endpoints as endogenous time-varying factors [34–37]. This joint modelling also enabled us to evaluate the validity of weight change as a surrogate endpoint for the time-to-event endpoints according to the three levels of surrogacy defined by Taylor and Elston [38]. The significant association between the outcome of weight change and the risk of acute exacerbation or death fulfils Taylor and Elston’s criteria for surrogacy at level two, but further validation is required. While several previous studies have shown that weight loss is associated with a greater risk of mortality in patients with pulmonary fibrosis [5, 7, 9, 10, 13, 18–21], we are not aware of prior studies suggesting an association between weight loss and acute exacerbations of ILD.
The adverse event profile of nintedanib was generally similar across the subgroups by baseline BMI, but adverse events of diarrhea, decreased appetite and weight decrease, and adverse events leading to dose reduction and treatment discontinuation, were more frequent in subjects who had a baseline BMI < 25 kg/m2 than a higher BMI. In the INPULSIS, SENSCIS and INBUILD trials in subjects with pulmonary fibrosis, the proportion of subjects with adverse events of weight loss over 52 weeks ranged from 9.7 to $12.3\%$ in the nintedanib groups compared to 3.3 to $4.2\%$ in the placebo groups [6, 24, 39]. The reported proportion of patients with IPF treated with nintedanib who experience weight loss in real-world studies is highly variable, likely reflecting differences in methodology and the patient populations studied [40–42]. Clinicians should be aware of weight loss as a potential adverse event of nintedanib, particularly in patients with low BMI, and consider nutritional interventions when required. Management of gastrointestinal adverse events associated with nintedanib therapy using symptomatic therapies such as loperamide and dose adjustment is important to minimize their impact and help patients remain on treatment [43, 44].
Strengths of our analyses include the robust collection of data on FVC and weight in the setting of a clinical trial, and the use of a joint modelling approach to assess the associations between weight loss and outcomes [34–37]. Limitations include that BMI is limited as a measure of nutritional status [45, 46] and that the subgroups with different BMI differed in factors beyond BMI. The lowest BMI subgroup included a greater proportion of subjects with autoimmune disease-related ILDs, which are associated with gastrointestinal complications that may lead to weight loss, and this may have influenced the risk of outcomes across the subgroups. The numbers of subjects with individual ILD diagnoses were too small for these subgroups to be analyzed separately. The number of subjects who were underweight was too small for this group to be analyzed separately. The number of acute exacerbations available for use in the time-to-event analyses was quite small. There were too few deaths for associations between weight change and death alone to be analyzed. Our analyses did not establish cause and effect. These analyses were post-hoc and should be considered exploratory.
## Conclusions
In conclusion, these analyses of data from the INBUILD trial suggest that in subjects with PPF, lower BMI at baseline and weight loss may be associated with worse outcomes. Nintedanib had a consistent effect on slowing ILD progression across subgroups by baseline BMI. Physicians should monitor weight in patients with PPF and consider interventions where necessary.
## Supplementary Information
Additional file 1: Table S1. Baseline characteristics in subgroups by body mass index (BMI) in the INBUILD trial. Table S2. Association between change in weight (slope) and risk of outcomes over 52 weeks in the INBUILD trial. Table S3. Baseline characteristics of subgroups of subjects by weight loss ≤$5\%$ and >$5\%$ over the whole INBUILD trial.
## References
1. Raghu G, Remy-Jardin M, Richeldi L, Thomson CC, Inoue Y, Johkoh T. **Idiopathic pulmonary fibrosis (an update) and progressive pulmonary fibrosis in adults: an official ATS/ERS/JRS/ALAT clinical practice guideline**. *Am J Respir Crit Care Med* (2022.0) **205** e18-47. DOI: 10.1164/rccm.202202-0399ST
2. Paterniti MO, Bi Y, Rekić D, Wang Y, Karimi-Shah BA, Chowdhury BA. **Acute exacerbation and decline in forced vital capacity are associated with increased mortality in idiopathic pulmonary fibrosis**. *Ann Am Thorac Soc* (2017.0) **14** 1395-1402. DOI: 10.1513/AnnalsATS.201606-458OC
3. Brown KK, Martinez FJ, Walsh SLF, Thannickal VJ, Prasse A, Schlenker-Herceg R. **The natural history of progressive fibrosing interstitial lung diseases**. *Eur Respir J* (2020.0) **55** 2000085. DOI: 10.1183/13993003.00085-2020
4. Nasser M, Larrieu S, Si-Mohamed S, Ahmad K, Boussel L, Brevet M. **Progressive fibrosing interstitial lung disease: a clinical cohort (the PROGRESS study)**. *Eur Respir J* (2021.0) **57** 2002718. DOI: 10.1183/13993003.02718-2020
5. Pugashetti J, Graham J, Boctor N, Mendez C, Foster E, Juarez M. **Weight loss as a predictor of mortality in patients with interstitial lung disease**. *Eur Respir J* (2018.0) **52** 1801289. DOI: 10.1183/13993003.01289-2018
6. Distler O, Highland KB, Gahlemann M, Azuma A, Fischer A, Mayes MD. **Nintedanib for systemic sclerosis-associated interstitial lung disease**. *N Engl J Med* (2019.0) **380** 2518-2528. DOI: 10.1056/NEJMoa1903076
7. Comes A, Wong AW, Fisher JH, Morisset J, Johannson KA, Farrand E. **Association of BMI and change in weight with mortality in patients with fibrotic interstitial lung disease**. *Chest* (2022.0) **161** 1320-1329. DOI: 10.1016/j.chest.2021.11.008
8. Alakhras M, Decker PA, Nadrous HF, Collazo-Clavell M, Ryu JH. **Body mass index and mortality in patients with idiopathic pulmonary fibrosis**. *Chest* (2007.0) **131** 1448-1453. DOI: 10.1378/chest.06-2784
9. Kishaba T, Nagano H, Nei Y, Yamashiro S. **Body mass index—percent forced vital capacity—respiratory hospitalization: new staging for idiopathic pulmonary fibrosis patients**. *J Thorac Dis* (2016.0) **8** 3596-3604. DOI: 10.21037/jtd.2016.12.49
10. Jouneau S, Crestani B, Thibault R, Lederlin M, Vernhet L, Valenzuela C. **Analysis of body mass index, weight loss and progression of idiopathic pulmonary fibrosis**. *Respir Res* (2020.0) **21** 312. DOI: 10.1186/s12931-020-01528-4
11. Awano N, Jo T, Yasunaga H, Inomata M, Kuse N, Tone M. **Body mass index and in-hospital mortality in patients with acute exacerbation of idiopathic pulmonary fibrosis**. *ERJ Open Res* (2021.0) **7** 00037-2021. DOI: 10.1183/23120541.00037-2021
12. Jouneau S, Rousseau C, Lederlin M, Lescoat A, Kerjouan M, Chauvin P. **Malnutrition and decreased food intake at diagnosis are associated with hospitalization and mortality of idiopathic pulmonary fibrosis patients**. *Clin Nutr* (2022.0) **41** 1335-1342. DOI: 10.1016/j.clnu.2022.05.001
13. Jouneau S, Crestani B, Thibault R, Lederlin M, Vernhet L, Yang M. **Post hoc analysis of clinical outcomes in placebo- and pirfenidone-treated patients with IPF stratified by BMI and weight loss**. *Respiration* (2022.0) **101** 142-154. DOI: 10.1159/000518855
14. Nagy A, Palmer E, Polivka L, Eszes N, Vincze K, Barczi E. **Treatment and systemic sclerosis interstitial lung disease outcome: the overweight paradox**. *Biomedicines* (2022.0) **10** 434. DOI: 10.3390/biomedicines10020434
15. Doubková M, Švancara J, Svoboda M, Šterclová M, Bartoš V, Plačková M. **EMPIRE Registry, Czech part: impact of demographics, pulmonary function and HRCT on survival and clinical course in idiopathic pulmonary fibrosis**. *Clin Respir J* (2018.0) **12** 1526-1535. DOI: 10.1111/crj.12700
16. Snyder L, Neely ML, Hellkamp AS, O'Brien E, de Andrade J, Conoscenti CS. **Predictors of death or lung transplant after a diagnosis of idiopathic pulmonary fibrosis: insights from the IPF-PRO Registry**. *Respir Res* (2019.0) **20** 105. DOI: 10.1186/s12931-019-1043-9
17. Brown KK, Inoue Y, Flaherty KR, Martinez FJ, Cottin V, Bonella F. **Predictors of mortality in subjects with progressive fibrosing interstitial lung diseases**. *Respirology* (2022.0) **27** 294-300. DOI: 10.1111/resp.14231
18. Nakatsuka Y, Handa T, Kokosi M, Tanizawa K, Puglisi S, Jacob J. **The clinical significance of body weight loss in idiopathic pulmonary fibrosis patients**. *Respiration* (2018.0) **96** 338-347. DOI: 10.1159/000490355
19. Kulkarni T, Yuan K, Tran-Nguyen TK, Kim YI, de Andrade JA, Luckhardt T. **Decrements of body mass index are associated with poor outcomes of idiopathic pulmonary fibrosis patients**. *PLoS ONE* (2019.0) **14** e0221905. DOI: 10.1371/journal.pone.0221905
20. Kalininskiy A, Rackow AR, Nagel D, Croft D, McGrane-Minton H, Kottmann RM. **Association between weight loss and mortality in idiopathic pulmonary fibrosis**. *Respir Res* (2022.0) **23** 377. DOI: 10.1186/s12931-022-02277-2
21. Kim TH, Shin YY, Kim HJ, Song MJ, Kim YW, Lim SY. **Impact of body weight change on clinical outcomes in patients with idiopathic pulmonary fibrosis receiving pirfenidone**. *Sci Rep* (2022.0) **12** 17397. DOI: 10.1038/s41598-022-22449-w
22. Sekine A, Wasamoto S, Hagiwara E, Yamakawa H, Ikeda S, Okabayashi H. **Beneficial impact of weight loss on respiratory function in interstitial lung disease patients with obesity**. *Respir Investig* (2021.0) **59** 247-251. DOI: 10.1016/j.resinv.2020.10.002
23. Schaeffer MR, Kumar DS, Assayag D, Fisher JH, Johannson KA, Khalil N. **Association of BMI with pulmonary function, functional capacity, symptoms, and quality of life in ILD**. *Respir Med* (2022.0) **195** 106792. DOI: 10.1016/j.rmed.2022.106792
24. Flaherty KR, Wells AU, Cottin V, Devaraj A, Walsh SLF, Inoue Y. **Nintedanib in progressive fibrosing interstitial lung diseases**. *N Engl J Med* (2019.0) **381** 1718-1727. DOI: 10.1056/NEJMoa1908681
25. Flaherty KR, Wells AU, Cottin V, Devaraj A, Inoue Y, Richeldi L. **Nintedanib in progressive interstitial lung diseases: data from the whole INBUILD trial**. *Eur Respir J* (2022.0) **59** 2004538. DOI: 10.1183/13993003.04538-2020
26. Cottin V, Martinez FJ, Jenkins RG, Belperio JA, Kitamura H, Molina-Molina M. **Safety and tolerability of nintedanib in patients with progressive fibrosing interstitial lung diseases: data from the randomized controlled INBUILD trial**. *Respir Res* (2022.0) **23** 85. DOI: 10.1186/s12931-022-01974-2
27. Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM. **Statistics in medicine–reporting of subgroup analyses in clinical trials**. *N Engl J Med* (2007.0) **357** 2189-2194. DOI: 10.1056/NEJMsr077003
28. 28.European Medicines Agency. Guideline on the investigation of subgroups in confirmatory clinical trials. 2019. EMA/CHMP/539146/2013. https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-investigation-subgroups-confirmatory-clinical-trials_en.pdf.
29. Garcia-Hernandez A, Rizopoulos D. **%JM: a SAS macro to fit jointly generalized mixed models for longitudinal data and time-to-event responses**. *J Stat Softw* (2018.0) **84** 29. DOI: 10.18637/jss.v084.i12
30. 30.Jouneau S, Lescoat A, Crestani B, Riemekasten G, Kondoh Y, Smith V et al. Efficacy and safety of nintedanib in patients with systemic sclerosis-associated interstitial lung disease (SSc-ILD) by body mass index (BMI) at baseline: subgroup analysis of the SENSCIS trial. Poster presented at the American Thoracic Society International Conference. 2020. https://www.usscicomms.com/respiratory/ATS2020/jouneau.
31. Moon SW, Choi JS, Lee SH, Jung KS, Jung JY, Kang YA. **Thoracic skeletal muscle quantification: low muscle mass is related with worse prognosis in idiopathic pulmonary fibrosis patients**. *Respir Res* (2019.0) **20** 35. DOI: 10.1186/s12931-019-1001-6
32. Nakano A, Ohkubo H, Taniguchi H, Kondoh Y, Matsuda T, Yagi M. **Early decrease in erector spinae muscle area and future risk of mortality in idiopathic pulmonary fibrosis**. *Sci Rep* (2020.0) **10** 2312. DOI: 10.1038/s41598-020-59100-5
33. Tracey KJ, Wei H, Manogue KR, Fong Y, Hesse DG, Nguyen HT. **Cachectin/tumor necrosis factor induces cachexia, anemia, and inflammation**. *J Exp Med* (1988.0) **167** 1211-1227. DOI: 10.1084/jem.167.3.1211
34. Arisido MW, Antolini L, Bernasconi DP, Valsecchi MG, Rebora P. **Joint model robustness compared with the time-varying covariate Cox model to evaluate the association between a longitudinal marker and a time-to-event endpoint**. *BMC Med Res Methodol* (2019.0) **19** 222. DOI: 10.1186/s12874-019-0873-y
35. Papageorgiou G, Mauff K, Tomer A, Rizopoulos D. **An overview of joint modeling of time-to-event and longitudinal outcomes**. *Annu Rev Stat Appl* (2019.0) **6** 223-240. DOI: 10.1146/annurev-statistics-030718-105048
36. Arbeeva L, Nelson AE, Alvarez C, Cleveland RJ, Allen KD, Golightly YM. **Application of traditional and emerging methods for the joint analysis of repeated measurements with time-to-event outcomes in rheumatology**. *Arthritis Care Res (Hoboken)* (2020.0) **72** 615-621. DOI: 10.1002/acr.23881
37. Chen Y, Postmus D, Cowie MR, Woehrle H, Wegscheider K, Simonds AK. **Using joint modelling to assess the association between a time-varying biomarker and a survival outcome: an illustrative example in respiratory medicine**. *Eur Respir J* (2021.0) **57** 2003206. DOI: 10.1183/13993003.03206-2020
38. Taylor RS, Elston J. **The use of surrogate outcomes in model-based cost-effectiveness analyses: a survey of UK health technology assessment reports**. *Health Technol Assess* (2009.0) **13** iii ix-xi, 1-50. DOI: 10.3310/hta13080
39. Corte T, Bonella F, Crestani B, Demedts MG, Richeldi L, Coeck C. **Safety, tolerability and appropriate use of nintedanib in idiopathic pulmonary fibrosis**. *Respir Res* (2015.0) **16** 116. DOI: 10.1186/s12931-015-0276-5
40. Brunnemer E, Wälscher J, Tenenbaum S, Hausmanns J, Schulze K, Seiter M. **Real-world experience with nintedanib in patients with idiopathic pulmonary fibrosis**. *Respiration* (2018.0) **95** 301-309. DOI: 10.1159/000485933
41. Antoniou K, Markopoulou K, Tzouvelekis A, Trachalaki A, Vasarmidi E, Organtzis J. **Efficacy and safety of nintedanib in a Greek multicentre idiopathic pulmonary fibrosis registry: a retrospective, observational, cohort study**. *ERJ Open Res* (2020.0) **6** 00172-2019. DOI: 10.1183/23120541.00172-2019
42. Noor S, Nawaz S, Chaudhuri N. **Real-world study analysing progression and survival of patients with idiopathic pulmonary fibrosis with preserved lung function on antifibrotic treatment**. *Adv Ther* (2021.0) **38** 268-277. DOI: 10.1007/s12325-020-01523-7
43. Bendstrup E, Wuyts W, Alfaro T, Chaudhuri N, Cornelissen R, Kreuter M. **Nintedanib in idiopathic pulmonary fibrosis: practical management recommendations for potential adverse events**. *Respiration* (2019.0) **97** 173-184. DOI: 10.1159/000495046
44. 44.Boehringer Ingelheim. Ofev® (nintedanib capsules) prescribing information. 2022. https://docs.boehringer-ingelheim.com/Prescribing%20Information/PIs/Ofev/ofev.pdf.
45. Madden AM, Smith S. **Body composition and morphological assessment of nutritional status in adults: a review of anthropometric variables**. *J Hum Nutr Diet* (2016.0) **29** 7-25. DOI: 10.1111/jhn.12278
46. Jouneau S, Kerjouan M, Rousseau C, Lederlin M, Llamas-Guttierez F, De Latour B. **What are the best indicators to assess malnutrition in idiopathic pulmonary fibrosis patients? A cross-sectional study in a referral center**. *Nutrition* (2019.0) **62** 115-121. DOI: 10.1016/j.nut.2018.12.008
|
---
title: Plasma gelsolin levels are associated with diabetes, sex, race, and poverty
authors:
- Nicole Noren Hooten
- Nicolle A. Mode
- Edward Kowalik
- Victor Omoniyi
- Alan B. Zonderman
- Ngozi Ezike
- Mark J. DiNubile
- Susan L. Levinson
- Michele K. Evans
journal: Journal of Translational Medicine
year: 2023
pmcid: PMC9999548
doi: 10.1186/s12967-023-04026-5
license: CC BY 4.0
---
# Plasma gelsolin levels are associated with diabetes, sex, race, and poverty
## Abstract
### Background
The growing epidemic of the inflammation-related metabolic disease, type 2 diabetes mellitus, presents a challenge to improve our understanding of potential mechanisms or biomarkers to prevent or better control this age-associated disease. A gelsolin isoform is secreted into the plasma as part of the extracellular actin scavenger system which serves a protective role by digesting and removing actin filaments released from damaged cells. Recent data indicate a role for decreased plasma gelsolin (pGSN) levels as a biomarker of inflammatory conditions. Extracellular vesicles (EVs), a heterogeneous group of cell-derived membranous structures involved in intercellular signaling, have been implicated in metabolic and inflammatory diseases including type 2 diabetes mellitus. We examined whether pGSN levels were associated with EV concentration and inflammatory plasma proteins in individuals with or without diabetes.
### Methods
We quantified pGSN longitudinally ($$n = 104$$) in a socioeconomically diverse cohort of middle-aged African American and White study participants with and without diabetes mellitus. Plasma gelsolin levels were assayed by ELISA. EV concentration (sub-cohort $$n = 40$$) was measured using nanoparticle tracking analysis. Inflammatory plasma proteins were assayed on the SomaScan® v4 proteomic platform.
### Results
pGSN levels were lower in men than women. White individuals with diabetes had significantly lower levels of pGSN compared to White individuals without diabetes and to African American individuals either with or without diabetes. For adults living below poverty, those with diabetes had lower pGSN levels than those without diabetes. Adults living above poverty had similar pGSN levels regardless of diabetes status. No correlation between EV concentrations and pGSN levels was identified (r = − 0.03; $$p \leq 0.85$$). Large-scale exploratory plasma protein proteomics revealed 47 proteins that significantly differed by diabetes status, 19 of which significantly correlated with pGSN levels, including adiponectin.
### Conclusions
In this cohort of racially diverse individuals with and without diabetes, we found differences in pGSN levels with diabetes status, sex, race, and poverty. We also report significant associations of pGSN with the adipokine, adiponectin, and other inflammation- and diabetes-related proteins. These data provide mechanistic insights into the relationship of pGSN and diabetes.
## Background
The growing epidemic of type 2 diabetes is rapidly becoming a world-wide problem. In 2019 it was estimated that 463 million adults were living with diabetes mellitus [1]. These numbers are forecasted to rise by $51\%$ in the next 26 years [1], indicating that diabetes mellitus is projected to continue to be a global public health problem. Therefore, it is important to fully understand the underlying mechanisms that drive diabetes and pursue novel therapeutic modalities to combat this prevalent disease. Diabetes is characterized by impaired insulin secretion coupled with insulin resistance. The pathophysiology of insulin resistance involves inflammatory mechanisms, mitochondrial dysfunction, and hormonal dysregulation (e.g., adipokines) [2]. Thus, examining plasma proteins that may play a role in inflammatory processes may advance our understanding of the risk profiles for this metabolic age-associated disease.
Gelsolin exists as both a cytoplasmic and extracellular form, but only the extracellular form contains a “plasma extension” 24 amino acid sequence and a disulfide bond that augments stability of this isoform in the extracellular environment [3, 4]. This extracellular isoform, called plasma gelsolin (pGSN), is part of the extracellular actin scavenger system (EASS) and functions to continuously scavenge and remove actin filaments in the extracellular space, including the circulation. Cellular injury can release cellular debris, including actin filaments and DNA, into the extracellular space. To avoid the toxicity of insoluble F-actin filaments, the increase in blood viscosity, platelet activation, and induction of proinflammatory molecules, compensatory mechanisms are in place to scavenge and clear extracellular actin as part of the EASS. pGSN functions to depolymerize actin by severing and capping actin filaments, thereby facilitating its removal from the circulation [3, 4]. Actin inhibits DNase and gelsolin disinhibits DNase; thus, gelsolin may also promote the clearance of cell-free DNA [5].
Recent data suggest that pGSN may have a vital role in both normal physiological and pathological processes. *In* general, lower levels of pGSN have been reported in response to trauma, sepsis, chronic inflammatory diseases including rheumatoid arthritis, specific neurological disorders including Alzheimer’s disease and multiple sclerosis, and some cancers [4, 6]. Substantial interest has arisen in utilizing decreased pGSN as a prognostic biomarker [7, 8]. Given the beneficial role of pGSN repletion in diverse animal models, a recombinant human pGSN is actively being developed for therapeutic use in a diverse variety of infectious and other inflammatory conditions.
Small exploratory studies have reported differences in pGSN with diabetes. A study using liquid chromatography-mass spectrometry (LC–MS)–based proteomics to identify plasma proteins reported lower levels of pGSN in adolescent/young adults (average age 19.4 yrs) with type 1 diabetes [9]. This study also found lower levels of pGSN in adolescents (average age 14.4 yrs) with type 2 diabetes compared to healthy controls (average age 20.7 yrs) [9]. In a small cohort of obese women with ($$n = 6$$) or without ($$n = 6$$) diabetes, pGSN levels were lower after bariatric surgery [10]. Another recent study reported lower levels of pGSN in women with type 2 diabetes ($$n = 10$$) compared to women without type 2 diabetes ($$n = 8$$) [11]. Similar results were reported for men with type 2 diabetes ($$n = 17$$) compared to men without type 2 diabetes ($$n = 7$$). In this study, it was also found that two different mouse models of diabetes, db/db (homozygous for the diabetes spontaneous mutation (Leprdb)) mice and mice on a High Fat Diet (HFD) + Streptozotocin, also reported reduced pGSN compared to control mice [11]. Therefore, these studies together suggest that there may be differences in pGSN levels with and without diabetes, but more studies are warranted to confirm these findings. For example, few studies have measured pGSN levels over time and in a large, diverse cohort of middle-aged adults.
Extracellular vesicles (EVs) represent another tool to examine underlying pathophysiology of diabetes mellitus [12]. EVs are a diverse group of cell-derived membranous structures that include exosomes, microvesicles, and apoptotic bodies that are released by cells [13, 14]. In decompression illness, microparticles increase with an associated drop in pGSN levels [15]. EVs may be found in numerous bodily fluids including plasma, serum, and urine [13, 14]. They contain molecular cargo, including several types of RNAs, nuclear DNA, mitochondrial DNA, proteins, and lipids [14–19]. EVs mediate intercellular communication and are also associated with multiple immune, metabolic, and neoplastic diseases including diabetes mellitus [12–14, 20]. Our previous work found higher levels of circulating EVs in White adults with diabetes compared to those without diabetes or to African Americans with or without diabetes, suggesting that EVs may play a role in diabetes mellitus differentially in certain demographic groups [21].
In this study, we examined the association of pGSN with diabetes mellitus in a large, diverse cohort of African American and White adults with or without diabetes mellitus. We also investigated whether there was a relationship between EV concentration and pGSN in diabetic individuals. Additionally, we examined how the social determinants of health, race and poverty, affect pGSN levels. Levels of pGSN were measured at two different time points to determine whether pGSN levels changed over time in this cohort.
## Study sample
Participants were chosen from the Healthy Neighborhoods of Diversity across the Life Span (HANDLS) study of the National Institute on Aging Intramural Research Program, National Institutes of Health. HANDLS is an ongoing, prospective study that examines how race and socioeconomic status influence aging and age-related disease [22]. The study consists of African American and White adults aged 30–64 at baseline living in Baltimore, Maryland, USA. The baseline accrual occurred between 2004 and 2009. Race was self-identified as African American or White. Participants were either above or below poverty as defined by $125\%$ of the 2004 U.S. Health and Human Services Poverty Guidelines at enrollment. Participants provided written informed consent. The HANDLS study is approved by the Institutional Review Board of the National Institutes of Health.
Each HANDLS visit occurs approximately every 5 years and consists of physical exams, structured medical history interviews, questionnaires, and collection of blood samples. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2). Individuals were classified based on BMI as underweight/normal (< 25 kg/m2), overweight/obese class I (25 to < 35 kg/m2), and obese class II/III (≥ 35 kg/m2). Individuals with diabetes met one of the following three criteria: [1] self-reported previous diagnosis by health care provider, [2] currently taking medication for diabetes, or [3] fasting serum glucose ≥ 126 mg/dL. For the study sample, we chose participants who had fasting blood samples at two times ~ 5 years apart. These time points corresponded to wave 3 and wave 4 of the HANDLS study, which we will refer to here as time 1 and time 2. We randomly selected 26 African American and 26 White participants with diabetes at time 1, and randomly selected 26 African American and 26 White euglycemic participants at time 1 who were matched on BMI category and 10-yr age groups. Participants were aged 40 and older at time 1 and were excluded if diagnosed with HIV. Therefore, the final cohort consisted of 104 participants: 52 euglycemic adults and 52 individuals with diabetes with longitudinal samples at two time points (Table 1). Among euglycemic participants living below poverty, $47\%$ were African American; among those living above poverty, $54\%$ were African American. Among participants with diabetes living below poverty, $47\%$ were African American; among those living above poverty $51\%$ were African American. Table 1HANDLS longitudinal subcohort for plasma gelsolin measurementsSubcohortEuglycemic ($$n = 52$$)Diabetes ($$n = 52$$)p-valueaMen, N (%)18 [35]24 [46]0.318African American, N (%)26 [50]26 [50]1.000Below Poverty, N (%)22 [42]29 [56]0.239Age at Time 1, mean (SD)55.5 (7.5)55.9 (6.4)0.778pGSN at Time 1, mean (SD)44.0 (7.9)40.2 (7.7)0.015BMI, mean (SD)30.7 (6.7)32.9 (9.0)0.158Data by time pointTime 1Time 2p-valueaAge, mean (SD)55.7 (7.0)59.6 (7.3) < 0.001BMI, mean (SD)31.8 (7.9)31.9 (8.4)0.943pGSN, mean (SD)42.1 (8.0)42.8 (9.1)0.565aTest of differences for categorial variables used the Pearson Chi-squared test. Test of differences for continuous variables used a Student’s t-testpGSN: plasma gelsolin; SD: standard deviation
## Gelsolin measurement
Fasting blood samples were collected between 9:30 and 11:30 a.m. into sodium heparin blood collection tubes. Histopaque®-1077 (2.5 ml) was added to the bottom of the tube containing 5 ml blood and centrifuged for 20 min at 610g at 4 °C without the brake. Tubes were visually inspected for proper separation and hemolysis. The top layer containing plasma was aliquoted and stored at − 80 °C.
Gelsolin was measured by BioAegis® Therapeutics by a proprietary ELISA using a primary antibody raised against the N terminal tail specific to plasma gelsolin [8]. Samples from the different visits were run on the same plate for each individual, with the exception of nine samples. Plates were balanced to contain individuals across race and diabetes diagnosis across all plates. Each sample was run in triplicate, and the sample average was used. The coefficient of variance (CV) for sample replicates was $3.73\%$ (time 1) and $3.49\%$ (time 2). Controls (QC) were produced by spiking recombinant human plasma gelsolin into a plasma sample at three different concentrations 10 ng/ml, 50 ng/ml and 100 ng/ml; the same donor plasma was used for the entire study. In addition, the 50 ng/ml spike-in control (QC-M) was aliquoted, stored at − 80 °C and run on each subsequent plate. The mean intra-assay CV was $3.64\%$ and $3.99\%$ for QC and QC-M respectively. The mean inter-assay CV was $6.05\%$ and $6.19\%$, respectively.
## Plasma EV isolation
Of the initial pGSN cohort, we chose a subset of 20 White participants with diabetes and 20 White participants without diabetes to examine plasma EV concentration. Plasma (0.5 ml) from fasting participants was thawed on ice and EVs were isolated using ExoQuickTM Exosome Precipitation Solution (System Bioscience Inc.). Details about EV isolation have been described previously [23]. The EV pellet was resuspended in PBS and diluted 1:300 in PBS for NTA. All samples were then stored at − 80 °C until use.
## Nanoparticle tracking analysis
Diluted EVs were thawed, and concentration was measured using nanoparticle tracking analysis on a NanoSight NS500 (Malvern Instruments Ltd.) according to manufacturer’s instructions by a single operator, blinded to sample identity. For each EV sample, five 20 s videos were recorded at camera level = 15. Analysis was performed using NTA 3.4 Build 3.4.4 software at detection level = 4. Calculations for plasma EV concentration were described previously [23].
## SomaScan®
Of the initial pGSN cohort, we chose a subset of White participants ($$n = 18$$) with diabetes and White participants without diabetes ($$n = 18$$) to utilize plasma samples for SomaScan assays. These individuals were also in our EV subcohort. Plasma samples were run on the SomaScan® v4 proteomic platform by SomaLogic, Inc (Boulder, CO). Quality control, sample calibration and normalization were performed by SomaLogic. Only markers for human proteins were considered. We then estimated the limit of detection (LoD) using SomaScan Assay qualification that variance for buffer samples is approximately the same as variance for low protein quantity samples [24]. Markers with ≥ $30\%$ of samples below the LoD were removed ($3\%$). Relative fluorescence units (RFU) for each protein were then log10 transformed and centered and scaled into z-scores. A total of 7069 markers were available for analysis.
## Statistics
Statistical analyses were performed using R, version 4.2.0 [25]. Differences between groups were tested using Student's t-test for continuous variables and chi-squared tests of independence for categorical variables. Linear mixed model regression (package lme4) examined the associations of diabetes status with longitudinal measurements of gelsolin accounting for race, sex, poverty status, BMI category and age. Backward elimination was used for model building starting with the full model of interactions among diabetes status with race, sex, age, and poverty status. Significance of fixed factors were determined by log likelihood ratio tests. Statistical significance was defined as p-value < 0.05.
SomaScan proteins which met the data criteria (see SomaScan section) were analyzed using Student’s t-tests to compare mean RFU values between participants with diabetes and euglycemic participants. Effect size was determined using Cohen’s D. Significant markers were those which resulted in a p-value < 0.01 and an effect size ≥ 1. Pearson’s correlation (r) was used to examine the relationship between protein markers and pGSN.
## Results
To examine the relationship between pGSN levels and diabetes, we designed a longitudinal cohort of African American and White individuals with or without diabetes (Table 1). Overall, 104 HANDLS participants with or without diabetes were included in this study, matched on race, sex, BMI category and 10-yr age groups (Table 1). Per the matched design, there were no differences between groups in the distributions of age and BMI (Table 1). There were also no differences in sex, race, and poverty status. Euglycemic participants had significantly higher pGSN levels than diabetic participants. Overall, pGSN levels were 42.08 ± 7.99 µg/ml and 42.77 ± 9.12 µg/ml at time 1 and 2, respectively.
Using linear mixed model regression, we found that pGSN levels were significantly lower in men than women ($$p \leq 0.007$$, Fig. 1A). We also found a significant race and diabetes interaction ($$p \leq 0.012$$; Fig. 1B). White individuals with diabetes had significantly lower levels of pGSN compared to White individuals without diabetes or compared to African American individuals either with or without diabetes. pGSN levels were similar for African American participants with and without diabetes. There was also a significant interaction between poverty status and diabetes status ($$p \leq 0.033$$; Fig. 1C). For adults living below poverty, those with diabetes had lower pGSN levels than those without diabetes. Adults living above poverty had similar pGSN levels regardless of diabetes status. There were no significant interactions with age, indicating that the rates of change were similar regardless of demographics or diabetes status. Fig. 1Association of plasma gelsolin, diabetes status and demographic variables. Plasma gelsolin (pGSN) levels were measured in a cohort of 104 euglycemic (Eu) individuals and 104 individuals with diabetes at 2 different time points ~ 5 years apart. Linear mixed model regression accounting for repeated measurements was used to analyze the relationship of pGSN, diabetes status, sex (A), race (B), and poverty status (C). AfrAm: African American To gain a better understanding of the mechanisms that may account for differences in pGSN levels in individuals with diabetes, we first examined whether pGSN levels were related to circulating levels of extracellular vesicles (EVs). Previously in a different cohort, we found that plasma EVs were higher in White individuals with diabetes compared to White individuals without diabetes or compared to African American individuals either with or without diabetes [21]. Here, we speculated that EVs may function as physical “sponges” resulting in reduced levels of pGSN. From our pGSN cohort, we chose to isolate EVs from White individuals as this group displayed differences in pGSN levels with diabetes status. Plasma EVs were isolated from 20 White individuals with diabetes and 20 euglycemic White individuals. We found that there was no correlation between EV concentration and pGSN (r = − 0.03; $$p \leq 0.85$$). In light of significant differences in each of these variables by diabetes status, the lack of correlation between EV and pGSN levels may be due to the different tests employed. Correlation assesses the preservation of rank order between continuous variables, while tests between diabetes groups assess overall mean differences. Therefore, circulating EVs levels are not related to pGSN levels in this subcohort.
We further investigated the relationship of pGSN to other circulating proteins by performing an exploratory, unbiased proteomic approach to examine > 7000 circulating proteins using SomaScan®, which is a sensitive, reproducible assay with high specificity for detecting proteins in a large dynamic range. Data processing details are listed in Materials and Methods. We found 47 proteins that were significantly different between euglycemic individuals and individuals with diabetes (Table 2). Of these 47, we then examined if any of these proteins were correlated with pGSN levels. We found that there were 19 proteins that were significantly correlated with pGSN (indicated in bold in Table 2). Except for insulin-like peptide INSL6, these proteins were positively correlated with pGSN levels (Table 2). One of these proteins, adiponectin, is an adipokine that is secreted from adipocytes and plays a key role in metabolic processes and in obesity-related diseases, including diabetes [26].Table 2Plasma proteins that differ significantly by diabetes status and their correlation with plasma gelsolinProtein nameEntrez gene symbolDiabetest-testp-valueEffect sizeCorrelation (r) with pGSNCorrelation p-valueAnthrax toxin receptor 2ANTXR20.00141.01420.60380.0001Matrix-remodeling-associated protein 8:Extracellular domainMXRA80.00051.09240.56920.0003Alpha-1,6-mannosylglycoprotein 6-beta-N-acetylglucosaminyltransferase AMGAT510.00071.06970.56550.0003Cartilage oligomeric matrix proteinCOMP0.00071.06890.55950.0004Alpha-1,6-mannosylglycoprotein 6-beta-N-acetylglucosaminyltransferase AMGAT520.00051.08880.51440.0013Histone-lysine N-methyltransferase EHMT2EHMT20.00131.01830.46490.0043Transmembrane protein 132C:Extracellular domainTMEM132C0.00031.12850.45600.0052AdiponectinADIPOQ0.00011.22300.45130.0057Brevican core proteinBCAN0.00071.06290.43340.0083Vesicular, overexpressed in cancer, prosurvival protein 1VOPP10.00031.12780.42990.0089Iduronate 2-sulfataseIDS0.00151.00720.41830.011172 kDa type IV collagenaseMMP20.00081.05380.38610.0200Peptidyl-glycine alpha-amidating monooxygenasePAM0.00071.07290.36250.0298Dihydrolipoyl dehydrogenase, mitochondrialDLD30.00001.24800.36020.03091-phosphatidylinositol-4,5-bisphosphate phosphodiesterase delta-1PLCD10.00171.00160.35990.0311ProsaposinPSAP0.00131.04600.35950.0313Neurogenic locus notch homolog protein 3NOTCH30.00081.06830.34630.0386AcetylcholinesteraseACHE0.00091.04240.33470.0460Complement C1q tumor necrosis factor-related protein 5C1QTNF50.00151.00950.32900.0501Calsyntenin-1CLSTN10.00111.04110.32660.0518Glypican-3GPC30.00161.00060.32290.0548Protein Wnt-5bWNT5B0.00031.12350.31260.0634Glutathione reductase, mitochondrialGSR0.00161.00560.30060.0748Dihydrolipoyl dehydrogenase, mitochondrialDLD40.00021.14950.26510.1182Arylsulfatase AARSA0.00071.08960.20500.2304DermokineDMKN0.00071.06380.14460.4002Glutathione synthetaseGSS0.00081.05420.04590.7905Cyclic AMP-responsive element-binding protein 3-like protein 4CREB3L40.00071.0592− 0.01580.9271Protein phosphatase 1 regulatory subunit 14APPP1R14A0.00061.0902− 0.07820.6505Sulfatase-modifying factor 1SUMF10.00041.1168− 0.09410.5853Ribonuclease 4RNASE40.00011.1939− 0.09700.5735SH2 domain-containing protein 1ASH2D1A0.00071.0685− 0.10720.5336Cyclin-dependent kinase 15; EC = 2.7.11.22CDK150.00091.0487− 0.13440.4345Apoptosis-associated speck-like protein containing a CARDPYCARD0.00071.0588− 0.13660.4270Protocadherin gamma-C3PCDHGC30.00111.0315− 0.15390.3702Trafficking protein particle complex subunit 3TRAPPC30.00131.0145− 0.15610.3631Ephrin-A1EFNA10.00101.0422− 0.15800.3575Neural proliferation differentiation and control protein 1NPDC10.00071.0657− 0.17690.3019Delta-like protein 1DLL10.00011.1820− 0.17920.2956Guanine nucleotide exchange factor VAV3VAV30.00041.1155− 0.20640.2272Protein FAM3BFAM3B0.00101.0383− 0.22710.1828Vascular non-inflammatory molecule 2VNN20.00061.0792− 0.24570.1485Calsequestrin-1CASQ10.00111.0292− 0.24780.1450Protein kinase C gamma typePRKCG0.00021.1413− 0.27230.1082Plasma serine protease inhibitorSERPINA50.00141.0203− 0.29720.0783Rho GTPase-activating protein 36ARHGAP360.00001.2910− 0.30370.0718Insulin-like peptide INSL6INSL60.00011.1802− 0.37690.0234SomaScan was used to quantify plasma proteins and a Student’s t-test was used to examine differences between euglycemic and diabetic individuals. Cohen effect size is indicated. Pearson’s correlation (r) and p-value are indicated for the relationships between plasma proteins and pGSN. Proteins significantly correlated with pGSN are indicated in bold. Values are sorted by correlation with high positive values at topMultiple markers can match to the same target protein and are uniquely identified by sequence numbers. 1. SomaScan seq. 21768.9;. 2. SomaScan seq. 21813.171; 3. SomaScan seq. 10025.1; 4. SomaScan seq. 15527.90
## Discussion
In this study, we examined pGSN levels in a large longitudinal cohort of African American and White adults with or without diabetes. We found that pGSN was associated with the social determinants of health, race and poverty status, in the context of diabetes. White adults with diabetes had lower pGSN levels compared to the White adults without diabetes, as well as compared to African American adults with or without diabetes. In addition, we report that for individuals living below poverty status, those with diabetes had lower levels of pGSN compared to euglycemic individuals. Given the beneficial role of pGSN in clearing toxic actin filaments and other inflammatory mediators from the circulation, these data suggest that White adults with diabetes and adults living below poverty with diabetes may be at higher risk for end organ complications given that these groups have the lowest level of pGSN.
The average pGSN levels in our cohort were 42.1 µg/ml at time 1 and 42.8 µg/ml at time 2. This level is lower than what has previously been reported for human cohorts and may be due to several reasons. First, the reported concentration of pGSN can vary widely depending on the methodology used for measurement [4]. Available commercial kits for measurement of gelsolin by ELISA have not reported validation data for clinical use, including matrix effects, and frequently do not standardize against purified protein. Western blot and other immunoblotting methods used to estimate pGSN tends to have higher values due to denaturing of the samples which removes matrix interference effects, but have high variability. Actin nucleation assays can also be used to indirectly measure pGSN, but these assays are technically more difficult to perform, and have not been successfully used in multiwell plates where quality control can be tested. Here, we used a sensitive, quantitative ELISA which is specific for the secreted, plasma form of gelsolin [8].
In our cohort, we found that pGSN levels were lower in men than in women. This is consistent with another study that reported lower pGSN in men compared to women in a cohort of ankylosing spondylitis patients undergoing anti-TNF-alpha therapy and controls [27]. However, very few reports have examined differences in pGSN with sex or race. Here we also report that White adults with diabetes have lower pGSN compared to White adults without diabetes as well as to African American adults either with or without diabetes. Our findings demonstrate that it is important to quantify pGSN levels in diverse cohorts to determine its predictive value as a biomarker of disease. In addition, we report differences in pGSN with poverty status. Individuals with diabetes living below poverty have the lowest level of pGSN. This finding is important since poverty and low socioeconomic status may enhance the risk for developing the well-known complications of diabetes mellitus, cardiovascular disease, chronic kidney disease and premature mortality [28–31]. Here we have identified pGSN as a potentially novel biological transducer of social disadvantage.
Previously, we reported higher levels of circulating EVs in White adults with diabetes, a finding not observed in African Americans with diabetes. As EVs can contain surface filamentous actin which can bind to pGSN [15], we postulated that EVs may bind and reduce circulating pGSN levels in White individuals with diabetes. Conversely, higher pGSN levels might reduce EV levels. However, contrary to our initial hypothesis, we did not find that pGSN levels were correlated with circulating EV levels in our cohort. This difference may be since EVs may contain surface filamentous actin only under specific injury or acute conditions. For example, EVs have been shown to bind to pGSN under stress conditions of high pressure and decompression [15]. In this study larger EVs were analyzed using flow cytometry. Therefore, there may be a subset of larger EVs that contain surface filamentous actin and bind to pGSN.
In our diverse cohort of euglycemic participants and participants with diabetes, we found that pGSN levels were significantly correlated with 19 plasma proteins. Of these proteins, several have previously been associated with diabetes. Most notably, adiponectin was positively correlated with pGSN. Both pGSN and adiponectin have roles in combating inflammation. Given the inflammatory component of diabetes, this data suggests that lower levels of pGSN and adiponectin in White adults with diabetes may lead to heightened inflammation in these individuals. Adiponectin and other proteins identified in our analysis, such as ARHGAP36, WNT5B, VAV3, Ephrin-A1 and MMP2 have roles in regulating and remodeling the vascular endothelium. Vascular dysfunction in individuals with type 2 diabetes contributes to the increased burden of cardiovascular disease and end organ complications [32, 33], which may suggest a mechanism whereby individuals with diabetes are at a higher risk for vascular comorbidities and end organ complications. In support of this idea, an exploratory study using iTRAQ proteomics followed by immunoblotting validation reported that pGSN were lower in individuals with type 2 incipient diabetic nephropathy compared to controls [34]. Since this study used immunoblotting as a method for semi-quantitative pGSN validation, this study is interesting but warrants further investigation and follow up.
One limitation to our study is that the SomaScan assays were performed in a smaller subset of the initial cohort. Therefore, the relationships that we discovered between pGSN and plasma proteins should be interpreted as exploratory. It is not known whether these relationships would persist in a larger cohort. Furthermore, the small sample size reduced statistical power to examine numerous covariates in the analysis. Nevertheless, as this is a large-scale proteomic profile examining pGSN with other plasma proteins, these data implicate a role of pGSN in diabetes. In addition, poverty status was ascertained at study baseline and may change over time. As is the case with many observational cohort studies, we cannot rule out residual confounding despite considering key variables when designing the cohort and also the inclusion of covariates into our analyses.
Our study has several notable strengths. First, we measured pGSN longitudinally at two different time points, which is important given that there are few studies with longitudinal measurements of pGSN. Second, our cohort consists of both African Americans and White adults living above and below poverty. In this study, race and poverty status were equally represented across diabetes groups. This enabled us to examine the effects of race and poverty status on pGSN in the context of diabetes.
## Conclusions
Here we report lower pGSN levels in White individuals with diabetes compared to White individuals without diabetes as well as compared to African American adults either with or without diabetes mellitus. In addition, pGSN levels were lower in individuals with diabetes living below poverty and in men compared to women. pGSN levels were significantly correlated with the adipokine, adiponectin and other inflammation- and diabetes-related proteins. These data may indicate a potential role of pGSN in diabetes and further our knowledge about the utility of pGSN as a biomarker of diabetes.
## References
1. 1.International Diabetes Federation. IDF Diabetes Atlas, 9th edn. Brussels, Belgium. 2019; http://www.diabetesatlas.org.
2. Galicia-Garcia U, Benito-Vicente A, Jebari S, Larrea-Sebal A, Siddiqi H, Uribe KB. **Pathophysiology of type 2 diabetes mellitus**. *Int J Mol Sci* (2020.0) **21** 6275. DOI: 10.3390/ijms21176275
3. Bucki R, Levental I, Kulakowska A, Janmey PA. **Plasma gelsolin: function, prognostic value, and potential therapeutic use**. *Curr Protein Pept Sci* (2008.0) **9** 541-551. DOI: 10.2174/138920308786733912
4. Piktel E, Levental I, Durnaś B, Janmey PA, Bucki R. **Plasma gelsolin: indicator of inflammation and its potential as a diagnostic tool and therapeutic target**. *Int J Mol Sci* (2018.0) **19** 2516. DOI: 10.3390/ijms19092516
5. Davoodian K, Ritchings BW, Ramphal R, Bubb MR. **Gelsolin activates DNase I in vitro and cystic fibrosis sputum**. *Biochemistry* (1997.0) **36** 9637-9641. DOI: 10.1021/bi9711487
6. Li GH, Arora PD, Chen Y, McCulloch CA, Liu P. **Multifunctional roles of gelsolin in health and diseases**. *Med Res Rev* (2012.0) **32** 999-1025. DOI: 10.1002/med.20231
7. Mounzer KC, Moncure M, Smith YR, Dinubile MJ. **Relationship of admission plasma gelsolin levels to clinical outcomes in patients after major trauma**. *Am J Respir Crit Care Med* (1999.0) **160** 1673-1681. DOI: 10.1164/ajrccm.160.5.9807137
8. Self WH, Wunderink RG, DiNubile MJ, Stossel TP, Levinson SL, Williams DJ. **Low admission plasma gelsolin concentrations identify community-acquired pneumonia patients at high risk for severe outcomes**. *Clin Infect Dis* (2019.0) **69** 1218-1225. DOI: 10.1093/cid/ciy1049
9. Zhang Q, Fillmore TL, Schepmoes AA, Clauss TRW, Gritsenko MA, Mueller PW. **Serum proteomics reveals systemic dysregulation of innate immunity in type 1 diabetes**. *J Exp Med* (2012.0) **210** 191-203. DOI: 10.1084/jem.20111843
10. Insenser M, Vilarrasa N, Vendrell J, Escobar-Morreale HF. **Remission of diabetes following bariatric surgery: plasma proteomic profiles**. *J Clin Med* (2021.0) **10** 3879. DOI: 10.3390/jcm10173879
11. Khatri N, Sagar A, Peddada N, Choudhary V, Chopra BS, Garg V. **Plasma gelsolin levels decrease in diabetic state and increase upon treatment with F-actin depolymerizing versions of gelsolin**. *J Diabetes Res* (2014.0) **2014** 152075. DOI: 10.1155/2014/152075
12. Noren Hooten N, Evans MK. **Extracellular vesicles as signaling mediators in type 2 diabetes mellitus**. *Am J Physiol Cell Physiol* (2020.0) **318** C1189-C1199. DOI: 10.1152/ajpcell.00536.2019
13. Kalluri R, LeBleu VS. **The biology, function, and biomedical applications of exosomes**. *Science* (2020.0) **367** eaau6977. DOI: 10.1126/science.aau6977
14. Yanez-Mo M, Siljander PR, Andreu Z, Zavec AB, Borras FE, Buzas EI. **Biological properties of extracellular vesicles and their physiological functions**. *J Extracell Vesicles* (2015.0) **4** 27066. DOI: 10.3402/jev.v4.27066
15. Bhopale VM, Ruhela D, Brett KD, Nugent NZ, Fraser NK, Levinson SL. **Plasma gelsolin modulates the production and fate of IL-1β-containing microparticles following high-pressure exposure and decompression**. *J Appl Physiol* (2021.0) **130** 1604-1613. DOI: 10.1152/japplphysiol.01062.2020
16. Greening DW, Xu R, Gopal SK, Rai A, Simpson RJ. **Proteomic insights into extracellular vesicle biology—defining exosomes and shed microvesicles**. *Expert Rev Proteomics* (2017.0) **14** 69-95. DOI: 10.1080/14789450.2017.1260450
17. Kim KM, Abdelmohsen K, Mustapic M, Kapogiannis D, Gorospe M. **RNA in extracellular vesicles**. *Wiley Interdiscip Rev RNA.* (2017.0) **8** e1413. DOI: 10.1002/wrna.1413
18. Thakur BK, Zhang H, Becker A, Matei I, Huang Y, Costa-Silva B. **Double-stranded DNA in exosomes: a novel biomarker in cancer detection**. *Cell Res* (2014.0) **24** 766-769. DOI: 10.1038/cr.2014.44
19. Amari L, Germain M. **Mitochondrial extracellular vesicles—origins and roles**. *Front Mol Neurosci* (2021.0) **14** 767219. DOI: 10.3389/fnmol.2021.767219
20. Buzas EI, Gyorgy B, Nagy G, Falus A, Gay S. **Emerging role of extracellular vesicles in inflammatory diseases**. *Nat Rev Rheumatol* (2014.0) **10** 356-364. DOI: 10.1038/nrrheum.2014.19
21. Freeman DW, Noren Hooten N, Eitan E, Green J, Mode NA, Bodogai M. **Altered extracellular vesicle concentration, cargo and function in diabetes mellitus**. *Diabetes* (2018.0) **67** 2377-2388. DOI: 10.2337/db17-1308
22. Evans MK, Lepkowski JM, Powe NR, LaVeist T, Kuczmarski MF, Zonderman AB. **Healthy aging in neighborhoods of diversity across the life span (HANDLS): overcoming barriers to implementing a longitudinal, epidemiologic, urban study of health, race, and socioeconomic status**. *Ethn Dis* (2010.0) **20** 267-275. PMID: 20828101
23. Eitan E, Green J, Bodogai M, Mode NA, Bæk R, Jørgensen MM. **Age-related changes in plasma extracellular vesicle characteristics and internalization by leukocytes**. *Sci Rep* (2017.0) **7** 1342. DOI: 10.1038/s41598-017-01386-z
24. Armbruster DA, Pry T. **Limit of blank, limit of detection and limit of quantitation**. *Clin Biochem Rev* (2008.0) **29** S49-52. PMID: 18852857
25. 25.R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2022; https://www.R-project.org/.
26. Fisman EZ, Tenenbaum A. **Adiponectin: a manifold therapeutic target for metabolic syndrome, diabetes, and coronary disease?**. *Cardiovasc Diabetol* (2014.0) **13** 103. DOI: 10.1186/1475-2840-13-103
27. Genre F, López-Mejías R, Miranda-Filloy JA, Ubilla B, Carnero-López B, Gómez-Acebo I. **Gelsolin levels are decreased in ankylosing spondylitis patients undergoing anti-TNF-alpha therapy**. *Clin Exp Rheumatol* (2014.0) **32** 218-224. PMID: 24351434
28. Stringhini S, Carmeli C, Jokela M, Avendano M, Muennig P, Guida F. **Socioeconomic status and the 25 x 25 risk factors as determinants of premature mortality: a multicohort study and meta-analysis of 17 million men and women**. *Lancet* (2017.0) **389** 1229-1237. DOI: 10.1016/S0140-6736(16)32380-7
29. Zonderman AB, Mode NA, Ejiogu N, Evans MK. **Race and poverty status as a risk for overall mortality in community-dwelling middle-aged adults**. *JAMA Intern Med* (2016.0) **176** 1394-1395. DOI: 10.1001/jamainternmed.2016.3649
30. Crews DC, Charles RF, Evans MK, Zonderman AB, Powe NR. **Poverty, race, and CKD in a racially and socioeconomically diverse urban population**. *Am J Kidney Dis* (2010.0) **55** 992-1000. DOI: 10.1053/j.ajkd.2009.12.032
31. Odutayo A, Gill P, Shepherd S, Akingbade A, Hopewell S, Tennankore K. **Income disparities in absolute cardiovascular risk and cardiovascular risk factors in the United States, 1999–2014**. *JAMA Cardiol* (2017.0) **2** 782-790. DOI: 10.1001/jamacardio.2017.1658
32. Fox CS, Coady S, Sorlie PD, D'Agostino RB, Pencina MJ, Vasan RS. **Increasing cardiovascular disease burden due to diabetes mellitus: the Framingham Heart Study**. *Circulation* (2007.0) **115** 1544-1550. DOI: 10.1161/CIRCULATIONAHA.106.658948
33. Shi Y, Vanhoutte PM. **Macro- and microvascular endothelial dysfunction in diabetes**. *J Diabetes* (2017.0) **9** 434-449. DOI: 10.1111/1753-0407.12521
34. Lu H, Deng S, Zheng M, Hu K. **iTRAQ plasma proteomics analysis for candidate biomarkers of type 2 incipient diabetic nephropathy**. *Clin Proteomics* (2019.0) **16** 33. DOI: 10.1186/s12014-019-9253-1
|
---
title: A structural equation model linking health literacy, self-efficacy, and quality
of life in patients with polycystic ovary syndrome
authors:
- Yunmei Guo
- Ying Liu
- Rui Ding
- Xin Yan
- Huiwen Tan
- Yousha Wang
- Xueting Wang
- LianHong Wang
journal: BMC Women's Health
year: 2023
pmcid: PMC9999555
doi: 10.1186/s12905-023-02223-4
license: CC BY 4.0
---
# A structural equation model linking health literacy, self-efficacy, and quality of life in patients with polycystic ovary syndrome
## Abstract
### Background
Health literacy is a crucial factor that affects health outcomes. Understanding the current status of health literacy among patients with polycystic ovary syndrome (PCOS) is the basis for helping patients better manage risk factors and improve their health outcomes. This study aimed to explore the status of and factors influencing health literacy in patients with PCOS, and to validate the pathway between health literacy, quality of life, and self-efficacy for these patients.
### Methods
A cross-sectional study was conducted using a convenience sample of 300 patients with PCOS in the gynecology outpatient clinic of a tertiary hospital in Zunyi from March to September 2022. Data on health literacy, demographic features, quality of life, and self-efficacy were collected. Multiple stepwise linear regression was conducted to assess the risk factors associated with health literacy for the study participants. A structural equation model was used to construct and validate the pathways.
### Results
Most participants exhibited low health literacy (3.61 ± 0.72), and only $25.70\%$ had adequate health literacy. Multiple regression analysis revealed that the main factors associated with health literacy among participants included Body Mass Index (BMI) (B = −0.95, $p \leq 0.01$), education ($B = 3.44$, $p \leq 0.01$), duration of PCOS ($B = 4.66$, $p \leq 0.01$), quality of life ($B = 0.25$, $p \leq 0.01$), and self-efficacy ($B = 0.76$, $p \leq 0.01$). Multiple fit values indicated that the model fit the data effectively. The direct effect of health literacy on self-efficacy and quality of life was 0.06 and 0.32, respectively. The indirect effect of health literacy on quality of life was −0.053, and the total effect of health literacy on quality of life was 0.265.
### Conclusions
Health literacy was low among patients with PCOS. Healthcare providers should pay more attention to health literacy and to developing the corresponding intervention strategies urgently needed to improve the quality of life and health behavior of patients with PCOS.
## Background
Polycystic ovary syndrome (PCOS) is a common endocrine disease in women of reproductive age with a prevalence of $4\%$–$22.5\%$ [1]. PCOS is a multisystem disease characterized by anovulation, hyperandrogenism, and polycystic ovarian morphology [2]. To date, the cause of PCOS remains unclear. PCOS may increase the risk of type 2 diabetes, cardiovascular disease, endometrial cancer, metabolic syndrome, and pregnancy complications [2, 3]. Furthermore, PCOS can also significantly negatively impact psychological function [4]; the complex nature of the disease and its associated complications can diminish the quality of life (QOL) and lead to decreased emotional well-being [5]. Treatment for PCOS includes pharmacological options, surgical options, and life management. Although life management is recommended as a first-line treatment [6], maintaining a life management approach is a common challenge.
A previous study suggests that the challenges and barriers to implementing life management in patients with PCOS may be related to health literacy (HL) [7]. The World Health Organization defines HL as cognitive and social skills that determine an individuals' motivation and ability to understand and use information in a way that promotes and maintains health [8]. Having an adequate level of HL allows individuals to achieve the knowledge, skills, and confidence needed to improve personal health through lifestyle changes and to promote public health. Moreover, low or inadequate HL places a heavy burden on healthcare systems as it is associated with many poor health indicators, such as increased hospitalization and emergency department use, poor health outcomes, and higher mortality rates [10]. However, numerous studies [11, 12] have demonstrated that improving HL can constitute an essential breakthrough in promoting positive changes in health behavior. Unfortunately, only two articles [7, 13] have reported data related to HL in patients with PCOS and, thus far, the status of HL among patients with PCOS in China has not been investigated. Furthermore, the factors that affect HL have not yet been explored.
The evidence suggest that women with PCOS have poorer QOL compared with women with other chronic conditions, such as arthritis, back pain, and diabetes [14]. Previous studies have demonstrated that patients with low HL may be less concerned about their health status and therefore engage in unhealthy behavioral patterns that lead to a reduced QOL [15]. Similar studies have also suggested that HL can influence patients' QOL through self-efficacy (SE) [16–18]. SE is an individual’s belief in their ability to plan and execute a specific course of action [19]. SE management programs have been widely used in treating chronic diseases because they can improve functioning and health status [20]. The current study aims to affirm the role of SE in modulating psychological status and life management in PCOS [21, 22]. However, existing research on the relationship between SE and QOL [21, 22] has only identified a direct relationship through regression or correlation, and no study has explored this mediating relationship for patients with PCOS using structural equation modeling (SEM). SEM is a multivariate statistical framework used to model the complex relationships between directly and indirectly observed (latent) variables [23]. SE in PCOS patients may be a potential mediator, that is, an effective way to enhance HL and improve QoL. Therefore, it is necessary to use SEM to explore how HL can improve patients' QOL through SE.
To our knowledge, no previous study has investigated the status and associated factors of HL and validated the direct and indirect association between HL and QOL using SEM among patients with PCOS. To address this limited evidence, we conducted a cross-sectional study to investigate the status and influencing factors of HL in patients with PCOS and to validate the relationship between HL, SE, and QOL.
## Design and data collection
A cross-sectional study was conducted in Zunyi between March and September 2022 among patients with PCOS attending a gynecology clinic. Patients were included if they met the following criteria: between 18–45 years of age with two of the following Rotterdam Criteria: a) hyperandrogenism, b) ovulatory dysfunction, and c) polycystic ovaries. Patients who were unable to read and understand the provided questionnaire, could not use a smartphone, or refused to sign the informed consent form were excluded from the study.
Three researchers conducted face-to-face data collection. After recruiting participants according to the inclusion criteria, the nature of the study, purpose, and investigation procedure were explained to them. All participants signed an informed consent form before participating in the study. Patients were instructed to complete a questionnaire using a smartphone scan code. While the participants were completing the questionnaire, one of the researchers checked the questionnaire filling status. To reduce the generation of invalid questionnaires, researchers checked and confirmed incorrect or incomplete responses in real time.
## Sample size calculation
We calculated the sample size using event-per-variable (EPV), assuming that p represents the prevalence of PCOS and K represents the number of predictors. Based on the above assumptions and the formula N = EPV × K/p ($k = 4$, $$p \leq 0.15$$), only an EPV of 10 or above is considered robust. According to the above formula, the sample size was 267, and the final sample size was 307, considering a $15\%$ sample loss rate.
## Demographic questionnaire
A checklist included questions on the following demographic features: waist circumference, age, height, weight, years of illness, place of residence, marital status, education level, body mass index (BMI), and occupation.
## Health literacy scale
In this study, the Chinese version of the Health Literacy Management Scale (HeLMS) was used to evaluate HL of patients [24, 25]. This questionnaire consists of 24 items in four dimensions: communication and interaction ability (9 items with a total score of 45 points), information acquisition ability (9 items with a total score of 45 points), willingness to improve health (4 items with a total score of 20 points), and willingness to financially support health (2 items with a total score of 10 points). Each item is rated on a 5-point Likert scale ranging from 1 (very difficult) to 5 (not at all difficult), with higher scores indicating a more advanced level of HL. The total score is 120 points. An average score of < 4 on all dimensions of this scale was considered as inadequate or low level of HL. The internal consistency of the HeLMS scale was good, with a Cronbach's α of 0.874. Dimension mean scores = total score per dimension/total participants; score per dimension item = total score per dimension/number of dimension items; item mean scores = total score per dimension item/total participants; and possessing rate = total number of items with a score of less than four points per dimension/total participants.
## Quality of life [26]
QOL was calculated using the Short Form 36 (SF-36) scale. It comprises eight subscales: general health, bodily pain, physical functioning, vitality, role limitations due to physical problems, mental health, role limitations due to emotional problems, and social function. A higher score indicates a better QOL and the scores range from 0 to 100. The internal consistency of the SF-36 scale was good, with a Cronbach’s α of 0.89.
## Self-efficacy scale
SE was assessed using the General Self-Efficacy Scale, which consists of 10 items. A four-point Likert scale was used to answer each question, ranging from 1 (not at all true) to 4 (exactly true). Total scores range from 10 to 40, with higher scores indicating better SE [27].
## Ethical considerations
This study was approved by the Ethics Committee of the Affiliated Hospital of the Zunyi Medical University (No. KLLY-2020-134). Informed consent was obtained from all participants, and procedures were conducted according to the Declaration of Helsinki. All participants had the right to withdraw at any time without any adverse effect on the clinical work.
## Data analysis
SPSS18.0 was used to analyze the data. Descriptive data is expressed as frequencies and mean ± standard deviation. Demographic differences in HL, SE, and QOL were analyzed using an independent samples t-test and ANOVA. Correlations between HL, SE, and QOL were analyzed using Pearson’s correlation analysis. Multiple stepwise linear regression was performed to examine the risk factors for HL. P values less than 0.05 indicated a significant difference. AMOS 26.0 was used to model the structural equations, with HL, QOL, and SE as latent variables and the corresponding entries set as observed variables. The models were continuously improved and re-estimated, and the most appropriate model was selected.
## Results
Seven patients declined to participate in this study because of time pressure ($$n = 4$$) and too many questionnaire items ($$n = 3$$). Ultimately, 300 patients with PCOS were included in the study. Table 1 presents participants' general demographic information. The average age was 24.78 ± 4.21 years, the average BMI was 24.69 ± 3.43, and the average waist circumference was 81.83 ± 7.84. The majority of women had been educated beyond middle school ($$n = 233$$, $77.70\%$), $52.70\%$ of the patients were married, and $52.30\%$ reported that they wished to have children. Further, $47.70\%$ reported having PCOS for less than 1 year and $33.30\%$ reported having PCOS for 1–3 years. The mean scores for HL, SE, and QOL of the patients were 87.20 (SD = 17.20), 26.55 (SD = 6.56), and 56.11(SD = 10.55), respectively. Table 1Demographic characteristics of participants ($$n = 300$$)VariableCategoriesMean (SD)Frequency (N)Percentage (%)Age24.78 ± 4.21BMI24.69 ± 3.43WC81.83 ± 7.84ResidenceCity15852.70Countryside14247.30Marital statusSingle13645.30Married15852.70Widowed/divorced62.00Ethnic groupHan-Nationality25183.70Ethnic minority4916.30EducationElementary6722.30Middle school5819.30High school5418.00College12140.30OccupationEmployed9933.00Unemployed5719.00Student6321.00Other8127.00Duration of PCOS < 1 year14347.701–3 years10033.304–6 years3612.00 > 7 years217.00Desire for pregnancyYes15752.30No14347.70Self-efficacy26.55 ± 6.56Health literacy87.20 ± 17.20Quality of life56.11 ± 10.55BMI body mass index, WC waist circumference Table 2 shows the total HL scale scores and mean scores for each sub-dimension item. The average total HL was 87.20 ± 17.01 (mean: 3.61 ± 0.72). The total score for information acquisition ability was 32.47 ± 7.71 (mean: 3.67 ± 0.86), communication and interaction ability was 32.46 ± 7.47 (mean: 3.61 ± 0.83), willingness to improve health was 15.23 ± 3.72 (mean: 3.81 ± 0.92), and willingness to financially support health was 6.85 ± 2.28 (mean: 3.43 ± 1.14). Only $25.70\%$ of patients had adequate health literacy. Table 2HL total and subscale mean scores ($$n = 300$$) of patients with PCOSItemDimension mean scoresItem mean scoresHighest and lowest obtainable scorePossessing rate (%)Information acquisition ability32.47 ± 7.713.67 ± 0.869–4532.70Communication and interaction ability32.46 ± 7.473.61 ± 0.839–4525.00Willingness to improve health15.23 ± 3.723.81 ± 0.925–2033.00Willingness to support financially6.85 ± 2.283.43 ± 1.142–1018.70Total mean scores87.20 ± 17.013.61 ± 0.7225–12025.70 As shown in Table 3, there were significant differences in the scores of HL and SE between married and unmarried and widowed/divorced patients ($p \leq 0.01$). There were significant differences in HL, QOL, and SE among patients with different levels of education ($p \leq 0.001$) and with different PCOS durations ($p \leq 0.001$). Additionally, as shown in Table 4, HL was positively and statistically associated with QOL and SE ($p \leq 0.01$). This study showed that HL was negatively associated with BMI.Table 3Demographic characteristics and their associations with health literacy, self efficacy, and quality of life($$n = 300$$)VariableCategoriesFrequency (N)HL (mean ± SD)QOL (mean ± SD)SE (mean ± SD)Residencecity15886.11 ± 16.1850.55 ± 10.8526.86 ± 6.68Countryside14288.42 ± 17.8752.78 ± 10.1126.19 ± 6.44P value0.1730.0930.328Marital statusSingle13684.47 ± 17.9451.63 ± 11.0525.87 ± 6.23Married15888.96 ± 15.8151.31 ± 10.1626.81 ± 6.18Widowed/divorced6102.66 ± 13.0058.65 ± 7.3334.67 ± 15.23P value0.0060.2470.004Ethnic groupHan-Nationality25187.48 ± 17.1951.78 ± 10.3726.63 ± 6.44Ethnic minority4985.75 ± 16.0950.71 ± 11.4726.14 ± 7.22P value0.6940.150.33educationElementary6774.82 ± 15.2347.98 ± 10.6324.31 ± 6.19Middle school5683.08 ± 17.5851.84 ± 11.0124.38 ± 5.37High school5493.38 ± 13.4652.96 ± 9.9827.57 ± 5.35College12193.26 ± 14.7852.89 ± 10.1728.36 ± 7.15P value0.0000.0140.000OccupationEmployed9984.78 ± 16.8350.56 ± 10.8426.55 ± 6.24Unemployed5784.47 ± 17.0950.64 ± 11.5225.17 ± 6.52Student6389.77 ± 15.9753.14 ± 9.9226.98 ± 5.81Other8190.06 ± 17.5052.36 ± 9.9227.16 ± 7.45P value0.0650.3630.325Duration of PCOS < 1 year14380.04 ± 15.4949.57 ± 10.6624.90 ± 6.181–3 years10089.21 ± 15.5651.31 ± 10.5327.00 ± 5.934–6 years36102.36 ± 12.6556.88 ± 8.6829.11 ± 5.73 > 7 years21100.38 ± 12.5857.81 ± 7.4131.19 ± 9.33P value0.0000.0000.000Desire for pregnancyYes15787.29 ± 17.4951.86 ± 10.1826.75 ± 6.90no14387.09 ± 16.5351.32 ± 10.9626.32 ± 6.19P value0.500.1710.618The bold definition is statistically significantHL health literacy, SE self-efficacy, QOL quality of lifeTable 4Associations and differences of HL mean scores with demographic variables ($$n = 300$$)Variableinformation acquisition abilitycommunication and interaction abilitywillingness to improve healthwillingness to support financiallyTotal scoreAge−0.086−0.069−0.005−0.08−0.084BMI−0.312**−0.222**−0.195**−0.181**−0.303**WC−0.156**−0.042−0.069−0.067−0.112SE0.383**0.422**0.361**0.328**0.484**PF0.405**0.487**0.245**0.292**0.494**RP0.202**0.177**0.230**0.240**0.246**BP−0.065−0.111−0.137**−0.164**−0.130**GH0.187**0.180**0.0620.0940.189**VT0.0730.135**0.0980.0830.128**SF−0.043−0.083−0.029−0.033−0.066RE0.176**0.204**0.174**0.203**0.235**MH0.1000.168**0.148**0.0190.158**QOL0.28**0.336**0.211**0.161**0.346****$p \leq 0.05$; BMI body mass index; WC Waist circumference; HL health literacy, SE self-efficacy; QOL quality of life, GH general health, BP bodily pain, PF physical functioning, VT vitality, RP role limitations due to physical problems, MH mental health, RE role limitations due to emotional problems, SF social function Table 5 shows the results of the stepwise multiple linear regression analysis, which revealed that factors including BMI (B = −0.95, $p \leq 0.01$), education ($B = 3.44$, $p \leq 0.01$), duration of PCOS ($B = 4.66$, $p \leq 0.01$), QOL($B = 0.25$, $p \leq 0.001$), and SE ($B = 0.76$, $p \leq 0.01$) were associated with HL. The model showed $47.40\%$ variance shared between the dependent and independent variables (R2 = 47.40, $F = 52.97$, $p \leq 0.001$).Table 5Multivariate analysis (stepwise) of predictors for HL($$n = 300$$) ModelVariableUnstandardised coefficientsStandardised coefficientstsign$95.0\%$ CI for BR2FPBSEBetaLower boundUpper boundBMI−0.950.21−0.19−4.430.00−1.37−0.5347.4052.970.00Education3.440.640.245.400.002.194.70Duration of PCOS4.660.860.255.430.002.976.35QOL0.250.070.163.520.000.110.39SE0.760.120.296.450.000.530.99BMI body mass index, HL health literacy, SE self-efficacy, QOL quality of life
The final model fit indicated that the χ2/degree of freedom was 1.433, goodness of fit index was 0.938, adjusted goodness to fit was 0.915, comparative fit index was 0.971, normed fit index was 0.912, Tucker-Lewis index was 0.965, incremental fit index was 0.972, and the root mean square error of approximation was 0.038 (Table 6). Further, SE mediated the influence of HL on QOL. The direct effect of HL on SE and QOL was 0.06 and 0.32, respectively. The indirect effect of HL on QOL was −0.053 and the total effect of HL on QOL was 0.265. These fit indices suggest that SEM better describes the pathway relationship between HL, SE, and QOL (Fig. 1).Table 6Model fit indexVariableχ2GFIRMSEANFITLICFIIFIAICFit index197.6880.9380.0380.9120.9650.9710.972401.23Reference value> 0.9< 0.05> 0.9> 0.9> 0.9> 0.9> 400GFI goodness of fit index, CFI comparative fit index, NFI normed fit index, TLI =Tucker-Lewis index, IFI incremental fit index, RMSEA root mean square error of approximation, AIC Akaike Information CriterionFig. 1Final model and standardized pathway coefficients among health literacy, self efficacy, and quality of life. Note:*$P \leq 0.05$; **$P \leq 0.01$; HL = health literacy; SE = self-efficacy; QOL = quality of life; A1 = Information acquisition ability; A2 = Communication and interaction ability; A3 = Willingness to improve health; A4 = Willingness to support financially; RP = Role limitations due to physical problems; BP = bodily pain; SF = social function; RE = role limitations due to emotional problems; MH = mental health; B1 = If I do my best, I can always solve the problem; B2 = Even if others are against me, I still have a way to get what I want.; B3 = *It is* easy for me to stick to my ideals and achieve my goals; B4 = I am confident that I can deal with anything that happens all of a sudden.; B5 = With my intelligence, I will be able to cope with unexpected situations.; B6 = If I do what I have to do, I will be able to solve most of the problems.; B7 = I can face difficulties calmly because I believe in my ability to deal with problems.; B8 =. When faced with a difficult problem, I can usually find several solutions; B9 = When there is trouble, I can usually think of some ways to deal with it; B10 = No matter what happens to me, I can handle it
## Discussion
This study is the first to investigate the status of HL in patients with PCOS in China, and the findings showed that most patients ($74.30\%$) had inadequate HL (87.20 ± 17.20), and only $25.70\%$ had adequate HL. In addition, to the best of our knowledge, this is the first study to use SEM to construct and validate the pathway between HL, SE, and QOL in patients with PCOS. HL directly impacted QOL and SE in patients with PCOS. It was revealed that SE played a mediating role between HL and QOL.
In the current study, we found that patients with PCOS had low HL (3.61 ± 0.72), which is consistent with previous studies [13]. However, compared with other chronic diseases, the health literacy of PCOS patients is lower than that of hemodialysis patients and diabetes patients [28, 29]. Recent research has shown that the level of health literacy in the eastern coastal provinces of *China is* higher than that in the central and western provinces [30]. Considering that all the participants included in this study were from western China, the factor of geographical location may have led to low HL in patients. Furthermore, recent research has shown that doctors in China, Europe, and America lack knowledge regarding PCOS [31–34], and insufficient disease education in doctors may reduce the abilities for information acquisition and HL in patients. Finally, low health literacy in patients may be due to the fact that the clinical symptoms of PCOS do not endanger patients' lives and that there is lack of knowledge about the long-term complications of the disease [31, 32]. Therefore, future studies need to address how to strengthen the continuing education about PCOS for Chinese obstetricians and gynecologists, improve the quality of medical services, improve the health system at all levels, and use the internet and digital technology to solve the changing health literacy needs of individuals and communities. Furthermore, research is needed on how to formulate appropriate intervention measures according to local culture and to design policies to continuously improve the HL of patients in the western region.
This study showed that HL was positively associated with education and disease duration and negatively associated with BMI. First, it was found that higher educational levels in patients with PCOS were associated with better HL scores. This shows that education level is a significant factor in enhancing and improving HL. PCOS is the most common endocrine disease that requires long-term management [35], and adequate HL for patients with low education levels may enhance the effect of life management on patients' reproductive health and metabolism and promote the maintenance of healthy behaviors. Second, it was revealed that a higher duration of PCOS resulted in higher HL scores. A longer disease duration may increase the accessibility of health education and eventually enhance the skills required for disease knowledge and information. Finally, the negative relationships between HL scores and BMI were congruent with previous findings [18]. The negative correlation between HL and BMI suggests that patients with high HL have healthier lifestyles [36]. Moreover, previous studies have suggested that a higher BMI is associated with poor health behaviors in the life management of patients with PCOS [21], which may explain the relationship between a high BMI and low HL. Furthermore, chronic disease prevalence increases with increasing BMI [37, 38] and, since PCOS is a chronic disease and most patients are obese, this may also contribute to the negative relationship between HL and BMI.
In the current study, we discovered that the direct effects of HL on patients' QOL by SEM were congruent with previous findings [37, 38]. Previous studies have demonstrated that patients with low HL may be less concerned about their health status and therefore have unhealthy behavioral patterns that lead to a reduced QOL [15]. In addition, PCOS patients with low health literacy may have difficulty in fully understanding the instructions from healthcare providers. They may also have difficulty in maintaining life management, potentially resulting in suboptimal symptom management (e.g., metabolic and reproductive disorders), all of which may exacerbate treatment-related symptoms, resulting in a poorer QOL [7, 39]. This may explain the negative association between HL and QOL. Patients with higher SE may develop more strength, greater awareness, and intrinsic motivation, allowing them to persist in treatment and life maintenance. This study confirms that SE is a mediator between HL and QOL. Patients with PCOS that are deficient in HL may have lower SE in their daily lives, thereby reducing their QOL, a relationship that echoes previously proposed theoretical frameworks [40]. Finally, doctors often ignore the evaluation and management of declines in patients’ psychosocial status and QOL [31, 32]. Doctors' neglect of patients' QOL and lack of education may be potential reasons for the decline in patients' QOL.
The strength of our study is that it is the first to investigate HL status and confirm its associated risk factors. Furthermore, it is the first study to investigate the pathways between HL, SE, and QOL, while also confirming the direct and indirect effects of HL on QOL among patients with PCOS. However, this study has some limitations. First, patients were recruited using convenience sampling from the gynecology outpatient clinic of only one tertiary hospital in China. Second, data for this study were collected from patients through self-report. Therefore, the possibility of a response bias could not be eliminated. Third, owing to the limitations of cross-sectional studies, no causal relationship can be inferred; only correlations between variables can be analyzed. Finally, the HL scales used to investigate the patients in this study were not specific, which may have led to biased results.
## Conclusions
In summary, the study revealed that patients with PCOS had low HL; the factors affecting HL mainly included BMI, education, duration of PCOS, SE, and QOL. Moreover, our research confirmed the direct and indirect effects of HL on QOL using SEM. Therefore, future studies should focus on improving and promoting patients' HL and enhancing medical staff awareness to improve patients' HL. The results of this study fill a gap in the literature and provide new perspectives for further improving the QOL and behavior management of patients with PCOS.
## References
1. Skiba MA, Islam RM, Bell RJ. **Understanding variation in prevalence estimates of polycystic ovary syndrome: a systematic review and meta-analysis**. *Hum Reprod Update* (2018.0) **24** 694-709. DOI: 10.1093/humupd/dmy022
2. Zeng X, Xie YJ, Liu YT. **Polycystic ovarian syndrome: Correlation between hyperandrogenism, insulin resistance and obesity**. *Clin Chim Acta* (2020.0) **502** 214-221. DOI: 10.1016/j.cca.2019.11.003
3. Cooney LG, Dokras A. **Beyond fertility: polycystic ovary syndrome and long-term health**. *Fertil Steril* (2018.0) **110** 794-809. DOI: 10.1016/j.fertnstert.2018.08.021
4. Deeks AA, Gibson-Helm ME, Paul E. **Is having polycystic ovary syndrome a predictor of poor psychological function including anxiety and depression?**. *Hum Reprod* (2011.0) **26** 1399-1407. DOI: 10.1093/humrep/der071
5. Turner-McGrievy G, Davidson CR, Billings DL. **Dietary intake, eating behaviors, and quality of life in women with polycystic ovary syndrome who are trying to conceive**. *Hum Fertil (Camb)* (2015.0) **18** 16-21. DOI: 10.3109/14647273.2014.922704
6. Teede HJ, Misso ML, Costello MF. **Recommendations from the international evidence-based guideline for the assessment and management of polycystic ovary syndrome**. *Fertil Steril* (2018.0) **110** 364-379. DOI: 10.1016/j.fertnstert.2018.05.004
7. Lim S, Smith CA, Costello MF. **Health literacy needs in weight management of women with Polycystic Ovary Syndrome**. *Health Promot J Austr* (2021.0) **32** 41-48. DOI: 10.1002/hpja.349
8. Nutbeam D. **Health promotion glossary**. *Health Promot* (1986.0) **1** 113-127. DOI: 10.1093/heapro/1.1.113
9. Buyuksireci D, Demirsoy UN. **Evaluation of the health literacy level of female fibromyalgia patients and relationship between health literacy level and disease activity**. *Arch Rheumatol* (2021.0) **36** 274-279. DOI: 10.46497/ArchRheumatol.2021.8387
10. Berkman ND, Sheridan SL, Donahue KE. **Health literacy interventions and outcomes: an updated systematic review**. *Evid Rep Technol Assess (Full Rep)* (2011.0) **199** 1-941
11. Harsch S, Jawid A, Jawid E. **Health Literacy and Health Behavior Among Women in Ghazni**. *Afghanistan Front Public Health* (2021.0) **9** 629334. DOI: 10.3389/fpubh.2021.629334
12. 12.Prihanto J B, Nurhayati F, Wahjuni E S, et al. Health Literacy and Health Behavior: Associated Factors in Surabaya High School Students, Indonesia. Int J Environ Res Public Health, 2021,18(15).
13. Al-Ruthia YS, Balkhi B, AlGhadeer S. **Relationship between health literacy and body mass index among Arab women with polycystic ovary syndrome**. *Saudi Pharm J* (2017.0) **25** 1015-1018. DOI: 10.1016/j.jsps.2017.04.003
14. Coffey S, Bano G, Mason HD. **Health-related quality of life in women with polycystic ovary syndrome: a comparison with the general population using the Polycystic Ovary Syndrome Questionnaire (PCOSQ) and the Short Form-36 (SF-36)**. *Gynecol Endocrinol* (2006.0) **22** 80-86. DOI: 10.1080/09513590600604541
15. Berkman ND, Sheridan SL, Donahue KE. **Low health literacy and health outcomes: an updated systematic review**. *Ann Intern Med* (2011.0) **155** 97-107. DOI: 10.7326/0003-4819-155-2-201107190-00005
16. Wang CEA. **The effect of health literacy and self-management efficacy on the health-related quality of life of hypertensive patients in a western rural area of China: a cross-sectional study**. *Int J Equity health* (2017.0) **1** 58. DOI: 10.1186/s12939-017-0551-9
17. Lee EH, Lee YW, Moon SH. **A structural equation model linking health literacy to self-efficacy, self-care activities, and health-related quality of life in patients with type 2 diabetes**. *Asian Nurs Res (Korean Soc Nurs Sci)* (2016.0) **10** 82-87. PMID: 27021840
18. Du S, Feng Z, Wang W. **A structural equation model linking health literacy, self efficacy and quality of life in adults with coronary heart disease**. *BMC Cardiovasc Disord* (2022.0) **22** 285. DOI: 10.1186/s12872-022-02720-8
19. Artino AJ. **Academic self-efficacy: from educational theory to instructional practice**. *Perspect Med Educ* (2012.0) **1** 76-85. DOI: 10.1007/S40037-012-0012-5
20. Martinez-Calderon J, Zamora-Campos C, Navarro-Ledesma S. **The role of self-efficacy on the prognosis of chronic musculoskeletal pain: a systematic review**. *J Pain* (2018.0) **19** 10-34. DOI: 10.1016/j.jpain.2017.08.008
21. Guo Y, Liu Y, Yan X. **Factors affecting the adoption of health-promoting behaviours in patients with polycystic ovary syndrome: a cross-sectional study**. *BMJ Open* (2022.0) **12** e56478. DOI: 10.1136/bmjopen-2021-056478
22. Deeks AA, Gibson-Helm ME, Teede HJ. **Anxiety and depression in polycystic ovary syndrome: a comprehensive investigation**. *Fertil Steril* (2010.0) **93** 2421-2423. DOI: 10.1016/j.fertnstert.2009.09.018
23. Stein CMMNNN. **Structural equation modeling**. *Methods Mol Biol* (2012.0) **850** 495-512. DOI: 10.1007/978-1-61779-555-8_27
24. 24.Haolin S. Development and preliminary application of scale on health literacy for patients with chronic disease. Fudan University, 2011.
25. Jordan JEBR. **The health literacy management scale (HeLMS): a measure of an individual's capacity to seek, understand and use health information within the healthcare setting**. *Patient Educ Counsel* (2013.0) **91** 228-235. DOI: 10.1016/j.pec.2013.01.013
26. Brazier JE, Harper R, Jones NM. **Validating the SF-36 health survey questionnaire: new outcome measure for primary care**. *BMJ* (1992.0) **305** 160-164. DOI: 10.1136/bmj.305.6846.160
27. Hu X, Tian C, Sun F. **Reliability and validity of the Chinese version of the General self-efficacy scale**. *Psychol Explor* (2014.0) **34** 53-56
28. Wen Xiuqin HJZJ. **Relationship between health literacy and utilization of basic public health services among diabetics**. *Chin J Health Educ* (2015.0) **4** 456-459
29. Tian Qing JYXJ. **The current status and influencing factors of Health Literacy among hemodialysis patients**. *Chin Nurs Manag* (2018.0) **18** 492-496
30. Li Y, Lv X, Liang J. **The development and progress of health literacy in China**. *Front Public Health* (2022.0) **10** 1034907. DOI: 10.3389/fpubh.2022.1034907
31. Dokras A, Saini S, Gibson-Helm M. **Gaps in knowledge among physicians regarding diagnostic criteria and management of polycystic ovary syndrome**. *Fertil Steril* (2017.0) **107** 1380-1386. DOI: 10.1016/j.fertnstert.2017.04.011
32. Piltonen TT, Ruokojarvi M, Karro H. **Awareness of polycystic ovary syndrome among obstetrician-gynecologists and endocrinologists in Northern Europe**. *PLoS ONE* (2019.0) **14** e226074. DOI: 10.1371/journal.pone.0226074
33. Ma R, Zou Y, Wang W. **Obesity management in polycystic ovary syndrome: disparity in knowledge between obstetrician-gynecologists and reproductive endocrinologists in China**. *BMC Endocr Disord* (2021.0) **21** 182. DOI: 10.1186/s12902-021-00848-w
34. Yan D, Yan-Fang W, Shi-Yang Z. **Is polycystic ovary syndrome appropriately diagnosed by obstetricians and gynaecologists across China: a nationwide survey**. *J Ovarian Res* (2021.0) **14** 25. DOI: 10.1186/s13048-021-00780-6
35. Ruan X, Li M, Mueck AO. **Why does polycystic ovary syndrome (PCOS) need long-term management?**. *Curr Pharm Des* (2018.0) **24** 4685-4692. DOI: 10.2174/1381612825666190130104922
36. Toçi E, Burazeri G, Kamberi H, Toçi D, Roshi E, Jerliu N, Bregu A, Brand H. **Health literacy and body mass index: a population-based study in a South-Eastern European country**. *J Public Health* (2021.0) **43** 123-130. DOI: 10.1093/pubmed/fdz103
37. Toci E, Burazeri G, Jerliu N. **Health literacy, self-perceived health and self-reported chronic morbidity among older people in Kosovo**. *Health Promot Int* (2015.0) **30** 667-674. DOI: 10.1093/heapro/dau009
38. Kearns K, Dee A, Fitzgerald AP. **Chronic disease burden associated with overweight and obesity in Ireland: the effects of a small BMI reduction at population level**. *BMC Public Health* (2014.0) **14** 143. DOI: 10.1186/1471-2458-14-143
39. Lim SS, Hutchison SK, Van Ryswyk E. **Lifestyle changes in women with polycystic ovary syndrome**. *Cochrane Database Syst Rev* (2019.0) **3** D7506
40. Paasche-Orlow MK, Wolf MS. **The causal pathways linking health literacy to health outcomes**. *Am J Health Behav* (2007.0) **31** S19-S26. DOI: 10.5993/AJHB.31.s1.4
|
---
title: Effects of dandelion root on rat heart function and oxidative status
authors:
- Kristina Radoman
- Vladimir Zivkovic
- Nebojsa Zdravkovic
- Natalia Vasilievna Chichkova
- Sergey Bolevich
- Vladimir Jakovljevic
journal: BMC Complementary Medicine and Therapies
year: 2023
pmcid: PMC9999560
doi: 10.1186/s12906-023-03900-5
license: CC BY 4.0
---
# Effects of dandelion root on rat heart function and oxidative status
## Abstract
This study aimed to examine the effects of dandelion root on rat heart function and oxidative status. At the beginning of the experimental protocol, Wistar albino rats were randomly classified into two groups (10 rats per group): 1. control group – animals that drank tap water; 2. experimental group – animals that drank dandelion root for four weeks. Every morning for four weeks, the animals received freshly boiled dandelion root in a volume of 250 ml. At the end of the dandelion administration, animals were sacrificed, and their hearts were isolated and retrogradely perfused according to the Langendorff technique at a gradually increasing perfusion pressure between 40 – 120 cm H2O. The following myocardial function parameters were measured: maximum rate of left ventricular pressure development (dp/dt max), minimum rate of left ventricular pressure development (dp/dt min), systolic left ventricular pressure (SLVP), diastolic left ventricular pressure (DLVP), heart rate (HR). In addition, the coronary flow (CF) was measured flowmetrically. Finally, blood samples were collected after sacrificing to determine oxidative stress biomarkers: nitrite (NO2−), superoxide anion radical (O2−), hydrogen peroxide (H2O2), the index of lipid peroxidation (TBARS), reduced glutathione (GSH), catalase (CAT) and superoxide dismutase (SOD). The present pioneer results indicated that dandelion root did not manifest a negative impact on functional aspects of isolated rat heart. In addition, dandelion consumption was not associated with promising results in terms of maintaining systemic redox balance.
## Introduction
Dandelion (Taraxacum officinale Weber) belongs to the Taraxacum genus, a member of the Asteraceae family and Cichoriodeae subfamily. It is widespread in the warm and humid zones characteristic of the northern hemisphere and has long been used in traditional medicine in the form of infusions and decoctions [1].
The medicinal raw materials of this plant consist of roots, leaves, and flowers. Environmental conditions, periods of plucking, different plucking methods, and drying methods significantly influence the chemical composition of the materials themselves. Dandelions are a rich source of various phyto compounds such as flavonoids, phenolic acids, terpenes, and polyphenolic compounds [2, 3]. The positive health effects of dandelions are a consequence of their phytochemical properties.
Due to such a composition, they exhibit potent antioxidant and anti-inflammatory effects. Thus, it has been observed that treating mice with a herbal mixture containing dandelion can cause a decrease in lipid peroxidation in serum and tissues and increase the activity of antioxidant protection enzymes (superoxide dismutase peroxidase and reduced glutathione) [4]. In addition, it has been observed in vitro and in vivo that dandelion can suppress the production of tumor necrosis factor (TNF-α) and interleukin-6 (IL-6). The animal model of diabetes mellitus–induced renal injury can be successfully treated with dandelion by reducing the production of interleukin-6 and TNF-α [5].
Although this plant is well known in traditional herbal medicine, there is only limited relevant scientific information on its pharmacological effects, with often contradictory results [6]. Recent research has presented dandelion as a new candidate for the fight against cancer since the aqueous extract from dandelion root has shown antineoplastic effects on aggressive and resistant cells of chronic myelomonocytic leukemia (CML) [7].
On the other hand, the cardiovascular effects of dandelion have been increasingly studied in recent years. They are based on its antiatherosclerotic potential resulting from antioxidant and anti-inflammatory properties. Recent experimental studies have shown that treatments with various dandelion extracts reduce adipogenesis and lipid accumulation, severity of atherosclerosis, serum concentrations of total cholesterol, triglycerides, and LDL cholesterol with an increase in HDL cholesterol [8–10].
However, despite the promising effects of dandelion in treating and preventing cardiovascular diseases, there is almost no data on the impact of this plant’s extracts on the heart. In that sense, this study aimed to examine the effects of dandelion root on rat heart function and oxidative status.
## Ethical aspects
This investigation was conducted in the Laboratory for cardiovascular physiology of the Faculty of Medical Sciences, University of Kragujevac, Serbia. The study protocol was approved by the Ethical Committee for the welfare of experimental animals of the Faculty of Medical Sciences, University of Kragujevac, Serbia. All experiments were performed following ARRIVE guidelines 2.0 for reporting animal research.
## Reagents
All reagents used in this study were of high purity and manufactured by Sigma-Aldrich Chemie GmbH, Germany.
## Experimental animals and groups
The study was carried out on 20 male Wistar albino rats (8 weeks old, body weight 250 ± 20 g). The animals consumed commercial rat food ($20\%$ protein rat food, Veterinary Institute Subotica, Serbia) and were housed under controlled environmental conditions at room temperature (22 ± 1 °C) with a 12-h light/day photoperiod. The rats had free access to food and tap water ad libitum. At the beginning of the experimental protocol, rats were randomly classified into two groups (10 rats per group): Control group – animals that drank tap water. Experimental group – animals that drank dandelion root for four weeks [6].
## Preparation of dandelion root
A total of 3 g chopped dandelion root is added to 300 ml of cold water, then heated and boiled for 5 min. After boiling, the root is left for ten minutes to cool and then filtered.
## Experimental protocol
Every morning for four weeks, the animals received fresh dandelion root in a 250 ml volume bottle [6]. In order to accurately record the intake of dandelion root, each animal was placed in a separate cage, while the volume of tea was recorded daily. The average daily dandelion root intake was 39.44 ± 2.67 ml in the experimental group, while the control group took tap water in an average amount of 42.85 ± 3.16 ml.
The animals were subjected to anesthesia at the end of the experimental protocol prior to sacrifice. A mixture of ketamine (Vet-Agro, Lublin, Poland) and xylazine (De Adelaar B.V, Venray, Holland) was prepared in a syringe. Administration of 25 µl/kg ketamine and 62.5 µl/kg xylazine was equivalent to the recommended dosage of 10 mg ketamine/kg and 5 mg xylazine/kg for rats [11]. The ketamine/xylazine mixture was administered i.p., and after 2 min, animals were sacrificed by decapitation.
## Evaluation ofex vivocardiac function
After decapitation, an emergency thoracotomy was performed, and rat hearts were isolated, attached via an aortic cannula, and retrogradely perfused using the Langendorff technique at a gradually increasing perfusion pressure between 40 – 120 cm H2O [12]. The hearts were perfused with Krebs–Henseleit solution (118 mM NaCl, 4.7 mM KCl, 2.5 mM CaCl2 2H2O, 1.7 mM MgSO4 H2O, 25 mM NaHCO3, 1.2 mM KH2PO4, 5.5 mM glucose, equilibrated with $95\%$ O2/$5\%$ CO2) and warmed to 37 °C (pH = 7.4). After heart perfusion commenced, a 30-min period was allowed for the hearts to stabilize. A transducer (BS473-0184, Experimetria Ltd., Budapest, Hungary) was used to monitor the following parameters of myocardial function: maximum rate of left ventricular pressure development (dp/dt max), minimum rate of left ventricular pressure development (dp/dt min), systolic left ventricular pressure (SLVP), diastolic left ventricular pressure (DLVP), heart rate (HR). The coronary flow (CF) was measured flowmetrically.
## Biochemical assay in blood
Blood samples were collected after decapitation in a vacutainer tube containing EDTA as an anticoagulant for the assay of pro-oxidative markers in the plasma and antioxidant markers in the lysate. The samples were centrifuged at 3000 rpm for 10 min at 4 °C using a Centurion centrifuge (K280R, UK). The plasma and erythrocyte lysate were then stored at -20 °C until analysis. All measurements were performed spectrophotometrically (Shimadzu UV-1800, Japan).
## Determination of oxidative status markers in blood
In plasma samples, the following oxidative stress markers were measured: nitrite (NO2), superoxide anion radical (O2−), hydrogen peroxide (H2O2), and the index of lipid peroxidation (measured as TBARS – thiobarbituric acid reactive substances).
Nitric oxide decomposes rapidly to form stable metabolite nitrite/nitrate products. The nitrite level was measured and used as an index of nitric oxide (NO) production using the Griess reagent. A total of 0.5 ml of plasma was precipitated with 200 μl of $30\%$ sulphosalicylic acid, vortexed for 30 min, and centrifuged at 3000 × g. Equal volumes of supernatant and Griess reagent containing $1\%$ sulphanilamide in $5\%$ phosphoric acid/$0.1\%$ naphthalene ethylenediamine dihydrochloride were added and incubated for 10 min in the dark, and the sample was measured at 543 nm. The nitrite levels were calculated using sodium nitrite as the standard [13].
The O2− concentration was measured after the reaction of nitro blue tetrazolium in Tris buffer with the plasma at 530 nm. Distilled water served as the blank [14].
The measurement of H2O2 is based on the oxidation of phenol red by H2O2 in a reaction catalysed by horseradish peroxidase (HRPO). Two hundred μl of plasma was precipitated with 800 ml of freshly prepared phenol red solution, followed by the addition of 10 μl of (1:20) HRPO (made ex tempore). Distilled water was used as the blank instead of the plasma sample. H2O2 was measured at 610 nm [15].
The degree of lipid peroxidation in the plasma samples was estimated by measuring TBARS using $1\%$ thiobarbituric acid in 0.05 NaOH, incubated with the plasma at 100 °C for 15 min, and measured at 530 nm. Distilled water served as the blank [16].
The activity of the following antioxidants in the lysate was determined: reduced glutathione (GSH), catalase (CAT), and superoxide dismutase (SOD). The level of reduced glutathione was determined based on GSH oxidation with 5,5-dithiobis-6,2-nitrobenzoic acid using a method by Beutler [17]. The CAT activity was determined according to Aebi [18]. The lysates were diluted with distilled water (1:7 v/v) and treated with chloroform-ethanol (0.6:1 v/v) to remove haemoglobin, and then 50 μl of CAT buffer, 100 μl of sample and 1 ml of 10 mM H2O2 were added to the samples. The detection was performed at 360 nm. SOD activity was determined by the epinephrine method of Beutler [19]. Lysate (100 μl) and 1 ml carbonate buffer were mixed, and then 100 μl of epinephrine was added. The detection was performed at 470 nm.
## Statistical analysis
The collected data were processed using the SPSS 21.0 software (SPSS Inc., Chicago, IL, USA). The Shapiro–Wilk test was used to examine data distribution normality. Depending on the data distribution, Student’s t-test and Kruskal–Wallis tests were applied to analyse parametric and nonparametric data, respectively. The confidence interval in all statistical analyzes is $95\%$, with a statistical significance $p \leq 0.05$ and a high statistical significance $p \leq 0.01.$ Data are described as mean ± standard deviation (SD).
## Ex vivoparameters of cardiac function
The mean values of the maximum rate of change in left ventricular pressure (dp / dt max) did not differ statistically significantly ($p \leq 0.05$) between the control and experimental group at all values of coronary perfusion pressure (Fig. 1). The mean values of the minimum rate of change in left ventricular pressure (dp / dt min) did not differ significantly ($p \leq 0.05$) between the control and experimental group at all values of coronary perfusion pressure (Fig. 2).Fig. 1Average values of the maximum rate of pressure change in the left ventricle (dp/dt max (mmHg/s)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SDFig. 2Average values of the minimum rate of pressure change in the left ventricle (dp/dt min (mmHg/s)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD The average values of systolic and diastolic pressure in the left ventricle (SLVP and DLVP) did not differ statistically significantly ($p \leq 0.05$) between the dandelion-treated and control group (Figs. 3 and 4). The mean values of the heart rate (HR) were very close between the groups, thus there was no statistical difference in this case either ($p \leq 0.05$) (Fig. 5).Fig. 3Average values of systolic left ventricular pressure (SLVP (mmHg)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SDFig. 4Average values of diastolic left ventricular pressure (DLVP (mmHg)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SDFig. 5Average values of heart rate (HR (bpm)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD The mean values of coronary flow during all values of coronary perfusion pressure were higher in the group that was administered dandelion root but without statistical confirmation ($p \leq 0.05$) (Fig. 6).Fig. 6Average values of coronary flow (CF (ml/min)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD
## Oxidative status markers in blood
In rats treated with dandelion root, there was a statistically significant increase ($p \leq 0.05$) in hydrogen peroxide (H2O2) values compared to the control group (Fig. 7). Lipid peroxidation index (TBARS) showed similar dynamics, whose values were significantly higher ($p \leq 0.05$) in the experimental group compared to the control group (Fig. 8). The superoxide anion radical (O2−) values were statistically significantly lower ($p \leq 0.05$) in the rats on dandelion consumption than in the control group (Fig. 9).Fig. 7Average values of the concentration of hydrogen peroxide (H2O2 (nmol/ml)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD. *— statistically significant, $p \leq 0.05$Fig. 8Average values of the concentration of thiobarbituric acid reactive substances (TBARS (µmol/ml)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD. *— statistically significant, $p \leq 0.05$Fig. 9Average values of the concentration of superoxide ion radical (O2.− (nmol/ml)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD. *— statistically significant, $p \leq 0.05$ *Unlike previous* biomarkers, nitrite (NO2−) values did not differ statistically significantly between groups ($p \leq 0.05$) (Fig. 10). On the other hand, catalase and superoxide dismutase activity was significantly lower ($p \leq 0.05$) in the dandelion-treated group (Figs. 11 and 12). In comparison, reduced glutathione activity was statistically higher (Fig. 13) than in the control group ($p \leq 0.05$).Fig. 10Average values of the concentration of nitrite ions (NO2− (nmol/ml)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SDFig. 11Average values of the concentration of superoxide dismutase (SOD (U/g Hb*10.3)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD. *— statistically significant, $p \leq 0.05$Fig. 12Average values of the concentration of the catalase enzyme (CAT (U/g Hb*10.3)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD. *— statistically significant, $p \leq 0.05$Fig. 13Average values of the concentration of glutathione (GSH (U/g Hb*10.3)) in the control and dandelion infusion-treated groups. Data is presented in the form of mean value ± SD. *— statistically significant, $p \leq 0.05$
## Discussion
The main goal of the present study was to examine the effect of dandelion root on rat heart function and oxidative status.
In the first part of the study, the effect of dandelion root on the cardiac muscle was studied through an assessment of the cardiodynamic parameters of an isolated rat heart. As previously mentioned, to our best knowledge, this is one of the rare studies covering this topic in the available literature. Therefore, we can consider this a "pioneering" investigation, i.e., a topic that requires further research.
The beneficial effects of dandelion in maintaining cardiovascular homeostasis have been known for a long time in traditional and folk medicine [9, 10]. However, in recent years investigations have become more complex and focused on different aspects of this field. The study by Majewski et al. was primarily concerned with examining the effect of plant extract on rat antioxidant status and lipid profile [20]. The single cardiodynamic parameter analyzed in this study was heart rate (HR). As in the present research, authors also used 8-week-old Wistar albino rats. The differences between the protocols are reflected in the fact that Majewski et al. used an ethanolic extract of dandelion leaves and flowers [20].
In contrast, we used an aqueous extract of dandelion root. The reason for using a root is our literature-based hypothesis that its phytochemical properties remain more stable when tea is made from this plant part [2, 3].
The results of this study, in terms of heart rate (HR), are in agreement with those of the above study. In both cases, there was no evident change in heart frequency after applying the dandelion extract, suggesting that dandelion root does not modify the function of the heart conduction system.
Langendorff preparation optimizes the detection of ventricular function and permits exceptionally accurate measurement intervals [21]. Indicators of contractile (dp / dt max) and relaxant (dp / dt min) force of the heart were also not affected by dandelion root. Similarly, dandelion did not impair systolic (SLVP) and diastolic (DLVP) function of the cardiac muscle as well as reactivity of coronary circulation (CF). Taken together, these results indicated that applied dandelion root did not negatively impact the function and perfusion of isolated rat heart. However, the exact mechanism of these effects requires a more complex experimental approach. In addition, considering the steady trends of non-statistical cardiodynamic changes, it should be pointed out that higher doses or extended time of exposure seem to be associated with different findings, which imposes the need for further research.
One of the rare studies of the influence of dandelion on any muscle was recently published [22]. Namely, ethyl acetate dandelion extract’s effects were evaluated on mouse airway smooth muscle. This extract was found to relax mouse smooth muscle via inhibition of L-type voltage-dependent calcium channel and non-selective cationic channel, which, at least theoretically, could be the site of action in rat cardiomyocytes [22].
On the other hand, in the second part of the research, we seek to examine whether, taking into account the antioxidant properties of dandelion, it can disrupt the redox homeostasis of rats, which can also be responsible for the obtained effects within the heart. It is well known that the accumulation of reactive oxygen species (ROS) leads to biochemical, structural, and functional disorders in cells [23]. It has been proven that dandelion successfully prevents synthesis and increases the removal of different ROS, especially H2O2 and O2−. The capacity to remove free radicals has been attributed to phenolic compounds in dandelion flowers [24].
Dandelion is widely used as a folk remedy against various disorders such as liver disease, bile, indigestion, and rheumatic diseases. In one research, dandelion leaf has been shown to possess a protective effect against acute pancreatitis caused by cholecystokinin octapeptide and acute lung and liver damage. This protective effect is due to components from dandelion leaves with flavonoids and polyphenols [25]. Plant flavonoids act as scavengers of free radicals and turn them into less reactive or bind metal ions preventing their production [26].
In our study, dandelion root affected the production of measured pro-oxidants oppositely. Namely, while the release of O2− was decreased, the concentration of H2O2 and TBARS were higher than in the control group. The explanation for this different impact on ROS generation from the point of this study is difficult to find. Literature data show that dandelion lowers the concentration of TBARS and H2O2, but only in in vitro conditions [27], while there are no data related to in vivo biological systems. It seems that the well-known antioxidant effects of dandelion directed towards individual ROS can only be achieved in cell lines, while in vivo systems require a higher dose or length of exposure to achieve them. In addition, the worrying increase of both investigated biomarkers after the use of dandelion (TBARS and H2O2) may be a consequence of the compensatory overproduction of other ROS in a situation where O2− generation is suppressed [23]. Nevertheless, the reduction of O2− we noted is a promising finding since it is one of the most toxic known ROS and completely correlates with literature findings [24, 25].
In addition, the concentrations of free radicals depend not only on their production but also on the expression and activity of antioxidant enzymes [28]. Flavonoids found in the dandelion extract have a beneficial effect on cardiovascular function based on their antioxidant features and the ability to increase the expression and activity of antioxidant enzymes [29, 30].
In the present investigation, the four-week dandelion consumption also impacted estimated antioxidant enzymes differently. While CAT and SOD activity was lower, reduced glutathione activity was improved compared to control. The drop in CAT values may be a reflection of the exhaustion of its activity as a consequence of the increased release of H2O2 we found. Furthermore, previous research examined the hypolipidemic and antioxidant potential of animals treated with dandelion leaf and root extract [31]. The activity of reduced glutathione was strongly improved in the group treated with dandelion root and leaf compared to the control group, while catalase activity was lower, which is in complete correlation with the results of our study [31].
A study by Park and associates compared the antioxidant and anti-inflammatory activity of methanolic and aqueous extracts of Taraxacum officinale. The activity of reduced glutathione and other antioxidant enzymes such as superoxide dismutase, catalase, glutathione peroxidase, and glutathione reductase were restored after using the extract. Methanol extract showed more potent antioxidant and anti-inflammatory abilities than aqueous extract, which can be attributed to the higher total content of phenol, luteolin, and cichoric acid [32].
Finally, the present research has some limitations. First, the longer duration of dandelion consumption and assessment of different doses could have a stronger impact on both heart and oxidative status of rats. Second, due to technical limitations, the study was limited in mechanistic approach, i.e., patch clamp assessment of ionic currents within the cardiomyocytes would provide potential mechanisms of dandelion effects. Pathohistological examination of heart tissue can also serve this purpose.
## Conclusion
To the best of our knowledge, this is the only experimental study investigating the effects of dandelion root on the mammalian heart and oxidative status in the available literature. The present pioneer results indicated that dandelion root did not manifest a negative impact on functional aspects of isolated rat heart. In addition, the impact of dandelion root on systemic oxidative status was variable and individually directed toward measured biomarkers. Therefore the global systemic antioxidant effect was not achieved. From a clinical perspective, these findings may be an excellent basis as the first step in developing an animal model of heart failure or other cardiovascular disease, where dandelion usage may have a practical benefit.
## Authors’ information
Non applicable.
## References
1. Kania-Dobrowolska M, Baraniak J. **Dandelion (Taraxacum officinale L.) as a source of biologically active compounds supporting the therapy of co-existing diseases in metabolic syndrome**. *Foods* (2022) **11** 2858. DOI: 10.3390/foods11182858
2. Jedrejek D, Lis B, Rolnik A, Stochmal A, Olas B. **Comparative phytochemical, cytotoxicity, antioxidant and haemostatic studies of Taraxacum officinale root preparations**. *Food Chem Toxicol* (2019) **126** 233-47. DOI: 10.1016/j.fct.2019.02.017
3. Jędrejek D, Kontek B, Lis B, Stochmal A, Olas B. **Evaluation of antioxidant activity of phenolic fractions from the leaves and petals of dandelion in human plasma treated with H**. *Chem Biol Interact* (2017) **262** 29-37. DOI: 10.1016/j.cbi.2016.12.003
4. Kim MY, Cheong SH, Kim MH, Son C, Yook HS, Sok DE, Kim JH, Cho Y, Chun H, Kim MR. **Leafy vegetable mix supplementation improves lipid profiles and antioxidant status in C57BL/6J mice fed a high fat and high cholesterol diet**. *J Med Food* (2009) **12** 877-84. DOI: 10.1089/jmf.2008.1125
5. Tian L, Fu P, Zhou M, Qi J. **Dandelion sterol improves diabetes mellitus-induced renal injury in in vitro and in vivo study**. *Food Sci Nutr* (2021) **9** 5183-97. DOI: 10.1002/fsn3.2491
6. Majewski M, Lis B, Juśkiewicz J, Ognik K, Jedrejek D, Stochmal A, Olas B. **The composition and vascular/antioxidant properties of Taraxacum officinale flower water syrup in a normal-fat diet using an obese rat model**. *J Ethnopharmacol* (2021) **265** 113393. DOI: 10.1016/j.jep.2020.113393
7. Menke K, Schwermer M, Felenda J, Beckmann C, Stintzing F, Schramm A, Zuzak TJ. **Taraxacum officinale extract shows antitumor effects on pediatric cancer cells and enhance mistletoe therapy**. *Complement Ther Med* (2018) **40** 158-64. DOI: 10.1016/j.ctim.2018.03.005
8. Olas B. **New perspectives on the effect of Dandelion, its food products and other preparations on the cardiovascular system and its diseases**. *Nutrients* (2022) **14** 1350. DOI: 10.3390/nu14071350
9. Ignat MV, Coldea TE, Salanță LC, Mudura E. **Plants of the spontaneous flora with beneficial action in the management of diabetes, hepatic disorders, and cardiovascular disease**. *Plants (Basel)* (2021) **10** 216. DOI: 10.3390/plants10020216
10. González-Castejón M, García-Carrasco B, Fernández-Dacosta R, Dávalos A, Rodriguez-Casado A. **Reduction of adipogenesis and lipid accumulation by Taraxacum officinale (dandelion) extracts in 3T3L1 adipocytes: an in vitro study**. *Phytother Res* (2014) **28** 745-52. DOI: 10.1002/ptr.5059
11. Barriga-Rivera A, Tatarinoff V, Lovell NH, Morley JW, Suaning GJ. **Long-term anesthetic protocol in rats: feasibility in electrophysiology studies in visual prosthesis**. *Vet Ophthalmol* (2018) **21** 290-7. DOI: 10.1111/vop.12507
12. Watanabe M, Okada T. **Langendorff Perfusion Method as an Ex vivo model to evaluate heart function in rats**. *Methods Mol Biol* (2018) **1816** 107-16. DOI: 10.1007/978-1-4939-8597-5_8
13. Green LC, Wagner DA, Glogowski J, Skipper PL, Wishnok JS, Tannenbaum SR. **Analysis of nitrate, nitrite and [15 N] nitrate in biological fluids**. *Anal Biochem* (1982) **126** 131-8. DOI: 10.1016/0003-2697(82)90118-x
14. Auclair C, Voisin E, Greenvvald RA. **Nitroblue tetrazolium reduction**. *Handbook of methods for oxygen radical research* (1985) 123-32
15. Pick E, Keisari Y. **A simple colorimetric method for the measurement of hydrogen peroxide produced by cells in culture**. *J Immunol Meth* (1980) **38** 161-70. DOI: 10.1016/0022-1759(80)90340-3
16. Ohkawa H, Ohishi N, Yagi K. **Assay for lipid peroxides in animal tissues by thiobarbituric acid reaction**. *Anal Biochem* (1979) **95** 351-8. DOI: 10.1016/0003-2697(79)90738-3
17. Beutler E, Beutler E. **Superoxide dismutase**. *Red cell metabolism. A manual of biochemical methods* (1984) 83-5
18. Aebi H. **Catalase in vitro**. *Methods Enzymol* (1984) **105** 121-6. DOI: 10.1016/s0076-6879(84)05016-3
19. Beutler E, Duron O, Kelly BM. **Improved method for the determination of blood**. *Glutathione J Lab Clin Med* (1963) **61** 882-8. PMID: 13967893
20. Majewski M, Lis B, Juśkiewicz J, Ognik K, Borkowska-Sztachańska M, Jedrejek D, Stochmal A, Olas B. **Phenolic fractions from Dandelion Leaves and Petals as modulators of the antioxidant status and lipid Profile in an in vivo study**. *Antioxid (Basel)* (2020) **9** 131. DOI: 10.3390/antiox9020131
21. Ozkanlar Y, Kijtawornrat A, Hamlin RL, Keene BW, Roche BM. **Acute cardiovascular effects of tacrolimus in the isolated guinea pig heart**. *J Vet Pharmacol Ther* (2005) **28** 313-6. DOI: 10.1111/j.1365-2885.2005.00631.x
22. Zhao P, Liu J, Ming Q, Tian D, He J, Yang Z, Shen J, Liu QH, Yang X. **Dandelion extract relaxes mouse airway smooth muscle by blocking VDLCC and NSCC channels**. *Cell Biosci* (2020) **10** 125. DOI: 10.1186/s13578-020-00470-8
23. Tien Vo TT, Vo QC, Tuan VP, Wee Y, Cheng HC, Lee IT. **The potentials of carbon monoxide-releasing molecules in cancer treatment: an outlook from ROS biology and medicine**. *Redox Biol* (2021) **46** 102124. DOI: 10.1016/j.redox.2021.102124
24. Wang L, Li L, Gao J, Huang J, Yang Y, Xu Y, Liu S, Yu W. **Characterization, antioxidant and immunomodulatory effects of selenized polysaccharides from dandelion roots**. *Carbohydr Polym* (2021) **260** 117796. DOI: 10.1016/j.carbpol.2021.117796
25. Liu N, Song M, Wang N, Wang Y, Wang R, An X, Qi J. **The effects of solid-state fermentation on the content, composition and in vitro antioxidant activity of flavonoids from dandelion**. *PLoS ONE* (2020) **15** e0239076. DOI: 10.1371/journal.pone.0239076
26. Sun Y, Wu Y, Wang Z, Chen J, Yang Y, Dong G. **Dandelion Extract Alleviated Lipopolysaccharide-Induced oxidative stress through the Nrf2 pathway in bovine mammary epithelial cells**. *Toxins (Basel)* (2020) **12** 496. DOI: 10.3390/toxins12080496
27. Jędrejek D, Kontek B, Lis B, Stochmal A, Olas B. **Evaluation of antioxidant activity of phenolic fractions from the leaves and petals of dandelion in human plasma treated with H2O2 and H2O2/Fe**. *Chem Biol Interact* (2017) **262** 29-37. DOI: 10.1016/j.cbi.2016.12.003
28. Ayoub KF, Pothineni NVK, Rutland J, Ding Z, Mehta JL. **Immunity, inflammation, and oxidative stress in Heart failure: emerging molecular targets**. *Cardiovasc Drugs Ther* (2017) **31** 593-608. DOI: 10.1007/s10557-017-6752-z
29. Jia JY, Zang EH, Lv LJ, Li QY, Zhang CH, Xia Y, Zhang L, Dang LS, Li MH. **Flavonoids in myocardial ischemia-reperfusion injury: therapeutic effects and mechanisms**. *Chin Herb Med* (2020) **13** 49-63. DOI: 10.1016/j.chmed.2020.09.002
30. Cebova M, Pechanova O. **Protective Effects of Polyphenols against Ischemia/Reperfusion Injury**. *Molecules* (2020) **25** 3469. DOI: 10.3390/molecules25153469
31. Choi UK, Lee OH, Yim JH, Cho CW, Rhee YK, Lim SI, Kim YC. **Hypolipidemic and antioxidant effects of dandelion (Taraxacum officinale) root and leaf on cholesterol-fed rabbits**. *Int J Mol Sci* (2010) **11** 67-78. DOI: 10.3390/ijms11010067
32. Park CM, Park JY, Noh KH, Shin JH, Song YS. **Taraxacum officinale Weber extracts inhibit LPS-induced oxidative stress and nitric oxide production via the NF-κB modulation in RAW 264.7 cells**. *J Ethnopharmacol* (2011) **133** 834-42. DOI: 10.1016/j.jep.2010.11.015
|
---
title: 'Maternal e-cigarette use can disrupt postnatal blood-brain barrier (BBB) integrity
and deteriorates motor, learning and memory function: influence of sex and age'
authors:
- Sabrina Rahman Archie
- Ali Ehsan Sifat
- Yong Zhang
- Heidi Villalba
- Sejal Sharma
- Saeideh Nozohouri
- Thomas J. Abbruscato
journal: Fluids and Barriers of the CNS
year: 2023
pmcid: PMC9999561
doi: 10.1186/s12987-023-00416-5
license: CC BY 4.0
---
# Maternal e-cigarette use can disrupt postnatal blood-brain barrier (BBB) integrity and deteriorates motor, learning and memory function: influence of sex and age
## Abstract
Electronic nicotine delivery systems (ENDS), also commonly known as electronic cigarettes (e-cigs) are considered in most cases as a safer alternative to tobacco smoking and therefore have become extremely popular among all age groups and sex. It is estimated that up to $15\%$ of pregnant women are now using e-cigs in the US which keeps increasing at an alarming rate. Harmful effects of tobacco smoking during pregnancy are well documented for both pregnancy and postnatal health, however limited preclinical and clinical studies exist to evaluate the long-term effects of prenatal e-cig exposure on postnatal health. Therefore, the aim of our study is to evaluate the effect of maternal e-cig use on postnatal blood-brain barrier (BBB) integrity and behavioral outcomes of mice of varying age and sex. In this study, pregnant CD1 mice (E5) were exposed to e‐Cig vapor ($2.4\%$ nicotine) until postnatal day (PD) 7. Weight of the offspring was measured at PD0, PD7, PD15, PD30, PD45, PD60 and PD90. The expression of structural elements of the BBB, tight junction proteins (ZO-1, claudin-5, occludin), astrocytes (GFAP), pericytes (PDGFRβ) and the basement membrane (laminin α1, laminin α4), neuron specific marker (NeuN), water channel protein (AQP4) and glucose transporter (GLUT1) were analyzed in both male and female offspring using western blot and immunofluorescence. Estrous cycle was recorded by vaginal cytology method. Long‐term motor and cognitive functions were evaluated using open field test (OFT), novel object recognition test (NORT) and morris water maze test (MWMT) at adolescence (PD 40–45) and adult (PD 90–95) age. In our study, significantly reduced expression of tight junction proteins and astrocyte marker were observed in male and female offspring until PD 90 ($P \leq 0.05$). Additionally, prenatally e-cig exposed adolescent and adult offspring showed impaired locomotor, learning, and memory function compared to control offspring ($P \leq 0.05$). Our findings suggest that prenatal e-cig exposure induces long-term neurovascular changes of neonates by disrupting postnatal BBB integrity and worsening behavioral outcomes.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12987-023-00416-5.
## Introduction
Maternal smoking during pregnancy remains one of the most important modifiable risk factors for poor pregnancy outcomes in the U.S. as well as globally [1–4]. Maternal smoking not only adversely impacts maternal health but also results in poor fetal outcomes including but not limited to tobacco-induced abortions and stillbirths, low birth weight (LBW), sudden infant death syndrome (SIDS), preterm birth, neurological and cognitive delays, congenital disabilities, colic, asthma and atopic pregnancies [3–5]. Considering the impact on the cerebrovascular system, several studies have demonstrated the long-term neurotoxic effects of maternal smoking on neonatal/postnatal rodent offspring including hyperactivity, decreased learning and memory function, increased depression, altered brain development and enhanced hypoxic ischemic brain injury [6–13].
Electronic cigarette or e-cigarette (e-cig), commonly known as vaping is a battery powered electronic nicotine delivery systems (ENDS), which has become extremely popular among all age groups and sexes since it entered the US market in 2007 [6, 14]. E-cig usage is considered as a safer alternative to conventional tobacco smoke because of their reportedly lower levels of nicotine and potential carcinogens and has been proposed to be utilized as a smoking cessation tool and for recreational purposes [15–17]. Surprisingly, recent studies have reported that e-cig usage is also popular in women of child-bearing age and up to $15\%$ of the pregnant women are now using e-cigs [17–19]. E-cigs mainly contain a solution of nicotine along with several additives including propylene glycol, vegetable glycerin, acrolein, formaldehyde, flavoring agents and other trace elements, some of which may be toxic for health including developing fetus and offspring [14, 20]. Although long-term toxic effects of prenatal tobacco smoking on postnatal health are well documented and well established, limited preclinical and clinical studies exist to evaluate the impact of maternal vaping on neonatal health and neurovascular effects. However, increasing evidence from animal studies have demonstrated that maternal vaping may also impact brain development of neonates and adolescents. It has been reported in clinical studies that e-cig use during pregnancy can impair fetal development through altered neurologic, pulmonary, and cardiovascular dysfunction [21]. Recent studies, including our lab, have shown that prenatal e-cig exposure is associated with several cerebrovascular and neurological dysfunctions including genetic/epigenetic alteration in developing brain, cognitive dysfunction, decreased brain glucose utilization and increased hypoxic-ischemic brain injury, alteration in neural regulators of energy homeostasis, and inflammation, disruption in brain excitatory/inhibitory neuron balance and calcium homeostasis, microglial cell death, alteration in gene expression in the frontal cortex and localized inflammation of the hippocampus [6, 14, 22–28]. A challenge and potential danger is that most of the e-cigs available in the market contain various concentrations of nicotine. As nicotine can cross the blood-placental barrier and accumulate in fetal blood, [29, 30] it is important to understand the neurologic effects of prenatal nicotine exposure to a developing fetus.
The blood-brain barrier (BBB) is the vital component of central nervous system (CNS) which plays a major role in maintaining the homeostasis of the brain microenvironment by controlling the passage of substances and regulating the trafficking of immune cells between the blood and the brain. BBB development starts during the fetal stage, and it is well constructed by the point of birth, especially for restriction of macromolecules and proteins. At the cellular level, the BBB consists of microvascular endothelial cells (EC) lining the luminal walls of brain microvessels alongside closely associated pericytes embedded within the basal membrane and surrounded by astrocytic end-feet processes. Studies have shown that BBB disruption has been associated with the onset and/or progression of major neurological disorders including ischemic stroke, multiple sclerosis, amyotrophic lateral sclerosis, traumatic brain injury, brain tumor, Alzheimer’s and Parkinson’s disease, epilepsy, edema and glaucoma [31]. Several preclinical studies, including our lab, have demonstrated that tobacco smoke and e-cig disrupt the BBB by decreased tight junction proteins, increased permeability [32–34], altered ion transporter expression and activity at BBB [35–37] and can induce oxidative stress and inflammation in brain [38] which may worsen brain injury and ischemic stroke outcomes [39]. Although there are several studies available demonstrating the neurotoxic impact of e-cig exposure on neurological and cerebrovascular systems, extensive research needs to be conducted to assess the long-term effect of maternal e-cig exposure on postnatal cerebrovascular dysfunction. To the best of our knowledge as of today, there is no such a study reported on the effect of maternal vaping on postnatal cerebrovascular disruption. Given the rapidly growing popularity of e-cig use in pregnant women and its potential toxic effect(s) on postnatal health, it is critical to conduct preclinical studies to evaluate the long-term effect of maternal e-cig exposure on postnatal brain health in different developmental ages and sex with understanding any underlying mechanisms. Therefore, the aim of this study is to evaluate the effect of maternal e-cig use during pregnancy on postnatal BBB integrity, motor, learning and memory function at different ages using an in-vivo mice model.
## Animals and surgical procedures
All studies were approved by the IACUC of Texas Tech University Health Sciences Center, Lubbock, Texas (IACUC protocol# 20026). Experiments were performed in accordance with relevant guidelines and regulations. Female CD1 pregnant mice (Charles River Laboratories, Inc., Wilmington, MA; Cat# CRL: 22, RRID: IMSR_CRL:22) and after delivery their offspring were kept under standardized light and dark conditions (12 h), humidity ($70\%$), and temperature (22 °C). Pregnant mice were singly housed. Offspring were separated into male and female after weaning (postnatal day 21–22) and housed in a group of 2–5. They were given ad libitum access to food and water. Animal behavior was monitored daily to minimize animal suffering. We applied the following exclusion criteria to our experiments: severe weight loss, infections, or significant behavioral deficits (decreased mobility, seizures, lethargy). No animal was excluded from this study. The research design is depicted in Fig. 1. A total number of 176 ($$n = 16$$ for mother and $$n = 160$$ for offspring) mice were used to perform this study. All experiments were conducted in compliance with the ARRIVE guidelines. Fig. 1Study design. Pregnant CD1 were exposed to Blu e-cigarette from gestational day 5 (E5) to postnatal day 7 (PD7). At the end of the exposure, plasma nicotine and cotinine level were measured by LCMS/MS, and body weight was measured at PD7, PD23, PD45 and PD90. Mice were sacrificed and brain was extracted at every time point to evaluate blood-brain barrier (BBB) integrity by western blot and immunofluorescence. Open field test, novel object recognition test and morris water maze test were conducted at adolescent and adult time point to evaluate hyperactivity and learning-memory function
## In-vivo e-cig vaping
Pregnant CD1 mice were exposed (via direct inhalation) to e-cig vapor containing $2.4\%$ nicotine (Blu™, 24 mg/ml nicotine) mixed with oxygenated air or oxygenated air alone, 6 times/day; 1 cartridge/day from gestational day 5 (E5) until delivery, and it was continued after delivery until the pups were 7 days old. After birth, the pups would be exposed to nicotine via lactation [40, 41]. This exposure model was adopted following a study by Sifat et al. who investigated the effects of prenatal electronic cigarette exposure on offspring in mice model [6, 8]. In our study, Blu™ was used as this is one of the most popular e-cig brands still on the market, the cig-a-like structure fits well in our smoking apparatus, and there have been previously reported studies using Blu™ [6, 34, 37]. A modified CORESTA (Cooperation Centre for Scientific Research Relative to Tobacco) standard smoking protocol adapted to study e-cig exposure (27.5 ml puff depth volume, 3 s puff duration, 2 puffs per 60 s, 32 puffs/session) was followed in the laboratory. E-cig vapor was generated using a Single Cigarette Smoking Machines (SCSM, CH Technologies Inc.) following a previously published method used by our laboratory [6, 34, 37]. This method was developed to mimic the smoking behavior of a human chronic/heavy smoker/vaper and yields plasma levels of cotinine (111 ng/ml) which is in the range of blood cotinine level (30–250 ng/ml) found in other preclinical rodent models of chronic e-cig exposure [42, 43]. The smoking exposure was done between 9 a.m. and 2 p.m.
## Nicotine and cotinine level measurement by LCMS/MS in offspring plasma
Concentration of nicotine and its principal metabolite cotinine were measured from prenatally e-cig exposed mice plasma at PD7 by LCMS/MS analysis using Cotinine-d3 (MilliporeSigma, St. Louis, MO, USA) as an internal standard (IS) following a previously published method [44]. In brief, samples were prepared by protein precipitation of 25 µL mouse plasma using acetonitrile at 1:8 ratio. Mass Spectrometer was operated in positive polarity under the multiple reaction monitoring mode using electrospray ionization technique. The transitions of m/z 163.2 → 132.1, 177.2 → 98.0 and 180.2 → 101.2 were used to measure the nicotine, cotinine, and IS, respectively. The elution of nicotine (MilliporeSigma), cotinine (MilliporeSigma), and IS were at 1.89, 1.77, and 1.76 min, respectively. This was achieved with a mobile gradient phase consisting of 5 mM ammonium bicarbonate, acetonitrile, and methanol (3:1, v/v) at a 0.3 mL/min flow rate on a Kinetex EVO C18 column (Phenomenex, Torrance, CA, USA).
## Body weight measurement
Weight of the mother was measured before e-cig exposure (E5) and post-delivery. Litter size and litter weight were also counted and measured respectively. The weight of offspring was measured at several time points including PD7, PD15, PD30, PD45, PD60, and PD90 to evaluate the body growth and development. To measure the brain to body weight ratio, brain was weighed after extraction at PD7, PD23, PD45, PD90 and brain-to-body weight was measured.
## Western blot
Prenatally e-cig exposed, or control mice were sacrificed at each time points (PD7, PD23, PD45 and PD90) and brains were isolated. Brains were lysed using RIPA buffer to isolate protein lysate. Protein concentration of protein lysates were determined using a bicinchoninic acid (BCA) assay. Exactly 30 μg of protein from each sample was loaded and separated using a $10\%$ Tris-glycine polyacrylamide precast gel (Bio-Rad Laboratories, Hercules, CA; Cat# 4568034). This method has been used previously to analyze Western blot immunoreactivity [6]. Protein samples were then transferred to a polyvinylidene difluoride membrane (Thermo Fisher; Cat# IPVH00010), and then membranes were incubated in blocking buffer ($0.2\%$ Tween-20 containing Tris-buffered saline (TBST) with $5\%$ bovine serum albumin) to block the nonspecific protein bands for 2 h at room temperature. Membranes were incubated with rabbit polyclonal anti-ZO-1 antibody (1: 2000, Thermo Fisher; Cat# 40–2200), mouse monoclonal anti-claudin-5 antibody (1: 2000, Thermo Fisher; Cat# 35–2500), rabbit monoclonal anti-occludin antibody (1: 1000, Cell Signaling; Cat# E6B4R), rabbit polyclonal anti-laminin α1 antibody (1:2000, Thermo Fisher; Cat# PA1-16730), rabbit polyclonal anti-laminin α4 antibody (1:2000, Sigma; Cat# SAB4501719), rabbit monoclonal anti-GFAP antibody (1: 2000, Cell Signaling; Cat# DIF4Q), Rabbit monoclonal anti-PDGFRβ antibody (1: 1000, Cell Signaling; Cat# 28E1), mouse monoclonal anti-AQP4 antibody (1: 1000, Santa Cruz; Cat# 2), rabbit monoclonal anti-NeuN antibody (1: 1000, Cell Signaling; Cat# D3S3I), rabbit monoclonal anti-Glut-1 antibody (1: 2000, Cell Signaling; Cat# D3J3A) and mouse monoclonal anti-beta-actin antibody (1: 10000 MilliporeSigma; Cat# A5441) in TBST with $5\%$ bovine serum albumin at 4 °C overnight. After 4 times washing with TBST for 15 min each cycle, membranes were incubated with anti-rabbit (Sigma Aldrich; Cat# GENA934- 1ML, RRID: AB_2722659) or anti-mouse (Sigma Aldrich; Cat# GENXA931-1ML, RRID: AB_772209) IgG-horseradish peroxidase secondary antibody (1:10000) in TBST with $5\%$ bovine serum albumin for 2 h at room temperature. After 4 times of 15 min wash with TBST, the protein signals were detected by enhanced chemiluminescence-detecting reagents (Thermo Fisher; Cat# 34577) and visualized in X-ray films in the dark. The protein bands were quantified relative to beta-actin in Image J software.
## Immunofluorescence
Immunofluorescence staining was performed as previously described with modifications [45, 46]. Mice were euthanized by isoflurane overdose at each time point. The brains were sectioned at 20 µM of thickness, fixed with $4\%$ paraformaldehyde (Thermo Fisher) for 15 min, then permeabilized with $0.1\%$ Triton X-100 for 10 min. After washing with the phosphate-buffered saline (PBS) for 15 min, the sections were blocked for 1 h and incubated overnight with primary antibodies for ZO-1 (1:100, Thermo Fisher) claudin-5 (1: 100, Thermo Fisher) and GFAP (1:100, Cell Signaling), respectively. Alexa fluorescent secondary antibodies (Thermo Fisher) were used at 1:400 dilutions for 1 h. After counterstaining with 4′,6-diamidino-2-phenylindole (DAPI) for nucleus and washing with PBS, the sections were mounted with Permount (Thermo Fisher). The whole sections were scanned with a Leica Stellaris SP8 Falcon microscope (Leica Microsystem) and the images (20X magnitude) were captured with the same microscope. Mean total fluorescence intensity was calculated for each color channel and intensity of green color (ZO-1/GFAP) and red color (claudin-5) was expressed relative to blue color (DAPI). Cortex and hippocampus of both hemispheres of each brain section were used to evaluate the expression levels of ZO-1, claudin-5 and GFAP. To minimize the subjective bias, all images for ZO-1, claudin-5 and GFAP expression analysis were captured under the same microscopic parameter (laser power, pinhole size, exposure time) setting.
## Open field test (OFT)
OFT was performed to evaluate the locomotor activity of the prenatally e-cig exposed or control mice both male and female at PD45 (adolescent) and PD90 (adult) following our previously published study [47, 48]. Versamax software (Accuscan Instruments., Columbus, OH) was used to automatically calculate the total distance traveled by the mice. Briefly, mice were introduced to 16″ × 16″ unobstructed glass chamber and their activities were monitored and recorded for 1 h. The first 10 min of 1 h was excluded as the acclimatization period. All experiments were performed between 8 and 10 am. Fecal boli was counted for each mouse after completing of the OFT to measure stress/anxiety level following published literatures [49, 50].
## Novel object recognition test (NORT)
NORT was performed to evaluate short-term memory retention. It was done by a slight modification of previously published literature [51]. For habituation, each mouse was placed in a wooden box without any object for 10 min, 24 h before the test. On the testing day, mouse was placed in that same box containing two identical green round blocks for 5 min for the familiarization phase. After a 30 min interval, during the test phase one of the objects was replaced with an orange rectangular shaped object. The time spent by the mice exploring each object was recorded by video capture and analyzed. The results are presented as the discrimination index which is calculated by: (time exploring the novel object—time exploring the familiar object)/(time exploring the novel object + time exploring the familiar object). It is common rodent behavior for a mouse to explore a novel object over a familiar one. The premise for this test is that a mouse with a cognitive deficit will not be able to remember the old object during the test phase, therefore will spend a similar amount of time exploring each object. All experiments were performed between 10 a.m. and 12 p.m.
## Morris water maze test (MWMT)
MWMT was performed to assess spatial learning and memory function [52]. A circular tank of 4 ft diameter was filled with water and the water was made opaque by the addition of non-toxic blue paint. The temperature of the water was maintained at 22 °C. Spatial cues of various shapes (round, rectangle, square, triangle) and colors (red, yellow, green, blue) were equally spaced and placed around the tank. An escape platform positioned 1 cm below the surface of the water and mice were trained to locate it. This study is of 5 days: day 1–4 are trial days and day 5 is probe test day. On trial days, each mouse had 3 training trials per day separated by 1 h. In each trial, the mouse was placed in one of the 3 start locations which were equally spaced around the perimeter of the tank. Start location was changed in each trial. The mouse was allowed to swim for 60 s or until it reached the platform. If the mouse could not locate the platform within 60 s, the mouse was placed on the platform by the experimenter for 10 s and then placed in a home cage after making them dry with gentle wiping and keeping under a heat lamp for 5 min. On day 5, a probe test was performed where in a 60 s trial the mice swim across the tank without the platform being present. This probe test measures reference memory of the mice as it would look for the platform from its previous memory and spend more time around the original platform location. Video capture and any-maze software were used to analyze data for this experiment. All experiments were performed between 12 p.m. and 4 p.m.
## Vaginal cytology
Estrous cycles were assessed at the same time every day during behavioral study performance day following published protocols [53, 54]. Female mice were properly handled to minimize stress, by gently lifting the animal by the base of tail a plastic pipette filled with about 1 ml of PBS was placed on the tip of the vagina and flushed 5 times with same PBS to allow proper collection of samples for vaginal cytology. Sample was then smeared on appropriate labeled microscope slides and after 1 h of drying time, a crystal violet ($0.1\%$) staining was performed on slides. After drying, the slides were observed under a light microscope to visualize cells. Images were obtained using NIS Elements imaging software version 4.0.
## Statistical analysis
The sample size for the animal study was estimated based on G-power analysis. Test for normality was performed to select the appropriate statistical method. All data are expressed as the mean ± SEM. The values were analyzed by ‘t’ test to compare between two groups (Prism, version 7.0; GraphPad Software Inc., San Diego, CA). P values less than 0.05 were considered statistically significant.
### offspring plasma confirms presence of nicotine and cotinine
The weight of the mother was measured at gestational day 5 (E5) before starting the e-cig exposure and after delivery. Before starting the treatment (e-cig exposure) at E5, there was no difference in weight between control and treatment group, however after delivery the e-cig exposed mother group had significant weight reduction compared to control group ($P \leq 0.001$) (Additional file 1: Fig S1).
We also counted the litter size after delivery and measured the weight of the offspring at PD0, PD7, PD15, PD30, PD45, PD60, and PD90. No difference was observed in litter size between control and exposed offspring (Fig. 2A), but significantly reduced body weight was observed in prenatally e-cig exposed both male and female offspring at PD0 ($P \leq 0.05$), PD7 ($P \leq 0.0001$), PD15 ($P \leq 0.0001$ for female and $P \leq 0.001$ for male), PD30 ($P \leq 0.0001$ for both male and female), PD45 ($P \leq 0.0001$ for female and $P \leq 0.05$ for male), PD60 ($P \leq 0.0001$ for female and $P \leq 0.001$ for male), and PD90 ($P \leq 0.0001$ for female and $P \leq 0.001$ for male) compared to control (Fig. 2C).Fig. 2Measurement of A litter size and B litter weight ($$n = 8$$) and C offspring body weight at PD7, PD15, PD30, PD45, PD60 and PD90. $$n = 20$$–40; *$P \leq 0.05$, ***$P \leq 0.001$ and ****$P \leq 0.0001$ Postnatal brain to body weight ratio was also measured at PD7, PD23, PD45 and PD90. Significantly low brain to body weight ratio was observed in e-cig exposed group at PD7 ($P \leq 0.05$) but not at other time points (Fig. 3).Fig. 3Measurement of brain to body weight ratio of offspring at PD7, PD15, PD30, PD45, PD60 and PD90. $$n = 10$$; *$P \leq 0.05$ After completing the maternal exposure at PD7, plasma was collected from randomly selected offspring ($$n = 4$$) to measure the nicotine and cotinine concentration and we observed 4.3 ± 1.001 ng/mL and 3.4 ± 0.6760 ng/mL of average nicotine and cotinine concentration in plasma respectively (Additional file 2: Fig S2). Previously, our lab also reported the presence of nicotine and cotinine in plasma and brain of offspring after maternal e-cig exposure, the level of which were comparable to those of mother [6].
## Prenatal e-cig exposure alters the expression of BBB component markers
We measured a total of 10 markers which are essential components of the BBB and neurovascular unit including tight junction proteins (ZO-1, claudin-5, occludin), astrocyte (GFAP), pericyte (PDGFR-β), basement membrane protein (laminin α1 and laminin α4), water channel protein (AQP4), glucose transporter (GLUT-1) and neuron specific marker (NeuN) by western blot at PD7, PD23, PD45 and PD90.
## Tight junction proteins
Significantly reduced expression of tight junction proteins- ZO-1 ($P \leq 0.05$), claudin-5 ($P \leq 0.05$), occludin ($P \leq 0.05$) was observed at PD7 and PD23 in prenatally e-cig exposed offspring (Fig. 4). Sex-dependent difference was observed at PD45 and PD90 where down regulation of ZO-1 ($P \leq 0.05$ for male and no difference for female) was observed in prenatally e-cig exposed male offspring, but not in female (Fig. 4). Significantly reduced expression level of occludin was observed in prenatally e-cig exposed offspring at PD45 ($P \leq 0.05$ for both male and female), but not observed at PD90 (Fig. 4). At every time point, significantly reduced expression of claudin-5 ($P \leq 0.05$), was found in both male and female prenatally e-cig exposed offspring (Fig. 4).Fig. 4Expression of tight junction proteins ZO-1, claudin-5 and occludin in prenatally e-cig exposed offspring compared to control at A PD7, B PD23, C PD45 and D PD90 by western blot; normalized to β-Actin; $$n = 6$$ for PD7 and $$n = 4$$ for PD23, PD45 and PD90; *$P \leq 0.05$, **$P \leq 0.01$
## Astrocyte, pericyte, and neuron specific marker
Downregulation of GFAP was observed in prenatally e-cig exposed offspring from PD7 ($P \leq 0.01$) to PD90 ($P \leq 0.05$) (Fig. 5). Interesting, significantly reduced expression of GFAP was only found in prenatally e-cig exposed male offspring at PD45 ($P \leq 0.05$) and PD90 ($P \leq 0.05$) (Fig. 5). No difference was observed in PDGFRβ and NeuN expression between control and prenatally e-cig exposed offspring at any time point (Fig. 5).Fig. 5Expression of astrocyte marker (GFAP), pericyte marker (PDGFRβ) and neuron specific marker (NeuN) in prenatally e-cig exposed offspring compared to control at A PD7, B PD23, C PD45 and D PD90 by western blot; normalized to β-Actin. $$n = 6$$ for PD7 and $$n = 4$$ for PD23, PD45 and PD90; *$P \leq 0.05$, **$P \leq 0.01$
## Basement membrane proteins
Laminin α1 is parenchymal and laminin α4 is endothelial basement membrane protein which are the major components of BBB. In our study, no significant difference was noticed in prenatally e-cig exposed offspring compared to control in any time point (Fig. 6).Fig. 6Expression of basement membrane proteins (laminin α1 and laminin α4) in prenatally e-cig exposed offspring compared to control at A PD7, B PD23, C PD45 and D PD90 by western blot; normalized to β-Actin. $$n = 6$$ for PD7 and $$n = 4$$ for PD23, PD45 and PD90; *$P \leq 0.05$
## Glucose transporter and water channel protein
Expression of the glucose transporter (GLUT-1) and aquaporin (AQP4) were evaluated in prenatally e-cig exposed offspring and control group at every time point (Fig. 7). Significantly decreased expression of GLUT-1 was observed only at PD7 ($P \leq 0.05$). In case of AQP4, downregulation of protein expression was observed at every time point ($P \leq 0.05$ and $P \leq 0.01$ for female offsprinf at PD23) in prenatally e-cig exposed male and female offspring except for PD45 ($P \leq 0.05$) and PD90 ($P \leq 0.05$) where decreased expression level was observed only in male (Fig. 7).Fig. 7Expression of glucose transporter (GLUT-1) and water channel protein (AQP4) in prenatally e-cig exposed offspring compared to control at A PD7, B PD23, C PD45 and D PD90 by western blot; normalized to β-Actin. $$n = 6$$ for PD7 and $$n = 4$$ for PD23, PD45 and PD90; *$P \leq 0.05$, **$P \leq 0.01$ Immunofluorescence: Our immunofluorescence results at PD90 also demonstrated that prenatally e-cig exposed offspring had reduced expression of GFAP, ZO-1 (in male) and Claudin-5 (in both male and female) (Fig. 8).Fig. 8Expression of ZO-1, Claudin-5 (cortex) and GFAP expression (hippocampus) at PD90 in control female (CF), treatment female (TF), control male (CM) and treatment male (TM), normalized to DAPI, $$n = 3$$; *$P \leq 0.05$, **$P \leq 0.01$, ****$P \leq 0.0001$
## Prenatal e-cig exposure causes hyperactivity in female adolescent and adult offspring
We measured the locomotor activity of control and prenatally e-cig exposed male and female offspring at their adolescent (PD45) and adult (PD90) time-point. At both time-points, prenatally e-cig exposed female offspring showed significantly higher locomotor activity compared to control counterpart; no significant difference was observed in prenatally e-cig exposed male offspring (Fig. 9A and B. Fecal boli was counted to measure stress/anxiety-like behavior after OFT completion and only prenatally e-cig exposed female offspring had higher fecal boli compared to control at adolescence (Fig. 9C) ($P \leq 0.01$) and no difference was observed in adulthood (Fig. 9D).Fig. 9Assessment of locomotor/hyperactivity by open filed test at A Adolescence and B Adult time point. Fecal boli was counted to evaluate stress/anxiety after OFT at C adolescence (PD 6 weeks) and D adult (PD 3 months) time point. $$n = 10$$; *$P \leq 0.05$, **$P \leq 0.01$
## Prenatal e-cig exposure deteriorates recognition memory function
NORT was conducted to evaluate short-term recognition memory of the adolescent (PD45) (Fig. 10A) and adult offspring (PD90) (Fig. 10D). Prenatal e-cig exposure significantly decreased recognition memory in both prenatally e-cig exposed male and female offspring at ($P \leq 0.05$ for both female and male) and adult time-point ($P \leq 0.001$ for female and $P \leq 0.05$ for male). The results have been shown in Fig. 10A and D.Fig. 10Assessment of learning and memory function in prenatally e-cig exposed offspring compared to control at adolescent and adult age. Novel object recognition test to evaluate short term memory function at A PD 6 weeks and D PD 3 months; Morris water maze test to assess special learning and memory function at B–C PD 6 weeks and E–F PD 3 months, $$n = 10$$; *$P \leq 0.05$, ***$P \leq 0.001$
## Prenatal e-cig exposure did not impact spatial acquisition and reference memory function
MWMT was performed to assess spatial acquisition and reference memory function in offspring at adolescent and adult age. No significant difference was observed in control and treated group (Fig. 10C and F) in reference memory however, prenatally e-cig exposed female and male group had slow learning process (learning the location of hidden platform) compared to control group on day 2–3 and day 3–4 respectively at adolescence ($P \leq 0.05$) (Fig. 10B). However, at adulthood only male offspring had impaired learning compared to control at day 3–4 ($P \leq 0.05$) (Fig. 10E).
## Reassessment of behavioral outcomes considering estrous cycle in female offspring
We collected vaginal secretions from offspring to evaluate the impact of estrous cycle on behavioral outcomes. Vaginal secretions are made up of three types of cells- leucocytes, anucleated cornified epithelial cells and nucleated epithelial cells. Estimation of the phase of estrous cycle is based on the proportion of these cells in the vaginal secretion. In proestrus stage, the proportion of nucleated epithelial cell is high whereas in estrus stage, cornified cells are abundant as you can see in Additional file 3: Fig S3. In metestrus stage, three types of cells are present and in diestrus stage, the prominent cell is the leukocyte (Additional file 3: Fig S3). In proestrus and estrus stages, the circulating estradiol level is high and in metestrus and diestrus stages, the circulating estradiol level is low. We reassessed open field test (Additional file 3: Fig S3A) and novel object recognition test (Additional file 3: Fig S3B) and observed that circulating estradiol level depending on estrous stages does not impact behavioral outcomes.
## Discussion
In recent years, e-cigs have become extremely popular among all age and sex groups, including pregnant women. The health impact of e-cig vaping is currently unclear, hence it should not be considered as a safe alternative to tobacco smoking during pregnancy as e-cig contains several toxic compounds including propylene glycol, vegetable glycerin, formaldehyde, acrolein, flavoring chemicals, and other trace elements, some of which may be neurotoxic to the developing fetus and offspring [14]. In addition to these chemicals, several studies have demonstrated that in-utero exposure of nicotine and tobacco smoke may impact neonatal brain development which can result in neuro-developmental abnormalities, including increased cell density in the hippocampus [55], reduced volumes of cortical gray matter [56], increased sensitivity to neonatal HI brain injury [8, 51], abnormalities in cell differentiation, disruption in neurotransmitter activity and neurobehavioral dysfunction [57]. Interestingly, post-birth exposure to nicotine did not show similar adverse effects suggesting differences in biological mechanisms during the gestation phase [57]. Although the neurotoxic effect of prenatal nicotine and tobacco exposure is well documented and well established, only a handful of studies have investigated the potential neurotoxic effects of prenatal e-cig exposure on neonates, mostly focusing on cognitive dysfunction [58, 59]. Previously, our lab published that maternal e-cig exposure is associated with decreased brain glucose utilization and worsened hypoxic–ischemic brain injury in offspring [6]. However, no study to date has looked at the effects of prenatal e-cig exposure on BBB integrity which is a critical determinant of cerebrovascular dysfunction in adulthood. In this study, we provided in-vivo evidence that maternal e-cig exposure disrupts some of the critical elements of BBB integrity and deteriorated motor, learning and memory function in male and female offspring at different stages of life.
In our study, we have investigated all the crucial markers for BBB integrity including tight junction proteins (ZO-1, claudin-5, occludin), astrocyte marker (GFAP), pericyte marker (PDGFRβ), neuron specific marker (NeuN), basement membrane protein (laminin α1, laminin α4), glucose transporter (GLUT-1) and water channel protein (AQP4) at different time points (PD7, PD23, PD45 and PD90) (Table 1). We have observed that prenatally, e-cig exposed offspring had reduced expression of tight junction proteins (mainly ZO-1 and claudin-5) until PD90. These results are consistent with previous literature demonstrating the downregulation of tight junction proteins mediated by nicotine and tobacco smoke [60–63]. Interestingly in our study, sex-dependent effect has been observed in TJ protein expression in adolescent and adult age. Prenatally, e-cig exposed male adolescent and adult mice had reduced expression of ZO-1 at PD45 and PD90 whereas reduced level of claudin-5 was observed in both male and female offspring at every time point. Claudin-5 is one of the important components of tight junction strand, particularly in endothelial cells of brain which selectively restricts the transport of ions and macromolecules through the tissue barrier [64–66]. Several studies have demonstrated that claudin-5 plays a pivotal role in BBB permeability and downregulation of claudin-5 may result in disrupted brain function [67, 68] by enhancing the passage of macrophages, leukocytes, endotoxins, bacteria, and drugs from the peripheral circulation into the brain [69]. Our study demonstrated decreases in claudin-5 expression in prenatally e-cig exposed male and female offspring at all time points which could be a concern in postnatal health considering its important role in BBB integrity and permeability. Decreased expression of occludin was also observed in prenatally e-cig exposed both male and female offspring in every time point except adulthood (PD90). Sex-dependent impact on BBB integrity and permeability have been reported in several studies. The iPSC-derived BMECs from pre-menopausal women had reduced permeability, and increased barrier strength, compared to iPSC-derived BMECs from men [70]. Increased mRNA expression of tight junction proteins (claudin-1, claudin-5, claudin-12, occludin, ZO-1), junction adhesion molecule A (JAMA), major facilitator superfamily domain containing 2, and brain-derived neurotrophic factor (BDNF) were also observed in female mice compared to male [71]. Considering the biological significance of sex-differences in BBB integrity and permeability, sex dependent BBB characteristics should be further investigated capitalizing on different in-vitro and in-vivo models to more precisely determine the role of sex as a biological determinant of neurovascular function in cerebrovascular diseases [72].Table 1Summary of changes in BBB integrity markers in prenatally e-cig exposed offspring at PD7, PD23, PD45 and PD90 Glial fibrillary acidic protein or GFAP is astrocyte specific marker. Astrocytes facilitate the formation of complex neocortical circuitries involving a complex process of synaptogenesis, maturation, and synaptic pruning in the developing brain [73, 74]. Studies have shown that loss of GFAP expression results in enhanced susceptibility to ischemic insult, increased hippocampal LTP, reduced cerebellar long-term depression (LTD) and decreased glutamate transport [75]. Aquaporin 4 (AQP4) is a water channel protein which is highly expressed on astrocytic endfeet in the brain and plays a pivotal role in modulating astrocytic function, regulating extravascular brain water, brain volume homeostasis, synaptic plasticity as well as producing cerebrospinal fluid. Studies have confirmed the association of AQP4 in the pathophysiology of cerebral disorders including but not limited to stroke, cerebral edema, traumatic brain injury, Parkinson's disease, epilepsy, and depression [76]. In our study, we have observed downregulation of GFAP and AQP4 expression at every time point in prenatally e-cig exposed offspring till adulthood which indicates a potential neurotoxic effect of prenatal e-cig exposure. Interestingly, downregulation of GFAP and AQP4 was only observed in male offspring at PD45 and PD90 which indicates a sex difference in GFAP and AQP4 expression. Studies have reported the association of sex hormone with astrocytes expression. Estradiol (E2) has been reported as a mediator of neuronal sprouting through its effects on astrocytes [77], and it regulates the expression of GFAP, the major intermediate filament protein of differentiated astrocytes, both in-vitro and in-vivo model in the hypothalamus and the hippocampus of the rat [78, 79]. Arias C et al. reported higher expression of GFAP in CA1, CA3, and dentate gyrus in proestrus females as compared with males and diestrus females suggesting the association of sex steroid hormones in the sexually dimorphic functions of the hippocampus, and changes in its activity during the estrous cycle. In our study, we have also observed the downregulation of GFAP in prenatally e-cig exposed male offspring, not in female in adolescent and adult time point indicating a role of sex steroid hormones in GFAP expression. AQP4 is a water channel protein which is essential in maintenance of cerebral water balance and highly expressed in astrocytic end foot at BBB [80]. In our study, we have observed downregulation of AQP4 similar to GFAP. In PD45 and PD90, significantly lower expression level of AQP4 was observed in prenatally e-cig exposed male offspring, but not in female. As AQP4 is abundantly expressed in astrocytic end foot, downregulation of GFAP may have resulted in lower expression of AQP4 in prenatally e-cig exposed male offspring in our study.
We also evaluated the expression of basement membrane proteins, laminin α1 (parenchymal) and laminin α4 (endothelial), pericyte marker (PDGFRβ) and neuron specific marker (neuN) in our study however we did not observe any significant difference between prenatally e-cig exposed offspring and control offspring in these proteins’ expression in offspring brain.
Moreover, we assessed the effect of maternal vaping on long-term locomotor, learning and memory functions of offspring at adolescent and adult stage. OFT measures the hyperactivity or locomotor activity in rodent model. Hyperactivity is one of the elements of attention deficiency hyperactivity disorder (ADHD) in children [81]. Several studies have demonstrated the association between maternal smoking and ADHD in offspring [82–84]. In our study we have observed that prenatally e-cig exposed female offspring had higher locomotor activity than control group at adolescence and adulthood. However, no difference has been observed in prenatally e-cig exposed and control male group. This result is consistent with previous finding where higher locomotor activity was observed in female rats compared to male rats at low dose of nicotine injection [85]. Another study demonstrated the anxiogenic response in female mice treated with chronic nicotine and no effect was observed in males [86]. A recent study also reported that female offspring with maternal nicotine exposure demonstrated an increase in anxiety-like behavior in open-field test which is consistent with our findings [87]. These studies clearly indicate the potential role of sex in differential hyperactivity between males and females. Our study also demonstrated a sex-difference in hyperactivity followed by prenatal e-cig exposure in male and female offspring at PD45 and PD90. We also measured defecation or fecal boli counting after OFT and prenatally e-cig exposed female offspring had significantly higher fecal boli at adolescence demonstrating higher stress/anxiety-like behavior compared to control group.
We evaluated short-term memory function by NORT, where decreased recognition memory was observed in prenatally e-cig exposed both male and female offspring at adolescence and adulthood. MWMT was performed to evaluate spatial learning and memory function in offspring. No significant difference has been observed in reference memory function in prenatally exposed offspring at adolescence and adult time points. However, prenatally e-cig exposed male and female offspring at PD45 and only male offspring at PD90 had worsened spatial acquisition measured by escape latency. These results indicate that prenatal e-cig exposure worsens learning and memory function in adolescent and adult offspring. These results are consistent with some of the previously published literature investing exposure to tobacco smoke. Impaired cognitive development and lower intelligence quotient were observed in prenatally tobacco smoke exposed offspring [88, 89]. A recent literature review has demonstrated that children prenatally exposed to smoking are more likely to require support resulting from the well-documented physical, socio-emotional, behavioral, mental, and neurocognitive consequences of exposure [90]. Interestingly, maternal e-cig exposure has been found to be associated with short-term memory deficits, reduced anxiety, and hyperactivity in adult offspring using novel object recognition and elevated plus maze tests [58]. A recent study from our lab also demonstrated that maternal e-cig exposure resulted in cognitive deficits in e-cig exposed offspring with hypoxic-ischemic (HI) brain injury compared to sham offspring group [6].
Sex differences have been observed in pharmacokinetics and pharmacodynamics of nicotine in several studies [72]. Decreased nicotine clearance and lower ratio of nicotine/cotinine was reported in women compared to men due to faster metabolism of nicotine [6, 91]. Previous study has also shown differential responses of cerebral cortex to nicotine in male and female [92]. Similarly, our study has also found sex-specific effect of nicotine containing e-cig on BBB and behavioral outcomes and these results might be explained by sex differences in nicotine metabolism.
To evaluate the role of estradiol in behavioral outcomes, we re-assessed the data based on estrous cycle. Estrous cycle in mice consists of 4 stages: Proestrus, estrus, metestrus and diestrus. In proestrus and estrus days, circulating estradiol level is high and in metestrus and diestrus days, the circulating estradiol level is low [53]. Studies have demonstrated a sex-dependent effect of nicotine in locomotor activity. Higher locomotor activity was observed in female rats compared to male rats at low dose of nicotine injection [93]. However, another study reported contradictory results demonstrating higher locomotor activity in adolescent males compared to females after single dose nicotine administration with minipumps [94]. Lower dose of nicotine administration was also reported to cause decreased locomotor activity in female than male, using osmotic mini pumps [95]. Even though the nicotine-exposed adolescents demonstrated contradictory results, it is apparent that females are more sensitive to the nicotine induced locomotor activity compared to male, and possibly, ovarian hormones play a role in this greater responsivity. Interestingly, in our study we did not find the role of estradiol level in behavioral outcomes (OFT and NORT). Therefore, future experimental design should focus on acute vs. chronic nicotine dosing through vaping when interpreting effects on locomotor activity.
## Conclusion
Tobacco smoking during pregnancy is well documented as one of the crucial preventable causes of adverse birth outcomes across the world while the widespread use of recently introduced e-cig products among pregnant women poses additional challenges due to limited studies. E-cig is considered as a safe alternative to conventional cigarette and has become a pressing issue regarding its safety. The safety issue is even more salient during pregnancy as maternal e-cig use can be a threat to postnatal health due to presence of nicotine and several known and unknown additives in e-cig liquid. Considering the prevalence of e-cig use during pregnancy and associated health risks, the aim of our study was to evaluate the long-term impact of maternal e-cig use on postnatal BBB integrity and behavioral outcomes in in-vivo model. In this study, we have shown that maternal e-cig exposure can decrease the expression of tight junction proteins, astrocyte marker and AQP4 at PD7, PD23, PD45 and PD90. We also observed sex-different effect in case of ZO-1, GFAP and AQP4 expression at PD45 and PD90. However, no difference was observed in case of PDGFRβ, laminin (α1, α4) and NeuN expression at any time point. Consistent reduction of tight junction protein, claudin-5 till PD90 in both prenatally e-cig exposed male and female offspring is a serious health concern as it could be associated with long term neurotoxicity and cerebrovascular diseases. In our study we have also observed low birth weight and significantly reduced body weight in prenatally e-cig exposed male and female offspring till PD90. Considering that low birth weight is associated with neurodevelopmental deficits, the combined effect of lower body weight and reduced BBB integrity could indicate the severity of prenatal e-cig toxicity on postnatal health including both sexes. Moreover, this study has shown that maternal e-cig exposure deteriorates motor, learning and memory function in adolescent and adulthood. As disruption of key components of BBB integrity can result in worsened cerebral and neurological dysfunction including ischemic stroke, further studies are warranted to evaluate the impact of maternal vaping on the pathogenesis of different cerebrovascular diseases in adult and aged offspring.
## Supplementary Information
Additional file 1: Figure S1. Measurement of mother weight at (A) E5 (before exposure) and (B) post-delivery (after exposure), $$n = 8$.$ *** $P \leq 0.001$Additional file 2: Figure S2. Measurement of plasma nicotine and cotinine concentration in prenatally e-cig exposed offspring at PD7; plasma nicotine and cotinine concentration are 4.3 ng/mL and 3.4 ng/mL respectively, $$n = 4$.$Additional file 3: Figure S3. Re-assessment of behavioral outcomes in prenatally e-cig exposed offspring compared to control at PD90 considering estrous cycle. ( A) Open field test to measure locomotor/hyperactivity; (B) Novel object recognition test to evaluate short term memory function. $$n = 4$$-6; *$P \leq 0.05$
## References
1. Hofhuis W, de Jongste JC, Merkus PJ. **Adverse health effects of prenatal and postnatal tobacco smoke exposure on children**. *Arch Dis Child* (2003.0) **88** 1086-1090. PMID: 14670776
2. Rauschert S. **Maternal smoking during pregnancy induces persistent epigenetic changes into adolescence, independent of postnatal smoke exposure and is associated with cardiometabolic risk**. *Front Genet* (2019.0) **10** 770. PMID: 31616461
3. Lange S. **National, regional, and global prevalence of smoking during pregnancy in the general population: a systematic review and meta-analysis**. *Lancet Glob Health* (2018.0) **6** e769-e776. PMID: 29859815
4. 4.U.S. Department of Health and Human Services, C.f.D.C.a.P., National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health. The health consequences of smoking: 50 years of progress. A Report of the Surgeon General. 2014.
5. Gould GS. **Exposure to tobacco, environmental tobacco smoke and nicotine in pregnancy: a pragmatic overview of reviews of maternal and child outcomes, effectiveness of interventions and barriers and facilitators to quitting**. *Int J Environ Res Public Health* (2020.0) **17** 6
6. Sifat AE. **Prenatal electronic cigarette exposure decreases brain glucose utilization and worsens outcome in offspring hypoxic–ischemic brain injury**. *J Neurochem* (2020.0) **153** 63-79. PMID: 31883376
7. Alkam T. **Evaluation of cognitive behaviors in young offspring of C57BL/6J mice after gestational nicotine exposure during different time-windows**. *Psychopharmacology* (2013.0) **230** 451-463. PMID: 23793357
8. Li Y. **Perinatal nicotine exposure increases vulnerability of hypoxic-ischemic brain injury in neonatal rats: role of angiotensin II receptors**. *Stroke* (2012.0) **43** 2483-2490. PMID: 22738920
9. Lacy RT. **Intravenous prenatal nicotine exposure alters METH-induced hyperactivity, conditioned hyperactivity, and BDNF in adult rat offspring**. *Dev Neurosci* (2016.0) **38** 171-185. PMID: 27287203
10. Lee H, Chung S, Noh J. **Maternal nicotine exposure during late gestation and lactation increases anxiety-like and impulsive decision-making behavior in adolescent offspring of rat**. *Toxicol Res* (2016.0) **32** 275-280. PMID: 27818729
11. Newman MB, Shytle RD, Sanberg PR. **Locomotor behavioral effects of prenatal and postnatal nicotine exposure in rat offspring**. *Behav Pharmacol* (1999.0) **10** 699-706. PMID: 10780512
12. Pinheiro CR. **Maternal nicotine exposure during lactation alters food preference, anxiety-like behavior and the brain dopaminergic reward system in the adult rat offspring**. *Physiol Behav* (2015.0) **149** 131-141. PMID: 26048299
13. Zhang L. **Attention and working memory deficits in a perinatal nicotine exposure mouse model**. *PLoS ONE* (2018.0) **13** e0198064. PMID: 29795664
14. Ruszkiewicz JA. **Neurotoxicity of e-cigarettes**. *Food Chem Toxicol* (2020.0) **138** 111245. PMID: 32145355
15. Baeza-Loya S. **Perceptions about e-cigarette safety may lead to e-smoking during pregnancy**. *Bull Menninger Clin* (2014.0) **78** 243-252. PMID: 25247743
16. McCubbin A. **Perceptions and use of electronic cigarettes in pregnancy**. *Health Educ Res* (2017.0) **32** 22-32. PMID: 28158490
17. Whittington JR. **The use of electronic cigarettes in pregnancy: a review of the literature**. *Obstet Gynecol Surv* (2018.0) **73** 544-549. PMID: 30265741
18. Liu B. **Prevalence and distribution of electronic cigarette use before and during pregnancy among women in 38 states of the United States**. *Nicotine Tob Res* (2021.0) **23** 1459-1467. PMID: 33693836
19. Obisesan OH. **E-cigarette use patterns and high-risk behaviors in pregnancy: behavioral risk factor surveillance system, 2016–2018**. *Am J Prev Med* (2020.0) **59** 187-195. PMID: 32362509
20. Kaisar MA. **A decade of e-cigarettes: limited research & unresolved safety concerns**. *Toxicology* (2016.0) **365** 67-75. PMID: 27477296
21. Zhang Y. **Maternal electronic cigarette exposure in relation to offspring development: a comprehensive review**. *Am J Obstet Gynecol MFM* (2022.0) **4** 100659. PMID: 35568317
22. Chen H. **Maternal e-cigarette exposure in mice alters dna methylation and lung cytokine expression in offspring**. *Am J Respir Cell Mol Biol* (2018.0) **58** 366-377. PMID: 28960086
23. Chen H. **Modulation of neural regulators of energy homeostasis, and of inflammation, in the pups of mice exposed to e-cigarettes**. *Neurosci Lett* (2018.0) **684** 61-66. PMID: 29981356
24. Dwyer JB, McQuown SC, Leslie FM. **The dynamic effects of nicotine on the developing brain**. *Pharmacol Ther* (2009.0) **122** 125-139. PMID: 19268688
25. Chen Z. **Single-nucleus chromatin accessibility and RNA sequencing reveal impaired brain development in prenatally e-cigarette exposed neonatal rats**. *iScience* (2022.0) **25** 104686. PMID: 35874099
26. Orzabal MR. **Chronic exposure to e-cig aerosols during early development causes vascular dysfunction and offspring growth deficits**. *Transl Res* (2019.0) **207** 70-82. PMID: 30653941
27. Lauterstein DE. **Frontal cortex transcriptome analysis of mice exposed to electronic cigarettes during early life stages**. *Int J Environ Res Public Health* (2016.0) **13** 4
28. Zelikoff JT. **Microglia activation and gene expression alteration of neurotrophins in the hippocampus following early-life exposure to e-cigarette aerosols in a murine model**. *Toxicol Sci* (2018.0) **162** 276-286. PMID: 29161446
29. Wickström R. **Effects of nicotine during pregnancy: human and experimental evidence**. *Curr Neuropharmacol* (2007.0) **5** 213-222. PMID: 19305804
30. Tahajjodi SS. **The effect of maternal nicotine on basement membrane collagen IV of brain microvessels changes in neonatal Balb/C mice**. *Iran J Reprod Med* (2014.0) **12** 275-280. PMID: 24976823
31. Archie SR, Al Shoyaib A, Cucullo L. **Blood-brain barrier dysfunction in CNS disorders and putative therapeutic targets: an overview**. *Pharmaceutics* (2021.0) **13** 1779. PMID: 34834200
32. Abbruscato TJ. **Nicotine and cotinine modulate cerebral microvascular permeability and protein expression of ZO-1 through nicotinic acetylcholine receptors expressed on brain endothelial cells**. *J Pharm Sci* (2002.0) **91** 2525-2538. PMID: 12434396
33. Kadry H. **Comparative assessment of in vitro BBB tight junction integrity following exposure to cigarette smoke and e-cigarette vapor: a quantitative evaluation of the protective effects of metformin using small-molecular-weight paracellular markers**. *Fluids and Barriers of the CNS* (2021.0) **18** 28. PMID: 34158083
34. Kaisar MA. **Offsetting the impact of smoking and e-cigarette vaping on the cerebrovascular system and stroke injury: is metformin a viable countermeasure?**. *Redox Biol* (2017.0) **13** 353-362. PMID: 28646795
35. Abbruscato TJ. **Regulation of blood-brain barrier Na, K,2Cl-cotransporter through phosphorylation during in vitro stroke conditions and nicotine exposure**. *J Pharmacol Exp Ther* (2004.0) **310** 459-468. PMID: 15051802
36. Paulson JR. **Tobacco smoke chemicals attenuate brain-to-blood potassium transport mediated by the Na, K,2Cl-cotransporter during hypoxia-reoxygenation**. *J Pharmacol Exp Ther* (2006.0) **316** 248-254. PMID: 16174793
37. Sifat AE. **Nicotine and electronic cigarette (E-Cig) exposure decreases brain glucose utilization in ischemic stroke**. *J Neurochem* (2018.0) **147** 204-221. PMID: 30062776
38. Li H. **Exposure to cigarette smoke augments post-ischemic brain injury and inflammation via mobilization of neutrophils and monocytes**. *Front Immunol* (2019.0). DOI: 10.3389/fimmu.2019.02576
39. Sifat AE. **Short-term exposure to JUUL electronic cigarettes can worsen ischemic stroke outcome**. *Fluids Barriers CNS* (2022.0) **19** 74. PMID: 36085043
40. Chan YL. **Impact of maternal cigarette smoke exposure on brain inflammation and oxidative stress in male mice offspring**. *Sci Rep* (2016.0) **6** 25881. PMID: 27169932
41. Primo CC. **Effects of maternal nicotine on breastfeeding infants**. *Rev Paul Pediatr* (2013.0) **31** 392-397. PMID: 24142324
42. McGrath-Morrow SA. **The effects of electronic cigarette emissions on systemic cotinine levels, weight and postnatal lung growth in neonatal mice**. *PLoS ONE* (2015.0) **10** e0118344. PMID: 25706869
43. Shi H. **The effect of electronic-cigarette vaping on cardiac function and angiogenesis in mice**. *Sci Rep* (2019.0) **9** 4085. PMID: 30858470
44. Kaisar MA. **A convenient UHPLC-MS/MS method for routine monitoring of plasma and brain levels of nicotine and cotinine as a tool to validate newly developed preclinical smoking model in mouse**. *BMC Neurosci* (2017.0) **18** 71. PMID: 29020944
45. Zhang Y. **Role of P53-senescence induction in suppression of LNCaP prostate cancer growth by cardiotonic compound bufalin**. *Mol Cancer Ther* (2018.0) **17** 2341-2352. PMID: 30166403
46. Zhang Y. **Potential role of astrocyte angiotensin converting enzyme 2 in the neural transmission of COVID-19 and a neuroinflammatory state induced by smoking and vaping**. *Fluids Barriers CNS* (2022.0) **19** 46. PMID: 35672716
47. Sifat AE. **Prenatal electronic cigarette exposure decreases brain glucose utilization and worsens outcome in offspring hypoxic-ischemic brain injury**. *J Neurochem* (2020.0) **153** 63-79. PMID: 31883376
48. Villalba H. **Potential role of myo-inositol to improve ischemic stroke outcome in diabetic mouse**. *Brain Res* (2018.0) **1699** 166-176. PMID: 30165043
49. Swiergiel AH, Dunn AJ. **Feeding, exploratory, anxiety- and depression-related behaviors are not altered in interleukin-6-deficient male mice**. *Behav Brain Res* (2006.0) **171** 94-108. PMID: 16677721
50. Ren J. **Anxiety-related mechanisms of respiratory dysfunction in a mouse model of rett syndrome**. *J Neurosci* (2012.0) **32** 17230. PMID: 23197715
51. Chan YL. **Maternal cigarette smoke exposure worsens neurological outcomes in adolescent offspring with hypoxic-ischemic injury**. *Front Mol Neurosci* (2017.0) **10** 306. PMID: 29018327
52. Vorhees CV, Williams MT. **Morris water maze: procedures for assessing spatial and related forms of learning and memory**. *Nat Protoc* (2006.0) **1** 848-858. PMID: 17406317
53. Byers SL. **Mouse estrous cycle identification tool and images**. *PLoS ONE* (2012.0) **7** e35538. PMID: 22514749
54. Cora MC, Kooistra L, Travlos G. **Vaginal cytology of the laboratory rat and mouse: review and criteria for the staging of the estrous cycle using stained vaginal smears**. *Toxicol Pathol* (2015.0) **43** 776-793. PMID: 25739587
55. Roy TS, Seidler FJ, Slotkin TA. **Prenatal nicotine exposure evokes alterations of cell structure in hippocampus and somatosensory cortex**. *J Pharmacol Exp Ther* (2002.0) **300** 124-133. PMID: 11752107
56. El Marroun H. **Prenatal tobacco exposure and brain morphology: a prospective study in young children**. *Neuropsychopharmacology* (2014.0) **39** 792-800. PMID: 24096296
57. Wakschlag LS. **Maternal smoking during pregnancy and severe antisocial behavior in offspring: a review**. *Am J Public Health* (2002.0) **92** 966-974. PMID: 12036791
58. Nguyen T. **Maternal e-cigarette exposure results in cognitive and epigenetic alterations in offspring in a mouse model**. *Chem Res Toxicol* (2018.0) **31** 601-611. PMID: 29863869
59. Smith D. **Adult behavior in male mice exposed to e-cigarette nicotine vapors during late prenatal and early postnatal life**. *PLoS ONE* (2015.0) **10** e0137953. PMID: 26372012
60. Hawkins BT. **Nicotine increases in vivo blood-brain barrier permeability and alters cerebral microvascular tight junction protein distribution**. *Brain Res* (2004.0) **1027** 48-58. PMID: 15494156
61. Kadry H. **Comparative assessment of in vitro BBB tight junction integrity following exposure to cigarette smoke and e-cigarette vapor: a quantitative evaluation of the protective effects of metformin using small-molecular-weight paracellular markers**. *Fluids Barriers CNS* (2021.0) **18** 28. PMID: 34158083
62. Sharma S. **Effects of nicotine exposure from tobacco products and electronic cigarettes on the pathogenesis of neurological diseases: impact on CNS drug delivery**. *Front Drug Deliv* (2022.0). DOI: 10.3389/fddev.2022.886099
63. Prasad S. **Impact of cigarette smoke extract and hyperglycemic conditions on blood–brain barrier endothelial cells**. *Fluids Barriers CNS* (2015.0) **12** 18. PMID: 26206552
64. Jia W. **The role of claudin-5 in blood-brain barrier (BBB) and brain metastases (review)**. *Mol Med Rep* (2014.0) **9** 779-785. PMID: 24366267
65. Liu C, Wu J, Zou MH. **Activation of AMP-activated protein kinase alleviates high-glucose-induced dysfunction of brain microvascular endothelial cell tight-junction dynamics**. *Free Radic Biol Med* (2012.0) **53** 1213-1221. PMID: 22796592
66. Balbuena P, Li W, Ehrich M. **Assessments of tight junction proteins occludin, claudin 5 and scaffold proteins ZO1 and ZO2 in endothelial cells of the rat blood–brain barrier: cellular responses to neurotoxicants malathion and lead acetate**. *Neurotoxicology* (2011.0) **32** 58-67. PMID: 20970449
67. Zlokovic BV. **The blood-brain barrier in health and chronic neurodegenerative disorders**. *Neuron* (2008.0) **57** 178-201. PMID: 18215617
68. Hartz AM, Bauer B. **Regulation of ABC transporters at the blood-brain barrier: new targets for CNS therapy**. *Mol Interv* (2010.0) **10** 293-304. PMID: 21045243
69. Zhang Q. **Increase in blood-brain barrier (BBB) permeability is regulated by MMP3 via the ERK signaling pathway**. *Oxid Med Cell Longev* (2021.0) **2021** 6655122. PMID: 33859779
70. Lippmann ES. **Derivation of blood-brain barrier endothelial cells from human pluripotent stem cells**. *Nat Biotechnol* (2012.0) **30** 783-791. PMID: 22729031
71. Kamm DR. *The Effects of diet and sex differences on cortical tight-junction protein expression in senescence-accelerated mouse-prone 8 (SAMP8) mice* (2019.0)
72. Archie SR. **Biological determinants impact the neurovascular toxicity of nicotine and tobacco smoke: a pharmacokinetic and pharmacodynamics perspective**. *Neurotoxicology* (2022.0) **89** 140-160. PMID: 35150755
73. Bosworth AP, Allen NJ. **The diverse actions of astrocytes during synaptic development**. *Curr Opin Neurobiol* (2017.0) **47** 38-43. PMID: 28938161
74. Kumar P. **Mechanisms involved in epigenetic down-regulation of Gfap under maternal hypothyroidism**. *Biochem Biophys Res Commun* (2018.0) **502** 375-381. PMID: 29852171
75. Hughes EG. **Loss of glial fibrillary acidic protein results in decreased glutamate transport and inhibition of PKA-induced EAAT2 cell surface trafficking**. *Brain Res Mol Brain Res* (2004.0) **124** 114-123. PMID: 15135219
76. Li Y-K. **Aquaporin-4 deficiency impairs synaptic plasticity and associative fear memory in the lateral amygdala: involvement of downregulation of glutamate transporter-1 expression**. *Neuropsychopharmacology* (2012.0) **37** 1867-1878. PMID: 22473056
77. Rozovsky I. **Estradiol (E2) enhances neurite outgrowth by repressing glial fibrillary acidic protein expression and reorganizing laminin**. *Endocrinology* (2002.0) **143** 636-646. PMID: 11796520
78. Day JR. **Gonadal steroids regulate the expression of glial fibrillary acidic protein in the adult male rat hippocampus**. *Neuroscience* (1993.0) **55** 435-443. PMID: 8377935
79. Stone DJ. **Bidirectional transcription regulation of glial fibrillary acidic protein by estradiol in vivo and in vitro***. *Endocrinology* (1998.0) **139** 3202-3209. PMID: 9645694
80. Mader S, Brimberg L. **Aquaporin-4 water channel in the brain and its implication for health and disease**. *Cells* (2019.0) **8** 2
81. Curatolo P, D'Agati E, Moavero R. **The neurobiological basis of ADHD**. *Ital J Pediatr* (2010.0) **36** 79. PMID: 21176172
82. Huang L. **Maternal smoking and attention-deficit/hyperactivity disorder in offspring: a meta-analysis**. *Pediatrics* (2018.0) **141** 1
83. He Y. **Maternal smoking during pregnancy and ADHD: results from a systematic review and meta-analysis of prospective cohort studies**. *J Atten Disord* (2020.0) **24** 1637-1647. PMID: 29039728
84. Gustavson K. **Smoking in pregnancy and child ADHD**. *Pediatrics* (2017.0) **139** 2
85. Elliott BM. **Effects of nicotine on elevated plus maze and locomotor activity in male and female adolescent and adult rats**. *Pharmacol Biochem Behav* (2004.0) **77** 21-28. PMID: 14724038
86. Caldarone BJ, King SL, Picciotto MR. **Sex differences in anxiety-like behavior and locomotor activity following chronic nicotine exposure in mice**. *Neurosci Lett* (2008.0) **439** 187-191. PMID: 18524488
87. Liu F. **Maternal nicotine exposure during gestation and lactation period affects behavior and hippocampal neurogenesis in mouse offspring**. *Front Pharmacol* (2020.0). DOI: 10.3389/fphar.2019.01569
88. Batty GD, Der G, Deary IJ. **Effect of maternal smoking during pregnancy on offspring’s cognitive ability: empirical evidence for complete confounding in the US national longitudinal survey of youth**. *Pediatrics* (2006.0) **118** 943-950. PMID: 16950984
89. Julvez J. **Maternal smoking habits and cognitive development of children at age 4 years in a population-based birth cohort**. *Int J Epidemiol* (2007.0) **36** 825-832. PMID: 17550944
90. Micalizzi L, Knopik VS. **Maternal smoking during pregnancy and offspring executive function: What do we know and what are the next steps?**. *Dev Psychopathol* (2018.0) **30** 1333-1354. PMID: 29144227
91. Benowitz NL, Hukkanen J, Jacob P. **Nicotine chemistry, metabolism, kinetics and biomarkers**. *Handb Exp Pharmacol* (2009.0) **192** 29-60
92. Hoegberg BG. **Regulation of α4β2α5 nicotinic acetylcholinergic receptors in rat cerebral cortex in early and late adolescence: Sex differences in response to chronic nicotine**. *Neuropharmacology* (2015.0) **99** 347-355. PMID: 26272110
93. 93.Brenda M, Elliott Martha M, Faraday Jennifer M, Phillips Neil E, Grunberg. Effects of nicotine on elevated plus maze and locomotor activity in male and female adolescent and adult rats Pharmacology Biochemistry and Behavior. 2004;77(1):21–28. 10.1016/j.pbb.2003.09.016.. DOI: 10.1016/j.pbb.2003.09.016
94. 94.Martha M, Faraday Brenda M, Elliott Neil E, Grunberg. Adult vs. adolescent rats differ in biobehavioral responses to chronic nicotine administration Pharmacology Biochemistry and Behavior. 2001;70(4):475–489. 10.1016/S0091-3057(01)00642-6.. DOI: 10.1016/S0091-3057(01)00642-6
95. 95.Trauth JA, Seidler FJ, Slotkin TA. Persistent and delayed behavioral changes after nicotine treatment in adolescent rats. Brain Res. 2000;880(1-2):167–72. 10.1016/s0006-8993(00)02823-7.. DOI: 10.1016/s0006-8993(00)02823-7
|
---
title: 'Aberrant lncRNA expression in patients with proliferative diabetic retinopathy:
preliminary results from a single-center observational study'
authors:
- Lan Zeng
- Minwen Zhou
- Xiaocong Wang
- Xiaofeng Long
- Meng Ye
- Yuan Yuan
- Wei Tan
journal: BMC Ophthalmology
year: 2023
pmcid: PMC9999565
doi: 10.1186/s12886-023-02817-4
license: CC BY 4.0
---
# Aberrant lncRNA expression in patients with proliferative diabetic retinopathy: preliminary results from a single-center observational study
## Abstract
### Background
Diabetic retinopathy (DR) is a leading cause of blindness. Vision threat is particularly severe in patients with retinal neovascularization. However, little is known about the role of long noncoding RNAs (lncRNAs) in proliferative diabetic retinopathy (PDR). The goal of this study was to identify lncRNAs involved in PDR.
### Methods
We compared lncRNA expression profiles in the vitreous between patients with PDR and those with idiopathic macular hole (IMH) and between patients with PDR who had received anti-vascular endothelial growth factor (VEGF) therapy and those who had not. Vitreous samples from patients with PDR and IMH were screened for lncRNAs using microarray-based analysis, and quantitative real-time polymerase chain reaction (qRT-PCR) was used to confirm the microarray results. Bioinformatic analysis was also performed. Moreover, the effect of anti-VEGF therapy was investigated in vitreous samples of patients with PDR treated with anti-VEGF therapy and those who were not.
### Results
A total of 1067 differentially expressed noncoding RNA transcripts were found during screening in the vitreous humor of patients with PDR than in those with IMH. Five lncRNAs were subjected to qRT-PCR. RP11-573 J24.1, RP11-787B4.2, RP11-654G14.1, RP11-2A4.3, and RP11-502I4.3 were significantly downregulated; this was validated by the comparison using the microarray data. In addition, 835 differentially expressed noncoding RNA transcripts were found during screening in the vitreous humor of patients with PDR treated with anti-VEGF therapy compared with untreated PDR patients. RP4-631H13.2 was significantly upregulated, which is consistent with the trend of the microarray analysis.
### Conclusions
There were systemic expression differences in the vitreous at the microarray level between patients with PDR and those with IMH and between patients with PDR after anti-VEGF treatment and those that did not receive anti-VEGF treatment. LncRNAs identified in the vitreous humor may be a novel research field for PDR.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12886-023-02817-4.
## Background
Diabetic retinopathy (DR) is the most common complication of diabetes mellitus and the leading cause of blindness in middle-aged and elderly individuals [1, 2]. It is classified as non-proliferative diabetic retinopathy (PDR) and PDR [3]. Retinal neovascularization is a key feature of PDR [4]. The main therapeutic strategy is the use of anti-vascular endothelial growth factor (VEGF) [5]. However, resistance to neovascularization remains challenging owing to its ineffectiveness.
Long noncoding RNAs (lncRNAs) are non-protein-coding transcripts larger than 200 nucleotides. They can participate in gene regulation at transcriptional, post-transcriptional, and translational, such as regulatory transcription factors and endogenous competitive RNA [6]. LncRNAs have been implicated in a wide range of physiological processes and in the pathophysiology of several diseases, and DR is no exception. They are increasingly being recognized as important players in the development of DR.
Aberrant Expression of lncRNAs in PDR may be relevant to the molecular etiology of DR. Some researchers performed microarray-based gene expression analysis designed to serve as a resource for elucidating lncRNA-mediated DR pathogenesis [7, 8]. The identification of dysregulated lncRNAs is a key step in understanding the significance of lncRNAs in DR. The vitreous humor of patients with PDR can represent a reservoir of pathological signaling molecules because of tissue accessibility [9]. The lncRNAs that are differentially expressed in the vitreous humor of patients with PDR remain inadequately explored [9]. Therefore, this is a novel research field for PDR.
In addition, it is necessary to identify lncRNA expression profiles in the vitreous fluid of patients with PDR after anti-VEGF therapy. Adjuvant intravitreal injection of anti-VEGF drugs before vitrectomy, which reduces the difficulty of surgery and recurrence of vitreous hemorrhage, is beneficial [10]. Nevertheless, in some cases, the patient’s eyes fail to respond adequately. Thus, a complete understanding of the relationship between lncRNAs and anti-VEGF drugs may enable add another dimension to its therapeutic directions [11].
In this study, we aimed to detect differences in the expression of lncRNAs and messenger RNAs (mRNAs) between patients with PDR and those with idiopathic macular hole (IMH) and between patients with PDR treated with anti-VEGF therapy and untreated patients with PDR. To reveal the functional significance of lncRNAs in DR, we performed bioinformatic analysis.
## Patient recruitment
This clinical study adhered to the provisions of the Declaration of Helsinki for research involving human subjects. This study was approved by the Ethical Review Committee of the First People’s Hospital of Zunyi (Zunyi, China; project number 2019–019). All patients who underwent pars plana vitrectomy surgery for IMH or PDR at the First People’s Hospital of Zunyi between October 2019 and September 2020 were enrolled consecutively.
The patients were divided into three groups: group A consisted of patients with IMH without diabetes, group B consisted of patients with PDR pretreated with conbercept 3–7 days before surgery, and group C consisted of patients with PDR who underwent surgery alone. All subjects underwent a complete ophthalmologic examination, including medical history, best-corrected visual acuity measurement, intraocular pressure measurement, slit-lamp examination, fundus examination, ocular ultrasonography, optical coherence tomography, and fundus fluorescein angiography (as necessary). According to the international classification, PDR is defined as neovascularization and/or vitreous/preretinal hemorrhage [3].
Subjects with other systemic diseases such as renal failure or malignant tumors were excluded. Subjects with other eye diseases, including glaucoma, uveitis, age-related macular degeneration, retinal artery occlusion, retinal vein occlusion, rhegmatogenous retinal detachment, endophthalmitis, or ocular trauma, were also excluded. Patients who had undergone one or more of the following: eye surgery, retinal laser photocoagulation, intravitreal steroids, or anti-VEGF therapy (only 3–7 days before surgery were allowed) were excluded from the analysis. Also, subjects were excluded if they showed any evidence of systemic or local inflammation within 6 months.
Finally, three patient samples from each group were analyzed using microarray technology. The remaining 39 patients (6 in group A, 8 in group B, and 25 in group C) were included in the confirmation cohort. There were no statistically significant differences in the age or sex ratios (Table 1).Table 1Clinical characteristics of the enrolled patientsCohortAge (y), mean ± SDFemale, n (%)hemoglobin A1c (%), mean ± SDScreening Cohort Group A ($$n = 3$$)55.67 ± 12.342 (66.67)4.90 ± 0.40 Group B ($$n = 3$$)57.67 ± 3.062 (66.67)10.43 ± 2.49 Group C ($$n = 3$$)58.67 ± 8.022 (66.67)9.17 ± 2.20 P0.911.000.02PDR Confirmation Cohort Group C ($$n = 11$$)56.55 ± 9.867 (63.64)8.26 ± 1.30 Croup A ($$n = 6$$)57.33 ± 10.505 (83.33)4.90 ± 0.45 P0.880.60< 0.01Anti-VEGF Confirmation Cohort Group B ($$n = 8$$)49.00 ± 19.545 (62.50)10.26 ± 4.04 Croup C ($$n = 14$$)58.07 ± 8.727 (50.00)8.69 ± 2.34 P0.150.680.34P < 0.05 was considered to be statistically significantGroup A consisted of patients with IMH; Group B consisted of patients with PDR pretreated with conbercept 3–7 days before surgery; Group C consisted of patients with PDR who underwent surgery alonePDR Proliferative diabetic retinopathy, IMH Idiopathic macular hole
## Sample processing
A blood sample (approximately 6 mL) was drawn from the forearm vein into a tube containing an anticoagulant following overnight fasting on the day of surgery. The mixture was immediately centrifuged at 4000×g for 10 min at 4 °C. A vitreous sample (approximately 1 mL) was carefully collected into a 2 mL sterile syringe using a 25-gauge vitreous cutter and manual suction before opening the intraocular irrigation system. If vitreous hemorrhage was present, the surgeon avoided collecting blood components as much blood as possible. All samples were stored in cryopreservation tubes and immediately cooled at − 80 °C until analysis.
After sample collection, total RNA was extracted using TRIzol LS reagent (Invitrogen, Carlsbad, CA, USA) combined with miRNeasy Micro Kit (Qiagen, Hilden, Germany). RNA quality and integrity were measured using Nanodrop (Thermo Fisher Scientific, Waltham, MA, USA) and Agilent 4200 TapeStation.
## Microarray analysis
The noncoding RNA and coding RNA transcriptome analysis of the vitreous humor were detected using Clariom D Pico Assay (Affymetrix, Bedford, MA, USA), which has 13,574 transcripts. After assessing RNA quality and quantity, chip analysis was performed by Gminix Biotechnology Company (Shanghai, China). Briefly, sample labeling, hybridization, and washing were performed according to the manufacturer’s protocol. The microarrays were scanned using a GeneChip Scanner 3000 7G (Affymetrix). Raw intensity CEL files generated by GeneChip™ Command Console™ were imported into Transcriptome Analysis Console 4.0.2 (TAC 4.0.2). The data were analyzed with the Robust Multi-chip Analysis algorithm using Affymetrix default analysis setting and global scaling as the normalization method. Quality control graphs were used to assess the quality of sample files. The Limma Bioconductor package (implemented in TAC 4.0.2) was used to analyze expression data based on linear models.
The final difference result was obtained according to the filter condition | fold change | ≥ 1.5 and P values < 0.05. Hierarchical clustering was performed to show distinguishable noncoding RNA and coding RNA transcript expression patterns among the samples. All the original data were uploaded to the Gene Expression Omnibus public database (https://www.ncbi.nlm.nih.gov/geo; accession number GSE191210).
## Gene ontology (GO) enrichment analysis and Kyoto encyclopedia of genes and genomes (KEGG) pathway enrichment analysis
The GO database describes our knowledge of the biological domain regarding molecular functions, cellular components, and biological processes. According to the relationship of pathways in the KEGG database, the interaction network of a significant pathway was constructed to find the core pathway that plays a key role [12–14]. Fisher’s exact test was used to select significant GO categories and KEGG pathways, and the significance threshold was defined as P values < 0.05.
## Co-expression network
Co-expression networks were constructed to determine the potential roles of noncoding RNA transcripts. To find the co-expression relationship between the differences, Pearson correlation was calculated to identify significantly correlated pairs. It was constructed using Cytoscape, according to the expression value distribution of differential RNA transcripts in different groups. The Pearson correlation value cut-off was = 0.95, with P values < 0.05.
The co-expression network comprised 59 differentially expressed noncoding RNA transcripts and 32 coding RNA transcripts (Fig. 4). The network indicated that coding RNAs could correlate with many target noncoding RNAs and vice versa. The correlation analysis strengthens this argument. Fig. 4The co-expression network of differential transcripts (Group C versus Group A) The co-expression network comprised 11 differentially expressed noncoding RNA transcripts and 10 coding RNA transcripts (Fig. S4).
## Analysis of neighbor genes of the noncoding RNA transcripts
We searched for noncoding RNA transcripts and their associated coding gene pairs, including the same strand gene with overlap, upstream gene with 10,000 bp, downstream gene with 10,000 bp, and complementary strand gene with overlap.
To investigate the possible functions of the noncoding RNA transcripts, we predicted the potential targets of noncoding RNA transcripts. Table 2 contains 39 differentially expressed noncoding RNA transcripts and their associated coding gene pairs. Table 2Analysis of neighbor genes of the noncoding RNA transcripts (Group C versus Group A)Gene SymbolFold ChangeP-valueRegulationSame Strand Gene with OverlapUp Stream Gene in 10,000 bpDown Stream Gene in 10,000 bpComplementary Strand Gene with OverlapAP000487.41.570.03upPPFIA1RP11-242O24.32.020.03upSRSF4RP5-956O18.3−2.76<0.01downGALNT2CITF22-24E5.11.760.01upPVALBAC005009.2−1.680.01downGRM3RP4-613B23.32.03<0.01upHHATLRP11-49O14.3−1.980.01downC9orf3AP000320.61.80.03upKCNE2RP4-631H13.2−1.890.01downZYG11ARP5-1125A11.41.780.02upRALYCTD-2023 N9.1−1.640.03downACTBL2CTD-2116 N20.11.920.03upADAMTS6CTD-2201E9.4−1.960.01downSEMA5ACTC-345 K18.21.710.04upMYOZ3CTC-348 L14.11.580.04upVCANRP11-654G14.1−2.16<0.01downSLC30A8SLC30A8RP11-279 L11.1−1.530.02downCSMD1KB-1083B1.1−1.660.03downSNX31RP11-173P15.3−1.570.04downMLECRP11-1018 J8.21.560.01upRERGRERGRP13-820C6.2−2.220.02downEP400RP11-310I24.11.550.01upTMTC1RP1-267 L14.32.130.04upNAA25RP11-344B23.2−1.670.03downPAX5RP11-138H8.62.430.03upLARP6LARP6RP11-133 K1.51.710.04upPLCB2RP11-538I12.3−1.730.02downADAMTS18AC004449.6−1.770.04downRNF126POLRMTRP11-327F22.5−1.780.04downCYLDRP11-30 L15.61.80.01upLYPLA1RP13-638C3.41.60.01upFOXK2RP11-227G15.21.730.03upRHBDL3C17orf75RP11-715F3.21.650.02upMETTL4METTL4METTL4RP11-640I15.1−2.170.04downAC005324.3 TRIM16RP11-973H7.32.840.03upCEP76CEP76PSMG2CTB-189B5.31.580.02upFBXO27RP11-91G21.1−1.540.02downU2SURPRP3-467 L1.6−1.850.02downVAMP3CAMTA1PER3LL09NC01-139C3.11.580.02upWDR5P < 0.05 was considered to be statistically significantGroup A consisted of patients with IMH; Group C consisted of patients with PDRIMH Idiopathic macular hole, PDR Proliferative diabetic retinopathy A total of 32 differentially expressed noncoding RNA transcripts and their associated coding gene pairs are shown in Table S2.
## Quantitative real-time polymerase chain reaction (qRT-PCR)
To further verify the gene chip results, the threshold for the differential expression of lncRNAs was set to a P value < 0.01. Differentially expressed lncRNAs were sorted according to the absolute value of the fold change. Several candidate lncRNAs with multi-variable shear or difficult primer designs were excluded. Five target genes were selected for subsequent qRT-PCR analysis. Total RNA was reverse-transcribed using a PrimeScript RT reagent kit (TaKaRa, Dalian, China), and qRT-PCR was performed using the CFX96 real-time PCR detection system (Bio-Rad, Hercules, CA, USA). Transcript levels were determined using the PCR Master Mix (Solarbio, Beijing, China). The primer pairs used are listed in Table S1. The specificity of the qRT-PCR products was estimated using a dissociation curve. qRT-PCR was performed in duplicate for each sample. The relative gene expression was calculated using the 2-ΔΔCt method. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) served as the internal control.
## Statistical analysis
All statistical analyses were performed using the SPSS software (version 18.0; SPSS Inc., Chicago, IL, USA). Using the Shapiro–Wilk normality test, the numeric variables were first tested for the normality of distributions. Comparisons between two groups were made using Student’s t-test or Mann–Whitney U test. Also, comparisons among three groups were analyzed using a one-way analysis of variance or the Kruskal–Wallis H rank sum test. Categorical variables were determined using Fisher’s exact test. Statistical significance was defined as a P value less than 0.05.
## Evaluation of RNA quality and quantity
The range of extracted RNA from the vitreous sample was 12.0–96.0 ng, and that of the plasma sample was 10.5–87.0 ng. To evaluate RNA quality, A260/A280 ratio and A260/A230 ratio were determined. The RNA samples had an A260/A280 ratio of 1.8 to 2.0, indicating the absence of contaminating proteins. Also, the RNA samples had an A260/A230 ratio > 2.0, indicating the absence of other organic compounds. We used the RNA integrity number obtained via microfluidic analysis to evaluate RNA integrity. All samples were sufficiently pure and had RNA integrity.
## Transcriptome analysis
A total of 1067 differentially expressed noncoding RNA transcripts (526 upregulated and 541 downregulated) and 514 differentially expressed coding RNA transcripts (327 upregulated and 187 downregulated) were identified in the vitreous humor of patients with PDR compared with those with IMH. Heatmaps were generated to show the differentially expressed noncoding RNA and coding RNA transcripts (Fig. 1). When anti-VEGF-treated patients with PDR were compared with those untreated, 835 differentially expressed noncoding RNA transcripts (455 upregulated and 380 downregulated) and 226 differentially expressed coding RNA transcripts (124 upregulated and 102 downregulated) were found (Fig. S1).Fig. 1Heatmaps were generated from the hierarchical cluster analysis (Group C versus Group A)
## GO enrichment analysis
GO enrichment analysis revealed 113 terms corresponding to upregulated transcripts and 148 terms corresponding to downregulated transcripts. The GOs targets with top enrichment scores [−log10(P-value)] by the upregulated transcripts were olfactory receptor activity (ontology; molecular function), symbiont-containing vacuole membrane (ontology; cellular component), and detection of chemical stimulus involved in sensory perception of smell (ontology; biological process) (Fig. 2A). The GOs targets with top enrichment scores for the downregulated transcripts were protein kinase C inhibitor activity (ontology; molecular function), spermatoproteasome complex (ontology; cellular component), and cartilage homeostasis (ontology; biological process) (Fig. 2B).Fig. 2GO Enrichment Analysis (Group C versus Group A) GO enrichment analysis showed that 136 terms corresponded to dysregulated transcripts (109 upregulated and 27 downregulated). The GOs targets with top enrichment scores for the upregulated transcripts were protein-arginine deiminase activity (ontology; molecular function), actin filament (ontology; cellular component), and ureteric bud invasion (ontology; biological process). The GOs targeted with top enrichment scores by the downregulated transcripts were olfactory receptor activity (ontology; molecular function) and detection of chemical stimulus involved in sensory perception of smell (ontology; biological process) (Fig. S2).
## KEGG pathway enrichment analysis
KEGG pathway enrichment analysis showed that one pathway corresponded to upregulated transcripts, and the enriched pathway was Olfactory transduction (Fig. 3A). Moreover, five pathways corresponded to downregulated transcripts, and the enriched pathways were Aldosterone-regulated sodium reabsorption, Alzheimer disease, Proteasome, Olfactory transduction, and Basal cell carcinoma (Fig. 3B).Fig. 3KEGG Pathway Enrichment Analysis (Group C versus Group A) KEGG pathway enrichment analysis showed that one pathway (Homologous recombination) corresponded to upregulated transcripts, and one pathway (Olfactory transduction) corresponded to downregulated transcripts (Fig. S3).
## qRT-PCR
Expression was assessed via individual qRT-PCR assays using vitreous humor and plasma samples from a confirmation cohort. The threshold for the differential expression of lncRNAs was set to a P value of < 0.01. All the dysregulated lncRNAs in microarray data analysis are listed in Tables 3 and S3. To verify the results of the microarray analysis independently, five dysregulated lncRNAs were manually selected. The levels of RP11-573 J24.1, RP11-787B4.2, RP11-654G14.1, RP11-2A4.3 (LINC01210 provided by The HUGO Gene Nomenclature Committee symbol), and RP11-502I4.3 were significantly lower in the vitreous humor of patients with PDR than in those with IMH, which was validated entirely by comparing them with microarray data (Fig. 5).Table 3Dysregulated lncRNAs in microarray data analysis (Group C versus Group A)Gene IDGene SymbolP-valueFold ChangeRegulationTC0100017689.hg.1RP5-956O18.3< 0.01−2.76downTC0X00007085.hg.1RP11-342D14.1< 0.012.59upTC2200008344.hg.1RP1-231P7P.10.01−2.37downTC1500009850.hg.1RP11-502I4.30.01−2.27downTC0300008929.hg.1RP11-2A4.3< 0.01−2.22downTC0800008635.hg.1RP11-654G14.1< 0.01−2.16downTC1000011831.hg.1RP11-381 K7.10.012.16upTC0900008542.hg.1RP11-787B4.2< 0.01−2.05downTC0300007168.hg.1RP4-613B23.3< 0.012.03upTC0800007818.hg.1RP11-573 J24.1< 0.01−2.00downTC0500006707.hg.1CTD-2201E9.40.01−1.96downTC0700011403.hg.1RP4-736H5.3< 0.01−1.92downTC2000009883.hg.1RP5-1009E24.8< 0.01−1.91downTC0100014213.hg.1RP4-631H13.20.01−1.89downTC0500009510.hg.1CTB-43E15.10.01−1.88downTC1400009841.hg.1RP11-299 L17.30.01−1.83downTC0500013160.hg.1CTD-2118P12.1< 0.01−1.82downTC2200007269.hg.1CITF22-24E5.10.011.76upTC0700011688.hg.1AC005009.20.01−1.68downTC2100007705.hg.1AP001171.10.01−1.63downTC0100018126.hg.1RP11-407H12.8< 0.01−1.57downTC1200010026.hg.1RP11-1018 J8.20.011.56upThe threshold for the differential expression of lncRNAs was set to a P value < 0.01Group A consisted of patients with IMH; Group C consisted of patients with PDRIMH Idiopathic macular hole, PDR Proliferative diabetic retinopathyFig. 5Expression levels of lncRNAs of the confirmation cohort, in vitreous and plasma (Group C versus Group A) Meanwhile, RP4-631H13.2 expression levels in the vitreous of patients with PDR treated with anti-VEGF therapy were significantly higher than those in untreated patients with PDR, which is consistent with the trend of microarray profiling. The other lncRNAs, RP11-407H12.8, CTD-2532 K18.1, RP11-116 N8.4, and RP11-370P15.2, showed no significant differences (Fig. 5). However, no statistical differences were found in the plasma levels of these lncRNAs (Fig. 5 and S5).
## Discussion
The multifactorial pathogenesis of DR is not completely understood. As a new class of modulatory molecules, the essential roles of lncRNAs in the etiology of a broad spectrum of diseases have attracted considerable attention [6, 15, 16]. PDR is usually considered to be a neovascular disorder, and several lncRNAs have been implicated in neovascularization. ANRIL regulates VEGF expression and function in DR mediated by the PRC2 complex, thereby promoting new vessel formation [17]. MALAT1 promotes high glucose-induced human retinal endothelial cells (HRECs) by upregulating endoplasmic reticulum stress [18] or via suppressing the VE-cadherin/β-catenin complex by targeting miR-125b [19]. MIR497HG is downregulated after high-glucose treatment, and it suppresses the neovascularization of HRECs by targeting the miRNA-128-3p/SIRT axis [20]. SNHG7 inhibits high-glucose-induced angiogenesis by regulating the miR-543-mediated SIRT1/VEGF pathway [21]. TDRG1 promotes neovascularization by upregulating VEGF during DR [22]. Thus, dysregulated lncRNA expression is relevant to the molecular etiology of PDR.
Some researchers have focused on the expression of lncRNAs in DR using high-throughput screening technologies. Yan et al. first reported that approximately 303 lncRNAs are differentially expressed in the retinas of diabetic rats [7]. Likewise, to demonstrate the relationship between lncRNAs and anti-VEGF drugs, Wang et al. reported that 427 lncRNAs were differentially expressed after anti-VEGF treatment [23]. However, existing data are insufficient for the vitreous in PDR [9]. The transcriptional landscape of the vitreous, a reservoir of pathological signaling molecules, is a novel research field for PDR.
We analyzed the noncoding RNA transcript expression profiles of three patient groups: IMH, PDR treated with conbercept, and PDR alone. As a control for cytokine analyses in PDR, vitreous samples with idiopathic epiretinal membrane and/or IMH have been used in many studies [24–26]. Nonetheless, it has been reported that the activation conditions of inflammation and fibrosis in eyes with idiopathic epiretinal membranes should be carefully considered as a control group [27]. This study only included patients with IMH in the control group.
We identified that 1067 noncoding RNA transcripts were aberrantly expressed in patients with PDR. We also evaluated the effects of anti-VEGF therapy on the expression of lncRNA in patients with PDR. Ultimately, 835 dysregulated noncoding RNA transcripts were identified. We followed up on the differential expression analysis with qRT-PCR tests in a separate validation cohort, including six patients in group A, eight patients in group B, and twenty-five patients in group C. Most transcriptomes by qRT-PCR ($\frac{6}{10}$) were consistent with the results of gene microarray analysis, verifying the reliability of the microarray data.
Plasma samples from the same patient were subjected to qRT-PCR, which offers an alternative noninvasive strategy and helps to analyze whether a differential expression is more attributable to the local effect of the ocular vitreous or to the mutual influence of systemic and vitreoretinopathy modifications. The differential expression of all qRT-PCR-verified transcripts in plasma was not statistically significant, suggesting that these transcripts are likely to result from local differential expression of the ocular vitreous.
In this study, LINC01210 expression levels were significantly lower in the vitreous humor of patients with IMH patients in those study. LINC01210 is associated with the proliferative, migratory, and invasive abilities of cells [28, 29]. We need further investigation to verify the potential role of LINC01210 in PDR. To date, the effects of most lncRNAs tested by qRT-PCR in our study are not explicitly understood. In the future, comprehensive studies on the function of lncRNAs in the pathogenesis of PDR will help determine new and effective diagnostic and therapeutic targets.
Most lncRNAs were poorly annotated. Bioinformatics analysis was used to investigate further differentially expressed lncRNAs. Using bioinformatics analysis, we found that the mRNA expression levels of frizzled class receptor 6 (FZD6) and proteasomal subunit α4s (PSMA8), which associated with the Wnt signaling pathway and Alzheimer disease, were downregulated in the vitreous of patients with PDR. Of interest, increased Wnt signaling is one of the causes of pathological ocular neovascularization of DR [30, 31]. Further, multiple factors of DR have been shown to play a vital role in the development of neurodegeneration in Alzheimer’s disease [32]. These potential correlations will be explored in future studies; in vivo and in vitro studies should be performed to elucidate the molecular mechanisms of lncRNA-mediated PDR pathogenesis.
This study had certain limitations. First, the number of patients included in this study was relatively small. However, these results were statistically significant. We provided one of the few lncRNA expression profiles in individual vitreous samples of patients with PDR. Second, vitrectomy surgery for patients with DR was performed out of necessity, such as in cases of vitreous hemorrhage and retinal detachment [33]. Therefore, this study did not exclude patients with vitreous hemorrhage. If vitreous hemorrhage was incorporated, the surgeon avoided collecting as much blood as possible. Blood-borne molecules in vitreous samples cannot be completely eliminated [9]. Finally, a selection bias may exist. The effects of anti-VEGF drugs were evaluated in different cohorts because it is difficult to collect dissimilar vitreous samples from same patient with PDR.
## Conclusions
There were systemic expression differences in the vitreous between patients with PDR and those with IMH and between patients with PDR who had undergone anti-VEGF treatment and those who had not. The levels of RP11-573 J24.1, RP11-787B4.2, RP11-654G14.1, RP11-2A4.3, and RP11-502I4.3 in the vitreous humor of patients with PDR were lower than those in patients with IMH. RP4-631H13.2 in the vitreous of patients with PDR after anti-VEGF treatment showed increased expression compared to that in untreated patients with PDR.
## Supplementary Information
Additional file 1: Table S1. Primer sequence. Additional file 2: Fig. S1. Heatmaps were generated from the hierarchical cluster analysis (Group B versus Group C).Additional file 3: Fig. S2. GO Enrichment Analysis (Group B versus Group C).Additional file 4: Fig. S3. KEGG Pathway Enrichment Analysis (Group B versus Group C).Additional file 5: Fig. S4. The co-expression network of differential transcripts (Group B versus Group C).Additional file 6: Table S2. Analysis of neighbor genes of the noncoding RNA transcripts (Group B versus Group C).Additional file 7: Table S3. Dysregulated lncRNAs in microarray data analysis (Group B versus Group C).Additional file 8: Fig. S5. Expression levels of lncRNAsof the confirmation cohort, in vitreous and plasma (Group B versus Group C).
## References
1. Vujosevic S, Aldington SJ, Silva P. **Screening for diabetic retinopathy: new perspectives and challenges**. *Lancet Diabetes Endocrinol* (2020.0) **8** 337-347. DOI: 10.1016/S2213-8587(19)30411-5
2. Qin X, Zou H. **The role of lipopolysaccharides in diabetic retinopathy**. *BMC Ophthalmol* (2022.0) **22** 86. DOI: 10.1186/s12886-022-02296-z
3. Wilkinson CP, Ferris FL, Klein RE. **Proposed international clinical diabetic retinopathy and diabetic macular edema disease severity scales**. *Ophthalmology.* (2003.0) **110** 1677-1682. DOI: 10.1016/S0161-6420(03)00475-5
4. Mukwaya A, Jensen L, Lagali N. **Relapse of pathological angiogenesis: functional role of the basement membrane and potential treatment strategies**. *Exp Mol Med* (2021.0) **53** 189-201. DOI: 10.1038/s12276-021-00566-2
5. Crabtree GS, Chang JS. **Management of Complications and Vision Loss from proliferative diabetic retinopathy**. *Curr Diab Rep* (2021.0) **21** 33. DOI: 10.1007/s11892-021-01396-2
6. Chang X, Zhu G, Cai Z. **miRNA, lncRNA and circRNA: targeted molecules full of therapeutic prospects in the development of diabetic retinopathy**. *Front Endocrinol (Lausanne).* (2021.0) **12** 771552. DOI: 10.3389/fendo.2021.771552
7. Yan B, Tao ZF, Li XM. **Aberrant expression of long noncoding RNAs in early diabetic retinopathy**. *Invest Ophthalmol Vis Sci* (2014.0) **55** 941-951. DOI: 10.1167/iovs.13-13221
8. Thomas AA, Biswas S, Feng B. **lncRNA H19 prevents endothelial-mesenchymal transition in diabetic retinopathy**. *Diabetologia.* (2019.0) **62** 517-530. DOI: 10.1007/s00125-018-4797-6
9. Nawaz IM, Rezzola S, Cancarini A. **Human vitreous in proliferative diabetic retinopathy: characterization and translational implications**. *Prog Retin Eye Res* (2019.0) **72** 100756. DOI: 10.1016/j.preteyeres.2019.03.002
10. Chatziralli I, Loewenstein A. **Intravitreal anti-vascular endothelial growth factor agents for the treatment of diabetic retinopathy: a review of the literature**. *Pharmaceutics.* (2021.0) **13** 1137. DOI: 10.3390/pharmaceutics13081137
11. Chung SH, Frick SL, Yiu G. **Targeting vascular endothelial growth factor using retinal gene therapy**. *Ann Transl Med* (2021.0) **9** 1277. DOI: 10.21037/atm-20-4417
12. Kanehisa M, Goto S. **KEGG: Kyoto encyclopedia of genes and genomes**. *Nucleic Acids Res* (2000.0) **28** 27-30. DOI: 10.1093/nar/28.1.27
13. Kanehisa M. **Toward understanding the origin and evolution of cellular organisms**. *Protein Sci* (2019.0) **28** 1947-1951. DOI: 10.1002/pro.3715
14. 14.Kanehisa M, Furumichi M, Sato Y, et al. KEGG for taxonomy-based analysis of pathways and genomes. Nucleic Acids Res. 2023;51(D1):D587-92.
15. Rinn JL, Chang HY. **Genome regulation by long noncoding RNAs**. *Annu Rev Biochem* (2012.0) **81** 145-166. DOI: 10.1146/annurev-biochem-051410-092902
16. Guo J, Liu Z, Gong R. **Long noncoding RNA: an emerging player in diabetes and diabetic kidney disease**. *Clin Sci (Lond)* (2019.0) **133** 1321-1339. DOI: 10.1042/CS20190372
17. Thomas AA, Feng B, Chakrabarti S. **ANRIL: a regulator of VEGF in diabetic retinopathy**. *Invest Ophthalmol Vis Sci* (2017.0) **58** 470-480. DOI: 10.1167/iovs.16-20569
18. Wang Y, Wang L, Guo H. **Knockdown of MALAT1 attenuates high-glucose-induced angiogenesis and inflammation via endoplasmic reticulum stress in human retinal vascular endothelial cells**. *Biomed Pharmacother* (2020.0) **124** 109699. DOI: 10.1016/j.biopha.2019.109699
19. Liu P, Jia SB, Shi JM. **LncRNA-MALAT1 promotes neovascularization in diabetic retinopathy through regulating miR-125b/VE-cadherin axis**. *Biosci Rep* (2019.0) **39** BSR20181469. DOI: 10.1042/BSR20181469
20. Yang J, Yang FJ, Wang YG. **LncRNA MIR497HG inhibits proliferation and migration of retinal endothelial cells under high-level glucose treatment via miRNA-128-3p/SIRT1 axis**. *Eur Rev Med Pharmacol Sci* (2020.0) **24** 5871-5877. PMID: 32572899
21. Ke N, Pi LH, Liu Q. **Long noncoding RNA SNHG7 inhibits high glucose-induced human retinal endothelial cells angiogenesis by regulating miR-543/SIRT1 axis**. *Biochem Biophys Res Commun* (2019.0) **514** 503-509. DOI: 10.1016/j.bbrc.2019.04.141
22. Gong Q, Dong W, Fan Y. **LncRNA TDRG1-mediated overexpression of VEGF aggravated retinal microvascular endothelial cell dysfunction in diabetic retinopathy**. *Front Pharmacol* (2020.0) **10** 1703. DOI: 10.3389/fphar.2019.01703
23. Wang J, Gao X, Liu J. **Effect of intravitreal conbercept treatment on the expression of long noncoding RNAs and mRNAs in proliferative diabetic retinopathy patients**. *Acta Ophthalmol* (2019.0) **97** e902-e912. DOI: 10.1111/aos.14083
24. Mandava N, Tirado-Gonzalez V, Geiger MD. **Complement activation in the vitreous of patients with proliferative diabetic retinopathy**. *Invest Ophthalmol Vis Sci* (2020.0) **61** 39. DOI: 10.1167/iovs.61.11.39
25. Tomita Y, Cagnone G, Fu Z. **Vitreous metabolomics profiling of proliferative diabetic retinopathy**. *Diabetologia.* (2021.0) **64** 70-82. DOI: 10.1007/s00125-020-05309-y
26. Wu F, Phone A, Lamy R. **Correlation of aqueous, vitreous, and plasma cytokine levels in patients with proliferative diabetic retinopathy**. *Invest Ophthalmol Vis Sci* (2020.0) **61** 26. DOI: 10.1167/iovs.61.2.26
27. Zandi S, Tappeiner C, Pfister IB. **Vitreal cytokine profile differences between eyes with Epiretinal membranes or macular holes**. *Invest Ophthalmol Vis Sci* (2016.0) **57** 6320-6326. DOI: 10.1167/iovs.16-20657
28. Zhang C, Liu J, Zhang Y. **LINC01210 accelerates proliferation, invasion and migration in ovarian cancer through epigenetically downregulating KLF4**. *Biomed Pharmacother* (2019.0) **119** 109431. DOI: 10.1016/j.biopha.2019.109431
29. Luo J, Gao K, Chen M. **LINC01210 promotes malignant phenotypes of colorectal cancer through epigenetically upregulating SRSF3**. *Pathol Res Pract* (2011.0) **234** 153905. DOI: 10.1016/j.prp.2022.153905
30. Wang Z, Liu CH, Huang S. **Wnt signaling in vascular eye diseases**. *Prog Retin Eye Res* (2019.0) **70** 110-133. DOI: 10.1016/j.preteyeres.2018.11.008
31. Nguyen H, Lee SJ, Li Y. **Selective activation of the Wnt-signaling pathway as a novel therapy for the treatment of diabetic retinopathy and other retinal vascular diseases**. *Pharmaceutics.* (2022.0) **14** 2476. DOI: 10.3390/pharmaceutics14112476
32. Jash K, Gondaliya P, Kirave P. **Cognitive dysfunction: a growing link between diabetes and Alzheimer's disease**. *Drug Dev Res* (2020.0) **81** 144-164. DOI: 10.1002/ddr.21579
33. Tomita Y, Lee D, Tsubota K. **Updates on the current treatments for diabetic retinopathy and possibility of future Oral therapy**. *J Clin Med* (2021.0) **10** 4666. DOI: 10.3390/jcm10204666
|
---
title: Impact of oral statin therapy on clinical outcomes in patients with cT1 breast
cancer
authors:
- Koji Takada
- Shinichiro Kashiwagi
- Nozomi Iimori
- Rika Kouhashi
- Akimichi Yabumoto
- Wataru Goto
- Yuka Asano
- Yukie Tauchi
- Tamami Morisaki
- Kana Ogisawa
- Masatsune Shibutani
- Hiroaki Tanaka
- Kiyoshi Maeda
journal: BMC Cancer
year: 2023
pmcid: PMC9999569
doi: 10.1186/s12885-023-10631-w
license: CC BY 4.0
---
# Impact of oral statin therapy on clinical outcomes in patients with cT1 breast cancer
## Abstract
### Purpose
A previous meta-analysis examining the relationship between statin use and breast cancer reported that the inhibitory effect of statins on breast cancer may be more pronounced in early-stage cases. In this study, we aimed to investigate the effects of hyperlipidemia treatment at the time of breast cancer diagnosis and to examine its correlation with metastasis to axillary lymph nodes among patients with so-called cT1 breast cancer whose primary lesion was 2 cm or less and was pathologically evaluated by sentinel lymph node biopsy or axillary lymph node dissection. We also investigated the effects of hyperlipidemic drugs on the prognosis of patients with early-stage breast cancer.
### Methods
After excluding cases that did not meet the criteria, we analyzed data from 719 patients who were diagnosed with breast cancer, with a primary lesion of 2 cm or less identified by preoperative imaging, and who underwent surgery without preoperative chemotherapy.
### Results
Regarding hyperlipidemia drugs, no correlation was found between statin use and lymph node metastasis ($$p \leq 0.226$$), although a correlation was found between lipophilic statin use and lymph node metastasis ($$p \leq 0.042$$). Also, the disease-free survival periods were prolonged following treatment of hyperlipidemia ($$p \leq 0.047$$, hazard ratio: 0.399) and statin administration ($$p \leq 0.028$$, hazard ratio: 0.328).
### Conclusion
In cT1 breast cancer, the results suggest that oral statin therapy may contribute to favorable outcomes.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12885-023-10631-w.
## Background
Although various orally administered drugs are clinically used in the treatment of a wide variety of diseases, it has been reported that some may have unexpected effects on cancer. For example, in a systematic review and meta-analysis, the diabetes drug metformin reduced the risk of colorectal cancer and prostate cancer [1]. However, the analysis in that report did not reveal a reduction in breast cancer risk, whereas some studies have reported that metformin reduces breast cancer risk and improves prognosis [1, 2]. Another meta-analysis of observational studies reported that long-term use of angiotensin-receptor blockers / angiotensin-converting enzyme inhibitors for the treatment of hypertension might reduce the risk of breast cancer [3]. Among these drugs, statins, which are typically used for the treatment of hyperlipidemia, have also been reported to suppress the development of cancer and reduce the rate of recurrence [4–12]. These outcomes may be explained by many preclinical studies that have reported antiproliferative and anti-apoptotic effects in breast cancer [13–17]. In addition, based on the anti-invasive properties [18–22] and metastasis-suppressing effects of statins that have been demonstrated in preclinical studies, some reports have clinically examined their progression-suppressing effects in breast cancer [23–29].
Another meta-analysis examining the relationship between statin use and breast cancer reported that the inhibitory effect of statins on breast cancer may be more pronounced in patients with early-stage breast cancer [30]. Therefore, we hypothesized that statins may affect metastasis to lymph nodes in breast cancer cases involving a small primary lesion. In recent years, axillary surgery for early-stage breast cancer has been reduced due to the increased effectiveness of multidisciplinary treatment before and after surgery, and evaluation of axillary lymph node metastasis before treatment has become even more important. If our hypothesis is correct, statin administration may affect the evaluation. In this study, we aimed to investigate the treatment of hyperlipidemia at the time of breast cancer diagnosis and to examine its correlation with the metastatic status in axillary lymph nodes among patients with so-called cT1 breast cancer involving a primary lesion of 2 cm or less who underwent pathological evaluations of metastasis in an axillary lymph node by sentinel lymph node biopsy or axillary lymph node dissection. We also aimed to investigate the effects of hyperlipidemic drugs on the prognosis of patients with early-stage breast cancer.
## Patient background and classification
Seven hundred forty-two patients were diagnosed with breast cancer involving a primary lesion of 2 cm or less by preoperative imaging and underwent surgery without preoperative chemotherapy from April 2007 to March 2020 at Osaka City University Hospital. Pathological diagnosis of breast cancer was based on core needle biopsy (CNB) or vacuum-assisted biopsy (VAB). As an evaluation of their general condition before initiating treatment for breast cancer, the patients were confirmed to have a history of pre-treatment and oral medication use. We classified the drugs used to treat hyperlipidemia for further examination. The pharmacological classification of statins based on their hydrophilicity and lipophilicity was performed according to the classification system widely used in cardiovascular studies [31, 32]. Specifically, rosuvastatin and pravastatin are classified as hydrophilic statins, while atorvastatin, pitavastatin, simvastatin and fluvastatin are classified as lipophilic statins. Either mastectomy or breast-conserving surgery was performed because the preoperative imaging examinations such as ultrasonography (US), computed tomography (CT), and bone scintigraphy revealed that radical resection was possible. Axillary lymph node dissection was performed for cases in which axillary lymph node metastasis was suspected, and sentinel lymph node biopsy was performed for cases in which no metastasis was diagnosed. During surgery for breast cancer, the sentinel lymph node was identified using a combination of radioisotope and dye methods according to previous reports [33, 34]. Histopathological diagnosis of sentinel lymph node metastasis was conducted by slicing the entire sentinel lymph node into 2-mm-thick sections [35, 36]. Sentinel lymph node metastases were categorized by size according to previously reported parameters (macrometastasis: tumor diameter > 2 mm; micrometastasis: tumor diameter > 0.2 mm, ≤ 2 mm or < 200 tumor cells; for isolated tumor cells: tumor diameter < 0.2 mm or < 200 tumor cells) [37]. Axillary dissection was additionally performed in patients with macrometastasis that was confirmed via sentinel lymph node biopsy.
The expression levels of estrogen receptor (ER), progesterone receptor (PgR), human epidermal growth factor receptor 2 (HER2), and Ki67, a marker of proliferation, were examined immunohistochemically in both the biopsy tissue used for breast cancer diagnosis and the surgically removed tissue. Based on the results of the immunohistological staining, breast cancer was classified into the following three subtypes: triple-negative breast cancer (TNBC; negative for ER, PgR, and HER2); hormone receptor (HR)-HER2+ breast cancer (HR-negative and HER2-positive breast cancer; ER-, PgR-, and HER2+); and HR+ breast cancer (hormone receptor-positive breast cancer; ER+ and/or PgR+).
There were 742 preoperatively diagnosed cases of cT1 breast cancer. However, 15 cases did not undergo axillary lymph node dissection or sentinel lymph node biopsy, and eight cases were being treated with unknown medications at the time of diagnosis. Therefore, these 23 cases were excluded from this study, and data was analyzed from the remaining 719 cases.
## Statistical analysis
All statistical analyses were performed using the JMP software package (SAS, Tokyo, Japan). Each correlation was examined using Pearson’s chi-square test. The odds ratio (OR) and $95\%$ confidence interval (CI) were calculated by logistic analysis, and multivariable analysis was performed using the multivariable logistic regression model. Prognostic analyses, such as the calculation of recurrence-free survival (RFS) or overall survival (OS), were conducted using the Kaplan–Meier method and the log-rank test. The hazard ratios (HR) and $95\%$ CI were calculated using the Cox proportional hazards model. Multivariable analysis was performed using the Cox regression model. A p-value of < 0.05 was considered statistically significant.
## Clinicopathological features
Table 1 shows the clinicopathological features of the 719 patients with cT1 breast cancer who underwent surgery without receiving preoperative chemotherapy. The median age was 58 years (range, 29–79 years), and the median tumor diameter was 13 mm (range, 3.0–20.0 mm). A total of 612 patients ($85.7\%$) were positive for ER, 398 patients ($55.4\%$) were positive for PgR, and 621 patients ($86.3\%$) were classified as having HR+ breast cancer, which represented the majority of cases. There were 66 patients ($9.2\%$) with HER2-positive breast cancer, but only 27 patients ($3.8\%$) were classified as having HR-HER2+ breast cancer. Seventy-one patients ($9.9\%$) were classified as having TNBC. Ki67 was expressed at a level higher than $20\%$ in 133 patients ($18.5\%$).Table 1Clinicopathological features of 719 cT1 breast cancer patients who underwent surgery without preoperative chemotherapyParametersNumber of patients ($$n = 719$$) (%)Age at operation (years old)median 58 (range, 29–91)Tumor size (mm)median 13 (range, 3–20)Estrogen receptor Negative / Positive107 ($14.9\%$) / 612 ($85.1\%$)Progesterone receptor Negative / Positive321 ($44.6\%$) / 398 ($55.4\%$)HER2 Negative / Positive653 ($90.8\%$) / 66 ($9.2\%$)Ki67 ≤ $20\%$ / > $20\%$586 ($81.5\%$) / 133 ($18.5\%$)Intrinsic subtype HR + BC / HR-HER2 + BC / TNBC621 ($86.3\%$) / 27 ($3.8\%$) / 71 ($9.9\%$)Pathological axillary lymph node metastasis *No metastasis* / only isolated tumor cell / only micrometastasis / metastasis573 ($79.7\%$) / 5 ($0.7\%$) / 29 ($4.0\%$) / 112 ($15.6\%$)Lymph vascular invasion No / Yes528 ($73.4\%$) / 191 ($26.6\%$)Hyperlipidemia No / Yes572 ($79.6\%$) / 147 ($20.4\%$)Number of medicine types for hyperlipidemia 0 / 1 / 2572 ($79.6\%$) / 139 ($19.3\%$) / 8 ($1.1\%$)Statins Non-user / User587 ($81.6\%$) / 132 ($18.4\%$)Lipophilic statins Non-user / User658 ($91.5\%$) / 61 ($8.5\%$)Hydrophilic statins Non-user / User648 ($90.1\%$) / 71 ($9.9\%$)Fibrate Non-user / User709 ($98.6\%$) / 10 ($1.4\%$)Nicotinic acid (tocopherol acetate) Non-user / User712 ($99.0\%$) / 7 ($1.0\%$)Sterol absorption inhibitors (ezetimibe) Non-user / User713 ($99.2\%$) / 6 ($0.8\%$)HER2 Human epidermal growth factor receptor 2, HR + BC Hormone receptor-positive breast cancer (ER+ and/or PgR+), HR-HER2 + BC Human epidermal growth factor receptor 2-enriched breast cancer (ER-, PgR-, and HER2+), TNBC Triple negative breast cancer (ER-, PgR-, and HER2-) Postoperative pathological examinations revealed no axillary lymph node metastasis in 607 patients ($84.4\%$), including five patients ($0.7\%$) with isolated tumor cells and 29 patients ($4.0\%$) with micrometastases based on sentinel lymph node biopsies. The median number of lymph node metastases in 112 patients ($15.6\%$) with axillary lymph node metastases was two (range, 1–26). Lymphovascular invasion was detected in 191 patients ($26.6\%$).
At the time of breast cancer diagnosis, 147 patients ($20.4\%$) were undergoing treatment with orally administered drugs for hyperlipidemia. Among them, only eight patients ($1.1\%$) were taking multiple drugs, whereas most were treated with single drugs. Among the 132 patients ($18.4\%$) who were being treated with statins, 61 patients ($8.5\%$) were taking lipophilic statins, and 71 patients ($9.9\%$), about half, were taking hydrophilic statins. Specifically, rosuvastatin, one of the hydrophilic statins, users were 36 patients ($5.0\%$) and pravastatin users were 35 patients ($4.9\%$). On the other hand, the results for lipophilic statins were as follows: atorvastatin; 27 patients ($3.8\%$), pitavastatin; 20 patients ($2.8\%$), simvastatin; 13 patients ($1.8\%$), and fluvastatin 1 patients ($0.1\%$). There were 10 fibrate users ($1.4\%$), seven nicotinic acid (tocopherol acetate) users ($1.0\%$), and six sterol absorption inhibitors (ezetimibe) users ($0.8\%$).
## Correlations between clinicopathological features and axillary lymph node metastasis
The correlations between clinicopathological features and axillary lymph node metastasis are listed in Table 2. Metastasis occurred significantly more frequently when the breast cancer tumor diameter exceeded 10 mm ($p \leq 0.001$). Although the relationship was not statistically significant, metastases tended to be found in PgR-positive breast cancer cases ($$p \leq 0.063$$). Metastases occurred significantly more frequently in breast cancer cases involving lymphovascular invasion ($p \leq 0.001$). Regarding hyperlipidemia drugs, no correlation was found between statin use in general and lymph node metastasis ($$p \leq 0.226$$); however, a significant correlation was found between the use of lipophilic statins and lymph node metastasis ($$p \leq 0.042$$).Table 2Correlation between axillary lymph node metastasis and clinicopathological featuresParametersAxillary lymph node metastasisp valueNo metastasis, including even micrometasis ($$n = 607$$)metastasis ($$n = 112$$)Age at operation (years old)0.872 ≤ 60331 ($54.5\%$)62 ($55.4\%$) > 60276 ($45.5\%$)50 ($44.6\%$)Tumor size (mm)< 0.001 ≤ 10.0202 ($33.3\%$)15 ($13.4\%$) > 10.0405 ($66.7\%$)97 ($86.6\%$)Estrogen receptor0.441 Negative93 ($15.3\%$)14 ($12.5\%$) Positive514 ($84.7\%$)98 ($87.5\%$)Progesterone receptor0.063 Negative280 ($46.1\%$)41 ($36.6\%$) Positive327 ($53.9\%$)71 ($63.4\%$)HER20.920 Negative551 ($90.8\%$)102 ($91.1\%$) Positive56 ($9.2\%$)10 ($8.9\%$)Ki670.734 ≤ $20\%$496 ($81.7\%$)90 ($80.4\%$) > $20\%$111 ($18.3\%$)22 ($19.6\%$)Intrinsic subtype HR + BC0.201 No87 ($14.3\%$)11 ($9.8\%$) Yes520 ($85.7\%$)101 ($90.2\%$)Intrinsic subtype HR-HER2 + BC0.911 No584 ($96.2\%$)108 ($96.4\%$) Yes23 ($3.8\%$)4 ($3.6\%$)Intrinsic subtype TNBC0.162 No543 ($89.5\%$)105 ($93.8\%$) Yes64 ($10.5\%$)7 ($6.3\%$)Lymph vascular invasion< 0.001 No478 ($78.7\%$)50 ($44.6\%$) Yes129 ($21.3\%$)62 ($55.4\%$)Hyperlipidemia0.212 No478 ($78.7\%$)94 ($83.9\%$) Yes129 ($21.3\%$)18 ($16.1\%$)Multiple medicine types for hyperlipidemia0.460 No601 ($99.0\%$)110 ($98.2\%$) Yes6 ($1.0\%$)2 ($1.8\%$)Statins0.226 Non-user491 ($80.9\%$)96 ($85.7\%$) User116 ($19.1\%$)16 ($14.3\%$)Lipophilic statins0.042 Non-user550 ($90.6\%$)108 ($96.4\%$) User57 ($9.4\%$)4 ($3.6\%$)Hydrophilic statins0.746 Non-user548 ($90.3\%$)100 ($89.3\%$) User59 ($9.7\%$)12 ($10.7\%$)Fibrate0.624 Non-user598 ($98.5\%$)111 ($99.1\%$) User9 ($1.5\%$)1 ($0.9\%$)Nicotinic acid (tocopherol acetate)0.341 Non-user602 ($99.2\%$)110 ($98.2\%$) User5 ($0.8\%$)2 ($1.8\%$)Sterol absorption inhibitors (ezetimibe)0.941 Non-user602 ($99.2\%$)111 ($99.1\%$) User5 ($0.8\%$)1 ($0.9\%$)HER2 Human epidermal growth factor receptor 2, HR + BC Hormone receptor-positive breast cancer (ER+ and/or PgR+), HR-HER2 + BC Human epidermal growth factor receptor 2-enriched breast cancer (ER-, PgR-, and HER2+), TNBC Triple negative breast cancer (ER-, PgR-, and HER2-) Examination of the correlation between lipophilic statin use and clinicopathological factors revealed that the users were significantly older than the non-users ($p \leq 0.001$) (Table 3).Table 3Correlation between lipophilic statins user and clinicopathological featuresParametersLipophilic statinsp valueNon-user ($$n = 658$$)User ($$n = 61$$)Age at operation (years old)< 0.001 ≤ 60388 ($59.0\%$)5 ($8.2\%$) > 60270 ($41.0\%$)56 ($91.8\%$)Tumor size (mm)0.450 ≤ 10.0196 ($29.8\%$)21 ($34.4\%$) > 10.0462 ($70.2\%$)40 ($65.6\%$)Estrogen receptor0.247 Negative101 ($15.3\%$)6 ($9.8\%$) Positive557 ($84.7\%$)55 ($90.2\%$)Progesterone receptor0.548 Negative296 ($45.0\%$)25 ($41.0\%$) Positive362 ($55.0\%$)36 ($59.0\%$)HER20.228 Negative595 ($90.4\%$)58 ($95.1\%$) Positive63 ($9.6\%$)3 ($4.9\%$)Ki670.554 ≤ $20\%$538 ($81.8\%$)48 ($78.7\%$) > $20\%$120 ($18.2\%$)13 ($21.3\%$)Intrinsic subtype HR + BC0.367 No92 ($14.0\%$)6 ($9.8\%$) Yes566 ($86.0\%$)55 ($90.2\%$)Intrinsic subtype HR-HER2 + BC0.107 No631 ($95.9\%$)61 ($100.0\%$) Yes27 ($4.1\%$)0 ($0.0\%$)Intrinsic subtype TNBC0.992 No593 ($90.1\%$)55 ($90.2\%$) Yes65 ($9.9\%$)6 ($9.8\%$)Lymph vascular invasion0.504 No481 ($73.1\%$)47 ($77.0\%$) Yes177 ($26.9\%$)14 ($23.0\%$)Hyperlipidemia< 0.001 No572 ($86.9\%$)0 ($0.0\%$) Yes86 ($13.1\%$)61 ($100.0\%$)Multiple medicine types for hyperlipidemia0.003 No653 ($99.2\%$)58 ($95.1\%$) Yes5 ($0.8\%$)3 ($4.9\%$)Hydrophilic statins0.007 Non-user587 ($89.2\%$)61 ($100.0\%$) User71 ($10.8\%$)0 ($0.0\%$)Fibrate0.862 Non-user649 ($98.6\%$)60 ($98.4\%$) User9 ($1.4\%$)1 ($1.6\%$)Nicotinic acid (tocopherol acetate)0.055 Non-user653 ($99.2\%$)59 ($96.7\%$) User5 ($0.8\%$)2 ($3.3\%$)Sterol absorption inhibitors (ezetimibe)0.454 Non-user652 ($99.1\%$)61 ($100.0\%$) User6 ($0.9\%$)0 ($0.0\%$)HER2 Human epidermal growth factor receptor 2. HR + BC Hormone receptor-positive breast cancer (ER+ and/or PgR+). HR-HER2 + BC Human epidermal growth factor receptor 2-enriched breast cancer (ER-, PgR-, and HER2+), TNBC Triple negative breast cancer (ER-, PgR-, and HER2-) We examined the factors causing axillary lymph node metastasis in patients with cT1 breast cancer; tumor size ($p \leq 0.001$, OR = 3.225) and lymphovascular invasion ($p \leq 0.001$, OR = 4.595), as well as the use of lipophilic statins ($$p \leq 0.042$$, OR = 0.357) were the factors associated with axillary lymph node metastasis (Table 4) (Fig. 1). Even after performing the multivariate analysis, these remained independent factors (tumor size: $$p \leq 0.003$$, OR = 2.352; lymphovascular invasion: $p \leq 0.001$, OR = 3.891; lipophilic statin use: $$p \leq 0.048$$, OR = 0.384). Thus, lipophilic statin was the only factor that reduced axillary lymph node metastasis. Table 4Univariate and multivariate analysis with axillary lymph node metastasis for cT1 breast cancerParametersUnivarite analysisMultivarite analysisOdds ratio$95\%$ CIp valueOdds ratio$95\%$ CIp valueAge at operation (years old) ≤ 60 vs > 600.9670.645–1.4510.872Tumor size (mm) ≤ 10.0 vs > 10.03.2251.825–5.700< 0.0012.3521.337–4.3910.003Estrogen receptor Negative vs Positive1.2660.694–2.3120.441Progesterone receptor Negative vs Positive1.4830.978–2.2480.0631.4570.945–2.2690.089HER2 Negative vs Positive0.9650.477–1.9530.920Ki67 ≤ $20\%$ vs > $20\%$1.0920.656–1.8180.734Intrinsic subtype HR + BC No vs Yes1.5360.792–2.9790.201Intrinsic subtype HR-HER2 + BC No vs Yes0.9400.319–2.7730.911Intrinsic subtype TNBC No vs Yes0.5660.252–1.2690.162Lymph vascular invasion No vs Yes4.5953.018–6.995< 0.0013.8912.529–6.016< 0.001Hyperlipidemia No vs Yes0.7100.413–1.2180.212Multiple medicine types for hyperlipidemia No vs Yes1.8210.363–9.1400.460Statins Non-user vs User0.7050.400–1.2430.226Lipophilic statins Non-user vs User0.3570.127–0.9960.0420.3840.113–0.9870.048Hydrophilic statins Non-user vs User1.1150.578–2.1480.746Fibrate Non-user vs User0.5990.075–4.7720.624Nicotinic acid (tocopherol acetate) Non-user vs User2.1890.419–11.4260.341Sterol absorption inhibitors (ezetimibe) Non-user vs User1.0850.125–9.3730.941HER2 Human epidermal growth factor receptor 2, HR + BC Hormone receptor-positive breast cancer (ER+ and/or PgR+), HR-HER2 + BC Human epidermal growth factor receptor 2-enriched breast cancer (ER-, PgR-, and HER2+), TNBC Triple negative breast cancer (ER-, PgR-, and HER2-), CI Confidence intervalsFig. 1Forest plot showed odd ratios for the univariate association of the risk factors for axillary lymph node metastasis
## Effects of lipophilic statins on prognosis
We examined the prognosis of 719 patients with cT1 breast cancer included in this study. The median follow-up period was 1838 days (range, 54–4841 days). During that period, 42 patients ($5.8\%$) experienced recurrence, three patients ($0.4\%$) died from breast cancer, and 11 patients ($1.5\%$) died from other causes. Univariate analysis of disease-free survival (DFS) times showed that tumor size affected prognosis ($$p \leq 0.011$$, HR: 2.902) and that vascular infiltration tended to lead to a poor prognosis ($$p \leq 0.086$$, HR: 1.712) (Online Resource Supplementary Table 1). Among the factors, the treatment of hyperlipidemia ($$p \leq 0.047$$, HR: 0.399) and statin use ($$p \leq 0.028$$, HR: 0.328) were associated with prolonged DFS periods. In the multivariate analysis, only tumor size was an independent factor ($$p \leq 0.025$$, HR: 2.620). Similarly, in the univariate analysis for RFS, tumor size ($$p \leq 0.017$$, HR: 2.732) as well as statin use ($$p \leq 0.038$$, HR: 0.345) affected prognosis (Table 5) (Fig. 2). No clinicopathological factors significantly affected OS (Table 6).Table 5Univariate and multivariate analysis with recurrence-free survival for cT1 breast cancerParametersUnivarite analysisMultivarite analysisHazard ratio$95\%$ CIp valueHazard ratio$95\%$ CIp valueAge at operation (years old) ≤ 60 vs > 600.6070.310–1.1330.119Tumor size (mm) ≤ 10.0 vs > 10.02.7321.177–7.9460.0172.6581.14–7.7390.021Estrogen receptor Negative vs Positive1.1660.549–2.8700.707Progesterone receptor Negative vs Positive1.5090.808–2.9500.200HER2 Negative vs Positive0.8870.214–2.4490.839Ki67 ≤ $20\%$ vs > $20\%$0.7980.274–1.8540.626Intrinsic subtype HRBC No vs Yes1.2160.551–3.2130.651Intrinsic subtype HER2BC No vs Yes0.5360.030–2.4600.494Intrinsic subtype TNBC No vs Yes0.9470.325–2.2010.909Pathological axillary lymph node metastasis *No metastasis* vs Metastasis0.9450.385–2.0020.891Lymph vascular invasion No vs Yes1.6730.888–3.0760.109Hyperlipidemia No vs Yes0.4210.126–1.0470.0641.0140.057–4.6930.989Multiple medicine types for hyperlipidemia No vs Yes––0.294Statins Non-user vs User0.3450.083–0.9510.0380.3530.045–7.1510.413Lipophilic statins Non-user vs User0.3160.018–1.4510.166Hydrophilic statins Non-user vs User0.4020.065–1.3070.147Fibrate Non-user vs User1.97010.111–9.0590.545Nicotinic acid (tocopherol acetate) Non-user vs User––0.345Sterol absorption inhibitors (ezetimibe) Non-user vs User––0.409HER2 Human epidermal growth factor receptor 2, HRBC Hormone receptor-positive breast cancer (ER+ and/or PgR+), HER2BC Human epidermal growth factor receptor 2-enriched breast cancer (ER-, PgR-, and HER2+), TNBC Triple negative breast cancer (ER-, PgR-, and HER2-), CI Confidence intervalsFig. 2Kaplan–Meier method comparing recurrence-free survival (RFS) and overall survival (OS) by statin or lipophilic statin. There was no significant difference in RFS due to statin (A) and lipophilic statin (B). No significant difference was found in OS due to statin (C) and lipophilic statin (D)Table 6Univariate and multivariate analysis with overall survival for cT1 breast cancerParametersUnivarite analysisMultivarite analysisHazard ratio$95\%$ CIp valueHazard ratio$95\%$ CIp valueAge at operation (years old) ≤ 60 vs > 602.3510.809–7.6770.117Tumor size (mm) ≤ 10.0 vs > 10.01.3230.412–5.8610.660Estrogen receptor Negative vs Positive3.0920.612–56.2490.202Progesterone receptor Negative vs Positive2.8090.876–12.4250.0852.6540.824–11.7700.106HER2 Negative vs Positive0.8700.048–4.3830.892Ki67 ≤ $20\%$ vs > $20\%$2.3810.649–7.1780.174Intrinsic subtype HRBC No vs Yes2.5960.515–47.1940.292Intrinsic subtype HER2BC No vs Yes––0.243Intrinsic subtype TNBC No vs Yes0.5710.031–2.8790.559Pathological axillary lymph node metastasis *No metastasis* vs Metastasis0.8580.133–3.1500.838Lymph vascular invasion No vs Yes1.3240.406–3.8390.621Hyperlipidemia No vs Yes2.3730.727–6.9100.143Multiple medicine types for hyperlipidemia No vs Yes6.1090.335–31.1840.169Statins Non-user vs User2.6050.798–7.5780.1072.4250.742–7.0570.134Lipophilic statins Non-user vs User2.2870.355–8.4490.328Hydrophilic statins Non-user vs User2.2530.508–7.2470.251Fibrate Non-user vs User7.7650.424–40.3670.131Nicotinic acid (tocopherol acetate) Non-user vs User––0.591Sterol absorption inhibitors (ezetimibe) Non-user vs User––0.672HER2 Human epidermal growth factor receptor 2, HRBC Hormone receptor-positive breast cancer (ER+ and/or PgR+), HER2BC Human epidermal growth factor receptor 2-enriched breast cancer (ER-, PgR-, and HER2+). TNBC Triple negative breast cancer (ER-, PgR-, and HER2-), CI Confidence intervals The prognoses were examined among the 607 patients who did not have macrometastases, and similar results were obtained. The median follow-up period was 1825 days (range, 54–4841 days). During that period, 35 patients ($5.8\%$) experienced recurrence, two patients ($0.3\%$) died from breast cancer, and 10 patients ($1.6\%$) died from other causes. In the univariate analysis for DFS, tumor size ($$p \leq 0.084$$, HR: 1.888) and PgR status ($$p \leq 0.032$$, HR: 1.977) affected prognosis, whereas the use of hyperlipidemic drugs did not (Online Resource Supplementary Table 2). On the other hand, in the univariate analysis for RFS, tumor size ($$p \leq 0.036$$, HR: 2.493) and PgR status ($$p \leq 0.043$$, HR: 2.064) affected the prognosis (Online Resource Supplementary Table 3). The analysis revealed that statin use ($$p \leq 0.096$$, HR: 0.411) tended to affect prognosis, but this did not reach statistical significance (Online Resource Supplementary Fig. 1). In the univariate analysis for OS, statin use ($$p \leq 0.047$$, HR: 3.460) was poor prognostic factor; in the multivariate analysis, no independent factors were found (Online Resource Supplementary Table 4).
## Discussion
In a study examining the correlation between lymph node metastasis and clinicopathological features among 91,364 patients with T1 breast cancer using information from the “Surveillance, Epidemiology, and End Results Program (SEER)” study, age, race, primary site, tumor size, and ER, PgR, and HER2 status were influencing factors [38]. Tumor size and lymphovascular invasion are cited as risk factors for lymph node metastasis in most studies involving sentinel lymph node biopsy [39–46]. This result also shows that tumor size and lymphovascular invasion were strongly correlated with lymph node metastasis, which is consistent with previously reported results. Among the investigated factors, this study showed that the use of lipophilic statins may suppress lymph node metastasis. In preclinical studies, statins have been shown to exhibit anti-proliferative on cancer by being associated with mechanisms that drive cell cycle disruption in cancer cells [13–17]. Many studies have investigated the effects of factors capable of suppressing the risk of breast cancer and its recurrence, and there have also been some reports examining the effects of statins on suppressing the progression of breast cancer. For example, when examining the correlation between statin use and clinicopathological factors at the time of diagnosis in about 2000 and 3000 breast cancer patients, respectively, the rates of diagnosis for breast cancer with high pathological malignancy and for highly advanced breast cancer were significantly lower in statin users than in non-users [27, 28]. In addition, a study of approximately 130,000 postmenopausal women conducted by the Women’s Health Initiative reported that the use of lipophilic statins reduced the rate of diagnosis of highly advanced breast cancer [29]. However, the opportunity for patient consultation is likely to strongly influence these results. On the other hand, in this study, the tumor size based on the TNM classification was used as a condition for examination; this methodology is different from that of previous reports. In preclinical studies, anti-invasive properties have also been reported [18–22], as have metastasis-suppressing effects [23–26]. This study demonstrates the possibility of suppressing lymph node metastasis in clinical practice, which could improve prognosis.
Based on many results from preclinical studies, it is expected that statins should suppress the risk of breast cancer and its recurrence. However, in clinical practice, contradictory results have been reported regarding the suppressing effect of statins on breast cancer risk [6, 47, 48]. One reports have discussed why prospective studies with statins have not yielded the expected results [49]. On the other hand, many studies have reported that statins reduce the risk of breast cancer recurrence, and some groups have reported that only lipophilic statins are effective, not hydrophilic statins [4, 5, 7–10, 12, 42]. A report indicated that effects may vary considerably among lipophilic statins [49]. The classification of statins in this study was the same as that used in a meta-analysis that examined the correlation between statin type and breast cancer prognosis [50]. In this study, statins reduced OS in patients without lymph node metastases. However, this result is likely due to the fact that only two patients ($0.3\%$) died from breast cancer and 10 patients ($1.6\%$) died from other causes. Breast-cancer-specific survival could not be examined due to the low numbers of breast cancer-related deaths; therefore, the results pertaining to OS in this study should be considered for reference. However, statin use tended to prolong the RFS period, instead of the DFS period. Regarding this result, the event point was narrowed down to the day of recurrence / death from breast cancer, suggesting that statins may have a positive effect on the treatment of early-stage breast cancer.
This study has some limitations that should be considered. First, patients receiving preoperative chemotherapy were excluded, as the evaluation of axillary lymph node metastasis is uncertain based on diagnostic imaging alone. Since it is known that the therapeutic effect of preoperative chemotherapy is a predictor of prognosis in HER2-positive breast cancer and TNBC [51–54], preoperative chemotherapy is actively performed for those types of breast cancer. The number of patients with HER2-positive breast cancer and TNBC was low, which could have been a source of bias in this study. In addition, statin was correlated with age, although age itself had no clear effect on axillary lymph node metastasis or prognosis in this study, it may have a significant effect. Moreover, one of the limitations was the exclusion of cases involving a primary lesion of 20 mm or less, accompanied by advanced regional lymph node metastasis or distant metastasis. Another limitation was that the duration of oral treatment for hyperlipidemia was unknown for each patient. However, clinical data, rather than in vivo or in vitro data, suggest that lipophilic statins may suppress breast cancer metastasis to lymph nodes. Furthermore, it was suggested that statins may suppress postoperative recurrence. Regarding the examination and treatment of axillary lymph nodes, in recent years, even sentinel lymph node biopsy has been deemed an overly invasive procedure for early-stage breast cancer cases, so clinical trials are underway to omit sentinel lymph node biopsies from the protocols for cN0 breast cancer cases assessed using US [55, 56]. It is also possible that lipophilic statins may have affected the results of these clinical trials. Regarding the prognosis, some studies have reported that even if statins are administered after the diagnosis of breast cancer, they may suppress the recurrence of breast cancer [4, 5, 7, 9, 10, 30]. Especially in ER-positive breast cancer, the effects driving the suppression of the risk of recurrence are well-recognized [5, 30]. The fact that the prognosis was affected in this study may have been due to the fact that ER-positive breast cancer patients accounted for the majority of the cases. This study suggests the possibility of improving the prognosis of breast cancer patients through treatment with statins.
## Conclusions
In patients with cT1 breast cancer, the results suggest that oral statin therapy may contribute to favorable outcomes.
## Supplementary Information
Additional file 1: Supplementary Fig. 1. Kaplan–Meier method comparing recurrence-free survival (RFS) and overall survival (OS) by statin or lipophilic statin in patients without lymph node metastasis. There was no significant difference in RFS due to statin (A) and lipophilic statin (B). However, statin user had poor OS ($$p \leq 0.025$$, log-rank) (C). No impact on OS was found in lipophilic statin(D).Additional file 2: Supplementary Table 1. Univariate and multivariate analysis with disease-free survival for cT1 breast cancer. Supplementary Table 2. Univariate and multivariate analysis with disease-free survival for cT1 breast cancer with no axillary lymph node metastasis pathologically. Supplementary Table 3. Univariate and multivariate analysis with recurrence-free survival for cT1 breast cancer with no axillary lymph node metastasis pathologically. Supplementary Table 4. Univariate and multivariate analysis with overall survival for cT1 breast cancer with no axillary lymph node metastasis pathologically.
## References
1. Coyle C, Cafferty FH, Vale C, Langley RE. **Metformin as an adjuvant treatment for cancer: a systematic review and meta-analysis**. *Ann Oncol* (2016.0) **27** 2184-2195. DOI: 10.1093/annonc/mdw410
2. Park YM, Bookwalter DB, O'Brien KM, Jackson CL, Weinberg CR, Sandler DP. **A prospective study of type 2 diabetes, metformin use, and risk of breast cancer**. *Ann Oncol* (2021.0) **32** 351-359. DOI: 10.1016/j.annonc.2020.12.008
3. Ni H, Rui Q, Zhu X, Yu Z, Gao R, Liu H. **Antihypertensive drug use and breast cancer risk: a meta-analysis of observational studies**. *Oncotarget* (2017.0) **8** 62545-62560. DOI: 10.18632/oncotarget.19117
4. Chae YK, Valsecchi ME, Kim J, Bianchi AL, Khemasuwan D, Desai A, Tester W. **Reduced risk of breast cancer recurrence in patients using ACE inhibitors, ARBs, and/or statins**. *Cancer Investig* (2011.0) **29** 585-593. DOI: 10.3109/07357907.2011.616252
5. Ahern TP, Pedersen L, Tarp M, Cronin-Fenton DP, Garne JP, Silliman RA, Sorensen HT, Lash TL. **Statin prescriptions and breast cancer recurrence risk: a Danish nationwide prospective cohort study**. *J Natl Cancer Inst* (2011.0) **103** 1461-1468. DOI: 10.1093/jnci/djr291
6. Cauley JA, McTiernan A, Rodabough RJ, LaCroix A, Bauer DC, Margolis KL, Paskett ED, Vitolins MZ, Furberg CD, Chlebowski RT. **Statin use and breast cancer: prospective results from the Women's Health Initiative**. *J Natl Cancer Inst* (2006.0) **98** 700-707. DOI: 10.1093/jnci/djj188
7. Kwan ML, Habel LA, Flick ED, Quesenberry CP, Caan B. **Post-diagnosis statin use and breast cancer recurrence in a prospective cohort study of early stage breast cancer survivors**. *Breast Cancer Res Treat* (2008.0) **109** 573-579. DOI: 10.1007/s10549-007-9683-8
8. Manthravadi S, Shrestha A, Madhusudhana S. **Impact of statin use on cancer recurrence and mortality in breast cancer: A systematic review and meta-analysis**. *Int J Cancer* (2016.0) **139** 1281-1288. DOI: 10.1002/ijc.30185
9. Cardwell CR, Hicks BM, Hughes C, Murray LJ. **Statin use after diagnosis of breast cancer and survival: a population-based cohort study**. *Epidemiology* (2015.0) **26** 68-78. DOI: 10.1097/EDE.0000000000000189
10. Murtola TJ, Visvanathan K, Artama M, Vainio H, Pukkala E. **Statin use and breast cancer survival: a nationwide cohort study from Finland**. *PLoS One* (2014.0) **9** e110231. DOI: 10.1371/journal.pone.0110231
11. Zhong S, Zhang X, Chen L, Ma T, Tang J, Zhao J. **Statin use and mortality in cancer patients: Systematic review and meta-analysis of observational studies**. *Cancer Treat Rev* (2015.0) **41** 554-567. DOI: 10.1016/j.ctrv.2015.04.005
12. Harborg S, Heide-Jorgensen U, Ahern TP, Ewertz M, Cronin-Fenton D, Borgquist S. **Statin use and breast cancer recurrence in postmenopausal women treated with adjuvant aromatase inhibitors: a Danish population-based cohort study**. *Breast Cancer Res Treat* (2020.0) **183** 153-160. DOI: 10.1007/s10549-020-05749-5
13. Demierre MF, Higgins PD, Gruber SB, Hawk E, Lippman SM. **Statins and cancer prevention**. *Nat Rev Cancer* (2005.0) **5** 930-942. DOI: 10.1038/nrc1751
14. Wachtershauser A, Akoglu B, Stein J. **HMG-CoA reductase inhibitor mevastatin enhances the growth inhibitory effect of butyrate in the colorectal carcinoma cell line Caco-2**. *Carcinogenesis* (2001.0) **22** 1061-1067. DOI: 10.1093/carcin/22.7.1061
15. Agarwal B, Halmos B, Feoktistov AS, Protiva P, Ramey WG, Chen M, Pothoulakis C, Lamont JT, Holt PR. **Mechanism of lovastatin-induced apoptosis in intestinal epithelial cells**. *Carcinogenesis* (2002.0) **23** 521-528. DOI: 10.1093/carcin/23.3.521
16. Spampanato C, De Maria S, Sarnataro M, Giordano E, Zanfardino M, Baiano S, Carteni M, Morelli F. **Simvastatin inhibits cancer cell growth by inducing apoptosis correlated to activation of Bax and down-regulation of BCL-2 gene expression**. *Int J Oncol* (2012.0) **40** 935-941. DOI: 10.3892/ijo.2011.1273
17. Cho SJ, Kim JS, Kim JM, Lee JY, Jung HC, Song IS. **Simvastatin induces apoptosis in human colon cancer cells and in tumor xenografts, and attenuates colitis-associated colon cancer in mice**. *Int J Cancer* (2008.0) **123** 951-957. DOI: 10.1002/ijc.23593
18. Goldstein JL, Brown MS. **Regulation of the mevalonate pathway**. *Nature* (1990.0) **343** 425-430. DOI: 10.1038/343425a0
19. Liao JK. **Isoprenoids as mediators of the biological effects of statins**. *J Clin Invest* (2002.0) **110** 285-288. DOI: 10.1172/JCI16421
20. Holstein SA, Wohlford-Lenane CL, Hohl RJ. **Isoprenoids influence expression of Ras and Ras-related proteins**. *Biochemistry* (2002.0) **41** 13698-13704. DOI: 10.1021/bi026251x
21. Duncan RE, El-Sohemy A, Archer MC. **Statins and cancer development**. *Cancer Epidemiol Biomark Prev* (2005.0) **14** 1897-1898. DOI: 10.1158/1055-9965.EPI-05-0027
22. Seasholtz TM, Majumdar M, Brown JH. **Rho as a mediator of G protein-coupled receptor signaling**. *Mol Pharmacol* (1999.0) **55** 949-956. DOI: 10.1124/mol.55.6.949
23. Denoyelle C, Vasse M, Korner M, Mishal Z, Ganne F, Vannier JP, Soria J, Soria C. **Cerivastatin, an inhibitor of HMG-CoA reductase, inhibits the signaling pathways involved in the invasiveness and metastatic properties of highly invasive breast cancer cell lines: an in vitro study**. *Carcinogenesis* (2001.0) **22** 1139-1148. DOI: 10.1093/carcin/22.8.1139
24. Beckwitt CH, Clark AM, Ma B, Whaley D, Oltvai ZN, Wells A. **Statins attenuate outgrowth of breast cancer metastases**. *Br J Cancer* (2018.0) **119** 1094-1105. DOI: 10.1038/s41416-018-0267-7
25. Beckwitt CH, Brufsky A, Oltvai ZN, Wells A. **Statin drugs to reduce breast cancer recurrence and mortality**. *Breast Cancer Res* (2018.0) **20** 144. DOI: 10.1186/s13058-018-1066-z
26. Mandal CC, Ghosh-Choudhury N, Yoneda T, Choudhury GG, Ghosh-Choudhury N. **Simvastatin prevents skeletal metastasis of breast cancer by an antagonistic interplay between p53 and CD44**. *J Biol Chem* (2011.0) **286** 11314-11327. DOI: 10.1074/jbc.M110.193714
27. Sendur MA, Aksoy S, Yazici O, Ozdemir NY, Zengin N, Altundag K. **Statin use may improve clinicopathological characteristics and recurrence risk of invasive breast cancer**. *Med Oncol* (2014.0) **31** 835. DOI: 10.1007/s12032-013-0835-x
28. Kumar AS, Benz CC, Shim V, Minami CA, Moore DH, Esserman LJ. **Estrogen receptor-negative breast cancer is less likely to arise among lipophilic statin users**. *Cancer Epidemiol Biomark Prev* (2008.0) **17** 1028-1033. DOI: 10.1158/1055-9965.EPI-07-0726
29. Desai P, Lehman A, Chlebowski RT, Kwan ML, Arun M, Manson JE, Lavasani S, Wasswertheil-Smoller S, Sarto GE, LeBoff M, Cauley J, Cote M, Beebe-Dimmer J, Jay A, Simon MS. **Statins and breast cancer stage and mortality in the Women’s Health Initiative**. *Cancer Causes Control* (2015.0) **26** 529-539. DOI: 10.1007/s10552-015-0530-7
30. Xu WH, Zhou YH. **The relationship between post-diagnostic statin usage and breast cancer prognosis varies by hormone receptor phenotype: a systemic review and meta-analysis**. *Arch Gynecol Obstet* (2021.0) **304** 1315-1321. DOI: 10.1007/s00404-021-06065-z
31. White CM. **A review of the pharmacologic and pharmacokinetic aspects of rosuvastatin**. *J Clin Pharmacol* (2002.0) **42** 963-970. DOI: 10.1177/009127000204200902
32. Corsini A, Bellosta S, Baetta R, Fumagalli R, Paoletti R, Bernini F. **New insights into the pharmacodynamic and pharmacokinetic properties of statins**. *Pharmacol Ther* (1999.0) **84** 413-428. DOI: 10.1016/s0163-7258(99)00045-5
33. McMasters KM, Tuttle TM, Carlson DJ, Brown CM, Noyes RD, Glaser RL, Vennekotter DJ, Turk PS, Tate PS, Sardi A, Cerrito PB, Edwards MJ. **Sentinel lymph node biopsy for breast cancer: a suitable alternative to routine axillary dissection in multi-institutional practice when optimal technique is used**. *J Clin Oncol* (2000.0) **18** 2560-2566. DOI: 10.1200/JCO.2000.18.13.2560
34. Kashiwagi S, Onoda N, Asano Y, Kurata K, Noda S, Kawajiri H, Takashima T, Ohsawa M, Kitagawa S, Hirakawa K. **Ambulatory sentinel lymph node biopsy preceding neoadjuvant therapy in patients with operable breast cancer: a preliminary study**. *World J Surg Oncol* (2015.0) **13** 53. DOI: 10.1186/s12957-015-0471-3
35. Lee A, Krishnamurthy S, Sahin A, Symmans WF, Hunt K, Sneige N. **Intraoperative touch imprint of sentinel lymph nodes in breast carcinoma patients**. *Cancer* (2002.0) **96** 225-231. DOI: 10.1002/cncr.10721
36. Khanna R, Bhadani S, Khanna S, Pandey M, Kumar M. **Touch imprint cytology evaluation of sentinel lymph node in breast cancer**. *World J Surg* (2011.0) **35** 1254-1259. DOI: 10.1007/s00268-011-1094-7
37. Houvenaeghel G, Nos C, Mignotte H, Classe JM, Giard S, Rouanet P, Lorca FP, Jacquemier J, Bardou VJ. **Micrometastases in sentinel lymph node in a multicentric study: predictive factors of nonsentinel lymph node involvement--Groupe des Chirurgiens de la Federation des Centres de Lutte Contre le Cancer**. *J Clin Oncol* (2006.0) **24** 1814-1822. DOI: 10.1200/JCO.2005.03.3225
38. Zhao YX, Liu YR, Xie S, Jiang YZ, Shao ZM. **A Nomogram Predicting Lymph Node Metastasis in T1 Breast Cancer based on the Surveillance, Epidemiology, and End Results Program**. *J Cancer* (2019.0) **10** 2443-2449. DOI: 10.7150/jca.30386
39. Capdet J, Martel P, Charitansky H, Lim YK, Ferron G, Battle L, Landier A, Mery E, Zerdoub S, Roche H, Querleu D. **Factors predicting the sentinel node metastases in T1 breast cancer tumor: an analysis of 1416 cases**. *Eur J Surg Oncol* (2009.0) **35** 1245-1249. DOI: 10.1016/j.ejso.2009.06.002
40. Viale G, Zurrida S, Maiorano E, Mazzarol G, Pruneri G, Paganelli G, Maisonneuve P, Veronesi U. **Predicting the status of axillary sentinel lymph nodes in 4351 patients with invasive breast carcinoma treated in a single institution**. *Cancer* (2005.0) **103** 492-500. DOI: 10.1002/cncr.20809
41. Reyal F, Rouzier R, Depont-Hazelzet B, Bollet MA, Pierga JY, Alran S, Salmon RJ, Fourchotte V, Vincent-Salomon A, Sastre-Garau X, Antoine M, Uzan S, Sigal-Zafrani B, De Rycke Y. **The molecular subtype classification is a determinant of sentinel node positivity in early breast carcinoma**. *PLoS One* (2011.0) **6** e20297. DOI: 10.1371/journal.pone.0020297
42. Zhang Y, Li J, Fan Y, Li X, Qiu J, Zhu M, Li H. **Risk factors for axillary lymph node metastases in clinical stage T1-2N0M0 breast cancer patients**. *Medicine (Baltimore)* (2019.0) **98** e17481. DOI: 10.1097/MD.0000000000017481
43. Ding J, Jiang L, Wu W. **Predictive Value of Clinicopathological Characteristics for Sentinel Lymph Node Metastasis in Early Breast Cancer**. *Med Sci Monit* (2017.0) **23** 4102-4108. DOI: 10.12659/msm.902795
44. Yoshihara E, Smeets A, Laenen A, Reynders A, Soens J, Van Ongeval C, Moerman P, Paridaens R, Wildiers H, Neven P, Christiaens MR. **Predictors of axillary lymph node metastases in early breast cancer and their applicability in clinical practice**. *Breast* (2013.0) **22** 357-361. DOI: 10.1016/j.breast.2012.09.003
45. Klar M, Foeldi M, Markert S, Gitsch G, Stickeler E, Watermann D. **Good prediction of the likelihood for sentinel lymph node metastasis by using the MSKCC nomogram in a German breast cancer population**. *Ann Surg Oncol* (2009.0) **16** 1136-1142. DOI: 10.1245/s10434-009-0399-3
46. Qiu PF, Liu JJ, Wang YS, Yang GR, Liu YB, Sun X, Wang CJ, Zhang ZP. **Risk factors for sentinel lymph node metastasis and validation study of the MSKCC nomogram in breast cancer patients**. *Jpn J Clin Oncol* (2012.0) **42** 1002-1007. DOI: 10.1093/jjco/hys150
47. Eliassen AH, Colditz GA, Rosner B, Willett WC, Hankinson SE. **Serum lipids, lipid-lowering drugs, and the risk of breast cancer**. *Arch Intern Med* (2005.0) **165** 2264-2271. DOI: 10.1001/archinte.165.19.2264
48. Boudreau DM, Yu O, Miglioretti DL, Buist DS, Heckbert SR, Daling JR. **Statin use and breast cancer risk in a large population-based setting**. *Cancer Epidemiol Biomark Prev* (2007.0) **16** 416-421. DOI: 10.1158/1055-9965.EPI-06-0737
49. Abdullah MI, de Wolf E, Jawad MJ, Richardson A. **The poor design of clinical trials of statins in oncology may explain their failure - Lessons for drug repurposing**. *Cancer Treat Rev* (2018.0) **69** 84-89. DOI: 10.1016/j.ctrv.2018.06.010
50. Liu B, Yi Z, Guan X, Zeng YX, Ma F. **The relationship between statins and breast cancer prognosis varies by statin type and exposure time: a meta-analysis**. *Breast Cancer Res Treat* (2017.0) **164** 1-11. DOI: 10.1007/s10549-017-4246-0
51. Rastogi P, Anderson SJ, Bear HD, Geyer CE, Kahlenberg MS, Robidoux A, Margolese RG, Hoehn JL, Vogel VG, Dakhil SR, Tamkus D, King KM, Pajon ER, Wright MJ, Robert J, Paik S, Mamounas EP, Wolmark N. **Preoperative chemotherapy: updates of National Surgical Adjuvant Breast and Bowel Project Protocols B-18 and B-27**. *J Clin Oncol* (2008.0) **26** 778-785. DOI: 10.1200/JCO.2007.15.0235
52. von Minckwitz G, Untch M, Blohmer JU, Costa SD, Eidtmann H, Fasching PA, Gerber B, Eiermann W, Hilfrich J, Huober J, Jackisch C, Kaufmann M, Konecny GE, Denkert C, Nekljudova V, Mehta K, Loibl S. **Definition and impact of pathologic complete response on prognosis after neoadjuvant chemotherapy in various intrinsic breast cancer subtypes**. *J Clin Oncol* (2012.0) **30** 1796-1804. DOI: 10.1200/JCO.2011.38.8595
53. Cortazar P, Zhang L, Untch M, Mehta K, Costantino JP, Wolmark N, Bonnefoi H, Cameron D, Gianni L, Valagussa P, Swain SM, Prowell T, Loibl S, Wickerham DL, Bogaerts J, Baselga J, Perou C, Blumenthal G, Blohmer J, Mamounas EP, Bergh J, Semiglazov V, Justice R, Eidtmann H, Paik S, Piccart M, Sridhara R, Fasching PA, Slaets L, Tang S, Gerber B, Geyer CE, Pazdur R, Ditsch N, Rastogi P, Eiermann W, von Minckwitz G. **Pathological complete response and long-term clinical benefit in breast cancer: the CTNeoBC pooled analysis**. *Lancet* (2014.0) **384** 164-172. DOI: 10.1016/S0140-6736(13)62422-8
54. 54.Bonnefoi H, Litiere S, Piccart M, MacGrogan G, Fumoleau P, Brain E, Petit T, Rouanet P, Jassem J, Moldovan C, Bodmer A, Zaman K, Cufer T, Campone M, Luporsi E, Malmstrom P, Werutsky G, Bogaerts J, Bergh J, Cameron DA, investigators EBS (2014) Pathological complete response after neoadjuvant chemotherapy is an independent predictive factor irrespective of simplified breast cancer intrinsic subtypes: a landmark and two-step approach analyses from the EORTC 10994/BIG 1-00 phase III trial. Ann Oncol 25 (6):1128–1136. 10.1093/annonc/mdu118.
55. Gentilini O, Botteri E, Dadda P, Sangalli C, Boccardo C, Peradze N, Ghisini R, Galimberti V, Veronesi P, Luini A, Cassano E, Viale G, Veronesi U. **Physical function of the upper limb after breast cancer surgery. Results from the SOUND (Sentinel node vs. Observation after axillary Ultra-souND) trial**. *Eur J Surg Oncol* (2016.0) **42** 685-689. DOI: 10.1016/j.ejso.2016.01.020
56. Reimer T, Stachs A, Nekljudova V, Loibl S, Hartmann S, Wolter K, Hildebrandt G, Gerber B. **Restricted Axillary Staging in Clinically and Sonographically Node-Negative Early Invasive Breast Cancer (c/iT1-2) in the Context of Breast Conserving Therapy: First Results Following Commencement of the Intergroup-Sentinel-Mamma (INSEMA) Trial**. *Geburtshilfe Frauenheilkd* (2017.0) **77** 149-157. DOI: 10.1055/s-0042-122853
|
---
title: Independent and interactive effect of type 2 diabetes and hypertension on memory
functions in middle aged adults
authors:
- Kinga Kálcza Jánosi
- Andrea Lukács
journal: BMC Endocrine Disorders
year: 2023
pmcid: PMC9999571
doi: 10.1186/s12902-023-01308-3
license: CC BY 4.0
---
# Independent and interactive effect of type 2 diabetes and hypertension on memory functions in middle aged adults
## Abstract
### Background
The study distinguishes the effect of type 2 diabetes and hypertension on cognitive functions when the two diseases are alone or when they occur together, compared to healthy individuals.
### Methods
A total of 143 middle-aged adults were screened using the Wechsler Memory Scale – Revised psychometric test (verbal memory, visual memory, attention/concentration and delayed memory). Participants were divided into four groups based on their diseases: patients with type 2 diabetes [36], patients with hypertension [30], patients having both diseases [33], and healthy controls [44].
### Results
This study found no differences among investigated groups in verbal and visual memory, however, hypertension and both-disease group performed unfavorably compared to patients with diabetes and to healthy individuals in attention/concentration and delayed memory.
### Conclusions
The findings of this study suggest that there is a relationship between hypertension and cognitive dysfunction, whereas type 2 diabetes without consequences was not proved to have an association with cognitive decline in middle-aged people.
## Background
Diabetes and hypertension are a worldwide epidemic, resulting in millions of deaths each year. These two diseases often occur together and exacerbate each other’s symptoms, and may share common pathways such as insulin resistance, inflammation, obesity and oxidative stress [1]. The coexistence of type 2 diabetes (T2D) and hypertension is increasing in the general population. Hypertension is disproportionately higher in patients with diabetes. Having one of the two conditions increases the risk of developing the other disease by 1.5–2.0 times [2]. The brain is one of the main target organs affected by high glucose level and hypertension, and both conditions might be associated with a wide variety of cognitive impairments. Especially type 2 diabetes and hypertension frequently coexist, and their combination may provide additive increases in the risk impairment of cognitive processes [3].
Cognitive disorders play an important role in diabetes and hypertension for two reasons: first, cognitive disorders are complications associated with the diseases and may indicate the patient has inadequate glucose metabolism and/or high blood pressure. Secondly, the cognitive dysfunctions have a major impact on the self-management of the disease and can influence the quality and progression of the diseases in the future. Awad’s et al. [ 4] proposed cognitive functions are more affected in patients with T2D, who have various associated complications such as hypertension. Other comorbid illnesses such as clusters of comorbidities that combine with hypertension may have a greater cognitive impact. Some studies suggest hypertension and diabetes, when combined, increase the risk of cognitive impairment [5, 6]. The main question, which has not yet been fully clarified, is whether having T2D combined with hypertension increases the risk of cognitive decline. Previous studies mainly evaluated older individuals with T2D [7], and knowledge regarding middle-aged people is scarce. To fill this gap, this study aims to answer the question of whether middle-aged individuals with independent T2D (without hypertension), independent hypertension (without T2D), or comorbid T2D and hypertension have some dysfunction in memory components (verbal memory, visual memory, attention and concentration, delayed memory) specific to the selected risk groups, compared to healthy individuals.
## Study design, ethics, and participants
A cross-sectional quantitative survey was performed in Harghita County (Transylvania, Romania). The patients were recruited from three randomly chosen settlements (Joseni, Ciumani and Lăzarea) with the help of general practitioners of these villages. All of the patients were informed about the purpose of the study and the voluntary nature of the participation. Participants gave written informed consent before taking part in the study. The protocol of the study was approved by the Ethics Committee of the University of Babeș-Bolyai (RO) and by the Code of Deontology for the professions of psychologists, elaborated by the Romanian College of Psychologists. Cognitive functions were assessed independently for each participant by accredited clinical psychologists. Inclusion criteria were: 1) ages between 35 and 65 years [8, 9], 2) diagnosed with T2D or hypertension or both (comorbid T2D and hypertension) according to the standards of the American Diabetes Association [2020] [10] and according to the standards of the American College of Cardiology and American Heart Association guidelines [11], respectively, and 3) diabetes and/or hypertension duration at least 5 years. From this study, patients were excluded with 1) any medical illnesses other than T2D, dyslipidemia, hypertension and obesity, 2) a history of hypoglycemic coma or complications of diabetes, 3) primary neurological condition as history of transient ischemic attacks, cerebrovascular stroke or epilepsy or psychiatric disease, 4) previous serious head injury, 5) any sensory or motor disorder that would preclude psychological testing (including blindness), 6) regular treatment with any medications known to have psychoactive effect, and 7) drug or alcohol abuse. The control group was made up of people who applied at the invitation of the municipality's management and did not have a chronic illness. Their health status was checked at the health clinic of the municipality.
The patients were divided into three groups: a group with T2D, a group with hypertension and a group with both diseases. The control group was specifically recruited from the settlement where the patients were registered.
## Demographics
Patients provided data about their gender, age, and education duration in years, as well as the duration of the disease.
## Clinical parameters
Blood pressure was measured on the left arm in sitting posture during the visit by the physician using an aneroid sphygmomanometer with a stethoscope.
Glycemic control was explained by HbA1c (the gold standard measurement of glycemic control) [10]. The last three-month HbA1c value was extracted from the medical records of patients with diabetes, or was measured from a vein during the routine visit.
## Cognitive measures
Cognitive functions were assessed using the Wechsler Memory Scale – Revised (WMS-R) [12]. The test was administered during a routine visit to the general practitioner and it required around 1 h to complete. The WMS-R is a neuropsychological test designed to measure different memory functions such as verbal, visual memory, attention/concentration and delayed memory. The psychometric characteristics of the WMS-R are evaluated and related to its clinical utility in the Romanian population [13, 14]. The weighted scores (weighted raw score summaries) were calculated according to the scoring system from in the WMS-R administrative and scoring manual, with a higher score indicating better functioning.
## Data analysis
For data analysis SPSS (Statistical Package for the Social Sciences) version 25.0 was used (SPSS, Version 25, IBM Corporation, Armonk, NY). All p-values were two-tailed at the significance level of 0.05. In the first stage, the Z-score between -3 and 3 method of outlier detection was performed and removed from the database. Violations of the normality assumption were checked using Shapiro–Wilk’s test. As a result, the memory variables were transformed by square-root transformation (moderately positively and negatively skewed data). The differences between investigated groups on demographic and clinical information were tested with analysis of variance (one-way ANOVA) for continuous outcomes and Pearson’s chi-square test for dichotomous outcomes. Comparison of memory scores between groups was measured using multiple analysis of covariance (MANCOVA) with follow-up univariate ANOVAs and the Bonferroni post-hoc test. Age, sex and education duration were controlled, age and education as covariates and also gender as a factor. We checked the assumptions that are required for MANCOVA, assumption of independence, assumption of normality, assumption of homogeneity of variance and assumption of absence of multicollinearity.
## Power and sample size
A post hoc power analysis was conducted using the software package GPower. The recommended effect sizes used for this assessment were as follows: small (f 2 = 0.02), medium (f 2 = 0.15), and large (f 2 = 0.35) [15]. The alpha level used for this analysis was p ≤ 0.05. The power of the study was high to detect the main outcome of interest. The post hoc analysis revealed that the statistical power for this study was $99\%$ (0.99) for detecting medium effect (0.12) for statistical comparisons (MANCOVA, $$n = 143$$, 4 groups, 4 response variables).
## Patients and clinical variables
The initial group included 157 individuals but identifying outliers eventually resulted in 143 eligible persons. The control group consisted of 44 healthy adults. Participants were homogeneous by age, sex and education duration: there were no significant differences among the four investigated groups in age, gender or education duration, nor in clinical characteristics such as diabetes duration and glycemic control between groups with T2D and presenting both diseases. No significant differences were observed in duration of hypertension or systolic/diastolic blood pressure between the groups with hypertension and with both diseases. Characteristics of the participants stratified by disease are displayed in Table 1.Table 1Characteristics of the participants stratified by disease ($$n = 143$$)Mean, Standard Deviation (SD)Healthy group($$n = 44$$)Group with T2D ($$n = 36$$)Group with HTN ($$n = 30$$)Group with both diseases ($$n = 33$$)F(df)X2(df)pAge (years)46.40 (4.09)45.91 (5.16)45.83 (4.83)46.81 (5.48)0.295[3,139]-0.829Male sex, n (%)23 ($52.3\%$)17 ($47.2\%$)14 ($46.7\%$)17 ($51.1\%$)-.359[3]0.949Education (years)12.72 (1.70)12.02 (1.69)12.66 (1.98)12.36 (2.26)1.05[3,139]-0.371HbA1c-7.46 (0.54)-7.61 (0.55)1.35[1,67]-0.248Diabetes duration (years)-7.66 (2.44)-8.54 (2.57)2.11[1,67]-0.151Treatment regime (%) (frequency)-----Insulin11 [4]17 [6]OGLM57 [21]61 [20]Diet32 [11]22 [7]Antihypertensive med (%) (frequency)--34 [10]48 [16]---Hypertension duration (years)--8.96 (2.41)9.81 (2.93)1.56[1,61]-0.134SBP 24 h (mmHg)--144.47 (8.11)147.12 (7.41)1.84[1,61]-0.180DBP 24 h (mmHg)--90.73 (5.08)91.39 (4.19).319[1,61]-0.574T2D Type 2 diabetes, HTN Hypertension, HbA1c Glycosylated hemoglobin, OGLM Oral glucose-lowering medication, SBP Systolic blood pressure, DBP Diastolic blood pressure
## Neuropsychological functions (verbal memory, visual memory, attention and concentration, delayed memory)
There were statistically significant differences in memory functions based on disease, F[12, 344] = 4.240, p ≤ 0.01; Wilk's Λ = 0.694, partial η2 = 0.114, approximately $11.4\%$ of multivariate variance of the dependent variables is associated with the group factor (MANCOVA) after controlling for age (F[4, 130] = 1.220, $p \leq 0.05$), education duration (F[4, 130] = 5.914, p ≤ 0.01) and sex (F[4, 130] = 0.052, $p \leq 0.05$).
Follow-up univariate ANOVAs indicate significant differences between the groups in attention and concentration (F[3, 133] = 4.026, p ≤ 0.01; partial η2 = 0.083) and delayed memory (F[3, 133] = 5.665, p ≤ 0.01; partial η2 = 0.113), meaning that the disease has a statistically significant effect on these cognitive functions. Mean scores of attention and concentration were significantly different between the control group ($M = 8.79$, SD = 0.27) and group with hypertension ($M = 8.62$, SD = 0.19; p ≤ 0.05), and the status of the control group was better than that of the other two groups. For delayed memory, the control group ($M = 9.01$, SD = 0.39) showed a significantly better status than the group with hypertension ($M = 8.75$, SD = 0.30; p ≤ 0.05), while the group with T2D ($M = 9.02$, SD = 0.37) had a better status than the group with hypertension ($M = 8.75$, SD = 0.30; p ≤ 0.05) and the group with both diseases ($M = 8.79$, SD = 0.38; p ≤ 0.05). ( Table 2) Raw scores of the neuropsychological functions with mean and SD are displayed in Fig. 1.Table 2Neuropsychological functions with participants stratified by disease ($$n = 143$$)Mean (SD)memoryHealthy group($$n = 44$$)Group with T2D ($$n = 36$$)Group with HTN ($$n = 30$$)Group with both diseases ($$n = 33$$)F(df)ppartial η2Verbal9.13 (.40)8.89 (.47)8.97 (.44)8.86 (.47)2.515(3.133)0.0610.054Visual7.68 (.23)7.62 (.25)7.63 (.23)7.64 (.30).331(3.133)0.8030.007Attention and concentration8.79 (.27)8.75 (.28)8.62 (.19)8.63 (.28)4.026(3.133)0.0090.083Delayed9.01 (.39)9.02 (.37)8.75 (.30)8.79 (.38)5.665(3.133)0.0010.113T2D Type 2 diabetes, HTN HypertensionFig. 1Raw scores with mean and SD of the neuropsychological functions ($$n = 143$$). T2D—type 2 diabetes, HTN—hypertension, *p ≤ 0.05
## Discussion
The present study investigated whether there are any differences in memory functioning among adults with independent T2D (without hypertension), independent hypertension (without T2D), with comorbid hypertension and T2D, and healthy individuals. The literature is not clear on whether diabetes affects premature cognitive decline [16–19]. In our study, we found no differences between groups in the capacity to remember what the individual had previously learned (read or heard) and recall when needed (verbal memory). Similarly, no significant differences were found in visual memory between the investigated groups. Individuals had similar capacity to remember or recall information that had been previously viewed. The findings of our study cannot confirm that cognitive functions such as verbal memory, visual memory, attention and concentration, as well as delayed memory have a relationship with the presence of T2D, however, for the latter two, differences were found among groups. Groups with hypertension and with both diseases displayed unfavorable results compared to the group with diabetes and the healthy group in the functions of attention and concentration and in delayed memory. Moran et al. [ 20] also failed to prove a direct effect of T2D on cognitive function. Demakakos et al. [ 21] examined the combined effect of diabetes and elevated depressive symptoms and found they significantly accelerate cognitive decline over time, especially in people aged 50–64, but do not accelerate cognitive decline separately. The results of our study suggest hypertension has a negative impact on cognitive functions, and co-occurrence with diabetes does not exacerbate this. Muela et al. [ 22] evaluated patients with hypertension compared to healthy controls and found poorer cognitive performance in almost all cognitive tests. Only a few studies have explicitly studied the cognitive impacts of comorbid hypertension and T2D, reporting that the two conditions may interact to increase the risk of cognitive impairment [23, 24] or may not [25, 26]. We found no significant differences in cognitive components between the people with diabetes and the healthy adults, which indicate that cognitive deficits appear more pronounced in individuals who have hypertension, with or without T2D. It should be mentioned that we cannot unequivocally determine the direction of the association between cognitive function and hypertension and T2D. The relationship between cognitive changes and health outcomes is bidirectional. Medical aspects of hypertension and T2D can negatively affect cognitive functioning, while deterioration of cognitive functioning can negatively influence the self-management of disease. Those who are cognitively impaired have poorer command of the skills necessary to manage their chronic diseases. It is worth noting that our middle-aged patients with T2D had no vascular complications that might have resulted in cognitive impairment. Both T2D and hypertension are chronic psychosomatic diseases and the resulting complications can affect the entire body, requiring specific attention to both the cognitive and the somatic components of the diseases [27, 28].
Limitations of the study include a relatively small proportion of participants from one county which precludes generalizability. The cross-sectional design does not allow us to infer causality. Smoking status has not been investigated, which could help to understand its weight in the genesis of the change in memory. Dyslipidaemia was not assessed, and could play a possible role in the impaired memory pattern.
## Conclusion
Our results indicate that there is an association between hypertension and cognitive impairments in middle-aged patients with T2D, whereas this association is not proved between complication-free T2D and cognitive function. Regarding verbal and visual memory, no significant differences were found among groups, so it seems likely that these memory dimensions are not influenced by the investigated diseases. Regarding attention, concentration and delayed memory, the cognitive deficits are pronounced when hypertension is present. There is no evidence that co-occurrence of T2D amplifies the effect of hypertension on cognitive decline.
## References
1. Cheung BM, Li C. **Diabetes and hypertension: is there a common metabolic pathway?**. *Curr Atheroscler Rep* (2012.0) **14** 160-166. DOI: 10.1007/s11883-012-0227-2
2. Colussi G, Da Porto A, Cavarape A. **Hypertension and type 2 diabetes: lights and shadows about causality**. *J Hum Hypertens* (2020.0) **34** 91-93. DOI: 10.1038/s41371-019-0268-x
3. Tchistiakova E, Anderson ND, Greenwood CE, Macintosh BJ. **Combined effects of type 2 diabetes and hypertension associated with cortical thinning and impaired cerebrovascular reactivity relative to hypertension alone in older adults**. *NeuroImage Clin* (2014.0) **5** 36-41. DOI: 10.1016/j.nicl.2014.05.020
4. Awad N, Gagnon M, Messier C. **The relationship between impaired glucose tolerance, type 2 diabetes, and cognitive function**. *J Clin Exp Neuropsychol* (2004.0) **26** 1044-1080. DOI: 10.1080/13803390490514875
5. Hassing LB, Hofer SM, Nilsson SE, Berg S, Pedersen NL, McClearn G, Johansson B. **Comorbid type 2 diabetes mellitus and hypertension exacerbates cognitive decline: evidence from a longitudinal study**. *Age Ageing* (2004.0) **33** 355-361. DOI: 10.1093/ageing/afh100
6. Kelly A, Calamia M, Koval A, Terrera GM, Piccinin AM, Clouston S, Hassing LB, Bennett DA, Johansson B, Hofer SM. **Independent and interactive impacts of hypertension and diabetes mellitus on verbal memory: A coordinated analysis of longitudinal data from England, Sweden, and the United States**. *Psychol Aging* (2016.0) **31** 262-273. DOI: 10.1037/pag0000078
7. Bangen KJ, Werhane ML, Weigand AJ, Edmonds EC, Delano-Wood L, Thomas KR, Nation DA, Evangelista ND, Clark AL, Liu TT, Bondi MW. **Reduced Regional Cerebral Blood Flow Relates to Poorer Cognition in Older Adults With Type 2 Diabetes**. *Front Aging Neurosci* (2018.0) **2018** 270. DOI: 10.3389/fnagi.2018.00270
8. Lachman ME. **Development in midlife**. *Annu Rev Psychol* (2004.0) **2004** 305-331. DOI: 10.1146/annurev.psych.55.090902.14152143
9. Cohen P. *In Our Prime: The Invention of Middle Age* (2012.0)
10. **Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes-2021**. *Diabetes Care* (2020.0) **43** S14-S31. DOI: 10.2337/dc21-S002
11. 11.Whelton PK, Carey RM, Aronow WS, et al. 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 [published correction appears in Hypertension. 2018 Jun;71(6):e136-e139] [published correction appears in Hypertension. 2018 Sep;72(3):e33]. Hypertension. 2018;71(6):1269-1324. 10.1161/HYP.0000000000000066
12. Wechsler D. *Wechsler Memory Scale-Revised Manual* (1987.0)
13. Kálcza Jánosi K. **[Interpretation and diagnostic value of WMS-R (Wechsler Memory Scale-Revised) in people with type 1 and type 2 diabetes]**. *Transylv J Psychol* (2010.0) **3** 275-301
14. Kálcza Jánosi K, Pordea A. **Glucose metabolism and memory functions in adults with type 2 diabetes**. *Rom J Exp Appl Psychol* (2016.0) **7** 146-149. DOI: 10.15303/rjeap.2016.si1.a30
15. Cohen J. *Statistical power analysis for the behavioral sciences* (1977.0)
16. Cukierman-Yaffe T, Gerstein HC, Williamson JD, Lazar RM, Lovato L, Miller ME, Coker LH, Murray A, Sullivan MD, Marcovina SM, Launer LJ. **Action to Control Cardiovascular Risk in Diabetes-Memory in Diabetes (ACCORD-MIND) Investigators (2009). Relationship between baseline glycemic control and cognitive function in individuals with type 2 diabetes and other cardiovascular risk factors: the action to control cardiovascular risk in diabetes-memory in diabetes (ACCORD-MIND) trial**. *Diabetes Care* (2009.0) **32** 221-226. DOI: 10.2337/dc08-1153
17. van den Berg E, Reijmer YD, de Bresser J, Kessels RP, Kappelle LJ, Biessels GJ. **Utrecht Diabetic Encephalopathy Study Group. A 4 year follow-up study of cognitive functioning in patients with type 2 diabetes mellitus**. *Diabetologia* (2010.0) **53** 58-65. DOI: 10.1007/s00125-009-1571-9
18. Kumari M, Marmot M. **Diabetes and cognitive function in a middle-aged cohort: findings from the Whitehall II study**. *Neurology* (2005.0) **65** 1597-1603. DOI: 10.1212/01.wnl.0000184521.80820.e4
19. Nooyens AC, Baan CA, Spijkerman AM, Verschuren WM. **Type 2 diabetes and cognitive decline in middle-aged men and women: the Doetinchem Cohort Study**. *Diabetes Care* (2010.0) **33** 1964-1969. DOI: 10.2337/dc09-2038
20. Moran C, Beare R, Wang W, Callisaya M, Srikanth V. **Alzheimer's Disease Neuroimaging Initiative (ADNI). Type 2 diabetes mellitus, brain atrophy, and cognitive decline**. *Neurology* (2019.0) **92** e823-e830. DOI: 10.1212/WNL.0000000000006955
21. Demakakos P, Muniz-Terrera G, Nouwen A. **Type 2 diabetes, depressive symptoms and trajectories of cognitive decline in a national sample of community-dwellers: A prospective cohort study**. *PLoS ONE* (2017.0) **12** e0175827. DOI: 10.1371/journal.pone.0175827
22. Muela HC, Costa-Hong VA, Yassuda MS, Moraes NC, Memória CM, Machado MF, Macedo TA, Shu EB, Massaro AR, Nitrini R, Mansur AJ, Bortolotto LA. **Hypertension Severity Is Associated With Impaired Cognitive Performance**. *J Am Heart Assoc* (2017.0) **6** e004579. DOI: 10.1161/JAHA.116.004579
23. Petrova M, Prokopenko S, Pronina E, Mozheyko E. **Diabetes type 2, hypertension and cognitive dysfunction in middle age women**. *J Neurol Sci* (2010.0) **299** 39-41. DOI: 10.1016/j.jns.2010.08.057
24. Mount DL. **Hypertension associated with neurocognitive performance among persons with type 2 diabetes: a brief report**. *Ment Health Fam Med* (2011.0) **8** 243-247. PMID: 23205065
25. Cahana-Amitay D, Albert ML, Ojo EA, Sayers J, Goral M, Obler LK, Spiro A. **Effects of hypertension and diabetes on sentence comprehension in aging**. *J Gerontol B Psychol Sci Soc Sci* (2013.0) **68** 513-521. DOI: 10.1093/geronb/gbs085
26. Gregg EW, Yaffe K, Cauley JA, Rolka DB, Blackwell TL, Narayan KM, Cummings SR. **Is diabetes associated with cognitive impairment and cognitive decline among older women? Study of Osteoporotic Fractures Research Group**. *Arch Intern Med* (2000.0) **160** 174-180. DOI: 10.1001/archinte.160.2.174
27. Wasserman LI, Trifonova EA. **Diabetes mellitus as a model of psychosomatic and somatopsychic interrelationships**. *Span J Psychol* (2006.0) **9** 75-85. DOI: 10.1017/s1138741600005990
28. Murray E, Parati G. **Is essential hypertension sometimes a psychosomatic disorder?**. *J Hypertens* (2004.0) **22** 873-876. DOI: 10.1097/00004872-200405000-00003
|
---
title: The influence of socioeconomic aspects and hospital case volume on survival
in colorectal cancer in Saxony, Germany
authors:
- Andreas Bogner
- Jürgen Weitz
- Daniela Piontek
journal: BMC Cancer
year: 2023
pmcid: PMC9999591
doi: 10.1186/s12885-023-10672-1
license: CC BY 4.0
---
# The influence of socioeconomic aspects and hospital case volume on survival in colorectal cancer in Saxony, Germany
## Abstract
### Background
Colorectal cancer (CRC) is one of the most common types of cancer in Western civilization and responsible for a high number of yearly deaths. Long-term outcome is influenced by many factors, potentially including socioeconomic aspects like income, education, and employment. Furthermore, annual surgical case volume plays a major role in achieving good oncological results. In our retrospective study, we evaluated the effect of socioeconomic deprivation and hospital volume on overall survival (OS) in the federal state of Saxony, Germany.
### Methods
All patients with CRC who underwent surgery in Saxony, Germany between 2010 and 2020 and were living in Saxony at the time of diagnosis were included in our retrospective analysis. Uni- and multivariate analyses were conducted considering age, sex, tumor localization, UICC tumor stage, surgical approach (open/laparoscopic), number of resected lymph nodes, adjuvant chemotherapy, year of surgery, and hospital case volume. In addition, our model was adjusted for social disparity using the German Index of Socioeconomic Deprivation (GISD).
### Results
A total of 24,085 patients were analyzed (15,883 with colon cancer and 8,202 with rectal cancer). Age, sex, UICC tumor stage and tumor localization were distributed as expected for CRC. Median overall survival time was 87.9 months for colon cancer and 110.0 months for rectal cancer. Univariate analysis revealed laparoscopic surgery (colon and rectum $P \leq 0.001$), high case volume (rectum: $$P \leq 0.002$$) and low levels of socioeconomic deprivation (colon and rectum $P \leq 0.001$) to be significantly associated with better survival. In multivariate analyses, the associations of laparoscopic surgery (colon: HR = 0.76, $P \leq 0.001$; rectum: HR = 0.87, $P \leq 0.01$), and mid-low to mid-high socioeconomic deprivation (colon: HR = 1.18–1.22, $P \leq 0.001$; rectum: HR = 1.18–1.36, $P \leq 0.001$–0.01) remained statistically significant. Higher hospital case volume was associated with better survival only in rectal cancer (HR = 0.89; $P \leq 0.01$).
### Conclusion
In Saxony, Germany, better long-term survival after CRC surgery was associated with low socioeconomic deprivation, laparoscopic surgery and partly with high hospital case volume. Thus, there is a need to reduce social differences in access to high-quality treatment and prevention and increase hospital patient volume.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12885-023-10672-1.
## Background
Colorectal cancer (CRC) is the third most common cancer worldwide and the second leading cause of cancer-related deaths every year [1]. In Germany, CRC accounts for approximately 60,000 newly diagnosed tumors and around 25,000 tumor-related deaths annually. Due to the implementation of health programs offering colonoscopy as a widely available examination for early detection and screening, the incidence and mortality of CRC is only slowly declining [2]. Many diseases, including cancer, follow a socioeconomic gradient with higher rates in patients with lower socioeconomic status [3, 4]. In Germany, an aging society is being confronted with the risk of increasing social disparities. In view of the structure of the German healthcare system, which grants free access to healthcare for everyone, as well as several campaigns in the media, cancer therapy and outcome should theoretically be equally distributed. Since individual-level socioeconomic data are difficult to obtain, especially in large cohorts, the German Index of Socioeconomic Deprivation (GISD) can be used as an external indicator. It considers the dimensions of education, employment and income, with each contributing one-third to the overall score [5, 6]. Jansen et al. showed a worse relative survival of patients living in the most deprived districts compared to the survival of patients living in more privileged areas throughout Germany, regarding the 25 most common cancer sites in ten population-based registries (covering 32 million inhabitants).[7].
However, socioeconomic disparity is not the only factor influencing survival in CRC. High individual surgeon and institutional case volume is a well-documented factor influencing outcomes in oncological surgery. [ 8]. The centralization of care to “centers of excellence” in Europe has improved oncological outcomes over the last decades [9]. It has also been shown that treating patients in high-volume hospitals by high-volume surgeons is associated with better in-hospital mortality and oncological outcome [10, 11]. Even though centralization of cancer care is expected to yield superior results, the national strategy in *Germany is* still based on a voluntary certification process. Oncological colorectal surgery can still be performed in almost every hospital by every surgeon without restrictions [12, 13]. Nevertheless, only about one-third of CRC patients in Germany are treated in high-volume centers.
In our retrospective analysis, we wanted to examine socioeconomic disparities using the GISD, and the influence of annual hospital case volume for colorectal disease and their influence on survival in the federal state of Saxony, Germany.
## Study population
For the present retrospective analyses, we used data from the four clinical cancer registries in Saxony, Germany. This study does not include data of a clinical trial. In Germany, all inpatient and outpatient physicians as well as pathologists are obliged to report information on diagnosis, histological results, treatment and outcomes to the clinical cancer registries. Patients need to be informed about this process. Consent is not required but, patients have a right of objection, which is hardly used (fewer than five cases/year in Saxony). After documentation and validation of the data, analyses are conducted with an anonymized data set that does not allow identification of individual patients. Completeness of the registries has been estimated to be $98\%$ across all tumors since 2007 [14].
All cases with ICD-10 diagnoses C18 (excluding C18.1), C19 and C20, histologically verified adenocarcinoma and tumor-specific surgery in Saxony in the years 2010 to 2020 were included ($$n = 25$$,812). Patients had to be at least 18 years old and have their main residence in Saxony at the time of diagnosis ($$n = 24$$,306). We excluded 221 cases with missing information on UICC tumor stage. Thus, the final analytical sample comprised 24,085 cases (Fig. 1).
Fig. 1Flowchart of data collection, inclusion criteria and patient selectionAbbreviations: UICC, Union for International Cancer Control.
## Included variables and statistical analyses
Tumor localization was documented according to the International Classification of Diseases for Oncology (ICD-O; C18.0, C18.2 to C18.9, C19.9, C20.9). For multivariate analyses, colon cancer was further classified into right-sided (C18.0, C18.2–C18.4), left-sided (C18.5–C18.7) and others (C18.8, C18.9). We considered tumor stages defined by the Union for International Cancer Control (UICC; stages I, II, III, IV). Post-surgical histopathological classification of the UICC was used [15]. Regarding surgical treatment, only tumor-specific resections were considered. In the case of more than one resection, only the first procedure was analysed. The surgical approach was classified into open surgery, laparoscopic surgery, conversion and others/not specified. The urgency of surgery was documented as elective surgery vs. emergency surgery vs. unknown/missing. The documented number of resected lymph nodes was dichotomized into less than 12 and at least 12 lymph nodes. We further dichotomized patients based on the annual case volume of the hospital which was divided into low vs. high using a cut-off of 20 and 30 surgeries per year for rectal cancer (C19/C20) and colon cancer (C18), respectively. Adjuvant chemotherapy was documented if it was explicitly stated as a treatment option or based on the interval between surgery and start of chemotherapy (maximum time difference: 20 weeks). As an additional external indicator, a measure of area-based socioeconomic deprivation at the level of municipalities was used. The German Index of Socioeconomic Deprivation (GISD) considers the dimensions of education, employment and income, with each contributing one-third to the overall score [6]. Higher values indicate higher deprivation, i.e., lower socioeconomic status. For each tumor case, the GISD score of the patient’s place of residence at the time of diagnosis was used. The municipalities were then grouped into GISD quintiles based on our final analytical sample (1 – low level of deprivation, 2 – mid-low level of deprivation, 3 – medium level of deprivation, 4 – mid-high level of deprivation, 5 – high level of deprivation; see Fig. 2).
Fig. 2Distribution of German Index of Socioeconomic Deprivation (GISD) quintiles in Saxony, Germany For descriptive purposes, absolute numbers, and percentages (categorical variables) as well as medians (continuous variables) are presented. Absolute overall survival (OS) was calculated based on Kaplan–Meier estimation. Median survival time as well as 5-year survival rates with corresponding $95\%$ confidence intervals are reported. Univariate survival analyses were conducted for hospital case volume, surgical approach, and socioeconomic deprivation. For descriptive comparisons, the survival time of different groups of these categorical variables was plotted as Kaplan–Meier curves. Log-rank tests were used to test whether survival time differed between the included groups. Multivariate analyses were performed using weighted Cox regression, which provides unbiased average hazard ratio estimates even in the case of non-proportional hazards [16]. The resulting hazard ratios (HR) can be interpreted as relative risks. Survival analyses included only primary tumors and cases with a minimum follow-up time of one month ($$n = 19$$,845). Statistical analyses and graphical illustration were carried out using the R software package (version 3.6.0, https://www.R-project.org, The R Foundation).
## Patient characteristics
A total of 24,085 patients were included in the analysis (15,882 with colon cancer and 8,202 with rectal cancer). The description of the included colon and rectum cancer cases is shown in Table 1. The sample was predominantly male, also in terms of the proportion of cases of rectal cancer compared to colon cancer ($64.8\%$ vs. $54.6\%$). The median age was 70 and 74 years, respectively. The majority of colon cancer cases was localized on the right-side ($58.9\%$). Metastatic tumor stage (UICC IV) was present in $17.3\%$ (colon) and $13.9\%$ (rectum) of all cases. In most surgeries, at least 12 lymph nodes were resected (colon: $92.5\%$, rectum: $87.6\%$). The proportion of surgeries performed using a laparoscopic approach was higher for rectal cancer than for colon cancer ($24.8\%$ vs. $14.8\%$). Emergency surgery was more prevalent in colon cancer than in rectal cancer ($7.5\%$ vs. $1.8\%$). High hospital case volume was found in $56.1\%$ of colon cancer resections and $45.3\%$ of rectal cancer resections. Socioeconomic deprivation was evenly distributed due to the use of quintiles. The distribution of patient characteristics by case volume is shown in Supplementary Tables 1 and 2.
Colon resections were conducted in a total of 64 hospitals. There were 43 low-volume hospitals with a median number of 16.7 surgeries per year (minimum 0.1, maximum 29.1) and 21 high-volume hospitals with a median number of 42.7 surgeries per year (minimum 30.0, maximum 65.7). With regard to rectal cancer, the 64 hospitals were divided into 50 low-volume hospitals conducting a median of 9.5 resections per year (minimum 0.1, maximum 19.7) and 14 high-volume hospitals with a median of 25.4 resections per year (minimum 20.4, maximum 57.0).
Table 1Patient characteristics (2010–2020)Colon cancer (C18)n (%)Rectal cancer (C19/C20)n (%)Totaln (%)Sex Male8.675 (54.6)5.311 (64.8)13.986 (58.1) Female7.208 (45.4)2.891 (35.2)10.099 (41.9) Age (Median) 747073 Localization 1 Colon right-sided9.352 (58.9)-9.352 (38.8) Colon left-sided6.423 (40.4)-6.423 (26.6) Colon others*108 (0.7)-108 (0.4) Rectosigmoid junction-67 (0.8)67 (0.3) Rectum-8.135 (99.2)8.135 (34.0) UICC tumor stage I3.373 (21.2)1.542 (18.8)4.915 (20.4) II5.611 (35.3)1.904 (23.2)7.515 (31.2) III4.157 (26.2)3.612 (44.0)7.769 (32.3) IV2.742 (17.3)1.144 (13.9)3.886 (16.1) Surgical approach Open10.194 (64.2)4.147 (50.6)14.341 (59.6) Laparoscopic2.357 (14.8)2.033 (24.8)4.390 (18.2) Conversion259 (1.6)589 (7.2)848 (3.5) Others/n. a.3.073 (19.3)1.429 (17.4)4.502 (18.7) Urgency of surgery Elective surgery10.063 (63.4)5.723 (69.8)15.786 (65.5) Emergency surgery1.198 (7.5)148 (1.8)1.346 (5.6) Unknown/missing4.622 (29.1)2.331 (28.4)6.953 (28.9) Number of resected lymph nodes < 121.167 (7.5)1.006 (12.4)2.173 (9.2) 12+14.451 (92.5)7.085 (87.6)21.536 (90.8) Adjuvant chemotherapy no11.358 (71.5)4.963 (60.5)16.321 (67.8) yes4.525 (28.5)3.239 (39.5)7.764 (32.2) Hospital case volume 2 Low6.965 (43.9)4.484 (54.7)11.449 (47.5) High8.918 (56.1)3.718 (45.3)12.636 (52.5) Year of surgery 2010–20135.374 (33.8)3.104 (37.8)8.478 (35.2) 2014–20175.819 (36.6)2.974 (36.3)8.793 (36.5) 2018–20204.690 (29.5)2.124 (25.9)6.814 (28.3) GISD, Socio-economic deprivation 1 - low3.282 (20.7)1.584 (19.3)4.866 (20.2) 2 - mid-low3.104 (19.5)1.694 (20.7)4.798 (19.9) 3 - medium3.106 (19.6)1.696 (20.7)4.802 (19.9) 4 - mid-high3.624 (22.8)1.795 (21.9)5.419 (22.5) 5 - high2.767 (17.4)1.433 (17.5)4.200 (17.4) Total number of cases 15,883 8,202 24,085 1 Colon right-sided: C18.0, C18.2-C18.4, Colon left-sided: C18.5-C18.7, Colon others: C18.8-C18.9.2 C18: low < 30 surgeries/year, high ≥ 30 surgeries/year. C$\frac{19}{20}$: low < 20 surgeries/year, high ≥ 20 surgeries/year. Abbreviations: GISD, German Index of Socioeconomic Deprivation; UICC, Union for International Cancer Control.
## Univariate analysis
Median survival time was 87.9 months for colon cancer and 110.0 months for rectal cancer (Table 2). Five years after diagnosis, a total of $58.5\%$ (C18) and $65.6\%$ (C19/C20) of patients were still alive. Males and females had comparable 5-year survival rates (colon: $57.7\%$ vs. $59.5\%$, $$P \leq 0.1$$; rectum: $65.6\%$ vs. $65.6\%$, $$P \leq 0.2$$), whereas patients aged 70 years and younger had better 5-year survival than older patients (colon: $68.8\%$ vs. $51.6\%$, $P \leq 0.001$; rectum: $73.7\%$ vs. $55.4\%$, $P \leq 0.001$). Survival probability significantly decreased with higher UICC tumor stage (colon: stage I $80.0\%$ vs. stage IV $15.2\%$; rectum: stage I $80.3\%$ vs. stage IV $26.9\%$). Further univariate analyses revealed a significant association with surgical approach, hospital case volume (rectum only) and socioeconomic deprivation (Table 2; Fig. 3A–C). Laparoscopic surgery was associated with a better 5-year survival for both colon cancer ($71.2\%$ vs. $55.9\%$, $P \leq 0.001$) and rectal cancer ($73.2\%$ vs. $63.8\%$, $P \leq 0.001$). In addition, patients with rectal cancer had better 5-year survival when tumor resection was carried out in a high-volume hospital ($68.2\%$ vs. $63.5\%$, $$P \leq 0.002$$). Regarding socioeconomic deprivation, 5-year survival was worse for mid-low, mid and mid-high levels of deprivation compared to a low level of deprivation (overall $P \leq 0.001$). No difference could be observed for the highest level of deprivation compared to the lowest level of deprivation.
Table 2Results of overall and univariate survival analyses (Kaplan-Meier estimation)Colon cancer (C18)Rectal cancer (C19/C20)Median survival(Months, $95\%$ CI)5-year survival(%, $95\%$ CI)Median survival(Months, $95\%$ CI)5-year survival(%, $95\%$ CI) Overall 87.9 (84.5; 91.6)58.5 (57.6; 59.5)110.0 (105.0; 115.0)65.6 (64.4; 66.8) Sex $$p \leq 0.1$$ $$p \leq 0.2$$ Male84.1 (79.7; 88.9)57.7 (56.4; 59.0)105.0 (98.4; 112.0)65.6 (64.1; 67.2) Female92.2 (87.2; 101.1)59.5 (58.2; 61.0)121.0 (110.4; n. a.)65.6 (63.5; 67.7) Age $p \leq 0.001$ $p \leq 0.001$ ≤ 70 yearsn. a. (134.3.; n. a.)68.8 (67.4; 70.2)n. a. (135.9; n. a.)73.7 (72.2; 75.2) > 70 years63.7 (60.9; 67.3)51.6 (50.4; 52.9)71.7 (67.2; 76.3)55.4 (53.5; 57.3) UICC tumor stage $p \leq 0.001$ $p \leq 0.001$ I134.3 (132.5; n. a.)80.0 (78.2; 81.8)n. a. (130.0; n. a.)80.3 (77.8; 82.8) II121.2 (113.1; 131.4)69.9 (68.4; 71.4)116.3 (107.6; 128.7)70.5 (68.1; 73.0) III83.2 (76.8; 91.1)57.9 (56.1; 59.8)127.7 (115.2; 135.9)69.6 (67.9; 71.4) IV19.3 (17.8; 20.5)15.2 (13.6; 16.9)32.5 (30.3; 34.9)26.9 (24.0; 30.2) Surgical approach $p \leq 0.001$ $p \leq 0.001$ Open77.4 (73.3; 82.7)55.9 (54.7; 57.1)102.1 (94.5; 110.0)63.8 (62.1; 65.5) Laparoscopic138.5 (121.5; n. a.)71.2 (68.6; 73.9)129.9 (123.2; n. a.)73.2 (70.7; 75.8) Conversion122.7 (61.8; n. a.)61.4 (53.3; 70.8)90.8 (78.8; n. a.)62.8 (57.7; 68.3) Others/n. a.91.4 (85.4; 103.1)58.4 (56.3; 60.6)100.5 (84.9; 119.0)62.1 (59.3; 65.1) Hospital case volume 1 $$p \leq 0.8$$ $$p \leq 0.002$$ Low87.4 (82.2; 94.4)57.8 (56.3; 59.2)100.0 (95.2; 112.0)63.5 (61.8; 65.2) High88.0 (84.3; 93.4)59.2 (57.9; 60.4)115.0 (109.7; 128.0)68.2 (66.4; 70.0) GISD $p \leq 0.001$ $p \leq 0.001$ 1 – low104.7 (95.3; 118.3)62.1 (60.1; 64.2)129.3 (125.0; n. a.)71.7 (69.1; 74.4) 2 – mid-low87.4 (79.7; 93.6)58.6 (56.5; 60.8)98.8 (87.8; 114.0)64.0 (61.3; 66.8) 3 – mid83.2 (72.4; 91.1)56.3 (54.1; 58.5)96.3 (86.9; 105.0)62.3 (59.6; 65.1) 4 – mid-high79.9 (74.2; 86.7)57.0 (55.0; 59.0)108.6 (99.2; 123.0)64.4 (61.8; 67.1) 5 - high90.6 (83.3; 105.3)58.7 (56.4; 61.1)120.3 (110.0; n. a.)66.2 (63.3; 69.2)1 C18: low < 30 surgeries/year, high ≥ 30 surgeries/year. C$\frac{19}{20}$: low < 20 surgeries/year, high ≥ 20 surgeries/year. Abbreviations: CI, Confidence interval; GISD, German Index of Socioeconomic Deprivation; n. a., survival probability did not reach $50\%$; UICC Union for International Cancer Control.
Fig. 3Univariate survival analyses (Kaplan-Meier curves) and results of log-rank testFigure 3A: Univariate survival analysis for surgery in low-volume vs. high-volume hospitalsFigure 3B: Univariate survival analysis for surgical approachFigure 3C: Univariate survival analysis for socioeconomic deprivation (GISD-quintiles)Abbreviations: GISD, German Index of Socioeconomic Deprivation.
## Multivariate analysis
The results of multivariate analyses, including all relevant variables, are presented in Table 3. Females had better survival rates than males (colon: HR = 0.89, $95\%$ CI = 0.84–0.94, $P \leq 0.001$; rectum: HR = 0.85, $95\%$ CI = 0.78–0.92, $P \leq 0.001$) and older age was associated with worse survival (colon: HR = 1.91, $95\%$ CI = 1.79–2.03, $P \leq 0.001$; rectum: HR = 2.12, $95\%$ CI = 1.95–2.30, $P \leq 0.001$). Compared to early-stage cancer, increasing UICC was associated with worse survival (colon: HR = 10.81, $95\%$ CI = 9.65–12.11, $P \leq 0.001$ for stage IV; rectum: HR 7.30, $95\%$ CI = 6.24–8.53, $P \leq 0.001$). Laparoscopic surgery was associated with better survival compared to open surgery in colon cancer patients (HR = 0.76, $95\%$ CI = 0.68–0.83, $P \leq 0.001$) as well as rectal cancer patients (HR = 0.87, $95\%$ CI = 0.78–0.98, $P \leq 0.01$). Survival was worse in case of emergency surgery compared to elective surgery (colon: HR = 1.78, $95\%$ CI = 1.62–1.95, $P \leq 0.001$; rectum: HR = 2.14, $95\%$ CI = 1.68–2.74, $P \leq 0.001$). A higher number of resected lymph nodes was associated with better survival only in colon cancer patients (HR = 0.78, $95\%$ CI = 0.70–0.87, $P \leq 0.001$). Better survival was observed for patients receiving adjuvant chemotherapy (colon: HR = 0.49, $95\%$ CI = 0.46–0.53, $P \leq 0.001$; rectum: HR = 0.64, $95\%$ CI = 0.58–0.70, $P \leq 0.001$). Patients operated in a high-volume hospital showed better survival for rectal cancer (HR = 0.89, $95\%$ CI = 0.82–0.97, $P \leq 0.01$), whereas no significant effect was observed for colon cancer patients. Compared to patients living in a low deprivation municipality, those living in areas of mid-low to mid-high deprivation had worse survival outcomes (mid-low: HR = 1.18, $95\%$ CI = 1.08–1.30, $P \leq 0.001$; mid: HR = 1.18, $95\%$ CI = 1.07–1.29, $P \leq 0.001$; mid-high: HR = 1.22, $95\%$ CI = 1.12–1.33, $P \leq 0.001$). Rectal cancer patients showed almost similar results (Table 3).
Table 3Results of multivariate weighted Cox regressionColon cancer (C18)Rectal cancer (C19/C20)Hazard Ratio($95\%$ CI)pHazard Ratio($95\%$ CI)p Sex MaleReferenceReference Female0.89 (0.84; 0.94)< 0.0010.85 (0.78; 0.92)< 0.001 Age ≤ 70 yearsReferenceReference > 70 years1.91 (1.79; 2.03)< 0.0012.12 (1.95; 2.30)< 0.001 Localization 1 Colon right-sidedReference- Colon left-sided0.92 (0.86; 0.97)< 0.001- Colon others1.13 (0.84; 1.53)0.410- Rectosigmoid junction-Reference Rectum-1.01 (0.61; 1.68)0.969 UICC tumor stage IReferenceReference II1.51 (1.37; 1.67)< 0.0011.60 (1.39; 1.85)< 0.001 III2.98 (2.68; 3.31)< 0.0011.89 (1.64; 2.16)< 0.001 IV10.81 (9.65; 12.11)< 0.0017.30 (6.24; 8.53)< 0.001 Surgical approach OpenReferenceReference Laparoscopic0.76 (0.68; 0.83)< 0.0010.87 (0.78; 0.98)< 0.01 Conversion0.94 (0.73; 1.21)0.6321.09 (0.92; 1.28)0.318 Others/n. a.0.96 (0.89; 1.04)0.2881.05 (0.94; 1.17)0.383 Urgency of surgery Elective surgeryReferenceReference Emergency surgery1.78 (1.62; 1.95)< 0.0012.14 (1.68; 2.74)< 0.001 Unknown/missing1.06 (0.99; 1.14)0.8870.96 (0.88; 1.06)0.424 Resected lymph nodes < 12ReferenceReference 12+0.78 (0.70; 0.87)< 0.0011.05 (0.93; 1.19)0.412 Adjuvant chemotherapy noReferenceReference yes0.49 (0.46; 0.53)< 0.0010.64 (0.58; 0.70)< 0.001 Hospital case volume 2 LowReferenceReference High0.97 (0.91; 1.03)0.3030.89 (0.82; 0.97)< 0.01 Year of surgery 2010–2013ReferenceReference 2014–20170.97 (0.90; 1.03)0.3340.89 (0.81; 0.98)< 0.01 2018–20201.06 (0.97; 1.16)0.2330.83 (0.72; 0.95)< 0.001 GISD 1 – lowReferenceReference 2 – mid-low1.18 (1.08; 1.30)< 0.0011.36 (1.19; 1.55)< 0.001 3 – mid1.18 (1.07; 1.29)< 0.0011.34 (1.18; 1.53)< 0.001 4- mid-high1.22 (1.12; 1.33)< 0.0011.18 (1.03; 1.34)< 0.01 5 - high1.06 (0.96; 1.18)0.2401.04 (0.90; 1.21)0.5731 Colon right-sided: C18.0, C18.2-C18.4, Colon left-sided: C18.5-C18.7, Colon others: C18.8-C18.9.2 C18: low < 30 surgeries/year, high ≥ 30 surgeries/year. C$\frac{19}{20}$: low < 20 surgeries/year, high ≥ 20 surgeries/year. Abbreviations: CI, Confidence interval; GISD, German Index of Socioeconomic Deprivation; UICC Union for International Cancer Control.
## Conclusion
In accordance with other studies, we confirmed that characteristics like male sex, higher age and increasing UICC tumor stage are risk factors for worse survival [17, 18]. In addition, we found that a laparoscopic approach was associated with better survival. A mid-low, medium and mid-high GISD score was also an independent risk factor for worse survival. Only in rectal cancer patients, treatment in high volume hospitals was associated with better survival.
According to the literature, both the laparoscopic and open techniques seem to be equivalent in terms of tumor-specific survival [19, 20]. However, results from recent trials indicate that an open surgical approach was associated with a higher risk in terms of long-term mortality and OS [21, 22].Our findings also confirmed these data. Nevertheless, the reasons for this remain unclear and speculative. In our data, high-volume hospitals had almost double the percentage of laparoscopic surgeries compared to low-volume hospitals regarding rectal cancer. A more oncological adequate (better mesocolonic/mesorectal excision) resection may have been carried out. Compared with laparoscopic surgery, the open approach can result in postoperative morbidity (e.g., higher rates of incisional hernia, wound infection and other septic complications leading to reduced mortality and OS), which may contribute to the better outcomes following laparoscopic treatment [23]. The presence of a selection bias may also explain the superiority of the laparoscopic technique for overall survival. For elderly patients with higher risk profiles and secondary diseases, some hospitals are more likely to perform open procedures in order to avoid longer anesthesia and operating times. Unfortunately, we could adjust our analyses only for age, not for preexisting diseases or ASA (American Society of Anesthesiologists) risk classification or similar scores, due to missing data.
We are aware that our study has several weaknesses on account of its retrospective character. Due to lack of data (not reported to the clinical cancer registries), no statement about pre-existing illnesses or medication can be made. In Germany, older patients are more often operated in emergency and smaller hospitals,, which could result in a selection bias[24]. In addition, it is not possible to differentiate between high and low rectal cancer due to missing tumor height. No information on robotic surgery or tumor-specific survival is due to the current state of data collection. Use of the GISD is a necessary step because there is no valid better way to assess an individual socioeconomic score by income and educational level. Its limitations are set by the nature of GISD representing the patients’ socioeconomic status using habitation-based data with high individual variability. The fact that the lowest GISD strata did not show worse results is quite interesting. The explanation is speculative, but might occur due to the nature of GISD, which uses population-based and not individual data until 2012, whereas survival and effects of healthcare programs might result in the later episode. Also, you might state, that the quality of medical treatment in areas with highest GISD is quite good.
Only rectal cancer patients profited from treatment in high-volume hospitals achieving better survival. The highest level of evidence provided by the meta-analysis of Huo et al. indicates that the best outcomes occur in high-volume hospitals with high-volume surgeons, followed by low-volume hospitals with high-volume surgeons [9]. Thus, the individual surgeon is the main factor in achieving superior survival and oncological outcome. Unfortunately, the cancer registry data does not provide information on the specific surgeon. Rectal cancer surgery is more complex (narrow space in the pelvis) and less frequent (about one third of all colorectal operations) compared to colon cancer surgery. Training and experience of the surgeon is extremely important to achieve high quality of total mesorectal excision, which is directly associated with better patients’ survival and higher rates of sphincter preserving surgeries [25, 26]. Also, patients profit enormously from a laparoscopic or robotic surgical approach, which needs lots of practice and is more often carried out in high volume hospitals[27–29]. In addition, the fact that only one hospital specializing in rectal surgery had higher numbers than 50 operations per year is quite remarkable, and not competitive to other countries. For a better understanding of these problems in Germany, you need to look at the hospital allocation. In recent years, many hospitals in Germany, including Saxony, introduced evidence-based quality standards and improved their surgical and oncological outcomes [30]. However, since many hospitals are allowed to treat cancer patients, only few hospitals reach high patient volumes. In addition, in each hospital, there is not only one surgeon operating CRC patients, meaning the volume is shared by more surgeons, leading to low individual surgeon case load per year. Consequently, CRC care in Germany remains decentralized with high in-hospital morbidity and mortality rates compared to other western countries like USA, France, and UK [31]. While caseload can serve as a surrogate for treatment quality assessment in CRC surgery, our data do not suggest better survival for high-volume hospitals regarding colon cancer. [ 30]. Implementing independent controls and an auditing system surely helped maintaining standards [32]. Higher case numbers to achieve an obligatory certification as a pre-condition to treat CRC patients, and financial retribution could lead to more economical impact concerning the treatment of CRC patients [33]. There is a political and structural need for centralization and specialization in order to improve outcomes in the treatment of CRC patients in Saxony, Germany.
Our study also provides further evidence of social disparities in the treatment of CRC. Socioeconomic deprivation in Saxony, *Germany is* inversely associated with survival in CRC treatment. This association persists after demographic and cancer-related factors are considered. Relatively few data and studies exist on this topic in Germany. For the US, on the other hand, the existence of social and racial discrimination has been clearly demonstrated [34]. These major regional socioeconomic inequalities indicate a high potential for improving cancer care and survival worldwide. As many cancer entities have a long period of latency, changes in lifestyle, nutrition and physical activity, led to changes in incidence over the last decades. As obesity, one of the best-known risk factors for CRC, became epidemic and the nutrition worsened, colorectal cancer incidence increased, especially in rural areas with low income. In comparison to the 1950s, CRC changed from a prosperity disease to a disease of higher prevalence in low-income social stratum [35, 36]. This fact was clearly confirmed by our data indicating rural areas to have higher GISD and thus worse survival. An increased travelling distance to inpatient and outpatient care might be another reason for the high urban- rural disparities. Also, higher education and income, overrepresented in urban areas (lower GISD strata) showed better survival. This could be due to better access to prevention coloscopy and healthier lifestyle. Unfortunately, exact data on this topic is missing. Although, excessive food intake and insufficient physical activity are individual decision, political, economic, and social disparities play a major role in fighting adiposity and CRC[37]. We live in an environment where obesity is stimulated. This might be a key to further decrease CRC incidence and mortality. An efficient allocation of health resources therefore is indispensable.[38].
Our data analysis from Saxony could be extended to the other federal states and Germany as a whole. The data transfer process should also be expanded to create nationwide databases. Therefore, data from health insurance companies could also be added. Since individual data on income, employment status and educational level are often missing, future studies should address individual-level patient data with access to treatment information. This would allow more detailed examination of the reasons for these socioeconomic inequalities in cancer survival and could help establish improvements in access to health protection, diagnostic and therapy[7]. The social disparities leading to an elevated risk for cancer development and worse survival require targeted public health action and policy in order to address the complexity of these relationships [5].
## Summary
In the analyzed population in Saxony, Germany, we showed that laparoscopic surgical technique and lower socioeconomic deprivation were associated with better survival in the treatment of CRC. There was also some indication that high hospital case volume improves patient outcome in rectal cancer. Our study adds evidence that might help in changing national health policies with the aim of achieving a better outcome for patients with CRC. Only a change at the national level can help improve therapy and elevate the quality of colorectal cancer care delivery.
## Electronic supplementary material
Below is the link to the electronic supplementary material.
Supplementary Material 1
## References
1. 1.Ferlay J, Colombet M, Soerjomataram I, Parkin DM, Pineros M, Znaor A et al. Cancer statistics for the year 2020: An overview.International journal of cancer Journal international du cancer. 2021.
2. 2.2017/2018 Krebs in Deutschland. Robert Koch-Institut, Herausgeber und die Gesellschaft der epidemiologischen Krebsregister in Deutschland e.V, Herausgeber. Berlin., 2018. 12. Ausgabe.
3. 3.Rohleder S, Stock C, Bozorgmehr K. Socioeconomic deprivation is inversely associated with measles incidence: a longitudinal small-area analysis, Germany, 2001 to 2017.Euro Surveill. 2021;26(17).
4. Grundmann N, Mielck A, Siegel M, Maier W. **Area deprivation and the prevalence of type 2 diabetes and obesity: analysis at the municipality level in Germany**. *BMC Public Health* (2014.0) **14** 1264. DOI: 10.1186/1471-2458-14-1264
5. Hoebel J, Kroll LE, Fiebig J, Lampert T, Katalinic A, Barnes B. **Socioeconomic inequalities in Total and Site-Specific Cancer incidence in Germany: a Population-Based Registry Study**. *Front Oncol* (2018.0) **8** 402. DOI: 10.3389/fonc.2018.00402
6. 6.Kroll LE, Schumann M, Hoebel J, Lampert T. Regional health differences – developing a socioeconomic deprivation index for Germany. Robert Koch-Institut, Epidemiologie und Gesundheitsberichterstattung; 2017.
7. Jansen L, Eberle A, Emrich K, Gondos A, Holleczek B, Kajuter H. **Socioeconomic deprivation and cancer survival in Germany: an ecological analysis in 200 districts in Germany**. *Int J cancer J Int du cancer* (2014.0) **134** 2951-60. DOI: 10.1002/ijc.28624
8. Liu CJ, Chou YJ, Teng CJ, Lin CC, Lee YT, Hu YW. **Association of surgeon volume and hospital volume with the outcome of patients receiving definitive surgery for colorectal cancer: a nationwide population-based study**. *Cancer* (2015.0) **121** 2782-90. DOI: 10.1002/cncr.29356
9. Huo YR, Phan K, Morris DL, Liauw W. **Systematic review and a meta-analysis of hospital and surgeon volume/outcome relationships in colorectal cancer surgery**. *J Gastrointest Oncol* (2017.0) **8** 534-46. DOI: 10.21037/jgo.2017.01.25
10. Aquina CT, Probst CP, Becerra AZ, Iannuzzi JC, Kelly KN, Hensley BJ. **High volume improves outcomes: the argument for centralization of rectal cancer surgery**. *Surgery* (2016.0) **159** 736-48. DOI: 10.1016/j.surg.2015.09.021
11. Breugom AJ, Boelens PG, van den Broek CB, Cervantes A, Van Cutsem E, Schmoll HJ. **Quality assurance in the treatment of colorectal cancer: the EURECCA initiative**. *Annals of oncology: official journal of the European Society for Medical Oncology / ESMO* (2014.0) **25** 1485-92. DOI: 10.1093/annonc/mdu039
12. Jacob A, Albert W, Jackisch T, Jakob C, Sims A, Witzigmann H. **Association of certification, improved quality and better oncological outcomes for rectal cancer in a specialized colorectal unit**. *Int J Colorectal Dis* (2021.0) **36** 517-33. DOI: 10.1007/s00384-020-03792-8
13. Wesselmann S, Winter A, Ferencz J, Seufferlein T, Post S. **Documented quality of care in certified colorectal cancer centers in Germany: German Cancer Society benchmarking report for 2013**. *Int J Colorectal Dis* (2014.0) **29** 511-8. DOI: 10.1007/s00384-014-1842-x
14. 14.Tillack AWC, Buchali A, Klug SJ. Flächendeckende bevölkerungsbezogene klinische Krebsregister: Darstellung der Versorgungstransparenz in Brandenburg und Sachsen. Onkologe 2013 19:1065–1073. 2013.
15. 15.UICC. TNM classification of malignant tumours 7th edition2010.
16. Schemper M, Wakounig S, Heinze G. **The estimation of average hazard ratios by weighted Cox regression**. *Stat Med* (2009.0) **28** 2473-89. DOI: 10.1002/sim.3623
17. Yang Y, Wang G, He J, Ren S, Wu F, Zhang J. **Gender differences in colorectal cancer survival: a meta-analysis**. *Int J cancer J Int du cancer* (2017.0) **141** 1942-9. DOI: 10.1002/ijc.30827
18. Collaborators GBDCC. **The global, regional, and national burden of colorectal cancer and its attributable risk factors in 195 countries and territories, 1990–2017: a systematic analysis for the global burden of Disease Study 2017**. *Lancet Gastroenterol Hepatol* (2019.0) **4** 913-33. DOI: 10.1016/S2468-1253(19)30345-0
19. Study G, Buunen M, Veldkamp R, Hop WC, Kuhry E, Jeekel J. **Survival after laparoscopic surgery versus open surgery for colon cancer: long-term outcome of a randomised clinical trial**. *Lancet Oncol* (2009.0) **10** 44-52. DOI: 10.1016/S1470-2045(08)70310-3
20. 20.Lilley R, Chan E, Ng N, Orr A, Szostok M, Yeh GTT et al. Recurrence Kinetics after Laparoscopic Versus Open Surgery in Colon Cancer. A Meta-Analysis.J Clin Med. 2021;10(18).
21. Fahim M, Dijksman LM, Burghgraef TA, van der Nat PB, Derksen WJM, van Santvoort HC. **Increased long-term mortality after open colorectal cancer surgery: a multicentre population-based study**. *Colorectal disease: the official journal of the Association of Coloproctology of Great Britain and Ireland* (2021.0) **23** 2567-74. DOI: 10.1111/codi.15793
22. Lacy AM, Delgado S, Castells A, Prins HA, Arroyo V, Ibarzabal A. **The long-term results of a randomized clinical trial of laparoscopy-assisted versus open surgery for colon cancer**. *Ann Surg* (2008.0) **248** 1-7. DOI: 10.1097/SLA.0b013e31816a9d65
23. Iversen LH, Ingeholm P, Gogenur I, Laurberg S. **Major reduction in 30-day mortality after elective colorectal cancer surgery: a nationwide population-based study in Denmark 2001–2011**. *Ann Surg Oncol* (2014.0) **21** 2267-73. DOI: 10.1245/s10434-014-3596-7
24. Diers J, Baum P, Lehmann K, Uttinger K, Baumann N, Pietryga S. **Disproportionately high failure to rescue rates after resection for colorectal cancer in the geriatric patient population - A nationwide study**. *Cancer Med* (2022.0) **11** 4256-64. DOI: 10.1002/cam4.4784
25. Carlsen E, Schlichting E, Guldvog I, Johnson E, Heald RJ. **Effect of the introduction of total mesorectal excision for the treatment of rectal cancer**. *Br J Surg* (1998.0) **85** 526-9. DOI: 10.1046/j.1365-2168.1998.00601.x
26. Martling A, Holm T, Rutqvist LE, Johansson H, Moran BJ, Heald RJ. **Impact of a surgical training programme on rectal cancer outcomes in Stockholm**. *Br J Surg* (2005.0) **92** 225-9. DOI: 10.1002/bjs.4834
27. Muaddi H, Hafid ME, Choi WJ, Lillie E, de Mestral C, Nathens A. **Clinical outcomes of robotic surgery compared to Conventional Surgical Approaches (laparoscopic or open): a systematic overview of reviews**. *Ann Surg* (2021.0) **273** 467-73. DOI: 10.1097/SLA.0000000000003915
28. 28.Mann B, Kukies S, Krogh O, Virakas G. [Robotic-assisted surgery of rectal cancer-Technique, limitations and results]. Der Chirurg; Zeitschrift fur alle Gebiete der operativen Medizen. 2021;92(7):599–604.
29. Gachabayov M, You K, Kim SH, Yamaguchi T, Jimenez-Rodriguez R, Kuo LJ. **Meta-analysis of the impact of the learning curve in robotic rectal Cancer surgery on histopathologic outcomes**. *Surg Technol Int* (2019.0) **34** 139-55. PMID: 31037712
30. Trautmann F, Reissfelder C, Pecqueux M, Weitz J, Schmitt J. **Evidence-based quality standards improve prognosis in colon cancer care**. *Eur J Surg oncology: J Eur Soc Surg Oncol Br Association Surg Oncol* (2018.0) **44** 1324-30. DOI: 10.1016/j.ejso.2018.05.013
31. Baum P, Diers J, Lichthardt S, Kastner C, Schlegel N, Germer CT. **Mortality and complications following visceral surgery: a nationwide analysis based on the diagnostic categories used in German Hospital Invoicing Data**. *Deutsches Arzteblatt international* (2019.0) **116** 739-46. PMID: 31774053
32. Kowalski C, Graeven U, von Kalle C, Lang H, Beckmann MW, Blohmer JU. **Shifting cancer care towards Multidisciplinarity: the cancer center certification program of the german cancer society**. *BMC Cancer* (2017.0) **17** 850. DOI: 10.1186/s12885-017-3824-1
33. Ghadban T, Reeh M, Bockhorn M, Grotelueschen R, Bachmann K, Grupp K. **Decentralized colorectal cancer care in Germany over the last decade is associated with high in-hospital morbidity and mortality**. *Cancer Manag Res* (2019.0) **11** 2101-7. DOI: 10.2147/CMAR.S197865
34. 34.Poulson M, Geary AD, Annesi C, Dechert T, Kenzik K, Hall J. The impact of income and social mobility on colorectal Cancer outcomes and treatment: a cross-sectional study. Annals of surgery; 2021.
35. 35.Kogevinas M, Porta M. Socioeconomic differences in cancer survival: a review of the evidence.IARC Sci Publ. 1997(138):177–206.
36. Singh GK, Miller BA, Hankey BF. **Changing area socioeconomic patterns in U.S. cancer mortality, 1950–1998: part II–Lung and colorectal cancers**. *J Natl Cancer Inst* (2002.0) **94** 916-25. DOI: 10.1093/jnci/94.12.916
37. Brenner H, Mielck A, Klein R, Ziegler H. **The role of socioeconomic factors in the survival of patients with colorectal cancer in Saarland/Germany**. *J Clin Epidemiol* (1991.0) **44** 807-15. DOI: 10.1016/0895-4356(91)90133-T
38. Berrino F, De Angelis R, Sant M, Rosso S, Bielska-Lasota M, Coebergh JW. **Survival for eight major cancers and all cancers combined for european adults diagnosed in 1995-99: results of the EUROCARE-4 study**. *Lancet Oncol* (2007.0) **8** 773-83. DOI: 10.1016/S1470-2045(07)70245-0
|
---
title: Adherence to antidiabetic drug therapy and reduction of fatal events in elderly
frail patients
authors:
- Federico Rea
- Laura Savaré
- Valeria Valsassina
- Stefano Ciardullo
- Gianluca Perseghin
- Giovanni Corrao
- Giuseppe Mancia
journal: Cardiovascular Diabetology
year: 2023
pmcid: PMC9999593
doi: 10.1186/s12933-023-01786-8
license: CC BY 4.0
---
# Adherence to antidiabetic drug therapy and reduction of fatal events in elderly frail patients
## Abstract
### Background
To evaluate the protective effect of oral antidiabetic drugs in a large cohort of elderly patients with type 2 diabetes differing for age, clinical status, and life expectancy, including patients with multiple comorbidities and short survival.
### Methods
A nested case–control study was carried out by including the cohort of 188,983 patients from Lombardy (Italy), aged ≥ 65 years, who received ≥ 3 consecutive prescriptions of antidiabetic agents (mostly metformin and other older conventional agents) during 2012. Cases were the 49,201 patients who died for any cause during follow-up (up to 2018). A control was randomly selected for each case. Adherence to drug therapy was measured by considering the proportion of days of the follow-up covered by the drug prescriptions. Conditional logistic regression was used to model the risk of outcome associated with adherence to antidiabetic drugs. The analysis was stratified according to four categories of the clinical status (good, intermediate, poor, and very poor) differing for life expectancy.
### Results
There was a steep increase in comorbidities and a marked reduction of the 6-year survival from the very good to the very poor (or frail) clinical category. Progressive increase in adherence to treatment was associated with a progressive decrease in the risk of all-cause mortality in all clinical categories and at all ages (65–74, 75–84 and ≥ 85 years) except for the frail patient subgroup aged ≥ 85 years. The mortality reduction from lowest to highest adherence level showed a tendency to be lower in frail patients compared to the other categories. Similar although less consistent results were obtained for cardiovascular mortality.
### Conclusions
In elderly diabetic patients, increased adherence to antidiabetic drugs is associated with a reduction in the risk of mortality regardless of the patients’ clinical status and age, with the exception of very old patients (age ≥ 85 years) in the very poor or frail clinical category. However, in the frail patient category the benefit of treatment appears to be less than in patients in good clinical conditions.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12933-023-01786-8.
## Background
Randomized controlled trials have shown that, in patients with type 2 diabetes, old conventional antidiabetic drug therapy is associated with a reduction in the risk of microvascular complications [1] while meta-analyses of old conventional drug-based trials have shown that when data from several trials are pooled [2–6], the benefit of glucose-lowering treatment extends to macrovascular outcomes [7, 8]. Protection against macrovascular outcomes has been found to be even greater with use of newer antidiabetic agents such as sodium-glucose cotransporter 2 (SGLT2) inhibitors and glucagon-like peptide-1 receptor agonists (GLP1-RA), particularly for heart failure [9–12]. However, although randomized trial evidence on the protective effect of antidiabetic treatment has been extended to patients aged 65 years and beyond [13], knowledge is scanty for people aged 80 years or more [14] and especially limited for old patients with several comorbidities and a short life expectancy, i.e. those often referred to as “frail” patients [15], in part because frail patients tend to be excluded from recruitment in trials with a several year duration.
The objective of this observational study was to evaluate the protective effect of oral antidiabetic drugs in frail older adults (≥ 65 years) with type 2 diabetes. Frail individuals were identified via a multisource comorbidity score that accurately predicts the risk of mortality [16]. Analysis was extended to older adults in better clinical conditions.
## Setting
The data used for the present study were retrieved from the healthcare utilization databases of Lombardy, a region of Italy that accounts for about $16\%$ (almost 10 million people) of its population. All Italian citizens have equal access to healthcare services (e.g., hospitalizations, outpatient visits, instrumental examinations, laboratory tests, bioimaging, and drugs for chronic diseases) as part of the National Health Service (NHS). In Lombardy, these data are included in an automated system of databases that provides information on individual demographics, drug prescriptions (according to the Anatomical Therapeutic Chemical—ATC—system), medical or surgical interventions, and hospitalizations, according to diagnoses and procedures coded as in the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) system. Because a unique identification code was used for all databases, their linkage provided information on the complete care pathway supplied to NHS beneficiaries for several years. To preserve privacy, in the analyses of the Lombardy databases each individual identification code is automatically deidentified, the inverse process being allowed only to the Regional Health Authority upon request from judicial authorities. A detailed description of the healthcare utilization databases of the Lombardy region in the field of cardiovascular and metabolic diseases is available in previous studies [17, 18]. The ICD-9-CM and ATC codes used for the current study are reported in Additional file 1: Table S1.
## Cohort selection
The target population consisted of the Lombardy residents aged ≥ 65 years who were NHS beneficiaries. Of these, those who received ≥ 3 consecutive prescriptions of oral antidiabetic agents during 2012 were identified and the date of the third prescription was defined as the index date. We considered that three consecutive prescriptions within a year are indicative of regular prescription and use. Because insulin might require frequent changes in dose requirements over time, patients prescribed only insulin were not included in the study. Two additional categories of patients were excluded from the analysis, i.e., those who had not been NHS beneficiaries for at least 5 years before the index date and those who did not reach at least 6 months of follow-up. The remaining patients were included into the final cohort whose members accumulated person-years of follow-up from the index date until the earliest date among death, emigration or June 30th, 2018.
## Selection of cases and controls
When the effect of time-dependent exposure needs to be investigated in the context of large databases, the nested case–control design is a valid alternative to the cohort design [19]. The case–control study consists of four steps: (i) cohort selection, (ii) case definition and selection, (iii) for each case, identification of all possible controls, and (iv) random selection of m controls for each case [20]. In the present study, the cohort involved the oral antidiabetic drug users as described above. Death from any cause was the primary outcome of interest, and cases were thus the cohort members who died during follow-up. For each case patient, all cohort members who survived when the matched case died were identified (i.e. the incidence density sampling method was adopted). For each case patient, one control was randomly selected from the cohort members to be matched for sex, age at index date, clinical status (see below) and date of index prescription.
A secondary outcome was cardiovascular mortality, i.e., death for ischemic heart disease, cerebrovascular disease, or heart failure, which was addressed by another nested case–control study in which patients who died for cardiovascular causes were the cases, and patients matched for age, sex and clinical status and index date were the controls, as described above.
## Assessing the clinical category
For each cohort member, the clinical status was assessed by the Multisource Comorbidity Score (MCS), i.e., a prognostic score based on 34 morbidities identified by the ICD-9-CM and ATC codes, which has been shown to predict mortality better than the Charlson, Elixahauser and Chronic Disease Scores in the Italian population [16, 21]. We assigned to each morbidity a weight proportional to its strength in predicting mortality, and calculated the sum of the morbidity' weights suffered by a patient. Because all cohort members suffered from diabetes, the contribution of diabetes to the MCS was not considered. Further details on the calculation of MCS are available in the original manuscript [16]. The score was used to separate patients according to 4 categories of clinical status: good (MCS = 0), intermediate (1 ≤ MCS ≤ 4), poor (5 ≤ MCS ≤ 14), and very poor (MCS ≥ 15).
## Adherence to oral antidiabetic drug treatment
For each patient, all antidiabetic drugs prescribed during the follow-up were identified. The period covered by an individual prescription was calculated by dividing the total amount of the drug prescribed for the defined daily dose. For overlapping prescriptions, the patient was assumed to have taken all drugs contained in the first prescription before starting the second one. Adherence was measured by the cumulative number of days in which the drug was available divided by the days of the follow-up, i.e. by the proportion of days covered (PDC) by treatment [22]. We classified patients prescribed more than one antidiabetic drug class as “adherent” if they were covered by at least one drug prescription. Because information on drug therapies dispensed during hospitalization was not available, the exposure to antidiabetic treatment before hospital admission was assumed to be continued for the entire span of the-hospital stay [23]. Four categories of adherence with antidiabetic drug therapy were considered, i.e. very low (≤ $25\%$), low ($26\%$-$50\%$), intermediate ($51\%$-$75\%$) and high (> $75\%$) PDC values. These cut-off values were used because in previous studies on the Lombardy database these adherence levels showed a clear association with mortality among elderly patients in treatment with antihypertensive and lipid-lowering drugs [24, 25].
## Covariates
Additional information included (i) the class of antidiabetic drugs, (ii) the use of insulin, (iii) comedications, e.g., use of antihypertensive, lipid-lowering, antiarrhythmic and other cardiovascular agents, and (iv) comorbidities, i.e., cardiovascular, kidney, respiratory disease, mental disorder, and cancer. Comedications and comorbidities were identified from out-of-hospital prescriptions and in-hospital diagnoses within the 5 years prior to the index date.
## Data analysis
Survival curves were built by means of the Kaplan–Meier method according with categories of clinical status, and compared through the log-rank test. Linear regression and chi square for the trend were used to test trend in covariates along the categories of clinical status. In addition, standardized mean differences were used to test differences between cases and controls. Standardized mean differences < 0.10 were considered negligible [26].
Conditional logistic regression models were fitted to estimate the odds ratio, and its $95\%$ confidence interval (CI), of all-cause and cardiovascular mortality in relation to the PDC categories, using the lowest category (≤ $25\%$) as reference. Adjustments were made for the above-reported covariates. Odds ratio trends were tested according to the statistical significance of the regression coefficient of the recoded variable obtained by scoring the corresponding categories. All estimates were obtained by stratifying the cohort members according to the categories of clinical status. The impact of adherence on the outcomes according to categories of both clinical status and age (65–74 years, 75–84 years, and ≥ 85 years) was also measured.
## Sensitivity analysis
To verify the robustness of the main findings, five sensitivity analyses were performed. First, a different categorization of adherence was adopted: low (< $80\%$) and high (≥ $80\%$), as commonly used in the medical literature. Second, because the prescribed daily doses (not included in our database) might not be closely correspond to the defined daily doses [27], analyses were repeated by calculating the period covered by prescriptions from the number of tablets in the dispensed canisters, assuming a treatment schedule of one tablet per day. Third, the potential bias associated with unmeasured confounders was investigated by the rule-out approach described by Schneeweiss [28], which detects the extent of the unmeasured confounding required to fully account for the observed exposure–outcome association. We set the unmeasured confounder to exert a potentially marked effect on the results: (i) to have a $30\%$ prevalence in the study population; (ii) to increase the risk of death up to tenfold in patients exposed to the unmeasured confounder than in those unexposed; and (iii) to be up to tenfold less common in high than in low adherent patients. Fourth, we investigated the possible presence of “healthy user bias”, i.e. the possibility that more adherent patients were more likely to follow healthy lifestyle advices or seek other preventive services than less adherent patients. The number of outpatient services (e.g. outpatient visits, laboratory examinations, instrument-based examinations) provided by the NHS in the previous 2 years was considered as a proxy for the patient’s behaviour to search for preventive non-pharmacological services. We proceeded with the following two-stage procedure: (i) ordinal logistic regression was fitted to estimate the odds ratio of adherence to antidiabetic treatment in relation to the number of outpatient services, and (ii) the association between adherence to antidiabetic treatment and survival was further investigated by stratifying for the number of outpatient services. Fifth, because the study cohort included prevalent users, i.e. patients already taking antidiabetic drug therapy before cohort entry, results might be affected by selection bias [29]. The analyses were then repeated by restricting the cohort to new users, i.e. patients who did not receive a prescription of antidiabetic drugs in the 5 years before the index date [30].
The Statistical Analysis System Software (version 9.4; SAS Institute, Cary, NC, USA) was used for the analyses. For all hypotheses tested, p-values less than 0.05 were considered to be significant.
## Data and resource availability
The data that support the findings of this study are available from Lombardy Region, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the Lombardy Region upon reasonable request.
## Patients
The distribution of exclusion criteria is reported in Additional file 1: Fig S1. Among the 276,336 patients on treatment with antidiabetic agents during 2012, 188,983 met the inclusion criteria. The cohort members accumulated 1,072,151 person-years of observation (on average, 5.7 years) and generated 49,219 deaths, with a mortality rate of 45.9 per 1000 person-years. As shown in Fig. 1 (upper panel), in the group as a whole the 6-year survival decreased from 85 to $52\%$ from the group of patients with good to the group of patients with a very poor clinical status. Death incidence increased progressively as age increased and in each age group it was progressively greater as the clinical category changed from good to very poor or frail. More than $80\%$ of patients aged ≥ 85 years with a very poor clinical status died during the study follow-up (Fig. 1, lower panel).Fig. 1Kaplan–Meier survival curves for all-cause death according to the clinical category as determined by Multisource Comorbidity Score and 6-year mortality probabilities according to the clinical category and age strata The baseline characteristics of cohort members are reported in Additional file 1: Table S2, according to the clinical category. Age and men prevalence increased as the clinical category deteriorated, this being the case also for use of cardiovascular drugs, non-cardiovascular drugs and previous hospitalization for a variety of diseases. The increase was particularly large for patients in the very poor clinical category. Among antidiabetic drugs, use of metformin and sulfonylurea were by far the most widely used drugs. Use of metformin, sulfonylurea, and pioglitazone decreased from the good to the very poor clinical category. The opposite trend was observed for the other antidiabetic drugs, including insulin and the much less frequently used newer oral antidiabetic agents.
Overall, 49,219 patients of the selected cohort died during follow-up (cases), of whom 49,201 were matched with alive patients from the cohort (controls). As expected, cases and controls had superimposable age, sex representation and clinical status (Table 1). This was the case also for the index date (see Methods). There were only small differences in virtually all clinical and therapeutic characteristics between the two groups. Compared to controls, cases showed an overall lower adherence to antidiabetic drug therapy. Table 1Comparison of demographic, clinical and therapeutic characteristics of the cohort members who died (cases) or survived (controls)Cases ($$n = 49$$,201)Controls ($$n = 49$$,201)SMDBaseline Men26,035 ($52.9\%$)26,035 ($52.9\%$)MV Age (years): mean (SD)79.7 (7.2)79.6 (7.2)MV Clinical categoryaMV Good6285 ($12.8\%$)6285 ($12.8\%$) Intermediate11,378 ($23.1\%$)11,378 ($23.1\%$) Poor22,717 ($46.2\%$)22,717 ($46.2\%$) Very poor8821 ($17.9\%$)8821 ($17.9\%$)Antidiabetic agents at cohort entry Metformin33,078 ($67.2\%$)34,031 ($69.2\%$)0.042 DPP-4 inhibitor1869 ($3.8\%$)2348 ($4.8\%$)0.048 Sulfonylurea25,767 (52.4)25,127 ($51.1\%$)0.026 Pioglitazone1991 ($4.0\%$)2316 ($4.7\%$)0.032 GLP-1 RA293 ($0.6\%$)331 ($0.7\%$)0.010 Meglitinide6917 ($14.1\%$)6064 ($12.3\%$)0.051 Alpha glucosidase inhibitors1664 ($3.4\%$)1534 ($3.1\%$)0.015 Insulin3675 ($7.5\%$)2919 ($5.9\%$)0.061Other drugs Antihypertensive agents45,601 ($92.7\%$)45,320 ($92.1\%$)0.022 Lipid-lowering agents28,169 ($57.3\%$)30,063 ($61.1\%$)0.078 Antiarrhythmic agents4835 ($9.8\%$)4312 ($8.8\%$)0.037 Antiplatelet drugs33,169 ($67.4\%$)32,637 ($66.3\%$)0.023 Oral anticoagulant agents7638 ($15.5\%$)6153 ($12.5\%$)0.087 Digitalis4486 ($9.1\%$)3312 ($6.7\%$)0.088 Nitrates11,453 ($23.3\%$)10,988 ($22.3\%$)0.023 Anti-gout drugs10,466 ($21.3\%$)9674 ($19.7\%$)0.040 Antidepressant agents12,294 ($25.0\%$)10,732 ($21.8\%$)0.075 Drugs for respiratory disease17,010 ($34.6\%$)17,297 ($35.2\%$)0.012Previous hospitalizations Cardiovascular disease21,955 ($44.6\%$)20,014 ($40.7\%$)0.080 Kidney disease2928 ($6.0\%$)2143 ($4.4\%$)0.072 Metal disorders2159 ($4.4\%$)1658 ($3.4\%$)0.053 Respiratory disease7347 ($14.9\%$)5240 ($10.7\%$)0.128 Cancer7359 ($15.0\%$)7342 ($14.9\%$)0.001During follow-up Adherence with antidiabetic drugs b0.213 Very low4201 ($8.5\%$)3149 ($6.4\%$) Low10,272 ($20.9\%$)8400 ($17.1\%$) Intermediate13,437 ($27.3\%$)12,054 ($24.5\%$) High21,291 ($43.3\%$)25,598 ($52.0\%$)MV matching variable, SD standard deviation, SMD standardized mean differences, DDP-4 Dipeptidyl peptidase-4, GLP-1 RA glucagon-like peptide 1 receptor agonistsa Clinical frailty was assessed by the Multisource Comorbidity Score (MCS) and four categories were considered: good (MCS = 0), intermediate (1 ≤ MCS ≤ 4), poor (5 ≥ MCS ≤ 14) and very poor (MCS ≥ 15)b Adherence was measured by the ratio between the days with available antidiabetic drug prescriptions and all days of follow up. Adherence categories are: very low: ≤ $25\%$; low: 26 to $50\%$; intermediate: 51 to $75\%$; and high: > $75\%$
## Antidiabetic drug therapy and mortality
The association between adherence to drug treatment and all-cause mortality is shown in Fig. 2, top panel. A progressive increase of adherence to treatment was associated with a progressive decrease in the risk of all-cause mortality in all clinical categories. The reduction of all-cause mortality from the lowest to the highest adherence level was lowest in the very poor clinical category compared to the other clinical categories, i.e. $26\%$ ($95\%$ CI 17–$34\%$) vs $36\%$ (25–$46\%$), $50\%$ (44–$56\%$) and $38\%$ (33–$42\%$) for the good, intermediate and poor clinical category, respectively. Similar trends were observed for cardiovascular mortality, i.e. the reduction from the lowest to the highest adherence level were $26\%$ (-15–$52\%$), $48\%$ (28–$62\%$), $44\%$ (33–$53\%$), and $37\%$ (16–$52\%$) for the good, intermediate and poor clinical category, respectively (Fig. 2, lower panels).Fig. 2Effect of adherence with antidiabetic drugs on the odds ratio of all-cause and cardiovascular death according to the clinical category as measured by Multisource Comorbidity Score The results of the stratified analysis for clinical category and age are reported in Fig. 3. In all age strata, there was an association between adherence to treatment and all-cause mortality, i.e. a decrease in the risk of fatal events from any cause as adherence increased. An exception was the group of patients aged ≥ 85 years with a very poor clinical status (or frailty) in whom changes in adherence did not modify the total mortality risk (p-trend = 0.722). The results for adherence and cardiovascular mortality were similar but less consistent (Additional file 1: Table S3). In the group of frail patients aged ≥ 85 years, an increase of adherence was associated with a $27\%$ reduction of cardiovascular mortality, which did not achieve statistical significance (p-trend = 0.069). Furthermore, in the groups in good clinical conditions aged 65–74 years and aged ≥ 85 years, an increase of adherence had a paradoxical effect, i.e. an increase of cardiovascular mortality, albeit not significant (p-trend = 0.308 and 0.286, respectively). In these groups, however, the number of lethal cardiovascular events was extremely small (Additional file 1: Table S4).Fig. 3Effect of adherence with oral antidiabetics on the risk of all-cause mortality according to clinical category and age
## Sensitivity analyses
The results on all-cause mortality did not change either by varying the criteria for categorization of adherence with drug treatment (Additional file 1: Table S5) or by using a different way to estimate the duration of each prescription (Additional file 1: Table S6). As shown by the rule-out approach analysis reported in Additional file 1: Fig S2, assuming that highly adherent patients had a three-fold lower odds of exposure to an unmeasured confounder than patients with a very low adherence, the confounder should have increased the outcome risk of all-cause mortality by three-fold for nullifying the observed protective effect of drug adherence in patients with a very poor clinical category. The required nullifying confounder–outcome associations had to be even greater in the other clinical categories.
Additional file 1: Table S7 shows that the number of outpatient services used by the patients in the 2 years before the beginning of the study observations was associated with adherence to antidiabetic treatment. However, the results did not substantially change after accounting for this variable (Additional file 1: Table S8).
Finally, 61,722 cohort members were new users, i.e. they did not receive a prescription of antidiabetic drugs in the previous 5 years. Among these, 16,915 patients died during follow-up, of whom 16,897 were matched with alive new-user patients. The results did not change by restricting the cohort to new users (Additional file 1: Table S9).
## Discussion
Our study provides the following main findings. First, adherence to antidiabetic drug therapy reduced all-cause mortality in patients aged ≥ 65 years with type 2 diabetes. Second, this was the case regardless of the patients’ clinical status, i.e. antidiabetic treatment showed a protective effect not only in patients exhibiting relatively good clinical conditions but also in those with a progressive increase in the number of comorbidities, co-treatments, previous hospitalizations for a variety of diseases and a progressively marked reduction in the chance of survival, justifying their definition as “frail” individuals. Third, the reduction of all-cause mortality associated with better adherence to treatment was less pronounced in frail patients than in the other clinical categories. Furthermore, in frail patients there was no effect of antidiabetic treatment on survival from 85 years of age and beyond. These findings extend to the real-life setting the results of randomized clinical trials on the protective effect of antidiabetic drug treatment in patients aged ≥ 65 years [13]. They further suggest that protection extends to a very advanced age and that includes patients with a wide range of background clinical conditions and life expectancy, including those characterized by a high number of comorbidities, hospitalizations and risk of lethal events, which justifies their definition as frail patients. They also suggest, however, not only that in frail patients the protective effect of treatment on survival may be less than that seen in patients with better clinical conditions but fail to extend at or beyond 85 years of age.
Several other aspects of our study deserve to be mentioned. One, it is important to mention that “frail” patients aged 85 years or more were only 2644 out of 188,983 patients, i.e., about $1\%$ of our cohort. Thus, only a very small fraction of old patients with diabetes may eventually fail to benefit from antidiabetic treatment and be candidates to a deprescribing decision [31]. Two, the results of our study that an increase of adherence to antidiabetic drug treatment was associated with a reduction of all-cause mortality in octogenarians patients expands available information which is scarce in people with diabetes in this age range. Three, the results obtained by the analysis of adherence to treatment and cardiovascular mortality exhibited trends that were in general similar to the trends exhibited by all-cause mortality, with, however, a lower level of consistency and some between clinical and adherence group differences that were not seen for all-cause mortality. We can speculate that a factor involved was the low number of cardiovascular lethal events, and thus the insufficient statistical power, in some groups, e.g. 4 and 22 lethal events in patients with very low adherence and an age of 65–74 and ≥ 85 years, respectively. Assuming that the proportion of highly adherent patients among controls is $89\%$ (i.e. what was observed) and accepting a type 1 error of 0.05 with a statistical power of $80\%$, our study needed 1770 outcomes to detect a $25\%$ significant reduction of mortality risk in people with high adherence to drug therapy. This number was not available in all subgroups and more frequently lower for cardiovascular than for all-cause mortality, making the latter a safer basis for conclusions. Four, the reduction of outcomes associated with higher levels of adherence to treatment may be originated by factors different from the increase of adherence to treatment, for example from the fact that patients adherent to drug treatment may also be also more prone to follow healthier lifestyles and control their health conditions via more frequent medical visits and laboratory or instrumental examinations. However, although the number of outpatient medical services utilized by more adherent patients was greater than that utilized by non-adherent ones, our findings provide evidence that the difference in the risk of mortality between low and high adherence to antidiabetic drugs did not disappear after accounting for this proxy measure of health seeking behaviour. This supports the conclusion that an increased adherence to antidiabetic drugs was responsible for the associated protective effect. Finally, albeit restricting cohort members to new users is considered one of the best approaches to reduce confounding in observational studies assessing the effectiveness of drug therapies [30], this reduces the generalizability of study results, especially among elderly frail patients. Indeed, only $33\%$ of our cohort members did not have any prescription of antidiabetic drugs in the preceding years. Because the main results did not change by applying the new-user study design [30], this reinforces the robustness of the results.
Our study has several elements of strength as well as limitations. The strengths are that the study was based on a large and unselected population, which was made possible by the extension of the Italian healthcare system to virtually all citizens [32]. Furthermore, the drug prescription database we used provides accurate data because pharmacists are required to report prescriptions in detail in order to obtain reimbursement, and incorrect reports have legal consequences [33]. Finally, adoption of the “user-only” design (i.e. comparison between patients with the same indication at baseline, but with a different level of exposure to the drug of interest) reduces the potential for confounding [34]. Also, the choice of all-cause mortality as the primary outcome avoided any uncertainty about diagnostic accuracy in hospital records or causes of death reported in our database [35]. The limitations are that adherence to treatment was derived from drug prescriptions, a widely employed method to assess drug use in large populations which requires the assumption that the days covered by a prescription correspond to the days of drug use [32]. Because this is obviously not the case in all patients, our data on adherence to treatment are overestimated true adherence. Second, in absence of recorded daily doses of antidiabetic agents (not provided by our database [32]), we adopted the defined daily doses based on the reports of the World Health Organization (https://www.whocc.no/atc_ddd_index/) to estimate the time coverage of each prescription. However, the defined daily doses may overestimate the prescribed daily doses [27], making our adherence values lower than real adherence. However, it is unlikely that this had a substantial impact on the results because the main findings of the study were confirmed by a sensitivity analysis in which the drug coverage was estimated by the number of tablets in the dispensed canisters. Third, the inclusion and follow-up periods (from 2012 to 2018) did not allow us to suitably investigate the impact of the newer antidiabetic agents on mortality in different clinical categories and ages [9–12]. Between 2012 and 2018 treatment was still largely based on conventional antidiabetic drugs and only 3157 and 6771 cohort members started treatment with SGLT inhibitors and GLP1-RA, respectively.
Finally, and most importantly, because several clinical data (e.g. blood glucose glycated haemoglobin, lipid profile, blood pressure, body mass index, smoking, diet, duration of diabetes) are not included in the Lombardy database, we cannot exclude that a clinical imbalance between the adherence groups affected the results. However, our data were adjusted for several potential confounders. In addition, the “rule-out” sensitivity analysis showed that only a highly prevalent confounder closely associated with survival and extremely unbalanced between adherence groups could nullify the observed protection provided by greater adherence to antidiabetic drug therapy.
## Conclusions
In conclusion, adherence to antidiabetic drug therapy reduced the risk of death in elderly patients with diabetes, regardless of their background clinical conditions. The protective effect of treatment included patients definable as frail because of their high level of comorbidities, previous hospitalizations and short survival in whom in these patients the protective effect of antidiabetic treatment was less than that of patients in better clinical conditions but appreciable at least up to 85 years of age.
## Supplementary Information
Additional file 1: Table S1. Diagnostic and therapeutic codes used in the current study. Table S2. Comparison of demographic, clinical and therapeutic characteristics of the cohort members according to the clinical category. Table S3. Effect of adherence with oral antidiabetics on the risk of cardiovascular mortality according to categories of clinical frailty and age. Table S4. Number of cardiovascular deaths according to categories of clinical frailty and age. Table S5. Effect of adherence with antidiabetic drug therapy on the odds ratio (OR) of all-cause and cardiovascular death according to categorization of drug adherence by a different criterion than that used in the main analysis. Table S6. Effect of adherence with antidiabetic drug therapy on the odds ratio (OR) of all-cause death by assessing the coverage of each prescription from the number of tablets in the dispensed canister. Table S7. Effect of the number of outpatient services provided by the National Health Service on the odds ratio (OR) of adherence with antidiabetic drug therapy. Table S8. Effect of adherence with oral antidiabetics on the risk of all-cause mortality according to clinical categories of clinical frailty and number of outpatient services provided by the NHS in the previous two years. Table S9. Effect of adherence with oral antidiabetics on the risk of all-cause mortality by restricting the cohort to new users (i.e. patients with no antidiabetic drug prescriptions in the five years before the cohort entry). Figure S1. Flow-chart of inclusion and exclusion criteria for patients considered for data analysis.
## References
1. Zoungas S, Arima H, Gerstein HC, Holman RR, Woodward M, Reaven P, Hayward RA, Craven T, Coleman RL, Chalmers J. **Effects of intensive glucose control on microvascular outcomes in patients with type 2 diabetes: a meta-analysis of individual participant data from randomised controlled trials**. *Lancet Diabetes Endocrinol* (2017) **5** 431-437. DOI: 10.1016/S2213-8587(17)30104-3
2. Gerstein HC, Miller ME, Byington RP, Goff DC, Bigger JT, Buse JB, Cushman WC, Genuth S, Ismail-Beigi F, Grimm RH, Probstfield JL, Simons-Morton DG, Friedewald WT. **Effects of intensive glucose lowering in type 2 diabetes**. *N Engl J Med* (2008) **358** 2545-59. DOI: 10.1056/NEJMoa0802743
3. Patel A, MacMahon S, Chalmers J, Neal B, Billot L, Woodward M, Marre M, Cooper M, Glasziou P, Grobbee D, Hamet P, Harrap S, Heller S, Liu L, Mancia G, Mogensen CE, Pan C, Poulter N, Rodgers A, Williams B, Bompoint S, de Galan BE, Joshi R, Travert F. **Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes**. *N Engl J Med* (2008) **358** 2560-72. DOI: 10.1056/NEJMoa0802987
4. Duckworth W, Abraira C, Moritz T, Reda D, Emanuele N, Reaven PD, Zieve FJ, Marks J, Davis SN, Hayward R, Warren SR, Goldman S, McCarren M, Vitek ME, Henderson WG, Huang GD. **Glucose control and vascular complications in veterans with type 2 diabetes**. *N Engl J Med* (2009) **360** 129-39. DOI: 10.1056/NEJMoa0808431
5. Dormandy JA, Charbonnel B, Eckland DJ, Erdmann E, Massi-Benedetti M, Moules IK, Skene AM, Tan MH, Lefèbvre PJ, Murray GD, Standl E, Wilcox RG, Wilhelmsen L, Betteridge J, Birkeland K, Golay A, Heine RJ, Korányi L, Laakso M, Mokán M, Norkus A, Pirags V, Podar T, Scheen A, Scherbaum W, Schernthaner G, Schmitz O, Skrha J, Smith U, Taton J. **Secondary prevention of macrovascular events in patients with type 2 diabetes in the PROactive Study (PROspective pioglitAzone Clinical Trial In macroVascular Events): a randomised controlled trial**. *Lancet* (2005) **366** 1279-89. DOI: 10.1016/S0140-6736(05)67528-9
6. **Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33)**. *Lancet* (1998) **352** 837-53. DOI: 10.1016/S0140-6736(98)07019-6
7. Ray KK, Seshasai SR, Wijesuriya S, Sivakumaran R, Nethercott S, Preiss D, Erqou S, Sattar N. **Effect of intensive control of glucose on cardiovascular outcomes and death in patients with diabetes mellitus: a meta-analysis of randomised controlled trials**. *Lancet* (2009) **373** 1765-1772. DOI: 10.1016/S0140-6736(09)60697-8
8. Turnbull FM, Abraira C, Anderson RJ, Byington RP, Chalmers JP, Duckworth WC, Evans GW, Gerstein HC, Holman RR, Moritz TE, Neal BC, Ninomiya T, Patel AA, Paul SK, Travert F, Woodward M. **Intensive glucose control and macrovascular outcomes in type 2 diabetes**. *Diabetologia* (2009) **52** 2288-98. DOI: 10.1007/s00125-009-1470-0
9. Ristic S, Byiers S, Foley J, Holmes D. **Improved glycaemic control with dipeptidyl peptidase-4 inhibition in patients with type 2 diabetes: vildagliptin (LAF237) dose response**. *Diabetes Obes Metab* (2005) **7** 692-698. DOI: 10.1111/j.1463-1326.2005.00539.x
10. Zinman B, Wanner C, Lachin JM, Fitchett D, Bluhmki E, Hantel S, Mattheus M, Devins T, Johansen OE, Woerle HJ, Broedl UC, Inzucchi SE. **Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes**. *N Engl J Med* (2015) **373** 2117-28. DOI: 10.1056/NEJMoa1504720
11. Marso SP, Bain SC, Consoli A, Eliaschewitz FG, Jódar E, Leiter LA, Lingvay I, Rosenstock J, Seufert J, Warren ML, Woo V, Hansen O, Holst AG, Pettersson J, Vilsbøll T. **Semaglutide and cardiovascular outcomes in patients with type 2 diabetes**. *N Engl J Med* (2016) **375** 1834-1844. DOI: 10.1056/NEJMoa1607141
12. Lin DS, Lee JK, Hung CS, Chen WJ. **The efficacy and safety of novel classes of glucose-lowering drugs for cardiovascular outcomes: a network meta-analysis of randomised clinical trials**. *Diabetologia* (2021) **64** 2676-2686. DOI: 10.1007/s00125-021-05529-w
13. Karagiannis T, Tsapas A, Athanasiadou E, Avgerinos I, Liakos A, Matthews DR, Bekiari E. **GLP-1 receptor agonists and SGLT2 inhibitors for older people with type 2 diabetes: A systematic review and meta-analysis**. *Diabetes Res Clin Pract* (2021) **174** 108737. DOI: 10.1016/j.diabres.2021.108737
14. Bollig C, Torbahn G, Bauer J, Brefka S, Dallmeier D, Denkinger M, Eidam A, Klöppel S, Zeyfang A, Voigt-Radloff S. **Evidence gap on antihyperglycemic pharmacotherapy in frail older adults: a systematic review**. *Z Gerontol Geriatr* (2021) **54** 278-284. DOI: 10.1007/s00391-020-01724-3
15. Tufan F, Soyluk O, Bahat G, Karan MA. **Under-representation of frail older people in meta-analysis of dipeptidyl peptidase-4 inhibitors and hypoglycaemia**. *BMJ* (2016) **353** i3183. DOI: 10.1136/bmj.i3183
16. Corrao G, Rea F, Di Martino M, De Palma R, Scondotto S, Fusco D, Lallo A, Belotti LMB, Ferrante M, Pollina Addario S, Merlino L, Mancia G, Carle F. **Developing and validating a novel multisource comorbidity score from administrative data: a large population-based cohort study from Italy**. *BMJ Open* (2017) **7** e019503. DOI: 10.1136/bmjopen-2017-019503
17. Corrao G, Monzio Compagnoni M, Rea F, Merlino L, Catapano AL, Mancia G. **Clinical significance of diabetes likely induced by statins: evidence from a large population-based cohort**. *Diabetes Res Clin Pract* (2017) **133** 60-68. DOI: 10.1016/j.diabres.2017.08.008
18. Rea F, Ciardullo S, Savaré L, Perseghin G, Corrao G. **Comparing medication persistence among patients with type 2 diabetes using sodium-glucose cotransporter 2 inhibitors or glucagon-like peptide-1 receptor agonists in real-world setting**. *Diabetes Res Clin Pract* (2021) **180** 109035. DOI: 10.1016/j.diabres.2021.109035
19. Essebag V, Platt RW, Abrahamowicz M, Pilote L. **Comparison of nested case-control and survival analysis methodologies for analysis of time-dependent exposure**. *BMC Med Res Methodol* (2005) **5** 5. DOI: 10.1186/1471-2288-5-5
20. Essebag V, Genest J, Suissa S, Pilote L. **The nested case-control study in cardiology**. *Am Heart J* (2003) **146** 581-590. DOI: 10.1016/S0002-8703(03)00512-X
21. Corrao G, Rea F, Carle F, Di Martino M, De Palma R, Francesconi P, Lepore V, Merlino L, Scondotto S, Garau D, Spazzafumo L, Montagano G, Clagnan E, Martini N. **“Monitoring and assessing care pathways (MAP)” of the italian ministry of health measuring multimorbidity inequality across Italy through the multisource comorbidity score: a nationwide study**. *Eur J Public Health* (2020) **30** 916-921. DOI: 10.1093/eurpub/ckaa063
22. Andrade SE, Kahler KH, Frech F, Chan KA. **Methods for evaluation of medication adherence and persistence using automated databases**. *Pharmacoepidemiol Drug Saf* (2006) **15** 565-574. DOI: 10.1002/pds.1230
23. Suissa S. **Immeasurable time bias in observational studies of drug effects on mortality**. *Am J Epidemiol* (2008) **168** 329-335. DOI: 10.1093/aje/kwn135
24. Rea F, Cantarutti A, Merlino L, Ungar A, Corrao G, Mancia G. **Antihypertensive treatment in elderly frail patients: evidence from a large italian database**. *Hypertension* (2020) **76** 442-449. DOI: 10.1161/HYPERTENSIONAHA.120.14683
25. Rea F, Mancia G, Corrao G. **Statin treatment reduces the risk of death among elderly frail patients: evidence from a large population-based cohort**. *Eur J Prev Cardiol* (2022) **28** 1885-1894. DOI: 10.1093/eurjpc/zwaa126
26. Austin PC. **Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples**. *Stat Med* (2009) **28** 3083-3107. DOI: 10.1002/sim.3697
27. Arfè A, Nicotra F, Ghirardi A, Simonetti M, Lapi F, Sturkenboom M, Corrao G. **A probabilistic bias analysis for misclassified categorical exposures, with application to oral anti-hyperglycaemic drugs**. *Pharmacoepidemiol Drug Saf* (2016) **25** 1443-1450. DOI: 10.1002/pds.4093
28. Schneeweiss S. **Sensitivity analysis and external adjustment for unmeasured confounders in epidemiologic database studies of therapeutics**. *Pharmacoepidemiol Drug Saf* (2006) **15** 291-303. DOI: 10.1002/pds.1200
29. Danaei G, Tavakkoli M, Hernán MA. **Bias in observational studies of prevalent users: lessons for comparative effectiveness research from a meta-analysis of statins**. *Am J Epidemiol* (2012) **175** 250-262. DOI: 10.1093/aje/kwr301
30. Ray WA. **Evaluating medication effects outside of clinical trials: new-user designs**. *Am J Epidemiol* (2003) **158** 915-920. DOI: 10.1093/aje/kwg231
31. Farrell B, Black C, Thompson W, McCarthy L, Rojas-Fernandez C, Lochnan H, Shamji S, Upshur R, Bouchard M, Welch V. **Deprescribing antihyperglycemic agents in older persons: Evidence-based clinical practice guideline**. *Can Fam Physician* (2017) **63** 832-843. PMID: 29138153
32. Corrao G, Mancia G. **Generating evidence from computerized healthcare utilization databases**. *Hypertension* (2015) **65** 490-498. DOI: 10.1161/HYPERTENSIONAHA.114.04858
33. Strom BL, Strom BL. **Overview of automated databases in pharmacoepidemiology**. *Pharmacoepidemiology* (2000) 219-222
34. Corrao G, Ghirardi A, Segafredo G, Zambon A, Della Vedova G, Lapi F, Cipriani F, Caputi A, Vaccheri A, Gregori D, Gesuita R, Vestri A, Staniscia T, Mazzaglia G, Di Bari M. **User-only design to assess drug effectiveness in clinical practice: application to bisphosphonates and secondary prevention of fractures**. *Pharmacoepidemiol Drug Saf* (2014) **23** 859-867. DOI: 10.1002/pds.3650
35. Hyeraci G, Spini A, Roberto G, Gini R, Bartolini C, Lucenteforte E, Corrao G, Rea F. **A systematic review of case-identification algorithms based on Italian healthcare administrative databases for three relevant diseases of the cardiovascular system: acute myocardial infarction, ischemic heart disease, and stroke**. *Epidemiol Prev* (2019) **43** 37-50. PMID: 31650805
|
---
title: Moderation effects of serotype on dengue severity across pregnancy status in
Mexico
authors:
- Esther Annan
- Uyen-Sa D. T. Nguyen
- Jesús Treviño
- Wan Fairos Wan Yaacob
- Sherry Mangla
- Ashok Kumar Pathak
- Rajesh Nandy
- Ubydul Haque
journal: BMC Infectious Diseases
year: 2023
pmcid: PMC9999597
doi: 10.1186/s12879-023-08051-z
license: CC BY 4.0
---
# Moderation effects of serotype on dengue severity across pregnancy status in Mexico
## Abstract
### Background
Pregnancy increases a woman’s risk of severe dengue. To the best of our knowledge, the moderation effect of the dengue serotype among pregnant women has not been studied in Mexico. This study explores how pregnancy interacted with the dengue serotype from 2012 to 2020 in Mexico.
### Method
Information from 2469 notifying health units in Mexican municipalities was used for this cross-sectional analysis. Multiple logistic regression with interaction effects was chosen as the final model and sensitivity analysis was done to assess potential exposure misclassification of pregnancy status.
### Results
Pregnant women were found to have higher odds of severe dengue [1.50 ($95\%$ CI 1.41, 1.59)]. The odds of dengue severity varied for pregnant women with DENV-1 [1.45, ($95\%$ CI 1.21, 1.74)], DENV-2 [1.33, ($95\%$ CI 1.18, 1.53)] and DENV-4 [3.78, ($95\%$ CI 1.14, 12.59)]. While the odds of severe dengue were generally higher for pregnant women compared with non-pregnant women with DENV-1 and DENV-2, the odds of disease severity were much higher for those infected with the DENV-4 serotype.
### Conclusion
The effect of pregnancy on severe dengue is moderated by the dengue serotype. Future studies on genetic diversification may potentially elucidate this serotype-specific effect among pregnant women in Mexico.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12879-023-08051-z.
## Introduction
In Mexico, it is estimated that 139,000 symptomatic dengue fever (DF) cases occur yearly on average, with an estimated yearly cost of $170 million and an average annual disease burden of 65 disability-adjusted life years (DALYs) per million population [1]. The overlap between DF symptoms and the physiological alterations seen among women during pregnancy may make the identification of warning signs difficult [2]. However, for identified cases, pregnancy increases the risk of hospitalization and the development of severe dengue [2].
Maternal mortality rates (MMR) vary across and within regions, in Mexico [3]. MMR is associated with factors like pregnancy-related hypertension, obstetric hemorrhage, quality of health care [4], and infections like DF [5]. Severe DF is associated with a high rate of fetal distress, intrauterine death, obstetric hemorrhage, preeclampsia and eclampsia, caesarian section deliveries, and death due to multiple organ failure days after delivery [5, 6]. Prior studies have also implicated DF in vertical transmission during late pregnancy, and implicated serotypes have been serotypes 1 and 2 (DENV-1 and DENV-2) [7]. DF-specific serotypes have been linked with severe outcomes of DF. Severer complications have mostly been associated with the DENV-2 serotype [8]. There is, however, a gap in literature portraying how the severity of DF in pregnancy is modified by the DF serotype in Mexico.
Other factors related to severe outcomes of DF are comorbidities. Adults with self-reported hypertension have 1.6 times the odds of developing dengue hemorrhagic fever (DHF) compared to non-hypertensives [9], while diabetes is associated with 2.75 times the odds of DHF [10]. Diabetes presents with a far worse prognosis in Mexico compared to high-income countries [11]. Approximately $20\%$ of preventable deaths in Mexico are attributable to diabetes [12] and account for one-third of all mortality between the ages of 35 and 74 years [11]. Prior studies assert that the risk of dying among individuals hospitalized for dengue increases 11-fold when there are underlying comorbidities like diabetes [13]. A recent study done in Mexico, Brazil, and Colombia explored mortality associated with DF and found that comorbidities increase case fatality rates 3–17-fold [14].
This study aimed to explore (i) the moderation effects of DF serotype on pregnancy in causing severe DF. ( ii) The spatial distribution of severe and non-severe dengue across pregnancy status in Mexico. Findings from this will inform policies regarding the management of DF, particularly among pregnant women.
## Data collection
The dataset used in the analysis was retrieved from Mexico’s Ministry of Health and contains non-identifiable health information collected from notifying health units across 2469 Mexican municipalities from 2012 to 2020. The total sample size was 94,832 women.
## Definition of variables
Figure 1 shows a directed acyclic graph (DAG) of factors associated with pregnancy and dengue severity. Analysis was restricted to women of their reproductive age and defined by an age range of 15 to 49 years [15]. At each municipality clinic, women presenting with febrile illness characteristic of dengue fever were further tested for dengue antigens. An individual was defined as having dengue if she had a clinical diagnosis of DF and there was laboratory-confirmed evidence of non-structural protein (NS1) of DENV or a positive immunoglobulin M (IgM). DENV serotypes were determined based on polymerase chain reaction RT-qPCR results [16]. An individual with confirmed dengue had either DENV-1, DENV-2, DENV-3, or DENV-4 serotype. An individual with dengue was reported as having non-severe dengue, severe dengue, dengue without warning signs, dengue with warning signs, or ‘other’. Individuals with no dengue classification or ‘other’ classification were excluded from the analysis. The World Health Organization’s revised 2009 classification of DF emphasizes the inclusion of warning signs as a diagnostic criterion for probable and potentially severe dengue [17]. However, this classification requires laboratory-confirmed results to prevent inflation of the number of severe dengue cases. Because our dataset contained both clinical and laboratory-confirmed diagnoses, severe dengue was defined as individuals with severe dengue or having dengue with warning signs, while non-severe dengue was defined as individuals ‘having non-severe dengue or dengue without warning signs’. A woman was identified as pregnant or not pregnant based on pregnancy status classification retrieved from the dataset. Region was categorized as Center, Center West, Northeast, Northwest, and Southeast. Classification of the region has been defined elsewhere [18]. Hypertension and Diabetes were binary variables with ‘1’ indicating the presence of disease and ‘0’ indicating the absence of disease. Fig. 1A directed acyclic graph of factors associated with pregnancy and dengue severity
## Analysis
i)Model selection exploring the moderation effect of DENV serotype on dengue severity across pregnancy status A multicollinearity test was performed and tolerance for hypertension (0.93) and diabetes (0.93) showed no multicollinearity between them. Similarly, there was no multicollinearity observed among the other covariates. Year and region were considered as potential random effects and were evaluated in a two-level hierarchical model. The Wald tests for random effects for the region ($$p \leq 0.049$$) and year ($$p \leq 0.032$$) were both significant. A comparison between random and fixed effects model using the Bayesian information criteria (BIC) revealed that the random-effects model performed better. Hence, both region and year were retained in the model as random effects. Based on the hierarchical model [1], a multiple logistic regression was performed with severe or non-severe DF as the outcome variable and pregnancy as the exposure variable.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$y=X*\beta +Z*u+{\varvec{\varepsilon}}$$\end{document}y=X∗β+Z∗u+ε where X a matrix (N *p) with p predictor variables and Z is a matrix (N*q) for q random effects [19].
Two-way interactions were explored, and multiple interactions were found to be statistically significant and retained. Other covariates in the model included dengue serotype, diabetes, hypertension, and age. The final model was chosen by comparing AIC and BIC between models. The AUC for the predicted probabilities of the final model was 0.7156 (Additional file 1: Fig. S1).
## Sensitivity analysis
The dataset shows the results of individuals who were tested for pregnancy, diabetes, and hypertension. However, individuals without a test could potentially have been misclassified as ‘negative’. To quantitatively assess for this kind of systematic error, a sensitivity analysis is recommended [20]. Using a misclassification spreadsheet [21], results from the regression analysis were explored for exposure bias. The misclassification spreadsheet provides adjusted bias data based on observed data on pregnant women stratified by dengue severity. As suggested for best practices [22], pairs of sensitivity and specificity were explored to study exposure misclassification.ii)*Spatial analysis* The sums of severe and non-severe dengue cases were calculated by pregnancy status. The spatial distribution of severe and non-severe dengue was visualized and compared across pregnancy status using ArcGIS. To measure spatial autocorrelation, Moran’s Index (I) was calculated for both pregnant and non-pregnant women with severe dengue. Moran’s I for the attributes pregnant (0.010, $$p \leq 0.13$$) and non-pregnant women (0.003, $$p \leq 0.54$$) were both not statistically significant, and the spatial distributions of these two attributes were random.
## Coding and environment
Data preprocessing, analysis, and generation of figures were done using SAS (version 9.4, SAS Institute Inc., Cary, NC, USA), R (version 4.1.2, The R Foundation, Vienna, Austria), and STATA/SE (Stata Corp LLC, College Station, TX, USA). All codes can be found in the Additional file.
## Sample size determination
A study conducted in Brazil found that 4 out of 707 (0.006) pregnant women and 19 out of 15,576 (0.001) non-pregnant had severe dengue [2]. To determine such an association with a power of 0.80, at an alpha of 0.05, a total sample size of 4547 was needed. The sample size was determined using the G*Power 3.1.9.4 software. Our study was comprised of 94,832 women of their reproductive age, of which 4943 were pregnant ($5.21\%$) and 25,018 ($26.38\%$) of them had severe dengue.
## Ethical approval
This study was reviewed and approved by the ethics and research committee of the Universidad Autónoma de Nuevo León and “North Texas Regional Institutional Review Board” as an exempt category (reference # 2021-035). All methods were performed in accordance with the DECLARATION OF HELSINKI guidelines for reporting observational studies. The need for informed consent was waived by the North Texas Regional IRB because the data analyzed was aggregated, de-identified and delinked, and therefore, obtaining informed consent was not applicable.
## Results
Table 1 shows the distribution of the sample of women from 2012 to 2020. The average age was 29 years old, with most women living in the Southeast region ($50.12\%$) of Mexico. Across regions, DENV-2 was the commonest serotype found among individuals with severe DF (Additional file 2: Fig. S2). Compared to the other four regions, pregnant women in the Northeast region had the highest proportion of DENV-2 serotype (Additional file 3: Fig. S3). The chi-square statistic for the differences observed across various regions for dengue severity was statistically significant (χ2 = 2782.29, $p \leq 0.0001$). Among pregnant women, $33.57\%$ had severe dengue, compared to $25.23\%$ of non-pregnant women (Additional file 6: Table S1). The difference observed between pregnancy status for dengue classification was statistically significant (χ2 = 187.12, $p \leq 0.0001$).Table 1Demographic characteristics by dengue classificationParameterDengue classification, n (%)Severe dengueNon-severe dengueAllRegion**Center1968 ($7.87\%$)6490 ($9.30\%$)8458 ($8.92\%$)Center-West4805 ($19.21\%$)18580 ($26.61\%$)23385 ($24.66\%$)North-East2084 ($8.33\%$)9856 ($14.12\%$)11940 ($12.59\%$)North-West1817 ($7.26\%$)8059 ($11.54\%$)9876 ($10.41\%$)South-East14344 ($57.33\%$)26829 ($38.43\%$)41173 ($43.42\%$)Pregnancy status**Pregnant1707 ($6.82\%$)3236 ($4.64\%$)4943 ($5.21\%$)Not Pregnant23311 ($93.18\%$)66578 ($95.36\%$)89889 ($94.79\%$)Age in years (SD)**29.46 (± 9.87)29.61 (± 9.75)29.57 (± 9.78)Year**20126031 ($24.11\%$)11560 ($16.56\%$)17591 ($18.55\%$)20136251 ($24.99\%$)15547 ($22.27\%$)21798 ($22.99\%$)20142873 ($11.48\%$)8667 ($12.41\%$)11540 ($12.17\%$)20151930 ($7.71\%$)8591 ($12.31\%$)10521 ($11.09\%$)20169 ($0.04\%$)20 ($0.03\%$)29 ($0.03\%$)2017825 ($3.30\%$)4580 ($6.56\%$)5405 ($5.70\%$)20181085 ($4.34\%$)3162 ($4.53\%$)4247 ($4.48\%$)20194471 ($17.87\%$)10325 ($14.79\%$)14796 ($15.60\%$)20201543 ($6.17\%$)7362 ($10.55\%$)8905 ($9.39\%$)Total (n)25018 ($26.38\%$)69814 ($73.62\%$)94832 ($100.00\%$)**Chi-square or t-test performed for group differences had $p \leq 0.001$ Dengue severity has had both downward and upward trends from 2012 to 2020 (Additional file 4: Fig. S4). Among women with severe dengue, DENV-2 was the commonest variant, while DENV-1 was the commonest variant among those with non-severe dengue (Table 2). There was a similar distribution by pregnancy category; while most pregnant women had DENV-2, non-pregnant women mostly had the DENV-1 variant (Additional file 6: Table S1). DENV-2 was the commonest serotype in the Southeast region, while DENV-1 was the commonest serotype among women in the other regions of Mexico (Additional file 6: Table S2). The Southeast region had the highest proportion of severe dengue cases compared to other regions and this difference was statistically significant ($p \leq 0.0001$) (Additional file 6: Table S2).Table 2Proportions and p-value for different Chi-squared test results with $95\%$ significanceParameterDengue classification, n (%)Chi-square p-valueNon-severe dengueSevere dengueSerotypeDENV-19825 ($53.65\%$)1887 ($31.16\%$)973 (<.0001)DENV-28145 ($44.48\%$)4081 ($67.40\%$)DENV-3167 ($0.91\%$)63 ($1.04\%$)DENV-4175($0.96\%$)24 ($0.40\%$)HypertensionHypertension249 ($0.36\%$)465 ($1.86\%$)556 (<.0001)No Hypertension69565 ($99.64\%$)24553 ($98.14\%$)DiabetesDiabetes342 ($0.49\%$)566 ($2.26\%$)610 (<.0001)No Diabetes69472 ($99.51\%$)24452 ($97.74\%$)IgGPositive12655 ($83.45\%$)2754 ($93.17\%$)183 (<.0001)Negative2509 ($16.55\%$)202 ($6.83\%$)IgMPositive18388 ($80.49\%$)11594 ($95.42\%$)1441 (<.0001)Negative4457 ($19.51\%$)556 ($4.58\%$) Additional file 5: Fig. S5 shows variations in severe dengue prevalence from 2012 to 2020 across Mexican states and pregnancy status. Both pregnancy strata showed a similar pattern of spread of severe dengue, although the number of cases in non-pregnant women was higher. A look at the proportions between severe and non-severe dengue for each pregnancy strata shows variations across states (Fig. 2). While most states recorded higher counts of non-severe dengue compared with severe dengue, Chiapas in the Southeast region, and Nayarit in the Center west region had a higher prevalence of severe dengue for pregnant women. This was contrasted with non-pregnant women who had similar proportions across severity strata in both Chiapas and Nayarit. Fig. 2Dengue distribution among pregnant and non-pregnant women from 2012 to 2020 Unadjusted odds of a pregnant woman experiencing severe dengue were 1.5 times the odds of that of non-pregnant women. When adjusted for in a multiple logistic regression model, serotype moderated the effect of pregnancy. Respectively, among individuals with DENV-1, DENV-2, and DENV-4, pregnant women had 1.45, 1.35, and 3.78 times the odds, of severe dengue, compared to non-pregnant women after adjusting for other variables. Compared to those living in the Southeast region, individuals living in the Center, Center West, Northeast, and Northwest regions, had 0.38, 0.31, 0.37, and 0.66 times the odds of severe dengue (Table 3). Based on the random effects from yearly variations, on average, severe DF cases were higher in 2014 and lower in 2017 and 2020, compared to non-severe dengue (Additional file 6: Table S3). Similarly, severe DF cases were significantly higher in the Southeast and Center west regions. A one-unit increase in age was associated with 0.992 times lower odds of severe dengue after adjusting for other variables. All else equal, individuals diagnosed with diabetes had 2.6 times the odds, while those with hypertension had about 3.0 times the odds of severe dengue compared to those without diabetes and hypertension respectively. The variance seen among individuals with DENV-4 between pregnant and non-pregnant women is higher compared to the other serotypes. At a specificity of $99\%$, and sensitivity of $90\%$ the adjusted OR for pregnancy was 1.63 (CI 1.52, 1.75) (Additional file 6: Table S4) compared to the unadjusted OR of 1.49 (CI 1.41, 1.59).Table 3Univariate and multivariate logistic regression showing predictors of the odds of dengue severityIndependent VariablesUnivariateMultivariateOR$95\%$ CI*p-valueAOR$95\%$ CIp-valuePregnant Yes, vs No1.4971.413, 1.5870.00011.420.95, 2.11<.0001Age0.9970.996, 0.9990.00030.990.99, 0.99<.0001Year0.9420.937, 0.9470.0001–––Serotype DENV-2 vs DENV-12.5702.419, 2.7310.00012.952.63, 3.30<.0001 DENV-3 vs DENV-11.7931.368, 2.3500.00010.940.56, 1.560.0128 DENV-4 vs DENV-10.6770.441,1.0380.07370.590.32, 1.08<.0001Diabetes Yes, vs No4.5654.014, 5.1930.00012.581.95, 3.40<.0001Hypertension Yes, vs No5.3654.620, 6.2290.00012.972.14, 4.10<.0001Region Center vs Southeast0.5650.535, 0.5970.0001––– Center West vs Southeast0.4820.464, 0.5000.0001––– Northeast vs Southeast0.3960.376, 0.4170.0001––– Northwest vs Southeast0.4220.400, 0.4460.0001–––Pregnant vs non-pregnant DENV-1–––1.451.2, 1.7<.0001 DENV-2–––1.331.2, 1.5<.0001 DENV-3–––0.550.2, 1.50.2438 DENV-4–––3.781.1, 12.60.0305*p-values are adjusted, using false discovery rate (FDR)
## Discussion
Our study found that the association between pregnancy and severe DF is moderated by the DENV serotype. The effect of DENV-4 in pregnant women may indicate effects of genetic diversification and the emergence of new serotype-specific genotypes [23, 24] and this warrants further investigation. Further, compared to other regions, the Southeast region had higher odds of severe DF. Dengue control programs and policies need to be expanded, using a multidisciplinary approach across Mexico.
Although a previous study found individuals with DENV-2 to have a lower risk of dengue hemorrhagic fever [25], the association between DENV-2 and higher risk of severe dengue, particularly when compared to DENV-1 [26], is consistent with most literature [26–29]. Similarly, DENV-2 and DENV-3 are more commonly associated with severe dengue compared to DENV-4 [26]. A recent study in Brazil, found that pregnant women had 1.92 times the odds of having DENV-4 serotype compared to non-pregnant women [2]. However, the authors concluded that the persistence of DENV-4 in a region and a higher number of cases in particular years may have explained the results found [2]. Our study’s finding of similar associations in Mexico may point to other potential mechanisms inherent in serotype-specific variations and/or their interaction with the pregnancy status. Another potential mechanism could involve the presence of DENV-4 as a heterologous infecting serotype. A high predominance of DENV-4 has previously been implicated in an outbreak in Jember, Indonesia, an area that did not frequently report DENV-4 but previously had outbreaks due to DENV-1, DENV-2, and DENV-3 serotypes [30].
Pregnant women are generally more at risk and predisposed to certain clinal conditions [31]. The risk of hospitalization as well as DF severity tends to be higher among pregnant women compared to non-pregnant women [2]. A higher proportion of IgG-positive serology compared to IgM in pregnant women might point to a higher risk of severe DF among those with secondary infection. While a host’s genetic background and immune status may influence disease presentation, it is suggested that certain viral structures may aid in replication in human target cells [32]. Differences among DENV serotypes may be attributed to genotype-specific (within serotypes) variances [25, 32]. One potential explanation is genetic diversity from clade replacements [33] which may be independent of pregnancy status. A clade replacement of a DENV serotype may be associated with a decrease or an increase in the prevalence of a heterologous DENV serotype [34]. Although the evidence does not support the transmission of antigenically aberrant strains, prior research suggests the displacement of DENV genotypes of less epidemiological significance by more virulent genotypes [32]. Genetic diversification and an emergence of a DENV-4 genotype-I have been found in a molecular analysis in Brazil and parts of tropical and subtropical America [23]. This may explain the moderation effect of DENV-4 serotype specifically among pregnant women compared to non-pregnant women. This is contrary to what is expected in the general population, where DENV-2 has been mostly associated with severer outcomes and DENV-4 has generally been associated with causing clinically mild diseases [25, 28]. It is also worth noting that although the effect of DENV-4 was statistically significant, the large confidence interval and smaller sample size may have influenced this finding. Hence, future prospective studies which may involve phylogenetic analysis or gene sequencing may further explore DENV-4 specific effects among pregnant women and the moderation effect of dengue serotypes in pregnancy.
The changing trend of increasing and decreasing cases of severe dengue may indicate a change in programs/policies associated with dengue fever eradication. The local health system mainly spearheads the charge toward dengue prevention programs, with minimal effort from other sectors like water and sanitation [35]. When comparing regions, women in the Southeast regions had higher odds of severe dengue compared to those in other regions. Central regions on average had lower proportions of severe dengue. This may be explained by the fact that cities like Mexico City in the Center region are free of endemic mosquito-borne viral diseases [36]. With a subtropical climate and high elevation, there is a lower occurrence of Aedes spp. in Central Mexico [36]. On the contrary, Pacific and Coastal regions tend to be at a higher risk of dengue [37]. For instance, states like Oaxaca in the pacific and southeastern region is one among the most affected states in Mexico, and the persistence of high dengue cases has been attributed to the presence of all four serotypes, favorable climate, and socioeconomic level of the population [38].
Our study had several limitations. Firstly, only a confirmed diagnosis of pregnancy status was reported in the dataset. Hence, to address potential misclassification bias, a sensitivity analysis was performed to assess the misclassification of pregnancy status. The analysis showed that at a sensitivity and specificity higher than $80\%$ and $97\%$ respectively, findings from our study were conservative. Another limitation was missing data for all variables. Particularly, prevalence estimates for pregnant women in Zacatecas and Chihuahua were missing across the period of study. However, our sample size of 94,832 women, satisfied the requirement for this study. Also, the cross-sectional nature of the study limits the inferential interpretation of results from the logistic regression model. The unavailability of information about vaccination status and behavioral factors that affect mosquito control also limits the study’s ability to control for these confounders. However, the spatial trend analysis provides additional reasons for further exploration in future studies. Furthermore, since pregnant women are more likely than other women to visit the clinic for antenatal care and be hospitalized, the likelihood of being diagnosed with dengue might be higher than non-pregnant women. However, this assumes a high antenatal care uptake and secondly, it also assumes that serological tests are performed routinely for pregnant women. Serological tests are performed upon clinical diagnosis of dengue. Since most dengue cases are asymptomatic and might go unnoticed, an acute presentation with a febrile illness among women is likely to present to the clinic regardless of pregnancy status. Lastly, restriction of the data to only women with serotype data means the likely exclusion of more people with non-severe dengue. However, the proportions of missingness among those with severe ($76\%$) and non-severe ($74\%$) dengue were similar. I higher proportion of missingness among non-pregnant women ($75\%$) compared to pregnant women ($60\%$), however, this supports the theory that more frequent access to healthcare may influence more diagnoses among pregnant women. Future studies may prospectively collect serotype data to ensure further limitation of potential bias.’ has been added to the limitations.
## Conclusion
Pregnancy increases a woman’s risk of severe dengue. However, this may be modified by the DENV-specific serotype. Of note is the DENV-4 serotype, which is otherwise the least severe serotype in the general population. Across Mexican regions, the southeast region had the highest number of severe dengue cases. Particularly, perinatal care in states like Chiapas and Nayarit may warrant further surveillance. This may especially be important for individuals with comorbid conditions like hypertension and diabetes and under the age of 24 years. An intersectoral approach is still needed across Mexico, particularly in the Southeast region to address the risk of DF severity. Further research is needed to fully understand the moderation effect of dengue serotype in pregnancy.
## Supplementary Information
Additional file 1: Figure S1. ROC Curve for Hierarchical logistic regression model. Additional file 2: Figure S2. Distribution of dengue serotype by Dengue Severity and region. Additional file 3: Figure S3. Distribution of dengue serotype by *Pregnancy status* and region. Additional file 4: Figure S4. Trend of dengue fever among Mexican women in their reproductive age. Additional file 5: Figure S5. Prevalence of severe dengue among pregnant and non-pregnant women from 2012 to 2020Additional file 6: Table S1. Proportions of parameters by pregnancy status and p-value for different Chi-Squared Test results with $95\%$ significance. Table S2. Proportions of parameters by region and p-value for different Chi-Squared Test results with $95\%$ significance. Table S3. Random effects output from logistic regression. Table S4. Sensitivity Analysis for the effect of pregnancy on severe dengue.
## References
1. Undurraga EA. **Economic and disease burden of dengue in Mexico**. *PLoS Negl Trop Dis* (2015.0) **9** e0003547. DOI: 10.1371/journal.pntd.0003547
2. Martin BM. **Clinical outcomes of dengue virus infection in pregnant and non-pregnant women of reproductive age: a retrospective cohort study from 2016 to 2019 in Paranám Brazil**. *BMC Infect Dis* (2022.0) **22** 5-5. DOI: 10.1186/s12879-021-06985-w
3. Pisanty-Alatorre J. **Inequity in maternal mortality in Mexico: analyzing inequality on a sub-state regional scale**. *Salud Publica Mex* (2017.0) **59** 639-649. DOI: 10.21149/8788
4. Ordaz-Martínez KY, Rangel R, Hernández-Girón C. **Risk factors associated with maternal mortality in the State of Morelos, Mexico**. *Ginecol Obstet Mex* (2010.0) **78** 357-364. PMID: 20931812
5. Machain-Williams C. **Maternal, fetal, and neonatal outcomes in pregnant dengue patients in Mexico**. *Biomed Res Int* (2018.0) **2018** 9643083. DOI: 10.1155/2018/9643083
6. Kariyawasam S, Senanayake H. **Dengue infections during pregnancy: case series from a tertiary care hospital in Sri Lanka**. *J Infect Dev Ctries* (2010.0) **4** 767-775. DOI: 10.3855/jidc.908
7. Phongsamart W. **Dengue virus infection in late pregnancy and transmission to the infants**. *Pediatr Infect Dis J* (2008.0) **27** 500-504. DOI: 10.1097/INF.0b013e318167917a
8. Chang K. **Differences in mortality and clinical manifestations of dengue hemorrhagic fever in Taiwan in different years: a comparison for cases in 2014 and 2015 epidemics**. *Am J Trop Med Hyg* (2017.0) **97** 361-368. DOI: 10.4269/ajtmh.16-1018
9. Teixeira MG. **Arterial hypertension and skin allergy are risk factors for progression from dengue to dengue hemorrhagic fever: a case control study**. *PLoS Negl Trop Dis* (2015.0) **9** e0003812. DOI: 10.1371/journal.pntd.0003812
10. Figueiredo MA. **Allergies and diabetes as risk factors for dengue hemorrhagic fever: results of a case control study**. *PLoS Negl Trop Dis* (2010.0) **4** e699. DOI: 10.1371/journal.pntd.0000699
11. Alegre-Díaz J. **Diabetes and cause-specific mortality in Mexico City**. *N Engl J Med* (2016.0) **375** 1961-1971. DOI: 10.1056/NEJMoa1605368
12. Bello-Chavolla OY. **Epidemiology of diabetes mellitus in Mexico**. *Nutr Rev* (2017.0) **75** 4-12. DOI: 10.1093/nutrit/nuw030
13. Werneck GL. **Comorbidities increase in-hospital mortality in dengue patients in Brazil**. *Mem Inst Oswaldo Cruz* (2018.0) **113** e180082. DOI: 10.1590/0074-02760180082
14. Macias AE. **Mortality among hospitalized dengue patients with comorbidities in Mexico, Brazil, and Colombia**. *Am J Trop Med Hyg* (2021.0) **105** 102. PMID: 33970884
15. 15.Organization, W.H. Women of reproductive age (15-49 years) population (thousands). 2022; https://www.who.int/data/gho/indicator-metadata-registry/imr-details/women-of-reproductive-age-(15-49-years)-population-(thousands). Accessed 25 Jan 2022.
16. 16.Secretaría de Salud, M. Lineamientos Para La Vigilancia Por Laboratorio Del Dengue Y Otras Arbovirosis, Indre (Instituto De Diagnóstico Y Referencia Epidemiológicos ‘Dr. Manuel Martínez BÁEZ’). 2021; https://www.gob.mx/cms/uploads/attachment/file/629265/Lineamientos_Dengue_Arb_V1-2021.pdf.
17. Srikiatkhachorn A. **Dengue–how best to classify it**. *Clin Infect Dis* (2011.0) **53** 563-567. DOI: 10.1093/cid/cir451
18. Ananth S. **Clinical symptoms of arboviruses in Mexico**. *Pathogens* (2020.0) **9** 964. DOI: 10.3390/pathogens9110964
19. 19.Analytics, U.A.R.C.S.M.a.D. Introduction to Generalized Linear Mixed Models. 2021; https://stats.oarc.ucla.edu/other/mult-pkg/introduction-to-generalized-linear-mixed-models/. 4 Mar 2022.
20. Lash TL, Fink AK. **Semi-automated sensitivity analysis to assess systematic errors in observational data**. *Epidemiology* (2003.0) **14** 451. DOI: 10.1097/01.EDE.0000071419.41011.cf
21. 21.Analysis, B. Applying quantitative bias analysis to epidemiologic data. n.d.; https://sites.google.com/site/biasanalysis/Home. 28 Jan 2022.
22. Lash TL. **Good practices for quantitative bias analysis**. *Int J Epidemiol* (2014.0) **43** 1969-1985. DOI: 10.1093/ije/dyu149
23. Hernández-García E. **Epidemiological implications of the genetic diversification of dengue virus (DENV) serotypes and genotypes in Mexico**. *Infect Genet Evol* (2020.0) **84** 104391. DOI: 10.1016/j.meegid.2020.104391
24. Shihada S. **Genetic diversity and new lineages of dengue virus serotypes 3 and 4 in returning travelers, Germany, 2006–2015**. *Emerg Infect Dis* (2017.0) **23** 272-275. DOI: 10.3201/eid2302.160751
25. Yung C-F. **Dengue serotype-specific differences in clinical manifestation, laboratory parameters and risk of severe disease in adults, singapore**. *Am J Trop Med Hyg* (2015.0) **92** 999-1005. DOI: 10.4269/ajtmh.14-0628
26. Fried JR. **Serotype-specific differences in the risk of dengue hemorrhagic fever: an analysis of data collected in Bangkok, Thailand from 1994 to 2006**. *PLoS Negl Trop Dis* (2010.0) **4** e617. DOI: 10.1371/journal.pntd.0000617
27. Balmaseda A. **Serotype-specific differences in clinical manifestations of dengue**. *Am J Trop Med Hyg* (2006.0) **74** 449-456. DOI: 10.4269/ajtmh.2006.74.449
28. Nisalak A. **Serotype-specific dengue virus circulation and dengue disease in Bangkok, Thailand from 1973 to 1999**. *Am J Trop Med Hyg* (2003.0) **68** 191-202. DOI: 10.4269/ajtmh.2003.68.191
29. Vaughn DW. **Dengue viremia titer, antibody response pattern, and virus serotype correlate with disease severity**. *J Infect Dis* (2000.0) **181** 2-9. DOI: 10.1086/315215
30. Aryati A. **dengue virus serotype 4 is responsible for the outbreak of dengue in East Java City of Jember, Indonesia**. *Viruses* (2020.0) **12** 913. DOI: 10.3390/v12090913
31. Leeper C, Lutzkanin A. **Infections during pregnancy**. *Prim Care* (2018.0) **45** 567-586. DOI: 10.1016/j.pop.2018.05.013
32. Rico-Hesse R. **Microevolution and virulence of dengue viruses**. *Adv Virus Res* (2003.0) **59** 315-341. DOI: 10.1016/S0065-3527(03)59009-1
33. Suzuki K. **Genotype replacement of dengue virus type 3 and clade replacement of dengue virus type 2 genotype Cosmopolitan in Dhaka, Bangladesh in 2017**. *Infect Genet Evol* (2019.0) **75** 103977. DOI: 10.1016/j.meegid.2019.103977
34. Zhang C. **Clade replacements in dengue virus serotypes 1 and 3 are associated with changing serotype prevalence**. *J Virol* (2005.0) **79** 15123-15130. DOI: 10.1128/JVI.79.24.15123-15130.2005
35. González Fernández MI, Orozco Núñez E, Cifuentes E. **Policy analysis of the dengue control program in Mexico**. *Rev Saude Publica* (2010.0) **44** 1079-1086. DOI: 10.1590/S0034-89102010005000040
36. Davalos-Becerril E. **Urban and semi-urban mosquitoes of Mexico City: a risk for endemic mosquito-borne disease transmission**. *PLoS ONE* (2019.0) **14** e0212987. DOI: 10.1371/journal.pone.0212987
37. 37.Travellers, I.A.f.M.A.t. Mexico General Health Risks: Dengue. 2020; https://www.iamat.org/country/mexico/risk/dengue#. 22 Feb 2022.
38. Günther J. **Distribution of dengue cases in the state of Oaxaca, Mexico, during the period 2004–2006**. *J Clin Virol* (2009.0) **45** 218-222. DOI: 10.1016/j.jcv.2009.05.007
|
---
title: 'A cross-sectional study: a breathomics based pulmonary tuberculosis detection
method'
authors:
- Liang Fu
- Lei Wang
- Haibo Wang
- Min Yang
- Qianting Yang
- Yi Lin
- Shanyi Guan
- Yongcong Deng
- Lei Liu
- Qingyun Li
- Mengqi He
- Peize Zhang
- Haibin Chen
- Guofang Deng
journal: BMC Infectious Diseases
year: 2023
pmcid: PMC9999612
doi: 10.1186/s12879-023-08112-3
license: CC BY 4.0
---
# A cross-sectional study: a breathomics based pulmonary tuberculosis detection method
## Abstract
### Background
Diagnostics for pulmonary tuberculosis (PTB) are usually inaccurate, expensive, or complicated. The breathomics-based method may be an attractive option for fast and noninvasive PTB detection.
### Method
Exhaled breath samples were collected from 518 PTB patients and 887 controls and tested on the real-time high-pressure photon ionization time-of-flight mass spectrometer. Machine learning algorithms were employed for breathomics analysis and PTB detection mode, whose performance was evaluated in 430 blinded clinical patients.
### Results
The breathomics-based PTB detection model achieved an accuracy of $92.6\%$, a sensitivity of $91.7\%$, a specificity of $93.0\%$, and an AUC of 0.975 in the blinded test set ($$n = 430$$). Age, sex, and anti-tuberculosis treatment does not significantly impact PTB detection performance. In distinguishing PTB from other pulmonary diseases ($$n = 182$$), the VOC modes also achieve good performance with an accuracy of $91.2\%$, a sensitivity of $91.7\%$, a specificity of $88.0\%$, and an AUC of 0.961.
### Conclusions
The simple and noninvasive breathomics-based PTB detection method was demonstrated with high sensitivity and specificity, potentially valuable for clinical PTB screening and diagnosis.
## Key messages
What is already known on this topic—Breath VOC analysis is a potential technology for PTB detection. However, it is still desirable for a real-time, robust, accurate, and simple breath analysis platform for clinical application. What this study adds—An online breath detection for PTB was proposed and demonstrated with high sensitivity and specificity in a large clinical cohort. How this study might affect research, practice, or policy—This study may promote the application of breath detection in clinical TB detection and related biomarker studies.
## Introduction
Tuberculosis (TB) continues to be a major global health threat, with an estimated 10 million incident cases and 1.4 million deaths per year globally. In 2019, only $57\%$ of pulmonary TB cases were confirmed by bacteriological examination. There is still a large gap, 2.9 million cases in 2019, between reported and estimated cases [1]. The absence of available technology for the timely and accurate detection of TB has been one of the major impediments to preventing and ending TB. Undiagnosed TB is associated with substantial morbidity and mortality and leads to ongoing TB transmission in the community, which makes improving the performance and delivery of diagnostic testing services a leading priority [2].
Sputum-based TB diagnostics are usually either inaccurate, expensive, or complicated in their usage [3]. Sputum specimens are difficult to collect, process, and transport, and only one-third of suspected TB patients can give adequate high-quality sputum samples [4], while it is even harder in children, HIV-infected patients, and those with extrapulmonary TB. Acid-fast bacilli staining of sputum has a high false-negative rate (up to $50\%$) [5]. The culture of sputum alone has a poor sensitivity of approximately $30\%$ [6, 7]. GeneXpert MTB/RIF (Xpert) achieved good performances in TB detection and drug resistance testing in the clinic and has been recommended by the WHO. However, it still requires good infrastructure and sputum samples [8, 9]. WHO has identified four high-priority test types for diagnostic development and created target product profiles (TPPs) for each, among which some non-sputum tests should be offered [10]. Thus, there is a greater need than ever for fast, accurate, and non-sputum TB detection technologies.
Breathomics, a branch of metabolomics, is a promising tool because of its significant advantages: good accessibility, noninvasiveness, and specificness [11, 12]. A breath test could diagnose TB by detecting volatile organic compounds (VOCs) produced by mycobacterium tuberculosis (M.tb) and the infected host, which has been approved by many studies [13]. The most commonly used breath detection methods for TB diagnosis include gas chromatography–mass spectrometry (GC–MS) [14, 15] and electric or chemical sensors [16]. For GC–MS based studies, Phillips et al. used GC–MS to detect the VOCs in the exhalation of pulmonary TB (PTB) patients with positive culture results and healthy controls (HC), and the headspace air of M.tb culture flask. They found that patients' expiratory VOCs were similar to culture VOCs in naphthalene, 1-methyl- and cyclohexane, 1,4-dimethyl-. Based on the small sample modeling on 12 identified VOCs, the author obtained a sensitivity of $82.6\%$ and specificity of $100\%$, which verified the feasibility of the breath test for PTB detection [17]. They further validated the VOCs-based PTB detection method within a larger transcontinental and ethnic group of 226 symptomatic high-risk patients in United States, Philippines, and United Kingdom, which achieves an overall accuracy of approximately $85\%$ [18]. Beccaria et al. also used GC–MS to analyze the VOCs of exhaled breath of patients with active PTB and health controls in South Africa, achieving a sensitivity of $100\%$ and specificity of $60\%$ via the random forest method [19]. In addition, they performed another validation study using two-dimensional GC–MS for breath analysis on PTB and PTB-free patients in Haiti and found that a random forest model based on 22 characteristics VOCs can distinguish well between PTB and PTB-free patients, in which 2-butyl-1-octanol was the most expressed in the breath of TB positive population and was detected in $85\%$ of this group ($\frac{12}{14}$), while only in $50\%$ in the control group ($\frac{10}{20}$) [20]. 2-butyl-1-octanol was also identified by fuzzy logic analysis as the best discriminator between patients whose sputum cultures were positive or negative for Mycobacteria in Phillips’s study [17]. Bobak et al. conducted an exploratory study on the exhaled diagnosis of PTB in 31 children in South Africa and found that PTB could be identified with $90\%$ accuracy from other respiratory infections based on four VOCs, including decane and 4-methyloctane[21]. Furthermore, the sensor based breath test method also achieved good performance on TB/PTB detection. For example, Marcel et al. constructed and evaluated a DiagNose (C-it BV) based TB diagnosis method on 194 participants, and achieved a sensitivity of $93.5\%$ and a specificity of $85.3\%$ in discriminating TB patients and HC, and got a sensitivity of $76.5\%$ and specificity of $87.2\%$ when identifying TB patient within the entire test-population [22]. Morad et al. evaluated a nano-sensor based TB detection method on 60 blinded validation datasets, and achieved a specificity, positive predictive value (PPV), and negative predictive value (NPV) of $88\%$, $76\%$, and $94\%$, respectively [23]. In 2017, Mohamed et al. distinguished TB patients [260] from HC participants [240] for multiple biological samples (blood, breath, sputum, and urine) with sensitive and specificity > $95\%$ via e-Nose analyses [24]. The above studies proved the feasibility of breath VOCs based PTB detection.
GC–MS has advantages in the qualitative and quantitative detection of substances. However, the selection of chromatography columns and the complex procedures limited the detection scope of GC–MS. Besides, the consistency of reported VOCs from different studies is poor, since GC–MS analysis requires complex procedures and specialized skills [13]. The sensor based solution usually uses a single or a series of sensor to identify the response pattern to breath without considering the specific compositions. It is fast but easily affected by other interference factors such as the environment [13]. Thus, it is still desirable for a real-time, robust, accurate, and simple breath analysis platform for VOC detection. The online mass spectrometry platform could meet such requirements.
Recently, different online mass spectrometry technologies have been developed to analyze exhaled breath, such as proton transfer reaction MS (PTR-MS) [25], secondary electrospray ionization MS (SESI-MS) [26, 27], and high-pressure photon ionization time-of-flight mass spectrometry (HPPI-TOF-MS) [28]. The HPPI-TOFMS platform is designed and developed by our team and has been used for lung cancer and esophageal cancer detection [29–31] and achieved good performances with sensitivity and specificity > $90\%$. In this study, we aimed to develop a breathomics based PTB detection and investigate its performance on the clinical data set in this study.
## Study design and participants
We conducted a cross-sectional study from 1 March 2020 to 31 March 2021 at The Third People's Hospital of Shenzhen. The study was approved by the Ethics Committee of The Third People's Hospital of Shenzhen (number: 2020-012). Written informed consent was obtained from all participants.
The total participants consisted of a case group and a control group. For the case group, confirmed PTB patients were prospectively and consecutively recruited based on the following criteria: [1] aged 18–70 years old; [2] diagnosed by Xpert and/or culture, with suggestive clinical and radiological findings; [3] anti-TB treatment not initiated or started less than 2 weeks. The control group consists of two parts: healthy controls with no pulmonary diseases (HC) and patients with pulmonary diseases (unhealthy controls, UHC) which could be noninfectious diseases or infectious diseases other than PTB. HCs were simultaneously recruited and underwent a physical examination with the following criteria: [1] aged 18–70 years old; [2] no respiratory symptoms (e.g., cough, sputum, hemoptysis, shortness of breath, dyspnea, or chest pain); [3] no pulmonary lesions by chest imaging (chest X-ray or computed tomography). For UHC, they should: [1] aged 18–70 years old; [2] have pathogenic confirmed infectious diseases or treatment response suggestive of pulmonary infectious diseases, or have chronic noninfectious diseases, without evidence of infection. Both the case group and the control group would be excluded if the airbag leaked or were unable to take enough breath volume. The participant enrollment flow is illustrated in Fig. 1a. A total of 518 PTB patients and 887 controls with 77 UHC and 810 HC were enrolled in this study. Fig. 1The flow of participants enrollment and PTB detection model construction and test The physicians were responsible for making a clinical diagnosis and for the collection of the breath samples. The other researchers performed the VOCs detection and ML modeling and were blinded to clinical data and other test results. Additionally, the physicians were also blinded to the breath test results. The demographic and clinical characteristics of all participants were collected and summarized in Table 1, including age, sex, and antituberculosis therapy. Table 1Demographic characteristics of participantsDiscovery data setTest data setPTB ($$n = 361$$)Control ($$n = 614$$)p-valuePTB ($$n = 157$$)Control ($$n = 273$$)p-valueAge Median (min.–max.)36 (18–70)28 (18–69) < 0.00132 (18–70)28 (18–70) < 0.001 < 30 (%)115 (31.9)345 (56.2)0.00864 (40.8)169 (61.9)0.258 ≥ 30 (%)246 (68.1)269 (43.8) < 0.00193 (59.2)104 (38.1)0.009Sex Male (%)223 (61.8)325 (52.9)0.009101 (64.3)142 (52.0)0.004 Female (%)138 (38.2)289 (47.1)–56 (35.7)131 (48.0)–Bold p-value shows that there are significant differences between PTB and controls
## Sampling procedures
All breath samples were collected using a predefined protocol and tested within twenty-four hours. The sampling apparatus was composed of a disposable gas nipple and a sampling bag made of polyether-ether-ketone (PEEK). In this study, we set standard sampling demands and protocols to minimize the influence of the daily diet. Firstly, we conducted sampling at a second visit if he/she was an inpatient and informed the participants to prepare for sampling in advance: no smoking, alcohol, or diets within an hour before sampling. Secondly, participants were required to rinse their mouths with purified water instantly before sampling to minimize the influence of diet, smoking, etc. Thirdly, all samples are required to be collected in the same environment, which could minimize the effects of environmental facts. With a deep nasal inhalation, participants completely exhaled the air into the sampling bag with over 1.2 L volume.
## Breath sample detection
HPPI-TOFMS, which consisted of a vacuum ultraviolet (VUV) lamp-based HPPI ion source and an orthogonal acceleration time-of-flight (TOF) mass analyzer, was used to detect and analyze the breath samples. A commercial VUV-Kr lamp with a photon energy of 10.6 eV was adopted in this platform. Most VOCs with an ionization potential lower than 10.6 eV were ionized in the ionization region directly [32]. Breath samples were directly introduced through a 250 μm i.d. 0.60 m long stainless-steel capillary. The HPPI ion source works in soft HPPI ionization mode, which will produce mostly radical cations (M+) by ionization reaction as M + hγ → M+ + e. Then, the ion transmission system effectively transferred these ions from the ion source into the orthogonal acceleration, reflection TOFMS mass analyzer. The TOFMS signals were recorded by a 400 ps time-to-digital conversion rate at 25 kHz, and all the mass spectra were accumulated for 60 s. Thus, it takes 1 min for one sample to go through a detection. A spectrogram with 31,666 data pairs was extracted from each exhaled breath sample. Based on the flight time and m/z calibration on the standard gas with nine compounds at a concentration of 1 ppmv, the timeline of flight can be transferred as m/z, which is in the range of [0, 350]. The TOFMS signals were positively correlated with the concentration of the VOC ions. The detection limit is down to 0.015 ppbv (parts per billion by volume) for aliphatic and aromatic hydrocarbons [28]. The gas-phase breath sample was directly inhaled into the ionization region through a 250 μm i.d. 0.60 m long capillary from the sampling bag. The TOF signals were recorded by a time-to-digital converter, and all the mass spectra were accumulated for 60 s. Mass spectrum peaks with m/z < 350 were detected by HPPI-TOFMS for each exhaled breath sample. The noise-reducing and base-line correction were implemented via anti-symmetric wavelet transformation, which was achieved by Python package pywavelets [33]. To transfer the discrete signal of mass spectra data to standard breathomics data, we calculate the area of the strongest peak in the range of [x − 0.1, x + 0.1) as the feature of VOC with m/z close to x. In this study, 1500 breathomics data were detected for machine learning (ML) model construction in the ions m/z range of [20, 320) with an interval of 0.2. A statistical analysis based feature selection was executed to avoid model over-fitting, in which the features without significant difference ($p \leq 0.05$) were excluded before model training.
## PTB detection model construction
As illustrated in Fig. 1b, all the enrolled participants were randomly split into two groups: $70\%$ of them for model construction and the remaining $30\%$ of them for model blinded testing. Thus, 361 PTB patients and 614 controls were randomly selected as the discovery data set. Through 100 times of 7:3 randomization, the discovery data set was further divided into a training subset and an internal validation subset. On the training subset, several popular ML models including Random Forest (RF) [34], Support Vector Machine (SVM) [35], Logistic Regression (LR) [36], eXtreme Gradient Boosting (XGB) [37], and Decision Tree (DT) [38] were employed as the classifier to distinguish PTB patients and controls. The descriptions and main parameter settings of these ML models are illustrated in Table 2. Then, the optimal classifier for distinguishing PTB patients and controls is selected according to the model performance in the internal validation subset, which is named as “BreaTB”. Table 2The descriptions and main parameter settings of the employed ML modelsML modelsDescriptionsMain parameter settingsaRFA meta estimator that fits a number of decision tree classifiers on various sub-samples of the dataset and uses averaging to improve the predictive accuracy and control over-fittingn_estimators = 100, max_features = 0.5, min_samples_split = 4, min_samples_leaf = 10, criterion = "entropy"SVMSolves the separation hyperplane which can divide the training data set correctly and has the maximum geometric intervalpenalty = "l2", loss = "squared_hinge", tol = 1e−5, $C = 5.0$, max_iter = 1e + 5LREstimates the probability of an event occurring based on a given dataset of independent variablestol = 1e−5, $C = 5.0$, max_iter = 1e + 4XGBA boosting algorithm based on gradient boosted decision trees algorithmbooster: "gbtree", max_depth: 8, n_estimators: 100, min_child_weight: 3, gamma: 0.15, lambda: 2DTEmploys a divide and conquer strategy by conducting a greedy search to identify the optimal split points within a treecriterion = "gini", splitter = "best", min_samples_split = 2, min_samples_leaf = 1aThese algorithms were achieved based on python packages: xgboost (https://xgboost.readthedocs.io/en/stable/python/python_intro.html) and sklearn (https://scikit-learn.org/stable/user_guide.html)
## Performance evaluation and statistical analysis
As BreaTB is constructed, the most important features can be confirmed based on the feature importance or coefficient in model training. Feature differences analysis was also implemented on the relative density of VOCs among different patient groups.
BreaTB was applied and evaluated on the blinded testing data set, which consisted of 157 PTB patients, 248 HC, and 25 UHC. The model detection results were compared with the clinically confirmed diagnosis results. Furthermore, we also assessed the performance of BreaTB stratified by clinical characteristics. We calculated the sensitivity, specificity, PPV, NPV, accuracy, AUC (the area under the receiver operating characteristic curve (ROC)), and the relative $95\%$ confidence interval (CI) were calculated to evaluate the performance of BreaTB.
All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and Origin software (version 2018). Descriptive statistics were reported as frequencies (percentages) for categorical variables or median (minima to maxima) for continuous variables. We compared the demographic characteristics among different patient groups using the Mann–Whitney U test for continuous variables and the chi-square test for categorical variables. A p-value < 0.05 was considered statistically significant in all analyses. All the tests were two-tailed.
## Results
For different ML models, the mean performance metrics of 100 models on randomly selected training sets were illustrated in Table 3. Since the scale of the dataset enrolled is relatively large in this study, these basic classifiers such as SVM, LR, and DT all perform well in the PTB detection task. As the meta and boosting classifiers of DT, the RF and XGB based PTB detection models have superior performances. Based on the validation results, the best-performing RF and XGB based PTB detection models were selected for further testing. The results in Table 3 showed the XGB model has better performance than the RF model in the validation data set. However, the RF model performs superior to the XGB model in the blinded test data set with an accuracy of $92.6\%$ ($95\%$ CI 90.1–$95.0\%$), a sensitivity of $91.7\%$ ($95\%$ CI 88.5–$95.0\%$), and a specificity of $93.0\%$ ($95\%$ CI 88.9–$97.2\%$). It implies that the RF model is more robust than XGB. Thus, we only further analyze the RF-based PTB detection model (termed as BreaTB). Figure 2 illustrated the prediction scores of BreaTB on all tested samples, which represent the probability of PTB infection. The cut-off line(threshold = 0.5) divides the PTB patients from controls well with fewer false positives and false negatives. Table 3Performance metrics (mean ± STD) of difference ML models for PTB detection in internal validation and blinded test datasetData setsModelsSensitivity (%)Specificity (%)PPV (%)NPV (%)Accuracy (%)AUCValidation ($$n = 295$$)RF90.6 ± 3.190.6 ± 2.485.1 ± 3.294.3 ± 1.790.6 ± 1.70.960 ± 0.011SVM67.7 ± 20.083.4 ± 12.674.3 ± 10.483.1 ± 7.677.6 ± 4.80.755 ± 0.061LR78.6 ± 4.482.0 ± 4.172.2 ± 4.786.8 ± 2.580.8 ± 3.10.856 ± 0.030XGB88.1 ± 3.093.6 ± 2.189.0 ± 3.293.1 ± 1.691.5 ± 1.60.969 ± 0.010DT76.1 ± 5.090.5 ± 2.882.6 ± 4.286.7 ± 2.485.2 ± 2.50.833 ± 0.028Test ($$n = 430$$)RF90.7 ± 1.592.1 ± 1.586.9 ± 2.194.5 ± 0.891.6 ± 1.00.970 ± 0.005SVM69.4 ± 20.483.6 ± 12.774.5 ± 9.784.3 ± 7.878.4 ± 5.20.765 ± 0.066LR82.5 ± 3.383.2 ± 4.074.1 ± 4.589.2 ± 1.882.9 ± 2.80.877 ± 0.021XGB88.1 ± 1.794.6 ± 1.290.5 ± 2.093.2 ± 0.992.2 ± 0.90.970 ± 0.004DT75.3 ± 4.189.4 ± 1.980.5 ± 3.086.3 ± 2.084.3 ± 1.90.824 ± 0.023Bold values represent the best performance metrics achieved among differences mahcine learning methodsFig. 2Predictive score of BreaTB on the test data set As shown in Table 1, in the training data set, the median age of PTB patients was significantly higher than that of controls (36 (18–70) vs. 28 (18–69) years old), and there were more males in PTB patients than in controls ($61.8\%$ vs. $52.9\%$). The distribution of age ≥ 30 and gender in the test data set is as same as that in the training data set, except for that in age < 30. Thus, it is necessary to evaluate the influence of these clinic characteristics on model performance. As illustrated in Fig. 3 and Table 4, the ROC curve showed that BreaTB achieved an AUC of 0.975 ($95\%$ CI, 0.961–0.998) in the overall test data set. The diagnostic performance of BreaTB was fairly consistent across different subgroups based on demographic and clinical baseline characteristics, such as age, gender, and anti-tuberculosis therapy. The results demonstrated that age, sex, and anti-tuberculosis therapy have no evident influence on BreaTB. In detail, BreaTB has superior performance on participants with age < 30 than those with age ≥ 30. For different genders, BreaTB also performs slightly differently with superior sensitivity and inferior specificity in females than in males. After the anti-TB therapy, the PTB patients are more difficult to be distinguished from the controls for BreaTB. Except for the general characteristics, we also analyzed the PTB distinguish performance against HC and UHC. BreaTB had a sensitivity of $91.7\%$ ($95\%$ CI 87.4–$96.0\%$), and a specificity of $93.5\%$ ($95\%$ CI 90.5–$96.6\%$) for the identification of confirmed PTB from HC, which is a quasi-screening scenario. In contrast, inferior specificity of $88.0\%$ ($95\%$ CI 75.3–$100\%$) was achieved by BreaTB in distinguishing TB from UHC, which is a quasi-diagnosis scenario. Fig. 3Performance of the BreaTB on different tuberculosis subgroupsTable 4Performance metrics ($95\%$ CI) of BreaTB on the test data set and on different subgroupsGroupsSensitivity (%)Specificity (%)PPV (%)NPV (%)Accuracy (%)AUCOverall Test ($$n = 430$$)91.7 (87.4–96.0)93.0 (90.0–96.1)88.3 (84.5–92.2)95.1 (91.8–98.4)92.6 (90.1–95.0)0.975 (0.961–0.990)Age, year < 30 ($$n = 233$$)90.6 (83.5–97.8)91.7 (87.6–95.9)80.6 (74.4–86.7)96.3 (91.9–100.0)91.4 (87.8–95.0)0.972 (0.951–0.993) ≥ 30 ($$n = 197$$)92.5 (87.1–97.8)95.2 (91.1–99.3)94.5 (90.2–98.8)93.4 (88.3–98.5)93.9 (90.6–97.2)0.978 (0.958–0.999)Sex Male ($$n = 243$$)90.1 (84.3–95.9)95.1 (91.5–98.6)92.9 (88.7–97.0)93.1 (88.1–98.1)93.0 (89.8–96.2)0.975 (0.955–0.994) Female ($$n = 187$$)94.6 (88.7–100)90.8 (85.9–95.8)81.5 (74.7–88.4)97.5 (93.8–100)92.0 (88.1–95.9)0.981 (0.962–1.000)Anti-TB Untreated ($$n = 385$$)92.9 (88.1–97.6)93.0 (90.0–96.1)84.6 (80.2–88.9)96.9 (93.9–100)93.0 (90.4–95.5)0.978 (0.964–0.993) Treated ($$n = 318$$)88.9 (79.7–98.1)93.0 (90.0–96.1)67.8 (62.1–73.5)98.1 (94.6–100)92.5 (89.5–95.4)0.968 (0.949–0.987)Controls HC ($$n = 405$$)91.7 (87.4–96.0)93.5 (90.5–96.6)90.0 (86.2–93.8)94.7 (91.2–98.2)92.8 (90.3–95.4)0.977 (0.962–0.991) UHC ($$n = 182$$)91.7 (87.4–96.0)88.0 (75.3–100.0)98.0 (93.3–100)62.9 (55.0–70.7)91.2 (87.1–95.3)0.961 (0.933–0.989) *In this* study, over 30 VOC ions were selected via statistical analysis for the BreaTB model training in each iteration. To analyze the importance of different VOC ions for PTB detection, we selected the best VOC ion combinations through RF model based feature selection for 100 iterations. Then, all selected VOC ions were ordered by the selection frequency in RF modeling. As shown in Fig. 4a, there are five VOC ions with m/z of 72, 68, 65, 67, and 65.2 selected at each iteration. There are eleven VOC ions selected in over 90 iterations. Thus, we analyzed the most important eleven VOC ions between confirmed PTB patients and controls. Figure 4b shows the mass spectrum examples of a PTB patient and control individual. It demonstrates that there are some differences in the top eleven VOC ions, which are shown in color bars. To further explore these VOC ions, we analyzed the group differences between PTB and controls and evaluated the performance of each VOC ion in discriminating the PTB and controls. As demonstrated in Fig. 4c, d, all these eleven VOC ions are significantly different between the PTB group and controls with a p-value < 0.05 (the blue line in Fig. 4c). The discernibility (AUC in discriminating PTB group and controls) of VOC ions is related to the scale and significance of group differences. The ROC curve in Fig. 4d shows the discrimination of a single VOC ion is limited (AUC < 0.75). However, the combination of all eleven VOC ions performs well on the test data set with an AUC of 0.905($95\%$ CI: 0.878–0.933). It implies that the panel of VOC ions is the basis for breathomics based PTB detection. The heat map in the PTB group, UHC, and HC illustrated the patterns of these eleven VOC ions are visually different. Fig. 4Investigations of breath VOC ions and PTB. a The volcano plot shows the group changes and differences in breath VOC ion intensity between PTB and controls. b The performances of the top eleven VOC ions in distinguishing PTB patients and controls. c The heatmap of the top eleven VOC ions in PTB, UHC, and HC, shows the pattern differences of VOC ions Since the qualitative ability of the TOF mass spectrometer is limited, we can just infer the possible chemicals of these PTB related VOC ions based on their m/z (72.0, 68.0, 65.0, 67.0, 65.2, 69.0, 66.0, 59.0, 61.0, 53.0, 58.0), correlation-ship (Fig. 4e), intensity distribution (Fig. 5), other published potential biomarkers, and the human breathomics database [39]. Considering the ions intensity distribution similarity and the relationship of m/z values, the VOC ions with m/z of 68 and 69 could be isoprene and its protonated cation. The VOC ions with m/z of 58 and 59 could be acetone and its protonated cation. Isoprene and acetone are common metabolites in human breath [40]. Isoprene is proven to be related to oxidative stress responses [41, 42]. Acetone is related to diabetes [43], and tuberculosis patients have a high incidence of diabetes [44]. The VOC ion with m/z of 72 could be 2-butanone, which is also found as the top eleven biomarkers for PTB in Machel Phillips’s study [17]. The VOC ion with m/z of 61 could be the protonated ions of acetic acid, which was proven related to tuberculosis in skin samples [45]. The VOC ion with m/z of 65, 65.2, and 66 could be the fragment ion of 4-nitrophenol and the corrspounding protonated cation, respectively. The VOC ion with m/z of 67 could be Pyrrole or 3-Butenenitrile. The low peak intensity VOC ions with m/z of 53 could be the fragment ion of other unknown VOC with low concentration. These VOCs would be potential biomarkers of TB.Fig. 5Intensity comparison of VOC ions between PTB group and controls
## Discussion
In this study, for the first time, we explore the diagnostic value of breathomics data detection on HPPI-TOF–MS for PTB in a large cohort. The results demonstrated that the developed BreaTB model performs well in distinguishing PTB individuals and control with high sensitivity and specificity of $91.7\%$ and $93.0\%$. It implies that the proposed breathomics method via online HPPI-TOF–MS could be a potentially feasible diagnostic or screening tool in the clinical setting.
In the past decades, no breathomics-based method has been translated into clinical practice for the diagnosis of TB, which is primarily due to the complexity and the high cost of existing spectrometers and the limitations of sensor technologies [13]. Compared with past research on PTB detection, there are several advantages in our study. Firstly, the diagnostic accuracy of our VOCs-based PTB detection method was high, with sensitivity and specificity of $91.7\%$ and $93.0\%$. Furthermore, our study was tested on a large-size patient cohort. As participants were stratified based on their demographic and clinical characteristics: age, sex, and anti-tuberculosis therapy, the diagnostic performance was fairly consistent. Thirdly, TB diagnostic methods using non-sputum samples are strongly advocated by the WHO [46]. Breath sampling has excellent clinical accessibility, especially for certain categories of patients whose sputum is difficult to collect. Fourthly, the breath sample detection on HPPI-TOF-MS only takes about one minute. Thus, the total time cost from breath sampling to getting PTB detection results is about five minutes.
However, there are several limitations in our study. Firstly, the qualitative and metabolic pathways of ions have not been defined. Thus, the logical and mechanistic evidence of the breathomics-based PTB detection method is not enough to make it clinically convincing, although it performs well in clinical data. Further chemical composition analysis via GC–MS is the focus of our future works. Fortunately, many studies have demonstrated the VOCs similarities and differences between the breath of PTB patients and M.tb culture-released gases. For example, Phillips et al. found the common compounds: 1-methyl- and cyclohexane, 1,4-dimethyl- in the breath of PTB patients and headspace air of culture [17]. Using computational approaches, Purva et al. proposed putative biosynthetic pathways in M.tb for three VOCs(methyl nicotinate, methyl phenylacetate, and methyl p-anisate), and methyl nicotinate was also found in the exhaled breath of patients with tuberculosis [47]. Kuntzel et al. detected and analyzed the headspace VOCs of 17 different mycobacteria and control strains. Their result demonstrated the feasibility of identifying M.tb from other pathogens based on their metabolism of VOC [48]. Our team is also working on finding the links between the VOCs in the breath samples of PTB patients and the VOCs in the headspace air of M.tb culture. Secondly, the control group contained only a small sample of patients with pulmonary diseases other than PTB. Thus, the performance needs to be further evaluated in detecting PTB from other pulmonary diseases. Thirdly, our enrollment was restricted to adults with possible PTB. Similar independent validation studies are needed for children whose diagnostic tools are even more urgently needed [21], as well as patients living with diabetes or HIV and patients suspected of EPTB. At last, this is a single-center study conducted in a TB specialist hospital, which may limit the universality of the research results.
## Conclusion
In conclusion, we developed a breathomics model: BreaTB for PTB detection, which achieved high diagnostic accuracy on clinical data set with a sensitivity and a specificity of $91.7\%$ and $93.0\%$, respectively. Due to its simplicity and low cost, the breathomics-based PTB detection model on online breath analysis platforms such as HPPI-TOF-MS has the potential to meet the ongoing demand for TB diagnosis that would not require sputum and may work in active case finding in large populations, especially in resource-limited settings where it is urgently needed [12]. However, more clinical and basic researches are needed to evaluate this method in patients with more complex health conditions and with various lung diseases. Last but not least, more studies are needed to confirm the TB-specific breath biomarkers and clarify their metabolic pathways.
## References
1. Chakaya J, Khan M, Ntoumi F, Aklillu E, Fatima R, Mwaba P, Kapata N, Mfinanga S, Hasnain SE, Katoto PDMC. **Global tuberculosis report 2020—reflections on the Global TB burden, treatment and prevention efforts**. *Int J Infect Dis* (2021.0) **113** S7-S12. DOI: 10.1016/j.ijid.2021.02.107
2. Keeler E, Perkins MD, Small P, Hanson C, Reed S, Cunningham J, Aledort JE, Hillborne L, Rafael ME, Girosi F. **Reducing the global burden of tuberculosis: the contribution of improved diagnostics**. *Nature* (2006.0) **444** 49-57. DOI: 10.1038/nature05446
3. 3.World Health O. WHO consolidated guidelines on tuberculosis: module 3: diagnosis: rapid diagnostics for tuberculosis detection, 2021 update edn. Geneva: World Health Organization; 2021.
4. Parsons LM, Somoskövi A, Gutierrez C, Lee E, Paramasivan CN, Abimiku AL, Spector S, Roscigno G, Nkengasong J. **Laboratory diagnosis of tuberculosis in resource-poor countries: challenges and opportunities**. *Clin Microbiol Rev* (2011.0) **24** 314-350. DOI: 10.1128/CMR.00059-10
5. Datta S, Evans CA. **The uncertainty of tuberculosis diagnosis**. *Lancet Infect Dis* (2020.0) **20** 1002-1004. DOI: 10.1016/S1473-3099(20)30400-X
6. Gopi A, Madhavan SM, Sharma SK, Sahn SA. **Diagnosis and treatment of tuberculous pleural effusion in 2006**. *Chest* (2007.0) **131** 880-889. DOI: 10.1378/chest.06-2063
7. Jeon D. **Tuberculous pleurisy: an update**. *Tubercul Respir Dis* (2014.0) **76** 153-159. DOI: 10.4046/trd.2014.76.4.153
8. Theron G, Peter J, Meldau R, Khalfey H, Gina P, Matinyena B, Lenders L, Calligaro G, Allwood B, Symons G. **Accuracy and impact of Xpert MTB/RIF for the diagnosis of smear-negative or sputum-scarce tuberculosis using bronchoalveolar lavage fluid**. *Thorax* (2013.0) **68** 1043-1051. DOI: 10.1136/thoraxjnl-2013-203485
9. Yang J, Shen Y, Wang L, Ju L, Wu X, Wang P, Hao X, Sun Q, Yu F, Sha W. **Efficacy of the Xpert**. *Int J Infect Dis* (2021.0) **107** 121-126. DOI: 10.1016/j.ijid.2021.04.040
10. 10.WHO. High priority target product profiles for new tuberculosis diagnostics. In: Report of a consensus meeting. Geneva, Switzerland; 2014.
11. Fowler SJ, Basanta-Sanchez M, Xu Y, Goodacre R, Dark PM. **Surveillance for lower airway pathogens in mechanically ventilated patients by metabolomic analysis of exhaled breath: a case–control study**. *Thorax* (2015.0) **70** 320-325. DOI: 10.1136/thoraxjnl-2014-206273
12. Rattray NJ, Hamrang Z, Trivedi DK, Goodacre R, Fowler SJ. **Taking your breath away: metabolomics breathes life in to personalized medicine**. *Trends Biotechnol* (2014.0) **32** 538-548. DOI: 10.1016/j.tibtech.2014.08.003
13. Saktiawati AMI, Putera DD, Setyawan A, Mahendradhata Y, van der Werf TS. **Diagnosis of tuberculosis through breath test: a systematic review**. *EBioMedicine* (2019.0) **46** 202-214. DOI: 10.1016/j.ebiom.2019.07.056
14. Beale DJ, Pinu FR, Kouremenos KA, Poojary MM, Narayana VK, Boughton BA, Kanojia K, Dayalan S, Jones OAH, Dias DA. **Review of recent developments in GC–MS approaches to metabolomics-based research**. *Metabolomics* (2018.0) **14** 152. DOI: 10.1007/s11306-018-1449-2
15. Papadimitropoulos MP, Vasilopoulou CG, Maga-Nteve C, Klapa MI. **Untargeted GC–MS metabolomics**. *Methods Mol Biol (Clifton, NJ)* (2018.0) **1738** 133-147. DOI: 10.1007/978-1-4939-7643-0_9
16. Mochalski P, Shuster G, Leja M, Unterkofler K, Jaeschke C, Skapars R, Gasenko E, Polaka I, Vasiljevs E, Shani G. **Non-contact breath sampling for sensor-based breath analysis**. *J Breath Res* (2019.0) **13** 036001. DOI: 10.1088/1752-7163/ab0b8d
17. Phillips M, Cataneo RN, Condos R, Ring Erickson GA, Greenberg J, La Bombardi V, Munawar MI, Tietje O. **Volatile biomarkers of pulmonary tuberculosis in the breath**. *Tuberculosis* (2007.0) **87** 44-52. DOI: 10.1016/j.tube.2006.03.004
18. Phillips M, Basa-Dalay V, Bothamley G, Cataneo RN, Lam PK, Natividad MPR, Schmitt P, Wai J. **Breath biomarkers of active pulmonary tuberculosis**. *Tuberculosis* (2010.0) **90** 145-151. DOI: 10.1016/j.tube.2010.01.003
19. Beccaria M, Bobak C, Maitshotlo B, Mellors T, Purcaro G, Franchina F, Rees C, Nasir M, Black A, Hill J. **Exhaled human breath analysis in active pulmonary tuberculosis diagnostics by comprehensive gas chromatography–mass spectrometry and chemometric techniques**. *J Breath Res* (2018.0) **13** 016005. DOI: 10.1088/1752-7163/aae80e
20. Beccaria M, Mellors TR, Petion JS, Rees CA, Nasir M, Systrom HK, Sairistil JW, Jean-Juste M-A, Rivera V, Lavoile K. **Preliminary investigation of human exhaled breath for tuberculosis diagnosis by multidimensional gas chromatography–time of flight mass spectrometry and machine learning**. *J Chromatogr B* (2018.0) **1074–1075** 46-50. DOI: 10.1016/j.jchromb.2018.01.004
21. Bobak CA, Kang L, Workman L, Bateman L, Khan MS, Prins M, May L, Franchina FA, Baard C, Nicol MP. **Breath can discriminate tuberculosis from other lower respiratory illness in children**. *Sci Rep* (2021.0) **11** 2704. DOI: 10.1038/s41598-021-80970-w
22. Bruins M, Rahim Z, Bos A, van de Sande WWJ, Endtz HP, van Belkum A. **Diagnosis of active tuberculosis by e-nose analysis of exhaled air**. *Tuberculosis* (2013.0) **93** 232-238. DOI: 10.1016/j.tube.2012.10.002
23. Nakhleh MK, Jeries R, Gharra AL, Binder A, Broza YY, Pascoe M, Dheda K, Haick H. **Detecting active pulmonary tuberculosis with a breath test using nanomaterial-based sensors**. *Eur Respir J* (2014.0) **43** 1522-1525. DOI: 10.1183/09031936.00019114
24. Mohamed EI, Mohamed MA, Moustafa MH, Abdel-Mageed SM, Moro AM, Baess AI, El-Kholy SM. **Qualitative analysis of biological tuberculosis samples by an electronic nose-based artificial neural network**. *Int J Tuberculosis Lung Dis* (2017.0) **21** 810-817. DOI: 10.5588/ijtld.16.0677
25. Trefz P, Schmidt M, Oertel P, Obermeier J, Brock B, Kamysek S, Dunkl J, Zimmermann R, Schubert JK, Miekisch W. **Continuous real time breath gas monitoring in the clinical environment by proton-transfer-reaction-time-of-flight-mass spectrometry**. *Anal Chem* (2013.0) **85** 10321-10329. DOI: 10.1021/ac402298v
26. Gaugg MT, Bruderer T, Nowak N, Eiffert L, Martinez-Lozano Sinues P, Kohler M, Zenobi R. **Mass-spectrometric detection of omega-oxidation products of aliphatic fatty acids in exhaled breath**. *Anal Chem* (2017.0) **89** 10329-10334. DOI: 10.1021/acs.analchem.7b02092
27. Singh KD, Del Miguel GV, Gaugg MT, Ibañez AJ, Zenobi R, Kohler M, Frey U, Sinues PM. **Translating secondary electrospray ionization-high-resolution mass spectrometry to the clinical environment**. *J Breath Res* (2018.0) **12** 027113. DOI: 10.1088/1752-7163/aa9ee3
28. Wang Y, Jiang J, Hua L, Hou K, Xie Y, Chen P, Liu W, Li Q, Wang S, Li H. **High-pressure photon ionization source for TOFMS and its application for online breath analysis**. *Anal Chem* (2016.0) **88** 9047-9055. DOI: 10.1021/acs.analchem.6b01707
29. Huang Q, Wang S, Li Q, Wang P, Li J, Meng S, Li H, Wu H, Qi Y, Li X. **Assessment of breathomics testing using high-pressure photon ionization time-of-flight mass spectrometry to detect esophageal cancer**. *JAMA Netw Open* (2021.0) **4** e2127042. DOI: 10.1001/jamanetworkopen.2021.27042
30. Meng S, Li Q, Zhou Z, Li H, Liu X, Pan S, Li M, Wang L, Guo Y, Qiu M. **Assessment of an exhaled breath test using high-pressure photon ionization time-of-flight mass spectrometry to detect lung cancer**. *JAMA Netw Open* (2021.0) **4** e213486. DOI: 10.1001/jamanetworkopen.2021.3486
31. Wang P, Huang Q, Meng S, Mu T, Liu Z, He M, Li Q, Zhao S, Wang S, Qiu M. **Identification of lung cancer breath biomarkers based on perioperative breathomics testing: a prospective observational study**. *EClinicalMedicine* (2022.0) **47** 101384. DOI: 10.1016/j.eclinm.2022.101384
32. Zhao X, Liu X, Liu J, Chen J, Fu S, Zhong F. **The effect of ionization energy and hydrogen weight fraction on the non-thermal plasma volatile organic compounds removal efficiency**. *J Phys D Appl Phys* (2019.0) **52** 145201. DOI: 10.1088/1361-6463/aafe8b
33. Lee G, Gommers R, Waselewski F, Wohlfahrt K, O'Leary A. **PyWavelets: a Python package for wavelet analysis**. *J Open Source Softw* (2019.0) **4** 1237. DOI: 10.21105/joss.01237
34. Breiman L. **Random forests**. *Mach Learn* (2001.0) **45** 5-32. DOI: 10.1023/A:1010933404324
35. Suthaharan S, Suthaharan S. **Support vector machine**. *Machine learning models and algorithms for big data classification: thinking with examples for effective learning* (2016.0) 207-235
36. Bewick V, Cheek L, Ball J. **Statistics review 14: logistic regression**. *Crit Care* (2005.0) **9** 112-118. DOI: 10.1186/cc3045
37. 37.Deng L, Sui Y, Zhang J. XGBPRH: prediction of binding hot spots at protein–RNA interfaces utilizing extreme gradient boosting. LID. 10.3390/genes10030242.242: 2073–4425.
38. 38.Jordan MI. A statistical approach to decision tree modeling. In: COLT '94: 1994; 1994.
39. 39.Kuo TC, Tan CE, Wang SY, Lin OA, Su BH, Hsu MT, Lin J, Cheng YY, Chen CS, Yang YC, et al. Human breathomics database. Database (Oxford) 2020; 1758–0463 (Electronic)).
40. Kinoyama M, Nitta H, Watanabe A, Ueda H. **Acetone and isoprene concentrations in exhaled breath in healthy subjects**. *J Health Sci* (2008.0) **54** 471-477. DOI: 10.1248/jhs.54.471
41. Arashiro M, Lin Y-H, Zhang Z, Sexton KG, Gold A, Jaspers I, Fry RC, Surratt JD. **Effect of secondary organic aerosol from isoprene-derived hydroxyhydroperoxides on the expression of oxidative stress response genes in human bronchial epithelial cells**. *Environ Sci Process Impacts* (2018.0) **20** 332-339. DOI: 10.1039/C7EM00439G
42. 42.Alkhouri N, Singh T, Alsabbagh E, Guirguis J, Chami T, Hanouneh I, Grove D, Lopez R, Dweik R. Isoprene in the exhaled breath is a novel biomarker for advanced fibrosis in patients with chronic liver disease: a pilot study. 2015; 6.
43. Wang Z, Wang C. **Is breath acetone a biomarker of diabetes? A historical review on breath acetone measurements**. *J Breath Res* (2013.0) **7** 037109. DOI: 10.1088/1752-7155/7/3/037109
44. Du Q, Wang L, Long Q, Zhao Y, Abdullah AS. **Systematic review and meta-analysis: Prevalence of diabetes among patients with tuberculosis in China**. *Tropical Med Int Health* (2021.0) **26** 1553-1559. DOI: 10.1111/tmi.13686
45. 45.Vishinkin R, Busool R, Mansour E, Fish F, Haick H. Profiles of volatile biomarkers detect tuberculosis from skin. Adv Sci. 2021.
46. Denkinger C, Kik S, Cirillo D, Casenghi M, Shinnick T, Weyer K, Gilpin C, Boehme C, Schito M, Kimerling M. **Defining the needs for next generation assays for tuberculosis**. *J Infect Dis* (2015.0) **211** S29-S38. DOI: 10.1093/infdis/jiu821
47. Bhatter P, Raman K, Janakiraman V. **Elucidating the biosynthetic pathways of volatile organic compounds in**. *Mol BioSyst* (2017.0) **13** 750-755. DOI: 10.1039/C6MB00796A
48. Küntzel A, Oertel P, Fischer S, Bergmann A, Trefz P, Schubert J, Miekisch W, Reinhold P, Köhler H. **Comparative analysis of volatile organic compounds for the classification and identification of mycobacterial species**. *PLoS ONE* (2018.0) **13** e0194348. DOI: 10.1371/journal.pone.0194348
|
---
title: 'Metabolomic profiles in night shift workers: A cross-sectional study on hospital
female nurses'
authors:
- Elisa Borroni
- Gianfranco Frigerio
- Elisa Polledri
- Rosa Mercadante
- Cristina Maggioni
- Luca Fedrizzi
- Angela Cecilia Pesatori
- Silvia Fustinoni
- Michele Carugno
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9999616
doi: 10.3389/fpubh.2023.1082074
license: CC BY 4.0
---
# Metabolomic profiles in night shift workers: A cross-sectional study on hospital female nurses
## Abstract
### Background and aim
Shift work, especially including night shifts, has been found associated with several diseases, including obesity, diabetes, cancers, and cardiovascular, mental, gastrointestinal and sleep disorders. Metabolomics (an omics-based methodology) may shed light on early biological alterations underlying these associations. We thus aimed to evaluate the effect of night shift work (NSW) on serum metabolites in a sample of hospital female nurses.
### Methods
We recruited 46 nurses currently working in NSW in Milan (Italy), matched to 51 colleagues not employed in night shifts. Participants filled in a questionnaire on demographics, lifestyle habits, personal and family health history and work, and donated a blood sample. The metabolome was evaluated through a validated targeted approach measuring 188 metabolites. Only metabolites with at least $50\%$ observations above the detection limit were considered, after standardization and log-transformation. Associations between each metabolite and NSW were assessed applying Tobit regression models and Random Forest, a machine-learning algorithm.
### Results
When comparing current vs. never night shifters, we observed lower levels of 21 glycerophospholipids and 6 sphingolipids, and higher levels of serotonin (+$171.0\%$, $95\%$CI: 49.1–392.7), aspartic acid (+$155.8\%$, $95\%$CI: 40.8–364.7), and taurine (+$182.1\%$, $95\%$CI: 67.6–374.9). The latter was higher in former vs. never night shifters too (+$208.8\%$, $95\%$CI: 69.2–463.3). Tobit regression comparing ever (i.e., current + former) and never night shifters returned similar results. Years worked in night shifts did not seem to affect metabolite levels. The Random-Forest algorithm confirmed taurine and aspartic acid among the most important variables in discriminating current vs. never night shifters.
### Conclusions
This study, although based on a small sample size, shows altered levels of some metabolites in night shift workers. If confirmed, our results may shed light on early biological alterations that might be related to adverse health effects of NSW.
## 1. Introduction
Shift work (SW) refers to any organization of work hours that differ from the traditional diurnal work period (from 7:00 a.m. to 6:00 p.m.) [1], including evening, night, and early morning shifts, as well as fixed or rotating schedules [2, 3]. In particular, night shift work (NSW) refers to any kind of work that covers at least 3 h of work between 11:00 p.m. and 6:00 a.m. [4, 5]. In industrialized countries, SW and NSW are common work schedules [6]. Indeed, according to the U.S. Bureau of Labor Statistics, ~$16\%$ of employees surveyed in 2017–2018 followed SW schedules, including $6\%$ of evening shifts workers and $4\%$ of night shifts workers [7].
SW, especially if including night shifts, has been found to be associated with several diseases, e.g., cardiovascular diseases [8], cancers [9], metabolic disorders such as obesity [10, 11] and type 2 diabetes [12, 13], sleep disturbances [14], gastrointestinal disorders [15], and impaired mental health [16]. However, the underlying mechanisms are not fully understood. Some might be mediated by psychosocial stress deriving from interference with social rhythms, but there are indications also suggesting that disruption of normal sleep-wake cycle (circadian rhythm) following SW leads to neuroendocrine and cardiometabolic stress, curtailed and disturbed sleep, and, as a consequence, altered immune functioning and cellular stress [17, 18].
In recent years, omics-based approaches have shown great potential to shed light on mechanisms underlying diseases and their possible association with exposure to relevant risk factors through the identification of biomarkers. Metabolomics, one of the omics-based methodologies, refers to the techniques used to quantify the metabolites present within a cell, tissue or organism [19]. These techniques are mainly divided into two strategies: i.e., untargeted and targeted metabolomics. Targeted metabolomics is the measurement of defined groups of chemically characterized and biochemically annotated metabolites [20].
To our knowledge, only a few studies investigated the effects of shift work on the human metabolome. One measured plasma metabolites in 49 male workers at the beginning and end of a rotating shift schedule including nights and observed an association between NSW and alterations in several metabolites [21]. Two were laboratory studies aimed at evaluating the impact of simulated 3- [22] and 4-day [23] night shift schedules on the metabolic profile of healthy volunteers, with a particular focus on sleep/wake and feeding/fasting cycles. Other two studies compared shifters and non-shifters. The first one analyzed urinary metabolites, and found altered long-chain acylcarnitines, three amino acids, and one sphingomyelin [24] in night shift workers as compared to day workers, based on both crude and adjusted models. The second study evaluated serum metabolites and identified 76 of them in shift workers (including L- tryptophan, acylcarnitines, and several fatty acids) which may represent important biomarkers of impaired lipid metabolism, leading to weight gain and central obesity [25]. As such, evidence on this topic is still limited and further studies are needed. The aim of the present study is thus to evaluate the effect of NSW on serum metabolites in a sample of female nurses, using a targeted metabolomics approach.
## 2.1. Study population, personal data, and biological samples
Procedures for recruitment of the study population and collection of personal data and biological samples have been described elsewhere [26]. Briefly, 46 female nurses working in night shifts at the Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico in Milan, Italy, were recruited on a voluntary basis and matched by age and length of services to 51 colleagues not working in night shifts. Inclusion criteria were Caucasian ethnicity, age 30–45 years and length of service ≥1 year. After signing informed consent, all participants filled in a questionnaire on demographics, lifestyle habits, personal and family health history, and work history (with a particular focus on SW) and donated a 12 mL blood sample. The sample was drawn in the morning, at the end of the night shift for night shifters (7:15–7:45 a.m.) and at the beginning of the working day for day shifters (8:30–9:00 a.m.), to try to maximize potential differences between the two groups. The metabolomics profile (see below) could not be assessed for six subjects, and we thus performed our analyses on a total of 91 nurses.
The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board of the Policlinico Hospital (approval number 702_2015).
## 2.2. Metabolomic analysis
The metabolomic profile was assessed with a validated targeted metabolomics approach, implementing liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS), and using the AbsoluteIDQ p180 kit (Biocrates Life Sciences AG AbsoluteIDQ® p180 Kit, Innsbruck, Austria), which benefits of an established good interlaboratory reproducibility [27]. Briefly, the serum samples were placed on a 96-well plate pre-loaded with the isotopic labeled internal standards, along with a phosphate buffer solution as blank sample, a calibration curve (7 levels), and three levels of quality control samples. Two different plates were implemented for this study. The sample preparation consisted in the derivatization of amino acids and biogenic amines with phenyl isothiocyanate, evaporation, extraction with 5 mM ammonium acetate in methanol, centrifugation, and dilution. Amino acids and biogenic amines were separated and analyzed through an analytical column before the mass spectrometry (LC-MS/MS), while lipids and the hexose were analyzed with a simple flow injection analysis (FIA-MS/MS). A total of 188 metabolites were measured, including 21 amino acids, 21 biogenic amines, the sum of hexoses, 40 acylcarnitine, 15 sphingolipids (SM), and 90 glycerophospholipids among which 14 lysophosphatidylcholines (LysoPC), 38 diacylphosphatidylcholine (PC aa), and 38 acylalkylphosphatidylcholine (PC ae). Further instrumental and analytical details have been previously reported [28].
## 2.3. Statistical analysis
Differences in the distribution of the main adjustment variables across categories of NSW were assessed through the analysis of variance (ANOVA) for age and BMI (continuous) and chi-squared test for smoking habit (categorical).
Metabolomic data (from both LC-MS/MS and FIA-MS/MS) were batch-normalized through the MetIDQ software (Biocrates) using, for each metabolite, the median values of three repetitions of a quality control (reference sample) analyzed on the same plate, according to the manufacturer's instructions [29]. Only metabolites with at least $50\%$ of the observations above the limit of detection (LOD) were considered for the statistical analyses. Among these, each remaining value below the LOD was replaced with a value equal to the minimum LOD (specific for each metabolite). Metabolite concentrations were then log-transformed (base e) and standardized (each value subtracted by the mean and divided by the standard deviation).
To visualize how metabolites correlate with each other, we performed network analyses where metabolites were considered as nodes, and correlation coefficients obtained from each pair of metabolites as edges; the Fruchterman-Reingold force-directed layout algorithm was used, and the values of r were set as edge weights; only statistically significant correlations with r > 0.4 were considered and metabolites with no connection were not considered.
To assess the association between each metabolite and NSW, we applied Tobit censored linear regression models, which are useful to estimate linear relationships when considering dependent variables with left- or right-censoring [30, 31]: in the present work, we considered metabolite concentrations lower than LOD as left-censored. We built a Tobit model for each metabolite, with the metabolite concentration as dependent variable, and NSW as the main independent variable. NSW was modeled both as current or former vs. never NSW and as ever (i.e., current + former) vs. never NSW. As a sensitivity analysis, we stratified current night shifters according to their shift schedule (see below). We also considered “number of years worked in night shifts” as a variable of interest (equal to 0 in never night shifters). Adjustment variables considered a priori as potential confounders were body mass index (BMI) (kg/m2), age, plate (plate 1 or 2: i.e., which of the two 96-well plates the serum sample was loaded on during sample preparation for metabolomic analyses), and smoking habit (current vs. former/never smokers). The models assessing the association between metabolites and number of years in night shifts also included the variable “never vs. ever night shift.” Before implementing all models, we imputed the few values missing from our database (one for age, three for BMI, three for smoking habit) using the k-nearest neighbors algorithm (k-NN) [32] with a k-value = 9 [32]. From each model, we estimated the standardized beta coefficients and calculated the percent variation (Δ%) using the following formula: (exp(β)−1) x 100, where β is the regression coefficient representing the variation in the metabolite level for a unit increase in the independent variable. The p-values were adjusted for multiple testing by controlling the false discovery rate (FDR) according to the method of Benjamini and Hochberg [33] and a FDR p-value lower than 0.1 was considered statistically significant. To have a visual representation of the Tobit models, Volcano plots were created, assigning a dot to each molecule and plotting the Δ% vs. the negative logarithm of the FDR p-value.
A confirmatory analysis was also conducted applying a supervised machine-learning algorithm called Random Forest (RF). A RF consists of many unpruned individual decision trees that operate as an ensemble. Individual trees are grown by bootstrapping a random sample of the original data set and by selecting at random, at each node, a small group of input variables to split on. Results from different decision trees are, subsequently, averaged to make final predictions [34]. We used RF to classify subjects into current or former vs. never night shifters and into ever (i.e., current + former) vs. never night shifters, considering correlations among metabolites.
K-fold cross-validation, a statistical method for evaluating a machine-learning model and testing its performance, was applied to assess RF performances and to tune parameters in order to obtain optimal predictions. In the present study, based on 5-fold cross-validation results, we implemented RF algorithms setting the number of trees at 10,000, and the number of variables from which to choose at each node at 11 (i.e., the approximate square root of the total number of metabolites included in the analysis).
Variables importance scores were then calculated. They are RF-derived measures that facilitate results interpretation by ranking the importance of each feature (i.e., metabolite), and can be computed mainly through two methods: [1] Mean Decrease Accuracy, indicating how much the accuracy (i.e., the number of data points out of all data points which are correctly predicted) decreases when the interested variable is excluded; and [2] Mean Decrease Gini, indicating how much the Gini score (which calculates the probability of a specific feature to be classified incorrectly when selected at random) decreases when a variable is chosen to split a node. The larger the scores, the greater the importance of a variable [34]. We evaluated variable importance, using both above-cited methods, in order to produce more accurate results.
All statistical analyses were performed using R (R version 4.1.2, R Foundation, Vienna, Austria) [35] with the Rstudio interface (Version 1.4.1717, RStudio Inc., Boston, MA, USA) and the packages “tidyverse” [36], “VIM,” “AER” [37], “tidygraph,” “ggraph” [38, 39], and “randomForest” [40].
## 3. Results
Mean age of our study subjects was similar across categories of night shift work, ranging from 35.1 years in current shift workers to 36.8 years in former shift workers. The majority of current night shifters ($67\%$) followed a counterclockwise, very rapidly rotating schedule (A), in details: day 1: morning (6:00 a.m.−2:00 p.m. or 7:00 a.m.−2:00 p.m.); day 2: either morning or afternoon (2:00 p.m.−10:00 p.m. or 2:00 p.m.−9:00 PM); day 3: both morning and night (10:00 p.m.−6:00 a.m. or 9:00 p.m.−7:00 a.m.), followed by three rest days (72 h). Eight nurses followed a clockwise, rapidly rotating schedule (B), in details: day 1: morning (7:00 a.m.−2:00 p.m.); day 2: afternoon (2:00 p.m.−9:00 p.m.); day 3: night (9:00 p.m.−7:00 a.m.), followed by two or three rest days (48–72 h). Only five nurses worked on a 12 h schedule (C): day – day – night – (night) – rest – rest. For one subject the information was not available. Never night shift workers had a lower BMI compared to former and current shift workers ($$p \leq 0.057$$). Percent of current smokers increased from $19\%$ among never night shifters to about $33\%$ in current night shifters (Table 1). Descriptive statistics of metabolite concentrations are reported in Supplementary Table S1.
**Table 1**
| Characteristic | Never night shift workers | Former night shift workers | Current night shift workers | P * |
| --- | --- | --- | --- | --- |
| N | 26 | 22 | 43 | |
| Type of night shift work ** | Type of night shift work ** | Type of night shift work ** | Type of night shift work ** | Type of night shift work ** |
| Schedule A—N (%) | - | - | 29 (67) | |
| Schedule B—N (%) | - | - | 8 (19) | |
| Schedule C—N (%) | - | - | 5 (12) | |
| Missing—N (%) | - | - | 1 (2) | |
| Age | Age | Age | Age | Age |
| Mean ± SD | 36.6 ± 5.4 | 36.8 ± 5.4 | 35.1 ± 5.4 | 0.382 |
| Missing | 0 | 0 | 1 | |
| BMI | | | | |
| Mean ± SD | 21.4 ± 2.4 | 23.2 ± 3.9 | 23.1 ± 3.0 | 0.057 |
| Missing | 0 | 0 | 2 | |
| Smoking habit | Smoking habit | Smoking habit | Smoking habit | Smoking habit |
| Former/never smokers—N (%) | 21 (81) | 14 (64) | 27 (63) | |
| Current smokers—N (%) | 5 (19) | 7 (32) | 14 (33) | 0.389 |
| Missing—N (%) | 0 (-) | 1 (4) | 2 (4) | |
Network analysis (Supplementary Figure S1) mainly showed that (i) metabolites belonging to the same category are highly correlated, (ii) serotonin is correlated with taurine, and (iii) taurine is also correlated with aspartic acid.
When comparing current vs. never night shift workers (Figure 1A), 6 SM and several glycerophospholipids, among which 12 PC aa and 9 PC ae, were significantly decreased; while taurine, serotonin, and aspartic acids were significantly increased (with a percent variation of +182.1, +171.0, and +$155.8\%$, respectively). When comparing former vs. never night shift workers (Figure 1B), only taurine emerged as a significantly different metabolite, with a percent variation of +$208.8\%$. The Tobit regression comparing ever (i.e., current + former) vs. never night shift workers returned similar results (Supplementary Figure S2). Comparable findings were also observed when stratifying current night shifters by shift schedule and focusing on nurses following schedule A (Supplementary Figure S3A). Only four metabolites were found to be significantly altered in night shift workers following schedule B (Supplementary Figure S3B) while no alteration was observed when inspecting shift schedule C (Supplementary Figure S3C). No metabolite was found to be associated with increasing number of years worked in night shifts (Supplementary Figure S4). Complete results from Tobit regression models are reported in Supplementary Table S2.
**Figure 1:** *(A, B) Volcano plots showing the results of the Tobit linear regression models considering the metabolites (dependent variables) in relation to night shift work: current vs. never night shift workers (A) and former vs. never night shift workers (B). The models are adjusted for BMI, age, plate, and smoking habit. Each dot represents a metabolite and is displayed based on the percentage variation of its concentration (x-axis) vs. the negative logarithm (base 10) of the FDR p-value (y-axis). The dashed line represents a FDR p-value equal to 0.1.*
Figure 2A shows variable importance scores from the RF algorithm for current vs. never night shift workers. Taurine and aspartic acid were the most important variables discriminating subjects in the two groups, according to both Mean Decrease Accuracy and Mean Decrease Gini. Several PC aa and some PC ae were found to be among the 30 most important metabolites. Figure 2B reports variable importance scores for former vs. never night shift workers: C12.1, aspartic acid and taurine were found to be the three most important metabolites, according to both indexes. ADMA, and several PC aa, PC ae, and sphingolipids were also observed among the 30 most important metabolites. Again, when pooling current and former night shifters, we obtained similar results (Supplementary Figure S5).
**Figure 2:** *(A, B) Variable importance scores plots of the 30 most important metabolites in predicting current vs. never night shift workers (A) and former vs. never night shift workers (B), according to both Mean Decrease Accuracy and Mean Decrease Gini.*
In Supplementary Table S3, 5-fold cross-validation results are reported. RF performance was high in the analysis of current vs. never night shift workers as both sensitivity and specificity were above 0.80, while it was medium-low for former vs. never shifters. When analyzing ever vs. never night shifters, sensitivity was 0.80 while specificity was 0.65.
## 4. Discussion
In the present study, we evaluated the association between exposure to night shift work and metabolites levels, using a targeted metabolomic approach, in a sample of 91 nurses. In particular, when compared to never night shifters, current night shift workers had higher levels of taurine, serotonin and aspartic acid, while lower levels of several glycerophospholipids and sphingolipids. Similar results were observed in ever night shift workers, while the observed associations disappeared when comparing former shift workers to never night shifters, except for taurine levels. Findings across type of night shift schedule, although mostly confirming the overall results, were hampered by the very small number of subjects following schedules B and C. Number of years worked in night shifts did not impact the levels of metabolites.
Serum metabolites in shift workers were previously investigated on 60 subjects from China [25]. Shift workers had altered levels of several lipids, including some glycerophospholipids and sphingolipids, as observed in our study. However, authors also found altered levels of some amines and androgens, and they did not observe any change in serotonin and taurine levels. Differences might relate to the fact that, in the Chinese study, shifters worked two shifts not including the night (i.e., 7:00 a.m.−3:00 p.m.; 3:00 p.m.−11:00 p.m.): as such, their occupational exposure cannot be directly compared to the one in our study. In addition, the applied statistical techniques were different: Huang and colleagues used linear regression models in combination with an Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA), while we used Tobit linear regression models in combination with Random Forests.
A second study [21] investigating male workers, which rotated through 3 weeks of night shifts (10:00 p.m.−6:00 a.m.), followed by 3 weeks of evening shifts (2:00 p.m.−10:00 p.m.) and 3 weeks of early morning shifts (6:00 a.m.−2:00 p.m.), did observe alterations in some lipids (e.g., glycerophospholipids and lysophospholipids) associated with night shifts (i.e., somehow similarly to our results). Nonetheless, the shift scheme is hardly comparable to that experienced by our study population which, in addition, consists of female workers only.
Two other studies conducted in experimental settings simulated night shifts protocols on a small number of healthy volunteers and collected blood samples at repeated time points [22, 23]. Both concluded that the observed rhythmicity of several metabolites was driven mainly by behaviors imposed by the simulated shift schedule rather than by the central circadian clock. Notwithstanding the peculiar differences characterizing the experimental settings of these investigations, their findings do confirm the relevant role of night shift work in influencing health, in particular for what concerns metabolic imbalance.
We observed higher levels of taurine in both current and former night shifters when compared to never night shifters. To exclude this finding could be related to energy drink use, we inspected the distribution of the available variable “drinks other than coffee” and found no differences across categories of night shift (chi-squared p-value = 0.78). Previous investigations documented higher levels of taurine in both humans and rats during periods of sleep deprivation (41–43), a condition typically related to night shift work [44]. In fact, it seems that increased levels of taurine activate the extrasynaptic GABA (A) receptors in the mouse ventrobasal thalamus [45], an area involved in the regulation of the transitions between sleep and wakefulness [46].
Current night shifters showed also higher levels of serotonin. This neurotransmitter is another important factor involved in sleep/wake regulation, functioning primarily to promote wakefulness [42]. In addition, altered levels of both serotonin and taurine have been found to be involved in depression onset (47–49). This is particularly interesting in light of a recent meta-analysis, which estimated a $33\%$ increased risk of depressive symptoms associated with shift work, that rose to more than $70\%$ when restricting the analyses to female workers [16]. In our study, we found serotonin to be positively associated with night shift work in Tobit regression models only and not in the Random Forests analysis: this inconsistency may be explained by the high correlation existing between serotonin and taurine.
Current night shift workers had lower levels of several glycerophospholipids. Decreased levels of such lipids were also found in the plasma sample of breast cancer patients [50], reflecting a higher activity of phospholipase A2 (PLA2), an important pro-inflammatory mediator [51]. On the other hand, higher concentrations of several glycerophospholipids were associated with decreased risk of prostate cancer subtypes, especially those in advanced stage [52]. Positive associations between night shift work and both these cancer types have been mentioned by IARC in supporting the evaluation of NSW as probably carcinogenic to humans (group 2A carcinogen) [5]. Higher levels of phospholipids were also documented by several investigations to be negatively associated with metabolic diseases, as altered concentrations of such lipids were found in subjects with dyslipidemia, hypertension, obesity, insulin resistance or type 2 diabetes (53–58). In particular, it was found that elevated levels of phosphatidylcholines showed a possibly anti-inflammatory role under different conditions (e.g., oxidative stress and ulcerative colitis) (59–61). Indeed, phosphatidylcholines inhibit the upregulation of the inflammatory cytokines tumor necrosis factor alpha and interleukin-6 as well as the actin-assembly in phagosomes and macrophages [59, 61]. In this multifaceted scenario, night shift work emerges as a potentially relevant player in the development of metabolic disorders and cancers.
Another class of lipids we found to be decreased in night shift workers are the sphingolipids. Lower levels of sphingolipids were found in patients with a diagnosis of major depressive disorder: Demirkan and colleagues identified significant negative associations between the sphingomyelin (SM) ratio 23:1 to SM 16:0 and a psychometric depression measure (Center for Epidemiological Studies-Depression Scale: CES-D) [62]. However, subsequent analysis of an independent replication dataset did not confirm previous results. Another study from Liu and colleagues found that several differential lipid species were significantly correlated with depression severity measured by the Hamilton Depression Scale (HAMD) [63]. Moreover, rats exposed to chronic stress had reduced sphingomyelin and dihydrosphingomyelin levels in the prefrontal cortex (PFC) [64]. This region vulnerability fits with previous studies showing that PFC is the brain region displaying major lipid alterations after the use of maprotiline, an antidepressant. In our study population, only four subjects (ever shifters) declared a prolonged use of psychotropic drugs (not better specified). Decreased levels of sphingolipids were also found in subjects with dyslipidemia [56] and with diabetes mellitus [53], even if some other publications showed opposite results [54], indicating that there is no clear pattern between sphingolipids and metabolic disorders.
Levels of aspartic acid were found to be elevated in current shift workers. This is a relative new result. Indeed, few publications on elevated levels of aspartate were published. A study from Guevara-Cruz conducted in Mexico found that levels of aspartate were elevated in a 20-years-old population affected by obesity and insulin resistance [65]. Similar results were also found in a study conducted by Yamada and colleagues, in a Japanese non-diabetic population [66]. Moreover, higher levels of aspartic acid were also found in subjects with epilepsy, as compared to disease-free controls [67]. However, the available research is still too limited and further studies are needed to draw robust conclusions.
The present study has some strengths. This is one of the few studies investigating the effects of night shift work on human serum metabolome. Blood samples were collected within a relatively narrow time window (7:15–9:00 a.m.) to minimize the 24-h variations of metabolites levels [68]. Moreover, we evaluated the investigated associations using two different statistical methodologies (i.e., Tobit linear regression models and Random Forests) to make more robust conclusions. The first ones allow to adjust for individual confounders but are not able to consider correlations among the different metabolites. On the contrary, Random Forests are able to take into proper consideration inter-metabolite correlations but not to adjust for individual confounders. As such, the use of both methods provides a more comprehensive picture of our findings. In addition, we considered observations with non-determinable metabolite levels as left-censored, and applied Tobit linear regression models which are particularly adequate when dealing with dependent variables with censored values [30].
This study has also some limitations. First, it is a cross-sectional study, thus preventing to assess causality. Second, sample size is relatively small, not allowing to obtain optimal prediction results in the application of machine-learning algorithms and to thoroughly investigate the potential role of night shift schedule in influencing our findings. Third, our sample was entirely composed of women, preventing the possibility to detect sex-related differences which have been previously observed with metabolomics data, even if in experimental sleep-deprivation settings [42, 69]. Fourth, all variables including the exposure of interest (i.e., shift work status) as well as all the confounders were self-reported, although the absence of a pathologic outcome should allow to avoid major distortions. Last, given the limited set of available information, we were not able to fully disentangle whether the observed metabolic alterations were related directly to NSW (i.e., by modification of the endogenous circadian clock) or rather due to changes in behavioral (e.g., sleep/wake or feeding/fasting) cycles.
In conclusion, our study, although based on a small sample size, shows an alteration of metabolites levels in night shift workers when compared to never night shifters. In particular, serum concentrations of taurine, serotonin, and aspartic acid were higher, while those of several glycerophospholipids and sphingolipids were lower, independently from number of years worked in night shifts. These findings may shed light on early biological alterations that might be related to adverse health effects of NSW, such as metabolic disorders, cancers, and mental diseases. However, further studies including a larger sample size and male workers are needed to confirm our results.
## 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 Policlinico Hospital (approval number 702_2015). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
Conceptualization: MC, GF, CM, ACP, and SF. Methodology: EB, GF, and MC. Software and formal analysis: EB, GF, and LF. Validation: LF and MC. Investigation and writing—original draft preparation: EB and GF. Resources and data curation: GF, EP, and RM. Writing—review and editing: MC, EP, RM, CM, LF, ACP, and SF. Visualization: MC and GF. Supervision: ACP, SF, and MC. Project administration: CM. Funding acquisition: MC and CM. 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/fpubh.2023.1082074/full#supplementary-material
## References
1. Costa G. **Factors influencing health of workers and tolerance to shift work**. *Theor Issues Ergon Sci.* (2003.0) **4** 4-263. DOI: 10.1080/14639220210158880
2. Vyas MV, Garg AX, Iansavichus AV, Costella J, Donner A. **Shift work and vascular events: systematic review and meta-analysis**. *BMJ* (2012.0). DOI: 10.1136/bmj.e4800
3. Wang XS, Armstrong MEG, Cairns BJ, Key TJ, Travis RC. **Shift work and chronic disease: the epidemiological evidence**. *Occup Med (Chic Ill).* (2011.0) **61** 78-89. DOI: 10.1093/occmed/kqr001
4. Härmä M, Ropponen A, Hakola T, Koskinen A, Vanttola P, Puttonen S. **Developing register-based measures for assessment of working time patterns for epidemiologic Studies**. *Scand J Work Environ Heal.* (2015.0) **41** 268-79. DOI: 10.5271/sjweh.3492
5. **Night shift work**. *IARC Monogr Identif Carcinog Hazards Hum* (2020.0) **124** 1-371
6. Perez JF, Traversini V, Fioriti M, Taddei G, Montalti M, Tommasi E. **Shift and night work management in European companies**. *Qual - Access to Success.* (2019.0) **20** 157-65
7. **Job Flexibilities and Work Schedules-−2017–2018 Data From the American Time Use Survey**. (2019.0)
8. Torquati L, Mielke GI, Brown WJ, Kolbe-Alexander T. **Shift work and the risk of cardiovascular disease. A systematic review and meta-analysis including dose-response relationship**. *Scand J Work Environ Heal.* (2018.0) **44** 229-38. DOI: 10.5271/sjweh.3700
9. Arafa A, Eshak ES, Iso H, Muraki I, Tamakoshi A. **Night work, rotating shift work, and the risk of cancer in japanese men and women: The jacc study**. *J Epidemiol.* (2021.0) **31** 585-92. DOI: 10.2188/jea.JE20200208
10. Sun M, Feng W, Wang F, Li P, Li Z, Li M. **Meta-analysis on shift work and risks of specific obesity types**. *Obes Rev.* (2018.0) **19** 28-40. DOI: 10.1111/obr.12621
11. Zhang Q, Chair SY, Lo SHS, Chau JPC, Schwade M, Zhao X. **Association between shift work and obesity among nurses: a systematic review and meta-analysis**. *Int J Nurs Stud.* (2020.0) **112** 103757. DOI: 10.1016/j.ijnurstu.2020.103757
12. Vetter C, Dashti HS, Lane JM, Anderson SG, Schernhammer ES, Rutter MK. **Night shift work, genetic risk, and type 2 diabetes in the UK biobank**. *Diabetes Care.* (2018.0) **41** 762-9. DOI: 10.2337/dc17-1933
13. Gan Y, Yang C, Tong X, Sun H, Cong Y, Yin X. **Shift work and diabetes mellitus: a meta-analysis of observational studies**. *Occup Environ Med.* (2015.0) **72** 72-8. DOI: 10.1136/oemed-2014-102150
14. Jang TW, Jeong KS, Ahn YS, Choi KS. **The relationship between the pattern of shift work and sleep disturbances in Korean firefighters**. *Int Arch Occup Environ Health.* (2020.0) **93** 391-8. DOI: 10.1007/s00420-019-01496-3
15. Nojkov B, Rubenstein JH, Chey WD, Hoogerwerf WA. **The impact of rotating shift work on the prevalence of irritable bowel syndrome in nurses**. *Am J Gastroenterol.* (2010.0) **105** 842-7. DOI: 10.1038/ajg.2010.48
16. Torquati L, Mielke GI, Brown WJ, Burton NW, Kolbe-Alexander TL. **Shift work and poor mental health: a meta-analysis of longitudinal studies**. *Am J Public Health.* (2019.0) **109** E13-20. DOI: 10.2105/AJPH.2019.305278
17. Harrington JM. **Health effects of shift work and extended hours of work**. *Occup Environ Med.* (2001.0) **58** 68-72. DOI: 10.1136/oem.58.1.68
18. Kecklund G, Axelsson J. **Health consequences of shift work and insufficient sleep**. *BMJ.* (2016.0) **355** 1-13. DOI: 10.1136/bmj.i5210
19. Kuehnbaum NL, Britz-Mckibbin P. **New advances in separation science for metabolomics: resolving chemical diversity in a post-genomic era**. *Chem Rev.* (2013.0) **113** 2437-68. DOI: 10.1021/cr300484s
20. Roberts LD, Souza AL, Gerszten RE, Clish CB. **Targeted metabolomics**. *Curr Protoc Mol Biol.* (2012.0) **1** 1-24. DOI: 10.1002/0471142727.mb3002s98
21. Harding BN, Skene DJ, Espinosa A, Middleton B, Castaño-Vinyals G, Papantoniou K. **Chronobiology international metabolic profiling of night shift work—The HORMONIT study Metabolic profiling of night shift work**. *Hormonit Study* (2022.0) **87** 1562. DOI: 10.1080/07420528.2022.2131562
22. Skene DJ, Skornyakov E, Chowdhury NR, Gajula RP, Middleton B, Satterfield BC. **Separation of circadian- and behavior-driven metabolite rhythms in humans provides a window on peripheral oscillators and metabolism**. *Proc Natl Acad Sci U S A.* (2018.0) **115** 7825-30. DOI: 10.1073/pnas.1801183115
23. Kervezee L, Cermakian N, Boivin DB. **Individual metabolomic signatures of circadian misalignment during simulated night shifts in humans**. *PLoS Biol* (2019.0) **17** 303. DOI: 10.1371/JOURNAL.PBIO.3000303
24. Rotter M, Brandmaier S, Covic M, Burek K, Hertel J, Troll M. **Night shift work affects urine metabolite profiles of nurses with early chronotype**. *Metabolites.* (2018.0) **8** 1-15. DOI: 10.3390/metabo8030045
25. Huang X, Chen X, Zhao S, Hou J, Huang L, Xu J. **Metabolomic profiles of shift workers and day workers a cross-sectional study**. *Obes.* (2021.0) **29** 1074-82. DOI: 10.1002/oby.23164
26. Carugno M, Maggioni C, Crespi E, Bonzini M, Cuocina S, Dioni L. **Night shift work, DNA methylation and telomere length: an investigation on hospital female nurses**. *Int J Environ Res Public Health* (2019.0) **16** 2292. DOI: 10.3390/ijerph16132292
27. Siskos AP, Jain P, Römisch-margl W, Bennett M, Asad Y, Marney L. **Inter-laboratory reproducibility of a targeted metabolomics platform for analysis of human serum and plasma**. *Anal Chem.* (2019.0) **89** 656-65. DOI: 10.1021/acs.analchem.6b02930.Inter-laboratory
28. Frigerio G, Favero C, Savino D, Mercadante R, Albetti B, Dioni L. **Plasma metabolomic profiling in 1391 subjects with overweight and obesity from the sphere study**. *Metabolites* (2021.0) **11** 194. DOI: 10.3390/metabo11040194
29. **“Appendix 7.1 Data Normalization – Based on QC Target Values.”**. *MetIDQ* (2020.0)
30. McDonald JF, Moffitt RA. **The uses of tobit analysis**. *Rev Econ Stat.* (1980.0) **62** 318. DOI: 10.2307/1924766
31. Tobin J. **Estimation of relationships for limited dependent variables**. *Econometrica.* (1958.0) **26** 24. DOI: 10.2307/1907382
32. Altman NS. **An introduction to kernel and nearest-neighbor nonparametric regression**. *Am Stat.* (1992.0) **46** 175-85. DOI: 10.1080/00031305.1992.10475879
33. Benjamini Y, Hochberg Y. **Controlling the false discovery rate: a practical and powerful approach to multiple testing**. *J R Stat Soc Ser B.* (1995.0) **57** 289-300. DOI: 10.1111/j.2517-6161.1995.tb02031.x
34. Breiman L. **Random Forests**. *Mach Learn.* (2001.0) **45** 5-32
35. 35.R Core Team. R: A Language and Environment for E. Vienna, Austria: R Found Stat Comput.. *R: A Language and Environment for E*
36. Wickham H, Averick M, Bryan J, Chang W, McGowan L, François R. **Welcome to the tidyverse**. *J Open Source Softw.* (2019.0) **4** 1686. DOI: 10.21105/joss.01686
37. Kleiber C, Zeileis A. *Applied Econometrics with R* (2008.0) **3** 6. DOI: 10.1007/978-0-387-78171-6
38. Pedersen T. *Ggraph: An Implementation of Grammar of Graphics for Graphs and Networks* (2022.0)
39. Pedersen T. *Tidygraph: A Tidy API for Graph Manipulation* (2023.0)
40. Breiman L, Cutler A. **Random Forest: Breiman Cutler's Random Forests for Classification Regression.**. (2022.0)
41. Mohammed HS, Aboul Ezz HS, Khadrawy YA, Noor NA. **Neurochemical and electrophysiological changes induced by paradoxical sleep deprivation in rats**. *Behav Brain Res.* (2011.0) **225** 39-46. DOI: 10.1016/j.bbr.2011.06.018
42. Davies SK, Ang JE, Revell VL, Holmes B, Mann A, Robertson FP. **Effect of sleep deprivation on the human metabolome**. *Proc Natl Acad Sci U S A.* (2014.0) **111** 10761-6. DOI: 10.1073/pnas.1402663111
43. Giskeødegård GF, Davies SK, Revell VL, Keun H, Skene DJ. **Diurnal rhythms in the human urine metabolome during sleep and total sleep deprivation**. *Sci Rep.* (2015.0) **5** 1-11. DOI: 10.1038/srep14843
44. Costa G. **Sleep deprivation due to shift work**. *Handb Clin Neurol.* (2015.0) **131** 437-46. DOI: 10.1016/B978-0-444-62627-1.00023-8
45. Jia F, Yue M, Chandra D, Keramidas A, Goldstein PA, Homanics GE. **Taurine is a potent activator of extrasynaptic GABAA receptors in the thalamus**. *J Neurosci.* (2008.0) **28** 106-15. DOI: 10.1523/JNEUROSCI.3996-07.2008
46. Coulon P, Budde T, Pape HC. **The sleep relay-the role of the thalamus in central and decentral sleep regulation**. *Pflugers Arch Eur J Physiol.* (2012.0) **463** 53-71. DOI: 10.1007/s00424-011-1014-6
47. Altamura C, Maes M, Dai J, Meltzer HY. **Plasma concentrations of excitatory amino acids, serine, glycine, taurine and histidine in major depression**. *Eur Neuropsychopharmacol.* (1995.0) **5** 71-5. DOI: 10.1016/0924-977X(95)00033-L
48. Perry TL, Bratty PJA, Hansen S, Kennedy J, Urquhart N, Dolman CL. **Hereditary mental depression and parkinsonism with taurine deficiency**. *Arch Neurol.* (1975.0) **32** 108-13. DOI: 10.1001/archneur.1975.00490440058009
49. Ressler J, Karry CB. **Role of serotonergic and noradrenergic systems in the pathophysiology of depression and axiety disorders**. *Depress Anxiety* (2000.0) **12** 2-19. DOI: 10.1002/1520-6394(2000)12:1+<2::AID-DA2>3.0.CO;2-4
50. Qiu Y, Zhou B, Su M, Baxter S, Zheng X, Zhao X. **Mass spectrometry-based quantitative metabolomics revealed a distinct lipid profile in breast cancer patients**. *Int J Mol Sci.* (2013.0) **14** 8047-61. DOI: 10.3390/ijms14048047
51. Cummings BS, McHowat J, Schnellmann RG. **Phospholipase A2s in cell injury and death**. *J Pharmacol Exp Ther.* (2000.0) **294** 793-9. PMID: 10945826
52. Schmidt JA, Fensom GK, Rinaldi S, Scalbert A, Appleby PN, Achaintre D. **Pre-diagnostic metabolite concentrations and prostate cancer risk in 1,077 cases and 1,077 matched controls in the European prospective investigation into cancer and nutrition**. *BMC Med.* (2017.0) **15** 1-14. DOI: 10.1186/s12916-017-0885-6
53. Meikle PJ, Wong G, Barlow CK, Weir JM, Greeve MA, MacIntosh GL. **Plasma lipid profiling shows similar associations with prediabetes and type 2 diabetes**. *PLoS ONE* (2013.0) **8** 4341. DOI: 10.1371/journal.pone.0074341
54. Weir JM, Wong G, Barlow CK, Greeve MA, Kowalczyk A, Almasy L. **Plasma lipid profiling in a large population-based cohort**. *J Lipid Res.* (2013.0) **54** 2898-908. DOI: 10.1194/jlr.P035808
55. Pietiläinen KH, Sysi-Aho M, Rissanen A, Seppänen-Laakso T, Yki-Järvinen H, Kaprio J, Orešič M. **Acquired obesity is associated with changes in the serum lipidomic profile independent of genetic effects—A monozygotic twin study**. *PLoS ONE* (2007.0) **2** 218. DOI: 10.1371/journal.pone.0000218
56. Ke C, Zhu X, Zhang Y, Shen Y. **Metabolomic characterization of hypertension and dyslipidemia**. *Metabolomics.* (2018.0) **14** 1-11. DOI: 10.1007/s11306-018-1408-y
57. Graessler J, Schwudke D, Schwarz PEH, Herzog R, Schevchenko A, Bornstein SR. **Top-down lipidomics reveals ether lipid deficiency in blood plasma of hypertensive patients**. *PLoS ONE* (2009.0) **4** 6261. DOI: 10.1371/journal.pone.0006261
58. Dietrich S, Floegel A, Weikert C, Pischon T, Boeing H, Drogan D. **Identification of serum metabolites associated with incident hypertension in the European prospective investigation into cancer and nutrition-potsdam study**. *Hypertension.* (2016.0) **68** 471-7. DOI: 10.1161/HYPERTENSIONAHA.116.07292
59. Jung YY, Nam Y, Park YS, Lee HS, Hong SA, Kim BK. **Protective effect of phosphatidylcholine on lipopolysaccharide-induced acute inflammation in multiple organ injury**. *Korean J Physiol Pharmacol.* (2013.0) **17** 209-16. DOI: 10.4196/kjpp.2013.17.3.209
60. Al-Orf SM. **Effect of oxidized phosphatidylcholine on biomarkers of oxidative stress in rats**. *Indian J Clin Biochem.* (2011.0) **26** 154-60. DOI: 10.1007/s12291-010-0064-4
61. Treede I, Braun A, Sparla R, Kühnel M, Giese T, Turner JR. **Anti-inflammatory effects of phosphatidylcholine**. *J Biol Chem.* (2007.0) **282** 27155-64. DOI: 10.1074/jbc.M704408200
62. Demirkan A, Isaacs A, Ugocsai P, Liebisch G, Struchalin M, Rudan I. **Plasma phosphatidylcholine and sphingomyelin concentrations are associated with depression and anxiety symptoms in a Dutch family-based lipidomics study**. *J Psychiatr Res.* (2013.0) **47** 357-62. DOI: 10.1016/j.jpsychires.2012.11.001
63. Liu X, Li J, Zheng P, Zhao X, Zhou C, Hu C. **Plasma lipidomics reveals potential lipid markers of major depressive disorder**. *Anal Bioanal Chem.* (2016.0) **408** 6497-507. DOI: 10.1007/s00216-016-9768-5
64. Oliveira TG, Chan RB, Bravo FV, Miranda A, Silva RR, Zhou B. **The impact of chronic stress on the rat brain lipidome**. *Mol Psychiatry.* (2016.0) **21** 80-8. DOI: 10.1038/mp.2015.14
65. Guevara-Cruz M, Vargas-Morales JM, Méndez-García AL, López-Barradas AM, Granados-Portillo O, Ordaz-Nava G. **Amino acid profiles of young adults differ by sex, body mass index and insulin resistance**. *Nutr Metab Cardiovasc Dis.* (2018.0) **28** 393-401. DOI: 10.1016/j.numecd.2018.01.001
66. Yamada C, Kondo M, Kishimoto N, Shibata T, Nagai Y, Imanishi T. **Association between insulin resistance and plasma amino acid profile in non-diabetic Japanese subjects**. *J Diabetes Investig.* (2015.0) **6** 408-15. DOI: 10.1111/jdi.12323
67. AlGawwam G, Mnaather AA, Al-Mahdawi A, Fawzi HA, Shareef LG. **serum aspartic acid ?as a marker of epilepsy**. *Glob J Health Sci.* (2019.0) **11** 142. DOI: 10.5539/gjhs.v11n2p142
68. Ang JE, Revell V, Mann A, Mäntele S, Otway DT, Johnston JD. **Identification of human plasma metabolites exhibiting time-of-day variation using an untargeted liquid chromatography-mass spectrometry metabolomic approach**. *Chronobiol Int.* (2012.0) **29** 868-81. DOI: 10.3109/07420528.2012.699122
69. Honma A, Revell VL, Gunn PJ, Davies SK, Middleton B, Raynaud FI. **Effect of acute total sleep deprivation on plasma melatonin, cortisol and metabolite rhythms in females**. *Eur J Neurosci.* (2020.0) **51** 366. DOI: 10.1111/EJN.14411
|
---
title: 'Acceptability of patient-centered, multi-disciplinary medication therapy management
recommendations: results from the INCREASE randomized study'
authors:
- Noah I. Smith
- Ashley I. Martinez
- Mark Huffmyer
- Lynne Eckmann
- Rosmy George
- Erin L. Abner
- Gregory A. Jicha
- Daniela C. Moga
journal: BMC Geriatrics
year: 2023
pmcid: PMC9999619
doi: 10.1186/s12877-023-03876-4
license: CC BY 4.0
---
# Acceptability of patient-centered, multi-disciplinary medication therapy management recommendations: results from the INCREASE randomized study
## Abstract
### Background
Polypharmacy and inappropriate medications may be a modifiable risk factor for Alzheimer’s Disease and Related Dementias (ADRD). Medication therapy management (MTM) interventions may mitigate medication-induced cognitive dysfunction and delay onset of symptomatic impairment. The objective of the current study is to describe an MTM protocol for a patient-centered team intervention (pharmacist and non-pharmacist clinician) in a randomized controlled trial (RCT) directed at delaying the symptomatic onset of ADRD.
### Methods
Community dwelling adults 65 + years, non-demented, using ≥ 1 potentially inappropriate medications (PIM) were enrolled in an RCT to evaluate the effect of an MTM intervention on improving medication appropriateness and cognition (NCT02849639). The MTM intervention involved a three-step process: [1] pharmacist identified potential medication-related problems (MRPs) and made initial recommendations for prescribed and over-the-counter medications, vitamins, and supplements; [2] study team reviewed all initial recommendations together with the participants, allowing for revisions prior to the finalized recommendations; [3] participant responses to final recommendations were recorded. Here, we describe initial recommendations, changes during team engagement, and participant responses to final recommendations.
### Results
Among the 90 participants, a mean 6.7 ± 3.6 MRPs per participant were reported. Of the 259 initial MTM recommendations made for the treatment group participants ($$n = 46$$), $40\%$ percent underwent revisions in the second step. Participants reported willingness to adopt $46\%$ of final recommendations and expressed need for additional primary care input in response to $38\%$ of final recommendations. Willingness to adopt final recommendations was highest when therapeutic switches were offered and/or with anticholinergic medications.
### Conclusion
The evaluation of modifications to MTM recommendations demonstrated that pharmacists’ initial MTM recommendations often changed following the participation in the multidisciplinary decision-making process that incorporated patient preferences. The team was encouraged to see a correlation between engaging patients and a positive overall response towards participant acceptance of final MTM recommendations.
### Trial registration
Study registration number: clinicaltrial.gov NCT02849639 registered on $\frac{29}{07}$/2016.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12877-023-03876-4.
## Background
Many prior studies have provided evidence that medication therapy management (MTM) can lead to improved health and economic outcomes [1–4]. MTM involves five core components: availability of a personal medication record, medication therapy review, development of a medication-related action plan, intervention and/or referral, and documentation and follow-up of medication changes or lack thereof [5–7]. Though most MTM services share these five basic elements, there is heterogeneity in how these services are operationalized. Specifically, there is variability in how potentially inappropriate medications (PIMs) are identified, whether certain medications are targeted specifically, the types of recommendations made, and patient’s acceptance of the proposed changes from an MTM intervention. Additionally, patient and pharmacist engagement with prescribing clinicians varies, [5–7] though evidence shows that pharmacist-prescriber-clinician teams engaging together in MTM activities results in better medication optimization outcomes [2, 8–10]. It is important to characterize MTM-related services in collaborative practices in order to estimate their impact on patient health outcomes, especially for MTM services targeting vulnerable populations such as older adults receiving PIMs.
We recently completed the INtervention for Cognitive Reserve Enhancement in delaying the onset of Alzheimer's Symptomatic Expression (INCREASE) study, a randomized controlled trial where we tested an MTM intervention that actively involved the patient, a board-certified geriatric pharmacy specialist (BCGP), and a non-pharmacist clinician [11, 12]. INCREASE was designed to evaluate the effect of the MTM intervention on changes in medication appropriateness and cognitive function; study data included comprehensive information on health history, medication use and experience with medication taking, as well as the process of implementing the MTM intervention. We previously reported on the successful implementation of the MTM intervention that translated into an improved medication appropriateness at the one-year follow-up [12]. The current study characterizes the stepwise process of delivering the MTM intervention in the INCREASE trial with the goal of helping to fill a qualitative gap in the literature surrounding MTM interventions, specifically focused on patient-centered, multidisciplinary approaches. The specific approach described, including details of the process, provides a model for future evidence based, multidisciplinary MTM interventions that may be implemented rationally in practice.
The objectives of the current manuscript are twofold: [1] describe the recommendations made by the study BCGPs using participant-reported medical and medication histories for all INCREASE participants, prior to randomization to either the MTM intervention (specific medication recommendations plus provision of educational materials on inappropriate medication use) or usual care (i.e., only provision of educational materials on inappropriate medication use), and [2] describe final recommendations for patients randomized to the MTM intervention. The second objective describes (a) revisions to the preliminary baseline MTM recommendations over the course of the intervention, and (b) participant response to the MTM recommendations following the intervention.
## INCREASE study overview
The INCREASE study was a randomized controlled trial enrolling community-dwelling adults 65 years and older who did not have dementia and were using at least one PIM as defined in the 2015 Beers Criteria (the most recent version at the time of the study) [13]. Complete details of the INCREASE protocol and results are available elsewhere and briefly described below [11, 12]. After 1:1 randomization that was stratified based on baseline amyloid burden, participants randomized to the control group received usual care with educational pamphlets on medication appropriateness for older adults and risks associated with polypharmacy. In addition to educational materials, participants randomized to the MTM intervention met with the BCGP and a non-pharmacist study clinician (e.g., nurse practitioner, neurologist) to discuss the baseline recommendations. This meeting allowed for 1) participant education on risks, benefits, and alternatives to optimize medication use; and 2) the collection of additional relevant information, including participant beliefs, preferences, and treatment goals. During the MTM team meeting, final recommendations were formalized, and the details of any relevant revisions to the baseline recommendation were noted in the pre-specified data collection forms.
The INCREASE study was approved by the University of Kentucky Institutional Review Board (IRB #43239) and all the study participants provided informed consent. The protocol for the study was registered on clinicaltrials.gov (NCT02849639) on $\frac{29}{07}$/2016, in accordance with the relevant guidelines and regulations or in accordance with the Declaration of Helsinki. Study data were collected and managed using the Research Electronic Data Capture (REDCap), a secure, web-based software platform designed to support data capture for research studies [14, 15].
## Baseline recommendations (all INCREASE study participants)
Before randomization, comprehensive medication reviews were conducted by BCGPs for all participants using participant-reported medical conditions and information on dose, frequency, indication, duration of treatment, tolerability, and adverse drug reactions for all prescription medications, vitamins, and supplements. The BCGP medication review process involved 1) assessing the clinical appropriateness of each medication using the Beers Criteria [13] and Medication Appropriateness Index (MAI); [16] 2) evaluating potential drug-drug and drug-disease interactions in accord with the above and also taking into account prescription label information; and 3) assessing whether medication regimens followed relevant disease-specific evidence-based guidelines [13, 17, 18]. Of note, blood laboratory work results, electronic medical records, and previous therapies (e.g., medication failures) were not available to BCGPs when devising baseline recommendations, but were available to the clinician member of the MTM team. Following randomization, the MTM recommendations were only shared with those participants randomized to the intervention group ($$n = 46$$). Recommendations for the control group were recorded in the study database but not shared with those participants.
During the INCREASE study period, the pharmacy team of two BCGPs utilized drug and health information resources (e.g., Lexicomp and UpToDate [Wolters Kluwer Health Inc. Riverwoods, IL]), Beers Criteria [13], relevant guidelines (e.g., Diabetes Standards of Care [17] and Clinical Practice Guidelines for Hypertension [18]), and clinical judgement to justify their recommendations. Each recommendation was reviewed by both BCGPs and a consensus pharmacy recommendation was decided via discussion. Detailed information for each recommendation was then entered into a series of pre-specified study protocol data collection forms, allowing for systematic categorization of recommendations as either: 1) medication discontinuation with or without tapering; 2) switch to a different medication; 3) dose adjustment (e.g., decrease dose, adjust dose for organ function/tolerability, or increase dose); 4) new medication initiation; 5) drug or disease monitoring recommendation (e.g., vital signs, falls risk, sedation); or 6) a non-pharmacologic recommendation (e.g., sleep hygiene, avoiding gastroesophageal reflux triggers, referral for diagnostic workup). Baseline recommendations were also categorized by pharmacologic class and over the counter (OTC) or supplement status of the medication prompting a baseline MTM recommendation. A full schematic for medication categorization is available in the supplementary material (see Supplementary Table S1).
## Final recommendations (MTM intervention group only)
After 1:1 stratified randomization, study pharmacists met with the participant and study clinician to deliver the MTM intervention. During the intervention, the team gathered further information from the patient and discussed baseline recommendations together, in-person, with additional context provided by the participant on their health status, needs, and preferences. Because health status and medication use in participants may have changed in the time between the baseline assessment and the initial MTM recommendation, comparison of baseline to final recommendations was limited to those baseline MTM recommendations that proposed medication changes at the time of the initial MTM study visit. The non-medication related recommendations (see supplementary material for additional information) were discussed during the team MTM intervention, but they were not included in the present analysis.
Participant responses to each final MTM recommendation for participants randomized to the MTM intervention were collected at the conclusion of the initial MTM intervention visit using a standardized form where the participant selected his or her response to the recommendation as 1) willing to change, 2) refusing to change, 3) needing to confer with a primary care provider or other specialist (e.g., cardiologist), or 4) not applicable (e.g., the participant had already discontinued the medication, dose adjustment was no longer warranted per clinical judgement). In this manuscript we are describing in detail the immediate participant response as recorded following the baseline intervention. The impact of the intervention on medication appropriateness is described in detail elsewhere [12].
## Baseline characteristics
Of the 104 participants screened, 90 were eligible and randomized in the INCREASE study. Of these, 46 participants were randomized to the MTM intervention group. The mean (SD) age at enrollment was 73.9 years (6.0). The majority of the participants reported female gender ($64\%$) and white race ($89\%$), with an average of 16.5 (2.8) years of education. The mean Charlson Comorbidity Index score was 1.9 (1.9), with participants reporting an average of 12.8 (4.8) total medications 2.4 (1.4) medications per participant were identified as PIMs per 2015 Beers Criteria. Supplementary Table S2 provides additional information on baseline characteristics for all the participants in the INCREASE study as well as for those randomized to the MTM intervention.
## Baseline recommendations (all INCREASE study participants)
A total of 602 pre-randomization recommendations were made across the 90 INCREASE participants, averaging 6.7 ± 3.3 MTM recommendations per participant and ranging from 1 to 17 baseline recommendations per participant (median [IQR] of 7 [4, 8.9]). Table 1 shows the distribution of medication categories associated with baseline recommendations and the types of recommendations provided. Table 1Baseline MTM recommendations by medication category† and recommendation type among all INCREASE trial participants ($$n = 90$$)RecommendationsMTM recommendations (total: $$n = 602$$)N%Medication category†Cardiometabolic$13822.9\%$Gastrointestinal$10216.9\%$Pain management$8714.5\%$Anticholinergics$7712.8\%$Vitamins and supplements$7612.6\%$Neuropsychiatric$6711.1\%$Other$559.1\%$Recommendation typeDose adjustment$17028.2\%$Switch to preferred agent$16627.6\%$Drug and disease monitoring$10116.8\%$Non-pharmacologic therapy$7712.8\%$Discontinuation$457.5\%$Initiation of new medication$437.1\%$†See supplementary Table S1 for full listing of medication categories The most common class of medications with recommendations were cardiometabolic agents ($$n = 138$$, $23\%$), followed by medications for gastrointestinal conditions ($$n = 102$$, $17\%$), pain management ($$n = 87$$, $15\%$), anticholinergics ($$n = 77$$, $13\%$), vitamins and supplements ($$n = 76$$, $13\%$), neuropsychiatric agents ($$n = 67$$, $11\%$), and other medications ($$n = 55$$, $9\%$). Across all baseline recommendations, one-third ($$n = 201$$, $33\%$) were prompted by use of PIMs available on the US market as over-the counter (OTC) products without a prescription. The most frequent OTC medications included proton pump inhibitors, vitamins/supplements, antihistamines, OTC non-steroidal anti-inflammatories, aspirin, and H2 receptor antagonists.
The most common type of baseline recommendation was continuation of therapy with dose adjustment (e.g., decrease pain medication dose, intensify antihypertensive medication dose) ($$n = 170$$, $28\%$). Second most common were therapeutic switches to a less risky pharmacotherapeutic alternative ($$n = 166$$, $28\%$; e.g., de-escalate from a proton pump inhibitor to a H2 receptor antagonist ± calcium-based antacid; switch from a first-generation to non-sedating second-generation antihistamine). Monitoring ($$n = 101$$, $17\%$) and non-pharmacologic recommendations ($$n = 76$$, $13\%$) accounted for about one-third of all baseline MTM recommendations. The most frequent monitoring recommendations involved recommending objective testing (e.g., blood pressure, blood chemistry/organ function tests) and recording self-reported measures (e.g., dizziness, pain). Non-pharmacologic recommendations most frequently involved counseling for fall prevention strategies with and without physical therapy referral, dietary and lifestyle changes for gastrointestinal conditions, non-pharmacologic pain management, and sleep hygiene.
Although recommendations to discontinue medications were relatively less frequent ($$n = 46$$, $8\%$), those medications most commonly associated with a baseline MTM recommendation to discontinue included vitamins/supplements and medications with therapeutic duplication (e.g., participant was taking two separate antihistamines for seasonal allergies). All recommendations for initiation of a new medication ($$n = 43$$, $7\%$) involved treating an unmet clinical need and/or initiating a preventative medication, most often a guideline-recommended statin or aspirin in the setting of cardiovascular risk factors.
## Final recommendations (Intervention Group only)
Following randomization, INCREASE participants who were assigned to MTM ($$n = 46$$) met with the BCGP and a non-pharmacist clinician. There were 296 baseline recommendations across the MTM arm’s participants. Of these, 37 recommendations ($12.5\%$) proposed at baseline did not relate directly to a medication change and were therefore excluded from the final recommendation analysis included in this manuscript. An account of these 37 excluded recommendations is provided in supplementary table S3. Finalized, unblinded MTM recommendations that were directly related to a medication change, comprised 259 of the original 602 blinded baseline recommendations, averaging 5.6 (SD 2.3) MTM recommendations per participant.
The distribution of final recommendations by medication category was as follows: cardiometabolic ($$n = 58$$, $22\%$), pain management ($$n = 42$$, $16\%$), vitamins and supplements ($$n = 38$$, $15\%$), anticholinergics ($$n = 32$$, $12\%$), gastrointestinal ($$n = 32$$, $12\%$), neuropsychiatric ($$n = 31$$, $12\%$), and other ($$n = 26$$, $10\%$). The distribution of by recommendation type was as follows: dose adjustment ($$n = 98$$, $38\%$), switch to preferred agent ($$n = 92$$, $36\%$), drug and disease monitoring ($$n = 30$$, $12\%$), discontinuation ($$n = 26$$, $10\%$), and initiation of a new medication ($$n = 13$$, $5\%$).
Table 2 shows the results of the patient-pharmacist-clinician team MTM interventions after randomization. Less than half of the baseline recommendations were revised through the team discussion and deliberation process ($$n = 104$$, $40\%$). Baseline recommendations were least likely to be revised for vitamins/supplements and cardiometabolic medications, or with a recommended dose adjustment or new initiation. Conversely, baseline recommendations were the most likely to be revised when involving GI therapy and pain management medications, or for recommended medication monitoring or discontinuation. The most frequent reasons for revisions were due to missing information relevant to the participant’s medical history (e.g., a missing diagnosis for Barrett’s esophagus warranting proton pump inhibitor use) and/or missing medication information (e.g., previous failure or intolerability of a guideline-preferred pharmacotherapeutic agent).Table 2Revision status of MTM recommendations through intervention delivery, and participant responses to final MTM recommendations by medication category and recommendation type ($$n = 259$$ final recommendations) among INCREASE intervention group participants ($$n = 46$$ participants) aRecommendationsRecommendation was revised during the MTM interventionParticipant response to final MTM recommendation after revision, if applicableWilling to change ($$n = 118$$)Must first confer with another provider* ($$n = 99$$)Refusal to change ($$n = 15$$)Not applicable** ($$n = 27$$)N (%)N (%)N (%)N (%)N (%)Medication categoryCardiometabolic ($$n = 58$$)19 ($33\%$)19 ($33\%$)27 ($47\%$)1 ($2\%$)11 ($19\%$)Pain management ($$n = 42$$)23 ($55\%$)20 ($48\%$)15 ($36\%$)4 ($10\%$)3 ($7\%$)Vitamins and supplements ($$n = 38$$)11 ($29\%$)20 ($53\%$)12 ($32\%$)4 ($11\%$)2 ($5\%$)Anticholinergics ($$n = 32$$)13 ($41\%$)19 ($59\%$)10 ($31\%$)1 ($3\%$)2 ($6\%$)Gastrointestinal ($$n = 32$$)19 ($59\%$)14 ($44\%$)15 ($47\%$)1 ($3\%$)2 ($6\%$)Neuropsychiatric ($$n = 31$$)13 ($42\%$)12 ($39\%$)16 ($52\%$)1 ($3\%$)2 ($6\%$)Other ($$n = 26$$)6 ($23\%$)14 ($54\%$)4 ($15\%$)3 ($12\%$)5 ($19\%$)Recommendation typeDose adjustment ($$n = 98$$)36 ($37\%$)38 ($39\%$)44 ($45\%$)4 ($4\%$)12 ($12\%$)Switch to preferred agent ($$n = 92$$)45 ($49\%$)48 ($52\%$)34 ($37\%$)3 ($3\%$)7 ($8\%$)Drug and disease monitoring ($$n = 30$$)9 ($30\%$)14 ($47\%$)8 ($27\%$)2 ($7\%$)6 ($20\%$)Discontinuation ($$n = 26$$)9 ($35\%$)13 ($50\%$)8 ($31\%$)5 ($19\%$)0 ($0\%$)Initiation ($$n = 13$$)5 ($38\%$)5 ($38\%$)5 ($38\%$)1 ($8\%$)2 ($15\%$)aCell percentages are displayed as a percent of the row total*Another provider could be a primary care physician, specialist such as a cardiologist, or other non-study prescribing clinician**Reasons for lack of applicability included medication use having been appropriately modified since baseline medication use information was collected, or the proposed medication change was no longer clinically relevant given additional information from the participant and/or MTM team discussion Upon receiving the finalized MTM recommendations, participants responded about half the time that they were willing to make the changes proposed ($$n = 118$$, $46\%$), and often needed to confer with a primary care provider or other clinical specialist ($$n = 99$$, $38\%$) before making a decision, but rarely refused to make the proposed changes ($$n = 15$$, $6\%$). In some cases ($$n = 27$$, $10\%$), the recommendation was no longer clinically relevant and participant responses were recorded as not applicable. Lack of applicability arose from medication use having been appropriately modified since baseline medication use information was collected ($$n = 11$$), or from the proposed medication change no longer being clinically relevant given additional information from the participant and/or MTM team discussion ($$n = 16$$). A full account of these 27 recommendations in provided in supplementary table S4.
Participant willingness to adopt recommended MTM changes was highest for vitamins/supplements and anticholinergic agents, and for recommendations involving a pharmacotherapeutic switch. Participants most often responded that they needed to confer with a primary care provider or other specialist when the MTM recommendations included psychiatric, GI, and cardiometabolic medications, or for dose adjustments or medication switches. Participant refusal to adopt final recommended changes ($$n = 15$$, $6\%$) was low across all medication categories and recommendation types in the INCREASE trial MTM intervention. Refusal was highest among recommendations involving vitamins and supplements ($$n = 4$$) or pain management ($$n = 4$$), as well as for recommendations involving medication discontinuation ($$n = 5$$).
## Discussion
This study describes MTM recommendations for participants enrolled in the INCREASE trial. The most common medication categories flagged at baseline included 1) cardiometabolic medications, 2) gastrointestinal medications, 3) pain management medications, 4) anticholinergics, and 5) vitamins/supplements. The most common types of recommendations made at baseline were 1) dose adjustments and 2) switches to more appropriate therapeutic alternatives.
Notably, BCGP recommendations were not strictly medication related. In this study, many MTM recommendations did not directly involve a medication change, but rather addressed other potential medical problems (e.g., disease monitoring, referral for diagnostic workup or physical therapy, addition of non-pharmacologic therapies). Each of the top five medication categories identified in the analysis for baseline recommendations included at least some OTC medication options, and one-third of baseline MTM recommendations involved a medication available OTC. OTC products are available without a prescription and were identified frequently as PIMs ($13\%$ of all baseline recommendations and $15\%$ of final recommendations). Thus, our study points to the importance of educating patients on the risk–benefit profile of OTCs and the role of pharmacists in OTC stewardship.
Among the final MTM recommendations analyzed, $40\%$ underwent revision compared to the baseline MTM recommendation provided. This reflects the potential for several factors to influence recommendations as more information is gathered in a multidisciplinary MTM intervention. Notably, input from the patient on previous therapies, medication tolerability, feasibility/adherence, and condition severity may help inform the MTM team’s final decision-making process. Our results demonstrate that engaging the patient in a team-based intervention may result in patient-motivated revisions to baseline recommendations. This comparison of pre-intervention recommendations to final recommendations after team deliberation has not been discussed in previous literature.
Participant responses indicated willingness to make recommended changes about half of the time and a need to confer with a primary care provider or other clinical specialist about one third of the time. This was interpreted as generally positive, since participants were most often willing to either accept the final recommendation as specific, or to further engage with another healthcare provider to seek additional medical advice. While participant refusal to change was generally low, our findings suggest that patients may be less willing to adopt MTM recommendations for certain categories of medications or for certain recommendation types. Previous literature has addressed acceptability of MTM recommendations [19–23]; however, the recommendation type and medication category have not been described in relation to participant willingness to make changes. Further research is needed to determine if willingness to adhere long-term to recommendations is impacted by the type of recommendation and medication in question.
Though extensive medication and medical histories were collected from participants, the baseline recommendations were limited to self-reported information before randomization, and complete clinically relevant information was not always available to the BCGP at baseline (e.g., renal function from an electronic medical record). This finding indicates that pharmacists engaged in MTM processes need access to relevant clinical information and an opportunity for direct engagement with prescribers and the patient who have first-hand knowledge of such clinical variables. In addition, chart review may not capture all information necessary to make a patient-centered MTM recommendations, which has not been reviewed in previous literature [19–23]. As health status, medication use and tolerability change over time, there is a need to routinely review previous recommendations and adjust them as needed to reflect the patient’s current needs.
There are several limitations to this study. The 2015 Beers Criteria [13]. was used in the study, which was the latest version at the time of the study. During the INCREASE trial, updated Beers Criteria were published by the American Geriatrics Society in 2019 [24]. As an example, the 2015 Beers Criteria recommended caution in aspirin use for primary cardiovascular event prevention among adults aged ≥ 80 years. In contrast, the 2019 Beers Criteria expanded the recommendation to caution in aspirin use for both primary cardiovascular event and colorectal cancer prevention for adults aged ≥ 70 years. This may limit generalizability as medical treatments, guidelines, and prescribing patterns evolve over time in response to scientific evidence. Another limitation of this descriptive study was that the INCREASE participant experiences may not be generalizable to populations that have different distributions of demographic and health characteristics. Similarly, local prescribing practices and the use of PIMs observed in the INCREASE trial may not be representative of the entire US population today. Additionally, the number of study pharmacists and prescribing clinicians was small. Ability for multiple pharmacists to independently review and adjudicate the categorization of MTM recommendations would strengthen future studies. Though this study adds to the body of descriptive literature on baseline MTM trial recommendations, further studies in diverse populations are needed to identify culturally appropriate MTM strategies, as well as to allow a more detailed examination of prescribing inequities that might influence MTM outcomes over time. When teams gather to critically evaluate an individual’s medication use process (i.e., diagnostician/prescriber, dispenser, and medication user), open dialogue may facilitate transparency in strategic medication use decisions. Negotiation of an evidence-based approach to medication use should be to the individual’s unique combination of diseases, medications, clinical status, and very importantly – personal preference that may have its roots in cultural/social/racial/ethnic diversity.
It is important to note that not all MTM interventions are equivalent. The INCREASE trial modeled its intervention on a foundation of multidisciplinary team interaction with active participant engagement. This is often beyond the scope of traditional, community pharmacy-based MTM models in practice today. The present results suggest that medical advice from a patient-centered team with multiple healthcare perspectives is both appealing to patients and may elicit a stronger patient acceptance of MTM recommendations. Further studies characterizing patient responses to different MTM models are needed to determine whether the qualities such as mode of delivery and multidisciplinary involvement impact long-term recommendation adherence, and/or influences patient outcomes.
## Conclusion
Multidisciplinary interventions such as the pharmacist-clinician-patient MTM team used in the INCREASE study may hold promise for improving health-related outcomes among community dwelling persons. Thorough characterization of MTM interventions is needed to specifically describe the nuances of MTM approaches for making recommendations. It is also critical for guiding future endeavors in the area of MTM science. The present data demonstrate that the recommendations suggested by patient-centered multidisciplinary healthcare teams can be dynamic and complex, and that participant responses may vary depending on the medication targeted and the type of recommendation proposed.
## Supplementary Information
Additional file 1: Supplementary Table S1. Medication categorization schematic for medications prompting baseline medication recommendations in the INCREASE trial. Supplementary Table S2. Baseline characteristics of all INCREASE trial participants, and those randomized to the MTM intervention. Supplementary Table S3. Full account of baseline recommendations for the MTM intervention arm that were excluded from final recommendation analysis ($$n = 37$$). Supplementary Table S4. Full account of final recommendations designated as not applicable ($$n = 27$$)
## References
1. Barnett M, Frank J, Wehring H. **Analysis of Pharmacist-provided medication therapy management (MTM) services in community pharmacies over 7 years**. *J Manag Care Pharm* (2009.0) **15** 18-31. DOI: 10.18553/jmcp.2009.15.1.18
2. Lin HW, Lin CH, Chang CK. **Economic outcomes of pharmacist-physician medication therapy management for polypharmacy elderly: A prospective, randomized, controlled trial**. *J Formos Med Assoc* (2018.0) **117** 235-243. DOI: 10.1016/j.jfma.2017.04.017
3. Liu M, Liu J, Geng Z, Bai S. **Evaluation of outcomes of medication therapy management (MTM) services for patients with chronic obstructive pulmonary disease (COPD)**. *Pak J Med Sci* (2021.0) **37** 1832-1836. DOI: 10.12669/pjms.37.7.4518
4. Hui RL, Yamada BD, Spence MM, Jeong EW, Chan J. **Impact of a Medicare MTM program: evaluating clinical and economic outcomes**. *Am J Manag Care* (2014.0) **20** e43-51. PMID: 24738554
5. Bluml BM. **Definition of medication therapy management: development of professionwide consensus**. *J Am Pharm Assoc* (2005.0) **45** 566-572. DOI: 10.1331/1544345055001274
6. 6.Agency for Healthcare Research and Quality. Evidence-based Practice Center Systematic Review Protocol Project Title: Medication Therapy Management. Published online July 24, 2013. https://effectivehealthcare.ahrq.gov/sites/default/files/pdf/medication-therapy-management_research-protocol.pdf. Accessed 1 Feb 2022.
7. Pellegrino AN, Martin MT, Tilton JJ, Touchette DR. **Medication Therapy Management Services: Definitions and Outcomes**. *Drugs* (2009.0) **69** 393-406. DOI: 10.2165/00003495-200969040-00001
8. Isetts BJ, Buffington DE, Carter BL, Smith M, Polgreen LA, James PA. **Evaluation of pharmacists’ work in a physician-pharmacist collaborative model for the management of hypertension**. *Pharmacother J Hum Pharmacol Drug Ther* (2016.0) **36** 374-384. DOI: 10.1002/phar.1727
9. St. Peter WL, Farley TM, Carter BL. **Role of collaborative care models including pharmacists in improving blood pressure management in chronic kidney disease patients**. *Curr Opin Nephrol Hypertens* (2011.0) **20** 498-503. DOI: 10.1097/MNH.0b013e32834902c8
10. Hwang AY, Gums TH, Gums JG. **The benefits of physician-pharmacist collaboration**. *J Fam Pract* (2017.0) **66** E1-E8. PMID: 29202145
11. Moga DC, Beech BF, Abner EL. **INtervention for Cognitive Reserve Enhancement in delaying the onset of Alzheimer’s Symptomatic Expression (INCREASE), a randomized controlled trial: rationale, study design, and protocol**. *Trials* (2019.0) **20** 806. DOI: 10.1186/s13063-019-3993-0
12. Moga DC, Abner EL, Schmitt FA. **Intervention for Cognitive Reserve Enhancement in Delaying the Onset of Alzheimer’s Symptomatic Expression (INCREASE) Study: Results from a Randomized Controlled Study of Medication Therapy Management Targeting a Delay in Prodromal Dementia Symptom Progression**. *J Prev Alzheimers Dis* (2022.0) **9** 646-654. PMID: 36281668
13. 13.American Geriatrics Society 2015 Beers Criteria Update Expert Panel. American Geriatrics Society 2015 Updated Beers Criteria for Potentially Inappropriate Medication Use in Older Adults. J Am Geriatr Soc. 2015;63(11):2227–2246. 10.1111/jgs.13702
14. 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
15. Harris PA, Taylor R, Minor BL. **The REDCap consortium: Building an international community of software platform partners**. *J Biomed Inform* (2019.0) **95** 103208. DOI: 10.1016/j.jbi.2019.103208
16. Hanlon JT, Schmader KE. **The Medication Appropriateness Index at 20: Where It Started, Where It Has Been, and Where It May Be Going**. *Drugs Aging* (2013.0) **30** 893-900. DOI: 10.1007/s40266-013-0118-4
17. **Standards of Medical Care in Diabetes-2019 Abridged for Primary Care Providers**. *Clin Diabetes* (2019.0) **37** 11-34. DOI: 10.2337/cd18-0105
18. Casey DE, Thomas RJ, Bhalla V. **2019 AHA/ACC Clinical Performance and Quality Measures for Adults With High Blood Pressure: A Report of the American College of Cardiology/American Heart Association Task Force on Performance Measures**. *Circ Cardiovasc Qual Outcomes* (2019.0) **12** e000057. DOI: 10.1161/HCQ.0000000000000057
19. Brown LM, Rashrash ME, Schommer JC. **The certainty in consumers’ willingness to accept pharmacist-provided medication therapy management services**. *J Am Pharm Assoc* (2017.0) **57** 211-216. DOI: 10.1016/j.japh.2017.02.012
20. Schumacher L, Coe AB, Lester CA, Rothermal M, Dorsch MP. **Factors that influence patient acceptance of clinical pharmacy services: A nationwide survey**. *J Am Pharm Assoc* (2020.0) **60** 1058-1067.e4. DOI: 10.1016/j.japh.2020.08.042
21. Carter SR, Moles RJ, White L, Chen TF. **Consumers’ willingness to use a medication management service: The effect of medication-related worry and the social influence of the general practitioner**. *Res Social Adm Pharm* (2013.0) **9** 431-445. DOI: 10.1016/j.sapharm.2012.07.001
22. Arnoldussen DL, Keijsers K, Drinkwaard J, Knol W, van Marum RJ. **Older Patients’ Perceptions of Medicines and Willingness to Deprescribe**. *sr care pharm* (2021.0) **36** 444-454. DOI: 10.4140/TCP.n.2021.444
23. Chock YL, Wee YL, Gan SL, Teoh KW, Ng KY, Lee SWH. **How Willing Are Patients or Their Caregivers to Deprescribe: a Systematic Review and Meta-analysis**. *J GEN INTERN MED* (2021.0) **36** 3830-3840. DOI: 10.1007/s11606-021-06965-5
24. **American geriatrics society 2019 updated AGS beers criteria® for Potentially inappropriate medication use in older adults**. *J Am Geriatr Soc* (2019.0) **67** 674-694. DOI: 10.1111/jgs.15767
|
---
title: Influence of health literacy on health outcomes of different social strata——
an empirical study based on the data of China's health literacy investigation
authors:
- Huifang Yu
- Siwen Sun
- Jie Ling
- Haixiao Chen
- Guilin Liu
journal: International Journal for Equity in Health
year: 2023
pmcid: PMC9999621
doi: 10.1186/s12939-023-01858-x
license: CC BY 4.0
---
# Influence of health literacy on health outcomes of different social strata—— an empirical study based on the data of China's health literacy investigation
## Abstract
### Background
Health literacy has always been considered as an important factor to promote people's health, but does it have a significant effect on health across all social strata and especially lower social strata? This study aims to analyze the influences of health literacy on health outcomes of different social strata, and then infer whether improving health literacy can reduce health disparities among different social strata.
### Methods
Utilizing health literacy monitoring data from a city in Zhejiang Province in 2020, the samples are divided into three social strata according to the socioeconomic status score: low, middle and high social stratum, to compare whether there are significant differences in health outcomes between population with lower and higher health literacy among different social strata. In the strata with significant differences, control the confounding factors to further verify the influence of health literacy on health outcomes.
### Results
In low and middle social strata, there are significant differences between population with lower and higher health literacy, when considering the two types of health outcomes (chronic diseases and self-rated health), but in high social stratum, this difference is not significant. After controlling the relevant variables, the influence of health literacy on the prevalence of chronic diseases is statistically significant only in low social stratum, and the health literacy is negatively correlated with the prevalence of chronic diseases(OR = 0.722, $$P \leq 0.022$$). In addition, there are statistical significances for positive impact of health literacy on self-rated health in both low and middle social strata (OR = 1.285, $$P \leq 0.047$$; OR = 1.401, $$P \leq 0.023$$).
### Conclusion
Compared with high social stratum, the influence of health literacy on health outcomes of low social stratum (chronic diseases) or both middle and low social strata (self-rated health) is more significant, and both are to improve the health outcomes. This finding suggests that improving residents' health literacy may be an effective way to alleviate the health disparities among different social strata.
## Introduction
Since the Black Report was published in 1980 [1], the health disparities among different social strata have gradually attracted the attention of scholars from all countries. Most studies have found that the health status of low social stratum is often worse [2, 3]. With the development of medical technology and the continuous promotion of public health measures, the life expectancy of people in all countries around the world has been increasing, but the health disparities among different social strata still exist and even tend to get worse. Ten years after the Black Report was published, Smith and other scholars conducted another survey in British society and found that the health disparities among different British social strata were still expanding [4], and similar findings were found in the studies conducted by Tetzlaff and Fors [5, 6].
In order to alleviate this phenomenon of health inequality, researchers began to look for the reasons why health disparities exist among different social strata. In the field of health and medicine, some researchers believe that one of the important reasons for the emergence and continuous expansion of such health disparities is the uneven distribution of benefits brought by the progress of medical technology and various health promotion policies and measures in the whole society [7, 8]. For example, Pavalko believed that the advantages and resources possessed by people with higher socioeconomic status will make it easier for them to access and utilize new health promotion mechanisms, which resulted in population with high socioeconomic status would benefit more, while the poorest and the least educated population will benefit least. In order to reduce the uneven distribution of benefits among different social strata, governments and academia in all countries have begun to taken measures to improve the health status of low social stratum. Most of the measures are committed to providing a healthy supportive environment for population with low socioeconomic status, so that they have "the opportunity" to make healthy choices, such as establishing medical insurance systems, basic public health service systems, medical resources sinking and other measures [9, 10]. However, besides a few mandatory measures, most of the health services need residents to actively participate and utilize, especially the cultivation of a healthy lifestyle needs long-term self-consciousness. Compared with the population with high socioeconomic status, the population with low socioeconomic status often lack the ability to acquire, distinguish and utilize health information and health services [11–13], which is just the manifestation of lack of health literacy. The existing researches also indicate that the level of health literacy1 of the population with low socioeconomic status is generally low [14].
Health literacy is defined as " The degree to which individuals have the capacity to obtain, process, and understand basic health information and services needed to make appropriate health decisions "[15]. The basic differences between improving health literacy and other public health measures lie in that it is an internalized process, and its purpose is to enable people to make healthy choices sincerely, voluntarily and willingly. It can be believed that improving health literacy is the internal driving force for other public health measures to play their role. Therefore, analyzing influence of health literacy on health outcomes plays an irreplaceable role in finding ways to reduce the health disparities among different strata.
The current researches on the relationship between health literacy and health outcomes can be roughly divided into the following aspects according to different health outcomes: [1] The influence on disease prevalence and prognosis. For example, the population with low health literacy have a higher prevalence of chronic diseases [16], and perform worse in disease control and complication prevalence [17]. [ 2] Influence on mortality. Lower health literacy is associated with higher mortality [18]. [ 3] Influence on self-rated health status. The self-rated health status of the population with low health literacy is worse [19, 20]. Most of these studies are conducted in the whole population or divided into different subgroups according to gender, age and other characteristics, respectively studying the influence of health literacy on health outcomes in each subgroup. However, few studies have considered social strata of samples according to their socioeconomic status to understand the relationship between health literacy and health outcomes in different social strata. Although health literacy plays a positive role in promoting health for most health outcomes in the study of the whole population, it is uncertain whether health literacy can also play a positive role in different social strata, especially in low social stratum. If we want to alleviate the health disparities among different social strata by improving health literacy, we must first understand the influence of health literacy on health outcomes in all social strata, and then further infer whether this method to improving health literacy can play desired role in reducing health disparities among different social strata.
In this study, three classical measurement indicators, education level, income level and professional status [21], are used to measure socioeconomic status, the scores of health literacy questionnaire are used to measure whether the samples have health literacy, and the prevalence of chronic diseases and self-rated health status are used as indicators to measure health outcomes. According to the score of socioeconomic status, the samples are divided into three social strata: low, middle and high social stratum, and the influence of health literacy on health outcomes in different social strata is analyzed, so as to provide scientific evidences to find effective ways to reduce the health disparities among different social strata.
## Respondent
This survey is part of a survey of residents' health literacy in Zhejiang Province conducted by the Zhejiang Center for Disease Control and Prevention. The respondents were found in seven counties of a city in Zhejiang Province. Residents aged 15–69 who had lived in the local area for more than 6 months totally from July 2019 to June 2020 are selected as the respondents, but do not include residents who collectively lived in hospitals, dormitories, nursing homes, etc.
## Sampling method
The samples are selected by stratified multistage random sampling. In the first stage, four townships are randomly selected from each county, and a total of 28 townships surveyed places are selected. In the second stage, two communities are randomly selected from each township. In the third stage, 100 households are selected from each community, and one resident aged 15–69 is selected from each household as the respondent. It is enough to complete 85 questionnaires in each community, and a total of 4,760 samples are obtained (Fig. 1). In this study,we excluded respondents aged 15 to 17. Because we need to know the professions of the respondents to measure people's socioeconomic status, yet most respondents did not work before the age of 18. After initial screening according to age, there are 4,693 respondents aged between 18 and 69.Fig. 1Sampling flow chart
## Survey method
In this survey, the method of questionnaire survey and household survey are both adopted. The questionnaire is completed by the respondents. If the respondents can not complete the questionnaire independently, it will be surveyed by face-to-face inquiry. Before the survey, investigators were trained standardly to ensure the consistency of survey method used. During the survey, the on-site coordinators will supervise and verify whether the investigators comply with the survey technical specifications. After the survey, the disease control department will conduct quality control by checking the answer time in the system background, extracting sound recordings and on-site review, exclude the unqualified questionnaires, and select new respondents again, so as to obtain all the qualified data finally.
The questionnaire includes three parts: General information survey, Health literacy survey and *Health status* survey. *The* general information survey mainly collects the age, gender, profession, education level, income level and other information of the respondents. Health literacy was assessed by the Chinese Citizen Health Literacy Questionnaire, which was designed by Delphi method [22]. Experts in the fields of public health, health education and promotion, and clinical medicine jointly designed this questionnaire. And the respondents of this questionnaire are permanent urban and rural residents aged 15–69 in China. The overall Cronbach’s alpha of the questionnaire was 0.95 and Spearman-Brown coefficient was 0.94 [23]. This questionnaire is not only used in the annual China Health Literacy Survey (CHLS) [24], but also in many studies on health literacy in China [25–27]. The health status survey part is used to investigate the health outcomes of recent chronic diseases and self-rated health status.
## Statistical method
SPSS 22.0 is used for statistical analysis. Because the variables in this study are categorical variables, the categorical variables are expressed as constituent ratio (%) in statistical description, and the chi-square test is adopted for the inter-group difference test. Logistic regression model is used to further determined the influence of health literacy on health outcomes, and the significance level is set at α = 0.05.
## Assignment standard
Use profession status, education level and income level to measure the socioeconomic status of the respondents. As for which of the three variables of education, income and occupation is more important, the opinions of various researchers are not consistent [28, 29], so this paper still adds these three variables with equal weight [30]. The health literacy questionnaire included three types of questions: true/false (correct response received 1 points), single-answer (correct response received 1 points), and multiple-answer (correct responses received 2 points). The total score of the health literacy questionnaire is 66, and those who reach $80\%$ or more of the total score are judged to have basic health literacy [31]. Chronic disease and self-rated health are selected as indicators to measure health status. See Table 1 for specific assignment standards of each variable. Table 1Variable definition and assignmentVariableDefinition and assignmentEducation level1 = Illiterate/Primary school, 2 = Junior high school, 3 = Senior high school/Vocational high school/Technical secondary school, 4 = Junior college/University, 5 = Postgraduate and higherIncome levelAnnual per capita household income = Total annual household income/Household size. 1 = Less than 10,000 yuan; 2 = 10,000–29,999 yuan; 3 = 30,000–49,999 yuan; 4 = 50,000–69,999 yuan; 5 = 70,000 yuan and higherProfessional status1 = The unemployed/Retiree; 2 = Farmer/Worker; 3 = Enterprise employee/Personnel of other public institutions/Businessman/College student; 4 = Teacher/Medical staff; 5 = Civil servantSocioeconomic statusThe individual's comprehensive socioeconomic status is measured by adding the scores of education level, income level and professional status. The higher the score is, the higher the status is. The actual lowest score in all samples is 4 and the highest score is 14. 4–7 of socioeconomic status score = Population with low socioeconomic status, 8–10 = Population with middle socioeconomic status, 11–14 = Population with high socioeconomic statusChronic disease1 = Suffering from any one or more chronic diseases, such as hypertension, diabetes, cerebrovascular disease, etc.; Otherwise = 0Self-rated health1 = Self-rated health is "good" or "better"; Otherwise = 0
## General information of respondents
Four thousand six hundred ninety-three samples are screened by logical test and outlier cleaning, and 4011 valid questionnaires are obtained, with an effective rate of $85.47\%$. Descriptive statistical analysis is made on 4011 valid samples after screening, and the general information is as follows: 1,981 males, accounting for $49.4\%$, and 2,030 females, accounting for $50.6\%$; The age distribution is dominated by middle-aged people aged 40–59, accounting for $47.8\%$, young people aged 18–39, accounting for $30.0\%$, and elderly people aged 60–69, accounting for $22.1\%$; The marital status is mainly married, accounting for $84.0\%$; Census register is dominated by local census register, accounting for $88\%$; After the sample is stratified according to the socioeconomic status score, the low, middle and high levels account for $61.6\%$, $28.5\%$ and $9.9\%$ respectively; The population with higher health literacy accounts for $31.2\%$ of the total sample. See Table 2 for details. Table 2General information of respondents ($$n = 4011$$)VariableFrequencyPercentage%Gender Male198149.4 Female203050.6Age 18 ~ 39 years120430.0 40 ~ 59 years191947.8 60 ~ 69 years88822.1Rural/Urban Rural241660.2 Urban159539.8Marital Status Single40610.1 Married336984.0 Separated210.5 Divorced1203.0 Widowed952.4Census register Local353088.0 Other places48112.0Socioeconomic status Low247161.6 Middle114228.5 High3989.9Health literacy Lower276168.8 Higher125031.2
## Differences in health outcomes and health literacy among different social strata
The statistical results in Table 3 show that there are significant differences between the two types of health outcomes and health literacy among different social strata. The prevalence of chronic diseases is $34.4\%$, $16.3\%$ and $7.5\%$ in low, middle and high social stratum respectively. The proportion of self-rated health as good and better in low social stratum is $65.5\%$, while that in middle and high social strata is $75.5\%$ and $80.7\%$ respectively. There is a significant stratum gradient in the prevalence of chronic diseases and self-rated health status of different social strata. There are also stratum differences in health literacy. The health literacy level of low social stratum is $15.3\%$, which is significantly lower than that in middle and high social strata ($49.1\%$ and $78.1\%$).Table 3Differences in health outcomes and health literacy among different social strataSocioeconomic statusHealth outcomesHealth literacy levelChronic diseasesGood self-rated health status%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ$2\%$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ$2\%$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ2Low34.4211.698**65.561.646**15.3871.128**Middle16.375.549.1High7.580.778.1**$p \leq 0.01$
## Differences in health outcomes between population with lower and higher health literacy in stratified samples
Based on the significant differences in health literacy and health outcomes among different social strata, it is speculated that the influence of health literacy on health outcomes may be different among different social strata. To test this hypothesis, in this study, whether there are differences between the two types of health outcomes in population with lower and higher health literacy in different social strata are compared at first (Table 4). The results show that there are significant differences in the prevalence of chronic diseases and self-rated health status between population with lower and higher health literacy in low and middle social strata. In high social stratum, although the prevalence of chronic diseases in population with higher health literacy is slightly lower than that in population with lower health literacy and the self-rated health status is better, the difference is not statistically significant. Therefore, it can't be considered that there is a difference in the two types of health outcomes between population with lower and higher health literacy in high social stratum. Table 4Differences in health outcomes between population with lower and higher health literacy in stratified samplesSocioeconomic statusHealth literacyChronic diseasesGood self-rated health status%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ$2\%$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ2LowLower36.628.238**64.38.285**Higher22.572MiddleLower22.431.098**71.311.409**Higher10.279.9HighLower9.20.439770.946Higher7.181.7**$p \leq 0.01$
## Influences of health literacy on health outcomes in stratified samples
In order to further verify the relationship between health literacy and health outcomes found in low and middle social strata, multivariate logistic regression analyses of health literacy and chronic diseases or self-rated health in low and middle social strata are conducted respectively. Before regression analysis, difference test is performed to determine confounders that might affect health outcomes (Table 5).Table 5Difference test for health outcomes in populations with different characteristicsVariableLow socioeconomic statusMiddle socioeconomic statusChronic diseasesGood or better self-rated health statusChronic diseasesGood or better self-rated health status%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ$2\%$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ$2\%$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ$2\%$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ2Gender Male38.113.572**67.23.00920.215.096**76.91.384 Female3163.911.873.9Age 18 ~ 39 years4.0320.986**79.132.356**3.3226.308**80.120.011** 40 ~ 59 years28.365.225.672.1 60 ~ 69 years57.160.757.661.6Rural/Urban Rural33.80.91364.61.49915.21.57174.31.14 Urban35.767.018.077.1Marital Status Single24.49.28962.21.4193.056.571**85.220.037** Married34.565.719.573.4 Separated40.070.014.385.7 Divorced24.265.715.961.4 Widowed4560.061.576.9Census register Local38.1106.331**63.529.392**17.913.999**74.91.689 Other places8.079.45.280.0**$p \leq 0.01$ The difference test results show that, in low social stratum, there are significant differences in the prevalence of chronic diseases among different genders, ages and census registers, and in self-rated health status in different ages and census registers. In middle social stratum, there are significant differences in the prevalence of chronic diseases among different genders, ages, marital statuses and census registers, and in self-rated health status among different ages and marital statuses. These variables, which may affect the prevalence of chronic diseases and self-rated health status, are included as control variables in the regression model with health literacy as independent variable and chronic diseases or self-rated health are included as dependent variable (Table 6).Table 6Analysis of the relationship between health literacy and health outcomes in different social strataVariableOR (Chronic disease)OR (Self-rated health)Socioeconomic statusLowMiddleLowMiddleConstant0.082**0.044**2.732**4.759**Health literacy(ref. Lower) Higher0.722*0.7901.285*1.401*Gender (ref. Male) Female0.760**0.583**Age (ref. 18 ~ 39 years) 40 ~ 59 years6.386**7.855**0.626**0.850 60 ~ 69 years19.238**26.967**0.550**0.521*Marital Status(ref. Single) Married1.4220.581* Live apart0.8471.545 Divorced1.2040.327** Widowed2.9511.102Census register(ref. Local) Other places0.305**0.433*1.760** Likelihood ratio testP < 0.001P < 0.001P < 0.001P < 0.001 Hosmer–Lemeshow testP = 0.221P = 0.353P = 0.886P = 0.536*$p \leq 0.05$**$p \leq 0.01$ The validity test results of models of health literacy and two kinds of health outcomes are shown in Table 6. All models have passed the Likelihood ratio test ($P \leq 0.05$) and Hosmer–Lemeshow test ($P \leq 0.05$). The fitting results of each regression model are good.
After stratification, it can be observed that the risk of chronic diseases in population with higher health literacy is lower than that in population with lower health literacy. It can be considered that, for low social stratum, having higher health literacy can reduce the risk of chronic diseases(OR = 0.722, $$P \leq 0.022$$). However, in the middle socioeconomic stratum, after controlling other related variables, the influence of health literacy on chronic diseases is no longer statistically significant.
The positive influence of health literacy on self-rated health status is statistically significant in low and middle socioeconomic strata (OR = 1.285, $$P \leq 0.047$$; OR = 1.401, $$P \leq 0.023$$). For low and middle social strata, having higher health literacy is helpful to improve self-rated health status.
## Influence of health literacy on health outcomes of low social stratum (chronic diseases) or low and middle social strata (self-rated health) is more significant than that of high social stratum.
The difference of chronic disease prevalence and self-rated health status between population with lower and higher health literacy is only significant in low and middle social strata. After controlling the related variables, the influence of health literacy on chronic diseases is still statistically significant in population with low socioeconomic status, but such significant influence is not found in population with middle and high socioeconomic status. The influence of health literacy on self-rated health is statistically significant in population with low and middle socioeconomic status, but the correlation between health literacy and self-rated health is not found in population with high socioeconomic status.
Based on these results, it can be preliminarily inferred that the influences of health literacy on the health outcomes in low social stratum (chronic diseases) or low and middle social strata (self-rated health) is more significant than that in high social stratum, which is similar to the results of research conducted by Gibney in Ireland [32]. Gibney found that the influence of health literacy on health outcomes, such as chronic diseases and hospital attendance rate, was significant in low or middle and low social strata, but not in high social stratum. However, he did not explain detailly for this finding.
In this paper, we will attempt to explain this phenomenon from the following perspectives: Some researchers have found that the population that people come into contact with in work and life are mostly people in similar social stratum [33]. Because people have social needs, they are often imperceptibly influenced by the values and behavioral norms of surrounding people [34]. Population with high socioeconomic status have a high level of health literacy ($78.1\%$). Even if population with high social stratum have not health literacy, they will still be influenced and restrained by the surrounding people and environment, which will encourage them to maintain a healthy lifestyle. In addition, most of the population with high socioeconomic status have a good living and working environment, and their chances to be exposed to the risk factors affecting their health are less [35, 36],which also weakens the role of health literacy to some extent. However, in low social stratum, the proportion of people with health literacy is very low ($15.3\%$), and they are more likely to be exposed to health risk factors than those with high socioeconomic status. Therefore, health literacy has a greater influence on the health outcomes of population in low social stratum.
## Having higher health literacy can improve health outcomes (chronic diseases, self-rated health)
Among the significant influences of health literacy on health outcomes found in low and middle social strata, health literacy will all play a role to improve health outcomes. Those with higher health literacy had lower rates of chronic disease and better self-rated health status than those with lower health literacy. It is consistent with previous findings [16, 25].
People with higher health literacy are more willing and able to acquire and understand health knowledge, and utilize it to improve their lifestyle. However, one of the important reasons for chronic diseases and many other health damage is the long-term accumulation of health damage caused by unhealthy lifestyles [37]. In addition, Parikh believed that people with lower health literacy were easy to feel ashamed and embarrassed about their ignorance, which would hinder them from seeking health help including medical care services and acquisition of health knowledge, thus affecting their health status [38]. These findings can explain the improvement of health literacy on health outcomes to some extent.
## There are significant differences in health outcomes among different social strata
There is a significant stratum gradient in the prevalence of chronic diseases and self-rated health status among low, middle and high social strata. The prevalence of chronic diseases in population with low socioeconomic status ($34.4\%$) is significantly higher than that in population with middle and high socioeconomic status ($16.3\%$,$7.5\%$), and their proportion of self-rated health as good or better ($65.5\%$) is significantly lower than that in middle and high social strata ($75.5\%$,$80.7\%$). This is consistent with many study conclusions. For example, Roberto found that in almost all of the 22 European countries he surveyed, socioeconomic status had a significant negative correlation with the mortality rate and self-rated health [2]; Wolff believed that low subjective social status was significantly related to poor/common health status [39]. The same findings are found in the researches conducted in China [40, 41]. Generally speaking, the lower the socioeconomic status is, the worse the health status is. There are many reasons leading to health disparities among different social strata, including poor living and working environment [42], unhealthy lifestyle [37] and unequal access to medical resources [43] of population with low socioeconomic status, and the lack of health drive force caused by the insufficient health literacy discussed in this paper.
Combining conclusion 1 and 2, it can be found that health literacy has a more significant influence on health outcomes of population in low social stratum than that in high social stratum, and health literacy is a protective factor for health outcomes. Therefore, it can be considered that improving residents' health literacy is an effective measure to alleviate the health gap among social strata. While the government is committed to creating a healthy supportive environment for population with low socioeconomic status and improving the fairness of medical resources, it should also pay attention to the improvement of residents' health literacy, so that population of low social stratum not only have the opportunity but also have the ability to make healthy choices.
## Conclusions and shortcomings
In this paper, the influences of health literacy on health outcomes in all population with different socioeconomic status are discussed, according to the monitoring data of health literacy from a city of Zhejiang Province in 2020. The main conclusions are as follows: [1] The influences of health literacy on health outcomes of population in low social stratum (chronic diseases) or low and middle social strata (self-rated health) is more significant than that in high social stratum, which suggests that improving residents' health literacy may be an effective way to alleviate the health gap among different social strata; [2] Health literacy will play an role to improve health outcomes (chronic diseases and self-rated health); [3] There are significant differences in health outcomes among different social strata.
This study also has the following limitations. Firstly, this study is a cross-sectional survey, which can only provide some clues for causal inference between health literacy and health outcomes, but can not verify the causal relationship. Further research is needed to verify the causal relationship. Secondly, the types of health outcomes selected in this study are limited, and it is unknown whether health literacy has the same influence on other health outcomes with social stratum differences. Finally, this paper measures socioeconomic status by simply adding education level, income level and professional status. However, the influences of education level, income level and professional status on socioeconomic status are probably different, so we should further look for a more accurate way to measure socioeconomic status.
## References
1. 1.Black D, Morris J, Smith C, Townsend P. Inequalities in health: a report of a research working group. 1980.
2. De Vogli R, Gimeno D, Kivimaki M. **Socioeconomic inequalities in health in 22 European countries**. *N Engl J Med* (2008.0) **358** 2468-2481. DOI: 10.1056/NEJMsa0707519
3. Nwosu CO, Oyenubi A. **Income-related health inequalities associated with the coronavirus pandemic in South Africa: A decomposition analysis**. *Int J Equity Health* (2021.0) **20** 21. DOI: 10.1186/s12939-020-01361-7
4. Smith GD, Bartley M, Blane D. **The Black report on socioeconomic inequalities in health 10 years on**. *BMJ* (1990.0) **301** 373-377. DOI: 10.1136/bmj.301.6748.373
5. Tetzlaff F, Epping J, Sperlich S, Tetzlaff J. **Widening income inequalities in life expectancy? Analysing time trends based on German health insurance data**. *J Epidemiol Community Health* (2020.0) **74** 592-597. PMID: 32277000
6. Fors S, Wastesson JW, Morin L. **Growing Income-Based Inequalities in Old-Age Life Expectancy in Sweden, 2006–2015**. *Demography* (2021.0) **58** 2117-2138. DOI: 10.1215/00703370-9456514
7. Pavalko EK, Caputo J. **Social Inequality and Health Across the Life Course**. *Am Behav Sci* (2013.0) **57** 1040-1056. DOI: 10.1177/0002764213487344
8. Chen R, Li NX, Liu X. **Study on the equity of medical services utilization for elderly enrolled in different basic social medical insurance systems in an underdeveloped city of Southwest China**. *Int J Equity Health* (2018.0) **17** 54. DOI: 10.1186/s12939-018-0765-5
9. Zhao P, Diao Y, You L, Wu S, Yang L, Liu Y. **The influence of basic public health service project on maternal health services: an interrupted time series study**. *BMC Public Health* (2019.0) **19** 824. DOI: 10.1186/s12889-019-7207-1
10. Tirgil A, Dickens WT, Atun R. **Effects of expanding a non-contributory health insurance scheme on out-of-pocket healthcare spending by the poor in Turkey**. *BMJ Glob Health* (2019.0) **4** e001540. DOI: 10.1136/bmjgh-2019-001540
11. Janković J, Simić S, Marinković J. **Inequalities that hurt: demographic, socio-economic and health status inequalities in the utilization of health services in Serbia**. *Eur J Public Health* (2010.0) **20** 389-396. DOI: 10.1093/eurpub/ckp189
12. Wang MP, Wang X, Lam TH, Viswanath K, Chan SS. **Health information seeking partially mediated the association between socioeconomic status and self-rated health among Hong Kong Chinese**. *PLoS ONE* (2013.0) **8** e82720. DOI: 10.1371/journal.pone.0082720
13. Yaya S, Oladimeji O, Oladimeji KE, Bishwajit G. **Prenatal care and uptake of HIV testing among pregnant women in Gambia: a cross-sectional study**. *BMC Public Health* (2020.0) **20** 485. DOI: 10.1186/s12889-020-08618-4
14. Sørensen K, Pelikan JM, Röthlin F. **Health literacy in Europe: comparative results of the European health literacy survey (HLS-EU)**. *Eur J Public Health* (2015.0) **25** 1053-1058. DOI: 10.1093/eurpub/ckv043
15. Nielsen-Bohlman L, Panzer AM, Kindig DA. **Health Literacy: A Prescription to End Confusion**. *What Is Health Literacy* (2004.0) 31-36
16. McNaughton CD, Kripalani S, Cawthon C, Mion LC, Wallston KA, Roumie CL. **Association of health literacy with elevated blood pressure: a cohort study of hospitalized patients**. *Med Care* (2014.0) **52** 346-353. DOI: 10.1097/MLR.0000000000000101
17. Saeed H, Saleem Z, Naeem R, Shahzadi I, Islam M. **Impact of health literacy on diabetes outcomes: a cross-sectional study from Lahore**. *Pakistan Public Health* (2018.0) **156** 8-14. DOI: 10.1016/j.puhe.2017.12.005
18. Smith SG, Jackson SE, Kobayashi LC, Steptoe A. **Social isolation, health literacy, and mortality risk: findings from the english longitudinal study of ageing**. *Health Psychol* (2018.0) **37** 160-169. DOI: 10.1037/hea0000541
19. Yoko F, Naoki K, Zentaro Y, Hideki H. **Health literacy, socioeconomic status and self-rated health in Japan**. *Health Promot Int* (2013.0) **30** 505-513. PMID: 24131729
20. Storey A, Hanna L, Missen K, Hakman N, Osborne RH, Beauchamp A. **The Association between health literacy and self-rated health amongst Australian University students**. *J Health Commun* (2020.0) **25** 333-343. DOI: 10.1080/10810730.2020.1761913
21. 21.Dutton, D, and S Levine. Overview, methodological critique, and reformulation. Menlo Park, CA: The Henry J. Kaiser Family Foundation; 1989.
22. Li Y, Lv X, Liang J, Dong H, Chen C. **The development and progress of health literacy in China**. *Front Public Health* (2022.0) **10** 1034907. DOI: 10.3389/fpubh.2022.1034907
23. Shen M, Hu M, Liu S, Chang Y, Sun Z. **Assessment of the Chinese resident health literacy scale in a population-based sample in South China**. *BMC Public Health* (2015.0) **15** 637. DOI: 10.1186/s12889-015-1958-0
24. Liu Z, Li YH, Cui ZY. **Prevalence of tobacco dependence and associated factors in China: findings from nationwide china health literacy survey during 2018–19**. *Lancet Reg Health West Pac* (2022.0) **24** 100464. DOI: 10.1016/j.lanwpc.2022.100464
25. Nie X, Li Y, Li C, Wu J, Li L. **The association between health literacy and self-rated health among residents of china aged 15–69 years**. *Am J Prev Med* (2021.0) **60** 569-578. DOI: 10.1016/j.amepre.2020.05.032
26. Chen C, Xu T, Chen Y. **Does health literacy promote COVID-19 awareness? Evidence from Zhejiang**. *China Front Public Health* (2022.0) **10** 894050. DOI: 10.3389/fpubh.2022.894050
27. 27.He T, Liu L, Huang J, Li G, Guo X. Health knowledge about smoking, role of doctors, and self-perceived health: a cross-sectional study on smokers’ intentions to quit. Int J Environ Res Public Health. 2021;18(7):3629.
28. Stringhini S, Dugravot A, Shipley M. **Health behaviours, socioeconomic status, and mortality: further analyses of the British Whitehall II and the French GAZEL prospective cohorts**. *PLoS Med* (2011.0) **8** e1000419. DOI: 10.1371/journal.pmed.1000419
29. Winkleby MA, Jatulis DE, Frank E, Fortmann SP. **Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease**. *Am J Public Health* (1992.0) **82** 816-820. DOI: 10.2105/AJPH.82.6.816
30. Stalling I, Albrecht BM, Foettinger L, Recke C, Bammann K. **Associations between socioeconomic status and physical activity among older adults: cross-sectional results from the OUTDOOR ACTIVE study**. *BMC Geriatr* (2022.0) **22** 396. DOI: 10.1186/s12877-022-03075-7
31. Nie XQ, Li YH, Li L. **Statistical analysis of 2012 Chinese residents’ health literacy monitoring**. *Chin J Health Educ* (2014.0) **30** 178-181
32. Gibney S, Bruton L, Ryan C, Doyle G, Rowlands G. **Increasing health literacy may reduce health inequalities: evidence from a national population survey in Ireland**. *Int J Environ Res Public Health* (2020.0) **17** 5891-5908. DOI: 10.3390/ijerph17165891
33. Bhattacharya H, Dugar S. **Partnership Formation: the role of social status**. *Manage Sci* (2014.0) **60** 1130-1147. DOI: 10.1287/mnsc.2013.1818
34. Kaisari P, Higgs S. **Social modelling of food intake The role of familiarity of the dining partners and food type**. *Appetite* (2015.0) **86** 19-24. DOI: 10.1016/j.appet.2014.09.020
35. Yin P, Zhang M, Li Y, Jiang Y, Zhao W. **Prevalence of COPD and its association with socioeconomic status in China: findings from China chronic disease risk factor surveillance 2007**. *BMC Public Health* (2011.0) **11** 586. DOI: 10.1186/1471-2458-11-586
36. Marshall IJ, Wang Y, Crichton S, McKevitt C, Rudd AG, Wolfe CD. **The effects of socioeconomic status on stroke risk and outcomes**. *Lancet Neurol* (2015.0) **14** 1206-1218. DOI: 10.1016/S1474-4422(15)00200-8
37. Barbaresko J, Rienks J, Nöthlings U. **Lifestyle indices and cardiovascular disease risk: a meta-analysis**. *Am J Prev Med* (2018.0) **55** 555-564. DOI: 10.1016/j.amepre.2018.04.046
38. Parikh NS, Parker RM, Nurss JR, Baker DW, Williams MV. **Shame and health literacy: the unspoken connection**. *Patient Educ Couns* (1996.0) **27** 33-39. DOI: 10.1016/0738-3991(95)00787-3
39. Wolff LS, Subramanian SV, Acevedo-Garcia D, Weber D, Kawachi I. **Compared to whom? Subjective social status, self-rated health, and referent group sensitivity in a diverse US sample**. *Soc Sci Med* (2010.0) **70** 2019-2028. DOI: 10.1016/j.socscimed.2010.02.033
40. 40.Yuan L, Liang H. Research on the influence of socio-economic status on residents’ health level. Chin J Health Educ. 2020;36(10):912–5.
41. Tang SN, Jian WY. **Effect of socioeconomic status on prevalence of hypertension in Chinese adults**. *Chin J Prev Control of Chronic Diseases* (2011.0) **19** 238-241
42. Evans GW, Kim P. **Multiple risk exposure as a potential explanatory mechanism for the socioeconomic status-health gradient**. *Ann N Y Acad Sci* (2010.0) **1186** 174-189. DOI: 10.1111/j.1749-6632.2009.05336.x
43. Nwosu CO, Ataguba JE. **Socioeconomic inequalities in maternal health service utilisation: a case of antenatal care in Nigeria using a decomposition approach**. *BMC Public Health* (2019.0) **19** 1493. DOI: 10.1186/s12889-019-7840-8
|
---
title: Different associations of general and abdominal obesity with upper and lower
extremity artery disease among a community population in China
authors:
- Yong Wang
- Xiaoyan Guo
- Yi Zhang
- Ruiyan Zhang
- Jue Li
journal: Nutrition & Metabolism
year: 2023
pmcid: PMC9999629
doi: 10.1186/s12986-023-00736-1
license: CC BY 4.0
---
# Different associations of general and abdominal obesity with upper and lower extremity artery disease among a community population in China
## Abstract
### Background
The associations between obesity and abnormalities of upper and lower extremity arteries remain to be elucidated. This study is aimed to investigate whether general obesity and abdominal obesity are associated with upper and lower extremity artery diseases in a Chinese community population.
### Methods
This cross-sectional study included 13,144 participants in a Chinese community population. The associations between obesity parameters and abnormalities of upper and lower extremity arteries were evaluated. Multiple logistic regression analysis was used to assess the independence of associations between obesity indicators and abnormalities of peripheral arteries. Nonlinear relationship between body mass index (BMI) and risk of ankle-brachial index (ABI) ≤ 0.9 was evaluated using a restricted cubic spline model.
### Results
The prevalence of ABI ≤ 0.9 and interarm blood pressure difference (IABPD) ≥ 15 mmHg in the subjects was $1.9\%$ and $1.4\%$ respectively. Waist circumference (WC) was independently associated with ABI ≤ 0.9 (OR 1.014, $95\%$ CI 1.002–1.026, $$P \leq 0.017$$). Nevertheless, BMI was not independently associated with ABI ≤ 0.9 using linear statistical models. Meanwhile, BMI and WC were independently associated with IABPD ≥ 15 mmHg respectively (OR 1.139, $95\%$ CI 1.100–1.181, $P \leq 0.001$, and OR 1.058, $95\%$ CI 1.044–1.072, $P \leq 0.001$). Furthermore, prevalence of ABI ≤ 0.9 was displayed with a U-shaped pattern according to different BMI (< 20, 20 to < 25, 25 to < 30, and ≥ 30). Compared with BMI 20 to < 25, risk of ABI ≤ 0.9 was significantly increased when BMI < 20 or ≥ 30 respectively (OR 2.595, $95\%$ CI 1.745–3.858, $P \leq 0.001$, or OR 1.618, $95\%$ CI 1.087–2.410, $$P \leq 0.018$$). Restricted cubic spline analysis indicated a significant U-shaped relationship between BMI and risk of ABI ≤ 0.9 (P for non-linearity < 0.001). However, prevalence of IABPD ≥ 15 mmHg was significantly increased with incremental BMI (P for trend < 0.001). Compared with BMI 20 to < 25, the risk of IABPD ≥ 15 mmHg was significantly increased when BMI ≥ 30 (OR 3.218, $95\%$ CI 2.133–4.855, $P \leq 0.001$).
### Conclusions
Abdominal obesity is an independent risk factor for upper and lower extremity artery diseases. Meanwhile, general obesity is also independently associated with upper extremity artery disease. However, the association between general obesity and lower extremity artery disease is displayed with a U-shaped pattern.
## Introduction
Atherosclerotic cardiovascular diseases (ASCVDs) which may involve coronary artery disease (CAD), atherosclerotic cerebral infarction, peripheral arterial disease (PAD), and atherosclerotic changes in other arteries, are the main causes of mortality worldwide [1–4]. PAD may include arterial disease of lower extremities, upper extremities, renal artery, carotid artery, or other peripheral arteries, and is one of the manifestation of systemic atherosclerosis [5]. PAD is an important component of the ASCVD, but is often underestimated by cardiologists. In fact, PAD was associated with higher risk of all-cause and cardiovascular disease (CVD) mortality in Chinese patients with high cardiovascular risk in our previous studies [6, 7]. Ankle-brachial index (ABI) ≤ 0.90 can be considered as a criterion for the diagnosis of lower extremity PAD [8]. At the same time, increased interarm systolic blood pressure difference (IABPD) often signifies the potential abnormalities of upper extremity arteries mainly including subclavian artery, brachiocephalic trunk, and axillary artery [9, 10]. Previous studies revealed that lower ABI and higher IABPD were associated with increased mortalities respectively in Chinese [6, 11].
Obesity is associated with a much higher prevalence of comorbidities such as diabetes, hypertension, and metabolic syndrome, which then increase the risk of ASCVD. In addition, obesity may also be an independent risk factor for the development of ASCVD [12]. With the improvement of living standards and change of lifestyle, the prevalence of obesity has been significantly elevated in China in recent years. Thus, more attention should be paid to obesity related metabolic and cardiovascular disorders in China. Obesity can be classified as general obesity and abdominal obesity. However, the associations between various kinds of obesity and abnormalities of upper and lower extremity arteries remain to be elucidated to date. It is worth noting that the data on the relationship between body mass index (BMI) and abnormalities of lower extremity arteries are controversial. A previous study found that the risk of lower extremity PAD was increased with incremental BMI [13]. But another study indicated that BMI did not increase the risk of developing lower extremity PAD [14]. Meanwhile, the association between obesity and abnormalities of upper extremity arteries was rarely investigated in previous studies. Thus, this study is aimed to investigate whether general obesity and abdominal obesity are associated with the prevalence of upper and lower extremity artery disease in a community population in China.
## Study subjects
The study subjects ($$n = 13$$,750) were enrolled through cluster multistage and random sampling to community population from several districts of Shanghai in China in this cross-sectional study. The participants aged more than 18 years old were investigated in each center from May to September in 2016. Exclusion criteria included history of aortic dissection, history of amputation surgery, atrial fibrillation, mental disorder or lack of compliance. After the subjects with incomplete data or exclusion criteria were removed, there were totally 13,144 participants left (Fig. 1).Fig. 1Flow chart of subjects enrollment The study complied with the Declaration of Helsinki. It was also approved by the ethics committee of Shanghai Jiao Tong University and informed consent was obtained from all the participants prior to enrollment.
## Four-limb blood pressure and ABI measurement
Four-limb blood pressure and ABI measurement was performed by trained technicians using a non-invasive vascular profiling system (Omron VP-1000 vascular profiling system, Japan) [3]. This system ensured accurate and reliable ABI measurement using advanced oscillometric technology. Simultaneous blood pressure measurement at all four limbs was included, using a dual chamber cuff system and a proprietary algorithm. Measurement was performed after a 10-min rest in the supine position with the upper body as flat as possible. The device simultaneously and automatically measured the blood pressures twice, and then we calculated the means to get final blood pressure values. Bilateral ankle and brachial artery pressures, and bilateral ABI were supplied after measurement. ACC/AHA guidelines recommend ABI ≤ 0.90 as the criterion for the diagnosis of lower extremity PAD [8]. Meanwhile, IABPD ≥ 15 mmHg was considered as the potential abnormalities of upper extremity arteries according to literatures in this study [9, 10].
## Clinical data collection
A case report form was developed to record general characteristics, clinical diagnosis, and biochemical examination. Waist circumference (WC) was measured at the middle point between the costal margin and iliac crest. BMI was calculated as body weight in kilograms divided by body height in meters squared (kg/m2). Smoking habit was categorized as current smoking, ever smoking, or no smoking. Current smoking was determined when subjects were smoking currently and more than one cigarette daily in at least one year continuously. Ever smoking was determined when subjects smoked more than one cigarette daily, but had quitted smoking at least one year before. Drinking habit was categorized as current drinking, ever drinking, or no drinking. Current drinking was determined when subjects were drinking liquor, beer or wine currently in at least one year. Ever drinking was determined when subjects drank previously, but had quitted drinking at least one year before. History of lipid disorders included that plasma total cholesterol was ≥ 5.7 mmol/l, or low-density lipoprotein cholesterol (LDL-C) was ≥ 3.6 mmol/l, or high-density lipoprotein cholesterol (HDL-C) < 1.04 mmol/l, triglyceride was ≥ 1.7 mmol/l, or treatment with antihyperlipidemic agents due to hyperlipidemia. Hypertension was diagnosed by systolic blood pressure (SBP) ≥ 140 mmHg, or diastolic blood pressure (DBP) ≥ 90 mmHg, or being actively treated with anti-hypertension drugs. Diabetes mellitus was diagnosed by a fasting plasma glucose ≥ 7.0 mmol/l, or by a random plasma glucose ≥ 11.1 mmol/l, or when they were actively receiving therapy using insulin or oral medications for diabetes. Chronic kidney disease was defined as an estimated glomerular filtration rate (eGFR) < 60 ml/min/1.73 m2.
## Statistical analysis
Data entry and management were performed using Epidata software, version 3.1 (Epidata Association, Odense, Denmark). All statistical analyses were conducted with SPSS 22.0 (IBM, Armonk, NY, USA) and R language software (version 4.1.1). Continuous variables were expressed as the mean ± standard deviation, and categorical variables as frequencies (percentages). The chi-square test was used to compare categorical variables. The linear tendency was evaluated among several groups using trend test. The independent-sample t-test and one-way analysis of variance (ANOVA) were used to compare continuous variables among two or more groups. Multiple logistic regression analysis was used to assess the independence of the associations between obesity indicators and various abnormalities of peripheral arteries, and the odds ratio (OR) and $95\%$ confidence interval ($95\%$ CI) was calculated. We also explored the nonlinear relationship between BMI and the risk of ABI ≤ 0.9 using a restricted cubic spline model by multivariable adjustment with three knots (at the 10th, 50th, and 90th percentiles). $P \leq 0.05$, which is two-sided, was considered significant.
## Study participants characteristics
General characteristics of the 13,144 participants by gender were shown in Table 1. The mean age was 52.2 ± 13.1 years old. 7181 subjects of them ($54.6\%$) were man. The average BMI of all participants was 25.2 ± 3.81 kg/m2, and the average WC was 88.5 ± 11.7 cm respectively. The average ABI was 1.08 ± 0.09, and the average IABPD was 3.55 ± 3.79 mmHg respectively. Furthermore, the prevalence of ABI ≤ 0.9 and IABPD ≥ 15 mmHg in this study population was $1.9\%$ and $1.4\%$ respectively. Table 1Clinical characteristics of study participants according to genderVariablesAll ($$n = 13$$,144)Man ($$n = 7181$$)Woman ($$n = 5963$$)P valueAge (years)52.2 ± 13.151.7 ± 13.352.8 ± 13.0 < 0.001BMI (kg/m2)25.2 ± 3.8125.5 ± 3.5724.8 ± 4.05 < 0.001WC (cm)88.5 ± 11.791.0 ± 10.685.5 ± 12.2 < 0.001Smoking––– < 0.001 Current smoking (n, %)3667 ($27.9\%$)3529 ($49.1\%$)138 ($2.3\%$)– Ever smoking (n, %)654 ($40.1\%$)619 ($8.6\%$)35 ($0.6\%$)– No smoking (n, %)8823 ($40.1\%$)3033 ($42.2\%$)5790 ($97.1\%$)–Drinking––– < 0.001 Current drinking (n, %)2612 ($19.9\%$)2496 ($34.8\%$)116 ($1.9\%$)– Ever drinking (n, %)485 ($3.7\%$)463 ($6.4\%$)22 ($0.4\%$)– No drinking (n, %)10,047 ($76.4\%$)4222 ($58.8\%$)5825 ($97.7\%$)–*Diabetes mellitus* (n, %)1399 ($10.6\%$)805 ($11.2\%$)594 ($10.0\%$)0.021Hypertension (n, %)5720 ($43.5\%$)3040 ($42.3\%$)2680 ($44.9\%$)0.003Lipid disorders (n, %)6212 ($47.3\%$)4133 ($57.6\%$)2079 ($34.9\%$) < 0.001Chronic kidney disease (n, %)335 ($2.5\%$)143 ($2.0\%$)192 ($3.2\%$) < 0.001Total cholesterol (mmol/l)4.83 ± 1.044.82 ± 1.044.85 ± 1.040.289Total triglyceride (mmol/l)2.01 ± 1.822.22 ± 2.041.63 ± 1.25 < 0.001LDL-C (mmol/l)2.74 ± 0.902.74 ± 0.922.74 ± 0.860.737HDL-C (mmol/l)1.20 ± 0.321.13 ± 0.301.33 ± 0.31 < 0.001Fasting plasma glucose (mmol/l)5.36 ± 1.715.39 ± 1.745.31 ± 1.650.046Serum creatinine (umol/l)81.0 ± 38.086.8 ± 37.872.7 ± 36.6 < 0.001eGFR (ml/min/1.73m2)95.1 ± 22.397.0 ± 21.592.3 ± 23.2 < 0.001ABI1.08 ± 0.091.10 ± 0.091.07 ± 0.08 < 0.001Systolic BP in left arm (mmHg)130 ± 20.1130 ± 18.0130 ± 22.30.438Diastolic BP in left arm (mmHg)78.8 ± 12.280.3 ± 11.677.1 ± 12.6 < 0.001Systolic BP in right arm (mmHg)131 ± 20.0131 ± 18.1131 ± 22.10.706Diastolic BP in right arm (mmHg)79.4 ± 12.280.8 ± 11.777.6 ± 12.6 < 0.001Systolic BP in left ankle (mmHg)146 ± 26.3147 ± 24.6144 ± 28.0 < 0.001Diastolic BP in left ankle (mmHg)76.6 ± 12.277.9 ± 11.974.9 ± 12.4 < 0.001Systolic BP in right ankle (mmHg)148 ± 26.7149 ± 25.1146 ± 28.3 < 0.001Diastolic BP in right ankle (mmHg)76.3 ± 12.277.8 ± 12.074.6 ± 12.3 < 0.001IABPD (mmHg)3.55 ± 3.793.50 ± 3.973.61 ± 3.550.092ABI ≤ 0.9256 ($1.9\%$)120 ($1.7\%$)136 ($2.3\%$)0.012IABPD ≥ 15 mmHg180 ($1.4\%$)82 ($1.1\%$)98 ($1.6\%$)0.014BMI body mass index, WC waist circumference, BP blood pressure, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, eGFR estimated glomerular filtration rate, ABI ankle-brachial index, IABPD interarm systolic blood pressure differenceValues are means ± SD, or numbers with percentage in parenthesis
## BMI and WC values according to different ABI and IABPD categories
The BMI and WC according to different ABI and IABPD categories were calculated and compared. WC was significantly higher in subjects with ABI ≤ 0.9 than that in subjects with ABI > 0.9 ($P \leq 0.001$, Table 2). However, BMI was not significantly different in subjects with ABI ≤ 0.9 and with ABI > 0.9 ($$P \leq 0.844$$, Table 2). At the same time, the WC and BMI were significantly higher in subjects with IABPD ≥ 15 mmHg than those in subjects with IABPD < 15 mmHg respectively (both $P \leq 0.001$, Table 2).Table 2BMI and WC values according to different ABI and IABPD categoriesVariablesABI > 0.9 ($$n = 12$$,888)ABI ≤ 0.9 ($$n = 256$$)P valueIABPD < 15 mmHg ($$n = 12$$,964)IABPD ≥ 15 mmHg ($$n = 180$$)P valueBMI (kg/m2)25.2 ± 3.7925.3 ± 4.750.84425.2 ± 3.7927.8 ± 4.66 < 0.001WC (cm)88.4 ± 11.691.0 ± 13.30.00188.4 ± 11.697.9 ± 12.2 < 0.001BMI body mass index, WC waist circumference, ABI ankle-brachial index, IABPD interarm systolic blood pressure differenceValues are means ± SD
## Independence of BMI and WC associated with different ABI and IABPD categories
In order to evaluate the independence of BMI and WC associated with different ABI and IABPD categories, multiple logistic regression analysis was used to calculate the OR and $95\%$ CI of BMI and WC associated with ABI ≤ 0.9 and IABPD ≥ 15 mmHg respectively with adjustment for other potential confounders including age, men, smoking, drinking, hypertension, diabetes mellitus, lipid disorders, and chronic kidney disease. These indicators of obesity entered regression equation as continuous variables respectively. We found that WC was independently associated with ABI ≤ 0.9 (OR 1.014, $95\%$ CI 1.002–1.026, $$P \leq 0.017$$, Table 3). Nevertheless, BMI was not independently associated with ABI ≤ 0.9 using this multiple logistic regression analysis. At the same time, the data showed that BMI and WC were independently associated with IABPD ≥ 15 mmHg respectively (OR 1.139, $95\%$ CI 1.100–1.181, $P \leq 0.001$, and OR 1.058, $95\%$ CI 1.044–1.072, $P \leq 0.001$, Table 3).Table 3Independence of BMI and WC associated with ABI ≤ 0.9 and IABPD ≥ 15 mmHgVariablesOR$95\%$ CIP valueABI ≤ 0.9––– BMI (kg/m2)0.9990.965–1.0340.954 WC (cm)1.0141.002–1.0260.017IABPD ≥ 15 mmHg––– BMI (kg/m2)1.1391.100–1.181 < 0.001 WC (cm)1.0581.044–1.072 < 0.001Multiple logistic regression analysis was used to calculate the odds ratio (OR) and $95\%$ CI of body mass index (BMI), and waist circumference (WC) (independent variables) associated with ankle-brachial index (ABI) ≤ 0.9, or interarm systolic blood pressure difference (IABPD) ≥ 15 mmHg respectively with adjustment for other potential confounders including age, men, smoking, drinking, hypertension, diabetes mellitus, lipid disorders, and chronic kidney disease. BMI and WC entered regression equation as continuous variables respectively
## Prevalence of ABI ≤ 0.9 and IABPD ≥ 15 mmHg with different categories of BMI
As we mentioned in the above section, though we cannot discover a linear relationship between BMI and ABI statistically, we still try to explore the prevalence of ABI ≤ 0.9 in study subjects when they were categorized as four groups according to BMI (< 20, 20 to < 25, 25 to < 30, and ≥ 30). As a result, we found that prevalence of ABI ≤ 0.9 was displayed with a U-shaped pattern according to different BMI categories (Fig. 2). Prevalence of ABI ≤ 0.9 in subjects with BMI < 20 and BMI ≥ 30 was significantly higher compared with that in subjects with BMI 20 to < 25 respectively (both $P \leq 0.001$).Fig. 2Prevalence of ABI ≤ 0.9 and IABPD ≥ 15 mmHg in different categories of BMI. ABI: ankle-brachial index; BMI: body mass index; IABPD: interarm systolic blood pressure difference. $$n = 925$$ for BMI < 20, 5643 for BMI ≥ 20 to < 25, 5203 for BMI ≥ 25 to < 30, and 1373 BMI ≥ 30. Prevalence of ABI ≤ 0.9 in subjects with BMI < 20 and BMI ≥ 30 was significantly higher compared with that in subjects with BMI ≥ 20 to < 25 respectively (both $P \leq 0.001$). Prevalence of IABPD ≥ 15 mmHg was significantly increased with incremental BMI (P for trend < 0.001) At the same time, we also tried to observe the prevalence of IABPD ≥ 15 mmHg when study subjects were categorized as four groups according to BMI. A different trend was discovered that prevalence of IABPD ≥ 15 mmHg was significantly increased with incremental BMI (P for trend < 0.001, Fig. 2).
## Relationship between BMI and abnormalities of peripheral arteries
The above data showed that, unlike WC, relationship between BMI and abnormalities of peripheral arteries appeared to be different. Thus, we further carefully evaluated whether various BMI categories (< 20, 20 to < 25, 25 to < 30, and ≥ 30) were associated with ABI ≤ 0.9 and IABPD ≥ 15 mmHg using multiple logistic regression analysis. The data displayed that, compared with BMI 20 to < 25, the risk of ABI ≤ 0.9 was significantly increased when BMI < 20 or ≥ 30 respectively (OR 2.595, $95\%$ CI 1.745–3.858, $P \leq 0.001$, and OR 1.618, $95\%$ CI 1.087–2.410, $$P \leq 0.018$$, Table 4). However, the risk of IABPD ≥ 15 mmHg tended to be increased when participants had bigger BMI. Compared with BMI 20 to < 25, the risk of IABPD ≥ 15 mmHg was significantly increased when BMI ≥ 30 (OR 3.218, $95\%$ CI 2.133–4.855, $P \leq 0.001$, Table 4).Table 4Various categories of BMI associated with ABI ≤ 0.9 and IABPD ≥ 15 mmHgVariablesOR$95\%$ CIP valueABI ≤ 0.9–– < 0.001 < 202.5951.745–3.858 < 0.001 20 to < 25ReferenceReferenceReference 25 to < 300.9950.736–1.3450.973 ≥ 301.6181.087–2.4100.018IABPD ≥ 15 mmHg–– < 0.001 < 200.2680.065–1.1060.069 20 to < 25ReferenceReferenceReference 25 to < 301.3570.944–1.9500.099 ≥ 303.2182.133–4.855 < 0.001Multiple logistic regression analysis was used to calculate the odds ratio (OR) and $95\%$ CI of body mass index (BMI) categories (< 20, $$n = 925$$; 20 to < 25, $$n = 5643$$; 25 to < 30, $$n = 5203$$; and ≥ 30, $$n = 1373$$), associated with ankle-brachial index (ABI) ≤ 0.9, or interarm systolic blood pressure difference (IABPD) ≥ 15 mmHg respectively with adjustment for other potential confounders including age, men, smoking, drinking, hypertension, diabetes mellitus, lipid disorders, and chronic kidney disease Furthermore, we also explored the nonlinear relationship between BMI and the risk of ABI ≤ 0.9 using a restricted cubic spline model by multivariable adjustment. Restricted cubic spline analysis (Fig. 3) indicated a significant U-shaped relationship between BMI and the risk of ABI ≤ 0.9 (P for non-linearity < 0.001).Fig. 3Nonlinear relationship between BMI and the risk of ABI ≤ 0.9. Y-axis stands for the odds ratio (OR) and $95\%$ CI of body mass index (BMI) (independent variable) associated with ankle-brachial index (ABI) ≤ 0.9 using restricted cubic spline analysis, with adjustment for other potential confounders including age, men, smoking, drinking, hypertension, diabetes mellitus, lipid disorders, and chronic kidney disease. A significant U-shaped relationship between BMI and the risk of ABI ≤ 0.9 was exhibited (P for non-linearity < 0.001)
## Discussion
The association between obesity and abnormalities of upper extremity arteries was rarely investigated in previous studies. The data in our study showed that both general and abdominal obesity parameters were independently associated with IABPD ≥ 15 mmHg respectively. A previous data demonstrated that BMI was in connection with IABPD ≥ 10 mmHg. However, WC was only statistically associated with inter-arm differences in diastolic blood pressure ≥ 10 mmHg [15]. The cut-off value of inter-arm differences was different from that in our study, and the sample size of the previous study was relatively small.
However, only abdominal obesity parameter was significantly higher in subjects with ABI ≤ 0.9. Meanwhile, BMI was not independently associated with ABI ≤ 0.9 using linear statistical models. In fact, previous studies also manifested that more patients with ABI ≤ 0.9 had abdominal obesity than those without lower extremity PAD [16]. Lots of data demonstrated that abdominal obesity was an independent risk factor for the development of ASCVD including PAD [12, 16]. These were possibly because the people with abdominal obesity tended to have more atherosclerotic plaques in arteries. Nevertheless, previous study provided controversial data on association between BMI and ASCVD including lower extremity PAD [17]. More abundant data on both abdominal obesity and general obesity associated with lower extremity PAD were supplied in our study, and we think the data in this study can be helpful to explore the relationship between obesity and lower extremity PAD.
In order to explore the relationship between BMI and lower extremity PAD further, subjects in this study were categorized as four groups according to BMI. As a result, we found that compared with BMI 20 to < 25, the risk of ABI ≤ 0.9 was significantly increased by more than 2.5-fold and 1.6-fold when BMI < 20 or ≥ 30 respectively. Additionally, a significant U-shaped relationship was observed between BMI and the risk of ABI ≤ 0.9 using restricted cubic spline analysis, which indicated that the risk of ABI ≤ 0.9 increased when BMI exceeded or less than the median value (i.e., 24.97). These data manifested that it was not a linear relationship, but a U-shaped pattern between BMI and ABI ≤ 0.9 in this Chinese population of our study. We speculated that when subjects had bigger BMI, they would possibly have more atherosclerotic plaques in lower extremity arteries. However, this study showed that underweight subjects also had increased prevalence of ABI ≤ 0.9. Similar results were found in a previous study [18], but the reason is not very clear yet. Some researchers considered that the underweight patients possibly had higher levels of inflammation which might promote the development of atherosclerosis [19]. In fact, a phenomenon called obesity paradox showed that a low body weight was also associated with cardiovascular disease and mortalities [20]. A previous study manifested that obesity was associated with lower in-hospital mortality in PAD patients relative to those with normal-weight/over-weight. This obesity survival paradox was independent of age, gender and comorbidities and observed for all obesity classes [21]. However, the precise mechanism is still not clear. We think that the obesity paradox between BMI and ABI needs to be further studied.
We not only studied the relationship between general obesity and lower extremity PAD, but also studied the relationship between general obesity and upper extremity PAD in this study. Univariate analysis and multiple logistic regression analysis indicated that prevalence of IABPD ≥ 15 mmHg was significantly increased with incremental BMI. This data was quite different according to the above analysis on the relationship between general obesity and ABI ≤ 0.9. However, the causes for this discrepancy were unknown. We speculated the possible causes as follows. First, ABI ≤ 0.90 can be considered as the presence of lower extremity PAD. However, IABPD ≥ 15 mmHg possibly signifies abnormalities in upper extremity arteries mainly including subclavian artery, brachiocephalic trunk, and axillary artery [10]. Risk factors for abnormalities of arteries at different anatomical locations might be different. Second, though atherosclerosis is the main cause of the upper or lower extremity PAD, there are also other divergent causes. Lower extremity PAD might be caused by atherosclerosis, takayasu arteritis, and so on. Meanwhile, more causes of the upper extremity PAD were found such as atherosclerosis, thoracic outlet syndrome, giant cell arteritis, takayasu arteritis, radiation artery fibrosis, fibromuscular dysplasia, and so on [10]. In fact, atherosclerosis in lower extremity PAD is possibly more frequently to be found compared with that in upper extremity PAD [10, 11]. The associations of these divergent pathogenic risk factors with BMI appear to be more complex. Of course, the future studies are still needed to explain the discrepancies.
In summary, this study demonstrates that abdominal obesity is an independent risk factor for abnormalities of upper and lower extremity arteries. Meanwhile, general obesity is also independently associated with abnormalities of upper extremity arteries. However, the association between general obesity and lower extremity PAD is displayed with a U-shaped pattern. We believe that the exploration of risk factors for abnormalities of upper and lower extremity arteries, and then the comprehensive control of abdominal obesity, will possibly help to improve the understanding and control of PAD.
## References
1. Yusuf S, Reddy S, Ounpuu S, Anand S. **Global burden of cardiovascular disease: part I: general considerations, the epidemiologic transition, risk factors, and impact of urbanization**. *Circulation* (2001) **104** 2746-2753. DOI: 10.1161/hc4601.099487
2. Robinson JG, Gidding SS. **Curing atherosclerosis should be the next major cardiovascular prevention goal**. *J Am Coll Cardiol* (2014) **63** 2779-2785. DOI: 10.1016/j.jacc.2014.04.009
3. Wang Y, Xu Y, Li J, Wei Y, Zhao D, Hou L. **Characteristics of prevalence in peripheral arterial disease and correlative risk factors and comorbidities among female natural population in China**. *Vasa* (2010) **39** 305-311. DOI: 10.1024/0301-1526/a000054
4. Zhao D, Liu J, Wang M, Zhang X, Zhou M. **Epidemiology of cardiovascular disease in China: current features and implications**. *Nat Rev Cardiol* (2019) **16** 203-212. DOI: 10.1038/s41569-018-0119-4
5. Hiatt WR. **Medical treatment of peripheral arterial disease and claudication**. *N Engl J Med* (2001) **344** 1608-1621. DOI: 10.1056/NEJM200105243442108
6. Wang Y, Mou Q, Zhao D, Xu Y, Hu D, Ma H. **Predictive value of ankle-brachial index and blood glucose on the outcomes of six-year all-cause mortality and cardiovascular mortality in a Chinese population of type 2 diabetes patients**. *Int Angiol* (2012) **31** 586-594. PMID: 23222938
7. Luo YY, Li J, Xin Y, Zheng LQ, Yu JM, Hu DY. **Risk factors of peripheral arterial disease and relationship between low ankle brachial index and mortality from all-cause and cardiovascular disease in Chinese patients with hypertension**. *J Hum Hypertens* (2007) **21** 461-466. DOI: 10.1038/sj.jhh.1002177
8. Gerhard-Herman MD, Gornik HL, Barrett C, Barshes NR, Corriere MA, Drachman DE. **2016 AHA/ACC guideline on the management of patients with lower extremity peripheral artery disease: a report of the American College of Cardiology/American Heart Association task force on clinical practice guidelines**. *Circulation* (2017) **135** e726-e779. PMID: 27840333
9. Clark CE, Taylor RS, Shore AC, Ukoumunne OC, Campbell JL. **Association of a difference in systolic blood pressure between arms with vascular disease and mortality: a systematic review and meta-analysis**. *Lancet* (2012) **379** 905-914. DOI: 10.1016/S0140-6736(11)61710-8
10. Aboyans V, Ricco JB, Bartelink MEL, Björck M, Brodmann M, Cohnert T. **2017 ESC Guidelines on the Diagnosis and Treatment of Peripheral Arterial Diseases, in collaboration with the European Society for Vascular Surgery (ESVS): Document covering atherosclerotic disease of extracranial carotid and vertebral, mesenteric, renal, upper and lower extremity arteriesEndorsed by: the European Stroke Organization (ESO)The Task Force for the Diagnosis and Treatment of Peripheral Arterial Diseases of the European Society of Cardiology (ESC) and of the European Society for Vascular Surgery (ESVS)**. *Eur Heart J* (2018) **39** 763-816. DOI: 10.1093/eurheartj/ehx095
11. Sheng CS, Liu M, Zeng WF, Huang QF, Li Y, Wang JG. **Four-limb blood pressure as predictors of mortality in elderly Chinese**. *Hypertension* (2013) **61** 1155-1160. DOI: 10.1161/HYPERTENSIONAHA.111.00969
12. Powell-Wiley TM, Poirier P, Burke LE, Després JP, Gordon-Larsen P, Lavie CJ. **Obesity and cardiovascular disease: a scientific statement from the American heart association**. *Circulation* (2021) **143** e984-1010. DOI: 10.1161/CIR.0000000000000973
13. Huang Y, Xu M, Xie L, Wang T, Huang XL, Lv XF. **Obesity and peripheral arterial disease: a Mendelian randomization analysis**. *Atherosclerosis* (2016) **247** 218-224. DOI: 10.1016/j.atherosclerosis.2015.12.034
14. Keller K, Hobohm L, Geyer M, Münzel T, Lavie CJ, Ostad MA. **Obesity paradox in peripheral artery disease**. *Clin Nutr* (2019) **38** 2269-2276. DOI: 10.1016/j.clnu.2018.09.031
15. Muñoz-Torres FJ, Andriankaja OM, Ruiz JI, Joshipura KJ. **Longitudinal association between adiposity and inter-arm blood pressure difference**. *J Clin Hypertens* (2019) **21** 1519-1526. DOI: 10.1111/jch.13678
16. Umuerri EM, Obasohan AO. **Obesity indices and peripheral artery disease measured by ankle brachial index in Nigerian out-patients**. *West Afr J Med* (2018) **35** 3-8. PMID: 29607470
17. Heffron SP, Dwivedi A, Rockman CB, Xia Y, Guo Y, Zhong J. **Body mass index and peripheral artery disease**. *Atherosclerosis* (2020) **292** 31-36. DOI: 10.1016/j.atherosclerosis.2019.10.017
18. Li J, Shichao Yu, Zhou W, Zhu L, Wang T, Bao H. **U-shaped association of body mass index with the risk of peripheral arterial disease in chinese hypertensive population**. *Int J Gen Med* (2021) **14** 3627-3634. DOI: 10.2147/IJGM.S323769
19. Kahraman S, Yilmaz R, Akinci D, Arici M, Altun B, Erdem Y. **U-shaped association of body mass index with inflammation and atherosclerosis in hemodialysis patients**. *J Ren Nutr* (2005) **15** 377-386. DOI: 10.1053/j.jrn.2005.07.004
20. Elagizi A, Kachur S, Lavie CJ, Carbone S, Pandey A, Ortega FB. **An overview and update on obesity and the obesity paradox in cardiovascular diseases**. *Prog Cardiovasc Dis* (2018) **61** 142-150. DOI: 10.1016/j.pcad.2018.07.003
21. Skilton MR, Chin-Dusting JP, Dart AM, Brazionis L, Lantieri O, O'Dea K. **Metabolic health, obesity and 9-year incidence of peripheral arterial disease: the D.E.S.I.R.study**. *Atherosclerosis* (2011) **216** 471-476. DOI: 10.1016/j.atherosclerosis.2011.02.032
|
---
title: Diagnostic value of combining preoperative inflammatory markers ratios with
CA199 for patients with early-stage pancreatic cancer
authors:
- Yuanlong Gu
- Qianjin Hua
- Zhipeng Li
- Xingli Zhang
- Changjie Lou
- Yangfen Zhang
- Wei Wang
- Peiyuan Cai
- Juan Zhao
journal: BMC Cancer
year: 2023
pmcid: PMC9999638
doi: 10.1186/s12885-023-10653-4
license: CC BY 4.0
---
# Diagnostic value of combining preoperative inflammatory markers ratios with CA199 for patients with early-stage pancreatic cancer
## Abstract
### Background
An early diagnosis of pancreatic cancer (PC) is extremely difficult because of the lack of sensitive liquid biopsy methods and effective biomarkers. We attempted to evaluate whether circulating inflammatory marker could complement CA199 for the detection of early-stage PC.
### Methods
We enrolled 430 patients with early-stage PC, 287 patients with other pancreatic tumors (OPT), and 401 healthy controls (HC). The patients and HC were randomly divided into a training set ($$n = 872$$) and two testing sets (n1 = 218, n2 = 28). The receiver operating characteristic (ROC) curves were investigated to evaluate the diagnostic performance of circulating inflammatory markers ratios, CA199, and combinations of the markers ratios in the training set, which would then be validated in the two testing sets.
### Results
Circulating fibrinogen, neutrophils, and monocytes in patients with PC were significantly higher while circulating albumin, prealbumin, lymphocytes, and platelets of patients with PC were significantly lower compared to those of HC and OPT (all $P \leq 0.05$). The fibrinogen-to-albumin (FAR), fibrinogen-to-prealbumin (FPR), neutrophil-to-lymphocyte (NLR), platelet-to-lymphocyte (PLR), monocyte-to-lymphocyte (MLR), and fibrinogen-to-lymphocyte (FLR) ratios were significantly higher while the prognostic nutrition index values (PNI) were lower in patients with PC than in HC and OPT (all $P \leq 0.05$). Combining the FAR, FPR, and FLR with CA199 exhibited the best diagnostic value for distinguishing patients with early-stage PC from HC with an area under the curve (AUC) of 0.964, and for distinguishing patients with early-stage PC from OPT with an AUC of 0.924 in the training sets. In the testing set, compared with HC, the combination markers had powerful efficiency for PC with an AUC 0.947 and AUC 0.942 when comparing PC with OPT. The AUC was 0.915 for the combination of CA199, FAR, FPR, and FLR for differentiating between patients with pancreatic head cancer (PHC) and other pancreatic head tumors (OPHT), and 0.894 for differentiating between patients with pancreatic body and tail cancer (PBTC) and other pancreatic body and tail tumors (OPBTT).
### Conclusion
A combination of FAR, FPR, FLR, and CA199 may serve as a potential non-invasive biomarker for differentiating early-stage PC from HC and OPT, especially early-stage PHC.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12885-023-10653-4.
## Introduction
Pancreatic cancer (PC) is the seventh leading cause of cancer-related deaths in both men and women with nearly equal rates of annual incidence and mortality [1]; it has been projected that by 2030, PC will be the second leading cause of cancer-related deaths, surpassing breast cancer, prostate, and colorectal cancers [2]. Surgical resection remains the primary form of treatment for patients with PC [3]. Currently, the diagnosis of PC is mainly based on clinical signs and symptoms, imaging techniques, serum CA199, and pathological features. However, most patients with PC are already at an advanced stage when they first visit the hospital, losing the opportunity for surgery, with a five-year survival rate of < $5\%$ [4]. Thus, more reliable diagnostic biomarkers are urgently needed to improve early diagnosis of PC.
In recent years, liquid biopsies to isolate circulating tumor DNA (ctDNA) [5], circulating tumor cells (CTCs) [6], circulating exosomal miRNA [7], and exosomal GPC1 [8] for the early detection of PC have re ceived much attention. However, these methods are complex, time-consuming, expensive, and difficult to perform. Tumor-promoting inflammation is the seventh most important feature of cancer cells [9]. Circulating inflammatory markers such as C-reactive protein (CRP) [10], neutrophils [11], lymphocytes, platelets, monocytes [12], and fibrinogen [13] play an essential role in the oncogenesis and development of cancer. Some studies have found that inflammation markers ratios could predict the prognoses of patients with PC. For example, CRP-to-albumin score, the Glasgow Prognostic Score (GPS) each have an independent prognostic value in patients with PC [14]. A high neutrophil-to-lymphocyte ratio (NLR) is associated with an adverse overall survival (OS) in pancreatic cancer [15]. A low fibrinogen-to-albumin ratio (FAR) was positively correlated with a good OS in locally advanced or metastatic PC [16].
Notably, inflammation is evident at the earliest stages of tumor progression and could promote the development of incipient tumors into full-blown cancers [17]. Therefore, we hypothesized that these circulating inflammatory markers change within the early stages of cancer and could act as reliable indicators for early diagnoses of PC. In this study, we assessed inflammation indicator values including FAR, fibrinogen-to-prealbumin ratio (FPR), NLR, platelet-to-lymphocyte ratio (PLR), monocyte-to-lymphocyte ratio (MLR), and prognostic nutritional index (albumin + 5 × lymphocyte count; PNI) in early-stage PC, healthy controls (HC), and other pancreatic tumors (OPT), with the aim of exploring whether inflammation indicators could be used as markers for the diagnosis of early-stage PC.
## Patients collection
This study included 422 patients with PC, 119 patients with benign pancreatic tumors (BPT; 39 chronic pancreatitis, 56 pancreatic serous cystadenomas, and 24 pancreatic mucinous cystadenomas), 98 patients with solid pseudo-papilloma of the pancreas (SPT), 59 patients with pancreatic neuroendocrine tumors (PNET), and 392 healthy controls (HC) from January 2015 to December 2021 at the Harbin Medical University Cancer Hospital. Eight patients with PC, 11 with other pancreatic diseases (OPT; two CP, two SPT, and seven pancreatic serous or mucinous cystadenoma), and nine HC from January 2017 to December 2021 in the Municipal Hospital Affiliated to Taizhou University were also enrolled in this study. The inclusion and exclusion criteria were as follows:1) age ≥ 18 years; 2) pathologically confirmed diagnoses of PC(adenocarcinoma, pancreatic ductal adenocarcinoma, and mucinous adenocarcinoma), neuroendocrine tumor (G1, G2, and G3), solid pseudopapillary neoplasm, chronic pancreatitis, pancreatic serous cystadenoma, and pancreatic mucinous cystadenoma; 3) R0 resection (radical surgical resection); 4) PC pathology at TNM stage I—II; 5) available clinical baseline information; 6) no antitumor therapy performed before surgery; 7) no second primary cancer; 8) no history of autoimmune disorders, hepatitis, nephropathy, coagulation disorders, or HIV infection; and 9) no acute inflammation before surgery.
Each disease group and HC from Harbin Medical University Cancer Hospital were randomly divided into training and testing sets 1 at a ratio of 4:1. The patients and HC from Municipal Hospital Affiliated to Taizhou University were used as testing set 2. Ethical approval for this study was granted by the Harbin Medical University Cancer Hospital and Municipal Hospital Affiliated to Taizhou University Ethics Committee, and all participants provided signed informed consent forms.
## Data collection
Detailed baseline and clinicopathological information, including sex, age, tumor location, tumor size, pathological type, differentiation, lymph node metastasis, and TNM stage of the patients with pancreatic diseases and HC, were obtained from the medical records of the inpatients or outpatients. The preoperative hematological parameters and liver function tests included neutrophils (× 109/L), lymphocytes (× 109/L), monocytes (× 109/L), platelets (× 109/L), plasma fibrinogens (g/L), serum albumins (g/L), prealbumin (mg/L), and CA199 (U/L) within seven days before surgery (average 2—7 days) were gathered from the medical records. TNM staging was performed using the 8th edition of the AJCC Cancer Staging Manual for Pancreatic Cancer.
## Inflammation markers ratios definitions
FAR, FPR, NLR, PLR, MLR, and FLR were defined as the plasma fibrinogen value divided by the serum albumin value, plasma fibrinogen value divided by the serum prealbumin value, neutrophil count divided by the lymphocyte count, platelet count divided by the lymphocyte count, monocyte count divided by the lymphocyte count, and plasma fibrinogen value divided by the lymphocyte count, respectively. PNI was defined as serum albumin value + 5 × lymphocyte count.
## Statistical analysis
Data were presented as mean ± standard deviation (SD). The differences in inflammatory markers and inflammatory markers ratios in different groups were examined using the Student’s t-test. A two-sided $p \leq 0.05$ was considered statistically significant. The receiver operating characteristic (ROC) curve and the area under the ROC curve (AUC) were used to evaluate the diagnostic accuracy of the inflammation indicator and CA199 for early-stage PC and the discrimination ability between early-stage PC and PNET, SPT, and BPT. ROC curve analysis was also used to determine the best cut-off values for FAR, FPR, NLR, PLR, MLR, PNI, FLR, and CA199 based on the maximum Youden index. AUC values < 0.7, 0.7—0.9, and > 0.9 were considered as low, medium, and high diagnostic power, respectively. All statistical analyses were conducted using SPSS (version23.0, IBM Corp., Armonk, NY, USA) and GraphPad Prism (version 5.0, La Jolla, CA, USA).
## Clinical characteristics and circulating inflammatory markers of patients with pancreatic diseases and HC
A total of 338 patients with early-stage PC, 96 with BPT, 78 with SPT, 47 with PNET, and 313 HC were assigned to the training set. Among the patients with PC, 187 ($55.3\%$) were male, and the average age was 57.5 ± 8.3 years, whereas among 78 patients with SPT, 66 ($84.6\%$) were female, and the average age was 35.5 ± 14.1 years. Most patients with SPT were young women. Most patients with PC had tumors located in the pancreatic head ($76.9\%$), whereas most patients with BPT, SPT, and PNET had tumors located in the pancreatic body and tail cysts (76, 67.9, and $72.3\%$, respectively). Most patients with PC had invasive ductal carcinomas ($91.1\%$). The clinical and pathological characteristics of the patients in the training and testing sets were similar. Detailed information on the patients and HC in the training and testing sets are listed in Table 1.Table 1Clinical characteristics of patients with pancreatic diseases and healthy controls in training and testing setsTraining setGroupsPC [338]HC [313]BPT [96]SPT [78]PNET [47]Gender Male187(55.3)164(52.3)25(26.0)12(15.4)27(57.4) Female151(44.7)149(47.6)71(74.0)66(84.6)20(42.6)Age ≤ 60215(63.6)205(65.4)70(72.9)70(89.7)20(42.6) > 60123(36.4)103(34.5)26(27.1)8(10.3)27(57.4)CA199 ≥ 37240(71.0)5(1.6)13(13.5)4(5.1)2(4.3) < 3798(29.0)308(98.4)83(86.5)74(94.9)45(95.7)Location: Head260(76.9)23(24.0)25(32.1)13(27.7) Body or Tail78(23.1)73(76.0)53(67.9)34(72.3)Tumor size > 4 cm81(24.0)40(41.6)40(51.3)16(34.0) ≤ 4 cm257(76.0)56(58.4)38(48.7)31(66.0)Pathological type Ductal adenocarcinoma308(91.1)31(32.3; chronic pancreatitis) others30(8.9)65(67.7; adenoma)Differentiation High and Moderate218(64.5)40(85.1; G1-G2) Poor120(35.5)7(14.94; G3)Lymph nodes + 107(31.7) -231(68.3)TNM stage I166(49.1) II172(50.9)Fibrinogen(g/L)3.44 ± 0.952.87 ± 0.562.76 ± 0.762.46 ± 0.692.54 ± 0.58albumin(g/L)38.52 ± 4.0443.75 ± 2.3839.62 ± 3.4540.79 ± 3.7240.69 ± 3.11prealbumin(mg/L)215.17 ± 69.86322.74 ± 58.73255.85 ± 61.72239.98 ± 56.07271.97 ± 65.09neutrophil(× 109/L)3.95 ± 1.643.45 ± 1.053.41 ± 1.953.70 ± 1.473.45 ± 1.24lymphocyte(× 109/L)1.62 ± 0.621.94 ± 0.562.03 ± 0.632.09 ± 0.541.98 ± 0.59platelet(× 109/L)224.43 ± 69.06238.48 ± 52.91227.42 ± 67.02248.58 ± 75.31212.57 ± 51.78monocyte(× 109/L)0.51 ± 0.190.37 ± 0.110.46 ± 0.190.52 ± 0.200.43 ± 0.16AFR (Mean ± SD)0.091 ± 0.030.066 ± 0.0130.71 ± 0.020.061 ± 0.0170.063 ± 0.01APR (Mean ± SD)0.019 ± 0.0120.009 ± 0.0030.011 ± 0.0050.011 ± 0.0060.009 ± 0.003NLR (Mean ± SD)2.78 ± 1.671.89 ± 0.741.81 ± 1.261.84 ± 0.771.87 ± 0.83PLR (Mean ± SD)156.32 ± 75.68131.15 ± 45.19119.15 ± 40.90123.36 ± 42.09115.21 ± 40.50MLR (Mean ± SD)0.36 ± 0.210.20 ± 0.060.24 ± 0.110.26 ± 0.100.23 ± 0.09PNI (Mean ± SD)46.62 ± 5.3353.47 ± 3.8149.79 ± 4.5051.23 ± 4.9250.59 ± 4.64FLR (Mean ± SD)2.46 ± 1.261.57 ± 0.541.47 ± 0.561.26 ± 0.511.38 ± 0.44Testing set 1GroupsPC [84]HC [79]BPT [23]SPT [20]PNET [12]Gender Male46(54.8)46(58.2)9(39.1)4(20.0)8(66.7) Female38(45.2)33(41.8)14(60.9)16(80.0)4(33.3)Age ≤ 6044(52.4)56(70.9)13(56.6)17(85.0)7(58.3) > 6040(47.6)23(29.1)10(43.4)3(15.0)5(41.7)CA199 ≥ 3763(75.0)2(2.6)9(39.1)1(5.0)2(16.7) < 3721(35.0)77(97.4)14(60.8)19(95.0)10(83.3)Location: Head58(69.0)4(17.4)5(25.0)5(41.7) Body or Tail34(31.0)19(82.6)15(75.0)7(58.3)Tumor size > 4 cm21(44.7)11(47.8)11(55.0)4(33.3) ≤ 4 cm63(55.3)12(52.2)9(45.0)8(66.7)Pathological type Ductal adenocarcinoma75(89.3)8(34.7; chronic pancreatitis) others9(10.7)15(65.2; adenoma)Differentiation High and Moderate43(51.2)10(83.3; G1-G2) Poor41(48.8)2(16.7; G3)Lymph nodes + 17(20.2) -67(79.8)TNM stage I50(59.5) II34(40.5)Fibrinogen(g/L)3.56 ± 0.942.82 ± 0.492.59 ± 0.582.42 ± 0.602.59 ± 0.69albumin(g/L)37.98 ± 3.0743.78 ± 2.0938.68 ± 3.2340.54 ± 3.2540.5 ± 3.28prealbumin(mg/L)212.73 ± 45.03312.30 ± 57.40269.65 ± 63.85231.74 ± 55.87280.17 ± 52.37neutrophil(× 109/L)3.88 ± 1.263.46 ± 0.953.37 ± 1.243.24 ± 1.123.17 ± 1.15lymphocyte(× 109/L)1.59 ± 0.452.01 ± 0.661.89 ± 0.442.05 ± 0.761.84 ± 0.42platelet(× 109/L)234.04 ± 77.48250.29 ± 61.98234.26 ± 78.76241.37 ± 49.04198.08 ± 46.20monocyte(× 109/L)0.49 ± 0.140.37 ± 0.100.42 ± 0.140.42 ± 0.130.40 ± 0.14AFR (Mean ± SD)0.095 ± 0.0280.065 ± 0.010.067 ± 0.0150.060 ± 0.0150.064 ± 0.015APR (Mean ± SD)0.018 ± 0.0090.009 ± 0.0030.010 ± 0.0030.011 ± 0.0050.009 ± 0.002NLR (Mean ± SD)2.57 ± 0.851.85 ± 0.651.83 ± 0.721.73 ± 0.751.74 ± 0.59PLR (Mean ± SD)157.09 ± 60.35133.85 ± 45.32128.29 ± 49.03135.00 ± 57.96112.36 ± 34.29MLR (Mean ± SD)0.33 ± 0.130.20 ± 0.060.22 ± 0.060.24 ± 0.150.22 ± 0.07PNI (Mean ± SD)45.91 ± 3.8853.83 ± 3.9548.16 ± 4.1150.78 ± 5.1349.70 ± 4.67FLR (Mean ± SD)2.45 ± 0.971.55 ± 0.551.45 ± 0.501.34 ± 0.591.46 ± 0.47Testing set 2GroupsPC [8]HC [9]BPT [11]Gender Male5(62.5)7(77.8)6(54.5) Female3(37.5)2(22.2)5(45.4)Age ≤ 605(62.5)6(66.7)6(54.5) > 603(37.5)3(33.3)5(45.4)CA199 ≥ 376(75.0)1(11.1)1(9.1) < 372(25.0)8(88.9)10(90.9)Location: Head6(75.0)4(36.4) Body or Tail2(25.0)7(63.6)Tumor size > 4 cm2(25.0)5(45.5) ≤ 4 cm6(75.0)6(54.4)Pathological type Ductal adenocarcinoma7(87.5)2(18.2; chronic pancreatitis)7 (63.6; adenoma)2(18.2, Solid pseudo papilloma) others1(12.5)Differentiation High and Moderate5(62.5) Poor3(37.5)Lymph nodes + 3(37.5) -5(62.5)TNM stage I4(50.0) II4(50.0)Fibrinogen(g/L)3.16 ± 0.552.55 ± 0.542.81 ± 0.52albumin(g/L)38.20 ± 3.0840.83 ± 4.4943.66 ± 3.62prealbumin(mg/L)205.25 ± 22.32287.78 ± 15.70278.72 ± 10.51neutrophil(× 109/L)3.69 ± 0.862.97 ± 0.712.75 ± 0.71lymphocyte(× 109/L)1.46 ± 0.341.95 ± 0.431.86 ± 0.54platelet(× 109/L)192.5 ± 54.5230.67 ± 48.95237.24 ± 97.47monocyte(× 109/L)0.48 ± 0.160.40 ± 0.160.36 ± 0.13AFR (Mean ± SD)0.082 ± 0.110.063 ± 0.0150.65 ± 0.13APR (Mean ± SD)0.015 ± 0.0020.009 ± 0.0020.010 ± 0.002NLR (Mean ± SD)2.60 ± 0.731.58 ± 0.431.54 ± 0.44PLR (Mean ± SD)135.35 ± 40.17121.40 ± 28.15133.67 ± 54.68MLR (Mean ± SD)0.33 ± 0.810.20 ± 0.060.20 ± 0.07PNI (Mean ± SD)45.51 ± 3.5850.57 ± 4.5952.90 ± 4.44FLR (Mean ± SD)2.24 ± 0.561.35 ± 0.341.56 ± 0.32PC pancreatic cancer, HC Healthy controls, BPT Benign pancreas tumors, SPT Solid pseudo papilloma of the pancreas, PNET Pancreatic neuroendocrine tumors, FAR Fibrinogen-to-albumin ratio, FPR Fibrinogen-to-prealbumin ratio, NLR Neutrophil-to-lymphocyte ratio, PLR Platelets-to-lymphocyte ratio, MLR Monocytes-to-lymphocyte ratio, PNI Albumin + 5 × the lymphocyte count, FLR Fibrinogen-to- lymphocyte ratio We compared the hematological and biochemical parameters of patients with PC, BPT, SPT, PNET, and HC. As shown in Fig. 1, in the training set, the average fibrinogen, neutrophil, platelet, and monocyte levels in patients with PC were 3.44 ± 0.95 g/L, 3.95 ± 1.64 × 109/L, 224.43 ± 69.06 × 109/L, and 0.51 ± 0.19 × 109/L, respectively; these were significantly higher compared to those of the HC and OPT groups, with P values < 0.05. In contrast, the average albumin, prealbumin, lymphocytes, and platelets of patients with PC were 38.52 ± 4.04 g/L, 215.17 ± 69.86 mg/L, 1.62 ± 0.62 × 109/L, 224.43 ± 69.06 × 109/L, respectively, which were significantly lower than those in the HC and OPT groups, with P values < 0.05. The results obtained from the testing set were consistent with those obtained from the training set (Supplementary Fig. 1). These results suggest that circulating inflammatory markers had already changed in the early stages of PC.Fig. 1The circulating inflammation markers in PC, HC, BPT, SPT, and PNET in training sets. The plasma fibrinogens (A), serum albumins (B), prealbumin (C), neutrophils (D), lymphocytes (E), platelets (F), and monocytes (G) in PC, HC, BPT, SPT, and PNET. Abbreviations: PC, pancreatic cancer; BPT, benign pancreas tumors; SPT, solid pseudo papilloma of the pancreas; PNET, patients with pancreatic neuroendocrine tumors; HC, healthy controls
## Inflammation markers ratios values in pancreatic diseases and HC
As shown in Table 1, in the training set, FAR, FPR, NLR, PLR, MLR, PNI, and FLR values of patients with PC were 0.091 ± 0.03, 0.019 ± 0.012, 2.78 ± 1.67, 156.32 ± 75.68, 0.36 ± 0.21, 46.62 ± 5.33, and 2.46 ± 1.26, respectively. FAR values were significantly higher in patients with PC than those in the HC, BPT, SPT, and PNET groups (Fig. 2A, $P \leq 0.0001$, $P \leq 0.0001$, $P \leq 0.0001$, and $P \leq 0.0001$, respectively). FPR values were significantly higher in patients with PC than those of HC, BPT, SPT, and PNET (Fig. 2B, $P \leq 0.0001$, $P \leq 0.0001$, $P \leq 0.0001$, and $P \leq 0.0001$, respectively). NLR values were significantly higher in patients with PC than those in the HC, BPT, SPT, and PNET groups (Fig. 2C, $P \leq 0.0001$, $P \leq 0.0001$, $P \leq 0.0001$, and $$P \leq 0.0003$$, respectively). PLR values were higher in patients with PC than those in the HC, BPT, SPT, and PNET groups (Fig. 2D, $P \leq 0.0001$, $P \leq 0.0001$, $$P \leq 0.0002$$, and $$P \leq 0.0003$$, respectively). MLR values were higher in patients with PC than those in the HC, BPT, SPT, and PNET groups (Fig. 2E, $P \leq 0.0001$, $P \leq 0.0001$, $P \leq 0.0001$, and $P \leq 0.0001$, respectively). FLR values were higher in patients with PC than those in the HC, BPT, SPT, and PNET groups (Fig. 2F, $P \leq 0.0001$, $P \leq 0.0001$, $P \leq 0.0001$, and $$P \leq 0.0008$$, respectively). In contrast, PNI values were lower in patients with PC than those in the HC, BPT, SPT, and PNET groups (Fig. 2G, $P \leq 0.0001$, $P \leq 0.0001$, $P \leq 0.0001$, and $P \leq 0.0001$, respectively). The results from the testing sets were consistent with those from the training set; the detailed data in the testing sets are shown in supplementary Fig. 2A—G. These results indicated that the inflammation markers ratios were significantly altered in patients with early-stage PC.Fig. 2The inflammation markers ratios in PC, HC, BPT, SPT, and PNET in training sets. The FAR (A), FPR (B), NLR (C), PLR (D), MLR (E), FLR (F), and PNI (G) in PC, HC, BPT, SPT, and PNET. Abbreviations: PC, pancreatic cancer; BPT, benign pancreas tumors; SPT, solid pseudo papilloma of the pancreas; PNET, patients with pancreatic neuroendocrine tumors; HC, healthy controls; FAR, fibrinogen-to-albumin ratio; FPR, fibrinogen-to-prealbumin ratio; NLR, neutrophil-to-lymphocyte ratio; PLR, platelets-to-lymphocyte ratio; MLR monocytes-to-lymphocyte ratio; PNI, albumin + 5 × the lymphocyte count, FLR, fibrinogen-to- lymphocyte ratio
## Diagnostic and differential diagnosis values of inflammation markers ratios values in PC
In the training sets, the ROC curve was used to evaluate the capabilities of CA199, FAR, FPR, NLR, PLR, MLR, and PNI in discriminating between early stage PC and HC. The AUC values were 0.868 for CA199 ($P \leq 0.0001$, cutoff 24.540, sensitivity 0.939, specificity 0.817), 0.776 for FAR ($P \leq 0.0001$, cutoff 0.080, sensitivity 0.885, specificity 0.556), 0.869 for FPR ($P \leq 0.0001$, cutoff 0.012, sensitivity 0.837, specificity 0.775), 0.686 for NLR ($P \leq 0.0001$, cutoff 2.252, sensitivity 0.780, specificity 0.527), 0.584 for PLR ($$P \leq 0.0002$$, cutoff 177.218, sensitivity 0.879, specificity 0.299), 0.818 for MLR ($P \leq 0.0001$, cutoff 0.249, sensitivity 0.830, specificity 0.678), 0.748 for FLR ($P \leq 0.0001$, cutoff 1.864, sensitivity 0.773, specificity 0.639), and 0.860 for PNI ($P \leq 0.0001$, cutoff 49.025, sensitivity 0.907, specificity 0.707) (Fig. 3A, Table 2). The AUC was 0.942 for a combination of CA199 and FAR, 0.964 for CA199 and FPR, 0.940 for CA199 + MLR, 0.955 for CA199 + PNI, 0.964 for CA199 + FAR + FPR, 0.964 for CA199 + FAR + FPR + FLR, and 0.976 for CA199 + FAR + FPR + MLR + PNI (Fig. 3B, Table 2). To determine whether inflammation indicator values could differentiate PC from other pancreatic diseases (OPT), we generated ROC curves. As shown in Fig. 3C-D and Table 3, the AUC was 0.846 for CA199 ($P \leq 0.0001$, cut-off 32.205, sensitivity 0.887, specificity 0.772), 0.778 for FAR ($P \leq 0.0001$, cut-off 0.070,sensitivity 0.701, specificity 0.734), 0.779 for FPR ($P \leq 0.0001$, cut-off 0.013, sensitivity 0.778, specificity 0.666), 0.716 for NLR ($P \leq 0.0001$, cut-off 1.961, sensitivity 0.674, specificity 0.642), 0.648 for PLR ($P \leq 0.0001$, cut-off 128.575, sensitivity 0.679, specificity 0.565), 0.697 for MLR ($P \leq 0.0001$, cut-off 0.271, sensitivity 0.733, specificity 0.607), 0.714 for PNI ($P \leq 0.0001$, cut-off 47.225, sensitivity 0.774, specificity 0.595), and 0.813 for FLR ($P \leq 0.0001$, cut-off 1.631, sensitivity 0.747, specificity 0.743). The AUC was 0.914 for a combination of CA199 + FAR, 0.915 for CA199 + FPR, 0.917 for CA199 + FAR + FPR, and 0.924 for CA199 + FAR + FPR + FLR. We calculated the ROC curves and AUC for the testing set 1 and testing set 2 using the best cut-off value from the ROC curve in the training set. In testing set 1, the AUC was 0.941 for a combination of CA199 + FAR + FPR, 0.947 for CA199 + FAR + FPR + FLR, 0.975 for CA199 + FAR + FPR + MLR + PNI to distinguish patients with PC from HC; 0.925 for CA199 + FAR + FPR, and 0.942 for CA199 + FAR + FPR + FLR to differentiate patients with PC from those with OPT. The results revealed that combinations of CA199 and inflammation indicator values had a strong capability for differentiating patients with PC from the HC and OPT groups, especially the combination of CA199 + FAR + FPR + FLR (Fig. 3E-H, and supplementary Tables 1 and 2). In testing set 2, the AUC was 0.993 for combination of CA199 + FAR + FPR + FLR to distinguish patients with PC from HC, and 0.994 for combination of CA199 + FAR + FPR + FLR to differentiate patients with PC from those with OPT (supplementary Fig. 3).Fig. 3Diagnostic value of single and combined inflammation markers ratios in early-stage PC. A The ROC curve analysis of FAR, FPR, NLR, PLR, MLR, FLR, PNI, and CA199 between PC and HC in the training set. B The ROC curve analysis of combined inflammation markers ratios and CA199 in PC and HC in the training set. C The ROC curve analysis of FAR, FPR, NLR, PLR, MLR, FLR, PNI, and CA199 between PC and OPT in the training set. D The ROC curve analysis of combined inflammation indicator and CA199 between PC and OPT in the training set. E The ROC curve analysis of FAR, FPR, MLR, FLR, PNI, and CA199 between PC and HC in testing set 1. F The ROC curve analysis of combined inflammation markers ratios and CA199 between PC and HC in testing set 1. G The ROC curve analysis of FAR, FPR, FLR, and CA199 between PC and OPT in testing set 1. H The ROC curve analysis of combined inflammation markers ratios and CA199 between PC and OPT in testing set 1. Abbreviations: PC, pancreatic cancer; OPT, other pancreas tumors; HC, healthy controls; FAR, fibrinogen-to-albumin ratio; FPR, fibrinogen-to-prealbumin ratio; NLR, neutrophil-to-lymphocyte ratio; PLR, platelets-to-lymphocyte ratio; MLR monocytes-to-lymphocyte ratio; PNI, albumin + 5 × the lymphocyte count; FLR, fibrinogen-to- lymphocyte ratio; ROC, receiver operating characteristicTable 2ROC curve results based on FAR, FPR, NLR, PLR, LMR, PNI, FLR, and CA199 for distinguishing PC patients from HC in training setMarkerAUC ($95\%$CI)P—valuecut-offSensitivitySpecificityFAR0.776(0.740–0.811)< 0.00010.0800.8850.556FPR0.869(0.842–0.896)< 0.00010.0120.8370.775NLR0.686(0.645–0.726)< 0.00012.2520.7800.527PLR0.584(0.540–0.628)0.0002177.2180.8790.299MLR0.818(0.786–0.850)< 0.00010.2490.8310.678PNI0.860(0.831–0.888)< 0.000149.0250.9070.707FLR0.748(0.711–0.785)< 0.00011.8640.7730.639CA1990.868(0.836–0.901)< 0.000124.5400.9390.817CA199 + FAR0.942(0.924–0.960)< 0.0001-0.4590.9710.814CA199 + FPR0.964(0.951–0.977)< 0.0001-0.4700.9780.840CA199 + MLR0.940(0.921–0.960)< 0.00010.3710.9460.870CA199 + PNI0.955(0.939–0.971)< 0.00010.0020.9490.870CA199 + FLR0.917(0.893–0.940)< 0.0001-0.0350.9650.793CA199 + FAR + FPR0.964(0.951–0.978)< 0.0001-0.2780.9650.858CA199 + FAR + FPR + FLR0.964(0.951–0.978)< 0.0001-0.4660.9740.849CA199 + FAR + FPR + MLR + PNI0.976(0.965–0.988)< 0.0001-0.2940.9740.891Abbreviations: PC Pancreatic cancer, HC Heathy controls, ROC Receiver operating characteristic, AUC Area under the receiver operating characteristic curve, CI Confidence interval, FPR Fibrinogen-to-prealbumin ratio, FAR Fibrinogen-to-albumin ratio, NLR Neutrophil-to-lymphocyte ratio, PLR Platelets-to-lymphocyte ratio, MLR Monocytes-to-lymphocyte ratio, PNI Albumin + 5 × the lymphocyte count, FLR Fibrinogen-to-lymphocyte ratioTable 3ROC curve results based on FAR, FPR, NLR, PLR, LMR, PNI, FLR, and CA199 for distinguish PC patients from OPT in testing set 1MarkerAUC ($95\%$CI)P—valuecut-offSensitivitySpecificityFAR0.778(0.740–0.817)< 0.00010.0700.7010.734FPR0.779(0.740–0.817)< 0.00010.0130.7780.666NLR0.716(0.674–0.759)< 0.00011.9610.6740.642PLR0.648(0.603–0.694)< 0.0001128.5750.6790.565MLR0.697(0.654–0.741)< 0.00010.2710.7330.607PNI0.714(0.671–0.757)< 0.000147.2250.7740.595FLR0.813(0.778–0.848)< 0.00011.6310.7470.743CA1990.846(0.812–0.880)< 0.000132.2050.8870.772CA199 + FAR0.914(0.891–0.937)< 0.0001-0.1820.8960.799CA199 + FPR0.915(0.891–0.938)< 0.00010.1010.8600.831CA199 + FLR0.915(0.829–0.938)< 0.0001-0.5230.9460.778CA199 + FAR + FPR0.917(0.895–0.940)< 0.0001-0.1250.8870.799CA199 + FAR + FPR + FLR0.924(0.903–0.946)< 0.0001-0.5060.9410.799Abbreviations: PC Pancreatic cancer, OPT Other pancreas tumors, ROC Receiver operating characteristic, AUC Area under the receiver operating characteristic curve, CI Confidence interval, FPR Fibrinogen-to-prealbumin ratio, FAR Fibrinogen-to-albumin ratio, NLR Neutrophil-to-lymphocyte ratio, PLR Platelets-to-lymphocyte ratio, MLR Monocytes-to-lymphocyte ratio, PNI Albumin + 5 × the lymphocyte count, FLR Fibrinogen-to- lymphocyte ratio
## Relationship between inflammation markers ratios values and clinical characteristics of patients with PC
The relationship between inflammation markers ratios and the clinical characteristics of patients with PC was analyzed. In the training set, patients with pancreatic head cancer had higher FAR, FPR, NLR, PLR, MLR, FLR, and lower PNI values than patients with pancreatic body or tail cancers (Fig. 4A-G; $P \leq 0.001$, $P \leq 0.001$, $P \leq 0.001$, $P \leq 0.001$, $P \leq 0.001$, and $P \leq 0.001$, respectively). Patients aged > 60 years had higher FAR values than those aged ≤ 60 years (Fig. 4A; $$P \leq 0.037$$). Male patients with PC had higher MLR values than female patients with PC (Fig. 4E, $$P \leq 0.011$$). In the testing set, the same trend was observed for the FAR, FPR, NLR, FLR, and PNI values (Supplementary Fig. 4A-C, F-G; $P \leq 0.001$, $$P \leq 0.007$$, $$P \leq 0.04$$, $P \leq 0.05$, and $$P \leq 0.004$$, respectively). Similarly, patients who were > 60 years of age had higher FAR, FPR, MLR, FLR, and lower PNI values than those aged ≤ 60 years (Supplementary Fig. 4A-B, E-F; $$P \leq 0.017$$, $$P \leq 0.007$$, $$P \leq 0.013$$, $P \leq 0.05$, and $$P \leq 0.02$$, respectively). In the two groups, the inflammation markers ratios values were independent of tumor size, differentiation, lymph nodes, TNM stage and sex ($P \leq 0.05$ in all inflammation markers ratios values).Fig. 4Comparison of inflammation markers ratios in different clinical characteristics early-stage PC. The FAR(A), FPR (B), NLR (C), PLR (D), MLR (E), FLR (F), and PNI (G), and in different tumor location, tumor size, differentiation, lymph nodes, stage, sex, and age in the training set. Abbreviations: PC, pancreatic cancer; FAR, fibrinogen-to-albumin ratio; FPR, fibrinogen-to-prealbumin ratio; NLR, neutrophil-to-lymphocyte ratio; PLR, platelets-to-lymphocyte ratio; MLR monocytes-to-lymphocyte ratio; PNI, albumin + 5 × the lymphocyte count; FLR, fibrinogen-to- lymphocyte ratio; Blue column: location, pancreatic head/tumor size ≤ 4 cm/ differentiation well / lymph node metastasis no /stage I/sex male/age ≤ 60; Red column: location, pancreatic body and tail/tumor size > 4 cm/ differentiation poor / lymph node metastasis yes /stage II/sex male/age > 60
## Differential diagnosis power of inflammation markers ratios values in different tumors location of PC
Patients with PC and OPT were classified into four subgroups according to the locations of the pancreatic lesions: pancreatic head cancer (PHC), pancreatic body and tail cancer (PBTC), other pancreatic head tumors (OPHT), and other pancreatic body and tail tumors (OPBTT). The AUC was 0.855 for CA199, 0.750 for FAR, 0.751 for FPR, 0.824 for FLR, 0.767 for NLR, 0.686 for PLR, 0.766 for MLR, and 0.709 for PNI, to differentiate between patients with PHC and OPHT (Fig. 5A). The AUC was 0.834 for FAR + FPR + FLR and 0.915 for CA199 + FAR + FPR + FLR, to differentiate between patients with PHC and those with OPHT (Fig. 5B). The AUC was 0.838 for CA199, 0.706 for FAR, 0.693 for FPR, 0.660 for FLR, 0.585 for NLR, 0.576 for PLR, 0.529 for MLR, and 0.601 for PNI, to differentiate between patients with PHC and OPHT (Fig. 5C). The AUC was 0.714 for FAR + FPR + FLR and 0.894 for CA199 + FAR + FPR + FLR, to differentiate between patients with PBTC and OPBTT (Fig. 5D). These results showed that a combination of CA199 + FAR + FPR + FLR could better help identify PHC and OPHT.Fig. 5Diagnostic value of single and combined inflammation markers ratios in different tumor location PC. A The ROC curve analysis of FAR, FPR, NLR, PLR, MLR, PNI, FLR, and CA199 between PHC and OPHT in the training sets. B The ROC curve analysis of combined inflammation markers ratios and CA199 between PHC and OPHT in the training sets. C The ROC curve analysis of FAR, FPR, NLR, PLR, MLR, PNI, FLR, and CA199 between PBTC and OPBTT in the training sets. D The ROC curve analysis of combined inflammation markers ratios and CA199 between PBTC and OPBTT in the training sets. Abbreviations: PHC, pancreatic head cancer; OPHT, other pancreas head tumors; PBTC, pancreatic body, and tail cancer; OPBTT, other pancreas body, and tail tumors; FAR, fibrinogen-to-albumin ratio; FPR, fibrinogen-to-prealbumin ratio; NLR, neutrophil-to-lymphocyte ratio; PLR, platelets-to-lymphocyte ratio; MLR monocytes-to-lymphocyte ratio; PNI, albumin + 5 × the lymphocyte count; FLR, fibrinogen-to- lymphocyte ratio; ROC, receiver operating characteristic
## Discussion
Cancer-associated inflammation has been reported as the seventh hallmark of cancer [9]. Nearly all human cancers harbor inflammatory reactions, which play an important role in tumor development, progression, and metastasis [18]. Systemic inflammation can play a supporting role in the evolution of PC. For example, chronic pancreatitis is a known risk factor for the development of PC [19]. Obesity, another risk factor for pancreatic cancer, can induce inflammation by promoting the release of IL-6, CCL2, and CCL5, and the infiltration of macrophages and immunosuppressive cells [20]. Smoking is an established risk factor for PC and can induce inflammation and immune activation [21]. In addition, cancer cells can promote systemic inflammation that can, in turn, support tumor growth and lead to a poor prognosis in PC [18]. Inflammatory cells and chemokines shape the inflammatory microenvironment leading to cancer [22]. For example, IL-6, IL-1β, and TNF-α were increased at cancer early stage and associated with disease severity [23]. In Pancreatic ductal adenocarcinoma (PDAC) mouse models, adipocyte-secreted IL-1β could promote obesity-induced pancreatic carcinogenesis and drug resistance through recruitment of tumor-associated neutrophils [24]. High intra-tumoral and serum IL-1β levels in PC patients correlate with poor overall survival and increased chemoresistance [25]. IL-6, a pro-inflammatory cytokine that shows diverse functions of cell multiplication, injury, infection, and inflammation, affects tumor cells to develop PC by controlling vascular endothelial growth factor (VEGF) secretion [26]. IL-8 which derived from macrophages, platelets, and epithelial cells could promote the angiogenesis of PC. Serum levels of IL-6, IL-8, IL-10, and IL-1RA were significantly increased in pancreatic cancer patients and were associated with worse survival rates, poor performance status. A panel of IP-10, IL-6, PDGF plus CA19-9 could discriminate PDAC patients from patients with pancreatic benign disease [27]. TNF-α is associated with acute and chronic inflammation and inflammation related to cancers [28]. In addition, increased expression of tumor-related inflammatory mediators and cytokines, such as TNF-α, IL-1, and IL-6 may stimulate the bone marrow to release neutrophils, resulting in an increase in the circulating neutrophil count and decrease in the circulating lymphocytopenia [29]. Neutrophils could promote growth and metastasis of tumors through secreting a variety of cytokines, including matrix metalloproteinase-9, chemokines and vascular endothelial growth factor (VEGF). It was reported that neutrophils could promote adhesion between circulating tumor cells and distant target organs through acting as an adhesive adapter, finally increasing the chance of distant metastasis. Moreover, neutrophil could also inhibit the antitumor immune function of natural killer cells and cytotoxic T cells [12]. Presently, it is believed that lymphocytes in the peripheral blood can cause synergistic cytotoxicity and play an anti-cancer role. Several subtypes of tumor infiltrating lymphocyte such as CD8 + T cells and memory T cells were associated with better outcomes of a variety of tumors, while regulatory T cells and Th17 cells were associated with progression and unfavorable prognosis of tumors [30]. Although different subset of T cells was associated with adverse prognosis of tumors, high level of absolute lymphocyte count was demonstrated to be associated with favorable prognosis of gastric cancer patients [31]. A study by Dominic et al. showed that inflammatory monocytes were lower in the bone marrow and higher in the blood of patients with resectable PC, and an increased blood-to-bone marrow monocyte ratio was a novel poor prognostic factor for PC [32]. Platelets are also involved in tumor development [33].
Meanwhile, patient’s nutritional status is associated with metabolic changes and immune status impairment. Circulating albumin and prealbumin are markers for evaluating nutritional status and immune status. Albumin can inhibit tumor progression by stabilizing DNA replication and enhancing the immune response [34]. The inflammatory factors may influence nutritional status through inhibition of appetite, alteration of gastrointestinal function, alteration of the carbohydrate metabolism and insulin resistance. Serum levels of IL-6 and IL-8 were inversely correlated to serum albumin and prealbumin. Serum IL-6 and IL-8 were highly expressed in patients with nutritional risk [35]. Genetic and pharmacological studies have revealed the key role of fibrinogen in determining the degree of local or systemic inflammation [36]. Fibrinogen is an important coagulation factor that can be recognized by a variety of integrin and non-integrin receptors on tumor, stromal, and inflammatory cells. These fibrinogen-mediated receptors are thought to control cell proliferation, apoptosis, cell migration, and the expression of inflammatory mediators [37].
In cancer, cytokines mediate signalling between cancer cells, and the cells of the TME, including PSCs, CAFs, endothelial cells, and a range of immune cells including macrophages, mast cells, neutrophils, and regulatory T-cells [38]. For example, glioblastoma (GBM) cells reduced lymphocyte infiltration by secreting immunosuppressive cytokines such as IL-10, IL-2, and TGF-β, and recruited and induced macrophages to become M2 phenotypes by secreting IL-10, IL-4, IL-6, macrophage–colony stimulating factor (M-CSF), TGF-β, and prostaglandin E2 (PGE2) [39]. Higher serum IL-8 and IL-6 levels were positively correlated with high NLR, modified glasgow prognostic score (mGPS), CRP-albumin ratio (CAR), and PLR [40, 41]. Fibrinogen induced the production of IL-6, IL-8, monocyte chemoattractant protein-1, vascular endothelial growth factor, angiopoietin-1 and type I collagen in pancreatic stellate cells [42]. CAR, NLR, and PNI were positively associated with IL-10, IL-23, and IL-1β [43]. Park et al. found moderate-to-strong correlations within circulating cytokines (TNF-α, IL-1β, IL-6, IL-8, IL-9, IL-10, and VEGF) as well as within systemic inflammatory markers (mGPS, NLR, and PNI) [44]. Higher mGPS was involved in increased plasma levels of IL-4, IL-6, IL-8 [45]. Patients with a low PNI exhibited high levels of TNF-αin advanced pancreatic cancer [46]. To sum up, there was a close relationship between systemic inflammatory markers and plasma cytokines.
Currently, routine measurement of serum inflammatory cytokines is not common in daily clinical practice. Many studies used inflammatory cell in the peripheral blood to reflect the systemic immune conditions of patients. In this study, we included HCs and patients with chronic pancreatitis, pancreatic serous/mucinous cystadenoma, solid pseudo-papilloma, and pancreatic neuroendocrine tumors. The results showed that serum albumin, prealbumin, and lymphocytes were significantly decreased, while fibrinogen, neutrophils, and monocytes were significantly increased in early-stage PC compared with HC and OPT. Our results provide supporting evidence that inflammation is emerging as a hallmark of early—stage cancer. Since neutrophil, monocyte, and lymphocyte counts are influenced by many factors, researchers are more inclined to use the ratio values between the two inflammation markers to explore the relationship between the ratio values and malignant tumor prognosis.
To date, many studies have shown that FAR, FPR, NLR, PLR, MLR, and PNI are predictive of outcomes in various types of cancer. For example, Michael et al. [ 47] found that an increased lymphocyte-to-monocyte ratio (LMR) was an independent prognostic factor for better cancer-specific survival in patients with PC (HR 0.70; $P \leq 0.001$). Qi et al. [ 36] showed that NLR, PLR, and LMR were independent predictors of survival in patients with advanced PC. Yi et al. [ 46] showed that a low PNI was associated with a systemic inflammatory response and was an independent poor prognostic factor for advanced PC. Fang et al. [ 16] reported that a high FAR was associated with poor OS in patients with locally advanced or metastatic PC. Xie et al. [ 48] found that high FPR was an independent poor prognostic factor for patients with stage I-III colorectal cancer (CRC). In addition, inflammatory indicators have important implications in cancer diagnosis. The combination of NLR, PLR, and CEA had a high diagnostic efficacy (AUC = 0.831, $95\%$ CI = 0.807–0.852) for early-stage CRC. Zheng et al. [ 39] found that NLR + LMR and the derived neutrophil-to-lymphocyte ratio (dNLR) + LMR had good diagnostic performance in patients with glioma (AUC = 0.777 and 0.778, respectively). Wu et al. [ 49] showed that a combination of PLR and CEA had a better AUC of 0.780 than CEA alone for diagnosing gastric cancer. Lu et al. [ 50] found that the combination of CA199 and AFR distinguished PC from HC with an ROC of 0.932. Liu et al. [ 51] showed that combined circulating dNLR and Alb was an effective diagnostic biomarker for early stage PC (AUC = 0.931), and that dNLR could distinguish early-stage PC from HC (AUC 0.895) and from additional cancers (AUC 0.794). Similar to the above results, this study found that FAR, FPR, NLR, PLR, MLR, and FLR were higher in early-stage PC than in HC and OPT, whereas PNI was lower in patients with early-stage PC. These results indicate that inflammatory indicators could act as early diagnostic markers for PC. Moreover, ROC analysis indicated that the FAR, FPR, PLR, MLR, and PNI were promising diagnostic indicators. Among these inflammation markers, a combination of FAR, FPR, FLR, and CA199 could be used to differentiate early-stage PC from HC and OPT with a better AUC (0.964 and 0.924 in training sets). The results obtained in the training set were confirmed for two independent testing sets.
The inflammation indicators were similar over differences in sex, age, tumor size, differentiation, lymph nodes, and TNM stage, but varied greatly for different tumor locations. PHC always obstructs bile ducts, which leads to the levels of albumin (38.2 g/L vs 39.6 g/L, $$P \leq 0.007$$) and prealbumin (207.7 g/L vs 240.1 g/L, $$P \leq 0.0003$$) that are lower than PBTC. We further explored the discriminating ability of inflammation indicators for different tumor locations in early-stage PC. The combination of CA199, FAR, FPR, and FLR could better distinguish PHC from OPHT (AUC = 0.915) than PBTC from OPBTT (AUC = 0.894). Hence, for patients with pancreatic head tumors at the first medical visit, a combination of FAR, FPR, FLR, and CA199 would significantly guide the initial clinical diagnosis and aid in a more accurate final diagnosis.
Our study had some limitations. First, it was a retrospective analysis of data from a clinical trial and lacked prospective data. Second, although all patients were from two single-centers, we enrolled only a small number of patients from one center. Third, the participants in our study had no measurements of serum inflammatory cytokines such as IL-2, IL-6 and so on, we have no way to compared correlations between cytokine levels and inflammation markers ratios. However, despite several limitations, this study confirmed that FAR, FPR, FLR, and CA199 have a potential as diagnostic markers for early-stage PC. These results need to be confirmed in a multicenter, large-scale, prospective study.
## Conclusion
This study established that circulating inflammation markers ratios, especially FAR, FPR, and FLR, could be used as cost-effective diagnostic biomarkers for early-stage PC that improve the diagnostic accuracy over CA199. The combination of FAR, FPR, FLR, and CA199 was a potentially effective biomarker for distinguishing early -stage PC patients from HC and in differentiating early -stage PC patients from patients with OPT. The combination of FAR, FPR, FLR, and CA199 may be useful as a differential diagnostic marker for patients with pancreatic head cancer.
## Supplementary Information
Additional file 1: Supplementary Figure 1. The circulating inflammation markers in PC, HC, BPT, SPT, and PNET in testing set 1.Additional file 2: Supplementary Figure 2. The inflammation markers ratios in PC, HC, BPT, SPT, and PNET in testing set 1.Additional file 3: Supplementary Figure 3. Diagnostic value of single and combined inflammation markers ratios in early-stage PC.Additional file 4: Supplementary Figure 4. Comparison of inflammation markers ratios in different clinical characteristics early-stage PC.Additional file 5: Supplementary Table 1. ROC curve results based on FAR, FPR, MLR, PNI, FLR and CA199 for distinguishing PC patients from HC in testing set 1.Additional file 6: Supplementary Table 2. ROC curve results based on FAR, FPR, FLR, and CA199 for distinguish PC patients from OPT in Testing set 1.Additional file 7: Supplementary Table 3. ROC curve results based onFAR, FPR, NLR, PLR, LMR, PNI, FLR and CA199 for distinguishing PC from OPT in training sets 1.
## 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* (2018.0) **68** 394-424. DOI: 10.3322/caac.21492
2. Ferlay J, Partensky C, Bray F. **More deaths from pancreatic cancer than breast cancer in the EU by 2017**. *Acta Oncol* (2016.0) **55** 1158-1160. DOI: 10.1080/0284186X.2016.1197419
3. Murphy JE, Wo JY, Ryan DP, Jiang W, Yeap BY, Drapek LC, Blaszkowsky LS, Kwak EL, Allen JN, Clark JW. **Total Neoadjuvant therapy with FOLFIRINOX followed by individualized Chemoradiotherapy for borderline Resectable pancreatic adenocarcinoma: a phase 2 clinical trial**. *JAMA Oncol* (2018.0) **4** 963-969. DOI: 10.1001/jamaoncol.2018.0329
4. Abe T, Blackford AL, Tamura K, Ford M, McCormick P, Chuidian M, Almario JA, Borges M, Lennon AM, Shin EJ. **Deleterious germline mutations are a risk factor for neoplastic progression among high-risk individuals undergoing pancreatic surveillance**. *J Clin Oncol* (2019.0) **37** 1070-1080. DOI: 10.1200/JCO.18.01512
5. Wang R, Zhao Y, Wang Y, Zhao Z, Chen Q, Duan Y, Xiong S, Luan Z, Wang J, Cheng B. **Diagnostic and prognostic values of KRAS mutations on EUS-FNA specimens and circulating tumor DNA in patients with pancreatic cancer**. *Clin Transl Gastroenterol* (2022.0) **13** e00487. DOI: 10.14309/ctg.0000000000000487
6. Cheng H, He W, Yang J, Ye Q, Cheng L, Pan Y, Mao L, Chu X, Lu C, Li G. **Ligand-targeted polymerase chain reaction for the detection of folate receptor-positive circulating tumour cells as a potential diagnostic biomarker for pancreatic cancer**. *Cell Prolif* (2020.0) **53** e12880. DOI: 10.1111/cpr.12880
7. Wang L, Wu J, Ye N, Li F, Zhan H, Chen S, Xu J. **Plasma-derived exosome MiR-19b acts as a diagnostic marker for pancreatic cancer**. *Front Oncol* (2021.0) **11** 739111. DOI: 10.3389/fonc.2021.739111
8. Melo SA, Luecke LB, Kahlert C, Fernandez AF, Gammon ST, Kaye J, LeBleu VS, Mittendorf EA, Weitz J, Rahbari N. **Glypican-1 identifies cancer exosomes and detects early pancreatic cancer**. *Nature* (2015.0) **523** 177-182. DOI: 10.1038/nature14581
9. Hanahan D, Weinberg RA. **Hallmarks of cancer: the next generation**. *Cell* (2011.0) **144** 646-674. DOI: 10.1016/j.cell.2011.02.013
10. Cook MB, Barnett MJ, Bock CH, Cross AJ, Goodman PJ, Goodman GE, Haiman CA, Khaw KT, McCullough ML, Newton CC. **Prediagnostic circulating markers of inflammation and risk of oesophageal adenocarcinoma: a study within the National Cancer Institute Cohort Consortium**. *Gut* (2019.0) **68** 960-968. DOI: 10.1136/gutjnl-2018-316678
11. Coffelt SB, Wellenstein MD, de Visser KE. **Neutrophils in cancer: neutral no more**. *Nat Rev Cancer* (2016.0) **16** 431-446. DOI: 10.1038/nrc.2016.52
12. Huang SH, Waldron JN, Milosevic M, Shen X, Ringash J, Su J, Tong L, Perez-Ordonez B, Weinreb I, Bayley AJ. **Prognostic value of pretreatment circulating neutrophils, monocytes, and lymphocytes in oropharyngeal cancer stratified by human papillomavirus status**. *Cancer* (2015.0) **121** 545-555. DOI: 10.1002/cncr.29100
13. Perisanidis C, Psyrri A, Cohen EE, Engelmann J, Heinze G, Perisanidis B, Stift A, Filipits M, Kornek G, Nkenke E. **Prognostic role of pretreatment plasma fibrinogen in patients with solid tumors: a systematic review and meta-analysis**. *Cancer Treat Rev* (2015.0) **41** 960-970. DOI: 10.1016/j.ctrv.2015.10.002
14. McMillan DC. **The systemic inflammation-based Glasgow Prognostic Score: a decade of experience in patients with cancer**. *Cancer Treat Rev* (2013.0) **39** 534-540. DOI: 10.1016/j.ctrv.2012.08.003
15. Templeton AJ, McNamara MG, Seruga B, Vera-Badillo FE, Aneja P, Ocana A, Leibowitz-Amit R, Sonpavde G, Knox JJ, Tran B. **Prognostic role of neutrophil-to-lymphocyte ratio in solid tumors: a systematic review and meta-analysis**. *J Natl Cancer Inst* (2014.0) **106** dju124. DOI: 10.1093/jnci/dju124
16. Fang L, Yan FH, Liu C, Chen J, Wang D, Zhang CH, Lou CJ, Lian J, Yao Y, Wang BJ. **Systemic inflammatory biomarkers, especially fibrinogen to albumin ratio, predict prognosis in patients with pancreatic cancer**. *Cancer Res Treat* (2021.0) **53** 131-139. DOI: 10.4143/crt.2020.330
17. Qian BZ, Pollard JW. **Macrophage diversity enhances tumor progression and metastasis**. *Cell* (2010.0) **141** 39-51. DOI: 10.1016/j.cell.2010.03.014
18. Stone ML, Beatty GL. **Cellular determinants and therapeutic implications of inflammation in pancreatic cancer**. *Pharmacol Ther* (2019.0) **201** 202-213. DOI: 10.1016/j.pharmthera.2019.05.012
19. Kirkegard J, Mortensen FV, Cronin-Fenton D. **Chronic pancreatitis and pancreatic cancer risk: a systematic review and meta-analysis**. *Am J Gastroenterol* (2017.0) **112** 1366-1372. DOI: 10.1038/ajg.2017.218
20. Kolb R, Sutterwala FS, Zhang W. **Obesity and cancer: inflammation bridges the two**. *Curr Opin Pharmacol* (2016.0) **29** 77-89. DOI: 10.1016/j.coph.2016.07.005
21. Fan X, Alekseyenko AV, Wu J, Peters BA, Jacobs EJ, Gapstur SM, Purdue MP, Abnet CC, Stolzenberg-Solomon R, Miller G. **Human oral microbiome and prospective risk for pancreatic cancer: a population-based nested case-control study**. *Gut* (2018.0) **67** 120-127. DOI: 10.1136/gutjnl-2016-312580
22. Farajzadeh Valilou S, Keshavarz-Fathi M, Silvestris N, Argentiero A, Rezaei N. **The role of inflammatory cytokines and tumor associated macrophages (TAMs) in microenvironment of pancreatic cancer**. *Cytokine Growth Factor Rev* (2018.0) **39** 46-61. DOI: 10.1016/j.cytogfr.2018.01.007
23. Gukovsky I, Li N, Todoric J, Gukovskaya A, Karin M. **Inflammation, autophagy, and obesity: common features in the pathogenesis of pancreatitis and pancreatic cancer**. *Gastroenterology* (2013.0) **144** 1199-1209 e1194. DOI: 10.1053/j.gastro.2013.02.007
24. Incio J, Liu H, Suboj P, Chin SM, Chen IX, Pinter M, Ng MR, Nia HT, Grahovac J, Kao S. **Obesity-induced inflammation and Desmoplasia promote pancreatic cancer progression and resistance to chemotherapy**. *Cancer Discov* (2016.0) **6** 852-869. DOI: 10.1158/2159-8290.CD-15-1177
25. Das S, Shapiro B, Vucic EA, Vogt S, Bar-Sagi D. **Tumor cell-derived IL1beta promotes Desmoplasia and immune suppression in pancreatic cancer**. *Cancer Res* (2020.0) **80** 1088-1101. DOI: 10.1158/0008-5472.CAN-19-2080
26. 26.Lee HM, Lee HJ, Chang JE. Inflammatory Cytokine: an attractive target for cancer treatment. Biomedicines. 2022;10(9):2116.
27. Shaw VE, Lane B, Jenkinson C, Cox T, Greenhalf W, Halloran CM, Tang J, Sutton R, Neoptolemos JP, Costello E. **Serum cytokine biomarker panels for discriminating pancreatic cancer from benign pancreatic disease**. *Mol Cancer* (2014.0) **13** 114. DOI: 10.1186/1476-4598-13-114
28. Wang Q, Tu H, Zhu M, Liang D, Ye Y, Chang DW, Long Y, Wu X. **Circulating obesity-driven biomarkers are associated with risk of clear cell renal cell carcinoma: a two-stage, case-control study**. *Carcinogenesis* (2019.0) **40** 1191-1197. DOI: 10.1093/carcin/bgz074
29. Wang SB, Chen JY, Xu C, Cao WG, Cai R, Cao L, Cai G. **Evaluation of systemic inflammatory and nutritional indexes in locally advanced gastric cancer treated with adjuvant chemoradiotherapy after D2 dissection**. *Front Oncol* (2022.0) **12** 1040495. DOI: 10.3389/fonc.2022.1040495
30. Ogiya R, Niikura N, Kumaki N, Bianchini G, Kitano S, Iwamoto T, Hayashi N, Yokoyama K, Oshitanai R, Terao M. **Comparison of tumor-infiltrating lymphocytes between primary and metastatic tumors in breast cancer patients**. *Cancer Sci* (2016.0) **107** 1730-1735. DOI: 10.1111/cas.13101
31. Feng F, Zheng G, Wang Q, Liu S, Liu Z, Xu G, Wang F, Guo M, Lian X, Zhang H. **Low lymphocyte count and high monocyte count predicts poor prognosis of gastric cancer**. *BMC Gastroenterol* (2018.0) **18** 148. DOI: 10.1186/s12876-018-0877-9
32. Sanford DE, Belt BA, Panni RZ, Mayer A, Deshpande AD, Carpenter D, Mitchem JB, Plambeck-Suess SM, Worley LA, Goetz BD. **Inflammatory monocyte mobilization decreases patient survival in pancreatic cancer: a role for targeting the CCL2/CCR2 axis**. *Clin Cancer Res* (2013.0) **19** 3404-3415. DOI: 10.1158/1078-0432.CCR-13-0525
33. Franco AT, Corken A, Ware J. **Platelets at the interface of thrombosis, inflammation, and cancer**. *Blood* (2015.0) **126** 582-588. DOI: 10.1182/blood-2014-08-531582
34. Sun DW, An L, Lv GY. **Albumin-fibrinogen ratio and fibrinogen-prealbumin ratio as promising prognostic markers for cancers: an updated meta-analysis**. *World J Surg Oncol* (2020.0) **18** 9. DOI: 10.1186/s12957-020-1786-2
35. Zhang M, Guo Y, Zhang X, Zhang Y, Fan Y, Liu Y, Zhao K, Fan R, Lu P. **Nutritional risk screening in malignant tumors: a study of 375 cancer inpatients**. *Support Care Cancer* (2022.0) **30** 8333-8337. DOI: 10.1007/s00520-022-07242-9
36. Davalos D, Akassoglou K. **Fibrinogen as a key regulator of inflammation in disease**. *Semin Immunopathol* (2012.0) **34** 43-62. DOI: 10.1007/s00281-011-0290-8
37. Zhang LP, Ren H, Du YX, Wang CF. **Prognostic value of the preoperative fibrinogen-to-albumin ratio in pancreatic ductal adenocarcinoma patients undergoing R0 resection**. *World J Gastroenterol* (2020.0) **26** 7382-7404. DOI: 10.3748/wjg.v26.i46.7382
38. van Duijneveldt G, Griffin MDW, Putoczki TL. **Emerging roles for the IL-6 family of cytokines in pancreatic cancer**. *Clin Sci (Lond)* (2020.0) **134** 2091-2115. DOI: 10.1042/CS20191211
39. Zheng SH, Huang JL, Chen M, Wang BL, Ou QS, Huang SY. **Diagnostic value of preoperative inflammatory markers in patients with glioma: a multicenter cohort study**. *J Neurosurg* (2018.0) **129** 583-592. DOI: 10.3171/2017.3.JNS161648
40. Rodrigues ISS, Martins-Filho A, Micheli DC, Lima CA, Tavares-Murta BM, Murta EFC, Nomelini RS. **IL-6 and IL-8 as prognostic factors in peritoneal fluid of ovarian cancer**. *Immunol Invest* (2020.0) **49** 510-521. DOI: 10.1080/08820139.2019.1691222
41. Kim YW, Kim SK, Kim CS, Kim IY, Cho MY, Kim NK. **Association of serum and intratumoral cytokine profiles with tumor stage and neutrophil lymphocyte ratio in colorectal cancer**. *Anticancer Res* (2014.0) **34** 3481-3487. PMID: 24982357
42. Masamune A, Kikuta K, Watanabe T, Satoh K, Hirota M, Hamada S, Shimosegawa T. **Fibrinogen induces cytokine and collagen production in pancreatic stellate cells**. *Gut* (2009.0) **58** 550-559. DOI: 10.1136/gut.2008.154401
43. Yu YL, Tseng WK, Fan CW, Chang PH, Kuo HC, Pan YP, Yeh KY. **Pretreatment nutrition-inflammation biomarkers correlated with differential cytokine profiles in Taiwanese patients with colorectal cancer**. *Nutr Cancer* (2022.0) **74** 1614-1624. DOI: 10.1080/01635581.2021.1957130
44. Park JW, Chang HJ, Yeo HY, Han N, Kim BC, Kong SY, Kim J, Oh JH. **The relationships between systemic cytokine profiles and inflammatory markers in colorectal cancer and the prognostic significance of these parameters**. *Br J Cancer* (2020.0) **123** 610-618. DOI: 10.1038/s41416-020-0924-5
45. Zou Z, Zhao L, Su S, Liu Q, Yu L, Wei J, Yang Y, Du J, Shen J, Qian X. **The plasma levels of 12 cytokines and growth factors in patients with gastric cancer**. *Medicine (Baltimore)* (2018.0) **97** e0413. DOI: 10.1097/MD.0000000000010413
46. Geng Y, Qi Q, Sun M, Chen H, Wang P, Chen Z. **Prognostic nutritional index predicts survival and correlates with systemic inflammatory response in advanced pancreatic cancer**. *Eur J Surg Oncol* (2015.0) **41** 1508-1514. DOI: 10.1016/j.ejso.2015.07.022
47. Stotz M, Szkandera J, Stojakovic T, Seidel J, Samonigg H, Kornprat P, Schaberl-Moser R, Seggewies F, Hoefler G, Gerger A. **The lymphocyte to monocyte ratio in peripheral blood represents a novel prognostic marker in patients with pancreatic cancer**. *Clin Chem Lab Med* (2015.0) **53** 499-506. DOI: 10.1515/cclm-2014-0447
48. Hailun X, Huang S, Yuan G, Tang S, Gan J. **Prognostic significance of preoperative fibrinogen-to-Prealbumin ratio in patients with stage I-III colorectal cancer undergoing surgical resection: a retrospective cohort study**. *Biomed Res Int* (2021.0) **2021** 3905353. PMID: 33521127
49. Wu Y, Jiang M, Qin Y, Lin F, Lai M. **Single and combined use of neutrophil-lymphocyte ratio, platelet-lymphocyte ratio and carcinoembryonic antigen in diagnosing gastric cancer**. *Clin Chim Acta* (2018.0) **481** 20-24. DOI: 10.1016/j.cca.2018.02.027
50. 50.Lu Y, Yu H, Qin S, Chen M, Chen H, Hu Z, Li S, Qin X. Single and combined use of preoperative in flammatory biomarkers and CA199 in diagnosing pancreatic cancer. Clin Lab. 2020;66(11):2197–204.
51. Liu JX, Li A, Zhou LY, Liu XF, Wei ZH, Wang XZ, Ying HQ. **Significance of combined preoperative serum Alb and dNLR for diagnosis of pancreatic cancer**. *Future Oncol* (2018.0) **14** 229-239. DOI: 10.2217/fon-2017-0339
|
---
title: Correlation and agreement between arterial and venous blood gas analysis in
patients with hypotension—an emergency department-based cross-sectional study
authors:
- Hari Prasad
- Nagasubramanyam Vempalli
- Naman Agrawal
- U. N. Ajun
- Ajmal Salam
- Soumya Subhra Datta
- Ashutosh Singhal
- Nishant Ranjan
- P. P. Shabeeba Sherin
- G. Sundareshan
journal: International Journal of Emergency Medicine
year: 2023
pmcid: PMC9999648
doi: 10.1186/s12245-023-00486-0
license: CC BY 4.0
---
# Correlation and agreement between arterial and venous blood gas analysis in patients with hypotension—an emergency department-based cross-sectional study
## Abstract
### Background
Blood gas analysis is integral to assessing emergency department (ED) patients with acute respiratory or metabolic disease. Arterial blood gas (ABG) is the gold standard for oxygenation, ventilation, and acid–base status but is painful to obtain. Peripheral venous blood gas (VBG) is a valuable alternative as it is less painful and easy to collect. The comparability of ABG and VBG was studied in various conditions. But in hypotension, previous findings were inconsistent. So, we studied the correlation and agreement between ABG and VBG in hypotensive patients.
### Methodology
The study was conducted at the emergency department of a tertiary healthcare center in Northern India. Patients with hypotension above 18 years who satisfied the inclusion criteria were clinically evaluated. Patients who require ABG as a part of routine care were sampled. ABG was collected from the radial artery. VBG was obtained from the cubital or dorsal hand veins. Both samples were collected within 10 min and were analyzed. All ABG and VBG variables were entered in premade proforma. The patient was then treated and disposed of according to institutional protocol.
### Results
A total of 250 patients were enrolled. The mean age was 53.25 ± 15.71 years. $56.8\%$ were male. The study included $45.6\%$ septic, $34.4\%$ hypovolemic, $18\%$ cardiogenic, and $2\%$ obstructive shock patients. The study found a strong correlation and agreement for ABG and VBG pH, pCO2, HCO3, lactate, sodium, potassium, chloride, ionized calcium, blood urea nitrogen, base excess, and arterial/alveolar oxygen ratio. Hence, regression equations were made for the aforementioned. There was no correlation observed between ABG and VBG pO2 and SpO2. Our study concluded that VBG could be a reasonable alternative for ABG in hypotensive patients. We can also mathematically predict values of ABG from VBG using regression equations derived.
### Conclusions
ABG sampling causes most unpleasant experiences to patients and is associated with complications like arterial injury, thrombosis, air or clotted-blood embolism, arterial occlusion, hematoma, aneurysm formation, and reflex sympathetic dystrophy. The study has shown strong correlations and agreements for most ABG and VBG parameters and can predict ABG mathematically using regression formulas formulated from VBG. This will decrease needle stick injury, consume less time, and make blood gas evaluation easy in hypotensive settings.
## Background
The first examination of gas content in blood dates back to the 1670 s when Magnus, Hooke, and Boyle obtained gas from the blood by employing a vacuum pump [1]. Hasselbalch adapted Henderson’s laws to the logarithmic form in 1917, creating the Henderson-Hasselbalch equation (pH = pK + log [HCO3 −]/[CO2]), which forms the foundation of clinical acid–base analysis [2, 3].
The clinical use of blood gas analysis originated from the poliomyelitis epidemic in the early 1950s, which showed unprecedented mortality rates in Denmark [4].
In emergency departments (ED), blood gas analysis is used for three primary purposes: establishing acid–base state (mainly pH but also to lesser extent bicarbonate and pCO2), assessing ventilation function (mainly pCO2 but also pH and to a lesser extent pO2), and obtaining lactic acid levels in sepsis. These parameters are measured by blood gas analysis. Historically, analyses were performed on arterial blood [5]. Since the original description in the nineteenth century, the techniques for assessing gas tensions in blood have undergone refinements resulting in accurate point-of-care machines. The physiological description of respiratory failure and metabolic status was initially defined in arterial gas tensions and pH measurements. As a result, various clinical guidelines for treating respiratory failure describe using ABG for evaluation and response to treatment [6].
The parameters commonly measured by modern blood gas analyzers include pH; PCO2; partial pressure of O2 (PO2); concentration of hemoglobin (Hb); dyshemoglobin—COHb (carboxyhemoglobin) and MetHb (methemoglobin); lactate; glucose; and electrolytes (sodium, potassium, and chloride). At the same time, HCO3 and base excess are calculated from measured values [7].
Arterial blood gas analysis is the gold standard to obtain information about oxygenation, ventilation, and acid–base status. Parameters like pH, pO2, pCO2, HCO3, lactate, and base excess obtained from ABG are considered the gold standard. Peripheral venous blood gas (VBG) sampling is a valuable alternative to arterial blood gas (ABG) sampling in the emergency department evaluating metabolic and acid–base disorders. It is easier to obtain venous blood, so obtaining venous blood gas (VBG) is less painful, and samples may be drawn along with sampling for other laboratory tests. Venous blood sampling reduces the risk of arterial hematoma, dissection, and thrombosis. So, it is increasingly performed these days in the emergency department [1, 2].
Peripheral, mixed, and central venous blood can also be sampled. Venous blood gas (VBG) measurements obtained from peripheral, mixed, or central venous blood can be used interchangeably with ABGs to assess acid–base status in hemodynamically stable critically ill patients [8]. Adrogue and Weil concluded in their study that in the presence of severe circulatory failure, there is a worse agreement between arterial and central or mixed venous values, with central or mixed venous blood having a higher CO2 concentration and lower pH than arterial blood because of impaired removal of generated CO2 from the tissues. This increase in the venous–arterial PCO2 difference occurs in states of decreased flow irrespective of the reason for the circulatory failure and has an inverse relationship with cardiac output [9].
The main complications of arterial blood gas measurements include arterial injury, thrombosis, air or clotted-blood embolism, arterial occlusion, hematoma, aneurysm formation, and reflex sympathetic dystrophy [10].
Turner et al. evaluated recall of patients’ collective experience in their ICU stay. He found that ABG sampling was rated by $48\%$ of the patients as the most unpleasant experience during admission, followed by tracheal suction in $44\%$ of the patients [11].
Till 2006, only 22 cases of radial artery aneurysms were reported in the literature [12]. Less severe adverse events such as hematoma after radial artery puncture occur in up to $59\%$ of the patients [13].
Many studies have found a strong correlation between arterial pH, partial pressure of CO2 (PCO2), and calculated bicarbonate and corresponding venous values in different clinical conditions [14]. In patients with diabetic ketoacidosis, pH and PCO2 levels obtained from venous blood gas reasonably correlate with ABG values [15]. Elborn et al. conducted a study on COPD patients and found no significant difference between the arterial and venous CO2 tensions, and the two were closely correlated [16]. In a study by Rees et al. on 40 patients with chronic lung disease, they supported the use of peripheral VBG to estimate PaO2 in a vast majority of patients. They also showed that peripheral venous carbon dioxide tension and pH correlate well with arterial values [17]. In another study, Malinoski et al. concluded that VBG and arterial PCO2, pH, and base excess values had a good correlation [18].
Hypotension is one condition where findings of ABG and VBG comparability are inconsistent. Few studies tried to study the correlation between arterial and venous samples in patients with hypotension, but their study findings are not uniform and vary widely. The various studies done on comparability between ABG and VBG are shown in Table 1.Table 1Various studies done on comparability between ABG and VBGStudy nameStudy settingParameters that showed correlationParameters that did not show correlationYear of publicationYildizdas et al. [ 19]ICUpH, pCO2, HCO3, base excesspO22004Malinoski et al. [ 18]ICUpH, pCO2, and base excesspO22005Kelly et al. [ 20]HCO3pCO22010Shirani et al. [ 21]EDpH, HCO3, base excesspCO22011Kim et al. [ 22]ICUpH, pCO2, HCO3-2013Byrne et al. [ 23]pHpCO22014Hynes et al. [ 24]ICUpH, HCO3, lactate, base excess-2015Zeserson et al. [ 25]ED, ICUpH, pCO2-2018White et al. [ 26]ICUpHpCO2, HCO32018Rudkin et al. [ 27]ED-pH, pCO22020Shin et al. [ 28]EDpH, HCO3, Ca2 +, Na +, K +, Cl −, lactate, glucosepCO22020Nanjayya et al. [ 29]ICUpHpCO22020Boon et al. [ 30]EDLactatepH, base deficit2021 ABG sampling is technically challenging and time-consuming, is necessitating sampling skill, and is occasionally associated by the risk of staff needle stick exposure. Because venous blood sample requires fewer punctures, the danger of needle stick injury in medical staff is reduced [21].
Therefore, we intend to do correlation and agreement between ABG and VBG in hypotensive patients. Also, there were no studies done correlating values of ABG and VBG in hypotensive patients in academic emergency medicine settings in India to our knowledge.
Our cross-sectional study aimed to determine the relationship and agreement between arterial and venous blood gas analysis for the parameters pH, pCO2, HCO3, lactate, and base excess in hypotensive patients in the emergency department and also to develop prediction models in measuring parameters like pH, pCO2, HCO3, lactate, and base excess in arterial and venous samples.
## Study design and settings
This cross-sectional study was conducted in the Emergency Department of All India Institute of Medical Sciences, Rishikesh, Uttarakhand, between January 2021 and June 2022 (patient recruitment period from March 2021 to March 2022).
## Inclusion criteria
Patients with systolic blood pressure less than 90 mmHg or MAP less than 65 mmHg in the emergency department who require ABG as a part of routine carePatients more than 18 years of age
## Exclusion criteria
Those who did not give consentExistence of contraindications for arterial blood sampling, including impalpable or negative Allen’s test in the upper extremities, infection or fistula at the desired site of puncture, or having severe coagulation disordersInterval of more than 10 min between arterial and venous sampling and inappropriate sample transfer to the laboratoryPostcardiac arrest patients
## Sample size
Patients with hypotension above 18 years of age in the emergency department who satisfied the inclusion criteria were clinically evaluated. The minimum sample size was found to be 14 [22]. Because of the very small sample size, we included as much as participants using the consecutive sampling method during the study course (a total of 250 patients were included).
## Operational definition of hypotension
Patients with systolic blood pressure less than 90 mmHg or mean arterial blood pressure less than 65 mmHg.
## Clinical evaluation
Clinical evaluations were performed on emergency room patients with hypotension who were older than 18 years and met the inclusion criteria. After explaining the study to the patient or relative, written consent was obtained. The patient’s demographic details, vitals, clinical details, and diagnosis were entered in the data collection proforma. Patients who required ABG as a part of routine care per treating physician were sampled. An arterial sample (0.5–1 mL) was collected using a heparinized syringe from the radial artery at the wrist level. The venous blood sample was obtained from the cubital or dorsal hand veins. Both samples were collected with minimum delay (less than 10 min). Both samples were analyzed as soon as possible using a blood gas analyzer Nova Biomedicals Stat profile pHOX ultra.
## Data collection
All variables of ABG and VBG were entered in premade data collection proforma. The patient was then treated and disposed of according to institutional protocol. The study flowchart is attached (Fig. 1).Fig. 1Flow chart
## Statistical analysis
All data were entered into an Excel sheet and analyzed using SPSS 23.0 version. Descriptive statistics for numerical variables were calculated as the mean and standard deviation for normal distribution and median (IQR) for non-normal distribution, whereas percentages for qualitative variables. Agreement between arterial and venous blood gas analysis parameters was done using the Bland–Altman plot. Pearson correlation coefficients were applied to estimate the correlation between arterial and venous blood gas analysis parameters for Gaussian distribution and the Spearman correlation coefficient for non-Gaussian distribution. Arterial parameters were predicted from venous samples using linear regression.
## Results
Clinical evaluations were performed on emergency room patients with hypotension who were older than 18 and met the aforementioned inclusion criteria. We obtained their consent and enrolled them in our research. Of all the patients, 250 patients were finally analyzed in the study.
## Baseline characteristics of the study population
The study population’s main baseline characteristics and comorbidities are shown in Table 2.Table 2Baseline characteristics of the study population Characteristics Values Age in years, mean (standard deviation)53.25 ± 15.71 Men, number (%)142 ($56.8\%$) Women, number (%)108 ($43.2\%$) *Final diagnosis* Number (%) Infectious diseases27 ($10.8\%$) Cerebrovascular accident11 ($4.4\%$) Acute gastroenteritis9 ($3.6\%$) Acute decompensated heart failure3 ($1.2\%$) Poisoning7 ($2.8\%$) Malignancy24 ($9.6\%$) Pneumonia29 ($11.6\%$) Chronic liver disease22 ($8.8\%$) Intestinal obstruction3 ($1.2\%$) Pulmonary thromboembolism2 ($0.8\%$) Cardiac tamponade2 ($0.8\%$) Acute on chronic kidney disease5 ($2.0\%$) Acute pancreatitis2 ($0.8\%$) Obstructive airway disease19 ($7.6\%$) Acute coronary syndrome17 ($6.8\%$) Diabetic ketoacidosis3 ($1.2\%$) Dilated cardiomyopathy7 ($2.8\%$) Trauma3 ($1.2\%$) Urosepsis8 ($3.2\%$) Others47 ($18.8\%$) Type of shock Number (%) Septic114 ($45.6\%$) Hypovolemic86 ($34.4\%$) Cardiogenic45 ($18.0\%$) Obstructive5 ($2.0\%$)
## Examination findings
The main examination findings are shown in Table 3.Table 3Examination findings Examination Values Pulse rate in beats/minute, mean (standard deviation)103.78 ± 16.52 Respiratory rate in breaths/minute, mean (standard deviation)21.67 ± 4.31 Systolic BP in mmHg, mean (standard deviation)81.40 ± 6.01 Diastolic BP in mmHg, mean (standard deviation)49.31 ± 4.42 MAP in mmHg, mean (standard deviation)59.98 ± 4.16 SpO2 to maintain target saturation of $94\%$ or more except 88–$92\%$ for obstructive airway disease Number (%) Room air174 ($69.6\%$) Oxygen with nasal prongs41 ($16.4\%$) Oxygen with face mask18 ($7.2\%$) Oxygen with non-rebreather mask1 ($0.4\%$) NIV4 ($1.6\%$) Ventilator12 ($4.8\%$) Temperature in F, mean (standard deviation)98.68 ± 0.44 GCS, mean (standard deviation)13.75 ± 3.11 POCUS findings Values Lung profile Number (%) A profile123 ($49.2\%$) B profile127 ($50.8\%$) Lung sliding (yes)249 ($99.6\%$) Effusion (yes)11 ($4.4\%$) Other findings None238 ($95.2\%$) Mild pleural effusion7 ($2.8\%$) Moderate pleural effusion3 ($1.2\%$) Gross pleural effusion1 ($0.4\%$) Mass lesion1 ($0.4\%$) Contractility Fair210 ($84.0\%$) Moderately reduced27 ($10.8\%$) Severely reduced13 ($5.2\%$) Ejection fraction, mean (standard deviation)51.52 ± 8.73 IVC size (cm), mean (standard deviation)1.53 ± 0.34
## Assessment of parameters
The assessment of various parameters of ABG and VBG with mean, standard difference, and p-value is shown in Table 4.Table 4Statistically significant correlation between ABG and VBGParametersABG (mean ± SD)VBG (mean ± SD)Absolute difference (mean ± SD) P-valuepH7.37 (0.11)7.34 (0.12) − 0.03 (0.03) < 0.001pO2 (mmHg)87.76 (20.78)37.60 (12.80) − 50.17 (21.81) < 0.001pCO2 (mmHg)32.76 (15.25)36.17 (16.02)3.41 (5.19) < 0.001HCO3 (mmol/L)18.57 (6.11)19.09 (6.27)0.52 (1.53) < 0.001Lactate (mmol/L)2.84 (2.70)2.97 (2.83)0.13 (0.82) < 0.001Sodium (mmol/L)139.09 (8.52)138.97 (8.77) − 0.12 (3.40)0.365Potassium (mmol/L)3.91 (0.71)3.92 (0.71)0.02 (0.31)0.114Chloride (mmol/L)108.79 (7.93)108.70 (7.88) − 0.10 (2.81)0.590Ionized calcium (mmol/L)0.93 (0.20)0.91 (0.19) − 0.01 (0.15)0.350BUN (mg/dL)35.89 (22.15)35.73 (21.90) − 0.16 (4.89)0.432Base excess (mmol/L) − 5.41 (5.78) − 5.12 (5.90)0.29 (1.45)0.001Arterial/alveolar oxygen ratio (a/A)0.87 (0.29)0.37 (0.41) − 0.50 (0.26) < 0.001Oxygen saturation (%)95.93 (2.68)59.90 (19.11) − 36.03 (19.16) < 0.001
## Assessment of correlation, agreement, and regression of various parameters
Correlation and agreement of various parameters studied are shown in Table 5. Scatterplots showing the correlation of various parameters are shown in Figs. 2, 3, and 4. Bland–Altman plots showing agreement of various parameters are shown in Figs. 5 and 6. Regression analysis done for parameters shown strong correlation and agreement are as follows:Table 5Correlation and agreement of various parametersParametersInterclass correlation coefficientLimits of agreementpH0.96 ± 0.07pO2 (mmHg)0.20 ± 42.74pCO2 (mmHg)0.95 ± 10.16HCO3 (mmol/L)0.97 ± 3.00Lactate (mmol/L)0.96 ± 1.61Sodium (mmol/L)0.92 ± 6.67Potassium (mmol/L)0.91 ± 0.60Chloride (mmol/L)0.94 ± 5.51Ionized calcium (mmol/L)0.70 ± 0.30BUN (mg/dL)0.98 ± 9.59Base excess (mmol/L)0.97 ± 2.84Arterial/alveolar oxygen ratio (a/A)0.72 ± 0.52Oxygen saturation (%)0.01 ± 37.55Fig. 2Scatterplot for correlation between ABG and VBG for various parameters: A pH, B pO2, C pCO2, and D HCO3Fig. 3Scatterplot for correlation between ABG and VBG for various parameters: A lactate, B sodium, C potassium, and D chlorideFig. 4Scatterplot for correlation between ABG and VBG for various parameters: A ionized calcium, B BUN, C base excess, D a/A, and E SO2Fig. 5Bland–Altman plot for agreement between ABG and VBG for various parameters: A pH, B pO2, C pCO2, D HCO3, E lactate, and F sodiumFig. 6Bland–Altman plot for agreement between ABG and VBG for various parameters: A potassium, B chloride, C ionized calcium, D BUN, E base excess, F a/A, and G SO2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A\mathrm{BG}:\mathrm{pH}=0.42+0.95\times\mathrm{VBG}:\mathrm{pH}(R^2=0.91)$$\end{document}ABG:pH=0.42+0.95×VBG:pH(R2=0.91)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm{pCO}2 (\mathrm{mmHg}) = 0.19 + 0.9 \times \mathrm{ VBG}:\mathrm{pCO}2 (\mathrm{mmHg}) ({R}^{2}=0.90)$$\end{document}ABG:pCO2(mmHg)=0.19+0.9×VBG:pCO2(mmHg)(R2=0.90)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A\mathrm{BG}:\mathrm{HCO}3 (\mathrm{mmol}/\mathrm{L}) = 0.53 + 0.95 \times \mathrm{ VBG}:\mathrm{HCO}3 (\mathrm{mmol}/\mathrm{L}) ({R}^{2}=0.94)$$\end{document}ABG:HCO3(mmol/L)=0.53+0.95×VBG:HCO3(mmol/L)(R2=0.94)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm{lactate }(\mathrm{mmol}/\mathrm{L}) = 0.12 + 0.92 \times \mathrm{ VBG}:\mathrm{lactate }(\mathrm{mmol}/\mathrm{L}) ({R}^{2}=0.92)$$\end{document}ABG:lactate(mmol/L)=0.12+0.92×VBG:lactate(mmol/L)(R2=0.92)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm{S}.\mathrm{ sodium }(\mathrm{mmol}/\mathrm{L}) = 14.44 + 0.9 \times \mathrm{ VBG}:\mathrm{S}.\mathrm{ sodium }(\mathrm{mmol}/\mathrm{L}) ({R}^{2}=0.85)$$\end{document}ABG:S.sodium(mmol/L)=14.44+0.9×VBG:S.sodium(mmol/L)(R2=0.85)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm{S}.\mathrm{ potassium }(\mathrm{mmol}/\mathrm{L}) = 0.34 + 0.91 \times \mathrm{ VBG}:\mathrm{S}.\mathrm{ potassium }(\mathrm{mmol}/\mathrm{L}) ({R}^{2}= 0.82)$$\end{document}ABG:S.potassium(mmol/L)=0.34+0.91×VBG:S.potassium(mmol/L)(R2=0.82)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm{S}.\mathrm{ chloride }(\mathrm{mmol}/\mathrm{L}) = 6.3 + 0.94 \times \mathrm{ VBG}:\mathrm{S}.\mathrm{ chloride }(\mathrm{mmol}/\mathrm{L}) ({R}^{2}=0.88)$$\end{document}ABG:S.chloride(mmol/L)=6.3+0.94×VBG:S.chloride(mmol/L)(R2=0.88)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm{ionized Ca}2+ (\mathrm{mmol}/\mathrm{L}) = 0.29 + 0.7 \times \mathrm{ VBG}:\mathrm{ionized Ca}2+ (\mathrm{mmol}/\mathrm{L}) ({R}^{2}=0.49)$$\end{document}ABG:ionizedCa2+(mmol/L)=0.29+0.7×VBG:ionizedCa2+(mmol/L)(R2=0.49)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm{BUN }(\mathrm{mg}/\mathrm{dL}) = 0.65 + 0.99 \times \mathrm{ VBG}:\mathrm{BUN }(\mathrm{mg}/\mathrm{dL}) ({R}^{2}=0.95)$$\end{document}ABG:BUN(mg/dL)=0.65+0.99×VBG:BUN(mg/dL)(R2=0.95)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm{ Base E}\times \mathrm{cess }(\mathrm{mmol}/\mathrm{L}) = -0.55 + 0.95 \times \mathrm{ VBG}:\mathrm{base excess }(\mathrm{mmol}/\mathrm{L}) ({R}^{2}=0.94)$$\end{document}ABG:BaseE×cess(mmol/L)=-0.55+0.95×VBG:baseexcess(mmol/L)(R2=0.94)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ABG}:\mathrm a/\mathrm $A = 0.67$+0.55\times\mathrm{VBG}:\mathrm a/\mathrm A(R^{\mathit2}=0.58)$$\end{document}ABG:a/$A = 0.67$+0.55×VBG:a/A(R2=0.58)
## Discussion
Our study enrolled 250 patients with hypotension and required ABG as a part of routine care and satisfied inclusion criteria.
## Demographics
Our study population comprised 250 patients with a mean (standard deviation) years of 53.25 (± 15.71). The study included $56.8\%$ [142] males and $43.2\%$ [108] females. A study by Kim et al. done on 34 patients had a mean age of 65.57 ± 12.4 years, and males were $58.8\%$ [20] and females $41.2\%$ [14] [22]. In a study done by Shirani et al., 192 patients had a mean age of 51.6 ± 23.6 years, and there were males $67.7\%$ and females $32.3\%$ [21], whereas in the study by Shin et al., 231 hypotensive patients had a mean age (SD) of 68.2 (16.6) years and males $64.5\%$ and females $35.5\%$ [28].
$50.8\%$ were residents of Uttarakhand, $46.8\%$ were residents of Uttar Pradesh, and $2.4\%$ were from other states. $45.6\%$ of patients were septic, $34.4\%$ hypovolemic, $18\%$ cardiogenic, and $2\%$ obstructive shock.
## Size and site of needle used
For ABG, 24G needles were used in all 250 patients, whereas for VBG, 18G was used for sampling in $59.6\%$ of cases, and in $40.4\%$ of cases, a 20G needle was used. The radial artery obtained all ABGs in our study, while VBG was obtained from forearm veins. In the study by White et al., ABG was collected by either needle puncture or withdrawal from an arterial catheter. The VBG was collected from the upper extremity, and a tourniquet could be left in place for no more than 1 min [26].
## pH
There was a very strong correlation between ABG:pH and VBG:pH, and this correlation was statistically significant (interclass correlation coefficient = 0.96, p ≤ 0.001) in our study, with $94.0\%$ of the observations having a difference which was within the limits of agreement (± 0.07). The r-value was 0.91 and the formula for ABG pH was devised as ABG:pH = 0.42 + 0.95 × VBG:pH. In the study by Yildizdas et al., pH showed a good interclass correlation of 0.907 with a p-value < 0.001. The r-value for pH for that study was 0.994 [19]. In the study by White et al., pH had an interclass correlation of 0.90 with a p-value < 0.001. Bias ± SD of ABG-VBG was 0.03 ± 0.04 [26]. In Zeserson et al. study, the mean difference for pH between VBG and ABG was 0.03 ($95\%$ confidence interval: 0.03–0.04) with a Pearson correlation of 0.94 [25]. In the study by Nanjayya et al., the mean bias for pH was + 0.036 with $95\%$ LOA ranging from − 0.005 to + 0.078 [29]. In a meta-analysis by Byrne et al., there was little difference between the pH obtained from the PVBG and the ABG, with the arterial pH typically 0.03 higher than the venous pH ($95\%$ confidence interval 0.029–0.038) [23]. Kim et al. found a good correlation between ABG and VBG pH with a Pearson correlation coefficient of 0.783 with a p-value of 0.0001. Regression equations were derived to predict ABG pH from peripheral VBG pH as follows: arterial pH = 0.763 × venous pH + 1.786 (R 2 = 0.544). The multivariate regression equations were found as follows: arterial pH = − 1.108 + 1.145 × venous pH + 0.008 × PCO2 – 0.012 × venous HCO3 + 0.002 × venous total CO2 ($r = 0.655$) [22]. In a study by Shirani et al., the average VBG-ABG amount of difference ($95\%$ limits of agreement) in the hypotension versus normotensive group were − 0.030 (− 0.09 to 0.03) vs. − 0.016 (− 0.1 to 0.068) for pH ($$p \leq 0.01$$) [21]. In studies by Boon et al. and Rudkin et al., pH was not correlating between ABG and VBG [27, 30].
## pO2
There was a weak correlation between ABG:pO2 (mmHg) and VBG:pO2 (mmHg), and this correlation was statistically significant (interclass correlation coefficient = 0.20, p ≤ 0.001 in our study). No correlation was found between ABG:pO2 and VBG:pO2 in previous studies.
## pCO2
There was a very strong correlation between ABG:pCO2 (mmHg) and VBG:pCO2 (mmHg), and this correlation was statistically significant (interclass correlation coefficient = 0.95, p = < 0.001) in our study. In the study by Yildizdas et al., there was a good correlation between ABG and VBG with an interclass correlation of 0.978 with a p-value < 0.001. The r-value found was 0.957 [19]. In the study by Malinoski et al., pCO2 values have a good correlation and agreement between ABG and VBG values with $R = 0.88$, $p \leq 0.001$, and $95\%$ LOAs of − 2.2 to 10.9 [18]. Kim et al., in their study, found a good Pearson correlation coefficient of 0.705 with a p-value of 0.0001 for pCO2. Regression equation derived for pCO2-arterial PCO2 = 0.611 × venous PCO2 + 9.521 (R 2 = 0.497). Multivariate regression equation derived was arterial PCO2 = 88.6 − 10.888 × venous pH + 0.150 × PCO2 + 0.812 × venous HCO3 + 0.124 × venous total CO2 ($r = 0.609$) [22], whereas in the study by Zeserson et al., the mean difference for pCO2 between VBG and ABG was 4.8 mm Hg ($95\%$ confidence interval: 3.7–6.0 mm Hg) with a Pearson correlation of 0.93 [25]. In our study, $95.6\%$ of the observations had a difference within the limits of agreement (± 10.16). R 2 = 0.90 for pCO2 (mmHg) and the formula was devised as ABG:pCO2 (mmHg) = 0.19 + 0.9 × VBG:pCO2 (mmHg).
## HCO3
There was a very strong correlation between ABG:HCO3 (mmol/L) and VBG:HCO3 (mmol/L), and this correlation was statistically significant (interclass correlation coefficient = 0.97, p ≤ 0.001) in our study. In the study by Shin et al., there was a good correlation and agreement between ABG and VBG HCO3 with a mean (SD) of 19.27 (5.39) with $95\%$ CI as − 1.927 to − 1.461 and LOA ± 3.53 [28]. Yildizdas et al. found that HCO3 has a good Pearson correlation coefficient of 0.976 with a p-value < 0.001 and $r = 1.676$ [19]. Venous–arterial difference of HCO3 − is − 0.37mmmol/L with good agreement in the study by Hynes et al. [ 24]. Pearson correlation coefficient of HCO3 in a study by Kim et al. was 0.846 with a p-value of 0.0001. The regression equation derived for HCO3 was arterial HCO3 = 0.822 × venous HCO3 + 2.815 (R 2 = 0.716). Multivariate regression equation derived was arterial HCO3 = − 89.266 + 12.677 × venous pH + 0.042 × PCO2 + 0.675 × venous HCO3 + 0.185 × venous total CO2 ($r = 0.782$) [22]. In a meta-analysis by Kelly et al., the weighted mean difference between arterial and venous values for bicarbonate was –1.41 mmol/L ($$n = 905$$), with $95\%$ limits of agreement of the order of ± 5 mmol/L [20]. Shirani et al. found the average VBG-ABG amount of difference ($95\%$ limits of agreement) in the hypotension versus normotensive group was 1.79 (− 1.91 to 5.49) vs. 1.32 (− 1.94 to 4.58) mEq/L for HCO3 ($$p \leq 0.032$$) [21]. In our study, $93.6\%$ of the observations had a difference within the limits of agreement (± 3.00). R 2 = 0.94 for HCO3 (mmol/L) and the formula was devised as ABG:HCO3 (mmol/L) = 0.53 + 0.95 × VBG:HCO3 (mmol/L).
## Lactate
There was a very strong correlation between ABG:lactate (mmol/L) and VBG:lactate (mmol/L), which was statistically significant (interclass correlation coefficient = 0.96, p = < 0.001) in our study. $97.6\%$ of the observations had a difference within the limits of agreement (± 1.61). R 2 = 0.92 for lactate (mmol/L) and the formula was devised as ABG:lactate (mmol/L) = 0.12 + 0.92 × VBG:lactate (mmol/L). In a study by Hynes et al., the venous–arterial difference for lactate was found to be 0.16 mmol/L [24]. In a study by Boon et al., venous lactate was clinically equivalent based on the pre-determined threshold limits of − 1.5 to 1.5 mmol/L, where $96.0\%$ of the values were within this acceptable range [30]. Shin et al., in their study, found the mean (SD) for lactate as 3.27 (3.23) with $95\%$ CI (− 0.406, − 0.241) and LOA ± 1.25 [28].
## Sodium
In the study by Shin et al., the mean (SD) for sodium was 136.21 (6.60) with $95\%$ CI (− 3.104, − 2.532) and LOA ± 4.33 [28]. There was a very strong correlation between ABG and VBG S. Sodium (mmol/L), and this correlation was statistically significant (interclass correlation coefficient = 0.92, p ≤ 0.001) in our study. $95.6\%$ of the observations had a difference within the limits of agreement (± 6.67). R 2 = 0.85 for sodium (mmol/L) and the formula was devised as ABG:S. sodium (mmol/L) = 14.44 + 0.9 × VBG:S. sodium (mmol/L).
## Potassium
There was a very strong correlation between ABG and VBG potassium (mmol/L), and this correlation was statistically significant (interclass correlation coefficient = 0.91, p ≤ 0.001) in our study. $95.6\%$ of the observations had a difference within the limits of agreement (± 0.60). The regression equation for potassium (mmol/L) was devised as ABG:S. potassium (mmol/L) = 0.34 + 0.91 × VBG:S. potassium (mmol/L) [R 2 = 0.82], whereas in the study by Shin et al., mean (SD) for potassium was 4.42 (1.18) with $95\%$ CI (− 0.365, − 0.257) and LOA ± 0.82 [28].
## Chloride
The study by Shin et al. found the mean (SD) for chloride as 102.94 (6.31) with $95\%$ CI (− 4.062, − 2.934) and LOA as ± 8.53 [28], whereas in our study, there was a very strong correlation between ABG and VBG chloride (mmol/L), and this correlation was statistically significant (interclass correlation coefficient = 0.94, p ≤ 0.001). $94.0\%$ of the observations had a difference within the limits of agreement (± 5.51). The regression equation for chloride was done using R2 = 0.88 formula was devised as ABG:S. chloride (mmol/L) = 6.3 + 0.94 × VBG:S. chloride (mmol/L). No other studies have done a correlation between ABG and VBG for chloride.
## Ionized calcium
Ionized calcium was found to have a strong correlation between ABG and VBG with an interclass correlation coefficient = 0.70 with p-value ≤ 0.001. $94.0\%$ of the observations had a difference which was within the limits of agreement (± 0.30). The regression equation devised for ionized calcium (mmol/L) was ABG:ionized Ca2 + (mmol/L) = 0.29 + 0.7 × VBG:ionized Ca2 + (mmol/L) [R 2 = 0.49] in our study, whereas in the study by Shin et al., the mean (SD) for ionized calcium was 1.21 (0.08) with $95\%$ CI (0.064, 0.089) and LOA as ± 0.18 [28].
## Blood urea nitrogen
Blood urea nitrogen was found to have a very strong correlation between ABG and VBG with an interclass correlation coefficient = 0.98 and p-value ≤ 0.001. $96.0\%$ of the observations had a difference which was within the limits of agreement (± 9.59). The regression equation derived for BUN (mg/dL) was ABG:BUN (mg/dL) = 0.65 + 0.99 × VBG:BUN (mg/dL) [R 2 = 0.95] in our study. We found no other study which correlates ABG:BUN and VBG:BUN.
## Base excess
There was a very strong correlation between ABG:base excess (mmol/L) and VBG:base excess (mmol/L) with an interclass correlation coefficient = 0.97 and p ≤ 0.001 in our study. $94.8\%$ of the observations had a difference within the limits of agreement (± 2.84). The regression equation derived for base excess (mmol/L) was ABG:base excess (mmol/L) = − 0.55 + 0.95 × VBG:base excess (mmol/L) [R 2 = 0.94]. The study by Yildizdas et al. found a good correlation with a Pearson correlation coefficient of 0.972 with a p-value < 0.001. R 2 was found to have 0.945 for base excess [19]. Malinoski et al. compared ABG and VBg values and found that base excess has $R = 0.96$, $p \leq 0.001$, and $95\%$ LOAs of − 2.2 to 1.8 in their study [18]. The average VBG-ABG amount of difference ($95\%$ limits of agreement) in the hypotension versus normotensive group found in a study by Shirani et al. was d 0.25 (− 3.73 to 4.23) vs. 0.79 (− 2.51 to 4.09) for BE ($$p \leq 0.036$$) [21]. In a study by Hynes et al., the venous–arterial difference for base excess was found to be 0.08 mEq/L [24].
## Arterial/alveolar oxygen ratio
Our study found a strong correlation between ABG:a/A and VBG:a/A, and this correlation was statistically significant (interclass correlation coefficient = 0.72, p ≤ 0.001). $98.4\%$ of the observations had a difference within the limits of agreement (± 0.52). The regression equation was derived with R 2 = 0.58 for a/A as ABG:a/$A = 0.67$ + 0.55 × VBG:a/A. Our study is the first to evaluate and find a correlation for arterial/alveolar oxygen ratio between ABG and VBG.
## Oxygen saturation
Our study found a weak correlation between ABG:SO2 (%) and VBG:SO2 (%), and this correlation was not statistically significant (interclass correlation coefficient = 0.01, $$p \leq 0.412$$). All other studies comparing ABG and VBG found weak or no correlation for oxygen saturation.
## Limitations
In our study, patients were recruited using convenience sampling as study investigators will not be available in all shifts. The study involved collecting single pair of arterial and peripheral venous samples from each patient. So, the homogeneity of ABG and peripheral VBG was not studied as multiple samples at different intervals were not done. We also had many cases where one patient had multiple diagnoses making it challenging to characterize into various subgroups. Subgroups of shock were defined at the time of “working diagnosis” when enrolling the patient and not the final diagnosis. The study design and setting did not allow for a follow-up on the mortality of the patients studied. Patients differed according to many demographic factors, which were not all the same.
## Conclusions
ABG and VBG assessments are essential tests for assessing ventilation, acid–base disturbances, and other metabolic parameters of patients. Obtaining ABG or VBG in hypotension is challenging as many previous studies have shown conflicting results. Our study has shown either strong or very strong correlations and agreements for most ABG and VBG parameters except pO2 and SO2. We can also predict an ABG mathematically using regression formulas devised from a VBG sample. This will decrease needle stick injury, consume less time, and make blood gas evaluation easy in hypotensive settings. Further studies are required to find the correlation between ABG and central, peripheral, and mixed VBGs and capillary blood gas. Our study is the first to incorporate all four types of shock in ABG and VBG analysis. Also, this is the first one to incorporate all the parameters of ABG and VBG. Also, this is the largest single-center study in ABG and VBG comparison on hypotensive patients. Our study is the first to incorporate details of the point-of-care ultrasound in hypotensive ABG v VBG studies. We have a standardized site and size of the needle to be used for ABG and VBG in this study.
## References
1. Breathnach CS. **The development of blood gas analysis**. *Med Hist* (1972.0) **16** 51-62. DOI: 10.1017/S0025727300017257
2. Henderson LJ. **Das Gleichgewicht zwischen Basen und Säuren im tierischen Organismus**. *Ergeb Physiol* (1909.0) **8** 254-325. DOI: 10.1007/BF02321087
3. Hasselbalch KA. *Die Berechnung der Wasserstoffzahl des Blutes aus der freien und gebundenen Kohlens??ure desselben, und die Sauerstoffbindung des Blutes als Funktion der Wasserstoffzahl* (1916.0)
4. Whitehead T, Slutsky AS, Fink M, Hayes M, Soni N. **Ventilation**. *Classic papers in critical care* (2008.0) 1-29
5. Kelly A. **Agreement between arterial and venous blood gases in emergency medical care: a systematic review**. *Hong Kong J Emerg Med.* (2013.0) **20** 166-71. DOI: 10.1177/102490791302000307
6. Severinghaus JW. **The invention and development of blood gas analysis apparatus**. *Anesthesiology* (2002.0) **97** 253-6. DOI: 10.1097/00000542-200207000-00031
7. Davis MD, Walsh BK, Sittig SE, Restrepo RD. **AARC clinical practice guideline: blood gas analysis and hemoximetry: 2013**. *Respir Care* (2013.0) **58** 1694-703. DOI: 10.4187/respcare.02786
8. Treger R, Pirouz S, Kamangar N, Corry D. **Agreement between central venous and arterial blood gas measurements in the intensive care unit**. *Clin J Am Soc Nephrol* (2010.0) **5** 390-4. DOI: 10.2215/CJN.00330109
9. Adrogué HJ, Rashad MN, Gorin AB, Yacoub J, Madias NE. **Assessing acid-base status in circulatory failure**. *N Engl J Med* (1989.0) **320** 1312-6. DOI: 10.1056/NEJM198905183202004
10. Criscuolo C, Nepper G, Buchalter S. **Reflex sympathetic dystrophy following arterial blood gas sampling in the intensive care setting**. *Chest* (1995.0) **108** 578-80. DOI: 10.1378/chest.108.2.578
11. Turner JS, Briggs SJ, Springhorn HE, Potgieter PD. **Patients’ recollection of intensive care unit experience**. *Crit Care Med* (1990.0) **18** 966-8. DOI: 10.1097/00003246-199009000-00012
12. Patel K, Gandhi S, Sutariya H. **Radial artery pseudoaneurysm: a rare complication after a single arterial puncture for blood-gas analysis**. *Indian J Crit Care Med* (2016.0) **20** 622-6. DOI: 10.4103/0972-5229.192066
13. Capewell S, Ali NJ, Makker H, Cockwell P, Davies DP, Rogers S. **Radial artery puncture: a comparison of threehaemostatic techniques**. *Respir Med* (1990.0) **84** 495-7. DOI: 10.1016/S0954-6111(08)80115-1
14. McKeever TM, Hearson G, Housley G, Reynolds C, Kinnear W, Harrison TW. **Using venous blood gas analysis in the assessment of COPD exacerbations: a prospective cohort study**. *Thorax* (2016.0) **71** 210-5. DOI: 10.1136/thoraxjnl-2015-207573
15. Brandenburg MA, Dire DJ. **Comparison of arterial and venous blood gas values in the initial emergency department evaluation of patients with diabetic ketoacidosis**. *Ann Emerg Med* (1998.0) **31** 459-65. DOI: 10.1016/S0196-0644(98)70254-9
16. Elborn JS, Finch MB, Stanford CF. **Non-arterial assessment of blood gas status in patients with chronic pulmonary disease**. *Ulster Med J.* (1991.0) **60** 164-7. PMID: 1785150
17. 17.Converting venous acid–base and oxygen status to arterial in patients with lung disease | European Respiratory Society. Available from: https://erj.ersjournals.com/content/33/5/1141.short. [cited 2022 Apr 12].
18. Malinoski DJ, Todd SR, Slone DS, Mullins RJ, Schreiber MA. **Correlation of central venous and arterial blood gas measurements in mechanically ventilated trauma patients**. *ARCH SURG* (2005.0) **140** 4. DOI: 10.1001/archsurg.140.11.1122
19. Yildizdas D. **Correlation of simultaneously obtained capillary, venous, and arterial blood gases of patients in a paediatric intensive care unit**. *Arch Dis Child* (2004.0) **89** 176-80. DOI: 10.1136/adc.2002.016261
20. Kelly AM. **Review article: Can venous blood gas analysis replace arterial in emergency medical care: venous blood gas analysis**. *Emerg Med Australas* (2010.0) **22** 493-8. DOI: 10.1111/j.1742-6723.2010.01344.x
21. Shirani F, Salehi R, Naini AE, Azizkhani R, Gholamrezaei A. **The effects of hypotension on differences between the results of simultaneous venous and arterial blood gas analysis**. *J Res Med Sci Off J Isfahan Univ Med Sci* (2011.0) **16** 188-94
22. Kim BR, Park SJ, Shin HS, Jung YS, Rim H. **Correlation between peripheral venous and arterial blood gas measurements in patients admitted to the intensive care unit: a single-center study**. *Kidney Res Clin Pract.* (2013.0) **32** 32-8. DOI: 10.1016/j.krcp.2013.01.002
23. Byrne AL, Bennett M, Chatterji R, Symons R, Pace NL, Thomas PS. **Peripheral venous and arterial blood gas analysis in adults: are they comparable? A systematic review and meta-analysis**. *Respirology* (2014.0) **19** 168-75. DOI: 10.1111/resp.12225
24. 24.Hynes D, Bates S, Loughman A, Klim S, French C, Kelly AM. Arteriovenous blood gas agreement in intensive care patients with varying levels of circulatory compromise: a pilot study. Crit Care Resusc. 2015. Available from: https://search.informit.org/doi/abs/10.3316/informit.704237279119400. [cited 2022 Apr 12].
25. Zeserson E, Goodgame B, Hess JD, Schultz K, Hoon C, Lamb K. **Correlation of venous blood gas and pulse oximetry with arterial blood gas in the undifferentiated critically ill patient**. *J Intensive Care Med.* (2018.0) **33** 176-81. DOI: 10.1177/0885066616652597
26. White HD, Vazquez-Sandoval A, Quiroga PF, Song J, Jones SF, Arroliga AC. **Utility of venous blood gases in severe sepsis and septic shock**. *Bayl Univ Med Cent Proc* (2018.0) **31** 269-75. DOI: 10.1080/08998280.2018.1460133
27. Rudkin SE, Anderson CL, Grogan TR, Elashoff DA, Treger RM. **Assessing acid–base status in circulatory failure: relationship between arterial and peripheral venous blood gas measurements in hypovolemic shock**. *J Intensive Care Med* (2020.0) **35** 511-8. DOI: 10.1177/0885066618762335
28. Shin H, Lee I, Kim C, Choi HJ. **Point-of-care blood analysis of hypotensive patients in the emergency department**. *Am J Emerg Med* (2020.0) **38** 1049-57. DOI: 10.1016/j.ajem.2019.158363
29. Nanjayya VB, McCracken P, Vallance S, Board J, Kelly PJ, Schneider HG. **Arterio-VENouS Intra Subject agreement for blood gases within intensive care: the AVENSIS study**. *J Intensive Care Soc* (2020.0) **21** 64-71. DOI: 10.1177/1751143719840259
30. Boon Y, Kuan WS, Chan YH, Ibrahim I, Chua MT. **Agreement between arterial and venous blood gases in trauma resuscitation in emergency department (AGREE)**. *Eur J Trauma Emerg Surg* (2021.0) **47** 365-72. DOI: 10.1007/s00068-019-01190-6
|
---
title: Extracellular vesicle distribution and localization in skeletal muscle at rest
and following disuse atrophy
authors:
- Ahmed Ismaeel
- Douglas W. Van Pelt
- Zachary R. Hettinger
- Xu Fu
- Christopher I. Richards
- Timothy A. Butterfield
- Jonathan J. Petrocelli
- Ivan J. Vechetti
- Amy L. Confides
- Micah J. Drummond
- Esther E. Dupont-Versteegden
journal: Skeletal Muscle
year: 2023
pmcid: PMC9999658
doi: 10.1186/s13395-023-00315-1
license: CC BY 4.0
---
# Extracellular vesicle distribution and localization in skeletal muscle at rest and following disuse atrophy
## Abstract
### Background
Skeletal muscle (SkM) is a large, secretory organ that produces and releases myokines that can have autocrine, paracrine, and endocrine effects. Whether extracellular vesicles (EVs) also play a role in the SkM adaptive response and ability to communicate with other tissues is not well understood. The purpose of this study was to investigate EV biogenesis factors, marker expression, and localization across cell types in the skeletal muscle. We also aimed to investigate whether EV concentrations are altered by disuse atrophy.
### Methods
To identify the potential markers of SkM-derived EVs, EVs were isolated from rat serum using density gradient ultracentrifugation, followed by fluorescence correlation spectroscopy measurements or qPCR. Single-cell RNA sequencing (scRNA-seq) data from rat SkM were analyzed to assess the EV biogenesis factor expression, and cellular localization of tetraspanins was investigated by immunohistochemistry. Finally, to assess the effects of mechanical unloading on EV expression in vivo, EV concentrations were measured in the serum by nanoparticle tracking analysis in both a rat and human model of disuse.
### Results
In this study, we show that the widely used markers of SkM-derived EVs, α-sarcoglycan and miR-1, are undetectable in serum EVs. We also found that EV biogenesis factors, including the tetraspanins CD63, CD9, and CD81, are expressed by a variety of cell types in SkM. SkM sections showed very low detection of CD63, CD9, and CD81 in myofibers and instead accumulation within the interstitial space. Furthermore, although there were no differences in serum EV concentrations following hindlimb suspension in rats, serum EV concentrations were elevated in human subjects after bed rest.
### Conclusions
Our findings provide insight into the distribution and localization of EVs in SkM and demonstrate the importance of methodological guidelines in SkM EV research.
## Background
Extracellular vesicles (EVs) are thought to play a role in the adaptive response of skeletal muscle and the ability to communicate with other tissues and organs [1]. EVs are endosomal or plasma membrane-derived vesicles released from cells [2] and are primarily classified into subtypes based upon their physical characteristics like size (50–1000 nm), density (low, middle, high), and/or biochemical composition (i.e., surface markers) [3]. EVs differ by their content, which can include mRNAs, microRNAs (miRNAs), proteins, lipids, and metabolites [4]. While it was initially thought that EVs were a mechanism to rid cells of unwanted material, it is now understood that the cargo carried by EVs can be delivered to local and distant cells and have biological and physiological effects on recipient cells and tissues [5, 6].
Skeletal muscle is not only the largest organ in the human body, playing a central role in whole-body energy metabolism, but also acts as a secretory organ, producing and releasing hundreds of products, including myokines, which can have autocrine, paracrine, and endocrine effects [7, 8]. Recently, mechanistic studies on EV-mediated cell-cell communication have shown the importance of EVs in organ crosstalk in normal physiology and diseased states [9], and this includes those derived from the skeletal muscle. However, EV research is complicated by the fact that EVs are produced nearly ubiquitously in all cells and tissues. While accumulating evidence supports the presence of vascular cell-derived EVs in circulation, the degree to which nonvascular cells release EVs across the vascular endothelium and into the bloodstream is not well-understood [10]. The lack of tissue-specific EV markers makes it even more difficult to track EVs from their “parent” cells or tissues.
Skeletal muscle is a complex heterogeneous tissue comprising not only multinucleated muscle fibers, but also several mononuclear cell populations, including immune cells, endothelial cells, fibro-adipogenic progenitors (FAPs), and satellite cells, which also play a role in EV production and release from the muscle [11]. Due to the heightened skeletal muscle energy metabolism during exercise and increases in skeletal muscle contractions, studies have used different exercise modalities to assess the changes in circulating EVs, which may be skeletal muscle-derived. Serum or plasma EVs are increased following acute bouts of exercise in both animals and human participants [5, 12–17], and detailed EV phenotypic analyses demonstrated that a majority of EVs released during exercise originate from immune cells, platelets, and endothelial cells [18]. It remains unclear whether the skeletal muscle contributes to these exercise-induced increases in circulating EVs. By contrast, mechanical unloading leads to skeletal muscle atrophy with important clinical ramifications such as decreased muscle force production and functional independence, and the etiology of disuse atrophy is not well understood [19]. Emerging evidence suggests that miRNAs can modulate muscle size in response to different conditions and may play a role in associated systemic consequences [20]. In vitro, dexamethasone treatment-induced atrophy led to a reduction in miR-23a levels in C2C12 myotubes via increased release into EVs [21]. Our laboratory has also previously shown that miR-203a-3p expression in circulating EVs was associated with skeletal muscle protein turnover and atrophy [22]. However, the source of miR-203a-3p was not determined due to the aforementioned issues with tracking circulating EVs in vivo. Thus, the purpose of this study was to investigate EV biogenesis factors, marker expression, and localization across cell types in the skeletal muscle; release of EVs from the atrophied skeletal muscle; and whether EV concentrations are altered by disuse atrophy in rats and human participants.
## Cell culture
Monolayer cultures of C2C12 and L6 myoblasts were grown in Dulbecco’s modified Eagle’s medium (DMEM) (Thermo Fisher, Waltham, MA) supplemented with $10\%$ (L6) or $20\%$ (C2C12) fetal bovine serum (FBS) (HyClone Laboratories, Logan, UT) and 1X Penicillin/Streptomycin (Thermo Fisher) at 37 °C in a humidified $10\%$ CO2-$90\%$ air atmosphere incubator, as previously described [23, 24].
## Animals and experimental procedures
All procedures were approved by the University of Kentucky’s Institutional Animal Care and Use Committee. Male Brown Norway/F344 rats at 10 months of age (National Institute on Aging, Bethesda, MD) were used in this study. Rats were randomly assigned into one of four groups: weight-bearing control conditions (WB), hindlimb suspension (HS) for 4 h (4h HS), HS for 24 h (24h HS), and HS for 7 days (7d HS). Rats were allowed free access to food and water at all times and were housed on a 12:12-h light-dark cycle. Hindlimb suspension was performed as previously described [22]. Briefly, a tail device containing a hook was attached with gauze and cyanoacrylate glue while the animals were anesthetized with isoflurane ($2\%$ by inhalation). The tail device was connected via a thin cable to a pulley sliding on a vertically adjustable stainless steel bar running longitudinally above a high-sided cage. The system was designed in such a way that the rats could not rest their hindlimbs against any side of the cage but could move around the cage on their front limbs and could reach water and food easily. Cages were randomly placed in the room, and the room temperature was 27 °C.
## Blood and tissue collection
At the end of the experimental period, rats were anesthetized with pentobarbital sodium, and blood was immediately collected through cardiac puncture. Rats were euthanized, and the soleus muscles were excised, weighed, and used for ex vivo collection of muscle EVs (right) or dissected, weighed, frozen in liquid nitrogen, and stored at − 80 °C for later biochemical analyses (left). For immunohistochemistry (IHC) analyses, the soleus muscles were covered in Tissue-Tek optimal cutting temperature compound (Sakura Finetek, Torrance, CA, USA), frozen in liquid nitrogen-cooled isopentane, and stored at − 80 °C. The soleus muscles were used in these experiments because in rats, the soleus, which is almost exclusively type I, is especially susceptible to hindlimb suspension-induced muscle atrophy [25, 26]. Furthermore, the smaller size of the soleus limits issues of oxygen diffusion during ex vivo muscle assessments [27]. The gastrocnemius muscles were used for single-cell RNA sequencing (scRNA-seq) as described in [28]. The serum was isolated by allowing blood to clot at room temperature for 30 min before centrifugation for 10 min at 2000g at 4 °C. The serum supernatant was collected and stored at − 80 °C until analysis. The number of animals used in each experiment is listed in the figure legends.
## Human participants
Serum and muscle biopsies obtained from the vastus lateralis from a previously published bed rest study (young group, age: 23 ± 3 years) were used for EV and RNA isolation, respectively [29]. Human participants were recruited at the University of Utah under an approved Institutional Review Board protocol, and the study conformed to the Declaration of Helsinki. Bed rest (5 days; Monday–Friday) took place according to the protocol and safety guidelines described in detail in the original publication [29]. For total RNA isolation from the muscle, samples were homogenized in TRIzol Reagent (Invitrogen, Waltham, MA), and 1-bromo-3-chloropropane was added for phase separation. Finally, 2-propanol was used to precipitate the RNA, and RNA was pelleted by centrifugation (12,000g for 10 min). The Bio-Rad iScript Reverse Transcription Supermix (1708841, Bio-Rad Laboratories, Hercules, CA) was used for cDNA synthesis from 1 μg of total RNA. Real-time PCR was used to determine the relative mRNA expression of the tetraspanins, CD63, CD9, and CD81. PCR reactions used primer sets and Applied Biosystems PowerUp SYBR Green Master Mix (A25742, Applied Biosystems, Waltham, MA).
## Ex vivo collection of muscle EVs
For ex vivo experiments, the soleus muscle was excised intact, rinsed with Krebs-Henseleit buffer (KHB) (118.5mM NaCl, 1.2mM MgSO4, 4.7mM KCl, 1.2mM KH2PO4, 25mM NaHCO3, 2.5mM CaCl2; pH 7.4), and then suspended prior to incubation in continuously gassed ($95\%$ O2/$5\%$ CO2) KHB supplemented with 5 mM glucose at 37 °C for 1 h. EV abundance in the KHB was measured using nanoparticle tracking analysis (NTA) (Zetaview®, Particle Metrix, Meerbusch, Germany) immediately after the 1-h incubation period. The Zetaview® instrument uses a laser scattering video microscope to track individual nanoparticle movement under Brownian motion, measuring the size and concentration [30].
## Serum EV isolation
EVs isolated from rat and human serum for miRNA analysis were isolated from 500 μL of serum with ExoQuick Exosome Precipitation Solution (System Biosciences (SBI), Palo Alto, CA). The serum was first centrifuged at 3000g for 15 min to remove debris, and the supernatant was collected and filtered through a 0.22-μm low-binding PVDF filter (Millex-GV; Millipore, Tullagreen, Ireland). Approximately 240 μL of ExoQuick was added to the sample and incubated at 4 °C overnight. The ExoQuick-serum mixture was centrifuged at 1500g for 30 min to pellet the EVs. The supernatant was removed, and the EV pellet was reconstituted in 300 μL of PBS.
To isolate a purer sample of EVs and better assess the contribution of EVs to the overall particle population in rat serum, we used a slightly modified density gradient ultracentrifugation (DGUC) protocol from Onodi et al. [ 31]. First, rat serum was centrifuged 2500g for 15 min at 4 °C to remove the debris, and the supernatant was collected. The supernatant was filtered through a 0.22-μm low-binding PVDF filter (Millex-GV; Millipore, Tullagreen, Ireland). The sample was layered on top of an iodixanol (OptiPrep™, BioVision Inc., Milpitas, CA) density gradient. The iodixanol was diluted to 50, 30, and $10\%$ in 0.25M sucrose/10mM Tris buffer, and a discontinuous gradient was formed by layering 3.66 mL of the 50, 30, and $10\%$ iodixanol solutions in a 13-mL ultracentrifuge tube (Beckman Coulter, Pasadena, CA). The volume of the filtered serum sample was brought up to 1 mL with PBS if necessary and then layered onto the top of the discontinuous gradient. The samples were centrifuged in a SW41 Ti rotor for 24 h at 120,000g at 4 °C. Twelve 1-mL fractions of the density gradient layers were collected (F1–F12).
## EV miRNA isolation and expression
Total RNA was isolated from EVs as previously described [22] using the commercial miRCURY RNA Isolation Kit (Exiqon, Woburn, MA). miRNA concentrations were quantified with a small RNA kit on an Agilent Bioanalyzer (Agilent, Santa Clara, CA), followed by reverse transcription of miRNA performed with 10 ng of total RNA using the miRCURY LNA RT kit (Qiagen, Hilden, Germany). RT-qPCR reactions used the miRCURY LNA SYBR Green PCR kit (Qiagen) and the appropriate miRCURY LNA primer sets for the miRNAs of interest (Qiagen). miRNA expression was normalized to the expression of UniSp6, an exogenous spike-in that resembles miRNAs, using the −ΔCT method [32].
## EV protein isolation and protein expression
Total protein was isolated from EVs using Pierce RIPA lysis buffer with Halt protease and phosphatase inhibitor cocktail (Thermo Fisher, Waltham, MA), and protein concentration was determined using the Pierce BCA protein assay kit (Thermo Fisher). For Western blotting, samples were prepared in Laemmli buffer, boiled at 95 °C for 5 min, and 5 μg protein was loaded. Proteins were separated by SDS-PAGE using 4–$15\%$ TGX Gels (Criterion, Bio-Rad, Hercules, CA) by running at 200 V at room temperature. Proteins were transferred for 60 min at 100 V on ice onto a nitrocellulose membrane in $20\%$ methanol Tris-glycine buffer. The Revert Total Protein Stain Kit (Li-Cor Biosciences, Lincoln, NE) or Ponceau S solution (Thermo Fisher) was used to stain total protein, and the membranes were imaged to verify transfer efficiency and loading. The membranes were subsequently blocked in $5\%$ nonfat dry milk in Tris-buffered saline-Tween (TBS-T, $0.1\%$ Tween-20) for 1 h at room temperature, then incubated overnight at 4 °C in primary antibody (anti-CD63, EXOAB-CD63A-1; System Biosciences, Palo Alto, CA and anti-Apolipoprotein A1 (ApoA1, 701239; Thermo Fisher) at a 1:1000 dilution in $5\%$ nonfat dry milk in TBS-T. The membranes were then washed before incubation in goat anti-rabbit secondary antibodies (EXOAB-CD63A-1; System Biosciences) (1:10,000 dilution) for 1 h at room temperature. Blots were developed with enhanced chemiluminescence (Clarity Western ECL Substrate, Bio-Rad), imaged, and quantified with ImageJ (National Institutes of Health).
## Fluorescence correlation spectroscopy (FCS) of EVs
To assess α-sarcoglycan protein levels, Western blotting was performed as described above for CD63, and the membranes were incubated with anti-α-sarcoglycan antibody (Santa Cruz, SC-271321) (1:1000 dilution). To further determine the number of EVs that are positive for α-sarcoglycan, we used fluorescence correlation spectroscopy (FCS). FCS is a powerful technique that can quantitatively evaluate picomolar concentrations, with sensitivity that can be up to a single-molecule level [33, 34]. Specifically, an anti-α-sarcoglycan antibody was used (Santa Cruz, SC-271321). The antibody was first labeled with CF488 dye using antibody labeling kits (Mix-n-Stain, Biotium) following the manufacturer’s antibody labeling protocol. Fifty ng/mL CF488 labeled antibody was added to each EV sample and allowed to incubate for 60 min at room temperature. The vesicles were purified from free dye using a 5000-molecular weight cutoff size exclusion column (PD Minitrap G25, GE Healthcare) as described previously [35]. Briefly, the binding of fluorescently labeled anti-α-sarcoglycan antibody to EVs was confirmed via FCS based on their diffusion times. All FCS measurements were done as reported previously by Fu et al. [ 36]. Briefly, 40 μL of fluorescently labeled EVs were placed onto a coverslip mounted on an Olympus IX83 microscope equipped with a PicoQuant PicoHarp 300 time-correlated single photon counting (TCSPC) system. We employed a 488-nm laser (50 μW) to excite the fluorescent labels, and a 60× water immersion objective was used to focus this laser beam into the sample solution. Two avalanche photodiodes (APDs) were used for photon detection, and the signal was directed to a PicoHarp 300 TCSPC module controller. All measurements were performed 30 μm above the glass surface in the sample solution. For the unconjugated fluorophore, the fitted autocorrelation functions (ACF) yield a diffusion time (τD) of 0.21 ± 0.02 ms. A longer diffusion time of 2.5 ± 0.2 ms was observed for the CF-488-labeled anti-α-sarcoglycan antibody. The immunolabeled (anti-α-sarcoglycan-CF488 antibody) EVs exhibited a diffusion time of 32 ± 5 ms. In order to calculate the average number of immunolabeled (anti-α-sarcoglycan-CF488 antibody) EVs within the focal volume, the FCS focal volume was first calibrated using commercially available 0.1-μm tetra speck beads with a known diffusion constant and concentration. The number of vesicles per mL of solution was determined using NTA, and the number of labeled vesicles per mL was determined using FCS and the calibrated size of the focal volume.
## Mononuclear cell isolation and scRNA-seq
Cell isolations were performed as previously described in mice [37] and modified slightly for rats [28]. Briefly, the gastrocnemius muscles from WB and HS male rats were excised and placed in muscle dissociation media (MDM) (Hams F-10 (Gibco, USA), $10\%$ Horse Serum (Thermo Fisher), $1\%$ penicillin/streptomycin (Gibco), 800 U/ml Collagenase II (Gibco)), and minced using sterilized surgical equipment. The muscle homogenate was then incubated in MDM for 1 h at 37 °C with gentle agitation. Following incubation, samples underwent further incubation in 1000 U/ml Collagenase II (Gibco) and 11 U/ml dispase (Gibco) for 30 min at 37 °C. The single-cell suspension was passed through an 18-gauge needle approximately 10 times prior to 0.2-μm filtration. Single cells were incubated in propidium iodide to identify dying/dead cells for removal via fluorescence-activated cell sorting (Sony Biotechnology, USA). Single-cell suspensions from each group were added to a Chromium Controller (10X Genomics, USA) using the Single Cell 3’ Reagent Kit per manufacturer’s instructions and sequenced on an Illumina HiSeq platform (Novogene, USA), yielding 200 million reads/sample.
## Data processing and cell population annotation
scRNA-seq data were processed using the Partek Genomics Suite (Partek, USA) as previously described [28]. Briefly, following data quality control, samples were aligned to the rn6 genome and low-quality cells and/or reads were excluded based on the following criteria: mitochondrial reads exceeding $20\%$, an indication of doublets via read counts/cell, lowly expressed genes in only $0.01\%$ of total cells, and high expression of myofiber-related RNA resulting from muscle mincing. Following dimensionality reduction, graph-based clustering was used in combination with known muscle mononuclear cell-related gene markers for population annotation [11, 28, 38].
## Mononuclear cell EV-related gene expression
Bubble plots were generated using the Extracellular Vesicle Biogenesis GO term (http://www.informatics.jax.org/vocab/gene_ontology/GO:0140112) in combination with the identified mononuclear cell populations. Following the filtration of genes represented by the selected GO term (GO: 0140112), a bubble plot was made with the average expression of the gene of interest represented by heatmap, and the percent of cells expressing each gene represented by the size of the bubble. Cell populations are grouped by sample for population-specific comparison.
## Immunohistochemistry (IHC)
The muscles were cut on a cryostat at − 23 °C (7 μm), air-dried, and stored at − 20 °C. Slides were air-dried, rehydrated, and fixed in $4\%$ paraformaldehyde (PFA) for 20 min at the time of staining. For CD63/DAPI/laminin staining, sections were incubated with mouse anti-CD63 IgG1 antibody (1:100 dilution, ab108950, Abcam, Cambridge, UK) and rabbit anti-laminin IgG antibody (1:100 dilution, L9393, Sigma-Aldrich, St. Louis, MO) overnight at 4 °C. Slides were washed in PBS, then incubated with Alexa Fluor 488 goat anti-mouse IgG1 (1:250 dilution, A11001, Invitrogen, Waltham, MA) and Alexa Fluor 594 goat anti-rabbit IgG (1:250 dilution, A11012, Invitrogen) secondary antibodies for 1 h at room temperature. Slides were washed in PBS and mounted with VectaShield fluorescent mounting media with DAPI (H-1200-10, Vector Laboratories, Newark, CA). For CD9/DAPI/dystrophin staining, sections were incubated with rabbit anti-CD9 IgG (1:100 dilution, SA35-08, Invitrogen) and mouse anti-dystrophin IgG2b (1:250 dilution, 08168, Sigma-Aldrich) overnight, followed by incubation with Alexa Fluor 594 goat anti-rabbit IgG (1:250 dilution, A11012, Invitrogen) and Alexa Fluor 647 goat anti-mouse IgG2b (1:250 dilution, A32728, Invitrogen) for 1 h at room temperature. For CD81/DAPI/dystrophin staining, sections were incubated with rabbit anti-CD81(1:100 dilution, SN206-01, Novus Biologicals, Centennial, CO) and mouse anti-dystrophin IgG2b (1:250 dilution, 08168, Sigma-Aldrich) overnight, followed by incubation with Alexa Fluor 594 goat anti-rabbit IgG (1:250 dilution, A11012, Invitrogen) and Alexa Fluor 647 goat anti-mouse IgG2b (1:250 dilution, A32728, Invitrogen) for 1 h at room temperature. For Pax7/CD9/DAPI/WGA staining, sections were subjected to epitope retrieval using sodium citrate (10 mM, pH 6.5) at 92 °C, followed by blocking of endogenous peroxidase activity with $3\%$ hydrogen peroxide in PBS. Sections were incubated overnight in mouse anti-Pax7 IgG1 (1:100 dilution, Developmental Studies Hybridoma Bank, Iowa City, IA) and rabbit anti-CD9 IgG (1:100 dilution, SA35-08, Invitrogen), followed by incubation in goat anti-mouse biotin-conjugated secondary antibody (dilution 1:1,000, 115-065-205; Jackson ImmunoResearch, West Grove, PA) and Alexa Fluor 647 goat anti-rabbit IgG (1:250 dilution, A32733, Invitrogen) for 1 h at room temperature. Next, sections were incubated with streptavidin-HRP (1:500 dilution, S-911, Invitrogen) and Texas Red-conjugated Wheat Germ Agglutinin (WGA) (1:50 dilution, W21405, Invitrogen) at room temperature for 1 h, before incubation in Tyramide Signal Amplification (TSA) Alexa Fluor 488 (1:500 dilution, B40953, Invitrogen). Sections were mounted with VectaShield fluorescent mounting media with DAPI (H-1200-10, Vector Laboratories).
Images were captured with a Zeiss upright microscope (AxioImager M1, Oberkochen, Germany). To quantify the percentage of nuclei (DAPI+) expressing CD63, MyoVision software was used for automated analysis of nuclear density in cross-sections [39], and nuclei-expressing CD63 (identified as DAPI+/CD63+ events) were counted manually in a blinded manner by the same assessor for all sections using the Zen Blue software.
## Statistical analysis
Differences between the two groups (HS vs WB) were analyzed by unpaired Student’s t-tests. When comparing 4 groups, a one-way ANOVA was used, with Tukey’s multiple comparisons test for post hoc analysis. A two-way ANOVA was used to assess the differences in particle concentrations between WB and HS across fractions of the density gradient. Paired t-tests were used to examine the changes in measures from pre- to post-immobilization in human samples. All statistical analyses were performed in GraphPad Prism (v7.00, GraphPad Software, La Jolla, CA), and statistical significance was set at an α < 0.05.
## Skeletal muscle-specific EV markers
Using Western blotting, α-sarcoglycan was detected in the heart and skeletal muscle, but not in other organs from the rat (Fig. 1A) and was absent from rat serum EVs (Fig. 1B). We were also unable to detect α-sarcoglycan in serum EVs from WB or HS rats or from EVs derived from ex vivo skeletal muscle experiments using FCS methods (Table 1). The only samples in which we detected EVs positive for α-sarcoglycan using FCS were EVs collected from conditioned media from L6 and C2C12 myotubes ($13.8\%$ and $28.6\%$, respectively) (Fig. 1C and Table 1).Fig. 1α-*Sarcoglycan is* expressed specifically in the muscle tissues and is not detectable in the serum or muscle-derived EVs. A Representative western blot image showing protein expression of α-sarcoglycan in rat tissues. B Western blot of α-sarcoglycan in rat serum EVs and rat skeletal muscle (positive control). Transfer efficiency verified by Ponceau S staining. C The normalized autocorrelations for the fluorescently labeled anti-α-sarcoglycan antibody and the immunofluorescently labeled EVs pooled from L6 myotubes ($$n = 3$$ biological replicates) and C2C12 myotubes ($$n = 3$$ biological replicates)Table 1Proportion of α-sarcoglycan positive EVs using FCSEV sampleParticles (EVs)/ml% of EVs positive for α-sarcoglycanWB rat serum2.80E+10No bindingWB rat serum3.00E+10No bindingHS rat serum4.30E+10No bindingHS rat serum4.10E+10No bindingL6 myotubes2.50E+$1113.8\%$C2C12 myotubes1.40E+$1128.6\%$Ex vivo muscle1.10E+10No bindingWB weight bearing, HS hindlimb suspension
## Skeletal muscle-specific miRNAs
As expected, miR-1 was detected at high abundance in the skeletal muscle, but not in the liver or kidney of rats (Fig. 2A). miR-1 was also detected at a higher abundance in the skeletal muscle than miR-23a-3p, miR-26a-5p, miR-27a-3p, and miR-29a-3p (Fig. 2A). miR-1 was not detected in serum EVs from rats or from humans under basal conditions in contrast to other miRs, such as miR23-3p and miR-29a (Fig. 2B, C).Fig. 2“Muscle-specific” miR-1 is not detectable in serum EVs under basal conditions in rats or humans. A miRNA abundance of selected miRNAs across the skeletal muscle, liver, and kidney from rats ($$n = 5$$). B miRNA abundance of selected miRNAs in serum EVs from rat serum ($$n = 12$$). C miRNA abundance of the selected miRNAs in serum EVs from human serum ($$n = 6$$)
## Single-cell analysis of EV biogenesis factors
Most EV biogenesis factors extracted from the Gene Ontology annotation “extracellular vesicle biogenesis” (GO: 0140112) had very low expression across all cell types in gastrocnemius muscle (Fig. 3). The tetraspanin CD63 was most highly expressed in most cell types, while CD9 was highly expressed only in some cell types, such as FAPS and tenocytes. CD81, a tetraspanin often used as a specific marker for EVs, however, was not highly expressed in cells present in skeletal muscle. It is also noted that differences between WB and HS exist in the expression of the abundant tetraspanins in some cell types, but not others. For example, CD9 is higher in HS than in WB in neutrophils, while CD63 is not different, but this is not observed in pericytes (Fig. 3).Fig. 3Gene expression of EV biogenesis factors across cell types in the skeletal muscle. Bubble plot of genes selected for supervised-based classification corresponding to each identified cell population from weight bearing (WB) and hindlimb suspended (HS) muscle. The average gene expression is denoted by heat map and the non-zero percent of cells expressing the gene is denoted by bubble size. FAPs, fibro/adipogenic progenitors; MCs, mast cells; TCs, T cells; SCs, satellite cells; APCs, antigen-presenting cells; NKs, natural killer cells; ECs, endothelial cells; SMCs, smooth muscle cells
## Cellular localization of tetraspanins in the skeletal muscle
We further investigated the CD63 expression at the protein level via Western blotting and found that it was undetectable in rat soleus muscle (Fig. 4A) despite its readily detectable levels at the mRNA level (Fig. 3). However, using IHC, we showed that CD63 protein is not detected in myofibers of the soleus but is highly abundant in a subset of cells residing in the muscle interstitial space (Fig. 4B–D, white arrows). The CD63+ cells reside outside the laminin border and are therefore not satellite cells. We found that, on average, CD63+/DAPI+ nuclei made up 2.30 ± $0.48\%$ of the total nuclei. Fig. 4CD63 protein is not detected in myofibers but is expressed in a small subset of cells residing in the muscle interstitial space. A Representative western blot of total protein and CD63 in the skeletal muscle. B Cross-section of the soleus muscle from a control F344/BN rat showing laminin and CD63+ cells. White arrows indicate CD63+ cells. C Image of the small subset of CD63+ cells in the interstitial space of the skeletal muscle. D Close-up image of the CD63+ cells in the interstitial space of the skeletal muscle Likewise, CD9 was not detected in the myofibers but was detected primarily in mononuclear cells residing in the extracellular space, albeit, at a much higher frequency than CD63 (Fig. 5A, white arrows). There was also a high amount of CD9 staining surrounding the blood vessels, nerves, and muscle spindles (Fig. 5B). We also observed a large amount of CD9 within the extracellular compartments around necrotic muscle fibers (Fig. 5C). Co-staining with Pax7 showed only a small subset of CD9+ satellite cells (Fig. 5D, white arrowheads). Most satellite cells did not express CD9 (Fig. 5D, white arrows). Notably, these results are consistent with the scRNA-seq analyses demonstrating low expression of CD9 by satellite cells (Fig. 3). Finally, CD81 exhibited a strikingly different distribution in rat skeletal muscle compared with CD63 and CD9. CD81 protein was abundant in the extracellular space (gray arrows) and residing mononuclear cells (white arrows) as well as in high abundance surrounding some of the myonuclei (white arrowheads) (Fig. 6A, D). There was also a significant overlap of CD81 with the dystrophin borders of myofibers suggesting the presence of CD81 protein in myofiber membranes (Fig. 6A–C).Fig. 5CD9 protein is not detected in myofibers but is abundantly expressed in the blood vessels, nerves, and cells residing in the muscle interstitial space. A Cross-section of the soleus muscle from a control F344/BN rat showing dystrophin and CD9+ cells. The magnified image shows CD9+ cells in the interstitial space (white arrows). B Image of the CD9 expression surrounding the blood vessels, nerves, and muscle spindles. C Image of the CD9 expression around the necrotic muscle fibers. D Staining for WGA as well as CD9+ and Pax7+ cells. Pax7+/CD9+ cells indicated by the white arrowhead, Pax7+/CD9− cells identified by white arrows, and Pax7−/CD9+ cells indicated by gray arrowsFig. 6CD81 protein is primarily detected in the interstitial space with some expression near the myonuclei and on myofiber membranes. A Cross-section of the soleus muscle from a control F344/BN rat showing dystrophin and CD81+ cells. The magnified image shows the CD81 protein in the extracellular space (gray arrows), in the mononuclear cells (white arrows), and in the surrounding myonuclei (white arrowheads). B Image of the CD81 expression in the myofiber membrane. C The same image as B, with dystrophin border shown. D Close-up image demonstrating the expression of CD81
## Serum EV concentrations after hindlimb suspension in rats
Separation of CD63-positive EVs from other serum components using an iodixanol density gradient revealed that CD63-positive EVs are present in only a few fractions (F5–F8, Fig. 7A), and these particular fractions only make up a very small percentage of the total vesicle number (Fig. 7B). Indeed, the largest percentage of detected particles is positive for ApoA1 (F1–F3, Fig. 7A, B). A high particle number was observed in fraction 1 of the iodixanol density gradient from WB serum, but there were no differences in the concentration of non-EV particles (ApoA1-rich fractions 1–4 of gradient) or CD63-positive EVs (fractions 5–7 of gradient) when comparing HS serum with WB serum (Fig. 7C–E).Fig. 7EVs are a small portion of particles in the serum, and the concentration is not changed with muscle atrophy in rats. A Western blot showing the separation of lipoprotein particles and CD63+ EVs in the serum using an iodixanol density gradient. B NTA quantification of the particle concentration from each fraction collected from the iodixanol density gradient shown in A ($$n = 3$$ WB and $$n = 3$$ HS). C Magnification of the NTA concentration data shown in B ($$n = 3$$ WB and $$n = 3$$ HS). D Summed particle concentration of fractions 1 through 4 of the density gradient. E Summed particle concentration of fractions 5 through 7 of the density gradient
## Ex vivo release of EVs from rat soleus muscle
Rat soleus muscle incubated in KHB released particles into the buffer in the expected size range of small EVs containing a mix of exosomes and other microvesicles with a mode in size of approximately 120 nm (Fig. 8A). In contrast to the in vivo findings (Fig. 7D, E), the number of EVs in KHB from the soleus muscles that had undergone hindlimb suspension for 24 h or 7 days was significantly higher than WB (Fig. 8B) ($$p \leq 0.01$$ for both).Fig. 8The muscle releases EVs ex vivo, and the release of EVs is elevated with disuse atrophy. A Size distribution of EVs collected in KHB measured by NTA. B Concentration of EVs in the KHB after incubation of a rat soleus muscle excised from WB ($$n = 7$$), HS 4 h ($$n = 4$$), HS 24 h ($$n = 4$$), and HS 7 days ($$n = 5$$) rats. * $p \leq 0.05$ compared with WB
## Serum and muscle EV concentrations after bed rest in human participants
Serum EV concentration was elevated by $31\%$ in human participants following 5 days of bed rest (Fig. 9A). There was a significant decrease in CD63 ($$p \leq 0.04$$) and CD9 ($$p \leq 0.001$$) mRNA abundance and no change in CD81 in the vastus lateralis muscle in response to bed rest (Fig. 9B).Fig. 9Bed rest for 5 days in human subjects elevates serum EV concentration but lowers mRNA abundance of vastus lateralis CD63 and CD9. A Serum EV concentration before and after 5 days of bed rest in human subjects ($$n = 13$$). B mRNA abundance of CD63, CD9, and CD81 in human vastus lateralis muscle ($$n = 9$$) before and after 5 days of bed rest. * $p \leq 0.05$ compared to pre, ***$p \leq 0.001$ compared to pre
## Muscle-specific EV markers
The primary goal of this study was to explore the expression and localization of markers related to EV assembly and secretion across cell types in the skeletal muscle. The lack of tissue-specific EV markers has made it impossible to determine whether skeletal muscle-derived EVs actually reach circulation and have systemic actions [40]. Guescini et al. first reported that EVs positive for α-sarcoglycan, an integral membrane protein localized to the sarcolemma of skeletal muscle, could be detected in 1–$5\%$ of EVs found in the plasma [16]. Although a number of other studies also reported the detection of α-sarcoglycan in circulation in vivo [15, 41, 42], studies using gold-standard, iodixanol-based density gradients for EV isolation fail to detect any α-sarcoglycan in plasma EVs, even after acute bouts of exercise [18, 43, 44]. Similarly, in the present study, we were unable to detect α-sarcoglycan in serum EVs by Western blot. Moreover, α-sarcoglycan was also undetected in serum EVs using FCS, which employs maximum sensitivity and can detect molecules at picomolar concentrations. Indeed, α-sarcoglycan was only detected in EVs in cell culture media from L6 and C2C12 myotubes and even then in only a small percentage. Overall, there is not enough evidence to suggest that α-sarcoglycan can be used as a skeletal muscle-specific EV marker in vivo. Beyond the lack of detection in pure serum or plasma-derived EV preparations, it is critical to note that α-sarcoglycan is not exclusively expressed in skeletal muscle, but is also expressed in cardiac muscle and in the lung, further questioning the designation of α-sarcoglycan-positive EVs as skeletal muscle EVs [45]. Similarly, APT2A1, β-enolase, and desmin have been suggested as marker proteins to identify skeletal muscle-derived EVs in vivo [46]; however, these proteins can be even more abundant in cardiac muscle tissue than skeletal muscle [47–49]. To address issues of tissue specificity, Estrada et al. recently developed a skeletal muscle myofiber-specific fluorescent reporter mouse [50]. Using this model, the researchers identified that myofiber-derived EVs do in fact reach circulation in vivo, and up to $5\%$ of circulating EVs may be derived from skeletal muscle myofibers [50]. However, in addition to tissue-specific markers, the lack of cell type-specific EV markers further limits the ability to distinguish EVs from the distinct subsets of cells residing in skeletal muscle. For example, although some groups have used markers such as platelet-derived growth factor receptor A (PDGFRα) to isolate fibro-adipogenic progenitor cells (FAPs) in the muscle [51], other cell types, including fibroblasts and smooth muscle cells, also express this protein [52]. Thus, the poor characterization of tissue- or cell-specific EV markers suggests that caution should be exercised in the interpretation of results when these nonspecific markers are used.
Another concept for tracking circulating EVs that may be skeletal muscle-derived is an analysis of myomiRs, muscle-enriched miRNAs, such as miR-1, which are expressed in much greater abundance in the muscle in comparison with other tissues [53, 54]. In this study, we were unable to detect miR-1 in EVs from either rat or human serum under resting, basal conditions. In contrast, other studies which used different EV RNA isolation methods and reverse transcription reactions for miR-1 have reported the detection of serum/plasma EV miR-1 in humans and mice [55, 56]. Notably, miR-1 abundance in circulating EVs has been reported to be low at rest and dramatically increased after mechanical overload in mice, as well as following an acute bout of high-intensity resistance exercise in humans [56]. This is in line with several studies which have demonstrated higher levels of circulating miR-1 following different exercise modalities (reviewed in [57–59]). Thus, miR-1 may be a useful circulating miRNA signature for response to acute or chronic exercise, but its utility may be less relevant for tracking EVs in sedentary or atrophic conditions. However, it is important to note that similar to α-sarcoglycan, miR-1 expression is not limited to skeletal muscle and is also expressed in other organs including cardiac tissue. In fact, increases in circulating miR-1 following exercise have been suggested to possibly be due to active heart remodeling rather than skeletal muscle secretion [60].
## Skeletal muscle expression of EV biogenesis factors
The tetraspanins CD63, CD9, and CD81 are most commonly used as ubiquitous markers of EVs to demonstrate isolation or enrichment [61]. Despite early research suggesting universal enrichment of the three tetraspanins in EVs across cell types, recent data suggest that there may be heterogeneous tetraspanin expression across EVs [62–64]. Our scRNA-seq data show large variations in tetraspanin and other EV biogenesis marker mRNA expression across different cell types in skeletal muscle. Variation in the expression of tetraspanins is thought to reflect distinct EV subpopulations, which may have functional differences. The expression of EV biogenesis-related genes was not significantly altered by HS, and the only notable changes induced by HS were small increases in CD9 expression of neutrophils and CD63 expression of macrophages, which may be related to stress and immune dysregulation that can accompany exposure to hindlimb unloading [65, 66]. Interestingly, of the three tetraspanins assessed, CD63 showed the highest mRNA expression across all cell types, including satellite cells. Despite this high expression, we were unable to detect CD63 protein in skeletal muscle homogenates. This finding prompted immunohistochemical staining of the tetraspanins in the skeletal muscle cross-sections, which demonstrated hardly any detection within myofibers and instead accumulation within the interstitium. This is consistent with the recent findings of Watanabe et al., showing EVs concentrated in the muscle interstitium, attached to extracellular matrix (ECM)-like structures by transmission electron microscopy [46]. Within muscle, satellite cell-derived EVs have been previously shown to influence early phases of myofiber growth in response to overload by regulating extracellular matrix (ECM)-related factors in both the myofiber and FAPs [67, 68]. These findings suggest that there is an interaction between EVs and ECM in the muscle, and in vitro and ex vivo analyses of EV secretion that fail to recapitulate this muscle microenvironment may not reflect in vivo conditions.
## Effects of mechanical unloading on skeletal muscle EVs
In the present study, we found no differences in serum EV concentrations following HS in rats. Although there tended to be a difference in fractions 1–4 of the iodixanol gradient between WB and HS rats, Western blot analysis of ApoA1 showed contamination of these fractions with lipoprotein particles. This finding suggests that future research should include ApoA1 analysis to indicate contamination of lipoproteins in EV-enriched fractions. In contrast to the in vivo findings, however, isolated muscles from rats that underwent HS for 24 h or 7 days secreted a greater amount of EVs into the media compared to WB muscles. Thus, ex vivo measurements of EV secretion may not reflect in vivo changes in the serum. One possibility for this difference is that EVs represent only a small portion of particles in the serum, and those that may be muscle-derived represent an even smaller portion. Therefore, changes due to muscle mechanical unloading may be harder to detect from a large pool of circulating particles. Alternatively, EVs released from the muscle ex vivo into the medium of an isolated muscle may not make it into the circulation in vivo. This is further supported by our finding that muscle-specific miRNAs, such as miR-1, are not detected in serum EVs in rats or humans.
Interestingly, 5 days of bed rest in human participants led to elevated serum EV concentrations in vivo. Notably, bed rest better represents physical inactivity compared with HS and results in whole-body physiological responses that affect different organ systems, including the cardiovascular, pulmonary, hepatic, and gastrointestinal systems [69]. The increase in EVs in human serum following bed rest may be due to the release from other tissues besides the skeletal muscle. In addition to the changes in circulating EV levels, we also found a reduction in the mRNA abundance of CD63 and CD9 in human muscles following bed rest. While these changes suggest that mechanical unloading may modulate skeletal muscle EV concentrations and composition, the clinical relevance of these changes remains unclear.
There has been a growing interest in studying the effects of different stimuli on EV release. For example, acute exercise has been shown to elevate the muscle mRNA content of CD9, CD63, and CD81 [18, 70]. Although tetraspanins have mostly been applied as markers of EVs, increasing evidence suggests that these proteins may also influence cellular communication and regulate aspects including cellular metabolism [71]. Importantly, tetraspanins have been shown to interact with immune receptors, which can lead to immune cell signaling and modulation of immune cell adhesion and proliferation [72]. In rat skeletal muscle, signaling via the CXCR4 receptor has been shown to improve skeletal muscle regeneration by upregulating CD9 expression and increasing stem cell mobilization to injured muscles [73]. Similarly, increases in CD9 and CD81 mRNA expression have also been shown to accompany muscle regeneration in rats [74], and mice lacking either CD9 or CD81 show abnormal muscle regeneration due to altered myogenic cell fusion [75]. However, it is not known whether tetraspanin expression by specific cell types mediates their function, and our scRNA-seq and immunohistochemistry data suggest that tetraspanins are, in fact, expressed by various cell types within the skeletal muscle. Further work should be done to determine the mechanisms and effects of CD63 and CD9 reductions in muscle disuse atrophy. Future studies are also required to unravel the effects of muscle disuse atrophy on EV concentrations and composition, as well as their role in tissue crosstalk and mode of action in target cells. Identification of EV signaling mechanisms associated with disuse atrophy may translate into future therapeutic applications.
## Conclusions
In conclusion, we provide evidence that traditional markers used to demarcate muscle-specific EVs, including α-sarcoglycan and miR-1, are not reliable in vivo. Furthermore, our scRNA-seq and IHC data demonstrate that EV biogenesis factors including the tetraspanins CD63, CD9, and CD81 are expressed by a variety of cell types in skeletal muscle and that the tetraspanins accumulate within the muscle interstitial space. Lastly, our studies demonstrate the importance of methodological guidelines, such as the separation of EVs from non-EV protein and lipoprotein contaminants as well as the need for caution in interpreting ex vivo findings on EV release by the muscle.
## Authors’ information
Current affiliations: ZRH is currently at the Department of Physical Medicine and Rehabilitation, Harvard Medical School, Boston, MA, and Department of Medicine, Division of Geriatric Medicine, University of Pittsburgh, Pittsburgh, PA.
IJV is currently at the Department of Nutrition and Health Sciences, University of Nebraska-Lincoln, Lincoln, NE.
## References
1. Darkwah S, Park EJ, Myint PK, Ito A, Appiah MG, Obeng G. **Potential roles of muscle-derived extracellular vesicles in remodeling cellular microenvironment: proposed implications of the exercise-induced myokine, Irisin**. *Front Cell Dev Biol* (2021.0) **9** 634853. DOI: 10.3389/fcell.2021.634853
2. Tkach M, Thery C. **Communication by extracellular vesicles: where we are and where we need to go**. *Cell.* (2016.0) **164** 1226-1232. DOI: 10.1016/j.cell.2016.01.043
3. Thery C, Witwer KW, Aikawa E, Alcaraz MJ, Anderson JD, Andriantsitohaina R. **Minimal information for studies of extracellular vesicles 2018 (MISEV2018): a position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines**. *J Extracell Vesicles* (2018.0) **7** 1535750. DOI: 10.1080/20013078.2018.1535750
4. Lai RC, Tan SS, Yeo RWY, Choo ABH, Reiner AT, Su Y. **MSC secretes at least 3 EV types each with a unique permutation of membrane lipid, protein and RNA**. *J Extracell Vesicles* (2016.0) **5** 29828. DOI: 10.3402/jev.v5.29828
5. Whitham M, Parker BL, Friedrichsen M, Hingst JR, Hjorth M, Hughes WE. **Extracellular vesicles provide a means for tissue crosstalk during exercise**. *Cell Metab* (2018.0) **27** 237-51 e4. DOI: 10.1016/j.cmet.2017.12.001
6. Valadi H, Ekstrom K, Bossios A, Sjostrand M, Lee JJ, Lotvall JO. **Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells**. *Nat Cell Biol* (2007.0) **9** 654-U72. DOI: 10.1038/ncb1596
7. Pedersen BK. **Muscle as a secretory organ**. *Compr Physiol* (2013.0) **3** 1337-1362. DOI: 10.1002/cphy.c120033
8. Pedersen BK, Febbraio MA. **Muscles, exercise and obesity: skeletal muscle as a secretory organ**. *Nat Rev Endocrinol* (2012.0) **8** 457-465. DOI: 10.1038/nrendo.2012.49
9. Rogers MA, Aikawa E. **MicroRNA extracellular vesicle stowaways in cell-cell communication and organ crosstalk**. *Arterioscler Thromb Vasc Biol* (2019.0) **39** 2448-2450. DOI: 10.1161/ATVBAHA.119.313533
10. Chatterjee V, Yang X, Ma Y, Wu MH, Yuan SY. **Extracellular vesicles: new players in regulating vascular barrier function**. *Am J Physiol Heart Circ Physiol* (2020.0) **319** H1181-H1H96. DOI: 10.1152/ajpheart.00579.2020
11. Rubenstein AB, Smith GR, Raue U, Begue G, Minchev K, Ruf-Zamojski F. **Single-cell transcriptional profiles in human skeletal muscle**. *Sci Rep* (2020.0) **10** 229. DOI: 10.1038/s41598-019-57110-6
12. Fruhbeis C, Helmig S, Tug S, Simon P, Kramer-Albers EM. **Physical exercise induces rapid release of small extracellular vesicles into the circulation**. *J Extracell Vesicles* (2015.0) **4** 28239. DOI: 10.3402/jev.v4.28239
13. Bei Y, Xu T, Lv D, Yu P, Xu J, Che L. **Exercise-induced circulating extracellular vesicles protect against cardiac ischemia-reperfusion injury**. *Basic Res Cardiol* (2017.0) **112** 38. DOI: 10.1007/s00395-017-0628-z
14. Nielsen MH, Sabaratnam R, Pedersen AJT, Hojlund K, Handberg A. **Acute exercise increases plasma levels of muscle-derived microvesicles carrying fatty acid transport proteins**. *J Clin Endocrinol Metab* (2019.0) **104** 4804-4814. DOI: 10.1210/jc.2018-02547
15. Rigamonti AE, Bollati V, Pergoli L, Iodice S, De Col A, Tamini S. **Effects of an acute bout of exercise on circulating extracellular vesicles: tissue-, sex-, and BMI-related differences**. *Int J Obes* (2020.0) **44** 1108-1118. DOI: 10.1038/s41366-019-0460-7
16. Guescini M, Canonico B, Lucertini F, Maggio S, Annibalini G, Barbieri E. **Muscle releases alpha-sarcoglycan positive extracellular vesicles carrying miRNAs in the bloodstream**. *PLoS One* (2015.0) **10** e0125094. DOI: 10.1371/journal.pone.0125094
17. Oliveira GP, Porto WF, Palu CC, Pereira LM, Petriz B, Almeida JA. **Effects of acute aerobic exercise on rats serum extracellular vesicles diameter, concentration and small RNAs content**. *Front Physiol* (2018.0) **9** 532. DOI: 10.3389/fphys.2018.00532
18. Brahmer A, Neuberger E, Esch-Heisser L, Haller N, Jorgensen MM, Baek R. **Platelets, endothelial cells and leukocytes contribute to the exercise-triggered release of extracellular vesicles into the circulation**. *J Extracell Vesicles* (2019.0) **8** 1615820. DOI: 10.1080/20013078.2019.1615820
19. Wilburn D, Ismaeel A, Machek S, Fletcher E, Koutakis P. **Shared and distinct mechanisms of skeletal muscle atrophy: a narrative review**. *Ageing Res Rev* (2021.0) **71** 101463. DOI: 10.1016/j.arr.2021.101463
20. Aoi W, Sakuma K. **Does regulation of skeletal muscle function involve circulating microRNAs?**. *Front Physiol* (2014.0) **5** 39. DOI: 10.3389/fphys.2014.00039
21. Hudson MB, Woodworth-Hobbs ME, Zheng B, Rahnert JA, Blount MA, Gooch JL. **miR-23a is decreased during muscle atrophy by a mechanism that includes calcineurin signaling and exosome-mediated export**. *Am J Physiol Cell Physiol* (2014.0) **306** C551-C558. DOI: 10.1152/ajpcell.00266.2013
22. Van Pelt DW, Vechetti IJ, Lawrence MM, Van Pelt KL, Patel P, Miller BF. **Serum extracellular vesicle miR-203a-3p content is associated with skeletal muscle mass and protein turnover during disuse atrophy and regrowth**. *Am J Physiol Cell Physiol* (2020.0) **319** C419-CC31. DOI: 10.1152/ajpcell.00223.2020
23. Ferry AL, Vanderklish PW, Dupont-Versteegden EE. **Enhanced survival of skeletal muscle myoblasts in response to overexpression of cold shock protein RBM3**. *Am J Physiol Cell Physiol* (2011.0) **301** C392-C402. DOI: 10.1152/ajpcell.00098.2011
24. Taylor JM, Dupont-Versteegden EE, Davies JD, Hassell JA, Houle JD, Gurley CM. **A role for the ETS domain transcription factor PEA3 in myogenic differentiation**. *Mol Cell Biol* (1997.0) **17** 5550-5558. DOI: 10.1128/MCB.17.9.5550
25. Stelzer JE, Widrick JJ. **Effect of hindlimb suspension on the functional properties of slow and fast soleus fibers from three strains of mice**. *J Appl Physiol (1985)* (2003.0) **95** 2425-2433. DOI: 10.1152/japplphysiol.01091.2002
26. Musacchia XJ, Steffen JM, Deavers DR. **Rat hindlimb muscle responses to suspension hypokinesia/hypodynamia**. *Aviat Space Environ Med* (1983.0) **54** 1015-1020. PMID: 6651726
27. Alkhateeb H, Chabowski A, Bonen A. **Viability of the isolated soleus muscle during long-term incubation**. *Appl Physiol Nutr Metab* (2006.0) **31** 467-476. DOI: 10.1139/h06-022
28. Hettinger ZR, Wen Y, Peck BD, Hamagata K, Confides AL, Van Pelt DW. **Mechanotherapy reprograms aged muscle stromal cells to remodel the extracellular matrix during recovery from disuse**. *Function (Oxf)* (2022.0) **3** zqac015. DOI: 10.1093/function/zqac015
29. Mahmassani ZS, Reidy PT, McKenzie AI, Stubben C, Howard MT, Drummond MJ. **Age-dependent skeletal muscle transcriptome response to bed rest-induced atrophy**. *J Appl Physiol (1985)* (2019.0) **126** 894-902. DOI: 10.1152/japplphysiol.00811.2018
30. Bachurski D, Schuldner M, Nguyen PH, Malz A, Reiners KS, Grenzi PC. **Extracellular vesicle measurements with nanoparticle tracking analysis - an accuracy and repeatability comparison between NanoSight NS300 and ZetaView**. *J Extracell Vesicles* (2019.0) **8** 1596016. DOI: 10.1080/20013078.2019.1596016
31. Onodi Z, Pelyhe C, Terezia Nagy C, Brenner GB, Almasi L, Kittel A. **Isolation of high-purity extracellular vesicles by the combination of iodixanol density gradient ultracentrifugation and bind-elute chromatography from blood plasma**. *Front Physiol* (2018.0) **9** 1479. DOI: 10.3389/fphys.2018.01479
32. Schmittgen TD, Livak KJ. **Analyzing real-time PCR data by the comparative C(T) method**. *Nat Protoc* (2008.0) **3** 1101-1108. DOI: 10.1038/nprot.2008.73
33. Chen HM, Farkas ER, Webb WW. **In vivo applications of fluorescence correlation spectroscopy**. *Methods Cell Biol* (2008.0) **89** 3-35. DOI: 10.1016/S0091-679X(08)00601-8
34. Yu L, Lei YZ, Ma Y, Liu M, Zheng JJ, Dan D. **A comprehensive review of fluorescence correlation spectroscopy**. *Front Phys Lausanne* (2021.0) **9** 644450. DOI: 10.3389/fphy.2021.644450
35. Neupane KR, McCorkle JR, Kopper TJ, Lakes JE, Aryal SP, Abdullah M. **Macrophage-engineered vesicles for therapeutic delivery and bidirectional reprogramming of immune cell polarization**. *ACS Omega* (2021.0) **6** 3847-3857. DOI: 10.1021/acsomega.0c05632
36. Fu X, Song Y, Masud A, Nuti K, DeRouchey JE, Richards CI. **High-throughput fluorescence correlation spectroscopy enables analysis of surface components of cell-derived vesicles**. *Anal Bioanal Chem* (2020.0) **412** 2589-2597. DOI: 10.1007/s00216-020-02485-z
37. Liu L, Cheung TH, Charville GW, Rando TA. **Isolation of skeletal muscle stem cells by fluorescence-activated cell sorting**. *Nat Protoc* (2015.0) **10** 1612-1624. DOI: 10.1038/nprot.2015.110
38. Oprescu SN, Yue F, Qiu J, Brito LF, Kuang S. **Temporal dynamics and heterogeneity of cell populations during skeletal muscle regeneration**. *iScience* (2020.0) **23** 100993. DOI: 10.1016/j.isci.2020.100993
39. Wen Y, Murach KA, Vechetti IJ, Fry CS, Vickery C, Peterson CA. **MyoVision: software for automated high-content analysis of skeletal muscle immunohistochemistry**. *J Appl Physiol (1985)* (2018.0) **124** 40-51. DOI: 10.1152/japplphysiol.00762.2017
40. Rome S, Forterre A, Mizgier ML, Bouzakri K. **Skeletal muscle-released extracellular vesicles: state of the art**. *Front Physiol* (2019.0) **10** 929. DOI: 10.3389/fphys.2019.00929
41. Fulzele S, Mendhe B, Khayrullin A, Johnson M, Kaiser H, Liu Y. **Muscle-derived miR-34a increases with age in circulating extracellular vesicles and induces senescence of bone marrow stem cells**. *Aging (Albany NY)* (2019.0) **11** 1791-1803. DOI: 10.18632/aging.101874
42. Conkright WR, Beckner ME, Sahu A, Mi Q, Clemens ZJ, Lovalekar M. **Men and women display distinct extracellular vesicle biomarker signatures in response to military operational stress**. *J Appl Physiol (1985)* (2022.0) **132** 1125-1136. DOI: 10.1152/japplphysiol.00664.2021
43. Cantin R, Diou J, Belanger D, Tremblay AM, Gilbert C. **Discrimination between exosomes and HIV-1: purification of both vesicles from cell-free supernatants**. *J Immunol Methods* (2008.0) **338** 21-30. DOI: 10.1016/j.jim.2008.07.007
44. Van Deun J, Mestdagh P, Sormunen R, Cocquyt V, Vermaelen K, Vandesompele J. **The impact of disparate isolation methods for extracellular vesicles on downstream RNA profiling**. *J Extracell Vesicles* (2014.0) **3** 24858. DOI: 10.3402/jev.v3.24858
45. Diniz G, Tosun Yildirim H, Gokben S, Serdaroglu G, Hazan F, Yararbas K. **Concomitant alpha- and gamma-sarcoglycan deficiencies in a Turkish boy with a novel deletion in the alpha-sarcoglycan gene**. *Case Rep Genet* (2014.0) **2014** 248561. PMID: 25050186
46. 46.Watanabe S, Sudo Y, Makino T, Kimura S, Tomita K, Noguchi M, et al. Skeletal muscle releases extracellular vesicles with distinct protein and miRNA signatures that function in the muscle microenvironment. PNAS Nexus. 2022;1(4):pgac173.
47. Goldfarb LG, Dalakas MC. **Tragedy in a heartbeat: malfunctioning desmin causes skeletal and cardiac muscle disease**. *J Clin Invest* (2009.0) **119** 1806-1813. DOI: 10.1172/JCI38027
48. Nomura M, Kato K, Nagasaka A, Shiga Y, Miyagi Y, Fukui R. **Serum beta-enolase in acute myocardial infarction**. *Br Heart J* (1987.0) **58** 29-33. DOI: 10.1136/hrt.58.1.29
49. Liu G, Li SQ, Hu PP, Tong XY. **Altered sarco(endo)plasmic reticulum calcium adenosine triphosphatase 2a content: targets for heart failure therapy**. *Diab Vasc Dis Res* (2018.0) **15** 322-335. DOI: 10.1177/1479164118774313
50. Estrada AL, Valenti ZJ, Hehn G, Amorese AJ, Williams NS, Balestrieri NP. **Extracellular vesicle secretion is tissue-dependent ex vivo and skeletal muscle myofiber extracellular vesicles reach the circulation in vivo**. *Am J Physiol Cell Physiol* (2022.0) **322** C246-CC59. DOI: 10.1152/ajpcell.00580.2020
51. Parker E, Mendhe B, Ruan L, Marshall B, Zhi W, Liu Y. **MicroRNA cargo of extracellular vesicles released by skeletal muscle fibro-adipogenic progenitor cells is significantly altered with disuse atrophy and IL-1beta deficiency**. *Physiol Genomics* (2022.0) **54** 296-304. DOI: 10.1152/physiolgenomics.00177.2021
52. Sugg KB, Korn MA, Sarver DC, Markworth JF, Mendias CL. **Inhibition of platelet-derived growth factor signaling prevents muscle fiber growth during skeletal muscle hypertrophy**. *FEBS Lett* (2017.0) **591** 801-809. DOI: 10.1002/1873-3468.12571
53. Sempere LF, Freemantle S, Pitha-Rowe I, Moss E, Dmitrovsky E, Ambros V. **Expression profiling of mammalian microRNAs uncovers a subset of brain-expressed microRNAs with possible roles in murine and human neuronal differentiation**. *Genome Biol* (2004.0) **5** R13. DOI: 10.1186/gb-2004-5-3-r13
54. McCarthy JJ. **The MyomiR network in skeletal muscle plasticity**. *Exerc Sport Sci Rev* (2011.0) **39** 150-154. DOI: 10.1097/JES.0b013e31821c01e1
55. 55.Xu B, Shen X, Yang Z, Zhao T, Liu B, Gao S, et al. Plasma miR-1, but not extracellular vesicle miR-1, functions as a potential biomarker for colorectal cancer diagnosis. Clin Lab. 2021;67(1).
56. Vechetti IJ, Peck BD, Wen Y, Walton RG, Valentino TR, Alimov AP. **Mechanical overload-induced muscle-derived extracellular vesicles promote adipose tissue lipolysis**. *FASEB J* (2021.0) **35** e21644. DOI: 10.1096/fj.202100242R
57. Siracusa J, Koulmann N, Banzet S. **Circulating myomiRs: a new class of biomarkers to monitor skeletal muscle in physiology and medicine**. *J Cachexia Sarcopeni* (2018.0) **9** 20-27. DOI: 10.1002/jcsm.12227
58. Ultimo S, Zauli G, Martelli AM, Vitale M, McCubrey JA, Capitani S. **Influence of physical exercise on microRNAs in skeletal muscle regeneration, aging and diseases**. *Oncotarget.* (2018.0) **9** 17220-17237. DOI: 10.18632/oncotarget.24991
59. Soplinska A, Zareba L, Wicik Z, Eyileten C, Jakubik D, Siller-Matula JM. **MicroRNAs as biomarkers of systemic changes in response to endurance exercise-a comprehensive review**. *Diagnostics.* (2020.0) **10** 813. DOI: 10.3390/diagnostics10100813
60. Ramos AE, Lo C, Estephan LE, Tai YY, Tang Y, Zhao J. **Specific circulating microRNAs display dose-dependent responses to variable intensity and duration of endurance exercise**. *Am J Physiol Heart Circ Physiol* (2018.0) **315** H273-HH83. DOI: 10.1152/ajpheart.00741.2017
61. Mizenko RR, Brostoff T, Rojalin T, Koster HJ, Swindell HS, Leiserowitz GS. **Tetraspanins are unevenly distributed across single extracellular vesicles and bias sensitivity to multiplexed cancer biomarkers**. *J Nanobiotechnology* (2021.0) **19** 250. DOI: 10.1186/s12951-021-00987-1
62. 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 U S A* (2016.0) **113** E968-E977. DOI: 10.1073/pnas.1521230113
63. Barranco I, Padilla L, Parrilla I, Alvarez-Barrientos A, Perez-Patino C, Pena FJ. **Extracellular vesicles isolated from porcine seminal plasma exhibit different tetraspanin expression profiles**. *Sci Rep* (2019.0) **9** 11584. DOI: 10.1038/s41598-019-48095-3
64. Koliha N, Wiencek Y, Heider U, Jungst C, Kladt N, Krauthauser S. **A novel multiplex bead-based platform highlights the diversity of extracellular vesicles**. *J Extracell Vesicles* (2016.0) **5** 29975. DOI: 10.3402/jev.v5.29975
65. Gaignier F, Schenten V, De Carvalho BM, Gauquelin-Koch G, Frippiat JP, Legrand-Frossi C. **Three weeks of murine hindlimb unloading induces shifts from B to T and from Th to Tc splenic lymphocytes in absence of stress and differentially reduces cell-specific mitogenic responses**. *PLoS One* (2014.0) **9** e92664. DOI: 10.1371/journal.pone.0092664
66. Wei LX, Zhou JN, Roberts AI, Shi YF. **Lymphocyte reduction induced by hindlimb unloading: distinct mechanisms in the spleen and thymus**. *Cell Res* (2003.0) **13** 465-471. DOI: 10.1038/sj.cr.7290189
67. Murach KA, Peck BD, Policastro RA, Vechetti IJ, Van Pelt DW, Dungan CM. **Early satellite cell communication creates a permissive environment for long-term muscle growth**. *iScience.* (2021.0) **24** 102372. DOI: 10.1016/j.isci.2021.102372
68. Fry CS, Kirby TJ, Kosmac K, McCarthy JJ, Peterson CA. **Myogenic progenitor cells control extracellular Matrix production by fibroblasts during skeletal muscle hypertrophy**. *Cell Stem Cell* (2017.0) **20** 56-69. DOI: 10.1016/j.stem.2016.09.010
69. Greenleaf JE. **Physiological responses to prolonged bed rest and fluid immersion in humans**. *J Appl Physiol Respir Environ Exerc Physiol* (1984.0) **57** 619-633. PMID: 6386766
70. Warnier G, De Groote E, Britto FA, Delcorte O, Nederveen JP, Nilsson MI. **Effects of an acute exercise bout in hypoxia on extracellular vesicle release in healthy and prediabetic subjects**. *Am J Physiol Regul Integr Comp Physiol* (2022.0) **322** R112-RR22. DOI: 10.1152/ajpregu.00220.2021
71. Toribio V, Yanez-Mo M. **Tetraspanins interweave EV secretion, endosomal network dynamics and cellular metabolism**. *Eur J Cell Biol* (2022.0) **101** 151229. DOI: 10.1016/j.ejcb.2022.151229
72. Cabanas C, Yanez-Mo M, van Spriel AB. **Editorial: functional relevance of tetraspanins in the immune system**. *Front Immunol* (2019.0) **10** 1714. DOI: 10.3389/fimmu.2019.01714
73. Brzoska E, Kowalski K, Markowska-Zagrajek A, Kowalewska M, Archacki R, Plaskota I. **Sdf-1 (CXCL12) induces CD9 expression in stem cells engaged in muscle regeneration**. *Stem Cell Res Ther* (2015.0) **6** 46. DOI: 10.1186/s13287-015-0041-1
74. Przewozniak M, Czaplicka I, Czerwinska AM, Markowska-Zagrajek A, Moraczewski J, Streminska W. **Adhesion proteins--an impact on skeletal myoblast differentiation**. *PLoS One* (2013.0) **8** e61760. DOI: 10.1371/journal.pone.0061760
75. Charrin S, Latil M, Soave S, Polesskaya A, Chretien F, Boucheix C. **Normal muscle regeneration requires tight control of muscle cell fusion by tetraspanins CD9 and CD81**. *Nat Commun* (2013.0) **4** 1674. DOI: 10.1038/ncomms2675
|
---
title: 'Factors associated with severity and anatomical distribution of diabetic foot
ulcer in Uganda: a multicenter cross-sectional study'
authors:
- Bienfait Mumbere Vahwere
- Robinson Ssebuufu
- Alice Namatovu
- Patrick Kyamanywa
- Ibrahim Ntulume
- Isaac Mugwano
- Theophilus Pius
- Franck Katembo Sikakulya
- Okedi Francis Xaviour
- Yusuf Mulumba
- Soria Jorge
- Gidio Agaba
- George William Nasinyama
journal: BMC Public Health
year: 2023
pmcid: PMC9999659
doi: 10.1186/s12889-023-15383-7
license: CC BY 4.0
---
# Factors associated with severity and anatomical distribution of diabetic foot ulcer in Uganda: a multicenter cross-sectional study
## Abstract
### Background
Diabetic foot ulcer (DFU) is a devastating complication of diabetes mellitus (DM) that is associated with increased mortality, morbidity, amputation rate and economic burden. This study aimed at identifying the anatomical distribution and factors associated with severity of DFU in Uganda.
### Methodology
This was a multicenter cross-sectional study conducted in seven selected referral hospitals in Uganda. A total of 117 patients with DFU were enrolled in this study between November 2021 and January 2022. Descriptive analysis and modified Poisson regression analysis were performed at $95\%$ confidence interval; factors with p-value < 0.2 at bivariate analysis were considered for multivariate analysis.
### Results
The right foot was affected in $47.9\%$ ($$n = 56$$) of patients, $44.4\%$ ($$n = 52$$) had the DFU on the plantar region of the foot and $47.9\%$ ($$n = 56$$) had an ulcer of > 5 cm in diameter. The majority ($50.4\%$, $$n = 59$$) of patients had one ulcer. $59.8\%$ ($$n = 69$$) had severe DFU, $61.5\%$ ($$n = 72$$) were female and $76.9\%$ had uncontrolled blood sugar. The mean age in years was 57.5 (standard deviation 15.2 years). Primary ($$p \leq 0.011$$) and secondary ($p \leq 0.001$) school educational levels, moderate ($$p \leq 0.003$$) and severe visual loss ($$p \leq 0.011$$), 2 ulcers on one foot ($$p \leq 0.011$$), and eating vegetables regularly were protective against developing severe DFU ($$p \leq 0.03$$). Severity of DFU was 3.4 and 2.7 times more prevalent in patients with mild and moderate neuropathies ($p \leq 0.01$), respectively. Also, severity was 1.5 and 2.5 higher in patients with DFU of 5–10 cm ($$p \leq 0.047$$) and in those with > 10 cm diameter ($$p \leq 0.002$$), respectively.
### Conclusion
Most DFU were located on the right foot and on the plantar region of the foot. The anatomical location was not associated with DFU severity. Neuropathies and ulcers of > 5 cm diameter were associated with severe DFU but primary and secondary school education level and eating vegetables were protective. Early management of the precipitating factors is important to reduce the burden of DFU.
## Background
Diabetic Foot Ulcer (DFU) is an advanced consequence of diabetes mellitus (DM). There is a $15\%$ lifetime chance of developing the ulcer among diabetic patients and when it occurs, it is associated with high mortality [1]. The 5-year survival rate among patients with DFU varies between $25\%$ and $45\%$ worldwide [2–4]. Globally, DFU has become one of the leading causes of lower limb amputation with over 1 million patients amputated annually, an average of a limb amputation every 20 s [5]. Following amputation for DFU, $85\%$ of patients will still develop chronic infection and other forms of gangrene which lead to poor quality of life and financial stress [6, 7]. One third of the management costs of diabetes is estimated to be linked to foot ulcers as compared to patients without DFU and the cost of care is estimated to be 5.4 and 2.6 times higher in the year of first episode and second episode, respectively [8]. The cost of treatment also increases with severity of the DFU with the highest grade DFU ulcers costing eight times higher than the lowest grade [7]. In addition to high management costs, the severity of DFU ultimately leads to high mortality [9]. The prevalence of diabetic foot has been reported to be higher among people with type 2 DM compared to those with type 1 diabetes worldwide [10]. Globally, the prevalence of DFU averages at $6.4\%$ with a higher predilection in men compared to women [11]. This prevalence varies between $3\%$ in Oceania to $13\%$ in North America, with a prevalence of $7.2\%$ across Africa; in Uganda it varies from 1 to $4\%$ [10]. In Ethiopia, a prevalence of $13.6\%$ of DFU was reported among type 2 diabetes mellitus patients and it was associated with rural residence, poor foot self-care practice, obesity, and neuropathy [12].
Studies have shown that factors usually associated with occurrence of DFU include older age, longer duration of DM, hypertension, diabetic retinopathy and smoking history [13]. Peripheral neuropathy, peripheral vascular disease and foot trauma were also reported risk factors in the pathophysiology of foot ulcer [14]. Other factors include low educational status, high body mass index (BMI)and inadequate foot self-care practice [15–17], however these factors may differ based on the patient’s socio-economic status, demographic characteristics, and the evolution of DFU within the facility [12]. The factors precipitating progression to severe DFU in Ugandan patients have not been well studied.
The management of DFU requires the participation of the patient [18] and the outcome is related to patient awareness and self-foot care behavior [19–21]. Self-management of DM has a significant impact on the outcome of blood sugar level control and complications such as DFU [18].
Although Wagner grade 3 and 4, metatarsal is the most common location of DFU and associated factors have been largely studied in developed countries [19–23], there is still a paucity of data regarding the anatomical distribution and factors associated with severe DFU among patients in LMICs such as Uganda. An evidence-based understanding of the factors associated with increasing severity of DFU is necessary to establish effective control measures to reduce its burden. The purpose of this study, therefore, was to determine the anatomical distribution and factors associated with increasing severity of DFU in patients with DM in Uganda.
## Study design and setting
This multicenter cross-sectional study was conducted in 7 hospitals in Uganda (Kampala International Teaching Hospital, Kitagata General Hospital, Mbarara Regional Referral Hospital, Fort Portal Regional Referral Hospital, Hoima Regional Referral Hospital, Jinja Regional Referral Hospital and Kiruddu Specialized Hospital) (Fig. 1). The hospitals were deliberately chosen based on the high prevalence of DM in Central, Western and Eastern Uganda where the hospitals are located [24].
Fig. 1Distribution of study sites according to district. RRH: Regional referral hospital; NRH: National referral hospital
## Patients and recruitment
All patients aged 18 years and above with DM type 1 and type 2, having a wound located below the ankle and attending the surgical department and/or DM clinics of selected hospitals in Uganda between November 1, 2021 and January 31, 2022 were recruited. Purposive.
consecutive sampling method was used until the desired sample size was reached. Patients provided written informed consent to participate in the study. Patients without mental capacity and those without an adult to consent for them were excluded from the study.
DFU patients with communication difficulty, such as those with severe cognitive impairment or those who could not consent, were excluded from the study.
The required sample size for the study patients with DFU was calculated using the Kish *Leslie formula* as cited by Singh [25]. Data about the prevalence of DFU in *Uganda is* still scanty, therefore we used the prevalence estimate of DFU in a cross-sectional study done in Egypt ($8.7\%$ among adult patients aged 18 years and above attending Alexandria University Teaching Hospital Diabetic clinic) to determine the sample size [26]. Using the prevalence estimate from Egypt, which is similar to the one in Kenya [27], resulted in a calculated sample size of 122 patients.
## Study procedure
We developed and piloted a data collection tool informed by relevant literature. The tool was translated by a language expert into the local languages spoken in each of the study areas, namely Luganda/Lusoga, Runyakore/Rukiga and Runyoro/Rutoro. A medical doctor, an ophthalmologist/ophthalmology clinical officer, a nurse and a surgical resident were recruited and trained as research assistants in each of the selected hospitals. Data collection was supervised by the principal investigator (BMV) and other authors (RS, GA, IM, and FKS).
Physical examination to assess the Wagner classification, anatomical distribution of DFU, neuropathy, and blood pressure was performed by a doctor. The ankle-brachial index (ABI) was calculated using the ratio of the blood pressure taken at the ankle of the affected foot and the blood pressure taken at the one third distal of the arm. The ABI was not measured for patients with DFU affecting the ankle and the variable was considered as “not applicable”. The key variables included socio-demographic variables such as age, sex, occupation, monthly income, tribe and residency, general characteristics of the patients such as type of DM, duration of DM, history of trauma, duration of the diabetic foot ulcer, type of therapy and behavior-related factors such as history of smoking, alcohol intake, type of diet, frequency of consulting the diabetic clinic, foot care and lifestyle. Patients were asked about their regular weekly diet and the food was categorized depending on the main constituent: carbohydrates (matoke/plantains, cassava, posho, rice, sweet potatoes, Irish potatoes, pumpkins, yams), proteins (meat, eggs, fish, beans, milk), lipids (ground nuts, sim sim paste, ghee, cooking oil) and fruits and vegetables. Frequency of consumption was categorized as: daily (if they ate the mentioned food in a category every day), usually (if they ate any of the foods in the category at least 4 to 6 times in a week), occasionally (if at least 2 to 3 times in a week) and rarely (if once a week or less frequent). Regarding alcohol intake, the patient was regarded as an alcoholic if they drink alcohol regularly irrespective of the type of alcohol and non-alcoholic if the patient does not take alcohol at all, if the patient drinks alcohol once in a while or if a patient had completely stopped drinking alcohol for two years.
For the smoking variable, a smoker was a patient who chewed or smoked tobacco irrespective of the type and number of times per day whilst a non-smoker status referred to a patient who has never smoked or one that has ceased smoking for at least two years. Diabetic clinic consultation was assessed as the frequency of consultation and as categorized as monthly, every 6 months, once in a year or not at all. Foot care of the patients specifically considered what the patient uses to cut the nails (nail cutter, knife, or razor blade),who cuts the nails for the patient (self or assisted by another person) and the type of shoes regularly worn by the patient (closed, open or fitted) while the foot score consisted of: washing feet - score 1; drying in between the toes - score 2; and using moisturizing products - score 3. Based on this scoring, a good score meant a patient practices all three, a medium score if patient practices any two and a poor score if the patient practices only one of the three.
Blood samples for HbA1c and fasting blood sugar (FBS) testing were collected from one of the main superficial veins of the cubital fossa (cephalic, basilic, median cubital, and median ante brachial).The blood sample was collected in an ethylenediaminetetraacetic acid (EDTA) grey top vacutainer for blood sugar tests and glycosylated hemoglobin (HbA1c) [22, 28] and transported in a cooler box (AL medical cool box, England UK) at 2 to 6 Celsius degree to the laboratory of Kampala International University-Teaching Hospital (KIU-TH) [29]. Four milliliters (ml) of blood were withdrawn from the anterior cubital fossa of each subject using a sterile disposable syringe and needle after cleaning the site with a swab soaked in $70\%$ alcohol.
HbA1c was analyzed using aIchroma II Machine [2017] and results were interpreted as follows: <$6.5\%$ HbA1c was considered as controlled DM and an HbA1c of $6.5\%$ and above were uncontrolled DM [30]. Each study patient received a printed copy of their results.
Patients’ weight in kilograms and height in meters were determined using a calibrated analogue weighing scale and wall mounted station meter manufactured by Southern Early Child Association (SECA). The body mass index was calculated using the formula: BMI = kg/m2 where kg was a patient’s weight in kilograms and m2 was their height in meters squared. Patients were categorized as normal (BMI of 18.5 to 24.9), overweight (BMI of 25.0 to 29.9) or obese (BMI of 30 and above) [31]. Blood pressure was taken using a digital automated sphygmomanometer (bosomedicus vital D-72,417 JUNGINGEN/ German) with an appropriate cuff size for the arm. High blood pressure was defined as systolic blood pressure ≥ l40 mmHg or diastolic pressure ≥ 90 mmHg (European Society of Cardiology/European Society of Hypertension, 2018) [32].
Pressure sensation was assessed using $\frac{5.07}{10}$ g monofilament (Semmes Weinstein monofilament test, made in China) at 4 of the 10 standard sites of the sole of the feet (plantar base of the big toe, 2nd and 5th toes and at the heel), avoiding areas with callosity [33]. Vibration sense was elicited using a 128 Hz tuning fork at the hallux [33, 34]. Diabetic neuropathy was classified as absent, minor, moderate or severe using the new classification by Picon and collaborators [35]. The Neuropathy Disability Score (NDS) was used to assess the grade of diabetic peripheral neuropathy (DPN) for each patient. The NDS system is a neuropathy scoring tool ranging from 0 to 10 which can also be used for assessment of severity of peripheral neuropathy by considering four parameters: vibration sense by using a 128 Hz tuning fork (0 = present,1 = reduced/absent for each foot), temperature sensation by using a cold tuning fork (0 = present,1 = reduced/absent for each foot), pin-prick sensation by a monofilament test (0 = present,1 = reduced/absent for each foot and ankle reflex), Achilles tendon reflex by using a patellar hammer (0 = normal, 1 = present with reinforcement, 2 = absent per side) [36]. Absence of neuropathy (normal) was considered when the score was 0 to 2. The classification of DPN disability was graded as follows: mild (score 3–5), moderate (score 6–8), and severe (score 9–10) as described by Dyck [37].
Patients’ feet were examined to determine the characteristics of the foot ulcer, number of ulcers, size of ulcers and location. The ulcer was classified using the Wagner classification [23].
The patients were classified into two [2] categories based on severity of DFU, as early DFU (less severe) (grade 1 and grade 2 foot ulcer) and severe or late DFU (grade 3 and above) using the Wagner classification [9].
Visual loss was assessed by either a trained ophthalmology clinical officer or ophthalmologist using a Snellen chart to assess the smallest letter a patient can read or orientation of letters at 20 feet or 6 m with one eye closed. Visual impairment was categorized according to the WHO classification where low vision is classified into three categories: mild, moderate, and severe. Mild visual impairment is visual acuity of ≥ $\frac{6}{18}$, moderate visual impairment is less than $\frac{6}{18}$ but equal or better than $\frac{6}{60}$, whilst severe visual impairment is < $\frac{6}{60}$ [38].
## Data processing and analysis
The raw data was entered into MS Excel spreadsheet software, cleaned and later exported to Stata version 15 (Stata Corp®) for analysis. Categorical variables were analyzed using the proportions whilst the difference in proportions was assessed using the Chi-square test and data was summarized and presented in form of frequencies and percentages. Continuous variables were presented as means and standard errors.
Bivariate and multivariable analyses were performed using modified Poisson regression to assess the association between severity of the DFU and the factors studied, with significance determined at $p \leq 0.05.$ Variables with a p-value < 0.2 in bivariate analysis were considered for multivariable analysis.
Poisson regression was chosen to obtain prevalence risk ratios (PRR) over logistic regression to avoid overestimation of the prevalence ratio and to allow for appropriate control of confounding variables, because the latter poorly estimates the standard errors of the estimated risk ratios especially when dealing with severe DFU which was a common outcome of interest among patients with DFU [39].
## Ethical considerations
Ethical approval was granted by the Research Ethics Committee of Kampala International University (KIU), reference KIU-REC-2021-57 and permission to access the patients was obtained from management of the selected hospitals before data collection.
## Results
Among 122 targeted study patients, 117 ($96\%$) patients with DFU had complete data and this was subsequently analyzed. Five ($4\%$) patients that had incomplete data were excluded from further analysis.
Among the 117 patients with DFU who formed the definitive sample, the majority ($$n = 70$$; $59.8\%$) had severe DFU and 47 ($40.2\%$) less severe DFU. The majority of the patients ($61.5\%$, $$n = 72$$) were female. The mean age of study patients was 57.1 years (49.5 for patients with less severe DFU and 67.5 for patients with severe DFU). The majority of the patients ($76.9\%$; $$n = 90$$) had poor blood sugar control. Most patients ($44\%$; $$n = 52$$) had a grade 3 diabetic foot ulcer followed by grade 2 ($28\%$, $$n = 33$$), grade 1 ($12\%$, $$n = 14$$), grade 4 ($8\%$, $$n = 9$$) and grade 5($8\%$, $$n = 9$$) according to the Wagner classification.
## Severity of the diabetic foot ulcer
As shown in Figs. 2 and 70 ($59.8\%$) of the study patients had severe DFU ($95\%$CI 50.4–68.8) and 47 ($40.2\%$) had less severe DFU ($95\%$CI 31.22–49.6). The proportions were significantly different ($p \leq 0.05$).
Fig. 2Proportion of patients with severe and less severe diabetic foot ulcers in selected referral hospitals in Uganda, November 2021 to January 2022 ($$n = 117$$)
## Socio demographic characteristics of DFU patients
Most ($33.3\%$) of the patients were in the age-group between 50 and 59 years; and the majority ($74.4\%$) were from rural areas (Table 1). The age group of 70–95 years was associated with severe DFU ($$p \leq 0.02$$).
Table 1Sociodemographic characteristics of patients with DFU in selected referral hospitals in Uganda, November 2021 to January 2022 ($$n = 117$$)VariablesLess severe DFU (G1&G2): n(%)Severe DFU (G3, G4, &G5): n(%)Totaln (%)P value Age group in years 0.02618–3910 (21.3)4 (5.7)14 (12.0)40–497 (14.9)8 (11.4)15 (12.8)50–5913 (27.7)26 (37.1)39 (33.3)60–6912 (25.5)13 (18.6)25 (21.4)70–955 (10.6)19 (27.1)24 (20.5) Sex 0.720Female28 (38.9)44 (61.1)72 (61.5)Male19 (40.4)26 (37.1)45 (38.5) Residence 0.682Urban13 (27.7)17 (24.3)30 (25.6)Rural34 (72.3)53 (75.7)87 (74.4) Region 0.555Western33 (70.2)51 (72.9)84 (71.8)Central6 (12.8)12 (17.1)18 (15.4)Eastern7 (14.9)5 (7.1)12 (10.3)Non-Ugandan1 (2.1)2 (2.9)3 (2.6) Occupation 0.502Peasant farmer28 (59.6)42 (60.0)70 (59.8)Business/self employed11 (23.4)13 (18.6)24 (20.5)Non-employed7 (14.9)9 (12.9)16 (13.7)Formerly employed1 (2.1)6 (8.6)7 (6.0) Average monthly income in UgShs 0.144< 10,00015 (31.9)21 (30.0)36 (30.8)10,000-100,00016 (34.0)14 (20.0)30 (25.6)100,001-500,00014 (29.8)22 (31.4)36 (30.8)500,001-1million1 (2.1)9 (12.9)10 (8.5)Above 1million1 (2.1)4 (5.7)5 (4.3)Ug. Shs: Ugandan Shillings; 1 United States (US) dollar = 3800Ug. Shs (exchange rate as of 1st September 2022)
## Medical characteristics of DFU patients
Table 2 shows that the majority of patients had DM type 2 ($94.9\%$, $$n = 111$$) and 65 ($55.6\%$) had suffered with DM for a duration of 9 years and above. In $55\%$ the DFU had lasted for a duration of 1–6 months, $25.6\%$ had a prior history of DFU and $18.8\%$ had a history of amputation. The majority of patients ($65\%$) were on insulin.
Table 2Medical characteristics of patients in selected referral hospitals in Uganda, November 2021 to January 2022 ($$n = 117$$)VariablesLess severe (G1&G2): n(%)Severe (G3, G4, &G5): n(%)Totaln (%)p-value Type of DM 0.174Type 14 (8.5)2 (2.9)6 (5.1)Type 243 (91.5)68 (97.1)111 (94.9) Duration of DM in years, mean (SD) 12.2 (11.0)10.0 (8.8)10.9 (9.7)0.236 Duration of DM in years 0.7440–312 (25.5)19 (27.1)31 (26.5)4–810 (21.3)11 (15.7)21 (17.9)9 and above25 (53.2)40 (57.1)65 (55.6)Method of treatment Insulin 0.834No17 (36.2)24 (34.3)41 (35.0)Yes30 (63.8)46 (65.7)76 (65.0) Herbal medicine 0.330No26 (55.3)45 (64.3)71 (60.7)Yes21 (44.7)25 (35.7)46 (39.3) Oral hypoglycemic agents (OHA) 0.783No20 (42.6)28 (40.0)48 (41.0)Yes27 (57.4)42 (60.0)69 (59.0) On treatment 0.683No45 (95.7)68 (97.1)113 (96.6)Yes2 (4.3)2 (2.9)4 (3.4) Number of types of medication received 0.17302 (4.3)2 (2.9)4 (3.4)120 (42.6)39 (55.7)59 (50.4)217 (36.2)13 (18.6)30 (25.6)38 (17.0)16 (22.9)24 (20.5) Duration of DFU in months, mean (SD) 4.9 (10.1)4.2 (7.3)4.5 (8.5)0.670 Duration of DFU in months 0.527Less than 117 (38.6)19 (28.4)36 (32.4)1–622 (50.0)39 (58.2)61 (55.0)Over 65 (11.4)9 (13.4)14 (12.6) History of trauma to an affected foot 0.604No40 (85.1)57 (81.4)97 (82.9)Yes7 (14.9)13 (18.6)20 (17.1) History of previous DFU 0.535No36 (76.6)50 (71.4)86 (73.5)Yes11 (23.4)20 (28.6)31 (26.5) History of amputation due to DFU 0.375No40 (85.1)55 (78.6)95 (81.2)Yes7 (14.9)15 (21.4)22 (18.8)SE: standard error
## Behavioral characteristics of patients with DFU
As shown in Table 3, only 20 ($17.1\%$, $$n = 117$$) of patients attended a diabetes clinic twice a month and only 22 ($18.8\%$) patients reported to be doing regular physical exercise. The prevalence of other factors associated with severity on DFU such as foot care, history of smoking, not doing physical exercise and eating a mixture of foods are also shown.
Table 3Behavioral characteristics of DFU patients in selected referral hospitals in Uganda, November 2021 to January 2022 ($$n = 117$$)VariablesEarly (G1&G2):n(%)Late (G3, G4, &G5): n (%)Totaln (%)p-value Taught About complications of DM 0.274No20 (42.6)37 (52.9)57 (48.7)Yes27 (57.4)33 (47.1)60 (51.3) Counselled about risk of DFU when you have DM 0.411No14 (29.8)26 (37.1)40 (34.2)Yes33 (70.2)44 (62.9)77 (65.8) Counselled by a friend 0.707No31 (66.0)41 (58.6)72 (61.5)Yes16 (4.3)29 (51.4)45 (38.5) Counselled by a healthcare worker 0.480No30 (63.8)37 (52.9)67 (57.3)Yes17(6.4)33 (47.1)50 (42.7) Counselled by social media 0.707No31 (66.0)41 (58.6)72 (61.5)Yes16 (34.1)29 (41.4)45 (39.0) Frequency of DM clinic attendance 0.577Never13 (27.7)25 (35.7)38 (32.5)*Once a* month22 (46.8)28 (40.0)50 (42.7)Every 2 months7 (14.9)13 (18.6)20 (17.1)Once yearly5 (10.6)4 (5.7)9 (7.7) *Smoking status* 0.040Non smoker44 (93.6)56 (80.0)100 (85.5)Ever smoked3 (6.4)14 (20.0)17 (14.5) Alcohol consumer 0.512No38 (80.9)53 (75.7)91 (77.8)Yes9 (19.1)17 (24.3)26 (22.2) Foot care score 0.464112 (25.5)25 (35.7)37 (31.6)25 (10.6)5 (7.1)10 (8.5)330 (63.8)40 (57.1)70 (59.8) Type of shoes worn 0.908Open shoes30 (63.8)47 (67.1)77 (65.8)Any type of shoes12 (25.5)17 (24.3)29 (24.8)Don’t wear shoes5 (10.6)6 (8.6)11 (9.4) How often nails are cut since DM diagnosed 0.791Always14 (29.8)22 (31.4)36 (30.8)Occasionally32 (68.1)45 (64.3)77 (65.8)Don’t1 (2.1)3 (4.3)4 (3.4) Individual who cuts the toenails 0.762Self30 (63.8)46 (65.7)76 (65.0)Other people17 (36.1)24 (34.3)41 (35.0)What do you use to cut your nails (multiple answers) Nail cutter 0.329No33 (70.2)43 (61.4)76 (65.0)Yes14 (29.8)27 (38.6)41 (35.0) Razor blade 0.152No14 (29.8)30 (42.9)44 (37.6)Yes33 (70.2)40 (57.1)73 (62.4) Knife 0.807No46 (97.9)68 (97.1)114 (97.4)Yes1 (2.1)2 (2.9)3 (2.6) Frequency of exercise 0.529Always8 (17.0)14 (20.0)22 (18.8)Occasionally27 (57.4)39 (55.7)66 (56.4)Never12 (25.5)17 (24.3)29 (24.8) Work in the garden 0.921No27 (57.4)37 (52.9)64 (54.7)Yes20 (42.6)33 (47.1)53 (45.3) Walk at least 30 min a day 0.625No25 (53.2)47 (67.1)72 (61.5)Yes22 (46.8)23 (32.9)45 (38.5) Housework 0.128No38 (80.9)56 (80.0)94 (80.3)Yes9 (19.1)14 (20.0)23 (19.7) Jogging 0.910No40 (85.1)67 (95.7)107 (91.5)Yes7 (14.9)3 (4.3)10 (8.5) No exercise 0.044No36 (76.6)57 (81.4)93 (79.5)Yes11 (23.4)13 (18.6)24 (20.5) Diet (how often food eaten:) Vegetables 0.103Daily26(0.2)32(0.3)58(0.5)Usually4(0.03)1(0.01)5(0.04)Occasionally2(0.02)8(0.1)10(0.1)Rarely15(0.1)29(0.2)44(0.4) Fruits 0.924Daily21(0.2)30(0.3)51(0.4)Usually1(0.01)1(0.01)2(0.02)Occasionally6(0.1)7(0.1)13(0.1)Rarely19(0.2)32(0.3)51(0.4) Fatty food 0.492Daily26(0.2)39(0.3)65(0.6)Usually1(0.01)6(0.1)7(0.1)Occasionally9(0.1)10(0.1)19(0.2)Rarely11(0.1)15(0.1)26(0.2) Proteins 0.215Daily30(0.3)47(0.4)77(0.7)Usually1(0.01)6(0.1)7(0.1)Occasionally8(0.1)5(0.04)13(0.1)Rarely8(0.1)12(0.1)20(0.2) Carbohydrates 0.209Daily33(0.3)56(0.5)89(0.8)Usually4(0.03)3(0.03)7(0.1)Occasionally4(0.03)1(0.01)5(0.04)Rarely6(0.1)10(0.1)16(0.1)
## Comorbidities of patients with DFU
The study showed that most patients ($47.9\%$) had a normal BMI whilst 38 ($32.5\%$) had pre-obesity. Hypertension was found in 69 ($59\%$) patients, moderate kidney failure in 23 ($19.7\%$), moderate neuropathy in 44 ($37.6\%$) and severe neuropathy in 38 ($32.5\%$) of patients with severe DFU (p-value 0.01). Hba1C levels above $6.5\%$ were found in 90 ($76.9\%$) of patients, compensated heart disease in 22 ($18.8\%$) and decompensated heart disease in 5 ($4.3\%$). Furthermore, claudication was the most common peripheral vascular disease, reported by 29 ($24.8\%$) patients, followed by gangrene ($6.0\%$) and deep venous thrombosis (DVT) ($4.3\%$) (Table 4).
Table 4Co-morbidities of patients with DFU in selected referral hospitals in Uganda, November 2021 to January 2022 ($$n = 117$$)VariablesEarly (G1&G2) n(%)Late (G3, G4, &G5)n(%)Totaln (%)p-value Neuropathy 0.013Absent16 (34.0)8 (11.4)24 (20.5)Mild3 (6.4)8 (11.4)11 (9.4)Moderate18 (38.3)26 (37.1)44 (37.6)Severe10 (21.3)28 (40.0)38 (32.5) Peripheral Vascular Disease 0.103No disease35 (74.5)41 (58.6)76 (65.0)Claudication10 (21.3)24 (27.1)34 (29.1)Gangrene0 (0.0)7 (10.0)7 (6.0)DVT2 (4.3)3 (4.3)5 (4.3) Hba1C, mean (SD) 8.3 (2.6)8.8 (2.6)8.6 (2.6)0.158 Hba1C levels Less than $6.5\%$14 (29.8)13 (18.6)27 (23.1)$6.5\%$ and above33 (70.2)57 (81.4)90 (76.9)0.158 BP class Normal19 (40.4)28 (40.0)47 (40.2)Hypertension28 (59.6)41 (58.6)69 (59.0)Hypotension0 (0.0)1 (1.4)1 (0.9) BMI, mean (SD) 25.9 (6.7)25.0 (5.2)25.3 (5.9)0.393 BMI group 0.614Underweight < 18.53 (6.4)5 (7.1)8 (6.8)Normal 18.5–24.921 (44.7)35 (50.0)56 (47.9)Pre-obesity 25.0–29.916 (34.0)22 (31.4)38 (32.5)Obesity Class I 30.0–34.93 (6.4)6 (8.6)9 (7.7)Obesity Class II 35.0–39.92 (4.3)0 (0.0)2 (1.7)Obesity Class III 40 and above2 (4.3)2 (2.9)4 (3.4) ABI group 0.192Normal (1.0 to 1.4)9(0.1)19(0.2)28(0.2)Acceptable (0.9 to less 1.0)4(0.03)3(0.03)7(0.1)Some Arterial Disease (0.8 to less 0.9)1(0.01)0(0.0)1(0.01)Moderate (0.5 to 0.7)0(0.0)2(0.02)2(0.02)Severe (Less than 0.5)0(0.0)2(0.02)2(0.02)Not assessed31(0.3)46(0.4)77(0.7)ABI: ankle brachial index; BMI: body mass index, Hba1C: glycosylated hemoglobin, DVT: deep venous thrombosis, BP: blood pressure, SD = standard deviation
## Anatomical distribution of DFU
Most patients ($47.9\%$, $$n = 56$$) had DFU of the right foot, with the majority ($54.7\%$) found on the dorsum of the foot. Location of the ulcer on the plantar region ($$p \leq 0.01$$), having more than 4 ulcers ($$p \leq 0.01$$) and a size of ulcer of > 10 cm of diameter ($$p \leq 0.00$$) were associated with higher severity of the DFU (Table 5).
Table 5Anatomical distribution of DFU among patients in selected referral hospitals in Uganda, November 2021 to January 2022 ($$n = 117$$)VariablesLess severe (G1&G2):n (%)Severe (G3, G4, &G5): n (%)Totaln(%)p-value Foot affected by DFU 0.633Right23 (48.9)33 (47.1)56 (47.9)Left20 (42.6)27 (38.6)47 (40.2)Both4 (8.5)10 (14.3)14 (12.0) Location of DFU Heel 0.409No38 (80.9)52 (74.3)90 (76.9)Yes9 (19.1)18 (25.7)27 (23.1) Dorsum 0.074No26 (55.3)27 (38.6)53 (45.3)Yes21 (44.7)43 (61.4)64 (54.7) Plantar 0.009No33 (70.2)32 (45.7)65 (55.6)Yes14 (29.8)38 (54.3)52 (44.4) Toes 0.788No22 (46.8)31 (44.3)53 (45.3)Yes25 (53.2)39 (55.7)64 (54.7) Number of ulcers 0.010129 (61.7)30 (42.9)59 (50.4)214 (29.8)15 (21.4)29 (24.8)34 (8.5)22 (31.4)26 (22.2)40 (0.0)3 (4.3)3 (2.6) Size of DFU in cm < 0.0011 to 534 (72.3)22 (31.4)56 (47.9)5 to 1013 (27.7)29 (41.4)42 (35.9)Over 100 (0.0)19 (27.1)19 (16.2)cm: centimeter
## Factors influencing severity of DFU
The patients with mild neuropathy (aPRR = 3.4; $95\%$ CI = 1.51–7.63; p-value = 0.003) and moderate neuropathy (aPRR = 2.65; $95\%$ CI = 1.34–5.23; p-value = 0.005) were more at risk of developing severe DFU as compared to those with normal foot sensation, when other factors were held constant. Although at bivariate analysis patients with severe neuropathy were twice more likely ($95\%$ CI = 1.21–4.03, $$p \leq 0.009$$) to develop severe DFU compared to those with normal foot sensations, this turned out not to be a significant risk factor at multivariable analysis. The age group of 70–95 years was significantly associated with increased risk of severe DFU (APR = 3.02; $95\%$ CI = 1.28–7.17; $$p \leq 0.02$$); the size of the ulcer of more than 5 cm (APR = 1.51; $95\%$CI= (1.01–2.27); $p \leq 0.05$) and size > 10 cm (APR = 2.46;$95\%$=1.38–4.39; $$p \leq 0.002$$). Regular eating of vegetables (APRR = 0.062; $95\%$ CI = 0.004–0.863; p-value = 0.038), education level: primary (APRR = 0.58; $95\%$ CI = 0.38–0.89; $$p \leq 0.011$$), secondary (APR = 0.53; $95\%$ CI = 0.29–0.97; $$p \leq 0.039$$); moderate (APRR = 0.48; $95\%$ CI =; $95\%$ CI = 0.29–0.78; $$p \leq 0.003$$) and severe visual loss (APR = 0.31; $95\%$=0.17–0.54; $$p \leq 0.011$$), for 2 ulcers on one foot (APR = 0.52; $95\%$CI = 0.32–0.86; $$p \leq 0.011$$) were significantly less likely associated with severe DFU (Table 6).
Table 6Bivariate and multivariable analysis for factors associated with severity of DFU in selected referral hospitals in Uganda, November 2021 to January 2022 ($$n = 117$$)VariableDFUBivariate analysisMultivariable analysisLess severen (%)Severen (%)c.PRR ($95\%$CI)p-valuea. PRR ($95\%$CI)p-value Age 47 (40.2)70 (59.8)1.01(1.01–1.02)0.0031.02(1.01–1.03)0.001 Age group in years 18–3910 (21.3)4 (5.7)1140–497 (14.9)8 (11.4)1.87(0.72–4.87)0.2022.5(0.91–6.91)0.07650–5913 (27.7)26 (37.1)2.33 (0.99–5.52)0.0541.43(0.63–3.26)0.39760–6912 (25.5)13 (18.6)1.82 (0.73–4.54)0.1991.67 (0.72–3.91)0.23470–955 (10.6)19 (27.1)2.77 (1.18–6.53)0.0203.023 (1.28–7.17)0.012 Sex Female28 (59.6)44 (62.9)11Male19 (40.4)26 (37.1)0.95 (0.69–1.29)0.724-- Residence Rural34 (72.3)53 (75.7)11Urban13 (27.7)17 (24.3)0.95 (0.69–1.29)0.724-- Profession Peasant farmer28 (59.6)42 (60.0)11Business/self employed11 (23.4)13 (18.6)0.9 (0.6–1.37)0.6300.84 (0.43–1.66)0.621Non employed7 (14.9)9 (12.9)0.94 (0.58–1.51)0.7901.02 (0.61–1.71)0.939Formerly employed1 (2.1)6 (8.6)1.43 (1–2.05)0.0521.13 (0.5–2.53)0.773 Level of education No formal3 (6.4)12 (17.1)11Primary29 (61.7)40 (57.1)0.73 (0.52–1)0.0520.58 (0.38–0.89)0.011Secondary14 (29.8)14 (20.0)0.63 (0.4–0.98)0.0410.53 (0.29–0.97)0.039Tertiary1 (2.1)4 (5.7)1 (0.6–1.66)1.0000.45 (0.18–1.17)0.102 Religion Muslim4 (8.5)9 (12.9)11Christian43 (91.5)61 (87.1)0.8 (0.52–1.38)0.5060.9 (0.54–1.45)0.629 Average monthly income in UgShs < 10,00015 (31.9)21 (30.0)0.73 (0.43–1.23)0.2340.42 (0.13–1.33)0.14210,000-100,00016 (34.0)14 (20.0)0.58 (0.33–1.05)0.0710.39 (0.12–1.31)0.128100,001-500,00014 (29.8)22 (31.4)0.76 (0.46–1.28)0.3030.47 (0.16–1.42)0.181500,001–1,000,0001 (2.1)9 (12.9)1.13 (0.69–1.83)0.6350.8 (0.3–2.14)0.650Above 1 M1 (2.1)4 (5.7)1.1. Smoking status Non smoker44 (93.6)56 (80.0)1.1.Ever smoked3 (6.4)14 (20.0)1.47 (1.11–1.95)0.0071.08 (0.68–1.71)0.751 Vegetables Rarely15(0.1)29(0.2)1.1.Daily26(0.2)32(0.3)0.69 (0.20–2.37)0.5584.50 (0.68–1.68)0.118Usually4(0.03)1(0.01)0.13 (0.01–1.26)0.0780.06 (0.00-0.86)0.038Occasionally2(0.02)8(0.1)0.85 (0.38–1.88)0.6850.69 (0.28–3.06)0.430 Carbohydrates Rarely6(0.1)10(0.1)1.1.Daily33(0.3)56(0.5)0.15 (0.01–1.68)0.1230.18 (0.01–2.40)0.193Usually4(0.03)3(0.03)0.45 (0.07–2.74)0.3861.08 (0.11–10.30)0.946Occasionally4(0.03)1(0.01)1.02 (0.34–3.06)0.9741.43 (0.40–5.09)0.581 Dorsum No26 (55.3)27 (38.6)1.1.Yes21 (44.7)43 (61.4)1.32 (0.96–1.81)0.0861.32 (0.88–1.98)0.176 Plantar No33 (70.2)32 (45.7)1.1.Yes14 (29.8)38 (54.3)1.48 (1.1–2)0.0091.46 (0.89–2.38)0.135 Number of locations for one Ulcer 129 (61.7)30 (42.9)1.1.214 (29.8)15 (21.4)1.02 (0.66–1.57)0.9380.52 (0.32–0.86)0.01134 (8.5)22 (31.4)1.66 (1.23–2.25)0.0010.57 (0.27–1.21)0.14540 (0.0)3 (4.3)1.97 (1.53–2.53)< 0.0010.72 (0.33–1.6)0.421 Visual loss Normal8 (17.0)18 (25.7)1.1.Mild7 (14.9)14 (20.0)0.96 (0.65–1.43)0.8530.55 (0.28–1.08)0.084Moderate15 (31.9)23 (32.9)0.87 (0.61–1.26)0.4700.48 (0.29–0.78)0.003Severe17 (36.2)15 (21.4)0.68 (0.43–1.06)0.0900.31 (0.17–0.54)< 0.001 Neuropathy Normal16 (34.0)8 (11.4)1.1.Mild3 (6.4)8 (11.4)2.18 (1.11–4.28)0.0233.4 (1.51–7.63)0.003Moderate18 (38.3)26 (37.1)1.77 (0.95–3.29)0.0702.65 (1.34–5.23)0.005Severe10 (21.3)28 (40.0)2.21 (1.21–4.03)0.0092.06 (0.96–4.43)0.063 Peripheral Vascular Disease No disease35 (74.5)41 (58.6)1.1.Claudication10 (21.3)19 (27.1)1.21 (0.87–1.7)0.2591.39 (0.88–2.2)0.162Gangrene0 (0.0)7 (10.0)1.85 (1.51–2.28)< 0.0011.26 (0.71–2.24)0.424DVT2 (4.3)3 (4.3)1.11 (0.53–2.35)0.7811.74 (0.44–6.89)0.433Hba1C $6.5\%$, n (%)Less than $6.5\%$14 (29.8)13 (18.6)1.6.5 and above33 (70.2)57 (81.4)1.32 (0.86–2.01)0.205-- Blood Pressure class Normal19 (40.4)28 (40.0)1.1.Hypertension28 (59.6)41 (58.6)1 (0.73–1.36)0.9871.05 (0.64–1.73)0.851Hypotension0 (0.0)1 (1.4)1.68 (1.33–2.13)< 0.0011.93 (0.56–6.64)0.296 Size of DFU in cm 1 to 534 (72.3)22 (31.4)1.1.5 to 1013 (27.7)29 (41.4)1.76 (1.2–2.58)0.0041.51 (1.01–2.27)0.047Over 100 (0.0)19 (27.1)2.55 (1.84–3.53)< 0.0012.46 (1.38–4.39)0.002UgShs: Ugandan shillings; 1US dollar ($) = 3800ugShs (01, September, 2022); M: million; DVT: *Deep venous* thrombosis, CPPR: crude prevalence risk ratio, APPR: adjusted prevalence risk ratio
## Discussion
This study assessed the anatomical distribution and factors associated with severity of DFU among patients with DM in Uganda. The majority of patients in our study had poor blood sugar control. The most common location of DFU was the plantar region of the foot unlike other studies which reported metatarsal as the common region [22, 23]. Factors associated with severity of DFU were size of the ulcer, neuropathy and age.
The burden of DM and DFU is increasing worldwide, especially in developing countries such as Uganda [8]. Adult patients have a 10–$15\%$ risk of developing DFU during their diabetic lifetime [1, 40]. Our study showed the mean age in years was 57.5 (SD15.2), and older age was significantly associated with higher severity of DFU, especially the 70–95 age group. This result is similar to what was reported in a systematic review where age was associated with severity of DFU [9].
Although our study had a female majority, sex was not significantly associated with severity of DFU. A similar result was reported by Agwu [39] and other authors in Uganda [39, 41]. However, a male majority and preponderance have been reported in other studies [2, 42]. Jupiter [40] in a worldwide systematic and meta-analysis review in found that male patients were significantly more affected by severe DFU than female patients. In addition, level of education, profession, level of income, duration of DM/DFU are significantly associated with the severity of the DFU although these factors have been reported to contribute to the genesis of the DFU [17].
History of smoking was not associated with severity of the DFU in this study although it has been reported to be associated with development of DFU. However, other studies have found history of smoking to be associated with severity of DFU [9, 22]. This finding can be explained by the relatively low number of smokers found in this study. High BMI has been reported to be associated with occurrence of DFU and with severe DFU [9]; however, our results did not support this association. This could be explained by the finding that most patients had normal BMI.
The majority of DFU patients in this study had poor blood sugar control, although this did not appear to be a risk factor associated with severe DFU. This result is different from what has reported in a recent systematic review and by Bekele in Ethiopia where uncontrolled glycemia was significantly associated with severe DFU [9, 42]. Although poor blood sugar control was not associated with severity of DFU in this study, there is still a need to stabilize the sugar levels of all DFU patients.
The majority of DFU patients presented with severe DFU, a finding similar to a report by Smith-Strøm in Norway [43]. However, contrary findings have been reported by Jalilian and colleagues in a systematic review where a higher proportion of patients had less severe DFU [9]. Development of clear guidelines on prevention and cure of DFU with evidence-based specifics for our setting could help reduce the burden of complications due to DFU.
Most patients had Wagner classification Grade 3 DFU which is in agreement with results of a study in Sri Lanka [22]. However, studies from Ethiopia and India, have reported Wagner classification Grade 2 DFU to be the most common among diabetic patients [44–46]. This calls for emphasis on education of patients living with DM to seek medical consultation early in case of any foot wound to reduce progression to severe DFU and its complications in our setting.
Living in a rural area, having a long duration of DM of type 2 or having a history of DFU for more than 6 months was not associated with severity of DFU in this study, although these factors have been reported elsewhere to be associated with development of severe DFU [9, 17, 22].
Our results showed that DFU among patients was mostly on the right foot which concurs with the results of Smith-Strøm H in Norway [43]. Most patients in this study had DFU measuring less than 10cm2 which was similar to a study in India where DFU measuring < 10cm2was the most common [47].
Many factors have been reported to be associated with the severity of DFU. In this study, a number of factors were identified to be associated with severe DFU and these included neuropathies and the size of the ulcer. This result is similar to a study in Ethiopia where neuropathies were 4 times more likely in patients with severe DFU [47]. This study showed that the larger the ulcer, the more likely it was to be severe. Therefore, DFU should be treated early and effectively to avoid progression towards severity. Ambageda in Sri Lanka did not find any significant association between any factor and severity of DFU [22]. Several studies have reported factors associated with severity of DFU among DM patients [22, 28, 41]. A study in Norway found the duration of the ulcer to be associated with the severity whilst a systematic review by Campbell reported that high HbA1c b was associated with increased severity of DFU [22, 28, 41] and in a systematic review by Jalalian, the location of DFU on the plantar region was associated with increased severity of DFU [9]. Since factors affecting the severity of DFU seem to vary from region to region, location-specific research is recommended in order to contextualize the mitigation of complications of DFU.
## Limitations and strength of this study
The study was hospital-based and this limits the generalization of the results to the entire Ugandan population. Patients interviewed may withhold some information but, in this study, they were assured of confidentiality to avoid this bias. The other limitation is the nature of being a cross-sectional study, where the causal relationship between the contributing factors and the severity of DFU could not be established. Assessment of the vascular insufficiency should have been done using a Doppler scan or CT angiogram which are the recommended tests, however we used the ABI, which was not performed on all patients because of extensive lesions around the ankle joint in some of these patients. The study did not capture data from northern Ugandan and from all major hospitals of Eastern Uganda due to limitation of funds. The strength of this study lies in its multi-center approach and as data was collected concurrently, there is a possibility of generalizing to the catchment area of the study. However, a nationwide prospective study on severity of the DFU and outcomes in the region is proposed.
More attention should be focused upon treating aggressively the neuropathy to prevent extension of the wound.
## Conclusion
Most patients were from western Uganda and have poor health-seeking behavior leading to late consultation in diabetic clinics when the condition is rather severe. The majority of patients had severe DFU and grade 3 was the most common. The dorsum of the right foot was most affected with ulcers and the majority of patients had poor blood sugar control. Primary and secondary educational level, moderate and severe vision loss and eating frequently vegetables were less likely to be associated with severe DFU. There was an association between age of 70–95 years, neuropathies, ulcer size of more than 5 cm and severity of the DFU. Therefore, there is a need to control blood sugar, effectively treat any neuropathy as early as possible and to focus upon patient education to encourage early consultation and prevent the complications associated with severe the DFU.
## References
1. 1.Yazdanpanah L, Shahbazian H, Nazari I, Arti HR, Ahmadi F, Mohammadianinejad SE, Cheraghian B, Hesam S. Incidence and Risk Factors of Diabetic Foot Ulcer: A Population-Based Diabetic Foot Cohort (ADFC Study)-Two-Year Follow-Up Study.Int J Endocrinol. 2018 Mar15;2018:7631659.
2. 2.Jupiter DC, Thorud JC, Buckley CJ, Shibuya N. The impact of foot ulceration and amputation on mortality in diabetic patients. I: from ulceration to death, a systematic review. Int Wound J. 2016 Oct;13(5):892–903.
3. 3.Jeyaraman K, Berhane T, Hamilton M, Chandra AP, Falhammar H. Mortality in patients with diabetic foot ulcer: a retrospective study of 513 cases from a single Centre in the Northern Territory of Australia. BMC EndocrDisord. 2019 Jan 3;19(1):1.
4. 4.Armstrong DG, Boulton AJM, Bus SA. Diabetic Foot Ulcers and their recurrence. N Engl J Med. 2017 Jun;15(24):2367–75.
5. 5.Hingorani A, LaMuraglia GM, Henke P, Meissner MH, Loretz L, Zinszer KM, Driver VR, Frykberg R, Carman TL, Marston W, Mills JL, Sr, Murad MH. 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 Feb;63(2 Suppl):3S–21S.
6. Rice JB, Desai U, Cummings AK, Birnbaum HG, Skornicki M, Parsons NB. **Burden of diabetic foot ulcers for medicare and private insurers**. *Diabetes Care* (2014.0) **37** 651-8. DOI: 10.2337/dc13-2176
7. 7.IDF. Eighth edition 2017. 8 TH. Suvi Karuranga, Joao da Rocha Fernandes, Yadi Huang BM, editor. 2017. 150 p.
8. 8.Driver VR, Fabbi M, Lavery LA, Gibbons G. The costs of diabetic foot: the economic case for the limb salvage team. J Vasc Surg. 2010 Sep;52(3 Suppl):17S–22S.
9. Jalilian M, Sarbarzeh PA, Oubari S. **Factors related to severity of diabetic foot ulcer: a systematic review**. *Diabetes Metab Syndr Obes Targets Ther* (2020.0) **13** 1835-42. DOI: 10.2147/DMSO.S256243
10. 10.Zhang P, Lu J, Jing Y, Tang S, Zhu D, Bi Y. Global epidemiology of diabetic foot ulceration: a systematic review and meta-analysis †. Ann Med. 2017 Mar;49(2):106–16.
11. 11.IDF. (2019). Idf diabetes atlas (9TH ed.).
12. Mariam TG, Alemayehu A, Tesfaye E, Mequannt W, Temesgen K, Yetwale F, Limenih MA. **Prevalence of Diabetic Foot Ulcer and Associated factors among adult Diabetic Patients who attend the Diabetic Follow-Up clinic at the University of Gondar Referral Hospital, North West Ethiopia, 2016: institutional-based cross-sectional study**. *J Diabetes Res* (2017.0) **2017** 2879249. DOI: 10.1155/2017/2879249
13. 13.Bus SA, Lavery LA, Monteiro-Soares M, Rasmussen A, Raspovic A, Sacco ICN, van Netten JJ, International Working Group on the Diabetic Foot. Guidelines on the prevention of foot ulcers in persons with diabetes (IWGDF 2019 update). Diabetes Metab Res Rev. 2020 Mar;36Suppl1:e3269.
14. 14.Amin N, Doupis J. Diabetic foot disease : From the evaluation of the “ foot at risk ” to the novel diabetic ulcer treatment modalities. 2016;7(7):153–64.
15. 15.Viswanathan. Pattern and Causes of Amputation in Diabetic Patients – A Multicentric Study from India.J Assoc Physicians India ·. 2011;59(March).
16. Deribe. **Prevalence and factors influencing Diabetic Foot Ulcer among Diabetic**. *J Diabetes Metab* (2014.0) **5** 1-6. DOI: 10.4172/2155-6156.1000322
17. Zulfiqarali. **Managing the diabetic foot in resource-poor settings : challenges and solutions**. *Chronic Wound Care Manag Res* (2017.0) **4** 135-42. DOI: 10.2147/CWCMR.S98762
18. 18.Reardon R, Simring D, Kim B, Mortensen J, Williams D, Leslie A. The diabetic foot ulcer. Aust J Gen Pract. 2020 May;49(5):250–5.
19. 19.Seid. Knowledge, Practice, and Barriers of Foot Care among Diabetic Patients Attending Felege Hiwot Referral Hospital, Bahir Dar, Northwest Ethiopia.Hindawi Publ Corp. 2015;2015.
20. 20.Mehmood MK, Parkar AZ, Nayab TM, Mustafa SS, Makin MA, Alawadi F, Farghaly S. (2019). Diabetic foot self-care: awareness and practice among type 2 diabetic patients in primary healthcare centers, Dubai Health Authority. Int J Community Med Public Heal. 2019;6(1):1–7.
21. Sari Y, Upoyo AS, Isworo A, Taufik A, Sumeru A, Anandari D, Sutrisna E. **Foot self-care behavior and its predictors in diabetic patients in Indonesia**. *BMC Res Notes* (2020.0) **13** 1-6. DOI: 10.1186/s13104-020-4903-y
22. 22.Ambegoda ALAMC, Wijesekera JR, Panditharathne KI, Gamage RT, Mudalige OM D. C. S., &, Piyasiri MDRM. (2016). Analysis of Severity and Anatomical Distribution of Diabetic Foot Ulcers-A Single Unit Experience. Int J Multidiscip Stud. 2015;2(I):12–21.
23. 23.Patil A, More D, Patil A, Jadhav KA, Mejia MEV, Patil SS, Clinical. Etiological, Anatomical, and Bacteriological Study of “ Diabetic Foot ”Patients : Results of a Single Center Study. 2018;10(4).
24. 24.Bahendeka S, Wesonga R, Mutungi G, Muwonge J, Neema S, Guwatudde D. Prevalence and correlates of diabetes mellitus in Uganda: a population-based national survey. Trop Med Int Health. 2016 Mar;21(3):405–16. 10.1111/tmi.12663. Epub 2016 Jan 21. PMID: 26729021.
25. Singh A, Masuku M. **Sampling techniques & determination of sample size in Applied Statistics Research: an overview**. *Ijecm Co Uk* (2014.0) **II** 1-22
26. Assaad-Khalil SH, Zaki A, Rehim AA, Megallaa MH, Gaber N, Gamal H. **Prevalence of diabetic foot disorders and related risk factors among egyptian subjects with diabetes**. *Prim Care Diabetes [Internet]* (2015.0) **9** 297-303. DOI: 10.1016/j.pcd.2014.10.010
27. Maingi W, Kikuvi G, Matheri J. **Prevalence and factors Associated with Diabetic Foot Ulcer among adult patients attending Diabetic Clinic at Nyeri Level 5 Hospita**. *Afr J Health Sci Volume* (2020.0) **33** 44-52
28. 28.Campbell L, Pepper T, Shipman K. HbA1c: a review of non-glycaemic variables.J Clin Pathol. 2019;12–9.
29. 29.Melaku T, Wondmagegn H, Gebremickael A, Tadesse A. Patterns of superficial veins in the cubital fossa and its clinical implications among southern Ethiopian population.Anat Cell Biol. 2022 Jun30;55(2):148–154. doi: 10.5115/acb.21.217
30. 30.Shi L, Wei H, Zhang T, Li Z, Chi X, Liu D et al. A potent weighted risk model for evaluating the occurrence and severity of diabetic foot ulcers.Diabetol Metab Syndr [Internet]. 2021;1–11.
31. Nuttall FQ. *Body Mass Index Nutr Res* (2015.0) **50** 126
32. 32.Bergler-Klein J. What’s new in the ESC 2018 guidelines for arterial hypertension: the ten most important messages. Wien KlinWochenschr. 2019 Apr;131(7–8):180–5.
33. Drechsel TJ, Monteiro RL, Zippenfennig C, Ferreira JSSP, Milani TL, Sacco ICN. **Low and high frequency vibration perception thresholds can improve the diagnosis of Diabetic Neuropathy**. *J Clin Med* (2021.0) **10** 13. DOI: 10.3390/jcm10143073
34. Lai S, Ahmed U, Bollineni A, Lewis R, City K, Angeles L. *HHS Public Access* (2016.0) **15** 96-101
35. Picon AP, Ortega NRS, Watari R, Sartor C, Sacco ICN. **Classification of the severity of diabetic neuropathy: a new approach taking uncertainties into account using fuzzy logic**. *Clinics* (2012.0) **67** 151-6. DOI: 10.6061/clinics/2012(02)10
36. 36.Yu Y. Gold Standard for Diagnosis of DPN. Front Endocrinol (Lausanne). 2021 Oct 26;12:719356. doi: 10.3389/fendo.2021.719356. PMID: 34764937; PMCID: PMC8576350.
37. Dyck PJ, Bushek W, Spring EM, Karnes JL, Litchy WJ, O’Brien PC. **Vibratory and cooling detection thresholds compared with other tests in diagnosing and staging diabetic neuropathy**. *Diabetes Care* (1987.0) **10** 432-40. DOI: 10.2337/diacare.10.4.432
38. Al Ashwal SM. **Prevalence and determinants of visual impairment among School Children in Qatar**. *Int Arch Public Health Community Med* (2021.0) **5** 069
39. 39.Yelland LN, Salter AB, Ryan P. Practice of Epidemiology Performance of the Modified Poisson Regression Approach for Estimating Relative Risks From Clustered Prospective Data. 2011;174(8):984–92.
40. 40.Agwu E, Dafiewhare EO, Ekanem PE. Possible Diabetic-Foot Complications in Sub-Saharan Africa. 2011;(June 2014).
41. Atwine. **Health-care seeking behaviour and the use of traditional medicine among persons with type 2 diabetes in south-western Uganda: a study of focus group interviews**. *pana Afr Med J* (2015.0) **8688** 1-13
42. Bekele F, Chelkeba L, Fekadu G, Bekele K. **Risk factors and outcomes of diabetic foot ulcer among diabetes mellitus patients admitted to Nekemte referral hospital, western Ethiopia: prospective observational study**. *Ann Med Surg [Internet]* (2020.0) **51** 17-23. DOI: 10.1016/j.amsu.2020.01.005
43. Smith-Strøm H, Iversen MM, Igland J, Østbye T, Graue M, Skeie S. **Severity and duration of diabetic foot ulcer (DFU) before seeking care as predictors of healing time: a retrospective cohort study**. *PLoS ONE* (2017.0) **12** e0177176. DOI: 10.1371/journal.pone.0177176
44. 44.Gebrekirstos. Prevalence and Factors Associated With Diabetic Foot Ulcer among Adult Diabetes & Metabolism Prevalence and Factors Associated With Diabetic Foot Ulcer among Adult Patients in Ayder Referral Hospital Diabetic Clinic Mekelle, North. Diabetes Metab. 2016;(January 2015).
45. Jeon BJ, Choi HJ, Kang JS, Tak MS, Park ES. **Comparison of five systems of classification of diabetic foot ulcers and predictive factors for amputation**. *Int Wound J* (2017.0) **14** 537-45. DOI: 10.1111/iwj.12642
46. Zubair M, Malik A, Ahmad J. **Clinico-microbiological study and antimicrobial drug resistance profile of diabetic foot infections in North India**. *Foot* (2011.0) **21** 6-14. DOI: 10.1016/j.foot.2010.10.003
47. Asegid R, Befikadu T, Esekezaw A, Busera S. **Magnitude of Diabetic Foot Ulcer and Associated factors among Diabetic Patients who attended Diabetic follow-up clinics in Gamo and Gofa Zones, Southern Ethiopia**. *Int J Diabetes Clin Res* (2021.0) **8** 1-9. DOI: 10.23937/2377-3634/1410139
|
---
title: 'Cognitive Frailty and Functional Disability Among Community-Dwelling Older
Adults: A Systematic Review'
authors:
- Kar Foong Tang
- Pei-Lee Teh
- Shaun Wen Huey Lee
journal: Innovation in Aging
year: 2023
pmcid: PMC9999676
doi: 10.1093/geroni/igad005
license: CC BY 4.0
---
# Cognitive Frailty and Functional Disability Among Community-Dwelling Older Adults: A Systematic Review
## Abstract
### Background and Objectives
This review aimed to summarize the association between cognitive frailty (presence of frailty and cognitive impairment) and the risk of disabilities in activities of daily living (ADL), instrumental ADL (IADL), mobility, or other functional disabilities among older adults.
### Research Design and Methods
PubMed, Embase, CINAHL Plus, and PsycINFO were searched from January 2001 to May 14, 2022, for observational studies that reported cognitive frailty among community-dwelling individuals aged 60 years and above. Results were narratively synthesized.
### Results
Eleven studies encompassing 44 798 participants were included, with a prevalence of cognitive frailty ranging from $1.4\%$ to $39.3\%$. Individuals with cognitive frailty were more likely to develop disabilities in ADL and IADL compared to robust (absence of frailty and cognitive impairment) individuals. Significant disability burden and elevated risk of combined ADL/IADL disability or physical limitation among participants with cognitive frailty were reported. There was limited evidence on the association between cognitive frailty and mobility disability.
### Discussion and Implications
Individuals with cognitive frailty were likely at higher risk of developing functional disability and incurring higher disability burden than robust individuals, but evidence remains limited for those with prefrailty with cognitive impairment. Further research on this gap and standardization of cognitive frailty assessments would facilitate comparisons across populations.
### PROSPERO Registration
CRD42021232222
## Background and Objectives
Frailty is a state of increased vulnerability due to the decline in the function of various physiologic systems [1,2]. This increases an older adult’s risk of developing adverse health outcomes (3–5) and represents a public health concern in light of the global aging population. However, to date, frailty remains a dynamic yet heterogeneous concept due to a lack of consensus on its definition (6–8). The most commonly used construct is the frailty phenotype proposed by Fried and colleagues, which is characterized by shrinking (unintentional weight loss), weakness, and poor endurance as indicated by exhaustion, slowness, and low physical activity level [1]. Individuals without these characteristics are considered robust, those with 1 or 2 characteristics are considered prefrail, and those with 3 or more characteristics are considered frail.
The construct primarily focuses on the physical domain of frailty, but some studies suggest that frailty represents a multidimensional concept and that the cognitive domain should be included in the definition of frailty [9,10], as both domains are thought to be interlinked [7,8,11]. As such, the concept of cognitive frailty was introduced to encompass a broader definition of frailty, where an older adult exhibits the simultaneous presence of physical frailty and cognitive impairment in the absence of dementia [10]. The International Academy on Nutrition and Aging and the International Association of Gerontology and Geriatrics were the first international consensus group to provide this definition of cognitive frailty in older adults in 2013 [10] to conceptualize cognitive impairment due to physical frailty and not the presence of neurological condition. The definition of cognitive impairment is heterogeneous, ranging from impairment in 1 or more cognitive domains, cognitive changes reported by the patient, caregiver or physician, or impairment based on clinical dementia rating (CDR). The consensus group suggested using a comprehensive cognitive and physical frailty assessment to identify cognitive frailty (physical frailty and cognitive impairment).
Physical frailty has been widely reported to be predictive of adverse health outcomes, ranging from mortality, hospitalization, falls, fractures, disability, cognitive decline, and institutionalization (12–14). Studies on the association between physical frailty and disability have indicated that older adults with physical frailty had higher odds of developing disability than robust older adults, although this association was not found in some studies [14,15]. In comparison, evidence on the association between cognitive frailty and adverse health outcomes other than mortality, such as functional disability, is still relatively underexplored [16]. In a previous systematic review on adverse outcomes of cognitive frailty [16], 2 studies identified in the review reported on the relationship between cognitive frailty and activities of daily living (ADL) or instrumental ADL (IADL) dependency, which showed a two- to fivefold increased risk of dependence. Another review [17] reported 6 other studies which evaluated the association between cognitive frailty and disability. Of the additional 6 studies, only 1 reported on mobility disability, but the association could not be determined because all 4 participants with cognitive frailty developed mobility disability during the study follow-up period [18]. In these 2 reviews, the authors found only a limited number of studies reporting disability as an adverse outcome of cognitive frailty whereas 2 other reviews [19,20] did not specify the types of disability measured. These reflect large gaps in cognitive frailty studies, which have mostly focused on well-known adverse outcomes and less on disability due to challenges in measuring and comparing disability, especially mobility disability, which represents an important health indicator for older adults.
In line with a rapidly aging population globally, an increasing proportion of older adults will have declining functional status and an increased risk of frailty and cognitive impairment, among other age-related conditions. There remains a gap in understanding how frailty and cognitive impairment affect functional ability, such as ADL or IADL, among older adults because frailty and cognitive impairment have usually been studied separately. Physical frailty [14] and cognitive impairment [21] are known independent risk factors of ADL or IADL disability, but it is unclear if the concurrent presence of both frailty and cognitive impairment will accentuate the risk of developing functional disability. Frailty [22,23], cognitive impairment, and functional disability [24] are expected to contribute to increased healthcare burden and cost due to the increased need for health and social care services, whereas the increase in cognitive frailty prevalence over time [25] is also likely to increase adverse outcomes among older adults.
Given the implications of frailty on health outcomes and healthcare utilization [8,22], there is a need for a focused review to understand the occurrence of functional disabilities in the presence of physical frailty and cognitive impairment (cognitive frailty). This can help guide frailty prevention strategies and interventions to maintain functional ability in older adults. This review aims to summarize the association between cognitive frailty and the risk of developing functional disability among community-dwelling older adults. In this review, we included functional disabilities, namely ADL, IADL, and mobility disabilities; limitations in physical function; and other functional disabilities to provide readers with an overview of the association between cognitive frailty and risk of functional disability.
## Research Design and Methods
The systematic review protocol was registered in PROSPERO (CRD42021232222). Reporting was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [26].
## Search Strategy
A systematic search was conducted in four databases: PubMed, EMBASE, CINAHL Plus, and PsycINFO from 2001 (corresponding to the year when the Fried frailty phenotype construct was published) until May 14, 2022. The search included a combination of keywords and Medical Subject Headings terms related to frailty, older adults, ADL, disability, mobility, physical function, and adverse outcome. The list of databases and search terms are available in Supplementary Material. Reference lists of included articles and previous systematic reviews on adverse outcomes of cognitive frailty were screened to identify additional studies.
## Selection Criteria
Observational studies (cross-sectional or longitudinal cohort studies) were included if they reported on community-dwelling older adults aged 60 years and above. This age cutoff point was selected because studies on frailty typically included participants aged 60 years and above [27]. Cross-sectional, prospective cohort studies were included due to the small number of longitudinal studies that reported on the association between cognitive frailty and disability. Cognitive frailty was defined by the presence of frailty or prefrailty, and concurrent cognitive impairment was identified using validated physical frailty and cognitive assessments. A preliminary literature search has identified that most studies on cognitive frailty have slightly modified the definition of cognitive frailty by the consensus group, defining this condition with the presence of mild cognitive impairment instead of a CDR of 0.5 with the exclusion of concurrent Alzheimer’s disease or other dementias, and physical frailty using the modified Fried frailty phenotype [28]. Thus, the utilization of CDR was not compulsory for study inclusion in this review if a validated cognitive assessment tool was reported. Studies must report the association between cognitive frailty and functional disability (ADL or IADL, mobility, physical function).
Studies were excluded if they included hospitalized or institutionalized older adults or those with neurological disorders or dementia. Conference abstracts, reviews, randomized controlled trials, protocols, and studies published in other languages besides English were excluded. Study titles and abstracts were screened based on the inclusion/exclusion criteria, and full texts of relevant studies were screened for eligibility. Data were extracted using a piloted data extraction form, including study and participant characteristics, frailty assessment and classification, and corresponding disability outcomes and measurement. Data extraction was conducted by 1 reviewer (K.F.T.) and checked by the second reviewer (S.W.H.L.), with discrepancies resolved by consensus.
## Quality Assessment
The methodological quality of included studies was evaluated using the Joanna Briggs Institute’s Critical Appraisal Checklist for cohort and cross-sectional studies. Two reviewers independently conducted the assessment, with disagreements resolved by consensus.
The methodological quality assessment of studies is shown in Supplementary Table 3. All studies were judged to be representative based on sampling methods but may be biased due to recruitment procedures (e.g., recruitment at community health centers). All studies reported covariate adjustments due to confounding such as age and gender. Two studies were assessed to have an inadequate follow-up of cohorts due to attrition [33,35]. The highest risk of bias was present in outcome assessment, given that disability was based on self-reporting by participants during interviews, and there would be a possible risk of recall bias, especially among participants with cognitive impairment. All studies have excluded participants with functional disability at baseline, but 7 studies did not have exclusion criteria for participants with Parkinson’s disease, which is known to increase the risk of functional disability.
## Data Analysis
The primary outcomes were ADL, IADL, and mobility disabilities. Other outcomes include the combination of disabilities and physical function limitations. Data were synthesized narratively and summarized using a Harvest plot [29,30] due to the small number of studies, heterogeneous outcome measurement, and cognitive frailty definition. Findings were reported based on fully adjusted statistical measures with a $95\%$ confidence interval (CI). In the reporting of results, participants were categorized into 4 groups defined previously: (a) robust (absence of frailty and cognitive impairment), (b) prefrailty with cognitive impairment, (c) cognitive frailty (physical frailty and cognitive impairment), and (d) combined cognitive frailty (only for studies which grouped both frail and prefrail participants with cognitive impairment). Comparison between groups was made by using the robust group as the reference.
## Study Characteristics
A total of 18 184 records were screened, 170 studies were selected for full-text screening, and 11 records were included in this review (Supplementary Figure 1). The 11 studies included 44 798 participants, with mean age ranging from 67.7 to 75.2 years. These include 9 cohort studies with a follow-up duration of between 1 and 11 years (3,18,31–37) and 2 cross-sectional studies(Supplementary Table 1) [38,39]. Nine studies were conducted in high-income countries, and 2 were in upper–middle-income countries. No studies were identified from low- and lower–middle-income countries.
## Definition of Frailty and Cognitive Frailty by Studies
Physical frailty was evaluated using the Fried frailty phenotype [1] or its modified version in all studies except for one that used walking and grip strength measurements [39]. Older adults were classified as physically frail in the presence of 3 or more phenotype criteria, prefrail in the presence of 1 or 2 phenotype criteria, and robust in the absence of phenotype criteria. Cognitive function was most commonly assessed using the Mini-Mental State Examination (31,33,36–38). Several studies used other measures, such as the National Center for Geriatrics and Gerontology–Functional Assessment Tool (Supplementary Table 1). The definition of cognition function/impairment varied across studies as classifications were based on cognitive test score distributions or cutoff points set for the population. The prevalence of cognitive frailty reported in 11 studies varied widely, ranging from $1.4\%$ to $39.3\%$ due to differences in cognitive frailty classification.
ADL disability was reported in 5 studies [3,18,31,32,37], IADL disability in 4 studies [18,31,37,39], and mobility disability in 3 studies [18,31,37]. Participants were classified as having a disability if they were free from disability at baseline but could not perform at least 1 activity without assistance during follow-up assessment. Other functional disabilities and physical function limitations were reported in 4 studies (33–36) and are summarized in Supplementary Table 2.
## ADL disability
Five longitudinal and prospective cohort studies [3,18,31,32, 37] involving 20 778 participants described the relationship between cognitive frailty and ADL disability (Table 1). These studies reported that older adults with cognitive frailty had a higher risk of incident ADL disability than robust older adults (Figure 1). In a national cohort study in the United States [3], participants with cognitive frailty had a higher risk of incident ADL dependency (sub-hazard ratio [sHR]: 2.00; $95\%$ CI: 1.60–2.60) compared with robust participants with normal cognition after 8 years of follow-up. Similar findings were reported in a 4-year longitudinal cohort study in China [18], where cognitively frail participants had a twofold higher risk of developing ADL disability compared to robust participants (odds ratio [OR]: 2.22; $95\%$ CI: 0.97–5.08). Older adults with cognitive frailty were reported to incur a higher burden of ADL disability over 11 years (adjusted rate ratio [aRR]: 20.60; $95\%$ CI: 15.70–26.90) compared to the group without cognitive frailty [37].
## IADL disability
The association between cognitive frailty and IADL disability was reported in 3 longitudinal/prospective cohort studies and 1 cross-sectional study [18,31,37,39] encompassing 20 697 older adults (Table 1). Similar to ADL, older adults with cognitive frailty had a higher incidence and burden of IADL disability (Figure 1). Two longitudinal cohort studies in France and China showed that cognitively frail individuals were at increased risk of IADL disability compared to robust individuals (OR: 3.17; $95\%$ CI: 1.47–6.83 in [31] and OR: 2.80; $95\%$ CI: 1.00–7.87 in [18], respectively). Similarly, older adults with prefrailty and cognitive impairment had a higher IADL disability risk, with this association being significant albeit weaker than cognitive frailty [31]. In a cross-sectional study involving 8 864 Japanese older adults from the National Center for Geriatrics and Gerontology–Study of Geriatric Syndromes cohort [39], cognitive frailty was associated with IADL limitation (OR: 2.63; $95\%$ CI: 1.74–3.97).
## Mobility disability
There was very limited evidence of the association between cognitive frailty and mobility disability (Table 1). Nevertheless, the burden of mobility disability was reported to be substantially higher in participants with cognitive frailty than those without (aRR: 3.10; $95\%$ CI: 2.60–3.60) [37].
## Other Outcomes
A longitudinal cohort study in China [36] showed that the odds of incident physical function limitation were increased in individuals with prefrailty and cognitive impairment or prefrailty alone compared with robust individuals after 4 years of follow-up, but this was higher in the prefrailty and cognitive impairment group (OR: 1.78; $95\%$ CI: 1.26–2.51). In 2 studies that combined cognitive frailty and prefrailty and reported disability outcomes using different measures [33,35], the risk of disability was at least 4 or 5 times higher in the combined cognitive frailty group compared with the robust group. The risk of incident disability based on the need for long-term care certification (LTCI) in Japan was higher among participants with cognitive frailty compared with robust participants (hazard ratio [HR]: 3.86; $95\%$ CI: 2.95–5.05) [34]. The mandatory social LTCI system in Japan categorizes older adults according to levels of need based on monthly follow-up for 24 months, and the onset of long-term care/support needs in movement, ADL, IADL, and other functions denotes incident disability.
A cross-sectional study in Italy revealed that cognitively frail individuals were more likely to develop ADL–IADL disability than robust individuals, but no differences in the association with disability were observed between prefrailty with cognitive impairment and robust groups [38].
## Discussion and Implications
In this review, we found 11 studies that examined ADL, IADL, or mobility disability among older adults with cognitive frailty. We observed that the measurement of cognitive frailty and prefrailty varied, resulting in the mixed prevalence of cognitive frailty in the community setting. Across all studies, community-dwelling participants with cognitive frailty were reported to have an increased risk of functional disability, particularly in ADL and IADL, compared with robust participants with normal cognition.
Findings are in line with previous reviews on cognitive frailty [16,17], where cognitive frailty was predictive for adverse outcomes, including functional disability. However, due to the limited number of cohort studies that reported on the use of similar statistical effects or outcome measurements, evidence remains limited for robust comparisons for functional disability outcomes in cognitive frailty. Our review identified 4 additional studies that reported the association between cognitive frailty and functional disability, but similar limitations hindered a pooled analysis of findings. There were another 4 studies that reported the association between cognitive frailty and functional disability but did not meet the inclusion criteria (age [4], population characteristics [28,40], and outcome combined with mortality [41]). Although there is an agreement that cognitive frailty is associated with an increased risk of ADL or IADL disability in these studies, the use of standardized cognitive frailty or prefrailty definition and measurements would better facilitate comparisons of frailty subtypes and the risk of disability. Previous studies have similarly reported differences in predictive values on adverse outcomes depending on the operationalization of frailty/phenotype criteria or cognitive function assessments [14,16,30,42].
The association between cognitive frailty and functional disability was stronger than physical frailty in 7 of 9 longitudinal cohort studies included in this review. Nevertheless, studies to date do not allow for a comparison that could show the higher importance of assessing cognitive frailty over physical frailty or cognitive impairment alone for this adverse outcome. Previous studies have suggested that the higher risk of adverse outcomes in cognitive frailty is attributed to the cumulative risk of adverse outcomes conferred when physical frailty and cognitive impairment coexisted [16,18]. This led to the suggestion for the inclusion of cognitive function in frailty assessment to improve the identification of adverse outcomes among frail older adults [3,4,16,36]. Despite this, the synergistic interaction of frailty/prefrailty and cognitive impairment on adverse health outcomes is still not well understood [17], necessitating further research in this area. This may be important in evaluating functional disability as an adverse outcome, given that frailty or cognitive impairment alone has been associated with an increased risk of functional disability (15,21,43–45).
Among prefrail older adults with cognitive impairment, results indicated an increased odds of developing ADL or IADL disability compared with robust older adults, but associations with the risk of disability were much weaker compared to cognitive frailty in studies that reported on both frailty and prefrailty. The presence of cognitive impairment did not appear to be a significant predictor of disability among those with prefrailty. As only a few studies reported on prefrailty with cognitive impairment, and its association with primary and other outcomes (31–33), we urge caution in the interpretation of review findings for this group. As such, more research is needed to understand the association between prefrailty with cognitive impairment and adverse outcomes, given that early prefrailty or cognitive impairment may be reversible with targeted interventions, thus allowing the prevention of transition to frailty and development of functional disability [9,36].
Only a few included studies reported on mobility disability or physical function limitation among older adults with cognitive frailty. Within the literature, studies have focused on mobility decline or disability as a risk factor for frailty (46–49), which may explain the lack of reporting of mobility disability as an adverse outcome of cognitive frailty. Further research will be needed to ascertain the association between cognitive frailty and mobility disability. The assessments for mobility disability or functional disabilities other than ADL or IADL disability were also varied, which limits comparisons across populations.
This study has several limitations. Studies included in the review were heterogeneous due to differences in inclusion and exclusion criteria, participant characteristics, frailty (modified phenotype criteria), and cognitive function assessments. We included observational studies, where a causal relationship between cognitive frailty and outcomes cannot be ascertained. Hospitalized and institutionalized participants are more vulnerable to adverse health outcomes irrespective of frailty status [50] and, thus, were not included in this review, but this might limit the generalizability of the findings. We did not include non-English language articles and, therefore, might have missed related studies.
Differences in cognitive frailty prevalence were observed due to differences in the study population and cognitive frailty classifications. Future research should focus on these gaps to estimate cognitive frailty burden and inform health and social care policy, especially for lower–middle- and low-income countries. With an increased risk of functional disability, early prevention of cognitive frailty or its progression through screening of frailty and cognitive impairment is vital. The effort to incorporate this into routine health care of older adults is needed [10], given that screening is often hindered by impracticability and not widely practiced in primary care. Above all, more longitudinal cohort studies and consensus on assessing and classifying cognitive frailty/prefrailty with cognitive impairment [17] would allow better comparisons across populations.
The present review found that older adults with cognitive frailty were likely to have an increased risk of developing disabilities than robust older adults. Further research is needed to ascertain the associations between prefrailty with cognitive impairment and the risk of disability. With functional disability and frailty being emerging public health issues, further research on cognitive frailty, especially in rapidly aging populations, is needed to advance our understanding of the burden of this condition, prevention of its adverse outcomes, and maintenance of functional ability among older adults.
## Funding
The AGELESS research program was funded by the Ministry of Higher Education Long Term Research Grant Scheme (LRGS/$\frac{1}{2019}$/UM/$\frac{01}{1}$/1, LR005-2019, GOV-000047). The financial sponsor had no role in the study design, analysis and interpretation of the data, writing, or decision to submit the paper for publication.
## Conflict of Interest
None declared.
## Author Contributions
K.F.T.: conceptualization; methodology; investigation; data curation; formal analysis; validation; visualization; writing—original draft.
P.L.T.: methodology; project administration; supervision; writing—review and editing.
S.W.H.L.: conceptualization; methodology; investigation; data curation; formal analysis; project administration; supervision; validation; visualization; writing—review and editing.
## Data Availability
All data relevant to the study are included in the article or uploaded as supplementary information.
## References
1. Fried LP, Tangen CM, Walston J. **Frailty in older adults: evidence for a phenotype**. *J Gerontol A Biol Sci Med Sci* (2001) **56** M146-M157. DOI: 10.1093/gerona/56.3.M146
2. Hoogendijk EO, Afilalo J, Ensrud KE, Kowal P, Onder G, Fried LP.. **Frailty: implications for clinical practice and public health**. *Lancet* (2019) **394** 1365-1375. DOI: 10.1016/s0140-6736(19)31786-6
3. Aliberti MJR, Cenzer IS, Smith AK, Lee SJ, Yaffe K, Covinsky KE.. **Assessing risk for adverse outcomes in older adults: the need to include both physical frailty and cognition**. *J Am Geriatr Soc* (2019) **67** 477-483. DOI: 10.1111/jgs.15683
4. Feng L, Zin Nyunt MS, Gao Q, Feng L, Yap KB, Ng TP.. **cognitive frailty and adverse health outcomes: findings from the Singapore Longitudinal Ageing Studies (SLAS)**. *J Am Med Dir Assoc* (2017) **18** 252-258. DOI: 10.1016/j.jamda.2016.09.015
5. Avila-Funes JA, Helmer C, Amieva H. **Frailty among community-dwelling elderly people in France: the three-city study**. *J Gerontol A Biol Sci Med Sci* (2008) **63** 1089-1096. DOI: 10.1093/gerona/63.10.1089
6. Gobbens RJ, Luijkx KG, Wijnen-Sponselee MT, Schols JM.. **Toward a conceptual definition of frail community dwelling older people**. *Nurs Outlook* (2010) **58** 76-86. DOI: 10.1016/j.outlook.2009.09.005
7. Bergman H, Ferrucci L, Guralnik J. **Frailty: an emerging research and clinical paradigm**. *J Gerontol A Biol Sci Med Sci* (2007) **62** 731-737. DOI: 10.1093/gerona/62.7.731
8. Buckinx F, Rolland Y, Reginster J-Y, Ricour C, Petermans J, Bruyère O.. **Burden of frailty in the elderly population: perspectives for a public health challenge**. *Arch Public Health* (2015) **73** 19. DOI: 10.1186/s13690-015-0068-x
9. Azzopardi VR, Beyer I, Vermeiren S. **Increasing use of cognitive measures in the operational definition of frailty—a systematic review**. *Ageing Res Rev* (2018) **43** 10-16. DOI: 10.1016/j.arr.2018.01.003
10. Kelaiditi E, Cesari M, Canevelli M. **Cognitive frailty: rational and definition from an (I.A.N.A./I.A.G.G.) International Consensus Group.**. *J Nutr Health Aging* (2013) **17** 726-734. DOI: 10.1007/s12603-013-0367-2
11. Maaravi Y, Stessman J, Jacobs JM, Ein-Mor E.. **Frailty, cognitive impairment and mortality among the oldest old**. *J Nutr Health Aging* (2011) **15** 678-682. DOI: 10.1007/s12603-011-0096-3
12. Chang SF, Lin PL.. **Frail phenotype and mortality prediction: a systematic review and meta-analysis of prospective cohort studies**. *Int J Nurs Stud* (2015) **52** 1362-1374. DOI: 10.1016/j.ijnurstu.2015.04.005
13. Kojima G.. **Frailty as a predictor of hospitalisation among community-dwelling older people: a systematic review and meta-analysis**. *J Epidemiol Community Health* (2016) **70** 722-729. DOI: 10.1136/jech-2015-206978
14. Vermeiren S, Vella-Azzopardi R, Beckwée D. **Frailty and the prediction of negative health outcomes: a meta-analysis**. *J Am Med Dir Assoc* (2016) **17** 1163.e1-1163.e17. DOI: 10.1016/j.jamda.2016.09.010
15. Kojima G.. **Frailty as a predictor of disabilities among community-dwelling older people: a systematic review and meta-analysis**. *Disabil Rehab* (2017) **39** 1897-1908. DOI: 10.1080/09638288.2016.1212282
16. Bu Z, Huang A, Xue M, Li Q, Bai Y, Xu G.. **Cognitive frailty as a predictor of adverse outcomes among older adults: a systematic review and meta-analysis**. *Brain Behav* (2021) **11** e01926. DOI: 10.1002/brb3.1926
17. Sugimoto T, Arai H, Sakurai T.. **An update on cognitive frailty: its definition, impact, associated factors and underlying mechanisms, and interventions**. *Geriatr Gerontol Int* (2022) **22** 99-109. DOI: 10.1111/ggi.14322
18. Chen C, Park J, Wu C. **Cognitive frailty in relation to adverse health outcomes independent of multimorbidity: results from the China health and retirement longitudinal study**. *Aging (Albany NY)* (2020) **12** 23129-23145. DOI: 10.18632/aging.104078
19. Chen B, Wang M, He Q. **Impact of frailty, mild cognitive impairment and cognitive frailty on adverse health outcomes among community-dwelling older adults: a systematic review and meta-analysis**. *Front Med (Lausanne)* (2022) **9** 1009794. DOI: 10.3389/fmed.2022.1009794
20. Zhang XM, Wu XJ, Cao J, Jiao J, Chen W.. **Association between cognitive frailty and adverse outcomes among older adults: a meta-analysis**. *J Nutr Health Aging* (2022) **26** 817-825. DOI: 10.1007/s12603-022-1833-5
21. Di Carlo A, Baldereschi M, Amaducci L. **Cognitive impairment without dementia in older people: prevalence, vascular risk factors, impact on disability. The Italian Longitudinal Study on Aging**. *J Am Geriatr Soc* (2000) **48** 775-782. DOI: 10.1111/j.1532-5415.2000.tb04752.x
22. Kasajima M, Eggleston K, Kusaka S. **Projecting prevalence of frailty and dementia and the economic cost of care in Japan from 2016 to 2043: a microsimulation modelling study**. *Lancet Public Health* (2022) **7** e458-e468. DOI: 10.1016/s2468-2667(22)00044-5
23. Bock J-O, König H-H, Brenner H. **Associations of frailty with health care costs—results of the ESTHER cohort study**. *BMC Health Serv Res* (2016) **16** 128. DOI: 10.1186/s12913-016-1360-3
24. Ansah JP, Malhotra R, Lew N. **Projection of young-old and old-old with functional disability: does accounting for the changing educational composition of the elderly population make a difference?**. *PLoS One* (2015) **10** e0126471. DOI: 10.1371/journal.pone.0126471
25. Qiu Y, Li G, Wang X. **Prevalence of cognitive frailty among community-dwelling older adults: a systematic review and meta-analysis**. *Int J Nurs Stud* (2022) **125** 104112. DOI: 10.1016/j.ijnurstu.2021.104112
26. Page MJ, McKenzie JE, Bossuyt PM. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews**. *BMJ* (2021) **372** n71. DOI: 10.1136/bmj.n71
27. Spiers GF, Kunonga TP, Hall A. **Measuring frailty in younger populations: a rapid review of evidence**. *BMJ Open* (2021) **11** e047051. DOI: 10.1136/bmjopen-2020-047051
28. Solfrizzi V, Scafato E, Lozupone M. **Additive role of a potentially reversible cognitive frailty model and inflammatory state on the risk of disability: the Italian Longitudinal Study on Aging**. *Am J Geriatr Psychiatry* (2017) **25** 1236-1248. DOI: 10.1016/j.jagp.2017.05.018
29. Ogilvie D, Fayter D, Petticrew M. **The harvest plot: a method for synthesising evidence about the differential effects of interventions**. *BMC Med Res Methodol* (2008) **8** 8. DOI: 10.1186/1471-2288-8-8
30. Hanlon P, Fauré I, Corcoran N. **Frailty measurement, prevalence, incidence, and clinical implications in people with diabetes: a systematic review and study-level meta-analysis**. *Lancet Healthy Longev* (2020) **1** e106-e116. DOI: 10.1016/s2666-7568(20)30014-3
31. Avila-Funes JA, Amieva H, Barberger-Gateau P. **Cognitive impairment improves the predictive validity of the phenotype of frailty for adverse health outcomes: the three-city study**. *J Am Geriatr Soc* (2009) **57** 453-461. DOI: 10.1111/j.1532-5415.2008.02136.x
32. Ma Y, Li X, Pan Y. **Cognitive frailty predicting death and disability in Chinese elderly**. *Neurol Res* (2021) **43** 815-822. DOI: 10.1080/01616412.2021.1939235
33. Rivan NFM, Singh DKA, Shahar S. **Cognitive frailty is a robust predictor of falls, injuries, and disability among community-dwelling older adults**. *BMC Geriatr* (2021) **21** 1-13. DOI: 10.1186/s12877-021-02525-y
34. Tsutsumimoto K, Doi T, Nakakubo S, Kim M., Kurita S., Ishii H., Shimada H... **Cognitive frailty as a risk factor for incident disability during late life: a 24-month follow-up longitudinal study**. *J Nutr Health Aging* (2020) **24** 494-499. DOI: 10.1007/s12603-020-1365-9
35. Wang W, Si H, Yu R. **Effects of reversible cognitive frailty on disability, quality of life, depression, and hospitalization: a prospective cohort study**. *Aging Ment Health* (2021) **26** 2031-2038. DOI: 10.1080/13607863.2021.2011835
36. Yu R, Morley JE, Kwok T, Leung J, Cheung O, Woo J.. **The effects of combinations of cognitive impairment and pre-frailty on adverse outcomes from a prospective community-based cohort study of older Chinese people**. *Front Med (Lausanne)* (2018) **5** 50. DOI: 10.3389/fmed.2018.00050
37. Liu Z, Han L, Gahbauer EA, Allore HG, Gill TM.. **Joint trajectories of cognition and frailty and associated burden of patient-reported outcomes**. *J Am Med Dir Assoc* (2018) **19** 304-309. DOI: 10.1016/j.jamda.2017.10.010
38. Roppolo M, Mulasso A, Rabaglietti E.. **Cognitive frailty in Italian community-dwelling older adults: prevalence rate and its association with disability**. *J Nutr Health Aging* (2017) **21** 631-636. DOI: 10.1007/s12603-016-0828-5
39. Shimada H, Makizako H, Lee S. **Impact of cognitive frailty on daily activities in older persons**. *J Nutr Health Aging* (2016) **20** 729-735. DOI: 10.1007/s12603-016-0685-2
40. Sugimoto T, Ono R, Kimura A. **Impact of cognitive frailty on activities of daily living, cognitive function, and conversion to dementia among memory clinic patients with mild cognitive impairment**. *J Alzheimers Dis* (2020) **76** 895-903. DOI: 10.3233/jad-191135
41. Okura M, Ogita M, Arai H.. **Self-reported cognitive frailty predicts adverse health outcomes for community-dwelling older adults based on an analysis of sex and age**. *J Nutr Health Aging* (2019) **23** 654-664. DOI: 10.1007/s12603-019-1217-7
42. Jung HW, Kim SW, Ahn S. **Prevalence and outcomes of frailty in Korean elderly population: comparisons of a multidimensional frailty index with two phenotype models**. *PLoS One* (2014) **9** e87958. DOI: 10.1371/journal.pone.0087958
43. Liu HX, Ding G, Yu WJ. **Association between frailty and incident risk of disability in community-dwelling elder people: evidence from a meta-analysis**. *Public Health (Elsevier)* (2019) **175** 90-100. DOI: 10.1016/j.puhe.2019.06.010
44. Han L, Gill TM, Jones BL, Allore HG.. **Cognitive aging trajectories and burdens of disability, hospitalization and nursing home admission among community-living older persons**. *J Gerontol A Biol Sci Med Sci* (2016) **71** 766-771. DOI: 10.1093/gerona/glv159
45. Shimada H, Makizako H, Doi T, Tsutsumimoto K, Lee S, Suzuki T.. **Cognitive impairment and disability in older Japanese adults**. *PLoS One* (2016) **11** e0158720. DOI: 10.1371/journal.pone.0158720
46. Davis DH, Rockwood MR, Mitnitski AB, Rockwood K.. **Impairments in mobility and balance in relation to frailty**. *Arch Gerontol Geriatr* (2011) **53** 79-83. DOI: 10.1016/j.archger.2010.06.013
47. Gill TM, Allore HG, Hardy SE, Guo Z.. **The dynamic nature of mobility disability in older persons**. *J Am Geriatr Soc* (2006) **54** 248-254. DOI: 10.1111/j.1532-5415.2005.00586.x
48. Fallah N, Mitnitski A, Searle SD, Gahbauer EA, Gill TM, Rockwood K.. **Transitions in frailty status in older adults in relation to mobility: a multistate modeling approach employing a deficit count**. *J Am Geriatr Soc* (2011) **59** 524-529. DOI: 10.1111/j.1532-5415.2011.03300.x
49. Cesari M, Demougeot L, Boccalon H, Vellas B.. **Prevalence of frailty and mobility limitation in a rural setting in France**. *J Frailty Aging* (2012) **1** 169-173. DOI: 10.14283/jfa.2012.26
50. Cunha AIL, Veronese N, de Melo Borges S, Ricci NA.. **Frailty as a predictor of adverse outcomes in hospitalized older adults: a systematic review and meta-analysis**. *Ageing Res Rev* (2019) **56** 100960. DOI: 10.1016/j.arr.2019.100960
|
---
title: 'Causes of Granulomatous Inflammation in Native and Allograft Kidneys: Case
Series from A Single Center and A Review of the Literature'
authors:
- Cihan Heybeli
- Berna Demir Yuksel
- Mehtat Unlu
- Mehmet Ası Oktan
- Hayri Ustun Arda
- Ozcan Uzun
- Filiz Yıldırım
- Serkan Yıldız
- Caner Cavdar
- Aykut Sifil
- Ali Celik
- Sulen Sarıoglu
journal: Turkish Journal of Pathology
year: 2022
pmcid: PMC9999700
doi: 10.5146/tjpath.2021.01561
license: CC BY 4.0
---
# Causes of Granulomatous Inflammation in Native and Allograft Kidneys: Case Series from A Single Center and A Review of the Literature
## Abstract
Objective: Granulomatous interstitial nephritis is a rare finding, and etiology differs by geography. We aimed to investigate the distribution of causes of granuloma/granulomata in the kidney and renal survival of these patients in a tertiary care hospital in Western Turkey.
Material and Method: Medical records of adults who underwent a kidney biopsy procedure in our institution between January 2000 and June 2019 were reviewed. Pathology reports were searched for biopsies where a granuloma was identified.
Results: Nineteen of 1121 ($1.7\%$) kidney biopsies included granuloma, 17 in native kidneys, and 2 in transplants. The majority of indications for native kidney biopsy was a rise in serum creatinine. Etiologies of granuloma included the following: pauci-immune vasculitis ($$n = 11$$, $64.7\%$), tuberculosis ($$n = 2$$, $11.8\%$), drug-induced ($$n = 2$$, $11.8\%$), tubulointerstitial nephritis/uveitis (TINU) syndrome ($$n = 1$$, $5.9\%$), and systemic-lupus erythematosus ($$n = 1$$, $5.9\%$). Despite treatment, 6 of 11 ($54.5\%$) patients with vasculitis developed end-stage kidney disease (ESKD) during the median follow-up of 16 months. Both of the patients with tuberculosis, and the patient with TINU syndrome developed ESKD months after the kidney biopsy, despite appropriate therapies. The only case with drug-induced granuloma and both cases with allograft kidney granuloma responded well to glucocorticoids, achieving a complete renal recovery.
Conclusion: The majority of our series had granuloma in the kidney secondary to vasculitis and renal outcomes appear considerably unfavorable despite treatment, probably related to the primary diagnosis. Multicenter studies are needed to better determine the etiology and outcome of each granuloma etiology at different geographic locations.
## INTRODUCTION
Granulomatous interstitial nephritis (GIN) is a histological diagnosis which comprises less than $1\%$ of all kidney biopsies [1]. The primary etiology is wide an includes various systemic inflammatory disorders such as granulomatous polyangitis (GPA), sarcoidosis (2–5), Crohn’s disease [6], Sjögren syndrome [7], malignancies such as chronic lymphocytic lymphoma [8,9], fungal infections [10,11], and mycobacterial infections [12,13]. Numerous drugs may also cause GIN and these include diuretics [14], proton-pump inhibitors [15,16], non-steroidal anti-inflammatory drugs [17,18], zoledronic acid [19], captopril [20], ciprofloxacin [21], vancomycin [22], anti-TNF agents [23,24], tramadol [25], atazanavir [26,27], and immunotherapy with ipilimumab and nivolumab [28].
It is not straightforward to differentiate between these causes using renal histology alone, and the diagnosis is usually made based on the clinical presentation and extrarenal findings [29,30]. The most common cause differs between geographical regions. Drug-induced GIN and renal involvement of sarcoidosis are quite common along with idiopathic GIN in Western countries [31,32], whereas tuberculosis is the most common cause of GIN in endemic areas such as India [12,13].
The number of studies on granuloma/granulomata formation in kidney is quite limited. Despite the endemicity of tuberculosis, there is no data on the prevalence of GIN in Turkey. We therefore aimed to investigate the causes of granuloma formation in the kidney and analyze long-term outcomes of these patients.
## MATERIALS and METHODS
Medical records of adults (≥18 years of age) who underwent a kidney biopsy procedure at Dokuz Eylul University Hospital between January 2000 through June 2019 were reviewed. Kidney biopsies of adult patients (≥18 years of age) with tissue sufficient to make the diagnosis were included in the study. Patients who had granuloma/granulomata formation in kidney biopsy specimens were determined. Given the numerous etiological factors for granuloma/granulomata formation, all subjects underwent a detailed evaluation in order to determine the cause. The following data were recorded: demographic details, comorbid diseases, drug exposures, clinical findings at the time of presentation, urinalysis, full blood count, serum biochemistry (creatinine, calcium, albumin, liver function tests), autoantibodies (antinuclear antibody [ANA], anti-neutrophil cytoplasmic antibody [ANCA]), complement C3 and C4, viral serology tests (hepatitis B, hepatitis C, human immunodeficiency virus), ultrasonography of the abdomen, and chest x-ray. The following work-up was carried out patients with no apparent cause for granuloma formation (excluding drug exposures): angiotensin-converting enzyme (ACE) levels, cytomegalovirus and Epstein-*Barr virus* serologies, acid fast bacilli in the urine, polymerase chain reaction and culture of mycobacterium tuberculosis on bronchoalveolar lavage specimens, and computed tomography of the chest and abdomen.
The date of kidney biopsy was recorded as the baseline for laboratory records. Acute kidney injury was defined according to consensus criteria [33]. Rapidly progressive glomerulonephritis was defined as the loss of kidney function within days to weeks along with remarkable findings in urinalysis [34]. Microscopic hematuria was accepted if >3 red blood cells per high power field were seen in urine microscopy [35]. End-stage kidney disease (ESKD) was defined according to consensus criteria [36]. Given that the majority of cases presented with a rise in serum creatinine, the definition of response to therapy was made as follows. Complete response was a return of serum creatinine to <0.35 mg/dL above the baseline value and partial response was a return of serum creatinine to >.0.35 mg/dL but less than twice the baseline value [37].
## Renal Histopathology
Histological data was retrieved from pathology reports. Renal biopsy specimens were evaluated using hematoxylin–eosin, Masson’s trichrome, periodic acid schiff, and methenamine silver stained sections by light microscopy. Immunofluorescent analysis was made after staining for antibodies against immunoglobulins G-A-M, complement components C3 and C1q, and kappa and lambda light-chains for immunofluorescence. Electron microscopic evaluation was not routinely performed. Previous studies used the term GIN if at least 1 granuloma in kidney sections was found [32]. For this study, we have used the term ‘’granuloma/granulomata formation’’ since there is no consensus for the definition of GIN. Given that tuberculosis is endemic in our country, Ziehl-Neelsen staining was performed on kidney sections of patients with a history of pulmonary tuberculosis, on allograft kidney with granulomatous inflammation, and on patients with no identified cause of granuloma formation.
## Statistical Analysis
Quantitative variables were expressed as median with the range (minimum-maximum). Qualitative variables were expressed as proportions. Overall renal survival was estimated using the Kaplan-Meier method. Statistical analysis was performed using SPSS 22.0 version (IBM SPSS, Chicago, IL).
## RESULTS
Granuloma was identified in 19 of 1121 ($1.7\%$) kidney biopsies performed in our hospital between 2007-2019. The median age at the time of kidney biopsy was 60 (range, 20-84), and 12 ($63.2\%$) were male. Two of the biopsies were allograft kidney biopsies and seventeen were native kidney biopsies.
## Native Kidney Granuloma/Granulomata
Of the 17 native kidney granulomata, 11 ($64.7\%$) were male and the median age was 60 (range, 21-84). At baseline, hypertension and diabetes mellitus constituted $52.9\%$ (9 patients) and $29.4\%$ (5 patients) of the cohort. Indications for kidney biopsy included acute kidney injury (AKI) in 12 ($70.6\%$), rapidly progressive glomerulonephritis (RPGN) in 4 ($23.5\%$), and asymptomatic urinary abnormalities in 1 ($5.9\%$) case. The following etiologies were captured after the detailed evaluation: pauci-immune vasculitis ($$n = 11$$, $64.7\%$), tuberculosis ($$n = 2$$, $11.8\%$), drug-induced ($$n = 2$$, $11.8\%$), tubulointerstitial nephritis/uveitis (TINU) syndrome ($$n = 1$$, $5.9\%$), and systemic-lupus erythematosus ($$n = 1$$, $5.9\%$). Detailed description of each native kidney granuloma is given in Table 1.
**Table 1**
| Age/Gender | Clinical presentation | Histology | Crescent | Diagnosis | Extrarenal involvement(s) | Treatment | Outcome |
| --- | --- | --- | --- | --- | --- | --- | --- |
| 84/M | RPGN, hemoptysis, hematuria, SCr:9.1 mg/dl, MPO-ANCA+ | Necrotizing granulomatous inflammation | Yes | ANCA-associated GN | Lung | Corticosteroids, cyclophosphamide | ESKD |
| 72/F | AKI, weight loss, proteinuria, SCr:2.8mg/dl | Non-necrotizing granulomatous inflammation | No | Drug-induced GIN (penicillin) | | Corticosteroids | CR |
| 74/M | AKI, weight loss, hematuria, hepatosplenomegaly, SCr:2,3mg/dL. ANCA-neg | Severe necrotizing granulomatous inflammation. Ziehl-Neelsen (+) | Yes | Miliary tuberculosis | Bone marrow | Ethambutol-Pyrazinamide-Isoniazid-Rifampin | ESKD |
| 69/M | AKI, weight loss, fever, SCr:2,6 mg/dl, MPO-ANCA+ | Severe neutrophilic necrotizing granulomatous inflammation | Yes | ANCA-associated GN | Liver | Corticosteroids, cyclophosphamide, AZA | PR |
| 21/M | AKI, hair loss, myalgia, uveitis, oral aphthae, proteinuria, SCr:2.65 mg/dl | Severe non-necrotizing granulomatous inflammation | Yes | TINU syndrome | Eye, skin | Corticosteroids, cyclophosphamide, RTX, MMF, IVIG | ESKD |
| 62/M | AKI, weight loss, fatigue, SCr:3.74 mg/dL. ANCA-neg | Severe necrotizing granulomatous inflammation | Yes | Drug-induced (non-steroid anti-inflammatory drug) | | Corticosteroids, cyclophosphamide | ESKD |
| 67/F | AKI, hemoptysis, SCr:1.69 mg/dl, PR3-ANCA+ | Necrotizing granulomatous inflammation | Yes | ANCA-associated GN | Lung | Corticosteroids, cyclophosphamide, AZA | CR |
| 64/M | RPGN, fever, hematuria, proteinuria, SCr:8.11 mg/dl, PR3-ANCA+ | Non-necrotizing granulomatous inflammation | Yes | ANCA-associated GN | Lung | Corticosteroids, PLEX, cyclophosphamide | ESKD |
| 60/M | AKI, dyspnea, hematuria, SCr:9.87mg/dl, MPO-ANCA+ | Non-necrotizing granulomatous inflammation | Yes | ANCA-associated GN | | Corticosteroids, cyclophosphamide, PLEX, AZA | ESKD |
| 59/F | RPGN, fatigue, fever, hematuria, SCr:6.46mg/dl, MPO-ANCA+ | Necrotizing granulomatous inflammation | Yes | ANCA-associated GN | | Corticosteroids, cyclophosphamide, PLEX, AZA | ESKD |
| 80/F | RPGN, fatigue, hematuria, SCr:4.61mg/dl, MPO-ANCA+ | Necrotizing granulomatous interstitial inflammation | Yes | ANCA-associated GN | | Corticosteroids, cyclophosphamide | ESKD |
| 49/F | RPGN, weight loss, fatigue, proteinuria, SCr:4.3 mg/dl, MPO-ANCA+ | Necrotizing granulomatous inflammation | Yes | ANCA-associated GN | | Corticosteroids, cyclophosphamide, PLEX, MMF | ESKD |
| 50/F | AKI, fatigue, weight loss, proteinuria, SCr:2.61mg/dl. ANCA-neg. ANA and Anti-dsDNA positive. | Severe necrotizing granulomatous inflammation | Yes | Systemic lupus erythematosus | Joints, skin | Corticosteroids, cyclophosphamide, PLEX | PR |
| 52/M | AKI, arthralgia, hematuria, SCr :1.05 mg/dl, PR3-ANCA+ | Non-necrotizing granulomatous inflammation | Yes | ANCA-associated GN | | Corticosteroids, cyclophosphamide, AZA | CR |
| 28/M | Microscopic hematuria, proteinuria, PR3-ANCA+ | Necrotizing granulomatous inflammation | Yes | ANCA-associated GN | | ACE-inhibitor | Stable |
| 50/M | AKI, proteinuria, SCr :3.32mg/dl | Necrotizing granulomatous inflammation | No | Tuberculosis | Lung | Ethambutol-Pyrazinamide-Isoniazid-Rifampin | ESKD |
| 60/M | RPGN, SCr 3.87 mg/dl, PR3-ANCA+ | Necrotizing granulomatous inflammation | Yes | ANCA-associated GN | Lung | Corticosteroids, PLEX, cyclophosphamide | PR |
## Pauci-Immune Vasculitis
The median age of these patients was 60 (range, 28-84), and 7 ($63.6\%$) were male. Clinical presentations were as follows: AKI (6 patients, $54.5\%$), RPGN (4 patients, $30.8\%$), and asymptomatic urinary abnormalities (1 patient, $7.7\%$). The median serum creatinine at baseline was 4.46 (range, 1.69-9.87) mg/dL. Six patients required hemodialysis at the time of diagnosis. Seven of these had positive antibodies against myeloperoxidase (MPO-ANCA), and 4 had positive antibodies against proteinase-3 (PR3-ANCA). All of these patients were pauci-immune based on immunofluorescence microscopy findings. Lung involvement was evident in 5 ($38.5\%$) of patients. Histology was remarkable for severe necrotizing granulomatous inflammation with crescent formation in the majority (Figure 1). All subjects received glucocorticoids while some were also treated with a mixture of cyclophosphamide, plasma exchange, azathioprine, and/or mycophenolate mofetil. Among the 10 patients who presented with AKI or RPGN, 6 had no response, 2 had partial response, and 2 achieved a complete response. The patient who had asymptomatic urinary abnormalities did not receive immunosuppressive therapy and findings in urine persisted following therapy with angiotensin-converting enzyme inhibitor. During the median follow-up of 16 (range, 1-84) months, 6 ($54.5\%$) patients developed ESKD and 2 ($18.2\%$) of them died.
**Figure 1:** *Necrotizing granulomatous interstitial inflammation in a case with granulomatous microscopic polyangiitis (H&E, x40).*
## Tuberculosis
Two patients had tuberculosis of the kidney. Both subjects already had a diagnosis of tuberculosis of the lung by polymerase chain reaction and sputum culture prior to kidney biopsy.
The first patient was a 74-year-old-male with long-standing hypertension. The clinical presentation was AKI, with a serum creatinine level of 2.3 mg/dL. Kidney functions deteriorated and hemodialysis was initiated. There was a history of lung tuberculosis and Ziehl-Neelsen staining was positive in the kidney. In addition to severe necrotizing granulomatous inflammation, there was also crescent formation in the histology. A subsequent bone marrow biopsy also showed severe granulomatous inflammation. The patient received a combination of rifampin, isoniazide, ethambutol, and pyrazinamide, but could not come off dialysis. Unfortunately, the patient died 3 months after the kidney biopsy.
Indication for kidney biopsy in the second patient, a 50-year-old male, was AKI, with a serum creatinine of 3.32 mg/dL. Nephrotic syndrome was also evident, with 7.9 grams/24 hours of urinary protein excretion. Kidney biopsy showed severe necrotizing granulomatous intersitial nephritis. The same treatment protocol was given for tuberculosis. Unfortunately, the patient developed ESKD within 3 months after the kidney biopsy. Although both cases had a rise in serum creatinine after hospitalization, baseline serum creatinine levels were not known. They were both HIV-negative. Kidney histology revealed moderate to severe interstitial fibrosis/tubular atrophy, suggesting a preceding chronic damage.
## Tubulointerstitial Nephritis/Uveitis (TINU) Syndrome
The patient with TINU syndrome was a 21-year-old-male, who was admitted to the hospital for allopecia totalis, red eye, and malaise. Examination of the eye was compatible with anterior uveitis. Serum creatinine was 3.3 mg/dL and proteinuria was subnephrotic. Urine sediment showed pyuria but no microhematuria. Serological work-up was unremarkable. Kidney histology showed severe tubulointerstitial granulomatous inflammation with mild interstitial fibrosis and crescent formation. There was non-necrotizing granulomatous inflammation (Figure 2). Following treatment with cyclophosphamide and glucocorticoids, the alopecia and uveitis completely resolved with a mild improvement in kidney functions. However, kidney functions deteriorated during the following year requiring permanent dialysis, despite therapy with mycophenolate mofetil, rituximab, and intravenous immunoglobulin.
**Figure 2:** *Non-necrotizing granulomatous interstitial inflamma-tion in a case with tubulointerstitial nephritis/uveitis (TINU) syndrome (H&E, x40).*
## Drug-Induced GIN
The first patient with drug-induced GIN, a 72-year-old woman, presented with AKI. Abnormal test results included a serum creatinine of 2.8 mg/dl with mild (<1 g/day) proteinuria and pyuria, but no microhematuria. Penicillin was given for upper respiratory tract infection a week before the AKI event. Kidney biopsy showed granulomatous inflammation and mild interstitial fibrosis and tubular atrophy (IFTA). Work-up for possible other etiologies such as tuberculosis, sarcoidosis, and inflammatory rheumatic diseases were unremarkable. Glucocorticoids provided a complete remission with no further relapse during the following 63 months of follow-up.
The second patient was a 62-year-old male with long-standing hypertension and diabetes. The presentation was AKI, with a serum creatinine of 3.74 mg/dL, which climbed from a baseline of 1.5 mg/dL. Urine sediment was active and the kidney biopsy showed severe necrotizing granulomatous interstitial nephritis with multiple multinucleated giant cells, and crescent formation. A detailed work-up including ANA and ANCA tests, serum calcium, hepatitis serology, ACE levels, and chest CT were unremarkable. The patient stated that he received daily non-steroid anti-inflammatory drugs within the last 2 weeks for headache. Due to crescents in histology, the patient received a combination of corticosteroids and cyclophosphamide. Unfortunately, he developed ESKD with no response to therapy during the follow-up of 6 months.
## Systemic Lupus Erythematosus
This was a 50-year-old woman with a history of chronic polyarthritis in small joints of the hand. Her presentation was with AKI, with a serum creatinine of 2.61 mg/dL. Baseline serum creatinine was told to be normal. There was approximately 3 g/24 hours of urinary protein excretion, but no hematuria. A kidney biopsy showed severe necrotizing granulomatous interstitial nephritis with crescent formation. Immunofluorescence was negative for immunoglobulins and complement. ANA and Anti-double stranded DNA (Anti-dsDNA) results were positive. Induction immunosuppression included corticosteroid and cyclophosphamide. The patient had partial response and she was maintained with azathioprine. After partial response, kidney functions remained stable with no further relapse during the follow-up of 52 months.
Among all patients with native kidney granuloma formation, the median follow-up was 24 months. Ten ($58.8\%$) patients developed ESKD, and 4 ($23.4\%$) died. Overall, the median estimated renal survival of 17 patients with granuloma/granulomata in the native kidney was 12 months.
## Granuloma in the Allograft Kidney
The first patient with granuloma in the allograft kidney, a 49-year-old female, had allograft dysfunction 7 years after the kidney transplantation. The primary etiology of ESKD was chronic pyelonephritis. Serum creatinine during allograft biopsy was 1.1 mg/dL, which increased from a baseline of 0.7 mg/dL. Urine tests showed a subnephrotic proteinuria and microhematuria. Biopsy showed granulomatous inflammation with crescents. There was no lung or upper respiratory-tract involvement. Detailed work-up including ANCA tests and other autoantibodies were all negative. Following glucocorticoid therapy (1 mg/kg), she achieved a complete remission with the serum creatinine returning to baseline levels around 0.7 mg/dl. Renal function remained stable for the following 68 months.
The second patient with granuloma in the allograft kidney was a 20-year-old male. He presented with a rising serum creatinine, from 1.2 mg/dL to 2.67 mg/dL. This was 11 years after the kidney transplantation and the primary etiology of the ESKD was chronic glomerulonephritis. Native kidney biopsy was compatible with immune-complex glomerulonephritis (including C1q positivity on immunofluorescence microscopy) but no granuloma, ANA and ANCA tests were negative. After a detailed evaluation, the cause of granuloma formation could not be found, and the etiology was deemed to be idiopathic. With the introduction of 1 mg/kg of glucocorticoids and continuation of mycophenolate mofetil with calcineurin-inhibitor, he achieved a complete response with a serum creatinine returning close to the baseline levels of 1.4 mg/dL.
## DISCUSSION
With this case series, we have observed that the ANCA-associated vasculitis was the most common cause of granuloma formation in the kidney. The number of patients with tuberculosis of the kidney is probably overlooked, since kidney biopsy is rarely performed in patients with tuberculosis of the lung. There were only a few patients who had acute onset disease with GIN, while the majority had more chronic onset diseases, such as chronic rheumatic conditions, vasculitis, and infections.
Previous reports included several cases with sarcoidosis of the kidney [38]; yet we have not observed any, despite detailed diagnostic tests such as angiotensin-converting enzyme levels, computed tomography, and bronchoscopy were performed. Excluding 2 cases of drug-induced GIN and 2 allograft biopsies, more than half of our patients developed ESKD. This is probably be due to the high frequency of glomerulonephritis with crescents in our cohort and low number of cases with drug-induced GIN and absence of sarcoidosis, rather than granuloma formation itself. Thus, diseases causing acute granulomatous inflammation in the kidney, which may respond better to treatment were less frequently observed in our cohort. Indeed, Zajjari et al. stated that the outcome was good in patients with drug-induced GIN or sarcoidosis [39]. In contrast to the outcomes of our patients, Joss et al. reported quite acceptable renal response to therapies [32]. However, the authors excluded cases with crescents from their study, as these were accepted as secondary GIN. We have not excluded these subjects, since GIN frequently occurs secondary to a systemic disease such as autoimmune disorders or particular infections. Crescent formation was also evident in one of our patients with tuberculosis and the one with TINU syndrome, which would be another argument to support the inclusion of patients with crescents.
Crescent formation with granulomatous inflammation is characteristic of granulomatous polianjitis (GPA), yet granulomatous inflammation is frequently seen in lung biopsies rather than kidney specimens [40]. It is not known whether renal survival is worse among subjects with GPA who have granulomatous inflammation in the kidney versus GPA without renal granulomatosis. This issue requires further study. In a multicenter study, kidney biopsies of patients with pauci-immune crescentic glomerulonephritides were classified according to the extent of the lesions in the Bowman space, and the authors used the term ‘’full moon’’ for those who had circumferential crescents [41]. The main message of the paper was that patients with full moon crescents had more unfavorable renal survival. Interestingly, granuloma formation was more common in patients with full moon crescents. Among the pauci-immune glomerulonephritides, GPA and eosinophilic GPA are known to cause granulomatous inflammation. Although ANCA against proteinase (c-ANCA) is usually the positive antibody found in GPA, p-ANCA positivity was more common in this cohort.
Another typical chronic disease that may cause granulomatous inflammation in the kidney is tuberculosis. Renal tuberculosis is easily overlooked, and the diagnosis sometimes made post-mortem [42]. The unfavorable outcomes of our 2 cases with renal tuberculosis may be due to the delayed diagnosis. Moreover, despite a high index of clinical suspicion, the diagnosis of GIN secondary to tuberculosis may be difficult and require PCR-based techniques [43]. Microorganisms may not be detected in histological examination of the kidney. Ziehl-Neelsen staining helped only in 1 of 9 cases in the study by Agrawal and co-workers [43]. The Ziehl-Neelsen stain result was positive in 1 of 2 our cases with tuberculosis. Some authors recommend combining the auramine O stain in order to increase sensitivity and specificity for the detection of tuberculosis [44]. Timely diagnosis and early treatment of tuberculosis is associated with more favorable outcomes [42]. GIN due to tuberculosis is even more common among subjects infected with HIV, and is associated with increased mortality in that case [45]. None of our cases were HIV-positive. Similar to our paper, tuberculosis was not the predominant etiology in some of the previous reports from endemic locations [46].
One chronic systemic inflammatory disease causing GIN is the TINU syndrome. The presentation as GIN is quite rare for this syndrome [30]. The presence of a crescent in our patient with TINU is also unusual. Another unexpected thing in our case is the unfavorable outcome, despite the prescription of a mixture of immunosuppressive drugs including glucocorticoids, cyclophosphamide, and rituximab. The patient uneventfully developed ESKD and was maintained on hemodialysis. Our case is the exception rather than the rule, and more studies are needed to delineate the prognostic impact of TINU syndrome on kidney outcomes.
Similar to native kidney GIN, the etiological factors are numerous in allograft kidney GIN, including several acute and chronic disorders. Infections are the most common cause of GIN in allograft kidneys [47]. For transplant patients, it is important to determine if the etiology of ESKD recurs after kidney transplantation. Data for transplant kidney GIN are more lacking and come from case reports (Table 2). It is not clear if GIN recurs after kidney transplantation. However, recurrences of sarcoidosis [48], idiopathic GIN [49], TINU [50,51], and crescentic GN [47] were reported. The potential to recur probably depends on the etiology but the majority of case reports and the data from our study indicate a favorable survival in most patients, although some developed graft loss. Concurrent rejection episodes may occur and contribute to graft failure in some patients with GIN [52].
**Table 2**
| Study | n | Time after KTx (mean) | Etiology | Recurrence | Outcome |
| --- | --- | --- | --- | --- | --- |
| Alsaad et al.(53) | 1 | 12 years | Adenovirus | No | Recovered |
| al Sulaiman et al. (54) | 2 | ~2 years | Tbc | No | Graft loss |
| Asim et al.(55) | 1 | 31 days | Adenovirus | No | Recovered |
| Aouizerate et al. (56) | 5 | Median 12 months | Sarcoidosis | Yes | 1 death, 1 stable, 1 improved |
| Baden et al.(57) | 1 | 9 years | Coccidioidomycosis | No | Dead |
| Bagnasco et al. (58) | 1 | 12 days | Candida | Donor-transmitted | Recovered |
| Barraclough et al. (59) | 1 | 14 days | Adenovirus | No | Improvement |
| Bijol et al. (31) | 3 | 3 weeks | Bactrim (1), Unknown (2) | | |
| Brown et al.(60) | 1 | 1 year | Sarcoidosis | Yes | Recovered |
| Farris et al.(47) | 22 | Mean 552 (range, 10-5898) days | Viral (5), bacterial (5), drugs (5), Idiopathic (4), fungal (2), GPA (1) | 1 (GPA) | 22.2% graft loss due to infections, others improved/recovered |
| Gaspert et al.(61) | 1 | 2 months | Adenovirus | No | Improvement |
| Gonçalves et al. (62) | 3 | | Tbc | | |
| Hatlen et al.(63) | 1 | 6 weeks | Adenovirus | Donor-transmitted | Recovered |
| Hotta et al.(64) | 3 | 6 (3-15) months | Idiopathic (2), Drug (1) | No | All recovered |
| Josephson et al. (65) | 2 | 24 months | Drug | No | Graft-loss due to rejection |
| Khaira et al.(52) | 3 | Range, 3-13 years | Tbc | ?* | Graft loss (1), stable/improved (2) |
| Kukura et al.(66) | 1 | 3 years | Sarcoidosis | Yes | Stable |
| Lachiewicz et al. (67) | 1 | 20 months | Adenovirus | No | No improvement |
| Lapasia et al.(68) | 3 | Median 4 (range, 1-6) weeks | Infection (1), Drug (1), Idiopathic (1) | No | Improved (Infection, Drug) Graft loss (Idiopathic) |
| Lorimer et al.(69) | 4 | Range, 12-26 months | Tbc | No | Graft loss (2) Improved (2) |
| Meehan et al.(70) | 3 | Median 7 (range, 1-24) months | Tbc (1), Candida (1) E. coli /antibiotics (1) | No | 1 death (Candida), 1 stable (Tbc), 1 transplant nephrectomy (E.coli) |
| Ozdemir et al. (71) | 3 | 7.3±4.6 months | Tbc (2), Candida albicans (1) | No | Graft loss |
| Parasuraman et al. (72) | 1 | 24 months | Adenovirus | No | Recovered |
| Park et al. (73) | 1 | 10 months | Adenovirus | No | Improved |
| Shen et al.(74) | 1 | 6 years | Sarcoidosis | Yes | Recovered |
| Storsley and Gibson (75) | 1 | 6 weeks | Adenovirus | No | Improved |
| Sujeet et al.(76) | 1 | 3 years | Adenovirus | No | Recovered |
| Teranishi et al. (49) | 1 | 8 months | Idiopathic | Yes | Recovered |
| Tse et al. (77) | 1 | 9 months | Rhodococcus | No | Stabled |
| Varma et al.(78) | 1 | 14 days | Adenovirus | No | Recovered |
| Vargas et al.(48) (pediatric) | 1 | 2 years | Sarcoidosis | Yes | Died of disseminated histoplasmosis |
| Veer et al. (79) | 1 | 1 year | Adenovirus | No | Recovered |
| Zhang et al.(80) | 3 | 6 (range, 4-12) months | BKV-associated GIN (3) | No | Recovered |
Our study contains a number of limitations. This was a descriptive study with a small sample size, and etiologies of renal granuloma were determined retrospectively. In the majority of our cases, GIN was probably caused by chronic kidney damage, and it may not be reasonable to compare outcomes of acute causes of GIN versus those with subacute/chronic damage. Despite these significant limitations however, there are only a few studies on GIN and our data shows the considerably unfavorable renal survival of patients, unlike previous case series (Table 3). This is particularly the case for patients with chronic causes GIN, such as GPA. All of our cases had granuloma/granulomata formation but it is not clear if all of them should be regarded as GIN. There is no consensus for the definition of GIN, and this issue should be studied. Effects of granuloma formation on kidney outcomes may not be the same for each etiology, and it may not be straightforward to use a common definition for all. Apparently, acute and chronic causes of GIN should be separately evaluated.
**Table 3**
| Study | n | Era | Male gender | Age, mean | SCr, mean | Etiology | Renal recovery |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Agrawal et al. (43) | 17 | 2004-2014 | 64.7% | 35 | 6 ±2.3 (tbc) 4.8±1.5 (sarcoid) 2.2±1.3 (idiopathic) | Tbc (52.9%), Idiopathic (23.5%), Sarcoidosis (17.6%), Fungal (5.9%) | The majority responded to therapy. Two dialysis-dependent (Tbc and sarcoidosis), 1 mortality (tbc) |
| Bijol et al.(31) | 35 (0.5% of all biopsies) | 1987-2004 | 50% | 52 (range, 21-84) | 4.1 (range, 1.5-12.8) | Sarcoidosis (28.9%), Drug (44.7%) GPA (15.9%) | |
| Gupta et al. (12) | 16 (1.08% of all biopsies) | 2009-2013 | 62.5% | 34 (range, 12-68) | 6.25 ± 3.53 | Tbc (56.3%), Cresc. GN (12.5%) Idiopathic (12.5%), Drugs (12.5%) Infection (6.2%) | The majority responded to therapy. Two dialysis-dependent (Tbc), 1 underwent transplantation (Tbc) |
| Javaud et al. (81) | 40 (1.37% of all biopsies) | 1991-2004 | 62.5% | 53 | median GFR 26 mL/min (range, 5-80) | Sarcoidosis (50%), Drug (17.5%) Tbc (7.5%), GPA (5%), Leprosy (2.5%), M. avium (2.5%), Crohn (2.5%) | 3 dialysis-dependent (GPA, Crohn, drug-induced), 1 transplant (sarcoidosis), 2 mortality (drug-induced GIN, Tbc) |
| Joss et al.(32) | 18 | 1990-2004 | 61% | 55 | 4.21 (1.15 to 15.41) | Idiopathic (50%), Sarcoidosis (28%), Drug (11%), TINU (11%) | None required long-term renal replacement therapy |
| Karmakar et al. (46) | 6 | | 33.3% | range, 14-65 | range, 0.9 - 7.13 | SLE (33.3%), Cresc. GN (33.3%) Idiopathic (33.3%) | |
| Mignon et al. (82) | 32 (0.9% of all biopsies) | | | 20-76 | | Drug (31.2%), GPA (25%), Idiopathic (25%), Sarcoidosis (9.3%), Tbc (9.3%) | Most recovered or stable, 5 died, 1 required long-term dialysis. |
| Naidu et al. (83) | 14 (0.5% of all biopsies) | 2000 to 2012 | 57.1% | 35 (range, 20-70) | 6.7±3.8 (2.3-14.7) | Tbc (64.3%), Drug (14.4%), SLE (7.1%), GPA (7.1%), IgAN (7.1%) | 5 dialysis, 1 transplant, 8 recovered/improved |
| Oliveira et al. (29) | 21 | 2000-2012 | 57% | 53 (range, 19-73) | GFR, range, 11-113 ml/min | Sarcoidosis (62%), Tbc (24%), Idiopathic (10%), Drugs (5%) | 1 death (idiopathic), 1 dialysis (tbc) |
| Viero and Cavallo (84) | 12 (5.9% of all biopsies) | 1974-1994 | 33.3% | 46 (range, 24-78) | 5.1 (range, 1.9-8.7) | Drugs (25%), Sarcoidosis (25%), Infections (25%), Oxalosis (8%), GPA (8%), Idiopathic (8%) | The majority were lost to follow-up. One mortality (infection). Three developed chronic renal failure. |
| Zajjari et al. (39) | 11 (2.7% of all biopsies) | | 36.4% | 44.2 | 3.91 ± 2.07 | Sarcoidosis (45.4%), Drugs (27.2%) | Patients with drug-induced GIN and sarcoidosis recovered, but no renal recovery in other etiologies |
In conclusion, ANCA-associated vasculitis appeared to be the most common cause of granuloma formation in the kidney in our study. Renal survival is significantly shortened and multicenter studies are needed in order to delineate the nature of different etiologies of granuloma formation in the kidney, and determine the best treatment option for each category.
## Conflict of Interest
The authors declare no conflict of interest.
## Ethics approval
This study was approved by the Ethics Committee of Dokuz Eylül University School of Medicine (IRB code: $\frac{2019}{17}$-230307).
## Informed consent
Informed consent was waived due to the retrospective design, confidentiality of patient identity, and absence of any invasive procedures.
## References
1. Shah Shivani, Carter-Monroe Naima, Atta Mohamed G.. **Granulomatous interstitial nephritis**. *Clin Kidney J* (2015) **8** 516. DOI: 10.1093/ckj/sfv053
2. O’Riordan E., Willert R. P., Reeve R., Kalra P. A., O’Donoghue D. J., Foley R. N., Waldek S.. **Isolated sarcoid granulomatous interstitial nephritis: review of five cases at one center**. *Clin Nephrol* (2001) **55** 297. PMID: 11334315
3. Robson Michael G., Banerjee Debasish, Hopster Debbie, Cairns Hugh S.. **Seven cases of granulomatous interstitial nephritis in the absence of extrarenal sarcoid**. *Nephrol Dial Transplant* (2003) **18** 280. DOI: 10.1093/ndt/18.2.280
4. Tsukada Hiroyuki, Shimizu Hideki, Kaname Shinya. **Bilateral nephromegaly due to isolated renal sarcoidosis**. *Clin Exp Nephrol* (2017) **21** 169. DOI: 10.1007/s10157-016-1264-x
5. Utaş C., Doğukan A., Patiroğlu T. E., Oymak S., Oymak O.. **Granulomatous interstitial nephritis in extrapulmonary sarcoidosis**. *Clin Nephrol* (1999) **51** 252. PMID: 10230559
6. Timmermans Sjoerd A. M. E. G., Christiaans Maarten H. L., Abdul-Hamid Myrurgia A., Stifft Frank, Damoiseaux Jan G. M. C., Paassen Pieter Van. **Granulomatous interstitial nephritis and Crohn’s disease**. *Clin Kidney J* (2016) **9** 556. DOI: 10.1093/ckj/sfw041
7. Bitik B., Gonul I. I., Haznedaroglu S., Goker B., Tufan A.. **Granulomatous interstitial nephritis associated with Primary Sjögren’s syndrome**. *Z Rheumatol* (2017) **76** 458. DOI: 10.1007/s00393-017-0269-5
8. Inoue Tsutomu, Sato Takahiko, Okada Hirokazu, Kayano Hidekazu, Watanabe Yusuke, Kikuta Tomohiro, Tsuda Masahiro, Sueyoshi Keita, Takenaka Tsuneo, Suzuki Hiromichi. **Granulomatous interstitial nephritis in chronic lymphocytic leukaemia**. *Nephrol Dial Transplant* (2010) **25** 4107. DOI: 10.1093/ndt/gfq489
9. Nasr Samih H., Shanafelt Tait D., Hanson Curtis A., Fidler Mary E., Cornell Lynn D., Sethi Sanjeev, Chaffee Kari G., Morris Joseph, Leung Nelson. **Granulomatous interstitial nephritis secondary to chronic lymphocytic leukemia/small lymphocytic lymphoma**. *Ann Diagn Pathol* (2015) **19** 130. DOI: 10.1016/j.anndiagpath.2015.03.003
10. Nasr Samih H., Koscica John, Markowitz Glen S., D’Agati Vivette D.. **Granulomatous interstitial nephritis**. *Am J Kidney Dis* (2003) **41** 714. DOI: 10.1053/ajkd.2003.50143
11. Ogura Makoto, Kagami Shino, Nakao Masatsugu, Kono Midori, Kanetsuna Yukiko, Hosoya Tatsuo. **Fungal granulomatous interstitial nephritis presenting as acute kidney injury diagnosed by renal histology including PCR assay**. *Clin Kidney J* (2012) **5** 459. DOI: 10.1093/ckj/sfs103
12. Gupta Pallav, Rana D. S., Bhalla A. K., Gupta Ashwini, Malik Manish, Gupta Anurag, Bhargava Vinant. **Renal failure due to granulomatous interstitial nephritis in native and allograft renal biopsies: experience from a tertiary care hospital**. *Ren Fail* (2014) **36** 1468. DOI: 10.3109/0886022X.2014.950975
13. Sampathkumar Krishnaswamy, Sooraj Yesudas S., Mahaldar Amol R., Ramakrishnan Muthiah, Rajappannair Ajeshkumar, Nalumakkal Seethalekshmy V., Erode Elango. **Granulomatous interstitial nephritis due to tuberculosis-a rare presentation**. *Saudi J Kidney Dis Transpl* (2009) **20** 842. PMID: 19736486
14. Enríquez R., Cabezuelo J. B., González C., Lacueva J., Teruel A., Fernández J., Arenas M. D.. **Granulomatous interstitial nephritis associated with hydrochlorothiazide/amiloride**. *Am J Nephrol* (1995) **15** 270. DOI: 10.1159/000168845
15. Geevasinga Nimeshan, Coleman Patrick L., Roger Simon D.. **Rabeprazole-induced acute interstitial nephritis**. *Nephrology (Carlton)* (2005) **10** 7. DOI: 10.1111/j.1440-1797.2005.00365.x
16. Nadri Quaid, Althaf Mohammed Mahdi. **Granulomatous tubulointerstitial nephritis secondary to omeprazole**. *BMJ Case Rep* (2014) **2014** bcr2014203842. DOI: 10.1136/bcr-2014-203842
17. Jung Kyungsuk, Zeng Xu, Bilusic Marijo. **Nivolumab-associated acute glomerulonephritis: a case report and literature review**. *BMC Nephrol* (2016) **17** 188. DOI: 10.1186/s12882-016-0408-2
18. Schwarz A., Krause P. H., Keller F., Offermann G., Mihatsch M. J.. **Granulomatous interstitial nephritis after nonsteroidal anti-inflammatory drugs**. *Am J Nephrol* (1988) **8** 410. DOI: 10.1159/000167627
19. Avguštin Nuša, Kovač Damjan, Kojc Nika, Mlinšek Gregor, Lindič Jelka. **Acute granulomatous interstitial nephritis and acute rejection in a kidney transplant recipient after zoledronic acid therapy - a case report and review of the literature**. *Clin Nephrol* (2017) **88** 97. DOI: 10.5414/CNP88FX22
20. Islam S., Dubigeon M. P., Guenel J.. **[Acute and reversible interstitial granulomatous nephropathy after treatment with captopril]**. *Rev Med Interne* (1990) **11** 231. PMID: 2096422
21. Goli R., Mukku K. K., Raju S. B., Uppin M. S.. **Acute Ciprofloxacin-Induced Crystal Nephropathy with Granulomatous Interstitial Nephritis**. *Indian J Nephrol* (2017) **27** 231. DOI: 10.4103/0971-4065.200522
22. Hong Susana, Valderrama Elsa, Mattana Joseph, Shah Hitesh H., Wagner John D., Esposito Michael, Singhal Pravin C.. **Vancomycin-induced acute granulomatous interstitial nephritis: therapeutic options**. *Am J Med Sci* (2007) **334** 296. DOI: 10.1097/MAJ.0b013e3180a6ec1e
23. Korsten Peter, Sweiss Nadera J., Nagorsnik Ulf, Niewold Timothy B., Gröne Hermann-Josef, Gross Oliver, Müller Gerhard A.. **Drug-induced granulomatous interstitial nephritis in a patient with ankylosing spondylitis during therapy with adalimumab**. *Am J Kidney Dis* (2010) **56** e17. DOI: 10.1053/j.ajkd.2010.08.019
24. Sandys Vicki, Moloney Brona, Lane Louise, Qazi Junaid, Doyle Brendan, Barry Maurice, Leavey Sean, Conlon Peter. **Granulomatous interstitial nephritis secondary to adalimumab therapy**. *Clin Kidney J* (2018) **11** 219. DOI: 10.1093/ckj/sfx104
25. Borrego Utiel Francisco J., Luque Barona Rafael, Pérez Del Barrio Pilar, Borrego Hinojosa Josefa, Ramírez Tortosa César. **Acute Kidney Injury due to granulomatous interstitial nephritis induced by tramadol administration**. *Nefrologia (Engl Ed)* (2018) **38** 227. DOI: 10.1016/j.nefro.2017.05.005
26. Chu Gerard Joseph, Henderson Christopher, Evans Louise, Howlin Kenneth, Murad Ari. **Chronic granulomatous interstitial nephritis and urothelial metaplasia associated with ritonavir-boosted atazanavir: a case study and literature review**. *Pathology* (2018) **50** 565. DOI: 10.1016/j.pathol.2018.01.009
27. Varghese Vipin, Rodriguez Rafael, Self Sally, Velez Juan Carlos Q.. **Atazanavir Crystal-Induced Chronic Granulomatous Interstitial Nephritis**. *Kidney Int Rep* (2020) **5** 1106. DOI: 10.1016/j.ekir.2020.04.007
28. Bottlaender Lea, Breton Anne-Laure, Laforcade Louis De, Dijoud Frederique, Thomas Luc, Dalle Stephane. **Acute interstitial nephritis after sequential ipilumumab - nivolumab therapy of metastatic melanoma**. *J Immunother Cancer* (2017) **5** 57. DOI: 10.1186/s40425-017-0261-2
29. Oliveira Ben, Jayawardene Satish, Shah Sapna. **Single-centre experience of granulomatous interstitial nephritis-time for a new approach?**. *Clin Kidney J* (2017) **10** 249. DOI: 10.1093/ckj/sfw119
30. Shah Kabeer K., Pritt Bobbi S., Alexander Mariam P.. **Histopathologic review of granulomatous inflammation**. *J Clin Tuberc Other Mycobact Dis* (2017) **7** 1. DOI: 10.1016/j.jctube.2017.02.001
31. Bijol Vanesa, Mendez Gonzalo P., Nosé Vânia, Rennke Helmut G.. **Granulomatous interstitial nephritis: a clinicopathologic study of 46 cases from a single institution**. *Int J Surg Pathol* (2006) **14** 57. DOI: 10.1177/106689690601400110
32. Joss Nicola, Morris Scott, Young Barbara, Geddes Colin. **Granulomatous interstitial nephritis**. *Clin J Am Soc Nephrol* (2007) **2** 222. DOI: 10.2215/CJN.01790506
33. **KDIGO clinical practice guideline for acute kidney injury**. *Kidney Int Suppl* (2012) **2** 1
34. Jennette J. Charles. **Rapidly progressive crescentic glomerulonephritis**. *Kidney Int* (2003) **63** 1164. DOI: 10.1046/j.1523-1755.2003.00843.x
35. Grossfeld G. D., Wolf J. S., Litwan M. S., Hricak H., Shuler C. L., Agerter D. C., Carroll P. R.. **Asymptomatic microscopic hematuria in adults: summary of the AUA best practice policy recommendations**. *Am Fam Physician* (2001) **63** 1145. PMID: 11277551
36. Levey Andrew S., Jong Paul E. De, Coresh Josef, El Nahas Meguid, Astor Brad C., Matsushita Kunihiro, Gansevoort Ron T., Kasiske Bertram L., Eckardt Kai-Uwe. **The definition, classification, and prognosis of chronic kidney disease: a KDIGO Controversies Conference report**. *Kidney Int* (2011) **80** 17. DOI: 10.1038/ki.2010.483
37. Cortazar Frank B., Kibbelaar Zoe A., Glezerman Ilya G., Abudayyeh Ala, Mamlouk Omar, Motwani Shveta S., Murakami Naoka, Herrmann Sandra M., Manohar Sandhya, Shirali Anushree C., Kitchlu Abhijat, Shirazian Shayan, Assal Amer, Vijayan Anitha, Renaghan Amanda DeMauro, Ortiz-Melo David I., Rangarajan Sunil, Malik A. Bilal, Hogan Jonathan J., Dinh Alex R., Shin Daniel Sanghoon, Marrone Kristen A., Mithani Zain, Johnson Douglas B., Hosseini Afrooz, Uprety Deekchha, Sharma Shreyak, Gupta Shruti, Reynolds Kerry L., Sise Meghan E., Leaf David E.. **Clinical Features and Outcomes of Immune Checkpoint Inhibitor-Associated AKI: A Multicenter Study**. *J Am Soc Nephrol* (2020) **31** 435. DOI: 10.1681/ASN.2019070676
38. Zammouri Asma, Barbouch Samia, Najjar Mariem, Aoudia Raja, Jaziri Fatima, Kaaroud Hayet, Hedri Hafedh, Abderrahim Ezzeddine, Goucha Rim, Hamida Fethi Ben, Harzallah Amel, Abdallah Taieb Ben. **Tubulointerstitial nephritis due to sarcoidosis: Clinical, laboratory, and histological features and outcome in a cohort of 24 patients**. *Saudi J Kidney Dis Transpl* (2019) **30** 1276. DOI: 10.4103/1319-2442.275471
39. Zajjari Y., Azizi M., Bahadi A., Montasser D., Aatif T., Alayoud A., El Kabbaj D.. **Granulomatous interstitial nephritis in the military hospital of Morocco: Causes and outcomes**. *Saudi J Kidney Dis Transpl* (2020) **31** 589. PMID: 32655045
40. Bajema I. M., Hagen E. C., Woude F. J. Van der, Bruijn J. A.. **Wegener’s granulomatosis: a meta-analysis of 349 literary case reports**. *J Lab Clin Med* (1997) **129** 17. DOI: 10.1016/s0022-2143(97)90157-8
41. Unlu Mehtat, Kiremitci Saba, Ensari Arzu, Ozluk Yasemin, Kilicaslan Isin, Ozdemir Binnaz Handan, Ates Deniz, Ertoy Baydar Dilek, Gonul Ipek Isik, Memis Leyla, Sarsik Banu, Sen Sait, Akkaya Bahar, Orhan Diclehan, Gonlusen Gulfiliz, Ellidokuz Hulya, Ada Sibel, Cavdar Caner, Akagun Tulin, Kamali Sevil, Aksu Kenan, Yazisiz Veli, Paydas Saime, Soylu Alper, Sarioglu Sulen. **Pauci-immune necrotizing crescentic glomerulonephritis with crescentic and full moon extracapillary proliferation: clinico-pathologic correlation and follow-up study**. *Pathol Res Pract* (2013) **209** 75. DOI: 10.1016/j.prp.2012.10.012
42. Eastwood John B., Corbishley Catherine M., Grange John M.. **Tuberculosis and the kidney**. *J Am Soc Nephrol* (2001) **12** 1307. DOI: 10.1681/ASN.V1261307
43. Agrawal Vinita, Kaul Anupama, Prasad Narayan, Sharma Kusum, Agarwal Vikas. **Etiological diagnosis of granulomatous tubulointerstitial nephritis in the tropics**. *Clin Kidney J* (2015) **8** 524. DOI: 10.1093/ckj/sfv071
44. Larsen Christopher P., Moreira Roger K., Hennigar Randolph A., Bijol Vanesa. **Kidney biopsy findings in a patient with fever, bilateral pulmonary infiltrates, and acute renal failure**. *Am J Kidney Dis* (2008) **51** 524. DOI: 10.1053/j.ajkd.2007.11.014
45. Davidson Bianca, Nel Debbie, Jones Erika S. W., Manning Kathryn, Spies Ruan, Bohmer Raphaela, Omar Aadil, Ash Samantha, Wearne Nicola. **Granulomatous interstitial nephritis on renal biopsy in human immunodeficiency virus positive patients: Prevalence and causes in Cape Town, South Africa**. *Nephrology (Carlton)* (2019) **24** 681. DOI: 10.1111/nep.13564
46. Karmakar Subhrajyoti, Basu Keya, Sengupta Moumita, Sircar Dipankar, Roychowdhury Arpita. **Granulomatous Interstitial Nephritis - A Series of Six Cases**. *Indian J Nephrol* (2020) **30** 26. DOI: 10.4103/ijn.IJN_364_18
47. Farris Alton B., Ellis Carla L., Rogers Thomas E., Chon W. James, Chang Anthony, Meehan Shane M.. **Renal allograft granulomatous interstitial nephritis: observations of an uncommon injury pattern in 22 transplant recipients**. *Clin Kidney J* (2017) **10** 240. DOI: 10.1093/ckj/sfw117
48. Vargas Federico, Gedalia Abraham, Craver Randall D., Matti Vehaskari V.. **Recurrence of granulomatous interstitial nephritis in transplanted kidney**. *Pediatr Transplant* (2010) **14** e54. DOI: 10.1111/j.1399-3046.2009.01173.x
49. Teranishi J., Hattori Y., Mochizuki T., Kawahara T., Makiyama K., Uemura H.. **Effective Treatment With Everolimus for Recurrent Granulomatous Interstitial Nephritis in a Renal Transplant Recipient: A Case Report**. *Transplant Proc* (2016) **48** 946. DOI: 10.1016/j.transproceed.2016.01.034
50. Murtaza Asam, Ali Hatem, Nair Beena, Ahmed Aimun K.. **Recurrence of Tubulointerstitial Nephritis Without Uveitis, in a Patient with Previous Tubulointerstitial Nephritis and Uveitis Syndrome After Transplant**. *Exp Clin Transplant* (2018) **16** 216. DOI: 10.6002/ect.2015.0329
51. Onyekpe Isioma, Shenoy Mohan, Denley Helen, Riad Hany, Webb Nicholas J. A.. **Recurrent tubulointerstitial nephritis and uveitis syndrome in a renal transplant recipient**. *Nephrol Dial Transplant* (2011) **26** 3060. DOI: 10.1093/ndt/gfr352
52. Khaira Ambar, Bagchi Soumita, Sharma Alok, Mukund Amar, Mahajan Sandeep, Bhowmik Dipankar, Dinda Amit K., Agarwal Sanjay K.. **Renal allograft tuberculosis: report of three cases and review of literature**. *Clin Exp Nephrol* (2009) **13** 392. DOI: 10.1007/s10157-009-0158-6
53. Alsaad Khaled O., Tobar Ana, Belanger Eric, Ahmad Mufazzal, Cattran Daniel C., Herzenberg Andrew M.. **Late-onset acute haemorrhagic necrotizing granulomatous adenovirus tubulointerstitial nephritis in a renal allograft**. *Nephrol Dial Transplant* (2007) **22** 1257. DOI: 10.1093/ndt/gfl843
54. Sulaiman M. H. Al-, Dhar J. M., Hasani M. K. Al-, Haleem A., Khader A. Al-. **Tuberculous interstitial nephritis after kidney transplantation**. *Transplantation* (1990) **50** 162. DOI: 10.1097/00007890-199007000-00032
55. Asim Muhammad, Chong-Lopez Agustin, Nickeleit Volker. **Adenovirus infection of a renal allograft**. *Am J Kidney Dis* (2003) **41** 696. DOI: 10.1053/ajkd.2003.50133
56. Aouizerate Jessie, Matignon Marie, Kamar Nassim, Thervet Eric, Randoux Christine, Moulin Bruno, Raffray Loic, Buchler Matthias, Villar Emmanuel, Mahevas Mathieu, Desvaux Dominique, Dahan Karine, Diet Carine, Audard Vincent, Lang Philippe, Grimbert Philippe. **Renal transplantation in patients with sarcoidosis: a French multicenter study**. *Clin J Am Soc Nephrol* (2010) **5** 2101. DOI: 10.2215/CJN.03970510
57. Baden Lindsey R., Digumarthy Subba R., Guimaraes Alexander S. R., Branda John A.. **Case records of the Massachusetts General Hospital. Case 35-2009. A 60-year-old male renal-transplant recipient with renal insufficiency, diabetic ketoacidosis, and mental-status changes**. *N Engl J Med* (2009) **361** 1980. DOI: 10.1056/NEJMcpc0900645
58. Bagnasco S. M., Subramanian A. K., Desai N. M.. **Fungal infection presenting as giant cell tubulointerstitial nephritis in kidney allograft**. *Transpl Infect Dis* (2012) **14** 288. DOI: 10.1111/j.1399-3062.2011.00676.x
59. Barraclough Katherine, Oliver Kimberley, Playford Elliot Geoffrey, Preston John, Campbell Scott, Johnson David W., Hawley Carmel, Mudge David, Eps Carolyn Van, Isbel Nicole. **Life-threatening adenovirus infection in a kidney transplant recipient**. *NDT Plus* (2009) **2** 250. DOI: 10.1093/ndtplus/sfp003
60. Brown J. H., Jos V., Newstead C. G., Lawler W.. **Sarcoid-like granulomata in a renal transplant**. *Nephrol Dial Transplant* (1992) **7** 173. DOI: 10.1093/oxfordjournals.ndt.a092094
61. Gaspert Ariana, Lüthi Beatrice, Mueller Nicolas J., Bossart Walter, Heim Albert, Wüthrich Rudolf P., Fehr Thomas. **Subacute allograft failure with dysuria and hematuria in a kidney transplant recipient**. *Am J Kidney Dis* (2009) **54** 154. DOI: 10.1053/j.ajkd.2008.11.008
62. Gonçalves A. R., Caetano M. A., Paula F. J., Ianhez L. E., Saldanha L. B., Sabbaga E.. **Tuberculous interstitial granulomatous nephritis in renal transplants: report of three cases**. *Transplant Proc* (1992) **24** 1911
63. Hatlen Timothy, Mroch Henry, Tuttle Katherine, Ojogho Okechukwu, Rooney Michele, Desmond Sara, Bani-Hani Samer. **Disseminated Adenovirus Nephritis After Kidney Transplantation**. *Kidney Int Rep* (2018) **3** 19. DOI: 10.1016/j.ekir.2017.08.004
64. Hotta Kiyohiko, Fukasawa Yuichiro, Sasaki Hajime, Seki Toshimori, Togashi Masaki, Harada Hiroshi. **Granulomatous tubulointerstitial nephritis in a renal allograft: three cases report and review of literature**. *Clin Transplant* (2012) **26 Suppl 24** 70. DOI: 10.1111/j.1399-0012.2012.01643.x
65. Josephson M. A., Chiu M. Y., Woodle E. S., Thistlethwaite J. R., Haas M.. **Drug-induced acute interstitial nephritis in renal allografts: histopathologic features and clinical course in six patients**. *Am J Kidney Dis* (1999) **34** 540. DOI: 10.1016/s0272-6386(99)70083-1
66. Kukura Stefan, Viklicky Ondrej, Lácha Jirí, Voska Ludek, Honsová Eva, Teplan Vladimír. **Recurrence of sarcoidosis in renal allograft during pregnancy**. *Nephrol Dial Transplant* (2004) **19** 1640. DOI: 10.1093/ndt/gfh197
67. Lachiewicz A. M., Cianciolo R., Miller M. B., Derebail V. K.. **Adenovirus causing fever, upper respiratory infection, and allograft nephritis complicated by persistent asymptomatic viremia**. *Transpl Infect Dis* (2014) **16** 648. DOI: 10.1111/tid.12248
68. Lapasia Jessica B., Kambham Neeraja, Busque Stéphan, Tan Jane C.. **Renal allograft granulomas in the early post-transplant period**. *NDT Plus* (2010) **3** 397. DOI: 10.1093/ndtplus/sfq081
69. Lorimer I., Botha J., Pontin A. R., Pascoe M. D., Kahn D.. **Tuberculosis isolated to the renal allograft**. *Transpl Infect Dis* (1999) **1** 83. DOI: 10.1034/j.1399-3062.1999.10109.x
70. Meehan S. M., Josephson M. A., Haas M.. **Granulomatous tubulointerstitial nephritis in the renal allograft**. *Am J Kidney Dis* (2000) **36** E27. DOI: 10.1053/ajkd.2000.17735
71. Ozdemir B. H., Sar A., Uyar P., Suren D., Demirhan B., Haberal M.. **Posttransplant tubulointerstitial nephritis: clinicopathological correlation**. *Transplant Proc* (2006) **38** 466. DOI: 10.1016/j.transproceed.2005.12.050
72. Parasuraman Ravi, Zhang Ping L., Samarapungavan Dilip, Rocher Leslie, Koffron Alan. **Severe necrotizing adenovirus tubulointerstitial nephritis in a kidney transplant recipient**. *Case Rep Transplant* (2013) **2013** 969186. DOI: 10.1155/2013/969186
73. Park U. J., Hyun S. K., Kim H. T., Cho W. H., Han S. Y.. **Successful treatment of disseminated adenovirus infection with ribavirin and intravenous immunoglobulin in an adult renal transplant recipient: a case report**. *Transplant Proc* (2015) **47** 791. DOI: 10.1016/j.transproceed.2014.11.054
74. Shen S. Y., Hall-Craggs M., Posner J. N., Shabazz B.. **Recurrent sarcoid granulomatous nephritis and reactive tuberculin skin test in a renal transplant recipient**. *Am J Med* (1986) **80** 699. DOI: 10.1016/0002-9343(86)90828-4
75. Storsley Leroy, Gibson Ian W.. **Adenovirus interstitial nephritis and rejection in an allograft**. *J Am Soc Nephrol* (2011) **22** 1423. DOI: 10.1681/ASN.2010090941
76. Sujeet K., Vasudev B., Desai P., Bellizzi J., Novoa-Takara L., He C., El-Meanawy A.. **Acute kidney injury requiring dialysis secondary to adenovirus nephritis in renal transplant recipient**. *Transpl Infect Dis* (2011) **13** 174. DOI: 10.1111/j.1399-3062.2010.00577.x
77. Tse Kai Chung, Chan Tak Mao, Wong Samson Sai-Yin, Chan Kwok Wah, Lam Man Fai, Chan Kingsley Hau-Ngai, Li Fu Keung, Choy Bo Ying, Lai Kar Neng. **Granulomatous interstitial nephritis of the allograft kidney associated with rhodococcal pulmonary infection**. *Nephrol Dial Transplant* (2004) **19** 486. DOI: 10.1093/ndt/gfg511
78. Varma M. C., Kushner Y. B., Ko D. S., Kawai T., Martins P. N., Martins P., Kaur P., Markmann J. F., Kotton C. N.. **Early onset adenovirus infection after simultaneous kidney-pancreas transplant**. *Am J Transplant* (2011) **11** 623. DOI: 10.1111/j.1600-6143.2010.03408.x
79. Veer Manik, Abdulmassih Rasha, Como James, Min Zaw, Bhanot Nitin. **Adenoviral nephritis in a renal transplant recipient: Case report and literature review**. *Transpl Infect Dis* (2017) **19**. DOI: 10.1111/tid.12716
80. Zhang Yang, Ahmed Hiba, Haririan Abdolreza, Ugarte Richard, Papadimitriou John C., Drachenberg Cinthia B.. **Granulomatous inflammation in BK polyomavirus-associated nephropathy**. *Transpl Infect Dis* (2018) **20** e12939. DOI: 10.1111/tid.12939
81. Javaud Nicolas, Belenfant Xavier, Stirnemann Jérôme, Laederich Joëlle, Ziol Marianne, Callard Patrice, Ronco Pierre, Rondeau Eric, Fain Olivier. **Renal granulomatoses: a retrospective study of 40 cases and review of the literature**. *Medicine (Baltimore)* (2007) **86** 170. DOI: 10.1097/MD.0b013e3180699f55
82. Mignon F., Méry J. P., Mougenot B., Ronco P., Roland J., Morel-Maroger L.. **Granulomatous interstitial nephritis**. *Adv Nephrol Necker Hosp* (1984) **13** 219. PMID: 6433667
83. Naidu G. D., Ram R., Swarnalatha G., Uppin M., Prayaga A. K., Dakshinamurty K. V.. **Granulomatous interstitial nephritis: Our experience of 14 patients**. *Indian J Nephrol* (2013) **23** 415. DOI: 10.4103/0971-4065.120336
84. Viero R. M., Cavallo T.. **Granulomatous interstitial nephritis**. *Hum Pathol* (1995) **26** 1347. DOI: 10.1016/0046-8177(95)90300-3
|
---
title: 'A cost-effectiveness analysis of risk-based intervention for prevention of
cardiovascular diseases in IraPEN program: A modeling study'
authors:
- Amirparviz Jamshidi
- Rajabali Daroudi
- Eline Aas
- Davood Khalili
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9999709
doi: 10.3389/fpubh.2023.1075277
license: CC BY 4.0
---
# A cost-effectiveness analysis of risk-based intervention for prevention of cardiovascular diseases in IraPEN program: A modeling study
## Abstract
### Background
IraPEN, a program developed in Iran based on the World Health Organization (WHO) package of essential noncommunicable (PEN) disease interventions for primary healthcare, was launched in 2015. Preventive interventions for cardiovascular diseases (CVDs) are based on the level of risk calculated using the WHO CVD risk chart.
### Objective
The main objective of this study was to measure the potential cost-effectiveness (CE) of IraPEN preventive actions for CVD in comparison with the status quo.
### Methods
A CE analysis from a healthcare perspective was conducted. Markov models were employed for individuals with and without diabetes separately. Based on the WHO CVD risk chart, four index cohorts were constructed as low (<$10\%$), moderate ($10\%$−$19\%$), high ($20\%$−$29\%$), and very high risk (≥$30\%$). Life years (LY) gained and quality-adjusted life years (QALY) were used as the outcome measures.
### Results
The intervention yields an incremental cost-effectiveness ratio (ICER) of $804, $551, and –$44 per QALY for moderate, high, and very high CVD risk in groups without diabetes, respectively. These groups gained 0.69, 0.96, and 1.45 LY, respectively, from the intervention. The results demonstrated an ICER of $711, $630, –$42, and –$71 for low, moderate, high, and very high-risk groups with diabetes, respectively, while they gained 0.46, 1.2, 2.04, and 2.29 years from the intervention.
### Conclusion
The IraPEN program was highly cost-effective for all CVD risk groups in the individuals without diabetes except the low-risk group. The intervention was cost-effective for all patients with diabetes regardless of their CVD risk. The results demonstrated that the IraPEN program can likely provide substantial health benefits to Iranian individuals and cost savings to the national healthcare provider.
## Background
Cardiovascular diseases (CVDs) are the leading cause of death worldwide and people die from CVDs more than any other causes. CVDs are considered a development issue as almost $75\%$ of global CVD deaths occur in low- and middle-income countries. However, the majority of CVDs can be prevented by reducing the burden of risk factors [1].
Iran with an 80 million population in 2016, as an upper-middle-income developing country, has acquired many achievements in the public health sector during the past three decades. A well-balanced referral system within a broad PHC network, even in far stretches of villages, could provide access to healthcare for $95\%$ of the community and control communicable diseases efficiently. This accomplishment resulted in a life expectancy of more than 75 years for men and more than 77 years for women. At present, the transition to chronic and noncommunicable diseases (NCDs) including CVDs, cancer, and mental disorders is the main problem in the health system. Based on the last report by the World Health Organization (WHO), NCDs are estimated to account for almost $80\%$ of total deaths in Iran, while almost half of them ($43\%$) are caused by CVDs. The comparison of Iranians' CVD mortality rate with other countries shows that not only it is substantially higher than high-income countries but also it is much higher than the countries in the region [2].
During the last two decades, many positive actions (e.g., public education, opportunistic finding of diabetic and hypertensive cases in network system, etc.) have been carried out to control NCDs in the country. Despite major achievements, NCDs and their subsequent burden have increased in the country [3].
In 2010, the WHO launched the package of essential noncommunicable (PEN) disease intervention for primary care in low-resource settings to deliver an adequate quality of care and, consequently, reduce the burden of these diseases in developing countries. WHO PEN has effective tools to facilitate early diagnosis and management of CVD, chronic respiratory diseases, diabetes, and cancer to prevent their upcoming morbidities and premature mortalities (e.g., stroke, myocardial infarction, renal failure, blindness, amputations, etc.) [ 4].
In 2015, IraPEN, an adaptation of WHO PEN, was launched as a part of the national Healthcare Reform Plan in Iran. Providing universal healthcare coverage and access to NCD prevention and treatment for all were the main goals of this reform. The first phase of the IraPEN program had been piloted in four cities, and the results were promising. In 2018, the program was expanded to all provinces of Iran. It was expected that at this phase, the program would cover up to four million people, and then based on the results, it would be expanded nationwide [4]. Due to the IraPEN project size and its impact on the national healthcare budget, it is essential to make a detailed evaluation of these pilot enforcements to pave the way for IraPEN national implementation. Therefore, there is a need for an economic analysis that can estimate costs and effects as well as an incremental cost-effectiveness ratio (ICER) of the IraPEN program in comparison with the status quo (no prevention). Therefore, the main objective of this analysis is to measure the cost per life-year (LY) gained as well as the cost per quality-adjusted life years (QALY) gained by the IraPEN program.
## Methods
This study evaluates the potential cost-effectiveness (CE) of the IraPEN program in comparison to the status quo through a health economic evaluation and the outcomes expressed in terms of QALY and LY gained for each CVD risk group. The target group of this analysis is all Iranian people aged older than 40 years and the evaluated intervention is the same as the recommended intervention of WHO PEN which is included screening, monitoring, and medications.
## Model structure
Two separate Markov decision models were developed to compare the long-term costs and health benefits of the IraPEN program (primary CVD prevention) with the status quo (no prevention) in two distinct scenarios. In the base case scenario, individuals without diabetes were included, while patients with diabetes were included in the alternative scenario. Each Markov model has four health states with transitions between the states according to age, sex, and the CVD risk characteristics of participants (Figure 1). In contrast to the usual Markov models, which are structured based on cohorts with average profiles, we decided to categorize the individuals based on their CVD risks. As the intervention (treatment) varied according to CVD risk level, it is logical to model them separately. In this way, we can take into account their specific characteristics. Therefore, based on WHO/ISH CVD risk prediction charts for EMR B, four index cohorts were constructed [5]. These hypothetical cohorts were used as a representative for individuals with low, moderate, high, and very high CVD risk profiles. The CVD risk state represents the starting point for all people who are 40 years old. It was assumed that people in this state may either remain in the same health state, move to the stroke state, or CHD (coronary heart disease) state, or die. As long as they are event-free, these individuals can stay in a healthy state, but after the first event, they move to the CHD or stroke state and stay there until their death.
**Figure 1:** *The structure of the Markov model used for the IraPEN analysis. CVD risk index cohort state, healthy individuals with different CVD risk; Stroke state, alive individuals after the first stroke event; CHD state, alive individuals after the first CHD event; Death state, dead individuals.*
In WHO/ISH CVD risk prediction charts, the CVD risk is calculated based on individuals' age and risk factors such as blood pressure, lipid profile, diabetes, and smoking status and categorized into the following five groups: below $10\%$ (low-risk group), between 10 and $19\%$ (moderate-risk group), between 20 and $29\%$ (high-risk group), between 30 and $39\%$, and above $40\%$ (very high-risk group). As the individuals in the two latter groups are treated the same, in the IraPEN program, whoever has a CVD risk above $30\%$ is categorized as the very high-risk group.
Therefore, considering what was mentioned earlier, all the Iranians aged older than 40 years who did not have CHD or stroke events before were eligible for this program. According to the recent census [2016], $31.16\%$ of Iranians were older than 40 years [6]. By adding individuals aged older than 30 years with the aforementioned risk factors, we can conclude that this program is going to screen at least 25 million people yearly.
The healthcare perspective and a 40-year time horizon were adopted for this analysis. As the analysis is a comparison between IraPEN (intervention) and status quo (no intervention) which both have the same Markov structure and transition probabilities, it is not expected that half cycle correction (HCC) approach makes any difference in ICER results; therefore, HCC was not applied to this analysis [7].
The hypothetical cohorts were used as a representative for individuals with low, moderate, high, and very high CVD risk profiles (Table 1). Progressively, a proportion of the cohort can go to the CHD state, who are the survivors of the first CHD event, or to the stroke state who are the survivors of the first stroke event. Those CHD and stroke events that were fatal moved to the death state. *In* general, the people in these two states are at a higher risk of dying from CHD or stroke, but they may die from any other causes like the normal population. Table 2 summarizes the assumptions of this analysis.
## Data input
This analysis tried to use the *Iranian data* wherever available. In case of a lack of local data, the inputs were derived from the global literature. Therefore, all transition probabilities of the models were extracted from available Iranian data, while medications' effects and states' utilities were driven from the literature of Western countries (refer to Supplementary material). No individual data have been used for this analysis.
## Transition probabilities
The annual incidence rate for CHD [8] and stroke was calculated from the Framingham study equations [9]. As four-index cohorts had been defined with specific characteristics, there was a need to calculate the risk based on those profiles. Based on the literature, one out of four CHD events are fatal in the first year [10], while $60\%$ of them are pre-hospital deaths [11]. Therefore, it was assumed that of those who have a CHD event in the model, $25\%$ die in the first year. Approximately $60\%$ of these deaths were costless as they are pre-hospital deaths. Regarding first-year stroke mortality, the range varies in different resources and is reported from 22 to $34\%$ [12]. For this analysis, the rate was applied from the largest available cohort [13]. Almost $25\%$ of stroke events are fatal in the first year, while half of them occur during the first 28 days. Therefore, it is assumed that although at the end of the cycle, they move to the death state, $40\%$ of the cycle cost should be considered for them. The fatality rate for stroke and CHD survivors was derived from a study that had been done on the Iranian population [14]. The background mortality rate from all causes other than stroke and CHD is calculated by excluding the total death attributed to these two diseases from the Iranian life table1. The total mortality rate of these two events had been calculated in Tehran Lipid and Glucose Study (TLGS). At first, the annual rates were derived from the life table and then the CHD- and stroke-attributed deaths were excluded.
## Intervention effect
The IraPEN's preventive actions are expected to reduce cardiovascular events. The relative risks (RRs) of these preventive actions and the medications that are used in the program were obtained from meta-analyses or randomized clinical trials (RCTs). By multiplying or adding up the RRs of different medications, there is a risk of effect overestimation, and a correction was made by using the formula below wherever multiple interventions were involved: This equation has been developed based on a study that compared the effect of controlling the risk factors separately vs. controlling all of them simultaneously [15].
Based on the field interviews, it was clear which medications are used for each index cohort. Almost in all cases, angiotensin-converting enzyme (ACE) inhibitors are the first choice for hypertension treatment. Enalapril is the most prescribed one as monotherapy. Thiazides (diuretics) are the second choice followed by beta-blockers. In case the hypertension is not controlled by monotherapy instead of increasing the dose, the second drug is added. As recommended by guidelines, small doses of various classes of antihypertensive medications are more useful than a high dose of one [16]. *In* general, the combination of ACE inhibitors and thiazide is the most common one. This pattern is aligned with Joint National Committee (JNC8) guidelines. Statins are prescribed for hyperlipidemia treatment. Among statins, *Atorvastatin is* the choice as it is one of the most potent ones. For diabetes, *Metformin is* started and increased to the maximum dose (2 g) and then the second medication that is *Glibenclamide is* added. Due to its potential harm and insufficient evidence of its efficacy, Aspirin was not recommended for primary prevention by PEN protocols. Therefore, *Aspirin is* not used in IraPEN as well. Here are the list of medications and their daily dosages which are used in IraPEN: The unit price of each of these medications was derived from the Iranian Annual Pharma Statistics file. For the calculation of the intervention's effects, it is assumed that the adherence of individuals to the treatment is $100\%$. Table 3 lists the RRs of different interventions (medications) for CHD and stroke.
**Table 3**
| Unnamed: 0 | Unnamed: 1 | RR | (95% CI) |
| --- | --- | --- | --- |
| Angiotensin-converting enzyme inhibitor (17) | Angiotensin-converting enzyme inhibitor (17) | Angiotensin-converting enzyme inhibitor (17) | Angiotensin-converting enzyme inhibitor (17) |
| RR | CHD | 0.81 | (0.70–0.94) |
| | Stroke | 0.65 | (0.52–0.82) |
| Thiazide diuretics (18) | Thiazide diuretics (18) | Thiazide diuretics (18) | Thiazide diuretics (18) |
| RR | CHD | 0.84 | (0.75–0.95) |
| | Stroke | 0.63 | (0.57–0.71) |
| Beta-blockers (19) | Beta-blockers (19) | Beta-blockers (19) | Beta-blockers (19) |
| RR | CHD | 0.90 | (0.78–1.03) |
| | Stroke | 0.83 | (0.72–0.97) |
| Statin (20) | Statin (20) | Statin (20) | Statin (20) |
| RR | CHD | 0.86 | (0.82–0.90) |
| | Stroke | 0.90 | (0.85–0.95) |
| Metformin (21) | Metformin (21) | Metformin (21) | Metformin (21) |
| RR | CHD | 0.67 | (0.51–0.89) |
| | Stroke | 0.80 | (0.50–1.27) |
| Sulfonylureas | Sulfonylureas | Sulfonylureas | Sulfonylureas |
| RR | CHD | 0.85 | (0.74–0.97) |
| | Stroke | 0.91 | (0.73–1.13) |
| Lifestyle counseling (22) | Lifestyle counseling (22) | Lifestyle counseling (22) | Lifestyle counseling (22) |
| RR | CHD | 0.86 | (0.81–0.91) |
| | Stroke | 0.86 | (0.81–0.91) |
## Costs
A healthcare perspective was adopted; therefore, we only included costs associated with healthcare such as direct medical costs (Table 4). The costs considered in the model are the cost of IraPEN screening, the cost of IraPEN monitoring, the cost of CHD survivors, and the cost of stroke survivors. It is assumed that the cost of individuals who are event-free in the status quo is zero as long as undiagnosed or untreated. These two facts were considered for the status quo costs. Furthermore, it was assumed that the cost of dying was equal to zero. According to PEN protocols, the needed resources for each index cohort were identified. Then, the items were quantified based on discussions with the physicians and supervisors of the visited centers. The cost of index cohorts consists of two different types. First, variable costs are different for each group based on the characteristics of each. Second, fixed cost is the same for all and consists of staff training, administration, IT, promotional stuff, and leaflets. The unit price of each item was derived from the last report of the Ministry of health [23]. The report estimated all the costs related to IraPEN implementation except the medications. In addition, the reported costs were adjusted by the 2018 inflation rate and the cost of each cohort was calculated.
**Table 4**
| Unnamed: 0 | Low-risk | Moderate-risk | High-risk | Very high-risk |
| --- | --- | --- | --- | --- |
| Behvarz's visit (Screening) | 108173 | 324519 | 432692 | 432692 |
| Physician visit | – | 216346 | 288462 | 432692 |
| Lab data (included in screening) | | – | – | – |
| Nutrition consultation | 38400 | 38400 | 38400 | 38400 |
| Psychiatrist consultation | – | – | 38400 | 38400 |
| Anti-hypertensive medication | – | One agent | Two agents | Three agents |
| Statin Therapy | – | + | + | + |
| Fixed costs | 19231 | 19231 | 19231 | 19231 |
| Cost of each group in 2017 (IRR) | 165804 | 598496 | 817185 | 961416 |
| The inflation rate was applied to the costs (IRR) | 216540 | 781636 | 1067243 | 1255609 |
| Cost of each index cohort without medication | $5.16 | $18.61 | $25.41 | $29.90 |
| Cost of medication of each index cohort | $16 | $38.85 | $48.27 | $55.39 |
| Cost of each index cohort for the model | $21.63 | $57.46 | $73.68 | $85.29 |
The cost of CHD state and stroke state was derived from an Iranian CE that had estimated the cost of these two states [24]. These two costs contain all the related medical costs such as hospital admissions and procedures, monitoring, follow-ups, medications, and secondary prevention (Table 5). Based on experts' opinions, it is assumed that the cost of CHD after the first year would be a third and the cost of stroke state after the first year would be a quarter. In addition, it is assumed that the standard error of costs for the consecutive year is $10\%$ of the mean.
**Table 5**
| Unnamed: 0 | Mean | SE |
| --- | --- | --- |
| CHD cost for the first year | $519 | $51 |
| CHD cost for consecutive years | $173 | $17 |
| Stroke cost for the first year | $5,691 | $569 |
| Stroke cost for consecutive years | $1,422 | $142 |
## Utilities
As all people who enter the model are healthy individuals, the utility for the first health state is considered 1. For the death health state utility, it was adopted the standard approach by setting the utility to zero. CHD and stroke state utilities were derived from the published literature. In the models after the first event, patients move to these states and stay there until they die. Although they remain in the same states (CHD or stroke), their utilities are different over time. From a medical view, acute post-event utilities are (much)lower than chronic post-event utilities. Therefore, it was essential to use the data from the study which assessed the acute and chronic utilities with the same participants at an appropriate time. For this purpose, the utilities were derived from the study which assessed the utilities within the first year and consecutive years [13]. Table 6 shows the utilities used for the model. All the costs and effects were discounted at the rate of $3.5\%$.
**Table 6**
| Unnamed: 0 | Utility mean | SE |
| --- | --- | --- |
| Utility of CHD survivors—first year | 0.67 | 0.024 |
| Utility of stroke survivors—first year | 0.33 | 0.033 |
| Utility of CHD survivors—second year onwards | 0.82 | 0.012 |
| Utility of stoke survivors—second year onwards | 0.52 | 0.027 |
## Sensitivity analysis
To quantify the level of confidence in the models' results, a deterministic sensitivity analysis (DSA) and a probabilistic sensitivity analysis (PSA) were performed. In the DSA, the input parameters were varied to the maximum and minimum possible values. This range is usually defined by the confidence interval of parameters. Therefore, for the examined parameters, a range of $95\%$ confidence interval was specified and then based on this range (maximum and minimum input values), one value in the model was varied manually each time. For the patients' adherence, the range of $50\%$−$100\%$ was considered. The results (new ICERs) were collected and expressed with tornado diagrams. The tornado diagrams depict the impact on the ICER whenever one single parameter changed.
For the PSA, as the main assumption, it was considered the deterministic input values in the parameter sheet as the mean values. As the standard errors of the cost items were not available, it was considered to mean value times by 0.1. Based on logical constraints, the probabilistic distribution for each of the different sources of uncertainty was defined. A gamma distribution for all cost items and a beta distribution for the utilities were defined. The PSA was conducted by drawing a random number for each of the input distributions and each time, the ICER was calculated by Excel. By running a macro, its action is repeated 1,000 times.
## Results
For the interpretation of the results, 1 GDP per capita was assumed as the CE threshold, which is equal to $4,091 [25]. In Table 7, the results of the “CVD risk model without diabetes” are reported. The IraPEN intervention for individuals having moderate CVD risk yields an ICER of $804 per QALY, while for high-risk groups, this intervention provides an ICER of $551 per QALY. The results showed that this intervention would be cost-saving and improve health if the IraPEN program targets only individuals with a CVD risk higher than $30\%$. The model yields the ICER of –$44 per QALY for this group. Moreover, the model results showed that individuals with higher CVD risks gained higher LY out of the intervention. The moderate, high, and very high CVD risk groups gained 0.69, 0.96, and 1.45 LY, respectively, from the intervention. Table 8 shows the model results for the “CVD risk model with diabetes”.
Here, the results demonstrate that the intervention would be cost-saving while improving health if target individuals with CVD risk comprise higher than $20\%$. The intervention yields an ICER of –$42 per QALY for the high-risk group and an ICER of –$71 per QALY for the very high-risk group. Similar to the previous model, individuals with higher risks gain more LY from the intervention.
The one-way sensitivity analysis reveals that the patients' adherence, the treatment effectiveness, and the total cost of the IraPEN program have the most impact on the ICER. The influence of patients' adherence is more noticeable in the higher CVD risk groups, while the results are more sensitive to treatment effectiveness in the lower risk groups (Figures 2, 3).
**Figure 2:** *One-way sensitivity analysis of modeled cost–effectiveness of CVD risk-based prevention for very high-risk without diabetes group.* **Figure 3:** *One-way sensitivity analysis of modeled cost–effectiveness of CVD risk-based prevention for low risk with diabetes group.*
In the scenario analysis of $50\%$ adherence, an ICER of $1,451, $1,141, and $329 was achieved for moderate, high, and very high CVD risk in groups without diabetes, respectively. In this scenario, the intervention yields an ICER of $1,029, $1,022, $236, and $199 for low, moderate, high, and very high CVD risk groups with diabetes, respectively. It was found that the ICERs are lower than the threshold at both the upper and lower limits of all examined parameters.
The result of 1,000 PSAs illustrates that the intervention for all groups, except the low CVD risk group without diabetes, is cost-effective while being cost-saving for at least half of high-risk and very high-risk patients (Figures 4, 5). By adopting 1 GDP per capita of Iran as the willingness to pay per quality-adjusted life-year (WTP/QALY) gained, $100\%$ of all the runs were cost-effective in these groups.
**Figure 4:** *Probabilistic sensitivity analysis of the Markov model without diabetes. Yellow color: represents the moderate-risk group (individuals with CVD risk between 10 and 19%). Orange color: represents the high-risk group (individuals with CVD risk between 20 and 29%). Red color: represents the very high-risk group (individuals with CVD risk more than 30%).* **Figure 5:** *Probabilistic sensitivity analysis of the Markov model with diabetes. Green color: represents the low-risk group (individuals with CVD risk <10%). Yellow color: represents the moderate-risk group (individuals with CVD risk between 10 and 19%). Orange color: represents the high-risk group (individuals with CVD risk between 20 and 29%). Red color: represents the very high-risk group (individuals with CVD risk more than 30%).*
Both models captured that men would be benefited more than women in terms of LY gained. In addition, the intervention generates a lower ICER for men than women in individuals with identical characteristics. For example, in the low CVD risk group with diabetes, the interventions yield an ICER of $239 per QALY for men, while it is $711 per QALY for women. Regarding the LY in this group, in equal circumstances, men saved 0.58 of a year and women saved 0.46 of a year. In higher risk groups, this difference is more prominent. For example, for very high-risk groups, regardless of their diabetes status, the intervention for men is cost-saving, while for women it is not (Tables 9, 10).
## Discussion
This analysis aimed to measure the potential CE of IraPEN preventive actions for CVD in comparison with the status quo. Our results illustrated that this intervention is not cost-effective for the low CVD risk group, whereas the other groups under study proved to be highly cost-effective. The reason why this group was not cost-effective could be justified by the fact that the low CVD risk group has lower CVD risk factors, which is why such individuals are just screened without being intervened.
In this study, we adopted the CE threshold, i.e., 1–3 GDP per capita, as proposed by WHO [26]. The study of the literature shows that by using this threshold, almost all interventions seem to be cost-effective [27]. It means that by adopting this CE threshold, there is a risk that the budgets are spent on interventions that should not and vice versa. The threshold which is recommended by WHO has received some criticism as it is believed that it does not reflect the true “opportunity cost.” *This is* more critical in low- to middle-income countries, because while they have a higher demand for health, in comparison with high-income countries, fewer resources are available to them. In 2016, Woods et al. [ 28] calculated the CE threshold based on the empirical estimates of opportunity cost, the relationship between a country's GDP per capita, and the value of statistical life. For that reason, they estimated the threshold for different countries with different levels of income. Based on their estimation, Woods et al. suggested the CE threshold to be about $50\%$ of GDP per capita. As appointing the precise CE threshold level is beyond the scope of our analysis, we interpret and discuss the results with the lowest recommended ICER and leave the decision to policymakers to choose an intervention that best fits their budget. The latest reported GDP per capita of *Iran is* $4,091. By comparing the results with this threshold, it is shown that all of them, except the low CVD risk group, are highly cost-effective, whereas if the program only targets the people in higher risk groups, it is both cost-saving and improves their health.
Based on the results, the intervention for the low-risk group was not cost-effective as the ICER was undefined. This could be explained by the fact that in this group, just screening is done without offering any intervention. So here, there is a cost for screening without any tangible effect assigned. Since this group does not receive any treatment annually, it is, by all means, sensible that no effects are observed. While it seems to be justifiable, we have every reason to believe that this is the only way to find the groups with a higher CVD risk. Another point that should be mentioned is that the individuals enter and are screened in our model at the age of 40 years. At this age, the proportion of individuals with low CVD risk is significantly higher than that of those with higher CVD risk. The older a person becomes, the more the probability of being in the higher-risk groups will be.
However, the fact that the intervention for the aforementioned group is not cost-effective needs to be approached more comprehensively and conservatively. Screening of this group is the first step and essential for all individuals in the other groups that cannot be disregarded. This means that the other groups can benefit from this, a fact that has not been considered in the analysis of the CE of this group in this study.
From another perspective, a closer observation reveals that screening has a wide range of benefits. For instance, according to the latest national data [29], the prevalence of people with diabetes in *Iran is* $11.4\%$, a quarter of whom are undiagnosed [30]. This means that at the moment, there are almost 1.5 Iranian people with undiagnosed diabetes. These undiagnosed individuals are discovered only when their complications have started to appear in them. Such complications as retinopathy, nephropathy, and neuropathy are very costly and can impose burdensome pressures on the healthcare system of the country. Other typical examples of this kind are blood hypertension and hyperlipidemia.
It is predicted that huge monetary resources should be allocated to the overall screening of all the individuals, which may not be conveniently supplied. Therefore, appropriate measures could be taken by the authorities to have the costs tailored. This can help manage the financial resources and distribute them as optimally as possible. If the available financial resources do not allow us to screen everyone, it is possible to screen all high-risk people. Although this may not sound optimal, still it has a lot of benefits to offer. In other words, when considered at a higher scale, it can be realized that since the proportion of high CVD risk group individuals outweighs those with lower risks, this may lead to much more favorable results.
It is applicable to have a paper pre-screening. The idea is that alternatively, paper questionnaires can be distributed among both households and health center visitors. These questionnaires aim to detect individuals with higher CVD risks. Typical examples might be those who are obese or have a positive history of CVDs in their intimidate relatives. Once identified, such participants can be invited to health centers to get screened. In this way, we can narrow the target population and screen those who are at higher risk levels.
The study of the literature suggested some good examples of this kind of practice. For example, Chamnan et al. [ 31], who conducted a modeling study using the data from a prospective cohort study (EPIC-Norfolk), concluded that adopting a stepwise screening approach can prevent the same number of CVD events annually. All the participants of the EPIC-Norfolk study had completed the questionnaire about their lifestyle and drug use and family history of diseases between 1993 and 1997. This population had been observed and followed for 10 years and all CVD events had been recorded. By adopting the Cambridge risk score (a British risk scoring tool) and based on the results of completed questionnaires, the Chamnan group ranked the population CVD risk. Then, they defined and modeled seven different stepwise screening strategies. By comparing the results of a 10-year follow-up of the population with their model, they found that inviting the individuals with a Cambridge risk score >60 might have had the same results as screening the population. According to their report, this strategy could have enabled them to have the same results about the whole population by screening only $60\%$ of the population. Similar results were observed by Móczár and Rurik [32] who concluded that performing screening for a selected target group is most likely to be more cost-effective than screening the whole population. It is essential to consider that although this approach is practical and likely more cost-effective in budget-strained situations, it has some serious ethical issues regarding equity because in this approach, a smoker would get screened and a non-smoker would not. The same is true for a person who has an unhealthy diet or lifestyle. Moreover, while it is conveniently applicable to rural areas, it is not easy to do in urban areas.
Since our literature search revealed no similar studies either in the WHO East Mediterranean region (EMR) or in the Middle East region, inevitably, we compared the results of our analysis with the European studies. Our findings in this analysis, in terms of CE and the trend between the different CVD risks, are similar to the results of Schuetz et al. 's [33] study. The researchers of this study estimated the CE of several different preventive strategies compared with a control scenario in six European countries. By using country-specific data from France, Germany, Denmark, Italy, Poland, and the United Kingdom, they generated six simulated populations of people aged 40–75 years eligible for preventive actions in those countries. Their model showed that the cost per QALY of offering these preventive services to the people in the study cohort ranged from €14,903 for France to €115 for Germany, while it was cost-saving for Poland. Their results showed that the health checks for detecting and managing CVDs at the early stages not only are highly cost-effective but also cost-saving in some scenarios. For example, their analysis for the UK showed that during the 30-year follow-up, the cost per QALY would be €2,426. This ICER in comparison with the UK threshold, which is between 20 and 30 thousand pounds, is highly cost-effective. Moreover, their results demonstrated that the program would be cost-saving if it targets only the top quartile of CVD risk groups. Furthermore, offering prevention checks after the pre-screening of individuals based on some characteristics such as higher age or obesity would pull the results in the direction of more favorable ICER [33].
Finally, while this screening program can have substantial benefits for individuals with CVD risk, it is essential to consider the potential disutility of lifelong preventive treatments. The search of the literature indicates that the disutility of medications' adverse events [34] and the disutility arising from taking daily medication [35] can play a key role in the decision that leads to non-adherence. It is vital to take into account that some individuals need to be on preventive treatments from their 40s for around 30–40 years. Although the sensitivity analysis demonstrated that this program can likely be cost-effective even with $50\%$ of adherence, it is crucial to enhance treatment compliance through patient education and take effective strategies to increase the engagement of target groups.
## Strengths and limitations
Based on our literature search, this analysis is the first CVD risk-based CE to be conducted in Iran, in the Middle East, and the WHO EMR so far. Furthermore, it is one of the few studies which model the individual with and without diabetes separately.
Similar to every CE analysis, this study has several limitations which were mainly caused by a large number of input parameters used in the model. Although the model was designed for the Iranian population, some input parameters were derived from various studies performed in different countries other than Iran directly because of the unavailability or a lack of Iranian data. Furthermore, it was assumed that the intervention effect is equal for all subgroups regardless of their initial CVD risk. Such an assumption can probably generate an underestimation of the intervention effect for high-risk groups and produce an overestimation of the treatment effect for lower CVD risk groups.
In this analysis, the second event of CVDs was not accounted for due to a lack of data. Based on the literature, almost $50\%$ of patients would experience the second or third event during their life once they have had the first event [36]. The second event may not necessarily be the same as the first one. For example, a patient who has had a CHD event could have the same event again or can even experience a stroke event. This cannot affect the result of our analysis unfavorably. As the IraPEN intervention causes the first CVD event to decrease or delay, it is logical to assume that the second event in the IraPEN group is lower than the status quo. Therefore, we could assume that adding the second CVD event to our model would be in favor of our ICER.
It could be better if we could adopt a lifetime horizon for this analysis. However, as the Iranian statistical data were just available for people aged below 80 years, inevitably 40 years/cycles were employed. The results of Kim et al. 's [37] systematic review, which was done on more than 750 CE analyses, showed that the usage of a lifetime horizon captures all consequences and health benefits most of the time and yields more favorable ICER. Therefore, it is logical to assume that by increasing the time horizon from 40 years to a lifetime, the ICER would be more favorable.
Finally, it should be expected that the effectiveness of the intervention would be lower in the real world than in the model. It could be explained by the fact that the intervention effects, which were used in this study, all had been derived/extracted from a controlled trial setting.
## Conclusion
In Iran, CVDs are the leading cause of mortality. Therefore, planning and implementing preventive actions are highly demanded. Our analysis results demonstrated that the IraPEN program implementation is highly cost-effective for all the CVD risk groups, except the low risk without diabetes group, whereas if the program only targets the people in higher risk groups, it is both cost-saving and improves their health.
## 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.
## Author contributions
AJ contributed to the design, carried out the analysis, and wrote the first draft. RD contributed to the analysis and reviewed the manuscript critically. EA contributed to the design and supervised the process of analysis reviewed the manuscript critically. DK designed the whole project and contributed to the analysis and drafting of the manuscript. All authors read the final manuscript and approved it.
## 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.2023.1075277/full#supplementary-material
## References
1. Chrysant S. **A new paradigm in the treatment of the cardiovascular disease continuum: focus on prevention**. *Hippokratia.* (2011.0) **15** 7-11. PMID: 21607028
2. Azizi F, Hadaegh F, Hosseinpanah F, Mirmiran P, Amouzegar A, Abdi H. **Metabolic health in the Middle East and north Africa**. *Lancet Diabetes Endocrinol.* (2019.0) **7** 866-79. DOI: 10.1016/S2213-8587(19)30179-2
3. Azadnajafabad S, Mohammadi E, Aminorroaya A, Fattahi N, Rezaei S, Haghshenas R. **Non-communicable diseases' risk factors in Iran; a review of the present status and action plans**. *J Diabetes Metab Disord* (2021.0) **22** 1-9. DOI: 10.1007/s40200-020-00709-8
4. Hadavand Siri F, Khalili D, Hashemi Nazari SS, Ostovar A, Mahdavi A. **Adherence to Iran's package of essential noncommunicable diseases (IraPEN) program for regular follow-up to reduce the risk of cardiovascular disease in healthcare centers**. *Iran J Endocrinol Metab.* (2020.0) **22** 116-26
5. 5.World Health Organization. WHO/ISH Cardiovascular Risk Prediction Charts. Geneva: World Health Organization (2011). Available online at: https://www.who.int/cardiovascular_diseases/guidelines/Chart_predictions/en/ (accessed February 20, 2019).. *WHO/ISH Cardiovascular Risk Prediction Charts* (2011.0)
6. 6.. Available online at: https://www.amar.org.ir/ (accessed June 10, 2019) .
7. Naimark DMJ, Kabboul NN, Krahn MD. **The half-cycle correction revisited**. *Med Decis Making.* (2013.0) **33** 961-70. DOI: 10.1177/0272989X13501558
8. Andreson, K, Wilson, P, Odell, P, Kannel, W. **An updated coronary risk profile a statement for health professionals**. *Circulation* (1991.0) **83** 356-62. DOI: 10.1161/01.CIR.83.1.356
9. Wolf P, D'agostino R, Belanger A, Kannel W. **Probability of stroke: a risk profile from the Framingham Study**. *Stroke* (1991.0) **22** 312-8. DOI: 10.1161/01.STR.22.3.312
10. Talaei M, Sarrafzadegan N, Sadeghi M, Oveisgharan S, Marshall T, Thomas GN. **Incidence of cardiovascular diseases in an Iranian population: the Isfahan Cohort Study**. *Arch Iran Med* (2013.0) **16** 138-44. PMID: 23432164
11. Grey C, Jackson R, Schmidt M, Ezzati M, Asaria P, Exeter DJ. **One in four major ischaemic heart disease events are fatal and 60% are pre-hospital deaths: a national data-linkage study (ANZACS-QI 8)**. *Eur Heart J* (2015.0) **38** 172-80. DOI: 10.1093/eurheartj/ehv524
12. Shoeibi A, Salehi M, Thrift AG, Kapral MK, Farzadfard MT, Azarpazhooh A. **One-year case fatality rate following stroke in the Mashhad Stroke Incidence Study: a population-based study of stroke in Iran**. *Int J Stroke* (2015.0) **10** 96-102. DOI: 10.1111/ijs.12611
13. Matza LS, Stewart KD, Gandra SR, Delio PR, Fenster BE, Davies EW. **Acute and chronic impact of cardiovascular events on health state utilities**. *BMC Health Serv Res* (2015.0) **15** 173. DOI: 10.1186/s12913-015-0772-9
14. Rabani SH, Sardarinia M, Akbarpour S, Azizi F, Khalili D, Hadaegh F. **12-year trends in cardiovascular risk factors (2002-2005 through 2011-2014) in patients with cardiovascular diseases: Tehran lipid and glucose study**. *PLoS ONE.* (2018.0) **13** e0195543. DOI: 10.1371/journal.pone.0195543
15. Dagenais GR, Pogue J, Fox K, Simoons ML, Yusuf S. **Angiotensin-converting-enzyme inhibitors in stable vascular disease without left ventricular systolic dysfunction or heart failure: a combined analysis of three trials**. *Lancet.* (2006.0) **368** 581-8. DOI: 10.1016/S0140-6736(06)69201-5
16. Xavier D, Noby M, Pradeep J, Prem P. **Letter to the editor. Pattern of drug use in hypertension in a tertiary hospital; a cross sectional study in the inpatients ward**. *Indian J Pharmacol.* (2001.0) **33** 456-7
17. Wright J, Musini V. **First-line drugs for hypertension**. *São Paulo Med J.* (2010.0) **128** 47. DOI: 10.1590/S1516-31802010000100011
18. Wright JM, Musini VM, Gill R. **First-line drugs for hypertension**. *Cochrane Database Syst Rev* (2018.0) **4** CD001841. DOI: 10.1002/14651858.CD001841.pub3
19. Law MR, Morris JK, Wald NJ. **Use of blood pressure lowering drugs in the prevention of cardiovascular disease: meta-analysis of 147 randomised trials in the context of expectations from prospective epidemiological studies**. *Br Med J.* (2009.0) **338** b1665. DOI: 10.1136/bmj.b1665
20. Brugts J, Yetgin T, Hoeks S, Gotto A, Shepherd J, Westendorp R. **The benefits of statins in people without established cardiovascular disease but with cardiovascular risk factors: meta-analysis of randomised controlled trials**. *BMJ.* (2009.0) **338** b2376. DOI: 10.1136/bmj.b2376
21. Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HAW. **10-year follow-up of intensive glucose control in type 2 diabetes**. *N Engl J Med.* (2008.0) **359** 1577-89. DOI: 10.1056/NEJMoa0806470
22. Lanier J, Bury D, Richardson S. **Diet and physical activity for cardiovascular disease prevention**. *Am Fam Physician* (2016.0) **93** 919-24. PMID: 27281836
23. 23.
Available online at: http://behdasht.gov.ir/ (accessed March 29, 2019).
24. Javanbakht M, Jamshidi AR, Baradaran HR. **Estimation and prediction of avoidable health care costs of cardiovascular diseases and type 2 diabetes through adequate dairy food consumption: a systematic review and micro simulation modeling study**. *Arch Iran Med* (2018.0) **21** 213-22. PMID: 29738265
25. 25.GDP per Capita (current US$) - Iran Islamic rep. Data. Available online at: https://data.worldbank.org/indicator/NY.GDP.PCAP.CD?locations=IR (accessed January 21, 2023).. *GDP per Capita (current US$) - Iran Islamic rep. Data*
26. Marseille E, Larson B, Kazi DS, Kahn JG, Rosen S. **Thresholds for the cost–effectiveness of interventions: alternative approaches**. *Bull World Health Organ.* (2014.0) **93** 118-24. DOI: 10.2471/BLT.14.138206
27. Revill P, Walker SM, Madan J, Ciaranello A, Mwase T, Gibb DM et al. *Using Cost-effectiveness Thresholds to Determine Value for Money in Low-and Middle-income Country Healthcare Systems: Are Current International Norms fit For Purpose?* (2014.0)
28. Woods B, Revill P, Sculpher M, Claxton K. **Country-Level cost-effectiveness thresholds: initial estimates and the need for further research**. *Value Health.* (2016.0) **19** 929-35. DOI: 10.1016/j.jval.2016.02.017
29. Esteghamati A, Etemad K, Koohpayehzadeh J, Abbasi M, Meysamie A, Noshad S. **Trends in the prevalence of diabetes and impaired fasting glucose in association with obesity in Iran: 2005–2011**. *Diabetes Res Clin Pract.* (2014.0) **103** 319-27. DOI: 10.1016/j.diabres.2013.12.034
30. Noshad S, Afarideh M, Heidari B, Mechanick JI, Esteghamati A. **Diabetes care in Iran: where we stand and where we are headed**. *Ann Glob Health.* (2016.0) **81** 839. DOI: 10.1016/j.aogh.2015.10.003
31. Chamnan P, Simmons RK, Khaw KT, Wareham NJ, Griffin SJ. **Estimating the population impact of screening strategies for identifying and treating people at high risk of cardiovascular disease: modelling study**. *BMJ* (2010.0) **340** c1693. DOI: 10.1136/bmj.c1693
32. Móczár C, Rurik I. **Comparison of cardiovascular risk screening methods and mortality data among Hungarian primary care population: preliminary results of the first government-financed managed care program**. *Slovenian J Public Health.* (2015.0) **54** 154-60. DOI: 10.1515/sjph-2015-0022
33. Schuetz CA, Alperin P, Guda S, Herick AV, Cariou B, Eddy D. **A standardized vascular disease health check in Europe: a cost-effectiveness analysis**. *PLoS ONE* (2013.0) **8** e0066454. DOI: 10.1371/journal.pone.0066454
34. Epstein D, García-Mochón L, Kaptoge S, Thompson SG. **Modeling the costs and long-term health benefits of screening the general population for risks of cardiovascular disease: a review of methods used in the literature**. *Eur J Health Econ.* (2015.0) **17** 1041-53. DOI: 10.1007/s10198-015-0753-2
35. Cutler RL. **Fernandez-Llimos F, Frommer M, Benrimoj C, Garcia-Cardenas V. Economic impact of medication non-adherence by disease groups: a systematic review**. *BMJ Open* (2018.0) **8** e016982. DOI: 10.1136/bmjopen-2017-016982
36. Bansilal S, Castellano JM, Fuster V. **Global burden of CVD: focus on secondary prevention of cardiovascular disease**. *Int J Cardiol* (2015.0) **1** S1-7. DOI: 10.1016/S0167-5273(15)31026-3
37. Kim DD, Wilkinson CL, Pope EF, Chambers JD, Cohen JT, Neumann PJ. **The influence of time horizon on results of cost-effectiveness analyses**. *Expert Rev Pharmacoecon Outcomes Res.* (2017.0) **17** 615-23. DOI: 10.1080/14737167.2017.1331432
|
---
title: Liver fibrosis-4 score predicts outcome of patients with ischemic stroke undergoing
intravenous thrombolysis
authors:
- Davide Norata
- Simona Lattanzi
- Serena Broggi
- Chiara Rocchi
- Marco Bartolini
- Mauro Silvestrini
journal: Frontiers in Neurology
year: 2023
pmcid: PMC9999710
doi: 10.3389/fneur.2023.1103063
license: CC BY 4.0
---
# Liver fibrosis-4 score predicts outcome of patients with ischemic stroke undergoing intravenous thrombolysis
## Abstract
Some evidence suggests a possible influence of liver disease on stroke prognosis. We investigated the association between fibrosis-4 (FIB-4) score, a marker of liver disease, and the 3-month outcome in patients with ischemic stroke undergoing intravenous thrombolysis. We also evaluated the rate of symptomatic intracranial hemorrhage after thrombolysis. In this prospective cohort study, we enrolled consecutive patients with ischemic stroke treated with thrombolysis who had a 3-month follow-up. The FIB-4 score was calculated and the validated cut-off values were used to indicate high/low risk of advanced liver fibrosis. The primary outcome was 3-month poor prognosis estimated as a modified Rankin scale score ≥3. Of the 264 included patients, 131 ($49.62\%$) had a 3-month mRS ≥3, with a significantly higher FIB-4 score, compared to those with a mRS <3 score (adjp <0.001). When adjusted for possible confounders by multivariate logistic regression, FIB-4 score remained a significant predictor of poor outcome (OR 1.894, $$p \leq 0.011$$), along with history of atrial fibrillation (OR 3.488, $$p \leq 0.017$$), admission NIHSS score (OR 1.305, $p \leq 0.001$), and low values of hemoglobin (OR 0.730, $p \leq 0.001$). Mechanical thrombectomy had a favorable effect on patients' outcome (OR 0.201, $$p \leq 0.005$$). The risk of poor 3-month outcome was significantly higher among the 32 patients ($12.1\%$) with high risk of severe fibrosis ($$p \leq 0.007$$). FIB-4 score values were also related to symptomatic intracranial hemorrhage ($$p \leq 0.004$$), specifically among patients with high probability of advanced hepatic fibrosis ($$p \leq 0.037$$). FIB-4 score can be considered as a promising independent predictor of poor prognosis in patients with acute ischemic stroke undergoing intravenous thrombolysis.
## 1.1. Background
According to the World Health Organization, 15 million people worldwide suffer from stroke each year. With a mortality rate of approximately one-third, it is the second most common cause of death and a leading cause of disability [1]. Ischemic stroke is the most common type of stroke, accounting for approximately $80\%$ of all acute strokes [2]. Treatment approaches have been primarily directed at preserving neurons in the ischemic territory. The internationally approved treatments, recombinant tissue plasminogen activator (rt-Pa) and endovascular intervention, aim at rapid arterial recanalization to restore oxygen and nutrient supply to the affected area [3]. Early recanalization after stroke is associated with a greater likelihood of favorable outcome [4, 5].
Liver fibrosis, the histologic precursor of cirrhosis, is a chronic disease [6], often preceded and promoted by an inflammatory process in combination with the accumulation of extracellular matrix in the liver [7]. Several biomarkers have been proposed for the assessment of liver fibrosis. Among them, the fibrosis index (FIB)-4 has shown the best diagnostic accuracy for advanced hepatic fibrosis, as demonstrated by ultrasonographic studies in nonalcoholic fatty liver disease (NAFLD) [8, 9], the most common cause of liver dysfunction in Western countries [10]. In recent studies, liver disease has been shown to be a strong predictor of both in-hospital and long-term mortality in stroke patients [11, 12]. Moreover, it is independently associated with an increased risk of hemorrhagic complications [13], the most threated complication of intravenous thrombolysis, leading to poor outcome and increased risk of mortality [14]. It is not yet clear whether these findings can also be applied to subclinical liver disease, which may not be uncommon in patients with stroke [15]. In a recently published study, Fandler-Höfler et al. [ 16] showed that stroke patients with higher FIB-4 score values had worse clinical outcomes 3 months after mechanical thrombectomy but they didn't find any increased risk of postoperative parenchymal hematoma, hemorrhagic infarction and symptomatic intracerebral hemorrhage.
## 1.2. Objectives
The aim of the present study was to investigate the association of FIB-4 score with 3-month neurological outcome and symptomatic intracranial hemorrhage in patients with acute ischemic stroke treated with IV rt-Pa.
## 2.1. Study design, setting, and participants
We retrospectively identified consecutive patients admitted to the Stroke Unit of the University Hospital of Ancona, Italy, from January 2017 to April 2021 for acute ischemic stroke treated with IV thrombolysis. Each patient underwent routine blood sampling at admission (within 24 h of admission). Supplementary Table 1 provides an overview of the eligibility criteria.
The study was approved by the ethics committee of the Marche Polytechnic University (ID $\frac{57}{2020}$) and conducted according to the Declaration of Helsinki. Informed consent was obtained from all subjects involved in the study or their representatives.
## 2.2. Variables
Demographics, medical history, National Institutes of Health Stroke Scale (NIHSS) scores [17], and admission blood pressure were documented at baseline. Laboratory tests [including serum levels of creatinine, glucose levels, hemoglobin (Hb), platelet count (PLT), absolute neutrophil count (ANC), absolute lymphocyte count (ALC), absolute monocyte count (AMC), alanine aminotransferase (ALT), aspartate aminotransferase (AST) total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides, γ-glutamyltransferase (γGT), and creatine phosphokinase (CPK)] were determined by admission blood tests.
To quantify the extent of liver fibrosis, we used the noninvasive liver fibrosis score (FIB-4) for each patient at the time of admission.
The FIB-4 score was computed for every patient as follows: As validated in previous clinical trials, prediction of advanced liver fibrosis was indicated using a cut-off value ≥2.67, whereas a score value <1.30 was used to exclude severe liver fibrosis with high probability [18, 19].
## 2.3. Outcome measures
The primary outcome measure was functional status at 3 months, evaluated in the hospital's outpatient setting. Because of its ease of use and interpretability, the modified Rankin Scale (mRS) is a widely applied clinical measure of global disability. In particular, it is used to assess recovery from stroke and as a primary end point in randomized clinical trials of stroke treatments. In our study, poor outcome was defined as the occurrence of death or major disability (mRS≥3) [20].
We also considered symptomatic intracranial hemorrhage (sICH) as a secondary outcome. We defined this hemorrhagic complication usually linked to rt-Pa, through the European Cooperative Acute Stroke Study (ECASS) III criteria, as follows [21]. [ 1] Clinical deterioration: an increase of ≥4 points in NIHSS score or that led to death. [ 2] Radiographic features: any intracranial hemorrhage on CT/MRI performed at 22–36 h after stroke onset.
## 2.4. Biases and study size
We conducted this study on consecutive patients to avoid any selection bias. In order to address information bias, two aspects should be considered: the number of lost to follow-up was acceptable (Figure 1); the admission FIB-4 score was calculated only after the 3-month assessment, so the experimenter did not know the score value when assessing the 3-month mRS (primary outcome measure). Based on previous RCTs on Alteplase effectiveness [22], the minimum number of samples required to achieve a $95\%$ confidence level with a marginal error of 0.05 was 241.
**Figure 1:** *Patient selection flow diagram. EVI, endovascular treatment; IVT rt-PA, intravenous thrombolysis with recombinant tissue plasminogen activator; mRS; modified Rankin Scale.*
## 2.5. Statistical methods
We used standard statistical methods for descriptive statistics. Categorical variables were presented as frequencies and continuous variables as mean (standard deviation, SD) or median (interquartile range, IQR), when appropriate. Normality was assessed through the Shapiro–Wilk test. Depending on the normality of the distribution, comparisons were made by Student's t-test or Mann–Whitney test for continuous variables, and by Pearson χ2 test for categorical variables. The multivariate logistic regression was used to identify whether the FIB-4 score could be an independent predictor of poor 3-month outcome, and to establish the real prognostic value of demographic, clinical, and laboratory variables that reached statistical significance in the univariate analysis. To prevent biases, we did not include the variables already used for calculating the FIB-4 score in the logistic regression. An equivalent analysis was carried out for the secondary outcome, and is available in the Supplementary material. A two-tailed p-value of <0.05 was considered statistically significant for all tests. False discovery rate (FDR) correction was applied to deal with the multiple testing problem (results are expressed as adjusted p or adjp-values). Analysis was performed using JASP Team [2020]. JASP (version 0.14.1).
## 3.1. Participants and descriptive data
Of the 306 patients who suffered ischemic stroke treated with IV rt-Pa, 42 were excluded (Figure 1). Of the 264 enrolled patients, 131 ($49.62\%$) had a modified Rankin Scale score of ≥3 after 3 months and 35 ($13.3\%$) experienced a symptomatic intracranial hemorrhage (sICH; Table 1).
**Table 1**
| Baseline characteristics | Full cohort n = 264 | mRS < 3 n = 133 | mRS ≥3 n = 131 | adjp-value |
| --- | --- | --- | --- | --- |
| Demographics | Demographics | Demographics | Demographics | Demographics |
| Female sex | 123 (46.6%) | 53 | 70 | 0.074a |
| Age, years | 69.3 (13.8) | 65.9 (14.0) | 72.7 (12.7) | < 0.001b* |
| Clinical history | Clinical history | Clinical history | Clinical history | Clinical history |
| Hypertension | 157 (63.3%) | 73 | 84 | 0.080a |
| Diabetes mellitus | 35 (14.6%) | 18 | 17 | 0.963a |
| Actual smoking | 59 (24.9%) | 34 | 25 | 0.575a |
| Hypercholesterolemia | 90 (36.6%) | 51 | 39 | 0.337a |
| Atrial fibrillation | 45 (19.2%) | 11 | 34 | < 0.001a* |
| Ischemic heart disease | 31 (13.3%) | 16 | 15 | 0.986a |
| Prior stroke | 32 (13.5%) | 18 | 14 | 0.116a |
| Blood test variables | Blood test variables | Blood test variables | Blood test variables | Blood test variables |
| FIB-4 score | 1.284 (0.989) | 1.112 (0.734) | 1.436 (1.186) | < 0.001c* |
| FIB-4 score ≥2.67 | 32 (12.1%) | 9 | 23 | 0.021a* |
| FIB-4 score 1.31–2.66 | 95 (36.0%) | 42 | 53 | 0.244a |
| FIB-4 score < 1.30 | 137 (51.9%) | 82 | 55 | 0.004a* |
| Total cholesterol, mg/dl | 176.1 (40.34) | 178.0 (39.63) | 174.2 (41.14) | 0.653b |
| HDL cholesterol, mg/dl | 51.76 (14.67) | 53.21 (14.81) | 50.24 (14.44) | 0.233b |
| LDL cholesterol, mg/dl | 103.47 (32.64) | 105.24 (34.04) | 101.59 (31.13) | 0.575b |
| Triglycerides, mg/dl | 96.00 (63.00) | 96.00 (62.00) | 95.00 (59.25) | 0.760c |
| Creatinine, mg/dl | 0.860 (0.330) | 0.850 (0.265) | 0.880 (0.425) | 0.575c |
| Glucose, mg/dl | 108.00 (45.00) | 107.00 (43.25) | 109.00 (44.00) | 0.080c |
| Platelets, × 109/L | 207.00 (80.50) | 207.00 (80.00) | 207.00 (79.00) | 0.768c |
| Hemoglobin, g/dl | 13.05 (2.425) | 13.40 (2.10) | 12.60 (2.65) | < 0.001c* |
| ANC, × 109/L | 7.304 (3.338) | 6.616 (3.228) | 8.047 (3.312) | 0.007b* |
| ALC, × 109/L | 1.670 (1.828) | 1.722 (0.737) | 1.614 (2.527) | 0.768b |
| AMC, × 109/L | 0.697 (0.407) | 0.642 (0.255) | 0.757 (0.519) | 0.090b |
| AST, U/L | 17.00 (11.00) | 16.00 (8.00) | 19.00 (13.50) | 0.021c* |
| ALT, U/L | 22.00 (12.00) | 22.00 (10.00) | 22.00 (12.00) | 0.768c |
| γGT, U/L | 26.00 (24.00) | 26.00 (23.00) | 27.00 (23.00) | 0.714c |
| CPK, U/L | 99.00 (86.50) | 94.00 (84.00) | 108.50 (90.00) | 0.389c |
| In-hospital variables | In-hospital variables | In-hospital variables | In-hospital variables | In-hospital variables |
| Systolic blood pressure | 139.0 (20.93) | 139.0 (21.20) | 139.0 (20.74) | 0.986b |
| Diastolic blood pressure | 76.4 (10.98) | 76.4 (11.05) | 76.4 (10.96) | 0.986b |
| Admission NIHSS score | 12.49 (6.16) | 9.61 (6.17) | 15.54 (4.49) | < 0.001b* |
| EVT | 146 (55.3%) | 60 | 86 | < 0.001a* |
| sICH | 35 (13.3%) | 28 | 35 | < 0.001a* |
The mean FIB-4 score was significantly higher among patients with a poor prognosis compared with the other group (1,436 vs. 1,112, Student t-test −3.303, adjp < 0.001).
## 3.2. Main results
Demographic, clinical, and laboratory characteristics of patients are presented in Table 1. In univariate analyses, patients with poor prognosis more frequently had the following characteristics (Table 1): older age, history atrial fibrillation, high admission-NIHSS scores, high blood levels of ANC and AST, and low blood levels of Hb. As shown in Table 1, in the baseline study, adjuvant treatment with mechanical thrombectomy was associated with poor outcome.
We performed a multivariate logistic regression to assess the true predictive value of variables that apparently had an influence on prognosis on univariate analysis. SICH rates (Tables 1, 2) were not included in the calculation because they were not obtainable at baseline assessment. Although age and AST values were statistically significant in univariate testing, they neither were included in the logistic regression since they were already factored into the FIB-4 score, in order to avoid a distortion of FIB-4 score effect on prognosis (i.e., a confusion bias).
**Table 2**
| Unnamed: 0 | Full cohort n = 264 | No sICH n = 229 | sICH n = 35 | p- value |
| --- | --- | --- | --- | --- |
| FIB-4 score | 1.284 (0.989) | 1.249 (0.796) | 1.697 (1.348) | 0.004c* |
| FIB-4 score ≥2.67 | 32 (12.12%) | 24 | 8 | 0.037a* |
| FIB-4 score < 1.30 | 137 (51.89%) | 125 | 12 | 0.025a* |
On multivariate analysis (Table 3), FIB-4 score (OR 1.894, $$p \leq 0.011$$), history of atrial fibrillation (OR 3.488, $$p \leq 0.017$$), high admission NIHSS score (OR 1.305, $p \leq 0.001$) and low blood values of Hb (OR of high Hb levels OR 0.730, $p \leq 0.001$) remained significant predictors of poor prognosis. In spite of what was hypothesized with the univariate analysis, the regression demonstrated a protective effect of thrombectomy (OR 0.201, $$p \leq 0.005$$). Other variables (female sex, and ANC) were not significant prognostic predictors.
**Table 3**
| Coefficients | Odds ratio | p -value | 95% confidence interval | 95% confidence interval.1 |
| --- | --- | --- | --- | --- |
| Coefficients | Odds ratio | p -value | Lower bound | Upper bound |
| FIB-4 score | 1.894 | 0.011* | 1.160 | 3.094 |
| Female sex | 0.666 | 0.285 | 0.316 | 1.404 |
| Atrial fibrillation | 3.488 | 0.017* | 1.253 | 9.710 |
| Admission NIHSS | 1.305 | < 0.001* | 1.177 | 1.448 |
| EVT | 0.201 | 0.005* | 0.066 | 0.609 |
| Hb, g/dl | 0.730 | < 0.001* | 0.661 | 0.807 |
| ANC, × 109/L | 1.000 | 0.223 | 1.000 | 1.000 |
This statistical model showed good discriminatory power, with an area under the Receiver Operating Characteristic (ROC) curve of 0.877 (Supplementary Figure 1). It also produced a precision of $79.5\%$ and an accuracy of $79.3\%$.
Considering the FIB-4 cut-off values, we observed that the 32 patients ($12.1\%$) with a high risk of advanced fibrosis (i.e., FIB-4 score ≥2.67) were more frequently associated with a poor 3-month outcome (adjp = 0.021), whereas the 137 patients ($51.9\%$) with a high probability of exclusion of significant liver fibrosis (i.e., FIB-4 score <1.30) more frequently had a favorable 3-month outcome (adjp = 0.004; Figure 2).
**Figure 2:** *Patient proportional distribution based on the FIB-4 score cut-off values and the 3-month modified Ranking Scale.*
As with the primary outcome, univariate analyses were used to investigate the influence of variables on sICH (Table 2 and Supplementary Table 2). Regarding the secondary outcome, our study showed a statistically significant relationship between rates of sICH and FIB-4 values ($$p \leq 0.004$$), admission NIHSS (adjp = 0.035), EVT (adjp < 0.001), and serum levels of glucose (adjp < 0.001) and ANC (adjp = 0.009). However, multivariate analysis only confirmed the effect of admission NIH score on the secondary outcome (OR 0.901, $$p \leq 0.035$$, Supplementary Table 3).
Moreover, we divided the entire study population according to the cut-off values of the FIB-4 score, and we noted that patients with FIB-4 score <1.30 (exclusion of liver fibrosis) had lower probability of sICH (FIB-4 score <1.30, $$p \leq 0.025$$), whereas ischemic lesions from patients with high risk of advanced fibrosis (i.e., FIB-4 values ≥2.67) tended statistically to bleed more frequently ($$p \leq 0.037$$).
## 4. Discussion
The extension of indications for intravenous rt-Pa in patients with stroke and, in particular, the lengthening of the time window, has stimulated the search for reliable predictors able to provide early information on the risk/benefit ratio of the treatment. The ability to predict the outcome shortly after hospitalization can play an important role in the decision-making process regarding the best therapeutic approach in stroke patients and to plan a proper overall therapeutic care. Recently, some of the main predictors of outcome in patients with ischemic stroke treated with rt-Pa have been described [23]. High NIHSS scores, elevated systolic blood pressure values on admission, history of atrial fibrillation, and coronary artery disease were associated with poor outcome after 3 months. Another recent study on stroke patients undergoing rt-Pa reported that glycosylated hemoglobin blood levels were related to a poor early outcome but not to a poor functional prognosis at 3 months [24].
## 4.1. Interpretation of key results
In our study, in addition to clinical data, we considered using a simple and rapidly available index such as the FIB-4 score, based on laboratory parameters, to obtain prognostic information.
After adjustment for confounding factors by logistic regression analysis, we found that high values of FIB-4 score predicted outcome at 3 months in stroke patients treated with intravenous rt-Pa. Moreover, considering the validated cut-off values of this index, we were able to select a group of patients, characterized by a high risk of advanced liver fibrosis, who had a significantly higher probability of poor outcome than other patients. On the other hand, patients with exclusion of significant hepatic fibrosis had a higher probability of a favorable prognosis.
The FIB-4 score, which integrates blood levels of ALT, AST, and PLT, is not only a simple measure of patients' liver function but also reflects the complex systemic role of the liver itself. As highlighted by a cross-sectional study [11], liver dysfunction can lead to brain damage by several mechanisms, including small vessel disease or coagulopathy [25]. In addition, NAFLD is associated with systemic inflammation [26, 27], vascular inflammation [28] and atherosclerosis (25, 29–33). Advanced liver disease is associated with mixed coagulopathy [34], which increases the risk of both thrombotic and hemorrhagic stroke.
It's intuitive that worse outcomes may be the consequence of higher comorbidity in general, not a worse effect of thrombolysis. The selection of outcomes more specifically linked to this treatment should be considered in further dedicated works. For that very reason, we introduced the evaluation of the symptomatic intracerebral hemorrhage, a crucial mechanism involved in modulating the prognosis of patients with ischemic stroke undergoing fibrinolysis. Although the multivariate analysis would seem to exclude a role for the FIB-4 score in predicting bleeding complications, this hypothesis could not be entirely ruled out for two reasons: the limited number of patients with symptomatic cerebral hemorrhage, and the statistical model's inability to corroborate data from previous studies, which have also constantly indicated that admission hyperglycemia plays a significant role in predicting post-thrombolysis intracranial hemorrhagic events (35–37). Our findings from univariate analysis suggested that being affected by severe hepatic fibrosis may increase the risk of intracerebral hemorrhage. Based on these findings, the poor outcome at 3 months in patients with advanced hepatic fibrosis may be, at least in part, related to hemorrhagic complications. As demonstrated in previous studies, the intravenous use of rt-Pa significantly increases the risk of intracranial hemorrhage, which is otherwise uncommon in ischemic stroke [38]. Therefore, we hypothesize that for patients with severe hepatic fibrosis and ischemic stroke, the option for intravenous thrombolysis should be carefully evaluated considering the possible related risks.
In the present study, other indicators able to predict outcome at 3 months were identified. The negative prognostic role of atrial fibrillation in our patients was not unexpected although its significance has not been fully elucidated [39]. In the Virtual International Stroke Trials Archive, no significant association was found between atrial fibrillation and overall stroke outcome [40]. However, some studies found that atrial fibrillation was associated with favorable outcomes after thrombolysis for severe stroke, probably because of the effect of the thrombolytic agent on embolic arterial occlusion [37]. In agreement with our findings, most studies suggest that atrial fibrillation may increase the risk of symptomatic intracranial hemorrhage and early death, and decrease the likelihood of favorable outcome after thrombolysis [41, 42].
Our finding of negative predictive effects of high NIHSS scores (23, 43–45) and low serum levels of hemoglobin (46–49) on outcome confirm previous findings in patients undergoing thrombolysis for stroke.
The results of our multivariate analyses showing a favorable effect of endovascular therapy on stroke outcome are consistent with the results of a recent systematic review of 19 randomized clinical trials (RCTs) [50]. In this review, endovascular thrombectomy in patients with acute ischemic stroke due to occlusion of large arteries in the anterior circulation increased the chance of survival with good functional outcome (3-month mRS <3) with no negative effect on the risk of intracerebral hemorrhage or death. The predictive influence of anamnestic and laboratory variables on patients undergoing mechanical thrombectomy was recently investigated in a 2021 publication [51].
Toh et al. published an article at the beginning of 2023 addressing the same topic as the current study, with impactful results that confirm the significant influence of the FIB-4 score on the outcome of stroke patients undergoing thrombolysis in a highly representative sample of Asian population [52].
## 4.2. Strengths and limitations
Our study has some limitations. Because of its observational nature, this retrospective investigation does not reach the quality of evidence needed to draw definitive conclusions. Therefore, future prospective studies with established time points for blood sampling need to be conducted to assess the true cause-and-effect relationship between liver injury and stroke. In the event of a demonstration of a causal relationship, it will be critical to understand whether any improvement in liver condition can lower the risk of poor prognosis in stroke. Furthermore, it is not sufficiently clear whether the calculation of FIB-4 on admission can be considered reliable in expressing chronic liver damage, or is too influenced by stroke-related changes in blood levels of AST, ALT, and PLT. Further investigation is needed to obtain a clear answer with simultaneous assessment of the FIB-4 score and other markers of chronic liver disease.
On the other hand, the large sample of patients included and the easy usability of the score in a clinical setting with ordinary and cost-effective laboratory tests are the most important strengths of the study. We also used the already validated cut-off values of the FIB-4 score, that are strong indicators for the presence/absence of advanced liver fibrosis, significantly simplifying the calculation of the risk of poor outcome.
## 5. Conclusion
The results of the present study suggested that the FIB-4 score, a rapidly available and cost-effective parameter, can be considered as an independent predictor of poor prognosis, with high predictive accuracy, in patients with acute ischemic stroke undergoing intravenous thrombolysis.
In the new perspective of patient-centered medicine, identification of simple factors that predict treatment response is crucial to guide physicians in providing therapeutic strategies tailored to each single patient.
## 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 study was approved by the Ethics Committee of the Marche Polytechnic University (ID $\frac{57}{2020}$) and conducted according to the Declaration of Helsinki. Informed consent was obtained from all subjects involved in the study or their representatives.
## Author contributions
Conceptualization, software, formal analysis, writing—original draft preparation, project administration, and had full access to all the data in the study and takes responsibility for its integrity and the data analysis: DN. Methodology: SL. Investigation: DN, SB, and CR. Resources: SL and MB. Data curation: MB, DN, SB, and CR. Writing—review and editing: DN, SB, CR, and MS. Supervision: SL and MS. 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/fneur.2023.1103063/full#supplementary-material
## References
1. Lindsay MP, Norrving B, Sacco RL, Brainin M, Hacke W, Martins S. **World stroke organization (WSO): global stroke fact sheet 2019**. *Int J Stroke.* (2019.0) **14** 806-17. DOI: 10.1177/1747493019881353
2. Moskowitz MA, Lo EH, Iadecola C. **The science of stroke: mechanisms in search of treatments**. *Neuron.* (2010.0) **67** 181-98. DOI: 10.1016/j.neuron.2010.07.002
3. Fisher M, Saver JL. **Future directions of acute ischaemic stroke therapy**. *Lancet Neurol.* (2015.0) **14** 758-67. DOI: 10.1016/S1474-4422(15)00054-X
4. Molina CA, Montaner J, Abilleira S, Arenillas JF, Ribó M, Huertas R. **Time course of tissue plasminogen activator–induced recanalization in acute cardioembolic stroke: a case-control study**. *Stroke.* (2001.0) **32** 2821-7. DOI: 10.1161/hs1201.99821
5. Lattanzi S, Coccia M, Pulcini A, Cagnetti C, Galli FL, Villani L. **Endovascular treatment and cognitive outcome after anterior circulation ischemic stroke**. *Sci Rep.* (2020.0) **10** 18524. DOI: 10.1038/s41598-020-75609-1
6. Paul S, Davis AM. **Diagnosis and management of nonalcoholic fatty liver disease**. *JAMA.* (2018.0) **320** 2474-5. DOI: 10.1001/jama.2018.17365
7. Bataller R, Brenner DA. **Liver fibrosis**. *J Clin Invest.* (2005.0) **115** 209-18. DOI: 10.1172/JCI24282
8. Morling JR, Fallowfield JA, Guha IN, Nee LD, Glancy S, Williamson RM. **Using non-invasive biomarkers to identify hepatic fibrosis in people with type 2 diabetes mellitus: the Edinburgh type 2 diabetes study**. *J Hepatol.* (2014.0) **60** 384-91. DOI: 10.1016/j.jhep.2013.10.017
9. Xu H-W, Hsu Y-C, Chang C-H, Wei K-L, Lin C-L. **High FIB-4 index as an independent risk factor of prevalent chronic kidney disease in patients with nonalcoholic fatty liver disease**. *Hepatol Int.* (2016.0) **10** 340-6. DOI: 10.1007/s12072-015-9690-5
10. Walker AP. **Ischaemic stroke and liver fibrosis**. *Atherosclerosis.* (2017.0) **260** 153-5. DOI: 10.1016/j.atherosclerosis.2017.03.028
11. Parikh NS, VanWagner LB, Elkind MSV, Gutierrez J. **Association between nonalcoholic fatty liver disease with advanced fibrosis and stroke**. *J Neurol Sci.* (2019.0) **407** 116524. DOI: 10.1016/j.jns.2019.116524
12. Baik M, Kim SU, Kang S, Park HJ, Nam HS, Heo JH. **Liver fibrosis, not steatosis, associates with long-term outcomes in ischaemic stroke patients**. *Cerebrovasc Dis.* (2019.0) **47** 32-9. DOI: 10.1159/000497069
13. Yuan C-X, Ruan Y-T, Zeng Y-Y, Cheng H-R, Cheng Q-Q, Chen Y-B. **Liver fibrosis is associated with hemorrhagic transformation in patients with acute ischemic stroke**. *Front Neurol.* (2020.0) **11** 867. DOI: 10.3389/fneur.2020.00867
14. Whiteley WN, Slot KB, Fernandes P, Sandercock P, Wardlaw J. **Risk factors for intracranial hemorrhage in acute ischemic stroke patients treated with recombinant tissue plasminogen activator: a systematic review and meta-analysis of 55 studies**. *Stroke.* (2012.0) **43** 2904-9. DOI: 10.1161/STROKEAHA.112.665331
15. Moshayedi H, Ahrabi R, Mardani A, Sadigetegad S, Farhudi M. **Association between non-alcoholic fatty liver disease and ischemic stroke**. *Iran J Neurol.* (2014.0) **13** 144-8. PMID: 25422733
16. Fandler-Höfler S, Stauber RE, Kneihsl M, Wünsch G, Haidegger M, Poltrum B. **Non-invasive markers of liver fibrosis and outcome in large vessel occlusion stroke**. *Ther Adv Neurol Disord.* (2021.0) **14** 175628642110372. DOI: 10.1177/17562864211037239
17. Wityk RJ, Pessin MS, Kaplan RF, Caplan LR. **Serial assessment of acute stroke using the NIH Stroke Scale**. *Stroke.* (1994.0) **25** 362-5. DOI: 10.1161/01.STR.25.2.362
18. Shah AG, Lydecker A, Murray K, Tetri BN, Contos MJ, Sanyal AJ. **Comparison of noninvasive markers of fibrosis in patients with nonalcoholic fatty liver disease**. *Clin Gastroenterol Hepatol.* (2009.0) **7** 1104-12. DOI: 10.1016/j.cgh.2009.05.033
19. Staufer K, Halilbasic E, Spindelboeck W, Eilenberg M, Prager G, Stadlbauer V. **Evaluation and comparison of six noninvasive tests for prediction of significant or advanced fibrosis in nonalcoholic fatty liver disease**. *United Eur Gastroenterol J.* (2019.0) **7** 1113-23. DOI: 10.1177/2050640619865133
20. Banks JL, Marotta CA. **Outcomes validity and reliability of the modified Rankin scale: implications for stroke clinical trials: a literature review and synthesis**. *Stroke.* (2007.0) **38** 1091-6. DOI: 10.1161/01.STR.0000258355.23810.c6
21. Yaghi S, Willey JZ, Cucchiara B, Goldstein JN, Gonzales NR, Khatri P. **Treatment and outcome of hemorrhagic transformation after intravenous alteplase in acute ischemic stroke: a scientific statement for healthcare professionals from the American Heart Association/American Stroke Association**. *Stroke.* (2017.0) **48** e343-61. DOI: 10.1161/STR.0000000000000152
22. Campbell BCV, Meretoja A, Donnan GA, Davis SM. **Twenty-year history of the evolution of stroke thrombolysis with intravenous alteplase to reduce long-term disability**. *Stroke.* (2015.0) **46** 2341-6. DOI: 10.1161/STROKEAHA.114.007564
23. Mehrpour M, Afrakhteh M, Shojaei SF, Sohrabi A, Ashayeri R, Esmaeili S. **Factors predicting good outcome of intravenous thrombolysis in stroke patients before rt-PA administration**. *Casp J Intern Med* (2019.0) **10** 424-430. DOI: 10.22088/cjim.10.4.424
24. Han L, Hou Z, Ma M, Ding D, Wang D, Fang Q. **Impact of glycosylated hemoglobin on early neurological deterioration in acute mild ischemic stroke patients treated with intravenous thrombolysis**. *Front Aging Neurosci* (2023.0) **14** 1073267. DOI: 10.3389/fnagi.2022.1073267
25. Kim YD, Song D, Heo JH, Kim SU, Kim BK, Park JY. **Relationship between cerebral microbleeds and liver stiffness determined by transient elastography**. *PLoS ONE.* (2015.0) **10** e0139227. DOI: 10.1371/journal.pone.0139227
26. Al Rifai M, Silverman MG, Nasir K, Budoff MJ, Blankstein R, Szklo M. **The association of nonalcoholic fatty liver disease, obesity, and metabolic syndrome, with systemic inflammation and subclinical atherosclerosis: the Multi-Ethnic Study of Atherosclerosis (MESA)**. *Atherosclerosis.* (2015.0) **239** 629-33. DOI: 10.1016/j.atherosclerosis.2015.02.011
27. Targher G, Bertolini L, Scala L, Zoppini G, Zenari L, Falezza G. **Non-alcoholic hepatic steatosis and its relation to increased plasma biomarkers of inflammation and endothelial dysfunction in non-diabetic men. Role of visceral adipose tissue**. *Diabet Med.* (2005.0) **22** 1354-8. DOI: 10.1111/j.1464-5491.2005.01646.x
28. Lee HJ, Lee CH, Kim S, Hwang SY, Hong HC, Choi HY. **Association between vascular inflammation and non-alcoholic fatty liver disease: analysis by 18F-fluorodeoxyglucose positron emission tomography**. *Metabolism.* (2017.0) **67** 72-9. DOI: 10.1016/j.metabol.2016.11.004
29. Mellinger JL, Pencina KM, Massaro JM, Hoffmann U, Seshadri S, Fox CS. **Hepatic steatosis and cardiovascular disease outcomes: an analysis of the Framingham Heart Study**. *J Hepatol.* (2015.0) **63** 470-6. DOI: 10.1016/j.jhep.2015.02.045
30. VanWagner LB, Ning H, Lewis CE, Shay CM, Wilkins J, Carr JJ. **Associations between nonalcoholic fatty liver disease and subclinical atherosclerosis in middle-aged adults: the Coronary Artery Risk Development in Young Adults Study**. *Atherosclerosis.* (2014.0) **235** 599-605. DOI: 10.1016/j.atherosclerosis.2014.05.962
31. Sinn DH, Kang D, Chang Y, Ryu S, Gu S, Kim H. **Non-alcoholic fatty liver disease and progression of coronary artery calcium score: a retrospective cohort study**. *Gut.* (2017.0) **66** 323-9. DOI: 10.1136/gutjnl-2016-311854
32. Wong VW-S, Wong GL-H, Yeung JC-L, Fung CY-K, Chan JK-L, Chang ZH-Y. **Long-term clinical outcomes after fatty liver screening in patients undergoing coronary angiogram: a prospective cohort study**. *Hepatol Baltim Md.* (2016.0) **63** 754-63. DOI: 10.1002/hep.28253
33. Madan SA, John F, Pyrsopoulos N, Pitchumoni CS. **Nonalcoholic fatty liver disease and carotid artery atherosclerosis in children and adults: a meta-analysis**. *Eur J Gastroenterol Hepatol.* (2015.0) **27** 1237-48. DOI: 10.1097/MEG.0000000000000429
34. Tripodi A, Primignani M, Chantarangkul V, Dell'Era A, Clerici M, de Franchis R. **An imbalance of pro- vs anti-coagulation factors in plasma from patients with cirrhosis**. *Gastroenterology.* (2009.0) **137** 2105-11. DOI: 10.1053/j.gastro.2009.08.045
35. Liu M, Pan Y, Zhou L, Wang Y. **Predictors of post-thrombolysis symptomatic intracranial hemorrhage in Chinese patients with acute ischemic stroke**. *PLoS ONE.* (2017.0) **12** e0184646. DOI: 10.1371/journal.pone.0184646
36. Nisar T, Hanumanthu R, Khandelwal P. **Symptomatic intracerebral hemorrhage after intravenous thrombolysis: predictive factors and validation of prediction models**. *J Stroke Cerebrovasc Dis.* (2019.0) **28** 104360. DOI: 10.1016/j.jstrokecerebrovasdis.2019.104360
37. Sung S-F. **Atrial fibrillation predicts good functional outcome following intravenous tissue plasminogen activator in patients with severe stroke**. *Clin Neurol Neurosurg* (2013.0) **4** 892-5. DOI: 10.1016/j.clineuro.2012.08.034
38. Lindley RI, Wardlaw JM, Sandercock PAG, Rimdusid P, Lewis SC, Signorini DF. **Frequency and risk factors for spontaneous hemorrhagic transformation of cerebral infarction**. *J Stroke Cerebrovasc Dis.* (2004.0) **13** 235-46. DOI: 10.1016/j.jstrokecerebrovasdis.2004.03.003
39. Dang H, Ge W-Q, Zhou C-F, Zhou C-Y. **The correlation between atrial fibrillation and prognosis and hemorrhagic transformation**. *Eur Neurol.* (2019.0) **82** 9-14. DOI: 10.1159/000504191
40. Frank B, Fulton R, Weimar C, Shuaib A, Lees KR, VISTA Collaborators. **Impact of atrial fibrillation on outcome in thrombolyzed patients with stroke: evidence from the Virtual International Stroke Trials Archive (VISTA)**. *Stroke* **43** 1872-7. DOI: 10.1161/STROKEAHA.112.650838
41. Saposnik G, Gladstone D, Raptis R, Zhou L, Hart RG. **Atrial fibrillation in ischemic stroke: predicting response to thrombolysis and clinical outcomes**. *Stroke* **44** 99-104. DOI: 10.1161/STROKEAHA.112.676551
42. Yue R, Li D, Yu J, Li S, Ma Y, Huang S. **Atrial fibrillation is associated with poor outcomes in thrombolyzed patients with acute ischemic stroke: a systematic review and meta-analysis**. (2016.0) **95** 9. DOI: 10.1097/MD.0000000000003054
43. Liu X, Zhang J, Tian C, Wang J. **The relationship of leukoaraiosis, haemorrhagic transformation and prognosis at 3 months after intravenous thrombolysis in elderly patients aged ≥ 60 years with acute cerebral infarction**. *Neurol Sci.* (2020.0) **41** 3195-200. DOI: 10.1007/s10072-020-04398-2
44. Jantasri S, Tiamkao S, Sawanyawisuth K. **A 2-point difference of NIHSS as a predictor of acute ischemic stroke outcome at 3 months after thrombolytic therapy**. *Clin Neurol Neurosurg.* (2020.0) **198** 106206. DOI: 10.1016/j.clineuro.2020.106206
45. Murphy A, Symons SP, Hopyan J, Aviv RI. **Factors influencing clinically meaningful recanalization after IV-rtPA in acute ischemic stroke**. *AJNR Am J Neuroradiol* (2013.0) **34** 146-52. DOI: 10.3174/ajnr.A3169
46. Altersberger VL, Kellert L, Al Sultan AS, Martinez-Majander N, Hametner C, Eskandari A. **Effect of haemoglobin levels on outcome in intravenous thrombolysis-treated stroke patients**. *Eur Stroke J.* (2020.0) **5** 138-47. DOI: 10.1177/2396987319889468
47. Barlas RS, Honney K, Loke YK, McCall SJ, Bettencourt-Silva JH, Clark AB. **Impact of hemoglobin levels and anemia on mortality in acute stroke: analysis of UK regional registry data, systematic review, and meta-analysis**. *J Am Heart Assoc.* (2016.0) **5** e003019. DOI: 10.1161/JAHA.115.003019
48. Kimberly WT, Lima FO, O'Connor S, Furie KL. **Sex differences and hemoglobin levels in relation to stroke outcomes**. *Neurology.* (2013.0) **80** 719-24. DOI: 10.1212/WNL.0b013e31828250ff
49. Lasek-Bal A, Holecki M, Steposz A, Duława J. **The impact of anemia on the course and short-term prognosis in patients with first ever ischemic stroke**. *Neurol Neurochir Pol.* (2015.0) **49** 107-12. DOI: 10.1016/j.pjnns.2015.03.001
50. Roaldsen MB, Jusufovic M, Berge E, Lindekleiv H. **Endovascular thrombectomy and intra-arterial interventions for acute ischaemic stroke**. *Cochrane Database Syst Rev.* (2021.0) **6** CD007574. DOI: 10.1002/14651858.CD010995.pub3
51. Lasek-Bal A, Binek Ł, Zak A, Student S, Krzan A, Puz P. **Clinical and non-clinical determinants of the effect of mechanical thrombectomy and post-stroke functional status of patients in short and long-term follow-up**. *J Clin Med.* (2021.0) **10** 5084. DOI: 10.3390/jcm10215084
52. Toh EMS, Joseph Ravi PR, Ming C, Lim AYL, Sia C-H, Chan BPL. **Risk of liver fibrosis is associated with more severe strokes, increased complications with thrombolysis, and mortality**. *J Clin Med.* (2023.0) **12** 356. DOI: 10.3390/jcm12010356
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.